The
reference class problem is illustrated by what Artificial Intelligence
researchers call the Nixon Diamond.1 
Quakers are usually pacifists and Nixon is a Quaker.  Republicans are usually not pacifists and
Nixon is a Republican.  On the total
evidence, that Nixon is a Quaker and a Republican, what should we believe about
whether Nixon is a pacifist?

The
problem arises very generally, whenever statistical evidence is applied to an
individual case.  The case, Nixon, is a
member of two different “reference classes”—the class of Quakers and the
class of Republicans—and in these classes the
frequency of the attribute or feature to be predicted—i.e., being a pacifist—differs.  How then can we decide how the statistical
evidence bears on the case?  Should we
attempt to find a best, most relevant, reference class?  Or should we attempt to combine evidence from
the various reference classes of which the case is a member, and if so, how?

In “A
Practical Solution to the Reference Class Problem“2 Edward K. Cheng usefully surveys the ways in which the problem arises in legal
contexts.  In United States v. Shonubi,3 sentencing guidelines required
an estimate of how much heroin Charles Shonubi, a Nigerian drug smuggler, had
carried through New York’s John F. Kennedy Airport (JFK)
on seven previous trips during which he had been undetected.  The estimate was based on the average amounts of heroin found on Nigerian drug smugglers caught at JFK airport in the time
period.  Why should that be used as the
reference class relevant to the case, rather than, say, George Washington
Bridge tollbooth collectors (Shonubi’s
day job)?  Or, take a more typical case
involving valuation:  Valuing a house for
sale involves estimating its price from the sale records for “similar”
houses.  No other house is exactly the
same as the given one, so how widely or narrowly should one choose the
reference class of “similar” houses, and on what criteria?  Number of bathrooms?  Age? 
Street number?

Statistical
theory has been dealing with inference from quantitative data for a very long
time.  So it is reasonable to hope, as
Cheng argues, that the discipline of statistics will have available methods
applicable to the reference class problem.4

It is
true that traditional statistical theory tends to avoid the problem, taking for
granted in expositions that the reference class has been correctly identified
before inference begins.  The basic idea
of statistical inference is to observe counts—i.e., frequencies,
proportions—in some reference class and apply
the result as an estimate for a new, similar case.  For example, one might draw a line of best
fit through data of age and heights of trees in order to apply the
estimated relationship between age and height to trees not yet observed.  In expounding such techniques it is normally
assumed that one has an unproblematic set of measurements of an identifiable
and reasonably homogeneous set of trees. 
It is normally left to the statistician’s
good sense to choose a data set relevant to the problem.

But the
recent expansion of statistics into the “data
mining” of huge “data warehouses”5 has forced consideration of how
to identify what is relevant in large and mainly uninformative heaps of data,
typically not collected with the current problem in mind.  Cheng argues that a practical solution to the
problem, at least when opposing counsel have put forward differing clear
proposals on what the reference class should be, lies in modern “model selection”
methods which decide on the appropriate complexity of a model by a formula such
as the Akaike Information Criterion.6 
This Essay argues that a simpler area of recent statistics, the theory
of feature selection methods, is more relevant. 
Since they are more straightforward and do not require an understanding of
the issues concerning model complexity, they are explained first, in Part I of
this Essay.  Part II discusses model
complexity and argues that Cheng’s approach is workable, but that
the statistical literature provides some equally credible alternative approaches.

I.  Feature Selection Methods

The
methods most applicable to problems like those in the legal context, such as
real estate valuation, fall under the heading of “feature
selection,” also known as “variable selection”
or “attribute selection.”7  A
database is organized into many rows (the cases) and columns (the fields,
attributes, properties, or features of the cases), as in the following
schematic example:

Table 1:  Sample Real Estate Database

In general, one should imagine many more
cases (rows):  hundreds for real estate
but millions for many kinds of health and gene data and financial records, and
possibly thousands of features (columns). 
In such large cases, the great majority of features are expected to be
irrelevant to the task of prediction. 
For example, not every feature or measurement in a gene database will be
helpful in predicting cancer, and most features of financial records will be
irrelevant to determining creditworthiness. 
The more features in a database, the harder it is to evaluate each
feature’s relevance.

The aim
of feature selection methods is to determine from large amounts of data which
of the many properties or features of the individual cases are relevant to a
given classification or prediction task. 
For example, of the many features of houses in a real estate database—house size, lot size, number of bathrooms, street
number, age, zip code, etc.—which are relevant to predicting
the price?

One major
purpose of identifying relevant features is to prevent the computations from
becoming unfeasible, since computing with all the data, most of which is
irrelevant, would be impossible for databases of the size typically used.  But identifying the relevant features is also
an aim in itself, since it will enable researchers to understand the data and
form hypotheses as to what features are driving the system.  These features could then either be
classified as a symptom or identified as a cause and possibly changed.8

For the
present purpose, however, the main significance of feature selection is that it
largely solves the reference class problem. 
Choosing the relevant features determines the appropriate reference class
for a case:  Ideally, it is the class of
those items that share with it all the features that have been found relevant
for the task.  However, that ideal is not
always attainable, as discussed below.  There is an accepted definition of the
relevance of a feature to an outcome—for
example, of “number of bathrooms” to “house price.”  A feature is
relevant if it gives some information about the outcome—i.e., if “number of bathrooms” makes some difference to “house price”
in the sense that, on average, a different number of bathrooms goes with a
different house price.  Relevance is correlation.  In a very simple
example, the reason that traffic lights are informative is that green is very
highly correlated with it being safe to drive through the intersection:  green and it is safe, red and it is not.9 
The color of the car ahead, however, is not correlated with the safe
time to drive, so there is no point attending to it when deciding whether to
drive across the intersection.  There is
a standard definition of correlation and there are some alternative measures of
association to choose from, but they are all intended to measure the degree to
which one variable “goes with” another.

Features
may be irrelevant to prediction in two ways: 
A feature is either not correlated with the variable being predicted, or
it is correlated but is redundant because other features provide the same
information, as they are highly correlated with it.  Work in data mining concentrates on finding a
suitable small subset of features, all relevant and, as far as possible, not
redundant, which competently predict the target.  There are some subtleties about the relevance
of sets of features (as opposed to individual features), since two features
could be relevant in combination although they are not relevant individually.10

Knowledge
of the relevance of features can come in two ways.  Either one measures correlation in the data,
or one brings to bear prior knowledge of relevance—or perhaps more often, irrelevance.  In Shonubi, discussed in Cheng and earlier, where the amount of drugs smuggled by
Shonubi had to be estimated from data on “similar” drug smugglers, there was a reasonable prior belief
that being a Nigerian drug mule was relevant to the amount of drugs smuggled,
while having a day job as a toll collector was not.11 
Humans’ prior beliefs on frequencies—e.g., whether eating colored mushrooms is often
followed by illness, whether rocks typically have lions behind them—have been much studied and often found to be sound.  But they have some persistent biases, such as
a tendency to overweight very rare disasters such as air crashes.  Therefore, any alleged prior knowledge is an
important matter of evidential weight and thus important to reach a correct
decision.  So it should be subject to
scrutiny in the usual way, not by statistical formulas, but by such means as a
committee of experts or cross-examination in court.12

Having
selected the features relevant to the prediction task, one then wishes to
create a model.  That is, one decides on
the mathematical form of the relationship between the relevant features and the
target.  In the classic reference class
problem, the outcome is predicted by a very simple function of the statistics
of the reference class.  For example, in Shonubi, the estimate of drugs smuggled
by Shonubi was the average of drugs smuggled by members of the reference class:  Nigerian drug mules at JFK during the time
period.  The question then is, how should
the reference class be chosen, once the set of relevant—or strongly relevant—features
have been identified?

There is
a unique natural choice:  The correct
reference class is that defined by the intersection of the relevant
features.  If being Nigerian, being a
drug mule, being at JFK, and being in the time period are all reasonably
believed to be relevant to the amount of drugs smuggled, and there is no
evidence that any other feature on which data is available is relevant, then
the ideal choice of reference class is Nigerian drug smugglers at JFK in the
time period.

The
reference class problem then divides into two, depending on whether this
ideal choice of reference class is usable. 
It is usable if there is a sufficiently large number of cases in it for
a reliable estimate of the target.  A
data set that is too small, or even empty, will not support reliable estimates,
since there is too much chance involved in which few cases happened to land in
the set.13 
To know whether the data is enough to ensure reliability of the
estimate, one consults standard statistical theory on the variance or standard
deviation of the estimate in question. 
That is, one asks how variable the estimate is given the sample
size:  the smaller the sample, the more
variable and thus unreliable the estimate. 
For example, if n drug
smugglers are found with amounts x₁, . . .,
x_n of
drugs, of which the average is x, then the standard deviation of x is approximately the
standard deviation of the original n observations divided by √n.  Or, if the problem is
to estimate a probability based on a proportion in a small reference class—for example, eighteen of twenty cars on the
surveillance video went through a red light, so the chance the defendant’s car went through a red light is ninety percent—then the reliability of the estimate is given by
calculating a “confidence interval.”14 
This calculation again improves with the square root of the number of
instances.  Thus, we can quantify how much increasing the size of the reference
class increases the reliability of the estimate, and how unreliable a very
small reference class is.

If, on
the other hand, the reference class defined by the intersection of relevant
features is too small for reliable inference, or perhaps even empty, one is
still left with useful information in the wider classes defined by taking some
but not all of the relevant features. 
The statistics in those different classes will usually be
different.  For example, Nigerian drug
smugglers in general and drug smugglers at JFK may have different averages of
drugs smuggled, even though one may not have data on Nigerian drug smugglers at
JFK.  The problem then is how to combine
the statistics in the different classes in which an individual lies, in order
to make an estimate applicable to the individual case.  A paradigm of the problem is:

Suppose
we have under observation a certain Jones, who is found to be a Texan and a
philosopher.  We know that 99 per cent of
all Texans are millionaires, and that only 1 per cent of all philosophers are
millionaires (and we have no information about the class of Texan
philosophers).  On that evidence, what
should we conclude about whether Jones is a millionaire?  We would know the probability of Jones being
a millionaire, given either one of those pieces of information, but what should
be conclude when we have both?15

This
problem unfortunately is unsolved.  If
the estimates coming from the different classes are close to one another, then
of course it does not matter much which one is used, as the estimates in effect
concur.  If, as in the example, the
estimates conflict, then any combination of them, even if correct, is
unreliable.  In that case, the issue may
come down to whether one estimate is better on intuitive or other external
grounds.  For example, one may be reduced
to arguing in court whether being Texan or being a philosopher is known to be
more relevant to wealth.

We come
now to the choice and fitting of models, where it is determined how the data deemed relevant will be used to make the estimate.

II.  Models: 
Simple or Smooth?

Cheng
poses the reference class problem as one of “model
selection.”16 
Statistical models comprise a range of techniques for deciding the form
of the method to be applied to data before the parameters of the model are
chosen by fitting it to the data.  For
example, one may decide that a linear, line of best fit model is applicable,
and then find the slope of the line by fitting it to the data.  That is illustrated in the top two panels of
Cheng’s Figure 1, which show a typical
dataset of observations (Figure 1a) and a straight line of best fit to the data
(Figure 1b).17 The line expresses the best linear relationship between the two measured
quantities—here, study hours and GPA—allowing a prediction of GPA from
study hours for data not yet observed.

It is
possible to fit more complex models, with quadratic, fourth degree or other
formulas, as illustrated in Figures 1c and 1d, in the hope of obtaining more
accurate predictions.  That hope may or
may not be realized.  Sometimes more complex
models appear to do better than a simple straight line (as in Figure 1c), while
sometimes they appear to have become overcomplicated and do worse (as in Figure 1d).  Cheng recommends choosing
simple models—that is, ones with few parameters—with the degree of complexity chosen according to
Akaike’s Information Criterion or some
similar formula.18 
That is a respectable option, according to the statistical literature,
and gives reasonable results when it comes to prediction.  But it is not the only option, and it is
arguably not well grounded in theory.  An
alternative to that “Ockhamist” approach, as it might be called in view of its
emphasis on simplicity, is a “smoothing” approach, which does not care about the complexity of models as
long as they are smooth, that is, varying little from point to point.

Cheng
Figure 1:  Example Fits to Observed
Datapoints

To
explain the contrast between the two styles of models, let us go back to Cheng’s example of fitting a curve to noisy one-input,
one-output data.19  The problem is:  Given some points generated by an underlying
but unknown function, perhaps with noise, how can you fit the best curve to
them?  In this sense, “best” would mean the ability to predict
new points that would be generated by the same process.

One good
feature of the problem is that, on the whole, the eye is quite competent to
judge fit.  One can see that the line in Figure 1a is too straight, the curve in Figure 1d too irregular, and the smooth curve
in Figure 1c is about right.  The eye, of
course, does not see the formulas or the number of parameters in them, but only
the smoothness of the fit.  Another
positive feature of the problem is that there are several mathematical methods
available, implemented in software, that give more or less the same answer—polynomial regression, neural networks, smoothed
splines, and kernel smoothing, among others. 
(To say that they give the same answer does not mean that they give the
same or similar formulas, but similar graphs of the fitted curve.)  The disadvantage of the problem is that there
is still no agreement on the basics of the theory, the cause of the agreement
between methods, or what feature the answer has that makes it the right
answer.  There are, in fact, two
radically different views on the theory: 
an Ockhamist view and an anti-Ockhamist view.

According
to the view based on Ockham’s razor, it is a matter of
choosing the right number of parameters for the curve.  If one decides to fit a polynomial, one can
choose a line, a quadratic, a cubic, and so on. 
A line is described by two parameters, intercept and slope, that are to
be chosen by, or “fitted to,” the data.  A
quadratic needs three parameters, a cubic four, and so on.  If one chooses too few parameters—a too simple curve—it
fails to fit the data well.  On the other
hand, if one chooses too many, the curve easily fits the actual data, but
because it wobbles like jelly in response to the details of the actual data
set, it would have been different if fitted to another data set generated by
the same process.  Hence it predicts
poorly:  It is said to have “overfitt[ed]”
the data, to be “fitt[ed] to the noise,” or to have “high
variance.”20

If one
decides to fit one of a fixed family of parametric curves, such as polynomials,
there is a reasonably well-established theory on how to choose the right number
of parameters for a particular data set, and on why that number is
correct.  Bayesian statisticians with a
background in physics have provided an analysis based on the precision of a
prior distribution with few parameters versus the imprecision of one with
many:  A model with many trainable
parameters “hedges its bets,” so to speak, by being ready to fit anything, and
hence is less “falsifiable” than a more precise one.21 
One speaks of “O[ckham's] hill,”
which has a peak at the correct number of parameters.22 
Statisticians who are not card-carrying Bayesians have a similar theory,
which issues in a prescription called the Akaike Information Criterion for the
number of parameters.23

That
works well, and is reasonably convincing in the context of physics, where there
is a prior expectation that simple models can often be found.  It is not so clear that, where things are
expected to be complex, as in economic modeling, taking a simple model is the
only or best method of producing a falsifiable model, that is, one that will be
found to perform well on new data.

Figure 2:  Spline Smoothing of Observed Datapoints24

In response, the
anti-Ockhamist school makes two points. 
Firstly, “having few parameters” is not the same thing as “simple”:  If one takes any wobbly curve, such as the
cross-section of a piece of corrugated iron, it can stand in place of the
straight line in the above example and generate a family of curves.  It has not been explained why the bottom
level of the family should be, for example, straight.25  More
importantly, some other methods which give much the same answer graphically,
such as smoothed splines, achieve the result by finding a curve that is
sufficiently smooth, but not too smooth—but
the curve does not have few parameters. 
Smoothing methods work by replacing the datapoints with averages of
nearby datapoints:  The value of the
curve at any point is the weighted average of nearby datapoints.  Different methods are distinguished by
different technical decisions on how to weigh and determining how close is “nearby.”

Given a
smoothing method, the main problem is to decide on the right degree of
smoothing:  smooth too much and there is
just a straight line that has lost most of the structure of the data, smooth
too little and the result wobbles around fitting the idiosyncracies of the
individual data set.  In most
circumstances, the right degree of smoothing is determined by the method of
cross-validation, which calculates how well the estimate would predict a
datapoint if it were left out:  If the
curve would be little changed by leaving out a datapoint, and would also
predict that datapoint well, the degree of smoothing is correct.26

Further,
in some related contexts, although simplicity does lead to good results, one
can see that complexity is even better. 
Studies of machine learning, which is in principle a higher-dimensional
version of curve-fitting, show that complicating the result, while preserving
its behavior on the old data, can improve its performance.27 
Leo Breiman goes so far as to speak of “two
cultures” in statistics:  an older style that looks for simple and
explanatory models of data, and a more contemporary style that embraces large
and internally complex “black-box” predictors such as neural nets and random forests,
as long as they are equipped with methods of smoothing to prevent overfitting
of the data.28

The
natural conclusion to reach is that simplicity, or fewness of parameters, is
not in itself desirable in curve fitting and related contexts, but only works
because it is normally used in such a way as to correlate with smoothness,
which is what really enables prediction. 
Prediction is what counts, and simplicity may or may not help it.

Conclusion

Statistical methods do have
advice to offer on how courts should judge quantitative evidence, but in a way
that supplements normal intuitive legal argumentation rather than replacing it
by a formula.  In cases like Shonubi and
those involving real estate valuation or the measurement of environmental
risks, there is relevant quantitative data and hence a need for technical
statistical advice.  It is crucial to
know what properties of the data are statistically relevant to the inferential
task, the answer to which determines the appropriate reference class in which
to take counts.  Knowledge of the relevance
of properties is of two kinds.  The
first, commonsense or scientific knowledge of causes and symptoms, is subject
to the usual style of intuitive reasoning and challenge in the courtroom.  The second, obtained from statistical methods
such as variable selection and the fitting of models, is also crucial when
there is significant reliance on inference from a set of quantitative
data.  The statistical techniques should
neither dominate nor be dominated by intuitive reasoning.  Lawyers need to understand both styles of
reasoning in order to integrate them.

* Professor, School of Mathematics and Statistics, University of New South Wales, Sydney, Australia.

1 Raymond Reiter & Giovanni
Criscuolo, On Interacting Defaults: 
Proceedings of the 7th International Joint Conference on Artificial
Intelligence 270 (1981).

2 Edward K. Cheng, A Practical
Solution to the Reference Class Problem, 109 Colum. L. Rev. 2081 (2009).

3 895 F. Supp. 460 (E.D.N.Y. 1995),
discussed in Peter Tillers, If Wishes Were Horses:  Discursive Comments on Attempts to Prevent
Individuals from Being Unfairly Burdened by Their Reference Classes, 4 L.
Probability & Risk 33 (2005).

4 Cheng, supra note 2, at 2095 (“[T]he reference class problem is . . . a
subspecies of the model selection problem” and “model
selection criteria . . .
eliminate the reference class problem as it arises in legal contexts.”). 

5 See generally Daniel T. Larose,
Discovering Knowledge in Data:  An
Introduction to Data Mining (2005); George M. Marakas, Modern Data Warehousing,
Mining and Visualization:  Core Concepts
(2003).

6 Cheng, supra note 2, at 2093–94.

7 See generally Avrim L. Blum &
Pat Langley, Selection of Relevant Features and Examples in Machine Learning,
97 Artificial Intelligence 245 (1997) (discussing feature selection in machine
learning); Isabelle Guyon & André
Elisseeff, An Introduction to Variable and Feature Selection, 3 J. Machine
Learning Res. 1157 (2003) (same); Mark A. Hall & Geoffrey Holmes,
Benchmarking Attribute Selection Techniques for Discrete Class Data Mining, 15
IEEE Trans. on Knowledge & Data Engineering 1437 (2003) (same); Patricia E.N.
Lutu & Andries P. Engelbrecht, A Decision Rule-Based Method for Feature Selection
in Predictive Data Mining, 37 Expert Sys. with Applications 602 (2010) (same).

8 The problem then becomes what to
do with the identified relevant features in constructing a predictive model;
that is, more a task of model selection and will be discussed later.  See infra Part II (discussing different
approaches to model selection). 

9 See R.K. Templeton & J. Franklin, Adaptive Information
and Animal Behaviour, 10 Evolutionary
Theory 145, 145–46 (1992) (discussing traffic light example and
adaptive information). 

10 Hopefully such
problems are rare.  There are a number of
algorithms available for searching a database and finding sets of relevant
features.

11 See Cheng, supra note 2, at 2082 (discussing statistical comparison in United
States v. Shonubi, 895 F. Supp. 460, 466 (E.D.N.Y. 1995)). 

12 See, e.g., James Franklin et al., Evaluating Extreme Risks in Invasion
Ecology:  Learning from Banking
Compliance, 14 Diversity &
Distributions 581, 582 (2008) (discussing how expert committees “encourage[d]
care and transparency” in Australian import biosecurity agency analysis).

13 Reichenbach said when coining the
phrase “reference
class problem” that it should be “the narrowest class for which
reliable statistics can be compiled,” which is correct, except that one
does not narrow a relevant reference class by splitting it according to
irrelevant attributes.  Hans Reichenbach,
The Theory of Probability 374
(1949).

14 See generally Lawrence D. Brown,
T. Tony Cai & Anirban DasGupta, Interval Estimation for a Binomial
Proportion, 16 Stat. Sci. 101, 101 (2001) (discussing “interval
estimation of the probability of success in a binomial distribution”).  Calculators that measure the confidence
interval are available online.  See
Binomial Proportion Confidence Interval: 
Web-based Calculators, Wikipedia, at
http://en.wikipedia.org/wiki/Binomial_proportion_confidence_interval (last
visited Feb. 12, 2010) (on file with the Columbia
Law Review) (providing links to web-based calculators).

15 S.F. Barker, Induction and Hypothesis 76 (1957).

16 Cheng, supra note 2, at 2090–92.

17 Id. at 2091.

18 Id. at 2093–94.

19 Id. at Part II.A
(detailing model selection problem).

20 See, e.g., Wallace E. Larimore
& Raman K. Mehra, The Problem of Overfitting Data, 10 Byte 167, 168 (1985) (“[O]verfitting
lessens the predictive value of the model.”).

21 See Prasanta S. Bandyopadhayay,
Robert J. Boik & Prasun Basu, The Curve-Fitting Problem:  A Bayesian Approach, 63 Phil. Sci. S264 (1996) (using
Bayesian theorem to solve curve-fitting problem); William H. Jeffreys &
James O. Berger, Ockham’s Razor and Bayesian Analysis, 80 Am. Sci. 64, 68 (1992).

22 See David J.C. MacKay, Bayesian
Interpolation, 4 Neural Computation
415 (1992) (arguing Bayesian analysis infers values regularizing constants
and noise levels, leading to effective number of parameters determined by data
set).

23 See Malcolm Forster & Elliot
Sober, How to Tell When Simpler, More Unified, or Less Ad Hoc Theories Will
Provide More Accurate Predictions, 45 Brit.
J. for Phil. Sci. 1 (1994); I.A. Kieseppä, Akaike
Information Criterion, Curve-Fitting, and the Philosophical Problem of
Simplicity, 48 Brit. J. for Phil.
Sci. 21 (1997).

24 B.W. Silverman, Some Aspects of
the Spline Smoothing Approach to Nonparametric Regression Curve Fitting, 47 J.
Royal Stat. Soc’y B. 1, 9 (1985).

25 Kieseppä, supra note 23 (analyzing
solutions to problems of “bumpier curves” and “smoother
curves”); André Kukla,
Forster and Sober on the Curve-Fitting Problem, 46 Brit. J. for Phil. Sci. 248, 249 (1995)(same).

26 See Grace Wahba, Spline Models for Observational Data 45–49
(1990); Richard R. Picard & R.
Dennis Cook, Cross-Validation of Regression Models, 79 J. Am. Stat. Ass’n 575
(1984).

27 See Pedro Domingos, The Role of
Occam’s Razor
in Knowledge Discovery, 3 Data Mining and Knowledge Discovery 409 (1999); C. Schaffer, Overfitting Avoidance as Bias, 10 Machine Learning 153 (1993); G.I.
Webb, Further Experimental Evidence Against the Utility of Occam’s Razor,
4 J. Artificial Intelligence Res.
397 (1996).

28 Leo Breiman, Statistical
Modeling:  The Two Cultures, 16 Stat.
Sci. 199, 221–22 (2001) (advocating active monitoring to protect
against overfitting).

Preferred
Citation:  James Franklin, Feature Selection Methods for Solving the
Reference Class Problem:  Comment on
Edward K. Cheng, “A Practical Solution to the Reference Class Problem,” 110
Colum. L. Rev. Sidebar 12 (2010),

http://www.columbialawreview.org/sidebar/volume/110/12_franklin.pdf.

Regulating immigration ultimately requires lawmakers to reach
beyond a unilateral “gatekeeping” strategy defined by efforts to stop
migration through law enforcement and economic coercion.  Because states cannot effectively manage
migration in isolation from one another,1 the United
States must approach the issue by prioritizing cooperation
with actors outside the United
States. 
In their contributions to the policy debate, scholars increasingly have
emphasized the importance of addressing labor and illegal migration through
bilateral and transnational frameworks2—through accords that would recognize
the interdependence of the United States
and Mexico
and engage our neighbor to the south directly through joint efforts to channel
migratory flows.

In this Essay, I seek to contribute to this strand of
commentary by focusing on the actual mechanisms of transnationalism and the
avenues they open up for advancing a meaningful bilateralism.  I demonstrate that the cross-border
administrative law space created by these mechanisms is occupied not just by
international entities, but also by entanglements between the domestic
institutions of different countries. 
I emphasize the importance of identifying and then building the
mechanisms of bilateralism, or the cross-border institutional capacities needed
for managing migration, in a manner that promotes burden-sharing, or that
ensures that both sides of the bilateral relationship reap benefits and bear
costs, in rough proportion.

Many transnational mechanisms of governance already exist,
and I use this space to advance the conversation about transnational regulation
by mapping some of them and then offering initial suggestions for their development.  For the sake of simplicity, I focus on the
U.S.-Mexico relationship, not least because immigration to the United States is overwhelmingly Mexican (twenty
percent of the authorized population and fifty-six percent of the unauthorized
population),3 and because ninety-eight percent of Mexico’s out-migration is to the United States.4 
Our shared border is also itself the source and site of many of the
pressures that make immigration a significant public policy issue.

Bringing to light existing institutional frameworks might
also help to calm the passions that arise from the popular assumption that the United States and Mexico are locked in an adversarial
relationship in relation to migration. 
This relationship is not without its tensions, of course.  Actors on either side of the border have
different sets of priorities.  Both the United States and Mexico have interests in
controlling smuggling, drug trafficking (and the associated extreme violence),
and other criminal activity around the border. 
But on the subject of immigration, Mexico’s objectives revolve primarily
around protecting the interests of its nationals abroad and facilitating
remittance traffic back to Mexico,5 whereas the United States’s interests focus on simultaneously serving the
country’s labor market needs and preventing immigration from having adverse
effects on U.S. workers and federal, state, and local budgets.  This divergence of interest makes efforts to
cooperate fraught and gives rise to collective action difficulties.  The asymmetrical nature of the relationship
between the United States
and Mexico further
complicates the matter; the United
States possesses superior bargaining power
and economic strength.6  But it is precisely because of these
asymmetries that developing institutional frameworks for cooperation is vital
to meeting the regulatory challenge posed by the shared border, and the shared
interest in managing migration creates crucial commonality.

I.  The Mechanisms of
Bilateralism in the U.S.-Mexico
Relationship

The mechanisms of bilateralism exist in at least three forms:
 (1) diplomatic and information-sharing
networks that involve consultation and conferencing among cabinet officials and
agency heads in both the United States and Mexico; (2) actual cooperative
ventures between administrative officials on both sides of the border; and (3)
civil society networks (many of which do not involve state actors) developed to
serve the needs of Mexican migrants inside the United States.  These frameworks are defined by different
degrees of formality and thus have varying relationships to the hard legal
regimes of the United States
and Mexico.  Most of them do not have legal regulatory
authority, though they do reflect how the sovereign power and local interests
of each nation project across the border. 
In addition, these mechanisms make clear that migration management
depends on activity—sometimes collaborative, sometimes isolated—by officials at
all levels of government and in the private sector.

A.
 Diplomatic and Information-Sharing Networks

The U.S. Department of State and the Mexican Foreign Ministry
clearly communicate over issues of mutual concern, but it is also the case that
almost every agency head deals regularly and directly with his or her counterpart
on the other side of the border.7 
Perhaps the most prominent example of such interaction is the
U.S.-Mexico Binational Commission (BNC), originally established in 1981 and
envisioned as a forum for cabinet-level officials to meet once or twice a year
to discuss regulatory issues related to the binational relationship.  Over time, the BNC has become a one-day
conference chaired by the U.S. Secretary of State and the Mexican Secretary of
Foreign Relations and attended by cabinet-level officials and heads of
administrative agencies.  The Clinton
Administration, in particular, invested heavily in this project,8 and migration and border security have become prominent among the Commission’s
agenda items.9 
The governors of border states
in both countries have developed a similar network of their own,10 and mayors in neighboring towns maintain close working relationships.  State and local officials in both countries
increasingly reach across the border to attract new forms of investment.  In the United States, state and local officials
also work to respond to the transnational interests of their new popular
constituencies of Mexican origin, by facilitating trade, investment, and
cross-border traffic.

The tangible accomplishments of these gatherings are
few.  But formal networks of this kind at
least promote information sharing across governments, as well as the
articulation of joint priorities.  Most
important, when taken seriously, these networks can build and sustain the
political will necessary to advance bilateral cooperation through the creation
of actual legal regimes.  They can also
help project the particular interests of the United
States and Mexico
into the domestic political debate of the other country, to ensure that the
interests of U.S.
and Mexican citizens receive recognition in the national decisionmaking
processes over which they have no direct control.11

The Partnership for Prosperity—a public-private partnership
initiated by Presidents Bush and Fox in 2001—offers another example of an
information-sharing network created by high level officials to advance a
tangible reform agenda through reliance on cooperation between bureaucrats and
the private sector.  The Partnership
promotes development in Mexico,
particularly in areas with high rates of out-migration to the United States.12  Former Secretary of State Colin Powell has
listed the Partnership’s accomplishments as including lowering the fees for
transferring money between the United States and Mexico, developing innovative
methods for funding infrastructure projects, and establishing an Overseas
Private Investment Corporation in Mexico to provide over $600 million in
various types of financing to U.S. businesses operating in Mexico.13

B.
 The Cross-Border Work of Administrative
Agencies

Administrative agencies on both sides of the border,
particularly in the areas of law enforcement and public health, work together
to develop actual practices that advance mutual goals, as well as to streamline
operations that involve officials from both countries.  With respect to border security, cooperation
by state officials has been a longstanding practice.  Again, this relationship has been complicated
by different priorities, with Mexico
focused on matters of internal security and protecting the lives and rights of
its nationals at the border, and the United States focused on
terrorism-related screening and intelligence gathering, as well as drug
interdiction.14 
But Presidents Clinton and Zedillo poured considerable effort into
formalizing bilateral consultation between consulates and border agencies.  In
2002, the two countries entered into the so-called “Smart Border Agreement”
(SBA) designed to increase the number of secure documents for frequent border
crossers, improve intelligence sharing, and implement security strategies
focused not just on the border, but also on the areas where threats originate.
Initiatives also have included simplifying the execution of arrest warrants on
either side of the border by reducing the legal obstacles that prevent warrants
from either country from being honored in the other, as well as removing
obstacles to collaboration between law enforcement officials.  Perhaps most important, the SBA has brought
attention to the need to regularize migratory flows in order to free up larger
amounts of scarce law enforcement resources to address public safety and
national security issues.15

On the subject of public health, perhaps the most formal
existing mechanism of cooperation is the United States-Mexico Border Health
Commission, which was established under the auspices of Public Law 103-400 and
through a Memorandum of Agreement between the governments of the United States and Mexico.16  Membership consists of high-level government
officials from federal and state health agencies in both countries, as well as
political appointees.  The Commission’s primary objective is to deal with
mobile public health crises and to address the needs for internal and external
forms of communication and technical harmonization.17  The Commission also continues to work toward
facilitating binational health insurance to cover Mexican nationals working in
the United States and to
address legal barriers to coverage, such as a Texas state law that does not permit HMOs to
market insurance plans across the border.18

Perhaps the most notable initiative to emerge from the
Commission is its support for the Ventanillas de Salud program, through which
Mexican consulates (there are fifty inside the United
States) collaborate with U.S.
nonprofit health organizations to provide services and education to Mexican
citizens living and working in the United States.19  Twelve such programs existed as of 2008, and
they provide direct services to 500,000 “consulate clients,” reaching an
estimated 1.5 million people in the U.S.  A combination of Mexican state funds and
private grants provides financial support for the programs, which focus
primarily on disseminating occupational health and safety-related information
and referring clients for services in Mexico.20

C.
 Transnational Civil Society
Networks

Several transnational civil society networks have taken shape
in the last few decades and have created structures for the delivery of social
services and the protection of Mexican nationals’ interests abroad.21 
Some of these mechanisms are creatures of the state, such as the
Ventanillas de Salud program, and others emerge via networks of private
actors.  Indeed, emigration countries
generally have begun adopting policies to address the rights and interests of
migrants abroad, and to provide incentives for them to return home.22  The attempts by sending societies, such as
Mexico, to both provide for and control their nationals abroad simultaneously
alleviate some of the burden felt by U.S. institutions and inject the sending
state and its civil society into life in the United States.

Two examples of the increasingly vibrant civil society
networks that have emerged include nongovernmental hometown associations and
the Institute for Mexicans Abroad, which was formed in 2003 to replace the
Mexican Communities Abroad Program.  The
hometown associations and migrant federations have evolved as informal networks
that connect migrants in the United States
to citizens and organizations in towns in Mexico.  They consist of business associations, labor
unions, and churches23 on both sides of the border.  In addition
to organizing social and civic events, the associations develop positions on
binational issues, and their component parts maintain close relations with one
another.  Some associations actually participate
in and help fund development projects in both the United
States and Mexico,24 and the Mexican government finances association projects through its consulates
in the United States.25

The Institute for Mexicans Abroad similarly exists to provide
services to Mexicans living in the United States, while simultaneously
promoting understanding of Mexican culture in receiving communities and
facilitating immigrant integration.  Housed
in Mexico’s Ministry of
Foreign Affairs, the Institute works through the Mexican consulates in the United States
and partners with civic actors on both sides of the border.26  When it
was still the Communities Abroad Program, the Institute began the “Tres por Uno”
program, through which migrants abroad and Mexican government agencies finance
infrastructure projects in Mexico,
spending eighty million dollars per year by 2005.27  Today, the Institute’s advisory council
consists of Mexican community leaders, Latino organizations in the United States, and officials from the state
governments of Mexico.28  The Institute cosponsors social programs with
agencies at all levels of government.  In
addition to coordinating the Ventanillas de Salud program in the United States,29 the Institute works with school districts to
assess the needs of migrant school children and provides Mexican nationals with
financial literacy workshops and English-language and continuing adult
education programs.30  The Institute’s objectives also include public diplomacy efforts in the United States designed to enhance respect for
the culture and country of Mexico,
in order to address the hostility toward Mexicans and Mexican culture generated
by illegal immigration.31  This promotion of understanding helps to
create political space for the treatment of immigration as a humanitarian and
regulatory issue, rather than as a law enforcement problem.

II.  Shaping the Future
of Bilateralism

The existing mechanisms of
bilateralism form an intricate constellation of institutions designed to
address the multiple needs engendered by immigration between Mexico and the United States.  These mechanisms are largely ad hoc, however.32  Efforts to develop an overarching normative
framework could prove stultifying, but some theoretical cohesion will be required
to shift the focus of public debate toward bilateralism, given the many
different forms of governance needed to negotiate the U.S.-Mexico relationship.

The mechanisms just discussed all embody an ethic of burden
sharing and reflect the crucial insight that when countries have large
populations in common, both sides of the border should benefit from and bear
responsibility for those populations. 
The civil society networks, in particular, reflect the burden-sharing
ethos, demonstrating through actual practice a commitment to treating migration
as a binational responsibility requiring transnational mechanisms of
governance.  The concepts of “burden
sharing” (or mutual obligation) and “management” (or regulation) thus offer
general but sufficiently substantive umbrella ideas under which to develop the
mechanisms of bilateralism.  This shared
responsibility ultimately can be realized through an agenda that builds the
existing institutional capacities described above to (1) address the root
causes of migration and (2) manage migration’s effects to promote public
health, safety, and prosperity, as well as human rights.

With respect to the first set of priorities—addressing the
root causes of migration—the concepts of burden sharing and management ought to
be highlighted at the diplomatic level as governing principles for the
U.S.-Mexico relationship.  The foregoing
discussion should make clear that the Mexican government takes the interests of
its emigrating nationals and the impact of their movement on Mexico and the
United States quite seriously,33 but it remains crucial for high-ranking U.S.
officials to apply consistent pressure on Mexico to address the underlying
structural causes of migration, especially limited job opportunities and
structural inequalities at home.  Perhaps
more important, the United States
should commit diplomatically to assisting Mexico in this long-term,
cause-based approach to stabilizing and reducing migration.34

At the administrative level, this approach could require
diminished reliance by Mexico on remittances as a tool of development,
especially since monies sent home from abroad offer only short-term and
individualized (rather than systemic) development assistance.  It could include devising positive incentives
for Mexican migrants to return home, such as business loans or educational
grants, to be jointly funded and implemented by Mexico
and the United States.  Increased funding and technical support from U.S. agencies
to arrangements like the Partnership for Prosperity could help expand the reach
of preexisting development initiatives. 
Some commentators also have advocated increased utilization of the North
American Development Bank, an institution created in 1994 in a side agreement
to NAFTA.  To transform the Bank into a
collaborative mechanism of development, both its substantive and geographic
mandates would need to be expanded beyond their original focus on the
environment.35

The second set of priorities, which revolves around
regularization and management of existing flows, is crucial to creating a
climate in which migration is seen as a win-win proposition, rather than as an
imposition on the United States
and a drain on Mexico.  Regularization will require a labor accord of
some kind, the details of which are well beyond the scope of this Essay.36  But at the very least, transnational, interagency
cooperation should be part of any new labor regime.  For example, federal and state agencies37 in Mexico
can assist in providing crucial data and economic and demographic projections
regarding the Mexican labor market and the sources, numbers, and
characteristics of likely Mexican migrants.38  The network of Mexican consulates, which
already focuses on the rights and interests of Mexican nationals in the United
States, could be cultivated and enlisted in identifying and helping to satisfy
the particular needs of migrants, and perhaps also in monitoring
employer-employee relations.  Indeed,
federal policymakers, as well as state and local officials, should regard the
civil society networks that have emerged over the last two decades as crucial
resources for providing for immigrant welfare and promoting immigrant
integration.  The hometown associations’
and Mexican Institute’s simultaneous focus on promoting attachments to Mexico
and encouraging community participation and English-language learning in the
United States provides an essential model for addressing the inherent tension
between wanting to encourage some migrants to return and ensuring that those
who inevitably stay for long periods (or forever) integrate successfully.

* * *

To varying degrees, these transnational schemes amount to the
extension of sovereignty across borders and the intermingling of different
sovereign regimes.  They thus implicate
one of the central concerns surrounding global governance:  accountability to the public.  Migration scuttles the social contract on
both sides of the border, making one government’s decisions consequential for
the citizens of the other state.  This
discussion thus surfaces a question that must be addressed in the shift toward
a burden-sharing and management model of migration:  How do we simultaneously address cross-border
regulatory challenges and ensure that the publics on both sides of the border
have adequate voice in what is done?  Crucially, this accountability must extend not
just from the state to its own citizens, but from the United States government to the people of Mexico and the government of Mexico to the people of the United States.

Interdependence, which is often invoked in the rhetoric that
surrounds diplomacy, must ultimately be a core presumption of actual reform
debates, and lawmakers ought to make concerted efforts to solidify bilateralism
as a working practice.  Building
transnational networks with ties to public institutions on either side of the
border ultimately will promote accountability in immigration policy by ensuring
the participation of representatives from both nations in the myriad
decisionmaking processes that involve an increasingly transnational public.

* Henry L. Stimson Visiting Professor of Law, Harvard Law School; Professor of Law, NYU School of Law. I am grateful for outstanding research assistance from Amelia Rawls and Casey Schwarz.

1 Bimal Ghosh, Managing
Migration:  Interstate Cooperation at the
Global Level, in Interstate
Cooperation and Migration 109, 111 (2005).

2 See, e.g., Jorge A. Bustamante,
Mexico-United States Labor Migration Flows, 31 Int’l
Migration Rev. 1112, 1112 (1997) (noting that outmigration in Mexico is
perceived as economic and labor phenomenon from which United States reaps
benefits, whereas in United States, same migration is perceived as law and
order or public safety phenomenon); Jennifer Gordon, Transnational Labor
Citizenship, 80 S. Cal. L. Rev. 503, 509 (2007) (“I propose an opening up of the
fortress of labor and of the nation-state to accommodate a constant flow of new
migrants through a model that would tie immigration status to membership in
organizations of transnational workers rather than to a particular employer.”);
Alejandro Portes, The Fence to Nowhere, Am. Prospect, Oct. 2007, at 26, 28
(arguing that United States and Mexico must develop means to manage massive
flow of migration); Marc Rosenblum, The United States and Mexico:  Prospects for a Bilateral Migration Policy,
Border Battles:  The U.S. Immigration
Debates, Mar. 8, 2007, at
http://borderbattles.ssrc.org/Rosenblum/printable.html (on file with the Columbia Law Review) (detailing fraught
history of U.S.-Mexico bilateralism and laying out agenda for bilateral
accord).  For discussion of the history
of U.S.-Mexico bilateralism and its limitations as reflected in the World War
II-era Bracero program, see Mae M. Ngai, Impossible Subjects:  Illegal Immigrants and the Making of Modern
America 138–47 (2005); Adam B. Cox & Cristina M. Rodríguez,
The President and Immigration Law, 119 Yale L.J. 458, 485–91
(2009).

3 Andrew Selee, Woodrow Wilson
Ctr., More Than Neighbors:  An Overview
of Mexico and U.S.-Mexico Relations 5 (2007), available at
http://www.wilsoncenter.org/topics/pubs/Mexico.More%20Than%20Neighbors.pdf (on
file with the Columbia Law Review).

4 David Fitzgerald, A Nation of
Emigrants:  How Mexico Manages its
Migration 5–6 (2009).

5 In 2005, remittances amounted to
approximately twenty billion dollars.  See
id. at 63.  Though the current recession has led to a decline in this traffic of
late, see Joel Millman, Remittances
to Mexico Fall More Than Forecast, Wall St. J., Jan. 28, 2009, at A3 (noting
that amount of money sent home by Mexicans working in United States dropped by
3.6% in 2008—first decline recorded since Mexico began tracking
remittance traffic thirteen years ago), the decline in remittances to Latin
America may be “bottoming out,” Dilip Ratha, Sanket Mohapatra &
Ani Silwal, Migration and Development Brief 11, Migration and Remittance
Trends 2009, at 2 (2009), available at http://siteresources.worldbank.org/INTPROSPECTS/Resources/334934-1110315015165/MigrationAndDevelopmentBrief11.pdf (on file with the Columbia Law Review) (“Remittance
flows to Mexico declined by 13.4 percent in the first nine months of 2009 . . .
. However, the decline in flows appears to be bottoming out in most countries
across the region.”).

6 Selee, supra note 3, at iii; Ghosh, supra note 1, at 121–22.

7 Selee, supra note 3, at 13.

8 Augustin Escobar, Interstate
Cooperation:  The Americas, in Interstate Cooperation and Migration,
supra note 1, at 65, 80.

9 See, e.g., K. Larry Storrs, Cong.
Research Serv., Mexico-United States Dialogue on Migration and Border Issues,
2001–2005, at
5–6
(2005), available at http://www.fas.org/sgp/crs/row/RL32735.pdf (on file with
the Columbia Law Review) (discussing
meeting of November 25–26, 2002, during which talks addressed importance of
forging bilateral migration accord and joint border security initiatives).

10 The first border governors’
conference took place in Ciudad Juarez in June 1980 and resulted, in part, from
the efforts of Governor Clements of Texas, who had promised during his campaign
to improve relations with Mexico.  John
Kincaid, The American Governors in International Affairs, Publius:  J. of Federalism, Fall 1984, at 95, 111.  The conference is designed to enable
governors to develop proposals for reform on border-related matters that can be
submitted to their respective state and national governments, but through
consultation with their counterparts on the other side of the border.  The
conference consists of thirteen working groups focused on issue areas that
include agriculture, border security, crossings, economic development, education,
emergency matters, energy, environment, health, science, tourism, water, and
wildlife. See Border Governors
Conference, Worktables, at http://www.bordergovernors.ca.gov/worktables (last
visited Jan. 12, 2010) (on file with the Columbia
Law Review).

11 Another such network is the
Regional Conference on Migration (RCM), a multilateral regional meeting that has
occurred annually since 1996.  The RCM is attended by eleven member states,
including the United States, Mexico, and Canada, and its objectives include
promoting the orderly movement of persons and respect for the human rights of
migrants.  Reg’l Conference on Migration, Conference Description, at
http://www.rcmvs.org/pagina_n.htm (last visited Jan. 5, 2010) (on file with the
Columbia Law Review).

12 P’ship for Prosperity, Report
to President Vicente Fox and President George W. Bush:  Creating Prosperity Through Partnership 1 (2002),
available at http://www.state.gov/documents/organization/16197.pdf (on file
with the Columbia Law Review).

13 Storrs, supra note 9, at 9–10.

14 Andrés Rozental & Peter H. Smith,
Woodrow Wilson Ctr. Mex. Inst., The United States and Mexico:  Forging a Strategic Partnership 11 (2005),
available at http://www.wilsoncenter.org/topics/pubs/USMEXenglish%20copy1.pdf (on file with the Columbia Law Review).

15 Id. at 12.

16 United States-Mexico Border Health
Commission Act, Pub. L. No. 103-400, 108 Stat. 4169 (1994) (codified as amended
at 22 U.S.C. §§ 290n–290n-6 (2006)); Agreement to
Establish a United States-Mexico Border Health Commission, U.S.-Mex., July 14–24,
2000, T.I.A.S. No. 13107, available at http://www.state.gov/documents/organization/126990.pdf (on file with the Columbia Law Review).

17 For a history of such efforts, see
Julie Collins-Dogrul, Managing U.S.-Mexico “Border Health”:  An Organizational Field Approach, 63 Soc.
Sci. & Med. 3199 (2006).

18 See U.S.-Mex. Border Health Comm’n, Annual Meeting of
the United States-Mexico Border Health Commission, McAllen, Texas, March 3–4, 2008:  Summary, Actions and Next Steps 8 (2008),
available at http://www.borderhealth.org/extranet/files/file_1058.pdf (on file
with the Columbia Law Review)
(reporting on work of H.H.S. Working Group on Binational Health Insurance).

19 See Hilda Bogue, Ventanillas de Salud Program and Community Health
Centers, Migrant Health Newsline, Nov.–Dec. 2006, at
http://findarticles.com/p/articles/mi_6843/is_6_23/ai_n28465712 (on file with
the Columbia Law Review) (describing
goals and benefits of program to decrease barriers to access among consular
clients and their families).

20 U.S.-Mex. Border Health
Comm’n, NIOSH Information Dissemination on the U.S.-Mexico Border, at
http://www.borderhealth.org/files/res_1203.pdf (last updated Mar. 11, 2008) (on
file with the Columbia Law Review).

21 For a general account of this
transnational activity, see Gaspar Rivera-Salgado, Binational Organizations of
Mexican Migrants in the United States, Soc. Just., Fall 1999, at 27.  For a
framing of this civil society activity as political activity and organization
by Mexican migrants, see Jonathan Fox, Mexican Migrant Civic Participation in
the United States, Border Battles:  The
U.S. Immigration Debates, Aug. 15, 2006, at
http://borderbattles.ssrc.org/Fox/printable.html (on file with the Columbia Law Review).

22 See, e.g., Fitzgerald, supra note 4, at 3, 155–64 (describing ways in which Mexican
government has attempted to regulate emigration).

23 The Catholic Church, itself a
global, transnational network, has become an important agent for the promotion
of migrants’ interests in recent years.  The Mexican Catholic
Church has abandoned its former policy of discouraging emigration in favor of accepting
emigration and maintaining connections to Mexicans abroad via partnerships with
U.S. dioceses and a “binational migrant ministry,”
simultaneously encouraging Mexicans to retain ties to Mexican culture and
society and learn English and American customs.  See id.
at 96.

24 See Selee, supra note 3, at 18 (describing cross-border activities of
nongovernmental organizations).

25 Rivera-Salgado, supra note 21, at 30.

26 For a detailed discussion of the
Institute’s structure, see
Laureen Laglagaron, Migration Policy Inst., Protection Through Integration:  The Mexican Government’s Efforts to Aid
Migrants in the United States 10–14 (2010), available at
http://www.migrationpolicy.org/pubs/IME-Jan2010.pdf (on file with the Columbia Law Review).

27 Fitzgerald, supra note 4, at 58–59.

28 Id.

29 Laglagaron, supra note 26, at 29–31.

30 Id., at 1–2, 14–28; see also Rodolfo Figueroa-Aramoni, A
Nation Beyond its Borders:  The Program
for Mexican Communities Abroad, 86 J. Am. Hist. 537, 539–40
(1999) (discussing educational and cultural activities of Institute’s
predecessor).

31 See Claudia Keller Lapayre, The Institute of Mexicans Abroad as a
Mexican Institution of Public Diplomacy in the United States 24 (Mar. 22, 2006)
(paper presented at the annual meeting of the International Studies
Association), available at http://www.allacademic.com/meta/p99314_index.html (on file with the Columbia Law Review)
(noting that Institute’s programs, though targeted at Mexican nationals,
also seek to “improve Mexico’s image”).

32 See Ghosh, supra note 1, at 115 (observing that existing mechanisms of
interstate cooperation act “in an isolated and fragmentary manner and are too
narrowly focused to provide an adequate normative framework for a comprehensive
approach to migration management”).

33 See Fitzgerald, supra note 4, at 155 (noting that Mexico sought from 1900 to 1970
to control “volume, trip
duration, skills, and geographic origin of emigrants” through
mechanisms such as propaganda, withholding of travel documents, Bracero
temporary worker program, and coercion at border, but that failure of such
efforts led Mexico to shift strategy to management).

34 For an argument that substantial
investment by the United States in Mexico represents the best long-term
strategy for reducing illegal immigration, see Jorge Durand & Douglas S.
Massey, Borderline Sanity, Am. Prospect, Sept. 24–Oct. 8, 2001, at 28.

35 See Rozental & Smith, supra
note 14, at 16 (describing expanding the mandate of the North
American Development Bank as one approach to infrastructure development).  Of
course, this sort of measure could be politically unpalatable, given that NAFTA
is perceived on both sides of the border as having failed to deliver the growth
in Mexico that would prevent the erosion of jobs and wages in the United
States.  See, e.g., Douglas S. Massey,
Jorge Durand & Nolan J. Malone, Beyond Smoke and Mirrors:  Mexican Immigration in an Era of Economic
Integration 73–104 (2002) (discussing relationship between NAFTA
and illegal immigration).

36 For literature offering various
approaches, see sources cited supra note 2.  For a critique of temporary worker programs, see
generally Cristina M. Rodríguez, Guest Workers and Integration:  Toward a Theory of What Immigrants and
Americans Owe One Another, 2007 U. Chi. Legal F. 219.

37 State and local governments in
Mexico, which have become increasingly independent as the result of the federal
government’s devolution to them, in the mid-1990s, of authority
over education and health care, should not be overlooked as partners in
regulation and information gathering.  See
Selee, supra note 3, at 30 (“Most education and healthcare has
been decentralized to state governments, and municipalities are responsible for
most basic city and county services.”).

38 A reform proposal that has gained
prominence would create an administrative agency or independent commission to
set visa policy on an annual basis, in the interest of creating an agile labor
policy mechanism attuned to facts on the ground.  Interagency communication
across the U.S.-Mexico border will be crucial to the information-gathering work
of such a commission.  For a discussion of the commission proposal, see Cox
& Rodríguez, supra note 2, at 538–39.

Preferred Citation:  Cristina M. Rodríguez, Building Capacity for the Transnational Regulation of Migration,
110 Colum. L. Rev. Sidebar 1
(2010) http://www.columbialawreview.org/Sidebar/volume/110/1_Rodriguez.pdf.

In recent years, an increasing number of scholars and
commentators have turned their attention to the criminalization of migration in
the United States.1 
These scholars have focused on three distinct trends:  the increasingly harsh criminal consequences
attached to violations of laws regulating migration,2 the use of removal as an adjunct to
criminal punishment in cases involving noncitizens,3 and the rising reliance on criminal law enforcement actors and mechanisms in
civil immigration proceedings.4  One major effect of these three trends has
been the incorporation of criminal law methodologies into the realm of civil
immigration enforcement and adjudication. 
Recently, Stephen Legomsky has theorized the asymmetric nature of this
incorporation.5  As he explains, the “theories, methods,
perceptions, and priorities” of criminal law enforcement have been incorporated
into immigration proceedings, while the procedural protections of criminal adjudication
have been explicitly rejected.6  His analysis focuses on how the
criminalization of migration is reshaping the realm of civil immigration
proceedings.

In contrast, this Essay centralizes and attempts to theorize
the criminal prosecutions of migration-related offenses.7  Part I of this Essay describes this
trend.  Specifically, Part I.A highlights
the ways in which the regulation of migration has increasingly become a subject
of the criminal law.  Part I.B discusses
the explosion of migration-related criminal prosecutions over the past few
years.

Part II of this Essay provides several examples of the use of
criminal prosecutions in the migration context in order to explore an
undertheorized effect of this trend, namely, that the protective features of
criminal investigation and adjudication are melting away at the edges in
certain criminal cases involving migration-related offenses.  Part II.A explores the border-centered
prosecutions of Operation Streamline and the more geographically diffuse Fast
Track program aimed at felony reentries. 
Part II.B focuses upon the use of criminal prosecutions in worksite
immigration enforcement efforts.  Part
II.C diagnoses the ways in which these proceedings reflect declining procedural
protections in the realm of criminal prosecutions for immigration-related
offenses.

As these examples make clear, not
only are we seeing what Stephen Legomsky has termed the asymmetric
incorporation of criminal justice norms into civil removal proceedings, but we
are also witnessing the importation of the relaxed procedural norms of civil
immigration proceedings into the criminal realm.

I.  Regulating
Migration Through Crime

The regulation of migration has
long taken place primarily in the civil sphere.8 
In recent years, however, the U.S. government has increasingly handled
migration control through the criminal justice system.  Part I.A discusses the legislation that
Congress and various state legislatures have enacted to criminalize acts associated
with migration.  Part I.B. describes the
recent upward spike in prosecutions of these migration-related offenses.

A.  Creating and Enhancing Criminal Sanctions for
Offenses Relating to Migration

Since the 1980s, Congress has passed legislation subjecting
more and more acts associated with migration to criminal penalties, or
increasing the severity of criminal sanctions imposed for the commission of
those acts.9  Criminal offenses newly created in the 1980s
included the hiring of unauthorized noncitizen workers,10 reliance on false documents to evade employer sanctions laws,11 and marriage fraud.12  In the late 1980s and the 1990s, illegal
reentry provisions were added and strengthened,13 as were various fraud provisions relating to the processes of seeking
immigration benefits and citizenship.14  And in 2000, penalties were raised for
various offenses relating to trafficking in persons.15

Although the criminalization of
migration-related offenses used to be entirely federal in nature, in recent
years, states and localities have added a host of anti-loitering laws and other
similar ordinances that are clearly intended to—and have been used to—facilitate
the criminal prosecution of unauthorized migrants at the state and local level.16 
One example is Arizona’s version of the identity theft law.  The crime—entitled “Taking identity of
another person or entity”—creates criminal culpability for the use of an
alternate identity whether or not the defendant knows that he is using the
identity of an actual person and whether or not another person with such an
identity actually exists.17  This offense, which does not require
theft of an actual identity, can be deployed as a means of prosecuting
noncitizens who have used false identities to obtain employment in cases where
there is no loss to anyone as a result of the use of that identity.  Several other states have also enacted
provisions that mirror federal prohibitions on immigration crimes like
smuggling and harboring of unauthorized migrants and false proof of citizenship
or immigration status.18  These statutes take advantage of federally
created immigration categories to create a space for local enforcement.19

B.  Increasing Prosecutions for Immigration
Offenses

Related to, but even more striking than the steady increase
in migration-related criminal offenses is the rising tidal wave of
immigration-related criminal prosecutions of the past decade.  After remaining relatively flat in the period
from 1986 to 1996, the number of immigration prosecutions almost quadrupled over
the next ten years.20  The prosecution of migration-related offenses
exploded in the wake of September 11, 2001. 
In 2004, U.S. magistrates convicted 15,662 noncitizens of immigration
crimes, and U.S. district court judges convicted another 15,546.21  The numbers continued to climb thereafter.22  Since 2004, immigration prosecutions have
topped the list of federal criminal prosecutions, outstripping federal drug and
weapons prosecutions, and dwarfing many other forms of federal criminal
prosecutions.23  This trend has continued even with the change
in presidential administrations.24  And, as previously noted, states and
localities—long thought to be excluded from the enforcement of immigration law—have
found ways to use their own criminal laws to supplement these federal
prosecutions.

These trends have attracted the notice of immigration
scholars,25 but have not received much concerted attention from criminal law scholars.  Indeed, the restoration of some semblance of
rationality to the discourse on the war on crime has perhaps drawn attention
away from the parallel trend whereby the tools formerly used to fight the war
on crime are increasingly put to use against noncitizens.26 
The lack of attention to the unprecedented criminalization of migration
by scholars of criminal law and procedure is unfortunate because, like the war
on drugs that preceded it,27 the emerging use of the criminal justice system to attack the social problem of
unauthorized migration carries with it distinct procedural and social
consequences that are worthy of sustained attention.  In the next Part, I provide a few examples
to illustrate this point.

II.  Declining Procedural Protections in
the Criminal Sphere

The well-known constitutional maxim that deportation (now
“removal”)28 is not punishment provides longstanding precedent for important legal
distinctions between civil immigration proceedings and criminal proceedings.29  Noncitizens in removal proceedings are not
entitled to counsel at the government’s expense.30  Evidence obtained in violation of a
noncitizen’s constitutional rights generally is not subject to suppression in
civil removal proceedings.31  And immigration detention—which is also not legal “punishment”—is not
subject to the same constitutional constraints as criminal detention.32  These distinctions have frequently caused
immigration attorneys to yearn for the constitutional protections of criminal
proceedings,33 even while acknowledging the inadequacies of those protections.

Unfortunately, recent developments
suggest that the lower standards of procedural protections that apply in
removal proceedings have made ultra vires incursions into the criminal realm. 
For purposes of this Essay, a few examples suffice to illustrate the
problem.  Part II.A outlines two programs
aimed at addressing the crimes of entry without inspection and felony
reentry.  Part II.B discusses
prosecutions arising out of a worksite raid. 
Part II.C addresses the declining procedural protections in the realm of
criminal prosecutions for migration-related offenses.34

A. 
Criminal Prosecutions for Unlawful Entries

Because there are hundreds of thousands of unauthorized
border crossings each year,35 prosecuting every misdemeanor
unlawful border crossing would require a prohibitive outlay of additional
governmental resources.  Nevertheless, in
recent years the Department of Justice (DOJ) has made an effort to
significantly increase the number of immigration prosecutions.36  Consequently, the number of such prosecutions
ballooned—from just over 18,000 in 2001 to over 35,000 in 2007.37

1. 
Operation Streamline. — A significant portion of these
prosecutions have taken place under the auspices of Operation Streamline.38  Under the Operation Streamline program, all
unlawful entrants interdicted by Customs and Border Protection (CBP) in a
designated sector of the border region are criminally prosecuted.39 
In Tucson, Arizona, for example, about fifty to one hundred defendants
are prosecuted for illegal entry every single day.40 
The picture is similar in other jurisdictions where Streamline has been
implemented.41 
During these proceedings, defense counsel represents anywhere from six
to eight defendants to as many as thirty or forty defendants.42  Defense counsel typically converses briefly
with each defendant to establish whether they might have any defenses (such as
citizenship, authorization to enter, or claims of entering pursuant to a lawful
inspection), but if no such issue is raised, counsel generally participates in
the entry of mass pleas on behalf of his or her multiple clients.43

2.  “Fast
Track” Proceedings. — Illegal entry cases are not the only cases
fueling the upward spiral in immigration-related prosecution.  Previously removed individuals who are
apprehended after returning to the United States can be charged with felony
reentry—a crime that now carries a sentence of up to twenty years if the prior
removal was a result of conviction for an aggravated felony.44  Felony reentry prosecutions and convictions
are already rapidly on the rise throughout the United States.45  In districts that demonstrate to the DOJ that
they have “an exceptionally large number” of illegal reentry cases that will “significantly
strain[]” prosecutorial resources, the DOJ can authorize prosecutors to offer “Fast
Track” sentences in illegal reentry cases that are significantly below the
federal guidelines.46  As in the Streamline context, the relatively
light sentences generate ready pleas from the vast majority of those
apprehended and charged with felony reentry.

B. 
Postville:  Aggravated Identity
Theft Pleas

The routinized mass plea agreements that characterize border
justice are not limited to the southern border. 
Attorneys from ICE and CBP who have been cross-designated as “Special
Assistant United States Attorneys” (SAUSAs) have been assigned to the southern
border regions to assist with illegal entry and felony reentry prosecutions,
but they have also been dispatched to assist in criminal prosecutions at the
sites of interior workplace raids47 like the one that took place at the Agriprocessors plant in Postville, Iowa on
May 12, 2008.48

Of the approximately 1,000 workers in the Postville plant,
ICE officers arrested about 390 workers on the day of the raids.49 
A number of these arrestees were released for humanitarian reasons, but
the rest—just over 300 people—were detained for prosecution.50  Over the course of the next few days, 297 of
them pled guilty to aggravated identity theft based on their use of false
documents to obtain employment.51

In reality, many of those who pled guilty to aggravated
identity theft probably did not satisfy the mens rea requirement of the charge
because they had no knowledge that they were taking the identity of an existing
person.  The legal soundness of charging
identity theft in cases where there is no evidence of such knowledge was in
doubt even at the time of the Postville raids, and less than two years after
these events, the Supreme Court affirmed the lower courts that had required
such knowledge as an element of the federal identity theft provision.52

As with the border prosecutions,
however, the government proceeded on the theory—in hindsight, a seemingly
erroneous one—that few of the arrestees would have valid legal defenses to the
charges.  Yet the combined threat of
lengthy pretrial detention, coupled with the threat of a two-year prison
sentence, prompted almost all of the arrestees to plead guilty in exchange for
a five-month sentence.53 
The defendants who pled guilty were also ordered removed by the same
judge even though the provisions of immigration law allowing for such judicial
orders of removal did not actually apply in these cases.54  The Postville prosecutions are the clearest
example to date of the ways in which mass plea agreements can be deployed in
immigration-related prosecutions even outside of the border context.

C.  Diagnosing the Harms

Arguably, the mass plea agreements used as a means of
migration control can do little harm since many defendants probably have no
good legal defenses to the charges, particularly in cases involving entry
without inspection or felony reentry. 
Moreover, most defendants who have valid defenses such as derivative
citizenship or an invalid prior removal order will presumably be identified by
the defenders who meet briefly with the client prior to entry of the plea.55  Nevertheless, even if these assumptions are
true, these mass plea proceedings have a corrosive effect on the administration
of justice.  I address three specific
problems here, but there are many others.56

First, this approach ensures that abuses that take place at
the stage of investigation and detention are not addressed by the courts.  For example, with regard to Operation
Streamline, it may be fair to assume that many of the arrests made by CBP
officials comport with CBP’s internal regulations and the requirements of the
Fourth Amendment, but that is certainly not always the case.57  Yet, when allegations of misconduct surface
in the context of Streamline’s mass plea agreements, the affected migrant is
generally released and charges dropped—otherwise, the government has to take a
border patrol officer off the line to testify.58  With thousands of potential defendants
crossing the border each day, it is simply easier to drop charges against one
migrant than defend against allegations of constitutional violations.59  From a resource perspective, this makes
perfect sense.  The problem is that it
removes any possibility of deterrence through suppression in the course of
criminal proceedings.  Rogue agents have
a much greater chance of non-detection in this system, particularly because
there is almost no chance of an impoverished migrant bringing—much less winning—a
civil suit from outside of the country after removal.

Second, the proceedings engage the justice system in a
process that is, at best, tremendously dehumanizing.  Individuals picked up along the border, many
of them who have been in the elements for some time, are brought in shackles en
masse to a courthouse after a period of detention (usually overnight), often in
the same clothes they wore over the course of their journey.60  Their only individualized contact in a
foreign criminal justice system is a very brief conversation with a public
defender.  These prosecutions have
already changed the face of federal prisons. 
As a result of aggressive immigration prosecutions, Latinos are increasingly
overrepresented in U.S. prisons.61 
The “browning” of federal prisons62 ironically feeds the erroneous but
rampant perception that immigrants have a higher propensity to commit crimes,63 thus generating a feedback loop of popular pressure that drives even more
aggressive immigration enforcement.64

Finally, the group setting of the
Streamline and Postville style processes creates an inherently pressured
situation where individuals may well be reluctant to speak up to raise
individual concerns.  Such an effort
would be an aberration and would seem to run contrary to the preferences of
judges seeking to run expedient proceedings.65  Moreover, even if most pleas are legitimate,
serious questions remain as to whether all of these pleas are actually “considered
and intelligent.”66 

On December 2, 2009, the Ninth Circuit recognized some of these problems, concluding that the mass plea agreement procedures of Operation Streamline violated Federal Rule of Criminal Procedure 11, which requires the court to “address the defendant personally in open court . . . and determine that the plea is voluntary.”67 Reviewing the mass plea bargaining practice of Operation Streamline, the court concluded that “[n]o judge, however conscientious could have possessed the ability to hear distinctly and accurately fifty voices at the same time.”68 Nevertheless, in the absence of a finding of prejudice, the convictions on appeal were sustained.69 So while an appellate court has recognized that Rule 11 violations are taking place on a massive scale, it has concluded that, absent a showing of individual prejudice, the court will not intervene to correct the problem.

Even by the low standards of the American plea bargaining system, the proceedings discussed in this Essay seemingly lack the indicia of basic fairness that the Constitution and federal procedural rules purport to provide in criminal prosecutions.  Yet these proceedings have endured a change in administration, and have reshaped the federal criminal docket. Now, courts are showing little inclination to upset these practices, despite their acknowledged procedural flaws.
 

Conclusions:  A Cause
for Concern

The
retooling of the criminal justice system to manage migration has resulted in
some troubling trends.  In this brief
Essay, I have discussed some features of the plea bargaining systems by which
immigration convictions are obtained. 
Regardless of what one thinks of current restrictions on legal
immigration, the wholesale retooling of the criminal justice system to manage
migration that is evinced in these examples should raise a host of questions
that deserve serious and immediate attention: 
Is this an effective deterrent to migrants?70  Is it worth the monetary price tag?  Is it worth the procedural consequences?

The
ongoing erosion of the procedural rights of these criminal defendants thus far
has been effectively normalized.  Such
procedural moves can be framed as nothing more than an extension of
long-standing limitations on the due process rights of noncitizens in
immigration proceedings.  However, it is
important not to lose sight of the legal distinctions that separate the
criminal from the civil realm.  The
prosecution of these offenses should not be allowed to reshape the criminal
sphere to look more like the less rights-protective civil system where
immigration enforcement has typically been centered.  Unfortunately, at the moment, this is exactly
what is happening.

* Professor of Law, University of California,
Irvine, School of Law.  My thanks to Marc
Miller, Jack Chin, Kathie Barnes, Don Dripps, Carissa Hessick, Art LaFrance,
Doralina Skidmore and the other participants of the 2009 Southwest Criminal Law
Workshop for their useful comments on a sprawling piece that included this Essay, and to Law Librarians Adam Sexton and Dianna Sahar for their speedy
responses to my many requests.  Thanks to Dean Erwin Chemerinsky for his support of my research.  Finally, thanks to Jonathan Simon, whose book, Governing Through Crime: How the War on Crime Transformed American Democracy and Created a Culture of Fear (2007), is an obvious influence on this Essay.

1See, e.g., Jennifer M. Chacón,
Unsecured Borders:  Immigration
Restrictions, Crime Control and National Security, 39 Conn. L. Rev. 1827, 1827–32
(2007) [hereinafter Chacón, Unsecured Borders] (outlining “origins and
consequences of the blurred boundaries between immigration control, crime
control, and national security”); Daniel Kanstroom, Criminalizing the
Undocumented:  Ironic Boundaries of the
Post-September 11th “Pale of Law,” 29 N.C. J. Int’l L. & Com. Reg. 639, 640
(2004) [hereinafter Kanstroom, Criminalizing the Undocumented] (considering “convergence
between the immigration and criminal justice systems”); Stephen H. Legomsky,
The New Path of Immigration Law: 
Asymmetric Incorporation of Criminal Justice Norms, 64 Wash. & Lee
L. Rev. 469, 471–72 (2007) [hereinafter Legomsky, Asymmetric Incorporation]
(describing “growing convergence” of criminal justice and immigration control
systems); Teresa A. Miller, Blurring the Boundaries Between Immigration and
Crime Control After September 11th, 25 B.C. Third World L.J. 81, 83–86 (2005)
(describing consequences of interaction between criminal justice and
immigration law); Teresa A. Miller, Citizenship & Severity:  Recent Immigration Reforms and the New
Penology, 17 Geo. Immigr. L.J. 611, 616–20 (2003) [hereinafter Miller,
Citizenship & Severity] (describing criminalization of immigration law);
Juliet Stumpf, The Crimmigration Crisis: 
Immigrants, Crime, and Sovereign Power, 56 Am. U. L. Rev. 367, 376–77
(2006) [hereinafter Stumpf, Crimmigration Crisis] (noting “intense interest” in
“crimmigration law”).

2 See, e.g., Chacón, Unsecured
Borders, supra note 1,
at 1844 (noting 1996 immigration laws “impos[ed] a system of tough penalties
that favor removal even in cases involving relatively minor infractions or very
old crimes”); Kanstroom, Criminalizing the Undocumented, supra note 1,
at 640 (noting post-9/11 push to criminalize “unlawful presence in the United
States”); Legomsky, Asymmetric Incorporation, supra note 1,
at 476–82 (describing expansion of immigration-related criminal offenses);
Miller, Citizenship & Severity, supra note 1, at 617 (describing “stiff
criminal penalties” for immigration violations); Juliet P. Stumpf, States of
Confusion:  The Rise of State and Local
Power Over Immigration, 86 N.C. L. Rev. 1557, 1589–92 (2008) [hereinafter
Stumpf, States of Confusion] (describing imposition of criminal penalties for violations
of immigration-related laws).

3 See, e.g., Jennifer M. Chacón,
Whose Community Shield?:  Examining the
Removal of the “Criminal Street Gang Member”, 2007 U. Chi. Legal F. 317, 321–24
[hereinafter Chacón, Whose Community Shield?] (describing use of revised
immigration laws and removal proceedings to disrupt gangs); Kanstroom,
Criminalizing the Undocumented, supra note 1,
at 653 (“After the criminal justice system has completed its work, the removal
system begins.”); Legomsky, Asymmetric Incorporation, supra note 1,
at 482–86 (describing “proliferation of new crime-related deportation
grounds”); Miller, Citizenship & Severity, supra note 1,
at 614 (“[C]riminal grounds for deporting non-citizens
that were previously quite limited and enforced with laxity have been greatly
expanded in scope and are now strictly enforced through a variety of mechanisms
and institutional arrangements that have produced unprecedented cooperation
between criminal and immigration law enforcement.”).

4 See, e.g., Chacón, Whose
Community Shield?, supra note 3, at 339–43 (describing participation of
criminal enforcement officers in immigration proceedings); Anil Kalhan, The
Fourth Amendment and Privacy Implications of Interior Immigration Enforcement,
41 U.C. Davis L. Rev 1137, 1161–63 (2008) (describing increased role of state
and local officials in enforcement of immigration laws); Legomsky, Asymmetric
Incorporation, supra note 1, at 489–500 (describing importation of criminal law
enforcement strategies to immigration law); Stumpf, States of Confusion, supra
note 2, at 1595 (“Several post-September 11, 2001 federal actions have had the
effect of drawing state and local police into indirectly enforcing immigration
law.”); Michael J. Wishnie, State and Local Police Enforcement of Immigration
Laws, 6 U. Pa. J. Const. L. 1084, 1084–88 (2004) (discussing entry of civil
immigration violations into NCIC database).

5 See generally Legomsky,
Asymmetric Incorporation, supra note 1.

6 Id. at 472.

7 Legomsky discusses these
prosecutions, but they are not central to his analysis.  Id. at 481.

8 See Daniel Kanstroom,
Deportation, Social Control, and Punishment: 
Some Thoughts About Why Hard Laws Make Bad Cases, 113 Harv. L. Rev.
1890, 1899–913 (2000) [hereinafter Kanstroom, Hard Laws] (discussing early
history of migration regulation).

9 See sources cited supra note 2;
see also Hiroshi Motomura, Immigration Outside the Law, 108 Colum. L. Rev.
2037, 2087–88 (2008) (“Criminalization has been the trend since the 1990s, when
Congress increased penalties for existing immigration-related crimes, such as
smuggling and various types of document fraud, and added several new
immigration-related crimes.”).

10 Immigration Reform and Control
Act (IRCA) of 1986, 8 U.S.C. § 1324a (2006).

11 IRCA § 103(a).

12 Immigration Marriage Fraud
Amendment of 1986 § 2(d), 8 U.S.C. § 1325(c) (2006).

13 See, e.g., Anti-Drug Abuse
Act of 1988 § 7345, 8 U.S.C. § 1326 (2006) (increasing criminal sentences
for unlawful reentry); Violent Crime Control and Law Enforcement Act of
1994, Pub. L. No. 103-322, § 130001(b), 108 Stat. 1796, 2023 (codified as amended at 8 U.S.C. § 1326(b)(1) (2006))
(criminalizing reentry after commission of three or more enumerated
misdemeanors).

14 Illegal Immigration Reform and
Immigrant Responsibility Act (IIRIRA) of 1996, Pub. L. No. 104-208, Div. C. §§ 213–215, 110 Stat.
3009 (codified at 8 U.S.C. § 1324c(e) (2006), 18 U.S.C. §
1546(a) (2006), and 18 U.S.C. § 1015(e)–(f) (2006)).

15 See Jennifer M. Chacón, Misery
and Myopia:  Understanding the Failures
of U.S. Efforts to Stop Human Trafficking, 74 Fordham L. Rev. 2977, 2992–93
(2006) [hereinafter Chacón, Misery and Myopia] (discussing criminal provisions
added or enhanced by Trafficking Victims Protection Act (TVPA) of 2000).

16 See Legomsky, Asymmetric
Incorporation, supra note 1, at 496–98 (discussing IIRIRA encouragement of “use
of state and local criminal enforcement machinery to bolster the INS civil
immigration enforcement efforts”); Karla Mari McKanders, Welcome to
Hazelton!  “Illegal” Immigrants
Beware:  Local Immigration Ordinances and
What the Federal Government Must Do About It, 39 Loy. U. Chi. L.J. 1, 6–13
(2007) (discussing municipal legislation in Hazelton, PA, Altoona, PA, and San
Bernardino, CA); Michael A. Olivas, Immigration-Related State and Local
Ordinances:  Preemption, Prejudice, and
the Proper Role for Enforcement, 2007 U. Chi. Legal F. 27, 31–33 (“[F]rom
January through June, 2006, almost 500 immigration-related bills had been
introduced in state legislatures, and 44 had been enacted, in 19 states.”);
Christina M. Rodríguez, The Significance of the Local in Immigration
Regulation, 106 Mich. L. Rev. 567, 581–90 (2008) (discussing examples in
Illinois, North Carolina, and Iowa); Stumpf, States of Confusion, supra note 2,
at 1596–600 (discussing examples of local criminal enforcement in North
Carolina, Illinois, Oklahoma, California, Wyoming, New York, and Pennsylvania);
Rick Su, A Localist Reading of Local Immigration Regulations, 86 N.C. L. Rev.
1619, 1642–49 (2008) (using Town of Hazelton in Eastern Pennsylvania to discuss
theory of indirect regulation at local level).

17 See Ariz. Rev. Stat. Ann. §
13-2008 (Supp. 2008) (criminalizing “Taking identity of another person or
entity; knowingly accepting identity of another person; classification”); §
13-2009 (criminalizing “Aggravated taking identity of another person or entity;
classification”); § 13-2010 (criminalizing “Trafficking in the identity of
another person or entity; classification”).

18 See Stumpf, States of
Confusion, supra note 2,
at 1599 n.224 (citing examples from Oklahoma, Tennessee, California, Oregon, and
Wyoming).  An expanding number of states
and localities have also targeted employers and landlords who hire or enter
into contracts with unauthorized migrants. 
See id. (discussing ordinances in Escondido, California, Suffolk County,
New York, and Hazleton, Pennsylvania); see also McKanders, supra note 16,
at 6–13 (discussing ordinances in Hazelton, PA, Altoona, PA, and San
Bernardino, CA).

19 See Su, supra note 16,
at 1642 (describing trend as “indirect regulations of immigration”).

20 TRAC, Graphical Highlights:  DHS
Criminal Enforcement Trends (2005), at
http://trac.syr.edu/tracins/highlights/v04/dhstrendsG.html (on file with the Columbia Law Review).

21 TRAC, Graphical Highlights: 
Offenses Differ by Court (2005), at
http://trac.syr.edu/tracins/highlights/v04/dhsoffcourtG.html (on file with the Columbia Law
Review).  The vast majority of these
prosecutions are for illegal entry and illegal reentry.  Id.

22 Spencer S. Hsu, Immigration Prosecutions Hit New High, Wash.
Post, June 2, 2008, at A1.

23 TRAC, New Findings:  Department of
Homeland Security (2005), at http://trac.syr.edu/tracins/latest/131/ (on file with the Columbia Law Review)
(“[I]mmigration matters now represent the single largest group of all federal
prosecutions, about one third (32%) of the total.  By comparison, narcotics and drugs, for many
years the government’s dominant enforcement interest, dropped to about a
quarter of the total (27%) and weapons matters to slightly less than one out of
ten (9%).”).

24 TRAC, ICE Criminal Prosecutions Continue
to Rise Under Obama (2009), at
http://trac.syr.edu/immigration/reports/216/ (on file with the Columbia Law Review) (“[A]t least
through the first five months of the Obama Administration there has been no let
up in the increase in criminal prosecutions as a result of ICE’s enforcement
activities.”).

25 See, e.g., Chacón, Unsecured
Borders, supra note 1,
at 1847 (describing “new enforcement actions that . . . feed and
fuel the notion of dangerous classes of aliens”); Legomsky, Asymmetric
Incorporation, supra note 1,
at 479 (noting escalation of immigration-related criminal prosecutions
beginning in 1980s); see also Stumpf, The Crimmigration Crisis, supra note 1,
at 388 (“For the first time, immigration prosecutions outnumber all other types of
federal criminal prosecutions, including prosecutions for drugs and weapons
violations.”); Abby Sullivan, Note, On Thin ICE:  Cracking Down on the Racial Profiling of
Immigrants and Implementing a Compassionate Enforcement Policy, 6 Hastings Race
& Poverty L.J. 101, 117 (2009) (providing statistics showing “notable leap”
in immigrants serving federal prison terms).

26 For examples of literature
that examine the former phenomenon, see After the War on Crime 2 (Mary Louise
Frampton, Ian Haney López & Jonathan Simon eds. 2008).

27 For discussion of the
procedural and social consequences of the war on drugs, see, e.g.,
Erik Luna, Drug
Exceptionalism, 47 Vill. L. Rev. 753, 754–57 (2002) (discussing
procedural consequences of war on drugs); Tracey Maclin, Race and the Fourth Amendment, 51 Vand.
L. Rev. 333, 340–42 (1998) (addressing social impact of procedural
changes); William J. Stuntz, Local Policing After the War on Terror, 111 Yale
L.J. 2137, 2160 (2002) (noting war on crime transformed criminal procedure and
hypothesizing that war on terror will do same); Loïc Wacquant, The Place of the
Prison in the New Government of Poverty, in After the War on Crime, supra note 26,
at 27 (discussing social consequences of mass incarceration).

28 Until 1996, immigration
proceedings to prevent noncitizens from entering the country were termed “exclusion”
proceedings, while proceedings to remove a noncitizen that had already entered the
country were termed “deportation” proceedings. 
See Stephen H. Legomsky, Immigration and Refugee Law and Policy 420–21
(5th ed. 2009).  IIRIRA, Pub. L. No.
104-208, 110 Stat. 3009-546, Title III (1996), consolidated exclusion and
deportation, and labeled the resulting proceedings “removal” proceedings.  IIRIRA, Pub. L. No. 104-208, §§ 304, 308, 110 Stat. 3009, 3009–597 (codified at 8 U.S.C. § 1324c(e) (2006), 18 U.S.C. § 1546(a) (2006), and 18 U.S.C. § 1015(e)–(f) (2006)) §§ 304, 308.  Now, 8 U.S.C.
§ 1229a(3) indicates that the removal proceedings defined in that section are for determining “whether an alien may
be admitted to the United States or, if the alien has been so admitted, removed
from the United States.”

29 See Wong Wing v. United
States, 163 U.S. 228, 237 (1896) (distinguishing hard labor, which is
punishment, from deportation, which is not); Fong Yue Ting v. United States, 149 U.S. 698, 713 (1893) (suggesting congressional
power with regard to deportation is virtually limitless); see also
Kanstroom, Hard Laws, supra note 8,
at 1895 (discussing and contesting maxim).

30 INS v. Lopez-Mendoza, 468 U.S.
1032, 1038 (1984).

31 See id. at 1050 (holding
exclusionary rule generally does not apply in deportation proceedings, but
could apply if constitutional violations were “egregious”).  The Court also discussed the possibility of
applying the exclusionary rule in removal proceedings if violations became “widespread.”  Id. 
For arguments that this threshold has been reached, see Stella Burch
Elias, Good Reason to Believe: 
Widespread Constitutional Violations in the Course of Immigration
Enforcement and the Case for Revisiting Lopez-Mendoza, 2008 Wis. L. Rev. 1109,
1109 (arguing that “constitutional violations by immigration officers have
become . . . geographically and institutionally widespread in
the years since Lopez-Mendoza“);
Michael J. Wishnie, State and Local Police Enforcement of Immigration Laws, 6
U. Pa. J. Const. L. 1084, 1114 (2004) (noting that “[u]nder the logic of
Lopez-Mendoza itself, the exclusionary rule may now be appropriate in
immigration proceedings” given widespread evidence of racial profiling in
immigration enforcement).

32 Margaret H. Taylor, Dangerous
by Decree:  Detention Without Bond in
Immigration Proceedings, 50 Loy. L. Rev. 149, 150 (2004) (discussing uses of
immigration detention since September 11, 2001 to circumvent constitutional
limitations on detention).

33 See, e.g., Burch Elias, supra
note 31,
at 1114 (arguing for adoption of exclusionary rule in removal proceedings);
Wishnie, supra note 31,
at 1114 (same).

34 Not discussed here, but also
relevant, is the development of case law concluding that certain undocumented
migrants in the United States—such as felony reentrants—are not entitled to the
protections of the Fourth Amendment at all. 
See, e.g., United States v. Esparza-Mendoza, 265 F. Supp. 2d 1254, 1273–74
(D. Utah 2003) (denying motion for suppression of evidence under Fourth
Amendment due to alien defendant’s “lack of substantial sufficient connection”
to United States); see also M. Isabel Medina, Exploring the Use of the Word “Citizen”
in Writings on the Fourth Amendment, 83 Ind. L.J. 1557 (2008) (discussing
Fourth Amendment jurisprudence that incorrectly suggests right is limited to
citizens).

35 Jeffrey S. Passel & D’Vera
Cohn, Pew Hispanic Ctr., Trends in Unauthorized Immigration:  Undocumented Inflow Now Trails Legal Inflow,
at i (2008), available at http://pewhispanic.org/reports/report.php?ReportID=94 (on file with the Columbia Law Review)
(“[I]nflows of unauthorized immigrants averaged 800,000 a year from
2000 to 2004, but fell to 500,000 a year from 2005 to 2008 with a decreasing
year-to-year trend”).

36 John Grasty Crews, II, The
Executive Office for United States Attorneys’ Involvement in Immigration Law
Enforcement, U.S. Attorney’s Bull., Nov. 2008, at 1, 2, available at
http://www.usdoj.gov/usao/eousa/foia_reading_room/usab5606.pdf (on file with
the Columbia Law Review) (describing “increase[d]
misdemeanor prosecutions along the southwest border”).

37 Id.

38 See, e.g., Press Release, U.S.
Customs & Border Prot., Operation Streamline Nets 1,200-Plus Prosecutions
in Arizona (June 24, 2007), available at http://www.cbp.gov/xp/cgov/newsroom/news_releases/archives/2007_news_releases/072007/07242007_3.xml (on file with the Columbia Law Review)
(noting 1,200 of 2,800 immigration prosecutions in Yuma sector of Arizona took
place under auspices of Operation Streamline).

39 The designated sector is not
fixed, but changes over time, and usually covers a fifteen to twenty mile
stretch of the border region.  The notion
is that all entrants in that sector will be prosecuted, but because available
criminal detention facilities (and the courts) are not designed to accommodate
such a mass of pretrial inmates, apprehended individuals who pose particular
challenges (such as women and individuals who speak languages other than Spanish)
are often released without being prosecuted. 
Telephone Interview with Jon Sands, Fed. Pub. Defender, Dist. of Ariz.
(Oct. 21, 2009) [hereinafter Sands Interview] (notes on file with the Columbia Law Review).

40 Id.; see also United States v. Roblero-Solis, No.
08-10396, 2009 WL 4282022, at *1 (9th Cir. Dec. 2, 2009) (describing one particular mass
plea agreement and noting that “in twelve months’ time the court has handled 25,000″ of these pleas);

Brief of Defendant-Appellants at 6,
United States v. Roblero-Solis, Nos. 08-10396, 08-10397, 08-10466, 08-10509,
08-10512, 08-10543 (consolidated) (9th Cir. Apr. 6, 2009) (on file with the Columbia Law Review) [hereinafter Roblero-Solis Brief] (describing
Streamline plea proceeding); Oversight
Hearing on the Executive Office for United States Attorneys, Before the H.R. S.
Comm. on Commercial & Admin. Law of the H. Comm. on the Judiciary, 110th
Cong. 10 (2008) (Statement of Heather E. Williams, Federal Public Defender,
District of Arizona), available at http://judiciary.house.gov/hearings/pdf/Williams080625.pdf (on file with the Columbia Law Review)
[hereinafter Williams Testimony] (tracking prosecution caseloads after
implementation of Operation Streamline); David Bacon, Railroading Immigrants, The Nation,
Oct. 6, 2008, at 20 (same).

41 See Williams Testimony, supra
note 40, at 5–7 (discussing developments in Del Rio, TX, Laredo, TX, and
Yuma, AZ); see also David McLemore, Border Patrol Gets Tough in Laredo,
Dallas Morning News, Nov. 1, 2007, available at
http://www.dallasnews.com/sharedcontent/dws/news/world/mexico/stories/110207dntexlaredo.3584a01.html
(on file with the Columbia Law Review)
(describing twenty-seven defendants in South Texas Streamline prosecution “lined
up three deep” and entering plea in unison after brief discussion with federal
defender).

42 Sands Interview, supra note
39.

43 Id.; see also Roblero-Solis, 2009 WL 4282022, at *2 (describing
one such mass plea agreement);

Roblero-Solis Brief, supra note 40,
at 10–14 (describing mass plea system). 
Sentencing also takes place en masse, in smaller groups of fifteen or
twenty.  Id. at 15.

44 8 U.S.C. § 1326 (2006).

45 See, e.g., TRAC, Immigration
Convictions for July 2009, at
http://trac.syr.edu/tracreports/bulletins/immigration/monthlyjul09/gui/ (last
visited Nov. 13, 2009) (on file with the Columbia
Law Review) (noting convictions under § 1326 comprise about ten percent of
convictions by magistrate judges for immigration violations).

46 See Alison Seigler,
Disparities and Discretion in Fast-Track Sentencing, 21 Fed. Sent’g Rep. 299,
299, 303 (2009) (noting DOJ arguments in favor of Fast-Track program).  One result of the program is massive
sentencing disparities between Fast-Track and non-Fast-Track
jurisdictions.  Id at 299. 

47 Crews, II, supra note 36, at 3.

48 Nigel Duara, William Petrosky
& Grant Schulte, Claims of ID Fraud Lead to Largest Raid in State History,
Des Moines Reg., May 12, 2008, available at
http://www.desmoinesregister.com/apps/pbcs.dll/article?AID=/20080512/NEWS/80512012/1001 (on file with the Columbia Law Review).

49 Erik Camayd-Freixas,
Interpreting After the Largest ICE Raid in US History:  A Personal Account, 7 J. Latino Stud. 123,
125 (2009).  Warrants existed for an
additional 300 workers who were not present at the time of the raid and were
not arrested.  Id.

50 Id. 

51 See Julia Preston, 270 Illegal
Immigrants Sent to Prison in Federal Push, N.Y. Times, May 24, 2008, at A1 (“The
unusually swift proceedings, in which 297 immigrants pleaded guilty and were
sentenced in four days, were criticized by criminal defense lawyers, who warned
of violations of due process.”). 

52 Flores-Figueroa v. United
States, 129 S. Ct. 1886, 1888 (2009) (holding aggravated identity theft statute
requires showing that defendant knew identification used belonged to another
person).

53 Camayd-Freixas, supra note 49,
at 5.  But cf. Peter R. Moyers,
Butchering Statutes:  The Postville Raids
and the Misinterpretation of Federal Criminal Law, 32 U. Seattle L. Rev. 651,
673-74 (2009) (taking issue with Camayd-Freixas’s conclusion that pleas were “coerced,”
noting that this was not true in strict legal sense, although pleas “were the
product of a subtle systemic coercion”).

54 Moyers, supra note 53,
at 688 (noting entire plea arrangement relied upon misinterpretations of
federal identity theft provision and relevant removal provisions); see also
Sioban Albiol, R. Linus Chan & Sarah J. Diaz, Re-Interpreting Postville:  A Legal Perspective, 2 DePaul J. Soc. Just.
31, 64 (2008) (arguing reliance on judicial removal provision of
immigration law was inappropriate in these cases).

55 The amount of time spent by an
attorney with a particular client varies. 
In Arizona, public defenders typically represent six or seven defendants
at each Streamline hearing and are able to spend around fifteen minutes talking
to each client before a plea is entered. 
Williams Testimony, supra note 40, at 4; Sands Interview, supra note 39.  Resources are stretched more thinly in places
like the Southern District of Texas, where there are fewer panel lawyers to
represent Streamline defendants, and defense counsel can represent as many as
thirty to forty defendants in a proceeding. 
See Sands Interview, supra note 39; see also McLemore, supra note 41 (“27 people were brought before U.S. Magistrate Adriana Arce-Flores to enter a
plea for misdemeanor illegal entry.  A
federal public defender visited briefly with each one as they lined up three
deep . . . .”).

56 See Williams Testimony, supra
note 40, at 10–16
(listing numerous administrative costs, health risks and potential
constitutional violations resulting from Streamline system).

57 See, e.g., United States v.
Rangel-Portillo, No. 08-40803, 2009 WL 3429563, at *5 (5th Cir. Oct. 27, 2009)
(holding CBP lacked reasonable suspicion to stop defendant’s car and therefore
conducted illegal search).

58 Sands Interview, supra note 39. 

59 Id. 

60 One visitor to the courthouse
recalls being told by the bailiff to sit far away from the mass of defendants
because of “the smell.”  Interview with
Doralina Skidmore, President, Immigration Law Student Ass’n, Univ. of Ariz., in
Tucson, Ariz. (Oct. 3, 2009).  Defendants
“can smell pretty ripe.”  Sands
Interview, supra note 39.

61 Mark Hugo Lopez & Michael
T. Light, Pew Hispanic Ctr., A Rising Share: 
Hispanics and Federal Crime, at i (2009), available at
http://pewhispanic.org/files/reports/104.pdf (on file with the Columbia Law
Review) (noting by 2007, Hispanics accounted for forty percent of
federal prisoners, which was triple their representation in general population,
and tracing trend to rise in immigration prosecutions).

62 Ian Haney-López, Post-Racial
Racism:  Policing Race in the Age of
Obama 133 (U.C. Berkeley Pub. Law Research Paper No. 1418212, 2009),
available at http://ssrn.com/abstract=1418212 (on file with the Columbia Law
Review).

63 See, e.g., Ruben G. Rumbaut et
al., Debunking the Myth of Immigrant Criminality:  Imprisonment
Among First- and Second-Generation Young Men, Migration Information Source,
June 2006, at http://www.migrationinformation.org/Feature/display.cfm?ID=403 (on file with the Columbia Law Review)
(finding immigrants have lower rates of criminal convictions than
native-born Americans).

64 See Chacón, Whose Community
Shield?, supra note 3, at 348–49 (discussing ways in which migrants are
perceived as criminals and how those perceptions increase support for harsh
immigration laws); see also Legomsky, Asymmetric Incorporation, supra note 1, at
507 (describing disconnect between reality of immigrant propensity toward
criminal behavior and public opinion polls).

65 Williams Testimony, supra note
40, at 4.

66 Id. at 11.

67 United States v. Roblero-Solis, No. 08-10396, 2009 WL 4282022, at *7 (9th Cir. Dec. 2, 2009) (discussing requirements of Federal Rule of Criminal Procedure 11).

68 Id. at *8.

69 Id. at *9 (requiring showing of prejudice and finding “[n]one of these defendants has made such a showing or even attempted it”).

70 Although there are reports
that crossings are less numerous since the initiation of efforts like Operation
Streamline and the proliferation of workplace raids, see, e.g., Passel &
Cohn, supra note 35,
at i, the extent to which this is actually driven by enforcement-as opposed to
the economic downturn, or the rise in migrants who stay in the United States
without authorization once they have entered rather than risking multiple
border crossings-is difficult to ascertain. 
Id. at ii.  Given the fact that
Streamline, Fast Track, and the workplace prosecutions in raids like Postville
rely on vastly reduced sentences (in the case of Streamline, often
approximating time served) to induce plea agreements, it is not exactly clear
what deterrent effect is served by incarceration.  See also Williams Testimony, supra note 40,
at 17–22 (questioning Operation Streamline’s deterrent effect).

Preferred Citation: Jennifer M. Chacón, Managing Migration Through Crime, 109 Colum. L. Rev. Sidebar 135 (2009), http://www.columbialawreview.org/Sidebar/volume/109/135_Chacon.pdf.

Correspondence
accounts of the relationship between contract and promise hold either that
contract law is justified to the extent it enforces a corresponding moral
responsibility for a promise or unjustified to the extent it undermines
promissory morality by refusing to enforce a corresponding moral responsibility
for a promise.  In “The Correspondence of
Contract and Promise,” I claim that contract scholars have mistakenly presumed
that they can assess the correspondence between
contract and promise without first providing a theory of self-imposed moral
responsibility that explains and justifies the promise principle.1  To illustrate the dependence of
correspondence accounts of contract law on a theory of self-imposed moral
responsibility, I demonstrate how a “personal sovereignty” account of
individual autonomy—one of the most familiar and intuitive theories of
self-imposed moral responsibility—explains how and why, contrary to existing
correspondence theories, promissory responsibility corresponds to the objective
theory of intent, the doctrines of consideration and promissory estoppel, and most
remedial contract doctrines, including the bar against mandatory punitive
damages, the foreseeability limitation on consequential damages, the mitigation
doctrine, and expectation damages, the paradigm example of a contract doctrine
alleged to conflict with promissory morality. 
I conclude that correspondence theorists can defend their critiques of
contract law only by rejecting the personal sovereignty theory of self-imposed
moral responsibility, defending an alternative theory, and explaining why any
resulting divergence between contract law and its requirements is
objectionable.

The
personal sovereignty account of promising therefore plays a crucial role in my
analysis of correspondence theories of contract.  For purposes of that analysis, I described
how personal sovereignty explains promissory obligation:

[P]ersonal
sovereignty . . . recognizes the fundamental right of
individuals not only to choose their system of ends but also to choose how to
pursue those ends.  Promising constitutes
a particularly valuable means for pursuing ends. . . . [I]f
morality itself can provide individuals a valuable means of pursuing their ends
simply by recognizing the individual moral power to undertake self-imposed
moral responsibilities, a moral theory committed to personal sovereignty as a
fundamental moral value would have no grounds for refusing to recognize such a
power.  Personal sovereignty therefore
counts the moral capacity to undertake self-imposed moral responsibilities as a
basic individual liberty.  By affirming the fundamental right of
individuals to choose how to pursue their desired ends, personal sovereignty
necessarily affirms the category of moral responsibility that obligation
describes.  The moral power to make—and
thus the moral obligation to keep—a promise is therefore an axiom of personal
sovereignty.2

According
to this account of promising, individuals have the normative power to undertake
self-imposed moral responsibilities (i.e., moral obligations) because such a
power enhances personal sovereignty.3 
A moral theory with a foundational commitment to personal sovereignty
would therefore give moral effect to attempts to undertake such a
responsibility.  Although I find this an
intuitive understanding of the logic of moral justification, some philosophers
have doubted that morality can simply “give moral effect” to attempts to create
moral responsibility.  They argue that
this view begs the fundamental question of whether individuals have the
normative power to create moral responsibilities, as promises purport to do, by
simply communicating an intention to undertake such responsibility.  In Part I of this companion piece, I explain
the skeptical argument that has been leveled against other theories of
promissory obligation.  In Part II, I argue
that it has no force against the personal sovereignty account I offer.

I.  The Skeptical Argument

Joseph Raz’s
account of promissory obligation holds that promises are morally binding
because many intrinsically valuable special relationships are possible only if
they are.4  Raz’s view therefore holds that individuals
have the normative power to promise because such a power would be
valuable.  The contemporary version of
the skeptical argument rejects this argument. 
Thus, Michael Pratt argues: 

That
it is desirable to be able to bind oneself to another by means of communicating
an intention to do so, provides no reason to suppose that it is possible to
obligate oneself in that manner.  The
value of making binding promises does not, in other words, provide any reason
for thinking that the rule that promises ought to be kept is valid.5

Pratt
here echoes an objection that Don Regan years ago leveled against Raz’s account
of why consent is morally binding.  Regan
takes the claim that consent is morally binding to mean that consent provides
an individual with a reason for action that weighs in the ultimate balance of
his reasons for acting in accordance with his consent.  Regan explains, however, that Raz’s argument
proceeds from the premise that “[i]t would be a good thing if consent were
binding” to the conclusion that “[c]onsent is binding.”6 
But the argument form “It would be a good thing if X.  Therefore, X” is a nonsequitur.  Thus, Regan claims, the argument “It
would be a good thing if promises were binding. 
Therefore they are,” is invalid as well.7 
Although Regan acknowledges that no one believes the general argument
form is valid—no one believes that we can simply infer that X is true because
it would be a good thing if X were true—he speculates that philosophers believe
it is valid when applied to morality:

The
underlying idea would be that morality is not a set of facts about the
universe, but rather a set of ideas and practices we invent.  So, if we have a moral belief, and if it is a
good thing that we have that moral belief, that is all the warrant one could
possibly want for saying the moral belief is true.8

Regan
rejects this view, roughly, because he believes that whether an action is
morally right or wrong turns on moral facts independent of our beliefs about
our moral responsibilities.  For Regan,
an action is morally right or wrong because of its consequences:  “[P]eople ought to do acts which can be
expected to have good consequences.”9  The moral assessment of actions, therefore,
turns on whether they promote good or bad consequences.10 
Thus, Regan concludes that promises are not morally binding—they do not
provide the promisor with a reason for keeping the promise that always weighs
in the balance of his reasons for action. 
Instead, their moral force depends entirely on the consequences of
keeping them.  That the promisor promised
to do the act provides no independent reason for doing it.  Regan concludes that whether it would be a
good thing, from some point of view, that a promise created a moral obligation
to perform the promised act has no bearing on whether it in fact does create a
moral obligation.

David
Owens traces this fundamental skepticism about the moral force of promising
back to David Hume.  Owens reconstructs
Hume’s problem nicely:

What
makes breach of promise a wronging is that someone has communicated the
intention that it be a wronging . . . . Now something can
be declared to be wrongful in this way whether or not it is harmful or
constitutes unjust enrichment, or has any further feature in which human beings
might sensibly take an interest.  So such
wrongfulness raises the problem of bare wronging:  What sense is there in refraining from doing
something simply because it has been declared to be wrongful?  Conversely, how could bare wronging,
wrongings which have no adverse effect on anything that matters to us come into
being unless we do indeed have the power to create them by declaration?11

As
Owens explains, the puzzlement underlying this question stems from the
assumption that it “makes sense to do something because you are obliged to do
it only if the discharge of this obligation would serve some interest, where
the interest in question can be specified without using the notion of an
obligation.”12 
Most promise theorists presume that this interest must be a human
interest, something which it makes sense to want or value.13 
So the problem is to explain why keeping a promise serves some sensible
interest or value that human beings have. 
But unless the relevant interest or value is necessarily promoted by
keeping a promise, or necessarily undermined by breaking a promise, any account
of promising that traces its moral force to its effects on a distinct interest
or value will render promissory obligation contingent, dependent entirely on
whether keeping a promise in any given instance promotes that interest or
value.  So conceived, it seems impossible
to provide an account of promising that vindicates the common belief that
promises create moral obligations irrespective of the consequences of breach—that
promises always provide promisors with a reason to perform the promised act,
even if those reasons might sometimes be outweighed by other competing
reasons.  In short, the skeptical
argument originating with Hume and reformulated by Regan and Pratt claims that
promising could create a genuinely deontic obligation only if promise-breaking
necessarily undermines some independent human interest or value.14

II.  A Deontic Reply

In this
Part, I explain how the personal sovereignty account of promising I offer
provides a deontic foundation for promissory obligation that does not depend on
the effects of promise-keeping or promise-breaking.  In so doing, I explain why I reject the
implicit premise that underwrites Humean skepticism.

The
personal sovereignty account of promising has much in common with the account
of promising Raz defends.  According to
Raz:

[T]o
acknowledge the validity of voluntary obligations . . . is to
accept a rather unfashionable view of practical reason.  It is a view according to which what a man
ought to do depends not only on the ways things happen to turn out in the world . . . .
What one ought to do depends in part on oneself . . . [in part]
because the agent has the power intentionally to shape the form of his moral
world, to obligate himself to follow certain goals, or to create bonds and
alliances with certain people and not others. 
It seems to me that many have become so preoccupied with the way
considerations of human welfare affect what one ought to do that they become
blind to the existence of this other dimension to our practical life.15

I
share Raz’s conviction that promising is a crucial moral device for pursuing
one’s projects and creating special relationships.  This much, which few would doubt, is enough
to explain why individuals would care about having the power to undertake
self-imposed obligations.  But Raz also
argues that promises create moral obligations only “if the creation of . . .
special relationships between people is held to be valuable.”16 
As we’ve seen, for Raz, the justification of promissory obligation
depends on “the intrinsic desirability of forms of life in which people create
or acknowledge special bonds between them and certain other individuals.”17 
Thus,

The
right to promise is based on the promisor’s interest to be able to forge
special bonds with other people. . . . Those who assign sufficient
importance to the interest people have in being able to impose on themselves
obligations to other people as a means of creating special bonds with other
people believe in a right to promise. . . . [P]eople’s interest
in being able to bind themselves is the basis of a power to promise which they
possess and of an obligation to keep promises they make.18

The
personal sovereignty account, however, does not ground the normative power to
make, and the moral obligation to keep, a promise on its causal effects,
including their role in facilitating the pursuit of projects and forming
special relationships.  Instead, it
derives the normative power to make, and the moral obligation to keep, a
promise from the foundational normative premise that individuals are morally
entitled to decide how to live their lives as they see fit, consistent with a
like liberty for others.  The personal
sovereignty account understands Raz’s conclusion that “the agent has the power
intentionally to shape the form of his moral world”19 to follow not from the valuable
activities and relationships it makes possible but from the same prior
normative commitment to a conception of the individual as sovereign over all
matters exclusively affecting his own life. 
Just as individuals have the sole right to decide whether they will eat
meat, devote themselves to a meditative practice, or become a lawyer, so they
can decide for themselves whether to undertake a moral responsibility they are
otherwise free to avoid.

If
morality is committed to the value of personal sovereignty, then it affords
individuals the maximum morally permissible control over “the shape of their
moral world.”  To be sure, a moral theory
that recognizes the fundamental value of personal sovereignty cannot delegate
individuals’ control over the moral duties to which they are subject because these are grounded in the principle of equal
respect for the personal sovereignty of all individuals.  To enhance one individual’s personal
sovereignty by allowing him to avoid moral responsibility to others necessarily
and simultaneously diminishes respect for the personal sovereignty of the other
individuals to whom that individual would no longer be morally
responsible.  Moral duties therefore
define, rather than fall within, the realm over which individuals are
personally sovereign.  In contrast, by
recognizing the power of individuals to undertake moral obligations, morality enhances everyone’s control over their lives—their
power to “shape the form of their moral world”—without diminishing the personal
sovereignty of others.

Thus,
although this account of personal sovereignty does not rest on its role in
enabling individuals to realize valuable relationships or to pursue their
valuable projects, it is nonetheless animated by the same ideas that inform the
conception of autonomy that Raz embraces:

The
ruling idea behind the ideal of personal autonomy is that people should make
their own lives.  The autonomous person is a (part) author of his own life.  The ideal of personal autonomy is the vision of
people controlling, to some degree, their own destiny, fashioning it through
successive decisions throughout their lives.20

If
morality imposes duties and recognizes rights that it derives from the values
it takes to be fundamental, and personal sovereignty is among the fundamental
values morality affirms, then morality must recognize the duties and rights
derived from personal sovereignty.  The
ability to undertake self-imposed moral obligations enhances personal sovereignty
by affording individuals more control over the norms that apply to them.  A moral theory therefore cannot consistently
affirm the fundamental value of personal sovereignty and yet deny the power and
obligation of promising.

The
personal sovereignty account of promissory responsibility, however, appears to
commit precisely the fallacy that Hume, Regan, and Pratt have identified.  Having the normative power to create
self-imposed obligations by promising may well enhance personal sovereignty.  Yet the skeptical view denies that this constitutes
an argument to demonstrate that such a power exists.  As Regan puts the point, from the fact that
it would be desirable if something were true, it certainly does not follow as a
general matter that it is true.  Why
should this be otherwise when it comes to moral truth?  Perhaps personal sovereignty does not include
the power to undertake self-imposed moral obligations because self-imposed
moral obligations simply do not exist. 
One can no more demonstrate the existence of the normative power of promising
by observing that this power would enhance personal sovereignty than one could
demonstrate the existence of a million dollars in my bank account by observing
that this money would enhance my financial sovereignty.  What is needed is an argument explaining how
promissory obligation is possible, not an argument demonstrating why it would
be a good thing if it were.

The
skeptical argument proceeds, however, on the basis of a crucial suppressed
premise that Owens has identified:  “[T]he
problem of bare wronging arises only if we impose some substantive constraints
on what kinds of consideration can make sense of an action.”21 
Thus, many philosophers believe that an adequate account of promissory
obligation must provide an account that explains the moral force of a promise
in terms of more basic, normatively primitive, values and interests, such as
fairness, reciprocity, well-being, harm, and the like.  Hume accounts for promissory obligation by
explaining its role in providing valuable social coordination.  Regan would be satisfied by an account of
promissory obligation that demonstrated why making and keeping promises reduced
human suffering.22 
Pratt accounts for promissory obligation by explaining its role in
providing valuable assurance.23 
And even Raz ultimately traces the normative power of promising to its
role in facilitating valuable relationships. 
In this sense, all of these philosophers are deeply consequentialist
about the normative force of promissory morality.

The
personal sovereignty account, however, explains promissory morality not on the
consequentialist ground that it promotes some other moral value, but on the purely deontic ground that it derives from a fundamental moral value.24 
In Regan’s terms, it claims that personal sovereignty, understood as
entailing the power to undertake self-imposed moral responsibility, is a
fundamental moral value which no more stands in need of justification than does
the claim that reducing human suffering is morally good.  Does it explain, as Hume requires, how a
promise provides the promisor with a reason for action by explaining how
keeping the promise serves some human interest, without using the notion of
obligation?  Here I am once again
inclined to follow Raz’s lead:

[T]o
the extent that promises are a source of voluntary obligations they are made by
the exercise of normative powers.  The
obligatoriness of many promises can no doubt be explained on other grounds
which do not depend on the fact that promises yield voluntary
obligations.  But such explanations,
correct and useful as they are, miss the essential point in the common
conception of promises.25

Furthermore,

Because
all types of voluntary obligations are characterized by being mandatory norms
with content-independent justification, they are justified by the justification
of the general norm that promises . . . ought to be respected;
they are not justified by giving reasons for the desirability of each
obligatory act in its particular circumstances.26

Personal
sovereignty itself provides, in Raz’s terms, “the justification of the general
norms that promises ought to be respected.” 
The fundamental moral value of according individuals the maximum
permissible control over the moral norms that govern their lives explains why
individuals have the power to make promises and promises provide reasons for
action.  According to the personal
sovereignty account, promisors should keep their promises not because of the
consequences of performing or failing to perform the promised act or following
or breaking a general norm of promising, but because morality treats personal
sovereignty as a fundamental value that requires promises to be kept.  Simply put, promisors have reason to keep
their promises because morality requires that promises be kept.

* Visting Professor of Law, University of Michigan Law School; Robert E. Scott Distinguished Professor of Law, Professor of Philosophy, and Albert Clark Tate, Jr., Research Professor, University of Virginia School of Law. I thank Scott Hershovitz, Don Regan, and A. John Simmons for helpful comments and discussions.

1 Jody S. Kraus, The Correspondence
of Contract and Promise, 109 Colum. L. Rev. 1603 (2009).

2 Kraus, supra note 1, at 1609.

3 For the distinction between moral
obligations and duties, see id. at 1614 (“Moral
duties designate those responsibilities to which morality subjects individuals
solely by virtue of their status as moral agents alone, while moral obligations
designate those responsibilities to which morality subjects moral agents only
if they have voluntarily chosen to undertake them.  Unlike moral duties, moral obligations are
self-imposed.”).

4 According to Raz,

[A promise] creates
a special bond, binding the promisor to be, in the matter of the promise,
partial to the promisee.  It obliges the
promisor to regard the claim of the promisee as not just one of the many claims
that every person has for his respect and help but as having peremptory
force.  Hence, [promissory obligation]
principles can only be justified if the creation of such special relationships
between people is held to be valuable. . . .
[Promissory obligation] principles [depend] on the intrinsic desirability of
forms of life in which people create or acknowledge special bonds between them
and certain other individuals.

Joseph
Raz, Promises and Obligations, in Law, Morality, and Society:  Essays in
Honour of H.L.A. Hart 210, 227–28 (P.M.S. Hacker & J. Raz
eds., 1977) [hereinafter Raz, Promises and Obligations].

5 Michael Pratt, Promises,
Contracts and Voluntary Obligations, 26 Law & Phil. 531, 567 (2007).  Similarly, Pratt writes that “[e]lsewhere Raz writes that ‘promises are binding because it is desirable to make
it possible for people to bind themselves and give rights to others if they so
wish.’  If by ‘binding’
Raz means ‘morally binding’ then, again, the objection is manifest:  that it is desirable does not make it so.”  Id. at 567
n.84 (citation omitted) (quoting Joseph Raz, Voluntary Obligations and
Normative Powers, 46 Proc. of the Aristotelian Soc’y (Supplementary
Volumes) 79, 101 (1972))

. 

6 Donald H. Regan, Authority and
Value:  Reflections on Raz’s Morality of
Freedom, 62 S. Cal. L. Rev. 995, 1036–37
(1989).

7 Id. at 1037.

8 Id. at 1037–38. 

9 Donald H. Regan, Reasons,
Authority, and the Meaning of “Obey”:  Further
Thoughts on Raz and Obedience to Law, 3 Can. J.L. & Jurisprudence 3, 27
(1990) [hereinafter Regan, The Meaning of "Obey"].

10 For example, Regan suggests “[t]hat an act is one of relieving suffering is an intrinsic reason for doing the act—the relief of suffering matters in itself.”  Id. at 26. 
Regan does allow that promising might provide an evidentiary, and
therefore defeasible,
reason for performing the promised act if it could be shown that
promise-keeping is on average conducive to promoting good consequences.  But on this view, promises are only prima
facie binding.  They do not bind when the
promisor reasonably believes that performance does not promote good
consequences.  In such instances, rather
than providing a reason for action that is outweighed by other reasons, the
promise provides no reason for action at all. 

11 David Owens, The Problem with
Promising 7 (Feb. 12, 2009) (unpublished working paper, on file with the Columbia Law Review), available at
http://ssrn.com/abstract=1342060. 

12 Id. at 3.

13 Id.

14 I use the term “deontic”
here to describe any moral theory that does not reduce an action’s moral rightness or wrongness entirely to its
consequences.  See e.g., Larry Alexander
& Michael Moore, Deontological Ethics, in The Stanford Encyclopedia of Philosophy (Edward N. Zalta ed., 2007), at
http://plato.stanford.edu/entries/ethics-deontological/ (on file with the Columbia Law Review) (“[D]eontological theories are best understood in
contrast to consequentialist ones. . . .
[D]eontologists of all stripes hold that some choices cannot be justified by
their effects—that no matter how morally good
their consequences, some choices are morally forbidden. . . . For deontologists, what makes a
choice right is its conformity with a moral norm.”).

15 Raz, Promises and Obligations,
supra note 4, at 228. 

16 Id.

17 Id. 

18 Joseph Raz, The Morality of
Freedom 173–74 (1986) [hereinafter Raz,
Morality of Freedom].  Similarly, Raz
argues:

[T]he
power to promise and the right to promise are distinct notions.  But both stem from a common core, i.e. the
interest of persons to be able to forge normative bonds with others.  That is why they coexist, and one has the
power to promise if and only if one has the right to do so.

Id. at
174.

19 Raz, Promises and Obligations,
supra note 4, at 228.

20 Raz, Morality of Freedom, supra
note 18, at 369. 

21 Owens, supra note 11, at 7.

22 “That an act is one of relieving suffering is an intrinsic reason for doing the act—the relief of suffering matters in itself.  That is what we believe about the relief of
suffering . . . .” Regan, The Meaning of “Obey,” supra note 9, at 26 (internal quotation marks omitted).  However, Regan in fact offers no such account
and believes promises do not generate genuine moral obligations.

23 See Michael Pratt, Promises and
Perlocutions, in Scanlon and
Contractualism 93 (Matt Matravers ed., 2003).

24 Although the personal sovereignty
account of promissory morality is distinct from Kant’s, both derive promissory morality from a conception
of autonomy.  Kant argues that “freedom would be depriving itself of the use of its
choice” were it not possible to acquire
rights over “external objects of [one's] choice.”  See Immanuel Kant, The Metaphysics of Morals
68–69 (Mary Gregor trans., Cambridge
Univ. Press 1991) (1797).  According to
Kant, “another’s choice”
is included in the category of “external objects of choice” and called “contract
right”. 
Id. at 90–91.  Kant does not appear to claim that autonomy
specially requires the freedom to bind
oneself according to one’s will.  Instead, his claim is that it requires the
freedom to receive commitments from
others, and thus to have rights over their choices, analogous to our rights
over external goods (i.e., property rights), whose possibility is similarly
essential to full autonomy.  Thus, Kant
argues that we deprive the will of its full scope if we confine it to “internal”
objects of choice (i.e., our own actions), excluding external objects of choice
(such as other things and other people). 
Hence, for Kant, full freedom requires the possibility of property and
contracts.

25 Joseph Raz, Voluntary Obligations
and Normative Powers, 46 Proc. of the Aristotelian Soc’y (Supplementary Volumes) 79, 98 (1972).

26 Id.

Preferred
Citation:  Jody S. Kraus, Personal Sovereignty and Normative Power
Skepticism, 109 Colum. L. Rev. Sidebar 126 (2009), http://www.columbialawreview.org/Sidebar/volume/109/126_Kraus.pdf

In Rethinking Free Speech and Civil Liability,1 Daniel Solove and Neil Richards attempt something truly ambitious.  The authors seek to map coherent boundaries
for the First Amendment’s vast civil domain.  Their project merits serious attention.  Currently, different rules apply to civil
liability for speech depending on whether the liability arises in tort, contract,
or property.  Solove and Richards claim
that these boundaries are unworkable, under-theorized, and in some cases
destined to collide.  They develop a
framework for mapping the First Amendment’s
civil domain that is based upon a distinction regarding the type of power the state exercises in various
civil liability contexts.  This response
critically examines the choice and meaning of power, and the boundaries that a power-defining approach
would draw.

I.  Current Boundaries and Approaches

The
boundaries of the First Amendment’s
civil domain have not been systematically drawn.  The Court started mapping civil liability
boundaries in New York Times Co. v.
Sullivan,2 owing to the unique First
Amendment concerns raised by state libel laws. 
From that point forward, there appears to have been no master plan.  Indeed, the present boundaries might well
have been quite different.  With respect
to access to certain private properties, for example, the Court was in favor of
First Amendment applicability just a few years before it ruled against it.3

As Solove
and Richards observe, the current boundaries have not been adequately
justified.4 
Most tort claims seem to have been reflexively brought within the First
Amendment’s domain, while most contract and
property claims have remained beyond these borders.  As Solove and Richards note, however, civil
liability boundaries often overlap and intersect.  For example, breach of confidentiality has
both tort and contract characteristics.5 
Which rule ought to apply?

As the
authors note, tort, contract, and property liability all may substantially affect expressive interests.  “Private” law, whatever its specific form, might dictate or
distort public discourse, suppress the free flow of information, and limit
opportunities for public exchange. 
Moreover, all civil liability emanates from the state.  By what logic or principle, then, are only
certain claims to be excluded from the First Amendment’s civil domain?

Solove
and Richards do superb work culling various proposed answers to this question
from existing shards of judicial reasoning and academic commentary.6 
They contend, however, that each of the approaches is conceptually or
theoretically flawed, and that none coherently explains the existing boundaries
of the First Amendment’s civil domain.  Solove and Richards attempt to synthesize the
vast landscape of civil liability under a single First Amendment
framework.  They urge that power be the new principal boundary
marker.  The authors claim that the First
Amendment is substantively applicable whenever “(1)
the government defines the content of the civil duty; and (2) the speaker
cannot avoid accepting the duty, or the government exercises undue power in
procuring the speaker’s acceptance.”7 
Claims that satisfy both elements of this definition are examples of the
exercise of “duty-defining power,” which the authors contend merits serious First
Amendment scrutiny.  All other civil
claims arise from the exercise of “non-duty-defining
power,” which does not trigger any First
Amendment scrutiny.8

Solove and Richards note that
under this framework, the general boundaries of the First Amendment’s civil domain would be largely unchanged.  Thus, enforcement of most tort duties would
continue to receive serious First Amendment scrutiny, while enforcement of most
contractual duties would receive none.9 
As explained below, the extent to which the power-defining approach
would redraw the boundaries of the First Amendment’s civil domain is debatable.  In any event, the authors’ principal goal is to offer a more coherent
justification for both the boundaries that exist and for the treatment of
claims located at the borders, where civil forms of action sometimes overlap
and intersect.

II.  Autonomy,
Consent, and State Power

Before turning to the
power-defining framework, I want first to consider one of the approaches the
authors reject—the “consensual waiver”
approach.  Where a speaker voluntarily
agrees not to speak, as in a confidentiality agreement, why should the First
Amendment apply to the enforcement of that promise?  The speaker has a strong liberty interest in
making such decisions.10 
This liberty interest plausibly explains some portion of the First
Amendment’s current civil landscape.  In particular, it seems to solve the vexing
puzzle of confidentiality claims.  As
Solove and Richards acknowledge, principles of consent and autonomy play a
significant role in their power-defining framework.  Indeed, the rather substantial influence of
autonomy on the power-defining framework (it affects both elements of the
definition of “duty-defining”) is such that one might wonder why a new approach grounded in “power” is necessary at all.

Solove
and Richards claim that the consensual waiver approach fails to take into
account the rights of audiences to receive information.  But their approach might be subject to the same
criticism.  Under the power-defining
approach, so long as private parties voluntarily negotiate expressive limits or
enter relationships in which a duty of confidentiality is implicit, putative
audience members have no cognizable First Amendment objection to the loss of
what may in some cases be information of vital public concern.  As a theoretical matter, rejecting a pure
autonomy approach at least allows for some consideration of audience interests.  As a practical matter, however, the switch to
power would seem to benefit audiences only minimally, if at all.

Solove
and Richards also claim that the consensual waiver approach permits the state
to effectively purchase silence from speakers. 
Their example is a cash-for-silence contract, under which the government
can purchase the suppression of criticism of its own policies.11 
But as Solove and Richards note, under the unconstitutional conditions
doctrine, such an agreement would be unenforceable.12 
The state may, of course, attempt to purchase or coerce silence in more
subtle ways.  The First Amendment is applicable, however, whenever the sovereign acts—whether as regulator, subsidizer, purchaser,
contractor, employer, or property owner. 
Further, as the authors note, waivers of constitutional rights are strictly
construed by courts.13 
In other words, concerns regarding the exercise of “undue power”
by the state are, to some extent, built into existing First Amendment doctrine.

Thus, a pure autonomy approach
helps to untangle the civil liability knot, specifically as it relates to the
problem of confidentiality.  The
principal weakness of the autonomy approach is its lack of
comprehensiveness.  If the goal is to
explain the relationship between free speech and civil liability in an
expansive sense, principles of consent and autonomy only advance the project so
far.  Autonomy principles shed important
light on one region of the civil liability landscape.  They cannot justify or explain the remaining
boundaries.  The question is whether a
power-based approach, modified by principles of speaker autonomy, has greater
explanatory power than an autonomy approach, modified by concerns regarding
state power.

II Duty-Defining Power and Civil Discourse

As
Solove and Richards observe, the object of line drawing in the civil liability
context is to identify instances in which state-sponsored civil actions pose
the greatest threat to free speech. 
Mechanically and theoretically, “power” is better suited to this task than autonomy.  Speech regulations are ordinarily viewed,
often quite skeptically, through the prism of power.  And some civil liability, as the authors
correctly note, is in essence a form of regulatory power.14 
But it is not, as the authors suggest, the mere imposition of any mandatory duty regulating social
conduct that seriously threatens the First Amendment.  Rather, as is true with regard to any speech
regulation, it is the character or substance of the duty that ought to
determine the degree of the First Amendment threat.  This, ultimately, is what separates many tort
and statutory speech rules from contract rules; it is also, as I will suggest
below, one of the things that sets property-based liability apart from other
forms of civil liability.

A.  Civil Liability as Regulatory Power

Solove
and Richards note that civil liability is most troublesome from a First
Amendment perspective “when it inhibits or tries to
direct public discourse.”15 
Accordingly, the authors are primarily concerned with “[t]he government’s
role in shaping the speaker’s expression,”16 specifically instances in which
the state is “dictating, distorting, or
suppressing the terms or content of public discourse.”17 
They seek to identify forms of government power that are “particularly dangerous and should be curtailed as
abridgements of free expression.”18 
This is the appropriate benchmark. 
The question is whether the power-defining framework draws boundaries
that faithfully track it.

Solove
and Richards convincingly establish that, as a function of speaker autonomy and
consensual waiver, contractual claims that do not involve the exercise of undue
state power or influence properly lie outside the First Amendment’s domain.  But
that leaves a substantial landscape of tort and statutory liability.  Words are a potential basis for a staggering
amount of civil liability.  The state
imposes countless mandatory duties that have some impact on the act of
speaking.  In addition to libel and
privacy, common law duties imposed under assault, negligence, alienation of
affections, interference with prospective economic advantage, and even trespass
to chattels torts, all may incidentally impact speech.19 
Under the power-defining framework, all of these actions, and presumably
any others not based upon consensual waiver, are deemed “particularly dangerous”
threats to public discourse and public debate.20  As a result, in common law actions the rule
must either be altered, as in the case of libel, or courts must engage in ad
hoc balancing.21

This
approach would formally constitutionalize substantial areas of the common
law.  That would certainly be consistent with some
recent trends, as evident in areas from punitive damages to prison
litigation.  But it bears emphasizing
that Sullivan, from which this
line-drawing exercise emanates, was an anomaly. 
The presumption, as Richard Epstein has noted, “should be in favor of the constitutional
permissibility of the common law rules.”22 
On this view, the rules ought to be altered or displaced by
constitutional principles only where truly necessary to preserve core First
Amendment rights and values.  We ought to be
looking, as the authors suggest, for the “cases where the government is using the civil
liability system in ways that are especially dangerous.”23

In
drawing their boundaries, Solove and Richards have plainly opted for certainty
over flexibility.  The authors are
extremely skeptical of state power, so much so that irrespective of the
particular content of the duty being imposed, they perceive a serious threat to
public discourse.  Anyone who has
struggled with the definitional and theoretical difficulties inherent in this
area can appreciate their choice. 
Moreover, persuasive negative First Amendment justifications counsel
skepticism of state power.24 
There is no question that civil liability can and often does regulate the act
of speaking.  The question is whether this form of regulation can generally be equated with state
suppression of public discourse.

Consider
the great mass of potential tort claims. 
One of the things that distinguishes tort from contract claims is that
tort law consists of extrinsically imposed obligations or directives that
specify “public norms of conduct.”25 
The communicative torts—that is, those that regulate
speech as a primary subject rather than one among many means of violating some
general duty—are, as the courts have justifiably held, particularly dangerous
to free speech.  As descendants of
criminal speech provisions, their provenance alone provides some reason for
special scrutiny.  Actions that permit
the state, through judges and juries, to evaluate and ultimately define the
boundaries of public civil discourse raise special First Amendment
concerns.  Robert Post has described
privacy torts, for example, as “civility rules” that define persons and communities.26  Some civil liability rules are committed to “the task of constructing a common community through
the process of authoritatively articulating rules of civility.  The common law tort purports to speak for a community.”27  When
they speak to the substance of public debate, civil liability rules deserve
special scrutiny.  For similar reasons, we ought to be wary of civil claims like intentional
infliction of emotional distress, which may facilitate suppression of something
as critical to free speech as political satire.

Absent
some First Amendment scrutiny of these claims, governments would essentially be
empowered, through the imposition of certain tort and statutory duties, to “maintain what they regard as a suitable level of
discourse within the body politic.”28 
A public civility code that rests upon common law or statutory claims is
as threatening to the First Amendment as a campus speech code or a law that
proscribes public utterance of derogatory or offensive words.29

Not all
mandatory duties pose this sort of threat, however.  For example, negligence law requires in many
contexts that a person warn others of foreseeable dangers.30 
Enforcement of a mandatory duty to warn “dictates” or compels speech. 
Under the power-defining approach, the defendant who fails to comply
with a duty to warn would presumably be entitled to some First Amendment “defense.”31 
But imposition of a duty to speak under these circumstances does not
seriously threaten First Amendment values. 
The duty to disclose or to warn not only aims to make us all safer, but
creates a more informed citizenry with regard to certain hazards.  This particular duty, although mandatory and
relating directly to the content of speech, does not seek to evaluate or define
civil discourse, or “speak for” a community. 
Nor does it implicate core First Amendment concerns regarding compelled
belief or state ventriloquism.

As
the example shows, the character of the duty matters. 

The
duties we ought to be most concerned with are those that evaluate and define
the substance of public discourse and debate. 
We ought to be especially wary of these civil claims owing to the
primary state interests they serve, namely protecting public audiences from
uncivil speech and shielding persons from various dignitary harms associated with public disclosure.  These purposes directly conflict with the
individualism at the core of the contemporary First Amendment.

It
is not simply that some liability rules specify, in very general terms, “the content of duties that private actors owe to
each other,”32 or that they create rules of
social conduct that may impact speech, which poses a “particularly dangerous”
First Amendment threat.  It is, rather,
what some tort and statutory standards do—or
are capable of doing—to individuals that marks them as
serious threats to free speech.  Certain
duties press and impinge upon speakers and speech in a manner and to a degree
that others do not.  Some communicative or
expressive duties aim principally to regulate what can be said to another.  Others specify how information can properly (“civilly”)
be obtained and shared with the public. 
These duties are not merely duty-defining;
they are person-defining and expressive community-defining in a much broader sense.  This is what renders defamation, false light,
and privacy torts far more serious threats to free speech than the duty not to
interfere with possession of one’s chattels, marital relations, or
prospective economic advantages.  In the
latter actions, moreover, speech often occurs in more private settings and is regulated
not for its own sake, but as one means of accomplishing some other forbidden
end.  It is thus difficult to
characterize these forms of liability as “cases where the government is using the civil liability system in ways that are especially dangerous.”33

To be
clear, I am not suggesting that speech concerns are absent in any of these
contexts.  But just as the authors would
have contract law play the principal role in assessing “coercion,”34 courts could apply tort and
statutory liability in light of free speech concerns without holding the First
Amendment fully “applicable” any time a mandatory duty is imposed.35 
Where the common law or statutory duty is not itself constitutionally
tainted, perhaps it would be best to allow states to experiment with, develop,
or repeal doctrines that implicate speech concerns.36

The
power-defining framework improves upon the autonomy approach by asking what is
unique, and uniquely threatening, about certain forms of civil liability.  We ought to conceive of “duty-defining power”
as regulatory power that not only undermines or eliminates speaker autonomy, but authorizes
an evaluative process by which the state dictates the substance of public
discourse.  So
characterized, duty-defining power is a form of censorship or suppression that
merits serious First Amendment scrutiny.

B.  Property Lines

When, as
suggested above, we measure the substance of a duty against First Amendment
values and concerns, the property lines drawn by the duty-defining approach
seem somewhat incongruous.  The tort duty not to
trespass onto the land of another is defined by the state, is mandatory, and
may indeed affect speech.  Although they
prohibit speakers from converting the private property of another into a speech
forum, thus affecting the location of expression, property rules do not
generally purport to evaluate or dictate the substance of public discourse.  If they are to be congruent, the First Amendment’s property lines ought to mark off places in which
the state arguably has some duty to facilitate expression.  It is in such places that property exclusions
pose the greatest threat to free speech.

Solove
and Richards reject the Hudgens rule,
which holds that the First Amendment is not technically applicable on private
properties.  They contend that the First
Amendment is also substantively applicable
whenever a civil no trespassing duty is enforced.37 
This means that a trespassing backyard or living room protester
possesses a First Amendment interest in expressing herself in that
location.  The authors are clearly
uncomfortable with this result, which conflicts with significant residential
privacy interests and the basic First Amendment principle that speakers do not
have a right to convey messages “whenever and however and wherever they please.”38 
They retreat to the position that a categorical rule, namely that the
homeowner’s interests always outweigh the
trespasser’s, may be appropriate.39 
Notably, by contrast, their approach would permit those in gated
communities, condominium associations, and other private associations to enact
and enforce substantial speech restrictions by covenant or agreement.40 
Putting aside which of these limitations is actually the greater threat
to free speech,41 the boundaries here seem
anomalous; the trespasser in a private community would enjoy some level of
First Amendment protection, while the residents of the community may enjoy none
at all.

Property
is indeed critical to free speech.  The where of speech can be just as important
as what may be said or how information may be
disseminated.  But in terms of First
Amendment values, not all places are of equal significance.  The greatest threats to free speech in terms
of property rules are the public forum and time, place, and manner doctrines,
which have resulted in increasingly diminished opportunities for expression and
exchange even in traditional public forums.42 
The First Amendment is undoubtedly both technically and substantively
applicable to regulations of public expression in these places; the problem
lies in the balance that has been struck between state and speaker interests.

Solove
and Richards correctly reject the traditional state action frame, which
obscures more than it elucidates, with regard to private properties.  The more appropriate question, as Mark
Tushnet has recently observed, is whether the government has a substantive duty
to provide or protect the right in question.43 
With respect to properties that are generally open to the public, tend
to be heavily subsidized by the state, and facilitate access to large public
audiences, one could plausibly argue that government has a duty to facilitate
and protect expressive rights.  Exclusion
from quasi-public venues like large shopping centers, which have replaced the
town squares and public streets speakers have largely abandoned or been
displaced from, poses the greatest threat to the First Amendment.  Increasingly, it is only in such places that significant public audiences can be found.44 
With regard to these properties, trespass enforcement may well be “duty-defining,”
in the sense that it suppresses public discourse on a substantial portion of our
expressive topography.  By contrast,
backyards and living rooms are not significant speech venues; restricting
access to such places has little to do with public debate.  The state has no duty to extend
speech protections over backyard fences or through front doors.

Property-based duties not to
trespass on, interfere with, or convert private property certainly “shape social conduct in ways defined by the state.”45 
But as with the duty to warn, this is not sufficient to render the First
Amendment fully applicable.  The
substance of these duties, which protect against interference with possession
or use of real property and chattels, seems rather far removed from concerns
regarding state censorship or suppression of speech.  That is not to say that no First Amendment
concerns arise where property rules exclude speakers from preferred
venues.  But again, there are ways to
address such concerns short of imposing First Amendment standards on all
private properties.46

Conclusion

Rethinking
Free Speech and Civil Liability will enhance critical thinking about the boundaries of the First Amendment’s civil domain. 
The power-defining approach is an impressive attempt to blend principles
of state power and speaker autonomy into a coherent and workable formula.  Solove and Richards successfully untangle the
confidentiality knot.  Their approach
resolves difficult borderline cases in which different standards sometimes
collide.  It is determinate without, as
the authors show through various examples, being rigidly categorical.  Moreover, by focusing on power, Solove and
Richards remind us that some liability rules can be as dangerous to free speech
as ordinary laws and regulations.  I have
raised questions about the extent to which we ought to constitutionalize
speech-related civil actions.  But
disagreement with regard to where the boundaries of the First Amendment’s civil domain ought to be drawn is perhaps
inevitable with a project of this scope. 
Solove and Richards may not have drawn perfect boundaries.  But the lines they have drawn, and more
importantly the reasons for them, are more coherent and determinate than those
that currently exist.

* Professor of Law, William & Mary Law School.

1 Daniel J. Solove & Neil M.
Richards, Rethinking Free Speech and Civil Liability, 109 Colum. L. Rev. 1650
(2009).

2 376 U.S. 254 (1964).

3 Compare Amalgamated Food
Employees Union Local 590 v. Logan Valley Plaza, Inc., 391 U.S. 308, 319–20 (1968) (holding nonemployee union members had
right to peacefully picket on property owned by mall), with Hudgens v. NLRB,
424 U.S. 507, 521 (1976) (holding speakers had no First Amendment rights at
private shopping center).

4 See Solove & Richards, supra
note 1, at 1652–54
(discussing cases of civil liability and free speech that have different
outcomes under First Amendment). 

5 Id. at 1669–70.

6 Id. at 1673–85 (discussing various approaches).

7 Id. at 1692 (emphasis omitted).

8 See id. at 1687–90 (explaining
distinction between duty-defining and non-duty-defining power).

9 As noted infra Part III.B, the
landscape with respect to property claims would be revised.

10 See Wooley v. Maynard, 430 U.S.
705, 714 (1977) (holding that First Amendment protects “the right to refrain from speaking at all”).  See generally C. Edwin Baker, Human Liberty and Freedom of Speech (1989) (offering theory
of free speech grounded in liberty and autonomy).

11 Solove & Richards, supra note
1, at 1690.

12 Id. at 1690–91.

13 Id. at 1677 n.149.

14 See Richard A. Epstein, Was New York Times v. Sullivan Wrong?, 53 U.
Chi. L. Rev. 782, 789 (1986) (“The Constitution speaks about
freedom of speech, and liability rules can tread upon that freedom as much as
direct regulation can.”).

15 Solove & Richards, supra note
1, at 1689.

16^. Id.

17 Id.

18 Id. at 1686.

19 State and federal statutes,
including employment and intellectual property laws, are also part of this
landscape.

20 Solove & Richards, supra note
1, at 1686.

21 Id. at 1696–97.

22 Epstein, supra note 14, at 791.

23 Solove & Richards, supra note
1, at 1697.

24 See Frederick Schauer, Free
Speech:  A Philosophical Enquiry 86
(1982) (emphasizing deep distrust of government power to regulate expression).

25 Benjamin C. Zipursky, Civil
Recourse, Not Corrective Justice, 91 Geo. L.J. 695, 755 (2003).

26 Robert C. Post, Constitutional
Domains:  Democracy, Community,
Management 56 (1995).

27 Id. at 67.

28 Cohen v. California, 403 U.S. 15,
23 (1971).

29 See Papish v. Bd. of Curators of
the Univ. of Mo., 410 U.S. 667, 670–71 (1973) (holding state cannot proscribe speech
or conduct that is merely “offensive to good taste”).

30 See, e.g., Tarasoff v. Regents of
the Univ. of Cal., 551 P.2d 334, 345 (Cal. 1976) (holding psychotherapist has
duty to warn third parties threatened by patients).

31 The defense would undoubtedly
fail.  Indeed, Solove and Richards might
argue for a categorical rule to that effect. 
That adjustment may be warranted; but creating categorical rules cannot
resolve questions regarding the accuracy or viability of the power-defining
approach.

32 Solove & Richards, supra note
1, at 1686.

33 Id. at 1697 (emphasis added).

34 See id. at 1701 (suggesting that
coercion “would be an issue for contract
law, not the First Amendment”).

35 See, e.g., Cucinotti v. Ortmann, 159
A.2d 216, 217 (Pa. 1960) (“Words in themselves, no matter
how threatening, do not constitute an assault.”).

36 See Elaine W. Shoben, Uncommon
Law and the Bill of Rights:  The Woes of
Constitutionalizing State Common-Law Torts, 1992 U. Ill. L. Rev. 173, 179 (arguing that constitutionalization of assault and other torts would be
unwise).

37 Solove & Richards, supra note
1, at 1698–99.

38 Adderley v. Florida, 385 U.S. 39,
48 (1966) (emphasis added).

39 Solove & Richards, supra note
1, at 1698–99.

40 See id. at 1700–01 (discussing restrictive residential covenants).

41 See Timothy Zick, Speech Out of
Doors:  Preserving First Amendment
Liberties in Public Places 159–61 (2008) (discussing free speech
implications of “gated communities” and other forms of privatization).

42 See id. at 53–59 (criticizing “judicial
bureaucratization” of public places).

43 Mark Tushnet, State Action in
2020, in The Constitution in 2020, at 69,
70 (Jack M. Balkin & Reva B. Siegel, eds., 2009) (“[T]he state-action doctrine is not really about what
the state does, but what it has a duty to
do.”).

44 See N.J. Coal. Against War in the
Middle E. v. J.M.B. Realty Corp., 650 A.2d 757, 779 (N.J. 1994) (recognizing,
under state constitution, that “if the people have left for the
shopping centers, our constitutional right includes the right to go there too,
to follow them, and to talk to them”).

45 Solove & Richards, supra note
1, at 1686.

46 Courts might, for example, tighten
the requirements for a prima facie case where speech concerns are present.  See, e.g., Intel Corp. v. Hamidi, 71 P.3d
296, 303–04 (Cal. 2003) (requiring
recipient of noncommercial spam email to prove actual damage to computer to
state trespass to chattels claim).

Preferred Citation:  Timothy Zick, “Duty-Defining Power” and the First Amendment’s Civil Domain, 109 Colum. L. Rev. Sidebar 116 (2009),

http://www.columbialawreview.org/Sidebar/volume/109/116_Zick.pdf.

I.  The Supposed Analogy to Obscenity

Analogies
between the First Amendment and the Second (and comparable state constitutional
protections) are over 200 years old.1 
District of Columbia v. Heller itself makes them,2 and they can often make sense.3

Such
analogies might, for instance, yield the conclusion that (1) most guns (like
most speech) are fully protected by the Second Amendment, subject to some
restrictions that leave open “ample alternative channels” for effective self-defense,4 but (2) some narrow categories
of valueless or marginal weapons (like some speech) are unprotected.  Distinctions between the two Amendments can
make sense, too, though I leave them for other articles.5

But Guns as Smut does something
peculiar:  It analogizes a core category
of private arms to one of the least protected
and marginal categories of speech (obscenity).6 
It’s hard to see any justification
for such an analogy, other than a purely instrumental one.

The
premise of the First Amendment’s obscenity jurisprudence is that
obscenity is historically recognized7 as one of the “limited areas”8 of speech that “lack any serious literary, artistic, political, or
scientific value,”9 and are thus “not protected by the First Amendment.”10 
Obscenity, at home or elsewhere, is a marginal category of speech that
lacks the full protection that most speech gets.  Because of this, selling, buying, and
possessing obscenity in public places can be outlawed, and long has been outlawed.  Only the special “solicitude to protect the privacies of the life
within [the home]” leads to the prohibition on
punishment for mere home possession of obscenity.11

None of
this analysis applies to guns. 
Possessing guns is traditionally legal. 
Guns do serve the self-defense value that the Court has found to be
embodied in the Second Amendment.  And, Heller held, ordinary guns are at the
core of “arms,” not on the margin.12

Even
carrying guns in public places is traditionally legal (though often with
license requirements),13 and serves the constitutional
value of armed self-defense.  But I need
not rely on that:  The premise of the
Court’s obscenity decisions is that
obscenity lacks constitutional value without regard to the place in which it
may be present,14 though it may not be suppressed
via intrusions into the home.  That
premise does not extend to private gun ownership under Heller.

And
naturally Guns as Smut’s unsound premise leads to unsound results.  If guns were really like obscenity, the
government would be free to ban the buying of guns and not just their public
possession.  Guns as Smut’s conclusion indeed suggests that it “remain[s]
unresolved” whether the government could “so restrict[] the commercial availability of guns
that only guns in situ in the home, or those made by enterprising amateur
gunsmiths, would be beyond regulation”;
the Article’s interpretation of Heller “will
not, and cannot, provide [an] answer[]”
to that question.15

Yet Heller expressly holds that the Second
Amendment secures an individual right to possess handguns “for the core lawful purpose of self-defense.”16 
Whatever such a right might mean, it must include the right to
accomplish that core lawful purpose by acquiring the handgun.17 
No sensible interpretation of Heller can leave the status of that right “unresolved.”  And no
sensible analogy between the Second and First Amendments can analogize typical
privately owned arms to material that the Court has expressly held lacks First
Amendment value.

II.  The Supposed Analogy to Other Constitutional
Rights

Guns as Smut also claims that limiting the Second Amendment to
the home is supported by a special role for the home “located . . .
in ‘the Bill of Rights as a whole.’”18 
Yet most of the rights to which the Article points as support for that
proposition apply outside the home as well.

Warren
and Brandeis’s (nonconstitutional) “right of determining, ordinarily, to what extent [a
person's] thoughts, sentiments and
emotions shall be communicated to others,”
a right lost “only when the author himself communicates
his production to the public,” isn’t about the home: 
It would apply equally to thoughts communicated privately outside the
home as to those communicated within it.19 
The right “to educate one’s children”20 is usually exercised outside the
home (including in the very cases the Article cites, Pierce v. Society of Sisters21 and Meyer v. Nebraska22).  The right to decide “whether and when to have a family”23 is exercised partly by getting
contraceptives (or having abortions) outside the home, again including in one of the very
cases the Article cites, Eisenstadt v.
Baird.24 
Though the Fourth Amendment has been read as giving extra protection in
the home, dozens of cases make clear that it provides substantial protection
outside the home as well.25

Guns as Smut returns to the Fourth Amendment late in the Article, pointing out how convenient a home-only Second Amendment would be for
legislators who wouldn’t have to deal with any possible “fact intensive, arbitrary, and incomprehensible
[Second Amendment] doctrine akin to Fourth Amendment search and seizure
jurisprudence.”26 
Yes, radically limiting the scope of any constitutional constraint would
be convenient for legislators and courts. 
But fortunately, such a radical limitation (“privacy as smut”?)
has not happened with the Fourth Amendment. 
Despite the difficulties of articulating Fourth Amendment doctrine
outside the home, and despite the Amendment’s
textual reference to “homes,” courts have not thrown up their hands and concluded
that freedom from unreasonable searches and seizures should only cover the
home.

Two
rights do indeed apply only within the home, or similar places. Guns as Smut points to the Third
Amendment,27 but that Amendment is textually
limited to “house[s].”  The Article
also points to the right to consensual sexual activity,28 but this just illustrates why
analogies between constitutional rights, while often helpful, are often
limited.  A ban on public sexual activity
is, for nearly all people, a modest burden on the right, because it leaves
people free to shift to a private place. 
At most, it makes sex slightly less convenient and less
spontaneous.  (Exhibitionists might see
the restriction as burdensome, but the law doesn’t
treat such unusual tastes as deserving of accommodation.)

But
self-defense can’t be shifted to a more convenient
time or location.  You can’t invite a robber back to your place, where you
might have a gun available to defend against him, the way you can invite a
lover to your place to have sex.  To
borrow from the First Amendment’s “time,
place, and manner restrictions” doctrine, a ban on public sex
leaves open “ample alternative channels” through which one can derive the benefit of the
right—develop relationships, beget
children, and enjoy sex.29 
A ban on public possession of arms does not leave open ample channels to
defend oneself when the need arises.30

III.  The Supposed Historical Arguments for
Limiting Gun Rights to the Home

Guns as Smut likewise errs in its arguments that the right has
historically been seen as limited to the home. 
To take one example, the Article argues that “Blackstone lumped low or no-value speech acts and
the public carrying of firearms in the very same category in his Commentaries:  Libel and ‘challenges
to fight’ fall just behind riot, unlawful
hunting, and ‘riding or going armed, with
dangerous or unusual weapons’ as offenses to the public peace.”31

Blackstone,
though, was writing a brief summary of the Statute of Northampton, which he
rendered as “[t]he offense of riding or going
armed with dangerous or unusual weapons, is a crime against the public peace,
by terrifying the good people of the land.”32 
And the Statute was understood by the Framers as covering only those
circumstances where carrying of arms was unusual and therefore terrifying.

Thus,
Justice James Wilson—a leading drafter of the U.S.
Constitution,33 and the first prominent American
legal commentator—described the prohibition as
covering “a man arm[ing] himself with
dangerous and unusual weapons, in such a manner, as will naturally diffuse a
terrour among the people.”34 
This was a nearly literal quote from the leading English commentator
Serjeant Hawkins,35 who also went on,

[N]o
wearing of arms is within the meaning of this statute, unless it be accompanied
with such circumstances as are apt to terrify the people; from whence it seems
clearly to follow, That persons of quality are in no danger of offending
against this statute by wearing common weapons . . .
for their ornament or defence, in such places, and upon such occasions, in
which it is the common fashion to make use of them, without causing the least
suspicion of an intention to commit any act of violence or disturbance of the
peace. And from the same ground it also follows, That persons armed with privy
coats of mail, to the intent to defend themselves, against their adversaries,
are not within the meaning of this statute, because they do nothing in terrorem
populi.36

American
benchbooks for justices of the peace echoed this, citing Hawkins,37 though of course any class-based
limitation to “persons of quality” could no longer be sustained.38 
Only public carrying “accompanied with such
circumstances as are apt to terrify the people”
was thus seen as prohibited; “wearing common weapons” in “the common fashion” was legal.39 
And this is consistent with the pre-Civil-War American legal practice of
treating open carrying of weapons as not only legal but constitutionally
protected.40

IV.  Public Gun Possession as Threat to Public
Debate?

Guns as Smut also argues that “the
presence of a gun in public has the effect of chilling or distorting . . . public deliberation and
interchange. . . . [E]veryone is deterred from
free-flowing democratic deliberation if each person risks violence from a
particularly sensitive fellow-citizen who might take offense.”41

This is
an intriguing speculation.  One could
also engage in the intriguing rival speculation that people’s ability to defend themselves may support public interchange, by assuring
minority speakers that they can protect themselves against violent suppression.  Private gun ownership was sometimes used this
way during the civil rights era.42

But
fortunately we don’t need speculation; we have ample
experience.  In Vermont, people have long
been free to carry concealed weapons without a license.  In New Hampshire and the state of Washington,
law-abiding adults have been legally entitled to concealed carry licenses for
over 50 years.43 
Today, law-abiding adults can get such licenses in all states except
California, Delaware, Hawaii, Illinois, Maryland, Massachusetts, New Jersey,
New York, Rhode Island, and Wisconsin.44 
In many states, such as Arizona, Delaware, and Maine, law-abiding adults
may carry guns openly, even without licenses.45

Is public
debate on balance especially inhibited in any of these categories of
states?  Is speech in Arizona, Vermont,
and Washington less free than in Hawaii, Maryland, or New York, which try to
limit the supposed “smut” of guns to the home?  I know of no evidence for this, and Guns as Smut doesn’t point to any.

V.  Anxiety About Self-Defense

Guns as Smut also seems to express concern about public
self-defense more broadly.

The Article suggests that early American
law was “ambivalent”46 towards armed self-defense in
public.  No; there was no doubt at the
Framing of people’s right to defend themselves in
public using deadly force against murder, rape, or robbery, so long as the use
of deadly force was genuinely necessary. 
(The necessity requirement explains the “duty
to retreat” that’s imposed in many states—if one can defend oneself safely without killing,
then the killing isn’t necessary.47)

Nor did
Blackstone “specifically reject[] the Lockean
notion that a man had a right to kill an aggressor in public” when it came to genuine self-defense against
serious crime.48 
Blackstone expressly said that, “If any person attempts a robbery or murder of
another . . . and shall be killed in such
attempt, the slayer shall be acquitted,”
and likewise for “a woman killing one who attempts
to ravish her.”49 
Blackstone’s disagreement is simply with
Locke’s assertion that deadly force can
be used to resist “all manner of force,”50 including force other than
robbery, rape, or murder.  Justice Wilson
likewise made clear that homicide was protected “when
it is necessary for the defence of one’s
person,” and not just for the defense of
one’s home.51

The Article also argues that terrorist or revolutionary mobs may publicly carry
arms under the pretext of self-defense.52 
Yet the legal system has had, and should have, little difficulty
distinguishing individual citizens’
permissible legal possession for self-defense from mob action aimed at
attacking or terrorizing.  Past refusals
to suppress some mob action—which left victims to fend for
themselves, sometimes with their private arms53—have stemmed from deliberate governmental
underenforcement of normal criminal law, not from law-abiding citizens’ right to carry guns.

Finally,
the Article worries that a right to bear arms outside the home may lead to the
constitutionalizing of self-defense law.54 
And indeed the right to bear arms in self-defense has been used as a
support for the well-established criminal law of self-defense.  Justice James Wilson spoke of this in 1791,55 and courts have discussed it
since (as well as relying on explicit constitutional rights to self-defense
mentioned in state constitutions).56

But that’s not much of a horrible for Guns as Smut’s parade.  Even under the narrow Washington v. Glucksberg test, the Constitution protects
unenumerated “fundamental rights and liberties
which are, objectively, ‘deeply rooted in this Nation’s history and tradition.’”57 
The right to self-defense should indeed qualify under that test, even
without support from the Second Amendment. 
Nor should recognizing such a right cause much change in the law, given
that the core of the right—the right to use even deadly
force when necessary to protect against the most serious crimes—is uncontroversially recognized by statutes and the
common law.

* * *

I have
reached my allotted word limit, so I leave the rest of Guns as Smut’s 79 pages to others. Suffice it
to say that, whatever sound arguments there might (or might not) be for
limiting gun rights to the home, the arguments that Guns as Smut gives do not qualify.

* Gary T. Schwartz Professor of Law, UCLA School of Law (volokh@law.ucla.edu).

1 See,
e.g., Respublica v. Oswald, 1 U.S. (1 Dall.) 319, 330 n.* (Pa. 1788) (“The
right of publication, like every other right, has its natural and necessary
boundary; for, though the law allows a man . . . the
possession of a weapon, yet it does not authorize him to plunge a dagger in the
breast of an inoffensive neighbour” (relating statement by William
Lewis, then a Pennsylvania legislator and later a federal judge)); H.P.
Nugent, An Account of the Proceedings had in the Superior Court of the
Territory of Orleans, Against Thierry & Nugent for Libels and Contempt of
Court 43 (Philadelphia 1810) (“[A]s the liberty of keeping arms
is not the liberty of killing or maiming whom we please, so is not the liberty
of the press, the liberty of publishing libels” (relating statements by Judge
François
Xavier Martin, then a territorial judge and later Chief Justice of the
Louisiana Supreme Court)); Commonwealth v. Blanding, 20 Mass. 304, 314 (1825) (“The
liberty of the press was to be unrestrained, but he who used it was to be
responsible in case of its abuse; like the right to keep fire arms, which does
not protect him who uses them for annoyance or destruction.”).

2 128 S.
Ct. 2783, 2791, 2799 (2008).

3 See,
e.g., Eugene Volokh, Implementing the Right to Keep and Bear Arms for
Self-Defense:  An Analytical Framework
and a Research Agenda, 56 UCLA L. Rev. 1443, 1449, 1452, 1454–55
(2009); see also L.A. Powe, Jr., Guns, Words, and Constitutional
Interpretation, 38 Wm. & Mary L. Rev. 1311 (1997) (analogizing two
Amendments in the course of exploring whether Second Amendment secures an
individual right).

4 See,
e.g., Volokh, supra note 3, at 1454–59.

5 See,
e.g., id. at 1472, 1548.

6 Darrell
A.H. Miller, Guns as Smut:  Defending the
Home-Bound Second Amendment, 109 Colum. L. Rev. 1278, 1280 (2009).

7 See Roth v. United States, 354 U.S. 476,
483–84
(1957) (relying heavily on history in concluding obscenity is constitutionally
unprotected); Paris Adult Theatre I v. Slaton, 413 U.S. 49, 57, 61, 67, 69
(1973) (relying on Roth and on
history in reaffirming obscenity exception).

8 R.A.V.
v. City of St. Paul, 505 U.S. 377, 383 (1992).

9 Slaton, 413 U.S. at 67.

10 Id. at
54.

11 United
States v. 12 200-Foot Reels of Super 8mm. Film, 413 U.S. 123, 127 (1973)
(quoting Poe v. Ullman, 367 U.S. 497, 551 (1961) (Harlan, J., dissenting)).

12 District
of Columbia v. Heller, 128 S. Ct. 2783, 2791, 2817 (2008).

13 Steven
W. Kranz, A Survey of State Conceal and Carry Statutes, 29 Hamline L. Rev. 638,
647, 657 (2006).

14 12 200-Foot Reels of Super 8mm. Film,
413 U.S. at 126 (“[O]bscene material is not protected by the First
Amendment. . . . Stanley depended, not on any First
Amendment right to . . .
possess obscene materials, but on the right to privacy in the home.” (citing Stanley v. Georgia, 349 U.S. 557, 569
(Stewart, J., concurring))); Slaton, 413 U.S. at 67.

15 Miller,
supra note 6, at 1356.

16 Heller, 128 S. Ct. at 2797, 2818.

17 The
prohibition on banning possession of obscenity doesn’t
include the right to buy obscenity precisely because obscenity is seen as far from
the core of First Amendment protection. 
See, e.g., United States v. Reidel, 402 U.S. 351, 358–59
(1971) (Harlan, J., concurring).

18 Miller,
supra note 6, at 1305 (quoting Michael C. Dorf, Does Heller Protect a Right to Carry Guns Outside
the Home?, 59 Syracuse L. Rev. 225, 232 (2008)) .

19 Id. at
36 n.198; Samuel D. Warren & Louis D. Brandeis, The Right to Privacy, 4
Harv. L. Rev. 193, 198–99 (1890).

20 Miller,
supra note 6, at 1305.

21 268 U.S.
510, 536 (1925) (striking down law requiring public school attendance).

22 262 U.S.
390, 403 (1923) (striking down law prohibiting teaching of foreign languages in
school).

23 Miller,
supra note 6, at 1304.

24 405 U.S.
438, 454–55 (1972) (striking down law
prohibiting distribution of contraceptives to unmarried persons).

25 See
Miller, supra note 6, at 1304–05. 
For examples from just this past year, see Safford Unified School Dist.
No. 1. v. Redding, 129 S. Ct. 2633, 2635 (2009) (holding middle school search
of student’s underwear violated Fourth Amendment), and Arizona
v. Gant, 129 S. Ct. 1710, 1713 (2009) (finding search of car unreasonable under
Fourth Amendment where defendant was arrested for driving with suspended
license).

26 Miller,
supra note 6, at 1351.

27 Id. at
1304.

28 Id.

29 See
Volokh, supra note 3, at 1515, 1516 n.308.

30 Id.

31 Miller,
supra note 6, at 1308.

32 4
William Blackstone, Commentaries *148–*149.

33 See,
e.g., Ralph Rossum, Wilson, James, in Encyclopedia
of the American Constitution 2908 (Leonard W. Levy & Kenneth L. Karst eds.,
2d ed. 2000).

34 2 James
Wilson, The Works of James Wilson 654 (Robert McCloskey ed., Harvard Univ.
Press 1967) (1804) (printing one of Justice Wilson’s
lectures on law); id. at 67 (noting lectures were delivered in 1790 and 1791).

35 1
William Hawkins, A Treatise of the Pleas of the Crown 135, ch. 63, § 4
(photo. reprint 1978) (1716).

36 Id. at
136, ch. 63, § 9.

37 See,
e.g., William W. Hening, The New Virginia Justice, Comprising the Office and
Authority of a Justice of the Peace, in the Commonwealth of Virginia 17–18
(Richmond, T. Nicolson 1795); James Parker, Conductor Generalis; or the Office,
Duty and Authority of Justices of the Peace 11 (New York, John Patterson 1788);
see also Simpson v. State, 13 Tenn.
356 (1833).

38 See,
e.g., Hening, supra note 37, at 18 (“[T]he wearing of common weapons . . . merely
for ornament or defence, where it is customary to make use of them [is not
illegal].” (citing Hawkins, supra note 35, at 136, ch. 63, § 9)).

39 Hawkins,
supra note 35, at 136, ch. 63, § 9.

40 See
Volokh, supra note 3, at 1517–18 n.312.

41 Miller,
supra note 6, at 1309–10.

42 Robert
Cottrol & Ray Diamond, The Second Amendment:  Toward an Afro-Americanist Reconsideration,
80 Geo. L.J. 309, 355–58 (1991) (suggesting private gun ownership and gun
carrying emboldened some civil rights activists and deterred terrorists who
would have otherwise attacked the activists).

43 Kranz,
supra note 13, at 647, 657.

44 Id.;
Kan. Stat. Ann. § 75-7c03 (Supp. 2008) (statute enacted after Kranz article was published); Neb. Rev. Stat. § 69-2430
(Supp. 2008) (same).

45 Compendium
of State Firearms Laws, 217 PLI/Crim 203, 205 (2009)  (compiled by the NRA
Institute for Legislative Action, but submitted by Art Parker from the Office
of the Attorney General of the District of Columbia)

.

46 Miller,
supra note 6, at 1343.

47 See,
e.g., United States v. Travers, 28 F. Cas. 204, 206 (C.C.D. Mass. 1814) (No.
16,537).

48 Miller,
supra note 6, at 1343 n.410.

49 Blackstone, supra note 32, at *180–*181.

50 Id.

51 3
Wilson, supra note 34, at 84–85.

52 Miller,
supra note 6, at 1332–34.

53 See
supra note 42.

54 Miller,
supra note 6, at 1354.

55 3
Wilson, supra note 34, at 84–85.

56 See
generally Eugene Volokh, State Constitutional Rights of Self-Defense and
Defense of Property, 11 Tex. Rev. L. & Pol. 399 (2007).

57 521 U.S.
702, 720–21 (1997).

Preferred
Citation:  Eugene Volokh, The First and Second Amendments, 109 Colum. L. Rev. Sidebar 97 (2009),

http://www.columbialawreview.org/Sidebar/volume/109/97_Volokh.pdf.

Noah Zatz’s article, Managing
the Macaw:  Third-Party Harassers,
Accommodation, and the Disaggregation of Discriminatory Intent,1 is a brilliant addition to an important line of scholarship bringing
accommodation requirements into the fold of antidiscrimination law.  Using the colorful hypothetical of the
harassing macaw first introduced by Judge Easterbrook in Dunn v. Washington County Hospital,2 Zatz deftly shows that in cases
of third-party harassment employers must do more than merely treat their
employees equally; they must
affirmatively protect against harassment. 
Requiring employers to protect their employees from third-party
harassment—a position that is uniformly
accepted—is like requiring an
accommodation—a position that continues to be
highly contested.  In both cases, the
employer must avoid or correct a workplace harm that would have been caused by
an individual’s membership in a protected
group, and, in both cases, the employer bears this responsibility even if none
of its agents has treated individuals differently based on their protected
group status.

Despite
the importance of this contribution, I find myself much less willing than Zatz
to believe that a harasser in the workplace is ever like the hypothetical
macaw.  Workplace harassers, after all,
act within an organizational context created by the employer, and their victims
are treated differently at work on the basis of a protected
characteristic.  In my view, these
realities place third-party harassment much closer to what Zatz calls “internal”
membership causation than a biased police stop that results in a subsequent
work-related harm, or a construction drawing that omits ramps and elevators
from a work site.3 
Indeed, I am deeply troubled by Zatz’s
vague definition of internal causation (which he equates with disparate treatment)
as turning on whether the employee’s
protected trait entered the causal chain “through
the employer’s own decisionmaking process.”4  This seems to me far too narrow an
understanding of the employer’s role in the biased treatment of
women and minorities at work.

For
purposes of this brief Response, however, I will leave my concerns about the
precise contours of the distinction between internal and external membership
causation mostly to the side.  Instead, I
want to probe the role of agency principles in employment discrimination
law.  The distinction between direct and
vicarious employer liability for discrimination has been under-analyzed, and
this lack of attention, I think, leads Zatz to overstate the applicability of
his account—to suggest that it can be used to
determine the full extent of an employer’s
liability for actions by its agents.  In
doing so, Zatz puts the future of individual disparate treatment law at
risk.  He opens individual disparate
treatment law to considerations of employer “notice” and “feasibility” where it has traditionally imposed strict
liability.

I.  Overstating the Account:  Direct and Vicarious Liability

Zatz is on firm footing when he compares cases that
require employers to prevent and correct third-party harassment to cases that
require employers to accommodate disabilities, pregnancy, or acts of religious
observance.5 
The law in all of these cases requires the employer to protect against
workplace harms caused by membership in a protected group but resulting from acts
of individuals or groups that are not agents of the employer, or by
circumstances outside of the employment relationship (e.g., by a woman’s pregnancy or a person’s religious observance).  The employer cannot turn a blind eye to the
work-related harm caused by an individual’s
membership in a protected group on the ground that none of its agents
discriminated.  Zatz is on shaky footing,
however, when he extends his account to subordinate bias cases,6 for those cases involve biased
decisions made by agents of the employer, and they are therefore cases that trigger
employer vicarious liability, as well as the possibility of direct liability.

A.  Direct and Vicarious Liability Under Title
VII

Title VII
attaches liability to employers for their own acts of discrimination and for
the discriminatory acts of their “agents.”7 
When an employer adopts an express policy of discrimination, liability
attaches under the former theory.  The
employer is directly liable for its discriminatory policies.  In cases of individual disparate treatment,
in contrast, liability typically attaches under the latter, indirect
theory.  A manager denies a female worker
a promotion because she is a woman.  A
human resources officer declines to hire a black man because he is black.  Title VII holds the employer, not the manager
or the human resources officer, responsible for the discriminatory decisions in
these cases; furthermore, it does so regardless of the precautions taken by the
employer to prevent those discriminatory decisions and regardless of the
employer’s policies concerning
discrimination (regardless, in other words, of whether the employer would be
held directly liable).  The employer can,
of course, be held both directly and vicariously liable for the same action.  Think, for example, of an employer who is
vicariously liable for the tortious acts of its employee; it may also be
directly liable for negligent hiring of that employee.  Similarly, an employer might be both directly
and vicariously liable for a hiring decision made by a human resources officer
who has no policymaking power within the company but who declines to hire a
black man pursuant to company policy against hiring blacks.

As Zatz
explains, third-party harasser cases, unlike,
say, harassment by a supervisor, involve
only direct employer liability.8 
The third-party harasser is not considered an “agent” of the employer under principles
of respondeat superior.9 
To impose liability, then, the law must find a more primary route to the
employer.  The same is true of the other
third-party examples that Zatz identifies: 
discriminatory customer preference, discriminatory requests for employee
removal made to a temporary services agency, and an insurer’s discriminatory policy.10 
Courts have found employer liability in these cases through negligence,
a theory of direct liability.  The
employer is liable for its failure to take action to correct and prevent
discrimination by non-agents of which it knows or should have known.

The
accommodation cases also are typically analyzed as cases of primary liability—and this is the way that Zatz analyzes them. 
These cases are understood as cases of primary liability because the
individual decisionmaker, the person who denies the accommodation upon request
by the individual with a disability, is assumed to have acted in line with company
policy.  Moreover, the difficult question
in these cases is one of primary employer responsibility:  Should employers be expected to adopt policies
of accommodation?

Zatz
places subordinate bias cases in the company of the third-party harasser and
accommodation cases.11 
He is right that in determining the scope of primary employer liability
we can and should ask whether and to what extent employers should be expected
to prevent discriminatory employment decisions from occurring.  But, unlike the cases of third-party
harassment and accommodation, the subordinate bias cases involve biased acts by
agents of the employer.  This means that
both vicarious liability and direct liability are potentially in play.  In Zatz’s
view, as courts struggle to develop a coherent doctrine for resolving cases of
subordinate bias, they “are debating the extent of an
employer’s responsibility to insulate its
decisions from influences traceable to an employee’s race or sex”;12 they are deciding the contours,
in other words, of the employer’s primary or direct liability for
harm incurred because of biased actions of subordinate decisionmakers.  In the next section, I show that these cases
are better understood as vicarious liability cases.  They are cases in which courts hold (or do
not hold) the employer liable for the biased acts of its agents.  To the extent that courts or commentators see
these cases as only involving direct liability, they erroneously eliminate a
longstanding form of employer liability.

B.  Subordinate Bias Cases:  Delayed Action and Multiple Decisionmakers

As Zatz
explains, the prototypical case of subordinate bias is one where a middle
manager receives a report of employee misconduct from a subordinate who acted
with discriminatory intent in writing and submitting the report.13 
The report prompts the manager to review the entire personnel file of
the employee.  After conducting the
full-file review, the manager, who does not consider (and may not even know)
the employee’s race, decides to discharge the
employee.

Similar
cases can involve numerous variations on these facts, including variations that
do not involve biased decisionmaking by a subordinate of the ultimate
decisionmaker.  An immediate supervisor
of an employee might give an unduly negative, racially biased annual review of
that employee, and another supervisor might later deny the employee a promotion
based on the discriminatory review.  Or
an immediate supervisor might assign an employee to a particular task based on
the employee’s race, and another supervisor
might later discharge the employee based on the employee’s poor performance of that task.

Cases
like these raise several doctrinal difficulties for courts.  Because the initial discriminatory act—the one that involved discriminatory bias—cannot be immediately challenged under Title VII,14 there is a delay between the
biased decision and the adverse employment action challenged by the
plaintiff.  That delay can raise complex
statute of limitations issues.

But the
real detail vexing courts in these cases is the presence of multiple
decisionmakers:  An initial, biased
decisionmaker sets the wheels in motion, and a second, unbiased one takes the
adverse action challenged by the plaintiff. 
Notice that each of these decisionmakers serves as an agent of the
employer,15 and if the initial decisionmaker
who acted on his or her discriminatory bias had taken action sufficient to meet
the “adverse employment action” requirement, few would question the employer’s liability. 
The employer is vicariously liable in this instance for the
discriminatory action of its agent. 
Statute of limitations issues aside, the same should be true if all that
transpires between the initial discriminatory action and the adverse employment
action is delay.  For example, if a
manager later relies wholly and exclusively on a subordinate’s discriminatory review of an employee in firing
that employee, then the employer should be vicariously liable for that action.

The
addition of a second actor complicates this analysis only because events
subsequent to the initial discriminatory action, whether those events are
brought about by another agent of the employer or by the victim of the
initially biased decision herself, can attenuate the causal connection between
the initial discriminatory act and the adverse employment action.  For example, if, based on an initial
discriminatory performance evaluation by an immediate supervisor, a manager
undertakes an in-depth review of the employee’s
entire history of job performance and decides to
discharge the employee based on that in-depth review, then we might be less certain that the discriminatory
evaluation was a sufficiently immediate cause of the adverse employment action,
even as we accept that it was a but-for cause of the action.  Similarly, when an employee is
discriminatorily assigned a job and then fails to perform that job adequately
(assuming that the job duties are not discriminatorily stacked against the
employee), and the employee is fired for lack of job performance, the causal
nexus between the discriminatory decision and the adverse employment action
becomes attenuated, even as but-for causation remains.

I do not
attempt here to resolve the question of whether the employer should be liable
in these cases.16 
Instead, I want to highlight that the issue being debated is really one
of causation with respect to an individual decision—much like the debate that took center stage between
Justice Brennan and Justice O’Connor in Price Waterhouse v. Hopkins and was addressed by Congress in
the 1991 Civil Rights Act17—rather
than one of shaping the employer’s primary liability or, as Zatz
frames it, “the extent of an employer’s responsibility to insulate its decisions from
influences traceable to an employee’s
race or sex.”18 
If the employer is held liable for the adverse employment action in
these cases, it is being held vicariously liable for the actions of its agents.

It is
possible, of course, to imagine a direct liability inquiry in these cases.  But this is not the inquiry undertaken by
most courts.  Few cases, for example,
turn on whether the employer knew
about the biased action and/or whether the employer had put in place sufficient review processes to insulate promotion and
discharge decisions from biased performance reviews.19 
And, even if some courts do undertake such an inquiry, they are wrong to
do so unless they also understand that direct liability is an alternative to
vicarious liability.  If they instead
seek to reframe the plaintiff’s claim from vicarious to direct
liability, they substantially overreach and fundamentally alter the law of
employment discrimination.

II.  Why Overstating the Account Matters

Failure
to distinguish between direct and vicarious liability risks setting up
employment discrimination law for dramatic substantive change.  If existing vicarious liability is reshaped
as direct liability, particularly a duty-based direct liability that turns on
considerations of employer notice and feasibility,20 many individual instances of
discrimination will go unaddressed.  This
risk extends far beyond subordinate bias or other delayed action cases to
include all individual decisions that are motivated by bias but that are
considered, as Zatz would put it, “external” to the employer or not entering the causal chain “through the employer’s
own decisionmaking process.”21

Zatz
suggests that “’structural discrimination,’ in which systemic workplace practices facilitate
repeated acts of individual disparate treatment”
might fall in this category.22 
He thinks this presumably because he sees actions based on unconscious
biases as deriving entirely from social forces outside of work rather than, at
least in part, from factors within the workplace.  If he is right, then even decisions motivated
by animus or conscious stereotypes—once
considered the paradigmatic case of discrimination for which employers would be
liable—might be considered “external”
to the employer, and the employer will be held liable only if it failed to
exercise reasonable care to prevent or correct known acts of discrimination.23

The idea
that employer liability should be limited by considerations of employer notice,
feasibility, and the state of mind of upper-level managers who set company
policy is not new.  Employers have been
making this argument in various permutations for decades.24 
Extending the question of primary employer obligation into the realm of
actions taken by agents of the employer simply breathes new life into the
argument that courts have long rejected. 
It opens the door for courts to cut back on vicarious liability through
reframing the inquiry as one of direct liability.

One final
point, with which I think Zatz would agree but about which he is not entirely
clear:  Negligence is not the only or
even necessarily the best possible standard for employer direct liability.  In the individual context—where Zatz focuses his analysis—it might make sense to adopt a negligence-type
theory because these cases will involve individual acts of discrimination.  We might ask, “If
the employer is not strictly liable for this individual’s act, how else might the entity be liable for this
situation?  Maybe for failure to monitor,
failure to prevent, or failure to correct that individual’s behavior?”  This inquiry might naturally involve
considerations of employer notice about the individual’s behavior. 
But Title VII is not exhausted by individual disparate treatment; it
also encompasses systemic theories, including patterns and practices of
discrimination.  In the systemic context,
it is particularly important to recognize that direct liability need not hinge
on notice or the reasonableness of employer efforts, or even the state of mind
of high-level managers.  In a current
project, I examine these issues as they pertain to pattern or practice claims.25  It is my contention that the pattern or
practice claim is not based on vicarious liability, nor is it founded in
negligence; it holds employers directly liable for producing employment
discrimination.

Conclusion

Managing the Macaw is an important article.  It forges analytical connections between
disparate treatment and disparate impact theories of discrimination without
requiring group harm, and it pushes us to think more carefully about the bases
of employer liability available under Title VII.  My concerns with the article—although deeply troubling—lie more at the edges than at the core of Zatz’s argument.  I
agree with him that the questions of whether an employer should be expected to
police third-party harassment and whether it should be expected to provide
accommodation are fundamentally similar. 
The questions are not, however, identical, particularly taking into
account the fact that the third-party harasser acts within an organizational
context created by the employer, while the biased police officer making a
police stop, for example, does not.  This
issue of whether third-party harassment involves “external” or “internal” membership causation I leave to another day.  The danger explored in this Response is of
another kind and extends beyond the distinction between internal and external
causation; the danger is that courts and commentators will assume that
resolution of this issue determines the full extent of employer liability, even
in cases in which employers have long been held strictly liable under a theory
of respondeat superior.  I think Zatz is
likely to agree with me in this regard. 
We—commentators, courts,
legislators, and litigators—must be careful to distinguish
between the two forms of employer liability and to recognize when one or
another, or both, are applicable.

*Visiting
Professor of Law, USF Law School;  Professor
of Law, Seton Hall Law School. 

I owe thanks to Charles Sullivan, Michelle Travis, and especially to Noah Zatz for valuable comments and conversations.

1 Noah D. Zatz, Managing the
Macaw:  Third-Party Harassers,
Accommodation, and the Disaggregation of Discriminatory Intent, 109 Colum. L.
Rev. 1357 (2009).

2 429 F.3d 689, 691 (7th Cir. 2005)
(“Suppose a patient kept a macaw in
his room, that the bird bit and scratched women but not men and that the
Hospital did nothing.  The Hospital would
be responsible for the decision to expose women to the working conditions
affected by the macaw . . . .”).

3 See generally Tristin K. Green, A Structural Approach as Antidiscrimination
Mandate:  Locating Employer Wrong, 60
Vand. L. Rev. 849 (2007) (objecting
to position that there is no fundamental normative difference between
antidiscrimination and accommodation mandates and distinguishing the two based
on whether costs are imposed for employer wrongs in workplace).

4 Zatz, supra note 1, at 1377.

5 Id. at 1386–406.

6 Id. at 1422–27.

7 42 U.S.C. § 2000e(b) (2006) (defining “employer”
to mean “a person engaged in an industry
affecting commerce . . . and any agent of such a person”). 

8 Zatz, supra note 1, at 1380–82.  This “direct” entity liability for negligence might sometimes be
understood as a form of vicarious liability for the action of a high-level
decisionmaker, but the inquiry in these cases focuses on the entity’s knowledge and action, as controlled by high-level
decisionmakers, rather than on the knowledge and action of the employee who
engaged in the tortious act. 

9 Although courts have not imposed
vicarious liability on employers for third-party harassers, they have imposed
vicarious liability in similar circumstances in other contexts.  See,
e.g., Sword v. NKC Hosps., Inc., 714 N.E.2d 142, 150 (Ind. 1999) (describing “ongoing movement by courts to use apparent or
ostensible agency as a means by which to hold hospitals vicariously liable for
the negligence of some independent contractor physicians”).

10 Zatz, supra note 1, at 1416–22.

11 Id. at 1415.

12 Id. at 1426.

13 Id. at 1422–23.

14 Most courts require an “adverse employment action” before a discriminatory decision can be challenged
under Title VII, and many courts define an adverse employment action as one
that involves a material, economic change, such as a hiring, promotion, or
discharge decision.  See, e.g., Minor v.
Centocor, Inc., 457 F.3d 632, 634 (7th Cir. 2006) (defining adverse employment
action as involving material difference in terms and conditions of employment).

15 The initial decisionmaker, by
providing an evaluation or allocating job duties, is likely to have acted
within the scope of employment or to have been aided in the agency
relation.  The individual who engages in
these acts “brings the official power of the
enterprise to bear on subordinates.”  Burlington Indus., Inc. v. Ellerth, 524 U.S.
742, 761–62 (1998); see also id. at 762 (“The supervisor has been empowered by the company as
a distinct class of agent to make economic decisions affecting other employees
under his or her control.”).  Even if the initial decisionmaker is a
coworker, the official company act involved in subordinate bias cases
distinguishes those cases from cases involving coworker harassment.  See id.

16 There is a strong argument that
the employer should be liable upon the plaintiff’s
showing that his or her protected characteristic was a “motivating factor”
in the ultimate decision, both because that is what is required by the Civil Rights Act of 1991 § 703(m), 42 U.S.C. § 2000e-2(m) (2006), and because the existence of an “independent investigation” as a test for attenuated causation is so
indeterminate that it can be used to limit liability in almost all cases.  At the very least, we should expect a
defendant’s showing that it would have
taken the same adverse action had the initial biased decision been unbiased,
or that the causal connection is otherwise too attenuated, to affect only the
boundaries of remedial relief and not the liability determination.  Cf. McKennon
v. Nashville Banner Publ’g Co., 513 U.S. 352, 354 (1995)
(considering effect on plaintiff’s claim of after-acquired
evidence that would support same employment decision).

17 See Price Waterhouse v. Hopkins, 490 U.S. 228,
250 (1989) (Brennan, J.) (arguing that “the plaintiff who shows that an
impermissible motive played a motivating part in an adverse employment decision
has thereby placed upon the defendant the burden to show that it would have
made the same decision in the absence of the unlawful motive”); id. at 265–66
(O’Connor, J., concurring) (arguing
that only when plaintiff has shown “the
illegitimate criterion was a substantial factor in an adverse employment action” will burden shift to defendant to show it would
have made same decision anyway); see also 42 U.S.C. § 2000e-2(m) (superseding Price Waterhouse and stating liability is
established if plaintiff demonstrates that protected factor was “a motivating factor for any employment practice”); id. § 2000e-5(g)(2)(B) (stating
plaintiff’s remedies are limited if
defendant demonstrates it would have “taken
the same action in the absence of the impermissible motivating factor”).  The issue
is like the one debated in Price
Waterhouse, and resolved in the CRA of 1991, because it is about whether a
particular decision was caused by discriminatory bias; it is also different
because it is about the substantiality of discriminatory bias as a cause
rather than about whether discriminatory bias was a but-for cause. 

18 Zatz, supra note 1, at 1426.

19 Although courts use the term “employer,”
see, e.g., EEOC v. BCI Coca-Cola Bottling Co. of L.A., 450 F.3d 476, 488 (10th
Cir. 2006) (“[A]n employer can avoid liability
by conducting an independent investigation of the allegations against an
employee.”), the inquiry focuses on the
factual question of whether an independent investigation purged the decision of
bias rather than on whether the employer established sufficient safeguards against bias.

20 See Zatz, supra note 1, at 1403
(identifying “[n]otice and cost” as “leading considerations” in coupling membership causation to employer
responsibility); see also id. at 1415 (“[C]ountervailing
considerations involve notice and feasibility . . . .”).

21 Id. at 1377.

22 Id. at 1427 n.270.

23 This is also part of the danger
of Zatz’s vague, seemingly narrow
definition of “internal membership causation.”  See supra note
4 and accompanying text.

24 See, e.g., Slack v. Havens, No.
72-59-GT, 1973 WL 339 (S.D. Cal. May 15, 1973), aff’d as modified, 522 F.2d 1091 (9th Cir. 1975).  The defendant in Price Waterhouse v. Hopkins made a similar argument.  It argued that it could not be liable under
Title VII for its partners’ biased reviews of the plaintiff
because members of the Policy Board, the final decisionmaking body that relied
on the reviews in voting on the plaintiff’s
candidacy, did not act with purpose to discriminate.  See Hopkins v. Price Waterhouse, 825 F.2d
458, 468 (D.C. Cir. 1987) (“Price Waterhouse . . . contends
that Hopkins did not prove ‘intentional’ discrimination on the part of the Policy Board, but
only ‘unconscious’ sex stereotyping by unidentified partners who
participated in the selection process.”).

25 Tristin
K. Green, Discrimination, Asbestos, and Guns: 
Dangerous Analogies and the Future of Employer Liability for Patterns
and Practices of Discrimination (unpublished working draft on file with
author).

Preferred
Citation:  Tristin K. Green, On Macaws and Employer Liability:  A Response to Professor Zatz, 109 Colum. L. Rev. Sidebar 107 (2009),

http://www.columbialawreview.org/Sidebar/volume/109/107_Green.pdf.

Analogies are temperamental things.  If it strikes someone wrong, no matter how
scrupulously you explain yourself, no matter how defensible your position,
people who may otherwise agree with you half of the time never seem to get past
the analogy.  Arguments in hotly
contested areas of the culture wars tend to run against how the thing is
expressed, rather than what is expressed. 
Race, abortion, sexual orientation: 
Very often, discourse on these topics degenerates into debates about
legitimate ways to talk about the thing, rather than talking about the thing
itself.  The same phenomenon applies to talk
of guns.  Further evidence, in my
opinion, that Second Amendment discourse is not so much about guns or gun
policy, but “much ado about something else.”1

This is how I read Professor Volokh’s occasionally strident
response2 to my recent piece, Guns as Smut:  Defending the Home-Bound Second Amendment.3  Much of Professor Volokh’s rebuttal is a
mordant challenge to the accuracy of the analogy, rather than to arguments that
underpin the analogy and independently justify the home-bound Second
Amendment.  I gather that Professor
Volokh believes that if he can show that the facts supporting an analogy to
obscenity are faulty, then those same facts supporting a home-bound Second
Amendment must be faulty as well.  Fair
enough.

Nevertheless, I propose a thought experiment:  For those who flinch at the title, and can’t
get past the suggestion that a right to private possession and use of firearms
might look like the right to private possession and use of smut, read the piece as if
the analogy did not appear until the very end. 
Or, if you rather, read the piece in reverse.

The argument now goes roughly like this:  People want a voice in the debate on use of
guns in public, if polling is to be believed. 
The history of public bearing of arms for self-defense is deeply contested—especially
during the Reconstruction period when public bearing of arms was so politically
volatile.  Contrast this with the private
possession and use of firearms for defense of the home, which has almost
universal historical support for several centuries.  Judges generally should defer to political and local
branches of government when history does not provide definitive guidance as to
the scope of a constitutional right.  The
home is the one place where possession and use of arms has been universally
historically supported.  Therefore, the
federal constitutional right to keep and bear arms should be limited to the
home where the history is most certain, with political judgments or local
constitutions regulating (or protecting) guns everywhere else.  What other federal constitutional right is
limited to the home, with political judgments governing the right everywhere
else?

Obscenity. 

* Assistant
Professor of Law, University of Cincinnati
College of Law.  J.D., Harvard
Law School;
B.A., M.A., Oxford University; B.A. Anderson
University.

1 See Maxine Burkett,
Much Ado About . . . Something Else:  D.C. v.
Heller, The Racialized Mythology of the Second Amendment, and Gun Policy
Reform, 12 J. Gender Race & Just. 57, 58 (2008) (discussing gun debate as
method of racialized discourse); Dan M. Kahan, The Secret Ambition of
Deterrence, 113 Harv. L. Rev. 413, 451 (1999) (arguing gun debate is not about
guns or crime so much as it is about contesting groups’ attempts to use prestige
of law to confirm their worldview). 

2 Eugene Volokh, The First and
Second Amendments, 109 Colum. L. Rev. Sidebar 97 (2009), http://www.columbialawreview.org/Sidebar/volume/109/97_Volokh.pdf

.

3 Darrell A.H. Miller, Guns as
Smut:  Defending the Home-Bound Second
Amendment, 109 Colum. L. Rev. 1278 (2009).

Preferred Citation:  Darrell A. H. Miller, A Short Reply to Professor Volokh,
109 Colum. L. Rev. Sidebar 105 (2009), http://www.columbialawreview.org/Sidebar/volume/109/105_Miller.pdf.

Statistical
data are powerful, if not crucial, pieces of evidence in the courtroom.  Whether one is trying to demonstrate the
rarity of a DNA profile, estimate the value of damaged property, or determine
the likelihood that a criminal defendant will recidivate, statistics often have
an important role to play.  Statistics,
however, raise a number of serious challenges for the legal system, including
concerns that they are difficult to understand, are given too much deference
from juries, or are easily manipulated by the parties’ experts.  In
this preview piece, I address one of these challenges, known as the “reference class problem,” and sketch a solution that I develop at greater
length in my forthcoming Essay.1

I.  The Reference Class Problem

The
reference class problem arises from a basic observation:  When we make statistical inferences about a
specific case, those inferences depend critically on how we group or classify
that case.  To illustrate, imagine that
plaintiff contracts cancer after being exposed to a chemical spill of a known
carcinogen.  To establish that the spill
is the cause of her cancer, plaintiff attempts to show that her cancer risk
doubled after exposure.2 
So far, the litigation seems pretty straightforward, but then we face a
dilemma.  What statistic should we use to
estimate plaintiff’s cancer risk?  Should we use the risk for the general
population, or should we be more specific? 
White females under the age of fifty? 
Residents of Littleton County with no family history of cancer?  In other words, in describing cancer risk,
how should we break down the population: 
by age, gender, geography, profession, or something else?

In any
litigation, parties will invariably offer different classifications in the hope
of gaining some advantage.  To minimize
her background risk, our plaintiff may suggest using women under the age of
fifty with no family history of cancer as the relevant group.  In contrast, the defendant will focus on
other attributes, such as the fact that she is a smoker or takes hormone
supplements.  Faced with these
conflicting statistics, what is a jury to do? 
One natural response is to use all of the information about the
plaintiff—but that would result in a class
of one person, the plaintiff herself, and that singular class does not enable
us to make any statistical inferences at all.

The
reference class problem thus presents a serious issue.  The use of statistics is supposed to increase
objectivity and rigor, yet as I describe it above, statistics appear almost
infinitely malleable:  As long as counsel
manipulates the reference class sufficiently, he can arrive at any background
risk number he wants.  Indeed, rather
than fulfill their promise as a neutral basis for decisionmaking, statistics
suddenly appear to be nothing but rhetorical tricks that advocates can deploy
in court.

Worse
yet, this problem is not confined to toxic tort cases; it arguably infects
every use of statistics in the law.  For
example, when courts value property for eminent domain, taxation, or insurance,
one standard method is to look at comparable properties.  But which properties are in fact “comparable”
and what attributes of a home or lot should be used for the valuation?  The choice of reference class can affect the
valuation considerably.  In DNA cases,
prosecutors often emphasize the random match probability (RMP), the probability
that a person chosen at random from the population will have the same profile
as the one found at the crime scene. 
Yet, what population is appropriate for calculating the RMP?  The entire human population?  The defendant’s
racial subgroup?  The city in which the
crime occurred?

From an
intuitive standpoint, the above discussion may seem somewhat alarmist.  After all, just because one can manipulate
statistical inferences by cleverly selecting reference classes does not
necessarily mean that a jury will buy them. 
Using the category of white females under age fifty to estimate cancer
risk seems natural and hence legitimate. 
Using the category “women who own blue handbags, like
sushi, and drive a red sedan” does not.  But relying on the jury’s powers of intuition carries two problems.  First, mindful of the jury’s skepticism, the parties will never offer
outrageous reference classes.  They will
instead choose plausible (but still conflicting) ones to advance their
case.  Under these conditions, the jury’s intuitive judgment is largely unhelpful, and its
choices effectively arbitrary.  Second,
to rely solely on intuition is to surrender the goal of using statistics to
inject greater objectivity and rigor into legal decisionmaking.  As Ron Allen and Mike Pardo recently noted,
if reference class selection ultimately boils down to subjective and intuitive
judgment, then statistical models of evidence have not advanced the field by
much.3

But
what if we could find a way to make this intuition about the “reasonableness”
of reference classes more rigorous? 
Providing a principled method for choosing one reference class over
another would arguably solve the reference class problem, or at least restrict
its potential for mischief.  To add this
rigor, my proposal draws a close analogy to the model selection problem in
statistics and applies those concepts and methods to the reference class
problem.

II.  Model Selection

A
straightforward way to understand model selection is to consider the problem of
fitting a line or curve to a set of points.4  For example, assume we would like to predict
a student’s GPA based on the number of
hours he/she studies.  We collect the
data shown in Figure 1a, and then the question becomes, what exactly is the
relationship?  The most obvious answer is
a simple linear relationship, as in Figure 1b. 
However, the slight curve in the data points might suggest a quadratic
relationship, as in Figure 1c.  We can
fit even more complex curves, such as the fourth degree polynomial in Figure
1d.  In any event, we have multiple
candidates for models and no obvious principle for choosing one over another.

Figure
1:  Example Fits to Observed Data Points

We
can of course select curves based on intuitive judgment.  For example, the fitted curve in Figure 1d is
obviously overcomplex:  Study hours and
GPA are unlikely to be related in this way. 
Indeed, this kind of intuition may be what underlies the time-honored
principle of Occam’s Razor.5 
But intuition does not tell us how or why the curve is excessively or unnecessarily complex. 
Intuition is neither precise nor objective.  It can exclude the fourth degree model with
ease, but has a harder time choosing between the linear and quadratic curves.

The
statistics literature, however, does offer a more rigorous perspective on the
model selection problem.  Complex models
like Figure 1d are problematic because they are “overfitted.”  The problem
with overfitted models is that they erroneously incorporate the random noise
that accompanies real world data.  As a
result, the predictions they make become less accurate than if they had simply
ignored the noise.  In the GPA example,
if presented with a new set of students and their study hours, the overfitted
model will make more errors in predicting GPA than a simpler model.  So we have a classic tradeoff.  Too simple a model, and it will fail to identify
the underlying relationship and have low accuracy.  Too complex, and it will incorporate too much
random noise and be similarly inaccurate.

To
perform this balancing between fit and complexity, statisticians have developed
various model selection criteria.6  These
criteria operate as rating systems, allowing researchers to compare different
models and select the “best” one.

III.  A Solution

At
this point, the deep parallels between model selection and the reference class
problem are probably evident.  Overly
narrow reference classes are essentially overly complex models—they take into account too many attributes and run
the risk of incorporating noise into their estimates or predictions.  Conversely, overly broad reference classes
are like underfit models—they fail to incorporate enough
of the information in the data.

Indeed,
as I argue in the Essay, the reference class and model selection problems are
precisely one and the same.  As a result,
model selection criteria can solve the reference class problem for all
practical purposes in legal proceedings. 
Choosing a reference class need not be arbitrary, subjective, or
intuitive, but rather can be relatively objective and quantifiable.  Juries do have principles for selecting which
statistics to use to estimate a plaintiff’s
background risk, a house’s market value, or a DNA profile’s random match probability.

Predictably,
this claim is subject to a number of limitations, the most important of which
is that the proposed solution only eliminates the reference class problem in
the legal context.  The reason is that no one has yet figured out
how to find the single best model for a given phenomenon.  (That problem is exceptionally difficult, if
not impossible, to solve.)  But as
lawyers, we do not need to find the absolute best reference class to resolve
issues in court.  The adversarial system
only requires courts or juries to mediate disputes between the parties, so they
just need to decide whose proposed reference class is better.  And model selection
criteria perform that comparative function handily.

Conclusion

Beyond
the academic aspects of the proposed solution, my hope is that this project
will alert practitioners and courts to two fundamental things.  For practitioners, as one philosopher of
science aptly said, “the reference class problem is
your problem too.”7 
Whenever you encounter a statistic, think deeply about the underlying
reference class.  Changing the reference
class may change the statistic, and thus allow you to challenge your opponent,
make a powerful rhetorical argument, or in the best case scenario, affect the
outcome.  For courts, the lesson is that
the reference class problem is not as intractable as it first seems.  The choice of reference class need not be
left entirely to a jury’s subjective or intuitive
judgment.  Rather, statistical tools
exist for making reference class selection more analytical, a development that
will hopefully make statistics more welcome in the future.

* Professor of Law, Brooklyn Law School.  Many thanks to Elina Shindelman for research
assistance in preparing this Sidebar companion piece, and the Brooklyn Law School Dean’s Summer Research Fund for
generous support.  For a comprehensive
list of acknowledgments, please see Edward K. Cheng, A Practical Solution to the Reference
Class Problem, 109 Colum. L. Rev. (forthcoming Dec. 2009).

1 Edward K. Cheng,
A Practical Solution to the Reference Class Problem, 109 Colum. L. Rev. (forthcoming Dec.
2009). 

2 See 3 David L. Faigman et al., Modern
Scientific Evidence § 23:27, at 249 (2008) (“Most courts . . .
have concluded a plaintiff can reach a jury if she can present epidemiological
studies indicating at least a doubling of the risk of injury due to exposure to
a substance . . . .”).

3 See Ronald J. Allen &
Michael S. Pardo, The Problematic Value of Mathematical Models of Evidence, 36
J. Legal Stud. 107, 115 (2007) (“[T]he question of which [reference class
is better will] . . . be the subject of argument and,
ultimately, judgment.”).

4 See generally Walter Zucchini, An Introduction to
Model Selection, 44 J. Mathematical Psychol. 41 (2000) (offering short and less
technical introduction to concepts in model selection).

5 See, e.g., Lewis S. Feuer, The
Principle of Simplicity, 24 Phil. Sci. 109, 109 (1957) (“Entities are not to be multiplied unnecessarily.” (emphasis
omitted)). 

6 See
Kenneth P. Burnham & David R. Anderson, Model Selection and Multimodel
Inference 31–37 (2d ed. 2002) (discussing the need to balance fit and
complexity and various model selection methods).

7 Alan Hájek, The Reference
Class Problem Is Your Problem Too, 156 Synthese 563 (2007). 

Preferred Citation:  Edward K. Cheng, Law, Statistics, and the Reference Class Problem, 109 Colum. L. Rev. Sidebar 92 (2009), http://www.columbialawreview.org/Sidebar/volume/109/92_Cheng.pdf.

On June
25, 2009, the civil rights case of Rodriguez
v. City of Houston concluded with a historic verdict in the Southern
District of Texas, finding that pervasive deficiencies in the City’s police crime laboratory had caused George
Rodriguez to be wrongly convicted on the basis of fabricated scientific
evidence.1 
That same day, the Supreme Court held in Melendez-Diaz v. Massachusetts that the Sixth Amendment prohibits
the prosecution in a criminal case from placing crime laboratory reports into
evidence in lieu of live testimony by the crime lab analyst.2

Though
litigated in the distinctive realms of civil and criminal adjudication, both
cases directly engaged current policy debates surrounding the use and misuse of forensic science in criminal
investigations.  Justice Scalia’s opinion for the Court in Melendez-Diaz observed:  “‘[B]ecause
forensic scientists often are driven in their work by a need to answer a
particular question related to the issues of a particular case, they sometimes
face pressure to sacrifice appropriate methodology for the sake of expediency.’”3  This account of scientific testimony’s
systemic vulnerability to fraud echoed the very theory of liability considered
by the Rodriguez jury, summarized in the plaintiff’s closing statements:

[Y]ou
are trying to work
on this crushing caseload . . . your technical skills are not
what they’re supposed to be.  You’re
messing up on your proficiency tests, and you’re
not being trained in a [sic] better, faster more accurate serology techniques
used to do your job. And then, one of the detectives, or maybe one of the
prosecutors, “I need your results.  We got . . .
Rodriguez in custody.” 
And you look at your work and you realize that maybe the serology doesn’t implicate the suspect, but there’s all that other evidence your fellow police
colleagues have developed.  You could
redo the work—there’s another month of waiting at least or you can fudge—you can tailor the results.  What’s
the real harm?  He’s guilty anyway, isn’t
he?4

The
decisions emerge from different remedial contexts—Rodriguez a civil
damages action, Melendez-Diaz a criminal appeal—but both have the
potential to generate systemic incentives for law enforcement in the use of
forensic science.5  Increasingly, the Supreme Court has recognized this
dual role of both criminal and civil remedies in vindicating criminal procedure
rights and, in turn, regulating law enforcement practices.  At the same time, however, the Court’s
harnessing of constitutional criminal procedure’s regulatory effects has been
stilted by what I call “remedial rationing,” in which enforcement of a given
criminal procedure right is committed either to the criminal or the
civil realm.  This Essay posits that
remedial rationing is misguided both in underestimating the structural
limitations of criminal and civil litigation to achieve regulatory goals, and
in disregarding potential synergies that may be generated by recursive criminal
procedure remedies.  Part I of this Essay
briefly describes remedial rationing. 
Parts II and III reflect on both the potential and the limitations of
the Melendez-Diaz and Rodriguez decisions for prospectively
shaping law enforcement conduct in the specific area of forensic science.  The discussion demonstrates that systemic
consequences of criminal or civil adjudication of criminal procedure rights are
uncertain but potentially coordinate. 
Part IV concludes with a preliminary critique of remedial rationing in
light of the lessons of Melendez-Diaz and Rodriguez.

I.  Remedial Rationing in Criminal Procedure

Two
interrelated trends can be gleaned from the Supreme Court’s recent criminal procedure jurisprudence.  One is an increased recognition that enforcement
of criminal procedure rights is effectuated both by defendants in criminal
proceedings as well as by plaintiffs in civil actions under federal remedial
schemes—chiefly, 42 U.S.C. § 1983.6 
Yet this recognition of multiple mechanisms for enforcing criminal
procedure rights has been accompanied by a move to cabin enforcement within one
remedial context or another—an approach I call “remedial rationing.”

The logic
of remedial rationing has been most prominently deployed in the context of
Fourth Amendment claims, where restrictions on remedies in criminal litigation
is increasingly justified by the purported availability of civil suits under § 1983.  Thus, Hudson v. Michigan held that the
exclusionary rule was not a constitutionally required remedy for violations of
the Fourth Amendment “knock-and-announce” rule; the asserted availability of federal civil
rights suits was, in the Court’s view, an adequate mechanism to
deter unconstitutional police entry.7 
More subtly, the Court has also refashioned remedial doctrine to make
the burdens of proof for criminal defendants and civil rights plaintiffs
essentially identical, thus effectively streamlining two potential remedial
routes into one.  This approach was
highlighted this term in Herring v.
United States, which collapsed the criteria for invoking the exclusionary
rule in a criminal trial into the burden of proof for civil rights claims
against localities or police departments: 
In both instances a “widespread pattern of violations” must be shown in the jurisdiction.8

But
remedial rationing is not limited to Fourth Amendment litigation.  In
Polk v. Dodson, the Court held that
public defenders may not be sued for Sixth Amendment deprivations as “state actors”
under 42 U.S.C. §
1983, noting the
availability of alternate remedies for ineffective assistance of counsel:  either state malpractice claims or habeas
relief.9 
Justice Kennedy’s partial concurrence in Chavez v. Martinez, which rejected an
arrestee’s §
1983 claim for his interrogators’ failure to administer Miranda warnings, reasoned that “[t]he exclusion of unwarned statements . . . is a complete and sufficient
remedy.”10

Thus a
trend emerges:  Where criminal and civil
litigation both afford mechanisms for enforcing criminal procedure rights, the
Court is likely to channel enforcement into one regime or the other.  Moreover, the Court appears to embrace
remedial rationing in full cognizance of the regulatory implications of
criminal procedure rights, viewing alternative rather than recursive remedies
as generally adequate to deter undesirable law enforcement conduct.

A
number of factors may explain this dynamic. 
Raw realism suggests that remedial rationing reflects a fundamental
hostility toward enforcement of criminal procedure rights that has pervaded the
post-Warren Court.  A more
charitable reading of the Court’s decisions might identify legitimate
institutional cost concerns driving the rationing approach.  The Fourth Amendment provides the classic
example:  The exclusionary rule’s detractors,
focused on the cost of lost opportunities to prosecute blameworthy individuals,
have employed rationing to cabin the criminal remedy while still purporting to
achieve deterrent effects through civil enforcement.11  Whatever
the rationale—a question warranting fuller assessment than this brief Essay
affords—remedial rationing overstates the regulatory promise of criminal and
civil remedies in isolation, and underappreciates the advantages of a recursive
remedial regime.  This is illustrated by
the Melendez-Diaz and Rodriguez cases.

II.  Melendez-Diaz

The right
of confrontation announced in Melendez-Diaz will surely have a range of varied and mutually-reinforcing systemic
consequences.12 
Here let us evaluate only the claim that the right to confront crime
laboratory analysts, enforced by the threat of a lost prosecution, will
systemically deter bad practices in forensic science.

Borrowing
criminal law terminology, the arguments in favor of this position sound in “specific deterrence”
and “general deterrence.”  Specific
deterrence operates at the level of individual prosecutions:  The right of confrontation might deter
prosecutors from introducing weak or faulty evidence at any given trial; threat
of cross-examination may deter a given analyst’s
impetus to falsity or even negligence.13  General
deterrence concerns the impact of the confrontation right beyond a specific
criminal case:  Prosecutors anticipating
cross-examination might increase their vetting of forensic science evidence and
witnesses; discredited forensic methodologies vulnerable to cross-examination
(for example, bullet-lead analysis14) might fall into disuse.

But
deterrence of any sort depends upon enforcement’s
adequacy and effectiveness.  The right of
confrontation must be invoked frequently enough to affect incentives, and the
enforcement mechanism—cross-examination, or loss of a
prosecution—must effectively expose poor
science or spur better forensic science practices.  The ability of criminal adjudication to deter
bad forensic science practice is hampered on both scores.

As for
adequacy, confrontation can only occur at trial—an
increasingly rare occurrence.15 
In addition, the right will not always be invoked—either for tactical considerations by the defense16 or as a symptom of poor or
underfunded defense advocacy.17 
Some deterrent effects might nevertheless be generated.  Prosecutors want to win those cases that are tried, and the lack of ability to predict ex
ante when confrontation rights will
be invoked might prompt wholesale efforts to improve analytical and testimonial
practices.  Or the defense bar, armed
with the confrontation right, might devote more attention to training or
information sharing on cross-examination of forensic scientists.18  Perhaps. 
But perhaps more likely is that, instead of investing time and political
capital in greater oversight of crime laboratory practices and testimony,
police and prosecutors will simply reduce reliance on scientific evidence
altogether.  In a universe of more crime
than there is time to investigate and prosecute, law enforcement might swap
low-science (think property crimes) for high-science (think DUI) cases.  Prosecutors might also sweeten plea deals in
drug cases to avoid trials.  Whether such
responses would be “good” or “bad”
is debatable.  But certainly, they thwart
Melendez-Diaz’s potential to generate systemic incentives for
improved forensic science.

These
substitution effects also, of course, hamper “effectiveness” to the extent that the right of cross-examination
triggers work-arounds rather than changes in forensic science practice.  Additionally, where cross-examination does
occur, its actual utility as a mechanism for exposing flaws in forensic science
is suspect.  Consider the anecdotal
data.  The Melendez-Diaz dissent pointed to six jurisdictions that already
required “confrontation of the results of
routine scientific tests or observations.”19 
Two, Texas and Mississippi,20 have seen some of the most
serious instances of forensic science fraud. 
In Texas, Houston’s police crime lab suffered decades-long
deficiencies which were finally uncovered in an independent investigation
commissioned by the city in 2005, long after the Texas Court of Criminal
Appeals mandated confrontation.21 And in Mississippi
the controversial “bite mark” opinions of the forensic dentist Michael West were
given subject to the scrutiny of cross-examination.22

Further
permutations of potential systemic consequences of Melendez-Diaz could be posed. 
But these examples suffice to illustrate that, while some adjustments will surely occur in the
aftermath of Melendez-Diaz, a
systemic realignment of incentives in support of improved forensic science
practices is far from assured.

III.  Rodriguez

For
George Rodriguez, neither the constitutional prohibition against fabricated
evidence nor cross-examination at trial deterred or uncovered the false
scientific testimony that led to his conviction.23 
His verdict may remediate the damages he personally suffered through
prosecution and imprisonment, but will it deter poor or fraudulent forensic
science practices in the future?

Supporters
of the notion that civil rights suits may be mechanisms for systemic change
assert that damages actions “can induce the government to
change its policies to avoid further liability.”24 
To be sure, a variety of factors call into question the operation of “specific deterrence”
as the term is defined above.  For
example, significant time lags between the rights violation and the civil suit—a factor in Rodriguez and other actions by former defendants who must obtain vacatur of their
convictions before bringing suit—limit the effect of a verdict on
those actually responsible for the flawed practices or policies at issue.  Nevertheless, general deterrence may be
possible.  As a consequence of the Rodriguez verdict, Houston is on notice that inadequacies in
oversight of its crime laboratory may trigger significant monetary outlays as
well as negative political exposure.25 
Notably, Houston
is currently debating a proposal to establish a regional crime lab that is
independent from law enforcement control, and there has been significant
wrangling over the City’s commitment to continue to
support the crime lab reforms that were recommended by its independent
investigator.26 
Costs are undoubtedly at the forefront of these debates, and the Rodriguez verdict might add decisive
fiscal and political heft to the reform side of the balance sheet.

But, as
with criminal litigation, civil rights actions are vulnerable to scrutiny on
grounds of both adequacy and effectiveness. 
From the standpoint of adequacy, suits like Mr. Rodriguez’s are difficult to bring and difficult to win.  Government officials enjoy an array of
immunities from suit, with special consequences in the arena of criminal
justice:  Not only do police, like all
officials, possess “qualified” immunity for “reasonably” inflicted constitutional deprivations, but
prosecutors and judges are absolutely immune from liability for essentially all
trial-based constitutional harm.27 
Proving municipal liability is an equally daunting task. A plaintiff
must establish not only a constitutional deprivation, but also, generally, that
the violation was deliberate and that
the municipality caused or deliberately failed to act to prevent the violation.28 
These standards are high, the burden of meeting them requires extensive
and expensive discovery, and plaintiffs rarely prevail.29 
Moreover, many “victims” of bad forensic science in the criminal justice
system will simply be unsuitable plaintiffs: 
Those who remain convicted are unable to bring suit;30 and those whose convictions were
vacated for procedural violations but who cannot establish innocence will not
have provable damages for their incarceration. 
The small pool of available “private
attorney[s] general”31 and low odds of success mean
that suits like Mr. Rodriguez’s, standing alone, cannot be
relied upon to deter poor forensic science practice.

Where
viable civil rights claims exist, however, their effectiveness may be questioned. 
Some have challenged the assertion that municipalities respond to civil
rights verdicts like private rational actors, enacting reforms that are less
costly than adverse judgments.32 
The Rodriguez case highlights
certain aspects of this critique.  For
example, what size civil verdict is necessary to prompt Houston to take proactive measures in regard
to its use of forensic science in criminal investigations?  It is difficult, perhaps impossible, to know
whether Houston’s policymakers view Rodriguez’s five million dollar verdict as a cost worth avoiding
or one they are willing to budget—particularly
upon comparing the political palatability of a court-ordered payment to an
innocent man with expenditures on reform proposals aimed, in some measure, to
benefit criminal defendants.

Additional
factors no doubt affect the ability of lawsuits like Rodriguez to generate incentives for law enforcement’s use of forensic science.  The above points begin to illustrate,
however, that while civil rights litigation may fare better than criminal adjudication in the category of “general deterrence,”
civil suits standing alone are far from predictable mechanisms for shaping law
enforcement conduct.

IV. 
Assessing Remedial Rationing

The
above-discussed prospects for the Melendez-Diaz and Rodriguez cases,
though necessarily preliminary, form the foundation for a critique of the
Court’s remedial rationing approach to criminal procedure rights.  Let us lay to one side the interest of
individual litigants in remedies for constitutional violations (exclusion or
reversal in the criminal context, damages on the civil side), and consider only
the consequences of remedial rationing for regulating law enforcement.  I conclude that rationing overlooks the
structural limitations of each remedy for generating systemic incentives in
isolation, and that it diminishes the potential for positive synergistic
effects between the criminal and civil remedial realms.

The
first point has been amply illustrated by the above discussion of the variety
of ways in which criminal and civil enforcement of criminal procedure rights
might fall short of deterring undesirable law enforcement conduct.  The claim that the two remedies together
might have synergistic effects warrants further comment.  The above discussion suggests that criminal
litigation, while perhaps unlikely to generate predictable general deterrence, possesses comparative advantages over civil litigation in the realm
of specific deterrence: 
Constitutional violations are remedied relatively contemporaneously, and
the consequences of the remedy are fairly likely to fall directly on the actors
most immediately involved in the deprivation.  Additionally, in the aggregate, criminal
adjudication of criminal procedure guarantees is a relatively cheap mechanism
for generating data about law enforcement conduct in a given jurisdiction.

On the
flip side, we might plausibly suppose that civil litigation provides a superior
mechanism for general deterrence. 
Even setting aside questions about the economic incentives actually
created by civil damages,33 civil rights suits have the capacity to generate political incentives for prospective reform. 
Significant verdicts or settlements have the potential to generate media
and public interest; the fact that the litigant is a civil plaintiff with a
plausible claim to victimization rather than a criminal defendant obtaining
relief on a “technicality” only enhances the comparative advantage of civil
rights litigation in resonating with popular sentiment.  The process of
litigation often generates public data, sheds public light on government
practices, or, as illustrated by the previously quoted Rodriguez closing
statements,34 generates a
roadmap for reform.  Indeed, even the
threat of litigation may be seen as a political pressure point:  Municipalities subject to suit might
proactively address questionable practices in the aftermath of alleged misconduct in
order to recapture a moral high ground or prevent current policymakers from
exacerbating liability through ongoing “deliberate indifference” to
constitutional violations.35

Thus criminal and civil remedies potentially
generate coordinate advantages.  But they
are also interdependent in critical respects that are defeated by
remedial rationing.  For example, the
Court advanced the view in Herring that application of the exclusionary
rule might be limited to instances where a pattern of Fourth Amendment
violations could be shown—a standard that tracks the showing required for
municipal liability under § 1983.36  Yet in the
absence of any criminal remedy for the assertion of a Fourth Amendment
violation, there is no incentive for defendants to litigate the issue; lack of
adjudication of the right in the criminal context curtails the data collection
that is required ever to meet the “pattern of violations” standard in the civil
context.

In
short, both remedial opportunities are systemically important in generating
ex ante norms and incentives for law enforcement actors, and in deterring
policymaker inaction, particularly in regard to constitutional protections that
primarily benefit those without political leverage, i.e., criminal
defendants.  The choice between
developing and enforcing criminal procedure rights either in criminal
adjudication or in civil rights actions would likely create a suboptimal
quantity and quality of legal development and enforcement.

Conclusion

This brief tour has
not touched on institutional and constitutional issues raised by the Court’s
cabining of congressionally created civil rights remedies, nor has it
considered what principles, if any, should guide a court in limiting
adjudication to one or another remedial regime. 
These and other aspects of remedial rationing warrant further inquiry,
including beyond the substantive arena of criminal procedure. Nevertheless,
preliminary conclusions are borne out by the foregoing analysis.  The prospects for the Rodriguez and Melendez-Diaz cases to generate positive systemic incentives for law enforcement are
fundamentally intertwined—a dynamic which, I assert, pertains to criminal
procedure remedies more generally.  The
Court’s current trend toward remedial rationing in criminal procedure would
decouple that regulatory dynamic.  Such a
parsimonious regime of constitutional litigation is misguided. 

* Assistant Professor, The University of Texas School of Law. From 2006 to June 2009 I was an associate at Cochran Neufeld & Scheck, LLP, and one of the attorneys representing Mr. Rodriguez.

1 See generally Special Verdict Form, Rodriguez v. City of Houston, No. H 06-2650
(S.D. Tex. June 25, 2009), 2009 WL 1978620. 
The five million dollar verdict against the City of Houston was, to the
author’s knowledge, the first ever finding of municipal liability for
constitutional violations attributable to forensic misconduct.

2 129 S. Ct. 2527, 2532 (2009), available at http://www.supremecourtus.gov/opinions/08pdf/07-591.pdf (on file with the Columbia Law Review).

3 Id. at 2536 (quoting Nat’l Research Council of the Nat’l Acads.,
Strengthening Forensic Science in the United States:  A Path Forward 23–24 (2009)).

4 Transcript of Record at 6-1410, Rodriguez, No. H 06-2650 (on file with
the Columbia Law Review). 

5 See, e.g., Pamela S. Karlan, The
Paradoxical Structure of Constitutional Litigation, 75 Fordham L. Rev. 1913,
1915–16 (2007) (describing constitutional litigation in defense of criminal
prosecution and through affirmative civil claims as dual mechanisms for setting
conduct incentives for government actors); William J. Stuntz, The Uneasy
Relationship Between Criminal Procedure and Criminal Justice, 107 Yale L.J. 1,
22–53 (1997) (discussing structural interplay among criminal procedure rights,
substantive criminal law, and behavior of criminal justice actors).

6 See Pearson v. Callahan, 129 S.
Ct. 808, 821–22 (2009) (noting most constitutional issues raised in § 1983
cases are also raised in criminal cases). 
This term the Court decided five civil rights cases stemming from
criminal investigations or prosecutions. 
Safford Unified Sch. Dist. No. 1 v. Redding, 129 S. Ct. 2633 (2009); Dist. Attorney’s Office v. Osborne, 129
S. Ct. 2308 (2009); Ashcroft v. Iqbal, 129 S. Ct. 1937 (2009); Van de Kamp v.
Goldstein, 129 S. Ct. 855 (2009); Pearson v. Callahan, 129 S. Ct. 808
(2009).  Only one such case was heard in
the 2007 term.  Rothgery v. Gillespie
County, 128 S. Ct. 2578 (2008).  Three
were decided in the 2006 term.  Los
Angeles County v. Rettele, 550 U.S. 609 (2007); Scott v. Harris, 550 U.S. 372
(2007); Wallace v. Kato, 549 U.S. 384 (2007).

7 547 U.S. 586, 596–99 (2006); see also
Nix v. Williams, 467 U.S. 431, 446 (1984) (following similar rationale to
preclude application of exclusionary rule to Sixth Amendment counsel
violations); Guzman v. City of Chicago, 565 F.3d 393 (7th Cir. 2009) (observing
that “in some ways it is easier to protect Fourth Amendment rights through
civil actions, rather than through” the exclusionary rule).

8 Compare Herring v. United States,
129 S. Ct. 695, 704 (2009) (holding that denial of motion to suppress was
correct where no systemic errors in warrant system were shown), with Bd. of
County Comm’rs of Bryan County v. Brown, 520 U.S. 397, 407–08 (1997) (noting
that “a pattern of tortious conduct” may establish deliberate indifference
necessary for finding of municipal liability).

9 454 U.S. 312, 325 & n.18
(1981). 

10 538 U.S. 760, 790 (2005) (Kennedy,
J., concurring in part and dissenting in part).

11 See, e.g., Nix, 467 U.S. at 445 (noting exclusionary rule would “wholly fail[]
to take into account the enormous societal cost of excluding truth in the
search for truth in the administration of justice”).

12 See generally Stuntz, supra note 5 (describing effects of judicial intervention in
criminal procedure).

13 See, e.g., Brief for National
Association of Criminal Defense Lawyers et al. as Amici Curiae Supporting
Petitioner at 10, Melendez-Diaz v. Massachusetts, 129 S. Ct. 2527 (2009) (No.
07-591), 2008 WL 2550612 [hereinafter NACDL Amicus] (advocating confrontation
right as creating “ex ante incentive” for prosecutors to “look preemptively”
for problematic scientific evidence); Brief for the National Innocence Network
as Amicus Curiae Supporting Petitioner at 17, Melendez-Diaz, 129 S. Ct. 2527 (No. 07-591), 2008 WL 2550614
[hereinafter Innocence Network Amicus] (“[E]xaminers who realize there is a
possibility their work—or lack thereof—will be subjected to adversarial
scrutiny can be expected to think twice before making up results and tests from
scratch”).

14 See Innocence Network Amicus,
supra note 13, at 10
(describing report finding that “courtroom testimony claiming bullet fragments
could be matched to a particular box of ammunition was so overstated that it
was misleading under the rules of evidence”).

15 Only about five percent of all
criminal convictions result from a trial. 
See Bureau of Justice
Statistics, U.S. Dep’t of Justice, Compendium of Federal Justice Statistics 59
(2004), available at http://www.ojp.usdoj.gov/bjs/pub/pdf/cfjs0404.pdf (on file
with the Columbia Law Review) (stating
that in 2004 ninety-six percent of federal felony convictions arose from pleas);
Bureau of Justice Statistics, U.S. Dep’t of Justice, Sourcebook of Criminal
Justice Statistics Online, tbl. 5.46.2004 (2004), available at
http://www.albany.edu/sourcebook/pdf/t5462004.pdf (on file with the Columbia Law Review) (reporting that in
2004 ninety-five percent of state felony convictions were obtained through
pleas).

16 See Innocence Network Amicus, supra note 13, at 31
(“[T]he accused will often choose to forego confrontation entirely, rather than
drive home in front of the fact-finder the accuracy and reliability of the
scientific evidence against him”); see also Melendez-Diaz,
129 S. Ct. at 2540 & n.10 (noting assumption that “every defendant
will . . . demand the appearance of the analyst” is “wildly
unrealistic”).

17 See Brandon L. Garrett & Peter
J. Neufeld, Invalid Forensic Science Testimony and Wrongful Convictions, 95 Va.
L. Rev. 1, 10–11, 89–90 (2009) (noting lack of cross-examination by defense
attorneys and lack of funding for defense experts as factors in forensic
science fraud in criminal prosecutions).

18 Professional organizations such as
the National Association for Criminal Defense Lawyers, and well-respected
public defender organizations such as D.C.’s Public Defender Service offer such
training programs. 

19 Melendez-Diaz, 129 S. Ct. at 2560 app. B (Kennedy, J., dissenting)
(listing Idaho, Mississippi, South Carolina, Texas, Utah, and Wisconsin as
states taking “minority view” that “state hearsay
rules . . . require confrontation of the results of routine
scientific tests or observations of medical personnel”).

20 Id.

21 See Cole v. State, 839 S.W.2d 798,
806 (Tex. Crim. App. 1990) (holding report of crime lab analyst was
inadmissible hearsay in criminal trial); Roma Khanna & Steve McVicker,
Doubt Cast on Hundreds More Cases From HPD Lab, Houston Chron., June 14, 2007,
at 1.

22 See Spears v. State, 241 So. 2d 148, 149 (Miss. 1970)
(holding doctors’ hearsay testimony concerning lab analysts’ results in
criminal trial violated right of confrontation); Radley Balko, Indeed, and
Without a Doubt, Reason Online, Aug. 2, 2007, at
http://www.reason.com/news/show/121671.html (on file with the Columbia Law Review) (discussing
unreliability of forensic odontology generally and allegations concerning
fraudulent testimony of West, who claims to link bite marks to teeth of
criminal defendants).

23 See supra note 1 and accompanying
text.

24 See Karlan, supra note 5, at 1918.

25 Indeed, principles of collateral
estoppel might permit the Rodriguez judgment,
or at least portions of the jury’s findings, to be used offensively by other
individuals who suffered constitutional violations as a result of crime
laboratory findings or testimony. 

26 Laurie Johnson, Houston Longs for
Regional Crime Lab, KUHF-Houston Public Radio News, June 17, 2009, at
http://app1.kuhf.org/houston_public_radio-news-display.php?articles_id=1245274030 (on file with the Columbia Law Review);
Khanna & McVicker, supra note 21.

27 See Harlow v. Fitzgerald, 457 U.S.
800, 818 (1982) (holding government officials performing discretionary
functions are not liable for civil damages if their conduct does not violate
clearly established rights of which reasonable
person would have known); Imbler v. Pachtman, 424 U.S. 409, 418–19 &
n.12, 431 (1976) (noting common law absolute immunity of judges was preserved
under § 1983 and holding prosecutors enjoy absolute immunity under § 1983 in
initiating and presenting state’s case).

28 See,
e.g., City of Canton v. Harris, 489 U.S. 378, 388–91 (1989) (announcing
“deliberate indifference” standard for municipal liability); Daniels v.
Williams, 474 U.S. 327, 335–36 (1986) (holding due process clause does not
afford remedy for negligent conduct of city official).

29 See Karlan, supra note 5, at 1920
(describing burden of proving municipal liability).

30 See Heck v. Humphrey, 512 U.S.
477, 486–87 (1994) (holding § 1983 plaintiff challenging conviction or
imprisonment must show conviction or sentence was reversed, expunged, declared
invalid, or called into question by issuance of writ of habeas corpus).

31 City of Riverside v. Rivera, 477 U.S. 561, 575 (1986) (“Congress expressly recognized that a plaintiff who obtains
relief in a civil rights lawsuit ‘does so not for himself alone but also as a private attorney general, vindicating a policy that Congress considered of the
highest importance.’”
(quoting H.R. Rep. No. 94-1558, at 2 (1976)) (additional internal quotation
marks and citation omitted)).

32 See, e.g., Daryl J. Levinson,
Making Government Pay:  Markets,
Politics, and the Allocation of Constitutional Costs, 67 U. Chi. L. Rev. 345,
347 (2000) (arguing “government actors respond to political, not market,
incentives”).

33 See, e.g., id. at 361–62 (arguing
accuracy of private firm logic of cost-benefit analysis as applied to
municipalities is speculative at best). 

34 See supra note 4 and accompanying text.

35 See, e.g., Larez v. City of Los
Angeles, 946 F.2d 630, 645 (9th Cir. 1990) (holding policymaker’s inadequate
response to constitutional violation may provide basis for finding of
deliberate indifference and collecting other cases so holding).

36 See supra note 8 and accompanying
text.

Preferred
Citation:  Jennifer E. Laurin, Melendez-Diaz
v. Massachusetts, Rodriguez v. City
of Houston, and Remedial Rationing,
109 Colum. L. Rev. Sidebar 82
(2009), http://www.columbialawreview.org/Sidebar/volume/109/82_Laurin.pdf.

While on the academic
job market, I presented Federalization
Snowballs to several stellar law faculties.1 
My argument, in short, was that: 
(1) federal healthcare spending allows the states to externalize onto
the federal government about forty percent of the utilization costs associated with
their medical malpractice policies (such as the cost of defensive medicine);
(2) such an externality systematically distorts a rational state’s incentive to reform medical malpractice; and (3)
federalization of medical malpractice is necessary to correct the
distortion.  In other words, I argued
that federalization of healthcare spending through Medicare, Medicaid, and
similar programs has snowballed into a need for federalization of medical
malpractice.  Federalization snowballs.

As I presented this argument
to faculties around the country, two questions commonly arose that I hadn’t intended to—and hadn’t
in fact—explicitly addressed in the Essay.  (Having been warned against “theoretical drift,”
I limited myself to one application of my theoretical idea, applying the
snowball concept only to my primary area of expertise:  healthcare law.)  The two questions were:  Given the ubiquity of federal spending, aren’t federalization snowballs much more common than the Essay
suggests?  And given the ubiquity of snowballs that must result from the
ubiquity of spending, isn’t the Essay’s theoretical idea much bigger and
therefore either much more important or much more implausible than the Essay
suggests?  (The implied conclusion of “much more important”
or “much more implausible” varied by questioner; some were highly skeptical, others
much more generous.)

This companion piece
addresses those two questions, further delineating the general theoretical idea
of the federalization snowball.  The
first part clarifies the scope of the snowball, demonstrating that the idea
is indeed bigger than medical malpractice but is not (yet) as big as the
federal budget.  The second part clarifies the legal underpinnings of the snowball, discussing its ties to a
constitutional debate that dates back to the framing era; the snowball idea provides
an important theoretical clarification for interpreting the Spending Clause.

I.  How Common is the Federalization Snowball?

Given the ubiquity of
federal spending, aren’t snowballs much more common than the
Essay suggests?  Well, yes and no.  There are probably many snowballs out there;
the phenomenon certainly extends beyond the examples provided in the Essay:  medical malpractice, ERISA, and Social
Security.2 
But not every penny of federal spending creates a federalization snowball.

The snowball has two
critical elements.  First, the problem
arises only if federal spending has an unintended effect on a state-based regulatory regime, which usually means that the federal
dollars must distort states’ incentives in a regime other than the
one that the federal government is directly subsidizing.  The point here is that federal subsidization
always distorts states’ incentives, but there are many federal
funding programs that are specifically intended to do just that—to encourage creation of a more robust program than the
states would create in the absence of federal funding.  This kind of distortion does not, in itself,
result in snowballing federalization because it is exactly what the federal government
hoped and expected would happen.3 
Medicaid itself is such a program. 
The federal government subsidizes the cost of providing health insurance
to the poor so that the states will provide more health insurance to more poor
people than they would in the absence of federal funding.  This distortion is not a problematic snowball;
it is the purpose of the federal subsidy. 
Medicaid, thus, does not create a snowball for social health insurance;
it does so only for the unintentionally affected medical malpractice regime.

This element of the
federalization snowball is important, but it might not knock out much federal
spending from the list of snowball suspects. 
We could imagine two categories of federal spending that would seem
least likely to impact states’ incentives unintentionally:  those over
which the federal government takes full control (such that the states have no
regulatory power that could be affected), and those over which the federal government
takes very little control or provides very little financial support (such that
the states’ incentives do not feel the
impact).  It could be, though—and this is one of many empirical questions that I do not
attempt to answer—that even those kinds of programs have
unintended effects on some state-based
regulatory regimes.

As an example of the
first category, take military spending. 
At first blush, military regulation looks like the rare case of a fully
federalized regime that cannot impact states’
incentives because the states have no power in foreign diplomacy or military
relations.  But this outcome might be a
completed snowball rather than a non-snowball.

Federal regulation of
militarily relevant criminal activity, for example, might have become
necessary, at least in part, because of the federal government’s responsibility for military funding.  The federal government first took a role in
criminal regulation—a departure from the states’ traditional policing powers—in
the late nineteenth century, justifying the intervention by reference to
interstate crime.  But that power quickly
expanded in World War I, as Congress gave the fledgling Bureau of Investigation
(later to become the FBI) control over crimes of espionage—crimes that, at the time, had drastic military consequences.4 
That precise federal role in controlling militarily relevant crimes has
continued into the present.  Most
recently, with the USA PATRIOT Act, Congress created federal crimes of
harboring terrorists, materially supporting terrorism, engaging in terrorism,
etc.5

Now, this particular
field of federal criminal regulation could be justified by reference to something like foreign affairs preemption:  a concern that embarrassing inconsistencies
or incompatibilities might arise among fifty different states’ interactions with foreign citizens.  But federalization here could also be
justified by reference to a snowball:  a
concern that the states lack sufficient incentive to avoid military costs that
would arise from such crimes since those costs accrue only to the federal
government.  In the end, then, even
something as uniformly federal as military spending might have had
unintentional effects on tangentially related state-based regulations and
might, therefore, have snowballed.

As an example of the
second category—i.e., federal spending that might be too insignificant to
affect states’ incentives—consider federal spending
on higher education.  The federal
government provides the following funding for higher education:  grants to low-income students, subsidized
loans to low-income and middle-class students, tax breaks to those paying
interest on student loans, and work study programs for part-time work while in
school.6 
The federal government’s total spending on higher education in
2008 was about $35.2 billion7—just .01% of total outlays that year.8 
In the meantime, substantive regulation of higher education remains
firmly in the states’ hands. 
So far, the federal government has not displaced any of the states’ regulatory powers over higher education, nor has it taken
financial responsibility (beyond the voucher-like programs described here) for
funding any nonmilitary public universities. 
Public universities are still state universities.

The question then is
whether this relatively tiny federal program—which
does not directly affect or displace any state-based regulation—has any effect on the states’
regulatory incentives.  An empirical
question, but the theoretical answer is certainly yes.  Consider tuition-setting at state
universities.  If a state raises the cost
of attending its public universities in the absence of federal funding, fewer
state residents will be able to attend; fewer residents will receive degrees of
higher education; and the state’s workforce and productivity will
weaken.  But because the federal
government diminishes some of those consequences by decreasing the effective
cost of education for the state’s residents, the state can be freer in
raising tuition costs.  It might also, as
a result, allow its universities to be less efficient in managing their
budgets.  Similarly, consider a state’s incentive to encourage parents to save money for their
children’s college.  That incentive, too, must be diminished by
the federal government’s willingness to bail out students
without savings.

The bottom line here
is that all federal spending will affect a rational state’s regulatory incentives, and most or all will probably
affect regulatory incentives other than those that the federal government
intended to affect.  In a rational state’s cost-benefit calculation—in
that state’s decision of whether and how to
regulate—federal subsidization will always have
an impact; it will always decrease the cost of some decision that we might not
want the state to make.9

So why is the
snowball theory not coextensive with
the federal budget?  The answer to that
question rests in the second critical element of the snowball:  the threshold requirement.  A distortion of states’ incentives justifies greater federalization—and thereby snowballs—only
if the distortion is so harmful that the cost of living with it exceeds the
cost of fixing it.  That is, the cost of
the current distortion must exceed the cost of further federalization.

Of course, it is
probably impossible to calculate the costs on either side with any
precision.  In the medical malpractice example
(and this is not for lack of trying), we don’t
know the exact cost of harms that arise from the distortion of states’ incentives (defensive medicine, iatrogenic injury,
frivolous litigation); we don’t know the exact cost of any future harms that might accompany
federal intervention; and, most importantly, we don’t know whether federalization would diminish existing harms
sufficiently to compensate for future harms.10 
My call for federalization, then, rests not on empirical certainty but
rather on a combination of theoretical certainty and circumstantial
evidence:  We know that the states
externalize about 40% of monetary utilization costs associated with medical
malpractice; we know that “40% of monetary utilization costs” in this industry amounts to a lot of dollars; we know—to a theoretical certainty—that
such an externalization would weaken a rational actor’s incentive to engage in cost-saving reforms; we know that
the states have done an appallingly terrible job of reforming medical malpractice;
and we know that the federal government is interested in intervening but has
not yet done so.  Combined, these factors
indicate that the externality is causing real harm and that federalization
might help—or at least is worth trying.

What, then, is an
example of federal spending that does not cross the threshold?  There are probably
many (including higher education), but I’ll
focus here on the one that came up most frequently in job-talk questions:  highway funding.

The federal
government pays for 90% of the interstate highway system’s costs, while the states contribute the remaining 10%.11 
Meanwhile, the states retain control over the vast majority of relevant
regulations:  speed limit laws,
wintertime maintenance (sanding, salting, plowing), auto insurance regulation,
weight restrictions, etc.  All of those
regulatory regimes affect wear and tear of the interstate highways, and 90% of
the cost of fixing that wear and tear—90%
of the monetary cost of under-regulation—accrues
to the federal government.  Surely, then,
highway funding gives rise to a snowball; the states externalize 90% of
monetary cost and must, therefore, underinvest in such regulation.

They probably do—another empirical question beyond my current scope.  But the circumstantial evidence indicates
that further federalization would be more costly than underregulation.  A historical fact for circumstantial
support:  The federal government was unwilling
or unable to enforce its speed limit law, its only attempt at such
regulation.  That law required states to
enforce a fifty-five mile per hour speed limit on their interstates or to
forego federal highway funding.12 
During the time that the law was in effect, the states and their citizens
largely ignored it; many states simply did not patrol for violators, while
others instituted significantly reduced fines for speeds between the federal
limit and the prior state limit.13 
Despite widespread noncompliance, the federal government never enforced
the law itself (at least not with something like FBI highway patrol, though it
did withdraw funding from a handful of states that got too cheeky).  Ultimately, Congress repealed the law.14

Based on this
circumstantial evidence (again, without any empirical certainty as to the
relevant costs or as to the causes of federal failure), it seems that
federalization of highway regulations would not have created enough savings to
be worth its cost.  At least, Congress
decided to repeal the speed limit rather than spend money on enforcing it,
implying that it deemed enforcement unjustifiably costly.  So even though we can be theoretically
certain that federal highway funding systematically distorts states’ incentives to engage in cost-saving regulations, the costs
associated with that distortion do not seem to be high enough to justify
further federalization.  Highway funding,
thus, does not seem to meet the threshold requirement of the federalization
snowball.

In
the end, then, the following hypothesis seems most plausible:  All federal spending gives rise to a potential snowball because all federal
funding unintentionally distorts states’ incentives, but at least some and probably
many federal dollars do not cause a costly enough distortion to justify further
federalization.

II.  How Important Is the Federalization Snowball?

If
all federal spending creates a potential snowball, isn’t the theoretical idea much bigger than the Essay
suggests?  I hope so!  At very least, the snowball idea contributes
to an important constitutional debate that is not broached in the Essay.

To start discussion,
allow me to restate history in my terms: 
In their views on the proper interpretation of the Spending Clause,
James Madison was concerned about snowballs, while Alexander Hamilton was enthusiastic about them.  The Spending Clause gives Congress power to
collect taxes in order “to pay the Debts and provide for the
common Defence and general Welfare of the United States.”15 
The Madison-Hamilton interpretive debate was whether that provision
gives Congress the power to spend on anything that promotes the “general Welfare”
(Hamilton’s view) or to spend only on those things that both promote general welfare and fit into other enumerated powers
(Madison’s view).16

Importantly, Hamilton was comfortable
with his interpretation of the Spending Clause because he was comfortable with
giving Congress full substantive power. 
That is, Hamilton
did not believe that a broad spending power would be distinct from a broad
regulatory power; rather, he believed that Congress should have broad spending
power so that it would have broad
regulatory power, limited by political forces rather than constitutional
constraints.17  Equally importantly, Madison’s
entire reason for wanting to limit the breadth of the Spending Clause was his
recognition that spending and regulating could not be disentangled; he wanted
to tether the spending power to the enumerated powers because he worried that
it would otherwise render the enumeration irrelevant.  That is, he recognized that the power to
spend on anything was tantamount to the power to regulate everything, and he
did not want Congress to have the power to regulate everything.18 
Both men, thus, believed that the power to spend necessarily implied the
power to regulate; their disagreement centered only on whether the new Congress
should possess such a regulatory power.

A century after this
framing-era debate (a century during which the debate went unsettled), a
snowball denier came onto the
scene:  Joseph Story.  In his Commentaries
on the Constitution of the United States, Story argued that the Spending
Clause gave Congress an untethered power to spend but that such a power did not imply that Congress could engage in
substantive regulations beyond those enumerated.19 
That is, Story denied that spending would or could snowball into greater
regulatory incursion.

The debate among the
three came to a head in the New Deal Era, when the federal government ramped up
its taxing and spending in an effort to combat the Great Depression.20 
In that critical moment, the Supreme Court adopted the denier’s account, holding that the federal government could spend
on anything that promotes the general welfare. 
In the first relevant case, United
States v. Butler, the Court laid out Hamilton’s, Madison’s, and Story’s interpretations and expressly adopted
Story’s.21 
But in that case the Court also held that appropriations are not constitutional if they are means to
an unconstitutional (i.e. non-enumerated) regulatory end.22 
Under Butler, thus, taxes could be imposed to
promote the general welfare, but appropriations could not be made to regulate
subjects beyond those enumerated.  After
the “switch in time,” however, the Court abandoned the Butler means-ends restriction on the spending power, and, in the next two relevant
cases, it upheld the Social Security tax despite the resulting appropriations’ creation of an unemployment program and a public pensions
program—two regulatory ends that Congress did
not have express power to pursue.23 
Indeed, in the first of those cases, the Court directly held that
Congress had power to address unemployment under the Spending Clause’s general welfare language, thereby fully abandoning any
need to tie an appropriation to an enumerated substantive power.24

In the end, then, the
Supreme Court adopted Story’s myth that unlimited federal spending
can coexist with limited federal regulation. Of course, the Supreme Court has
since curbed the myth somewhat, not by limiting spending but rather by allowing
substantive regulation.  The Court has
since explicitly permitted Congress to attach substantive regulatory conditions
to its grants, thereby acknowledging that appropriations necessitate
regulation.25 
But the legal community still fails to recognize what Hamilton and
Madison both knew:  that federal spending
ultimately is federal regulation.

What does the
snowball theory add to this story? 
Employing simple economic analysis, the snowball theory not only helps
Madison and Hamilton to rebut the myth, showing that the power to spend on X is
the power to regulate X, but goes one step further, showing that the power to
spend on X frequently necessitates the power to regulate Y.  In so doing, the snowball theory also
demonstrates that Madison and Hamilton understood the crux of the debate better
than we do today:  that our preferences
regarding the scope of the Spending Clause ought to align perfectly with our
preferences regarding the scope of federal regulatory power.  Because all spending distorts states’ incentives to regulate, all spending necessitates federal
regulation to some degree.  Of course, Hamilton had a point in
arguing that politics would limit the extent of federal incursion; the
threshold requirement of the snowball is just that—a political limit on further federalization.  But we cannot deny that our broad reading of
the Spending Clause (combined with a generally slack view of federalism limits)
has set us on a trajectory to full federalization.  And if we want to get off that trajectory, we need to do more than limit our reading
of the Commerce Clause; we also need to revert to Madison’s
view of the Spending Clause.

* Associate Professor of Law, Boston University
School of Law.  Many thanks to the law
faculties at Boston University, University of Miami, American University, Seton
Hall University, Brooklyn Law School, UCLA, and University of Virginia for
their terrific questions.  Thanks also to
Mike Henninger for excellent brainstorming sessions.

1 Abigail R. Moncrieff, Federalization
Snowballs:  The Need for National Action
in Medical Malpractice Reform, 109 Colum. L. Rev. 844 (2009).

2 The most significant example provided
in the Essay, of course, is medical malpractice.  The argument there is that federal financing
of healthcare utilization allows the states to externalize 40% of the
utilization costs associated with their medical malpractice policy choices, necessitating
federalization of medical malpractice policy. 
In Part V of the Essay, I offer two additional examples.  First, I point
out that federal funding of disability insurance through Supplemental Security
Income allows the states to externalize costs associated with unemployment
among disabled citizens, such that a snowball problem might justify federalization
of antidiscrimination regulations for the disabled.  And, indeed, such regulations were federalized
with the passage of the Americans with Disabilities Act.  Second, I point out that federal subsidization
of retirement income through Social Security allows the states to externalize a
portion of the costs associated with poor pension management among private
employers, such that a snowball problem might justify federalization of pension
management regulations.  And, indeed,
such regulations were federalized with the passage of the Employee Retirement
Income Security Act of 1974, 29 U.S.C. §§ 1001–1461 (2006).  See Moncrieff, supra note 1, at 890–91.

3 It does, however, probably necessitate
regulatory interventions in addition to financial ones, a point that I’ll develop more in
the second section.

4 See Espionage Act of 1917, Pub. L. No.
65-24, 40 Stat. 217 (repealed 1948); Fed. Bureau of
Investigation, FBI History:  Timeline of
FBI History, at http://www.fbi.gov/libref/historic/history/historicdates.htm (last visited June 30, 2009) (on file with the Columbia Law Review).

5 Uniting and Strengthening America by Providing Appropriate Tools Required
to Intercept and Obstruct Terrorism (USA PATRIOT) Act of 2001, Pub. L.
107-56, §§ 801–807, 115 Stat. 272, 374–86
(codified as amended in scattered sections of the U.S.C.).

6 New America Foundation, Federal
Education Budget Project:  Federal Higher
Education Programs—Overview, at
http://febp.newamerica.net/background-analysis/federal-higher-education-programs-overview (last visited June 30, 2009) (on file with the Columbia Law Review).

7 Id.  The total amount disbursed for these programs
in 2008 was about $95 billion, but $66.8 billion of that total went to
subsidized loans that must be paid back to the federal government.  Once we account for the principal and
interest that the government will recover, the public cost of those loans is
only about $7 billion.  New America
Foundation, Federal Education Budget Project: 
Student Loan Cost Estimates, at
http://febp.newamerica.net/background-analysis/federal-student-loan-cost-estimates (last visited June 30, 2009) (on file with the Columbia
Law Review).

8 The federal government spent nearly $3
trillion in 2008.  Fin. Mgmt. Serv., U.S. Dep’t
of the Treasury, Combined Statement of Receipts, Outlays and Balances 13–14
(2008),  available
at
http://www.fms.treas.gov/annualreport/cs2008/outlay.pdf (on file with the Columbia Law
Review).

9 None of this is to claim that the
snowball effect actually impacts states’ decisionmaking when legislators are
voting on regulatory change.  There are
undoubtedly many considerations that go into states’ lawmaking decisions,
and whether federal subsidization is a factor at all, much less a decisive one,
is a difficult empirical question that I do not try to answer here.

10 See Moncrieff, supra note 1, at 851
(noting “academics and politicians have continued to disagree—both between and
among themselves—on the significance of malpractice-related inefficiencies to
healthcare costs and on the causal relationship between those inefficiencies
and malpractice litigation”).

11 23 U.S.C. § 120 (2006) (“[T]he Federal share
payable on account of any project on the Interstate System . . . shall be 90 percent
of the total cost thereof . . . .”).

12 Emergency Highway Energy Conservation
Act of 1974, Pub. L. No. 93-239, § 2, 87 Stat. 1046, 1046 (repealed
1995).  The speed limit was later adjusted to 65 mph in rural areas.  Surface Transportation and Uniform Relocation
Assistance Act of 1987, Pub. L. No. 100-17, § 174, 101 Stat. 132, 218 (1987).

13 Paul Grimes, Practical Traveler:  The 55-M.P.H. Speed Limit, N.Y. Times, Dec.
26, 1982, at XX3 (reporting that 83% of travelers on New
York’s interstates drove over fifty-five mph).

14 National Highway System Designation Act
of 1995, Pub. L. No. 104-59, § 205, 109 Stat. 568, 577.

15 U.S. Const. art. I, § 8.

16 Jeffrey T. Renz, What Spending
Clause?  (Or the President’s Paramour):  An Examination of the Views of Hamilton,
Madison, and Story on Article I, Section 8, Clause 1 of the United States
Constitution, 33 J. Marshall L. Rev. 81, 87 (1999).

17 The Federalist No. 31, at 165
(Alexander Hamilton) (E.H. Scott ed., 1898) (arguing, in support of broad
spending power, that “[a] government ought to contain in itself every power
requisite to the full accomplishment of the objects committed to its care, and
to the complete execution of trusts for which it is responsible; free from
every other control, but a regard to the public good, and to the sense of the
people”); The Federalist
No. 32 (Alexander Hamilton), supra, at 168 (“I am persuaded that
the sense of the people, the extreme hazard of provoking the resentments of the
State Governments, and a conviction of the utility and necessity of local administrations,
for local purposes, would be a complete barrier against the oppressive use of
[the broad taxing] power.”).

18 The Federalist No. 41 (James Madison),
supra note 17, at 230 (arguing that other enumerated powers in art.
I, § 8 proved that the
Spending Clause did not amount “to an unlimited commission to exercise
every power, which may be alleged to be necessary for the common defence or general
welfare”).

19 2 Joseph Story, Commentaries on the
Constitution of the United
States § 920, at 383–85 (Boston, Hilliard,
Gray, & Co. 1833).

20 For a careful exploration of the three
men’s views as revealed
in their writings, see Renz, supra note 16.

21 297 U.S. 1, 65–66 (1936).

22 Id.
at 68.

23 Steward Machine Co. v. Davis, 301 U.S. 548, 585–90 (1937) (upholding Social Security’s unemployment
program); Helvering v. Davis, 301 U.S. 619, 640–42 (1937) (upholding
Social Security’s pension program, and holding that Congress has
unreviewable discretion in deciding what constitutes “general welfare”).

24 Steward Machine Co., 301 U.S. at 586–87.

25 South Dakota
v. Dole, 483 U.S.
203, 206 (1987) (“Congress has acted indirectly under its spending power to
encourage uniformity in the States’ drinking ages. . . . [W]e find this
legislative effort within constitutional bounds even if Congress may not
regulate drinking ages directly.”).

Preferred
Citation:  Abigail R. Moncrieff, A Closer Look at the Federalization Snowball,
109 Colum. L. Rev. Sidebar 73
(2009), http://www.columbialawreview.org/Sidebar/volume/109/73_Moncrieff.pdf.

David Ball’s article, Heinous,
Atrocious, and Cruel:  Apprendi, Indeterminate Sentencing, and the Meaning
of Punishment,1 merits a place on any top ten
list of must-read pieces concerning the Supreme Court’s modern sentencing jurisprudence.  Ball’s
article is valuable not only for its fresh conceptual and functional
perspectives on this jurisprudence, but also for its exploration of new and
important regions of the sentencing universe. 
In particular, Ball’s take on the Supreme Court’s work in Apprendi
v. New Jersey2 and its progeny is a major
contribution because, as he adventures through what Justice Scalia once called “Apprendi-land,”3 he spotlights what this
jurisprudential terrain could mean for parole decisionmaking, especially in
California.

It is a
pleasure to travel with Ball as he seeks to better understand the topography of
Apprendi-land.  I fear, however,
that Ball’s impressive work places undue
emphasis on a particular vision of juries which, while perhaps conceptually
appealing, is functionally problematic. 
I am also troubled that, like other commentators and even many Justices,
Ball allows an undue affinity for jury trial rights to dominate his view of Apprendi-land.  I believe Ball and others should focus much
greater attention on constitutional concepts other than the jury in their
efforts to articulate and advance sound procedural rules for modern sentencing
decisionmaking.

I.  Ball’s
Conceptual Vision of Juries and its Functional Problems

Ball’s
article explores the ramifications of Apprendi for indeterminate sentencing systems in which parole authorities continue to
play a significant role, and often engage in significant factfinding, in
deciding just how long a defendant will remain in prison.4 
Ball focuses particularly on California’s
sentencing system, in which the parole board can transform parole eligible
offenses into parole ineligible offenses based on their own findings of fact
about the defendant’s original crime.5 
Rejecting a purely mechanical reading of Apprendi that might require jury findings of fact for all
components of the parole release decision, Ball instead develops a
sophisticated argument that the combination of retributive and rehabilitative elements
in California’s indeterminate sentencing system call for the Apprendi jury right to apply to some,
but not all, factfinding by parole authorities.6

At the
heart of Ball’s analysis and assessment of Apprendi is his attraction to a
particular conceptual vision of juries. 
Specifically, Ball views jurors as community representatives who are
well positioned to make moral, retributivist judgments about criminal
wrongdoing.  As he explains early in his
article, Ball is seeking to give meaning to the “very
jury power that Apprendi established,” and he does so by “[l]ocating
the Apprendi right in the jury’s retributive role.”7 
Ball describes and praises juries as “the
moral representatives of the community,”
and he asserts that “the jury, [as] the conscience of
the community, is uniquely suited to make moral judgments.”8

Importantly,
Ball’s attraction to jury trial rights
and his conceptual vision of juries seems in accord with the Framers’ constitutional perspective.  As Professor Akhil Reed Amar has effectively
documented, the Framers viewed the jury as a critical democratizing force in
the judicial branch:  “Juries were, in a sense, the people themselves,
tried-and-true embodiments of late eighteenth century republican ideology.”9 
As such, Professor Amar explains, jurors during the Founding Era were
understood to have the “right and power to acquit against
the evidence” and “to consider legal as well as factual issues.”  They could even “refuse
to enforce any law that they deemed unconstitutional.”10 
Judge Jack Weinstein has similarly explained the Framers’ perspective on the jury’s role in this way:

The
authors known to the founders had a high respect for the wide powers of the
jury over law, fact and punishment.  In a
sense, the jury was, and remains, the direct voice of the sovereign, in a
collaborative effort with the judge.  It
expresses the view of a sometimes compassionate free people faced with an
individual miscreant in all of his or her tainted humanity, as opposed to the
abstract cruelties of a more theoretical and doctrinaire distant representative
government.11

In other
words, the Framers’ vision of juries as the people’s representative in the judicial branch with broad
powers seems in perfect harmony with Ball’s
modern vision of juries serving as the conscience of the community making
moral, retributivist judgments.

Unfortunately,
while Ball’s conceptual vision of juries may
have a great historical pedigree, this vision finds no functional expression in
current legal doctrines or in the practical realities of modern criminal
justice systems.  Though the Framers long
ago may have embraced juries as the community’s
representatives making retributivist judgments about criminal wrongdoing, just
as Ball does now, in modern times neither criminal law doctrine nor criminal
justice practices allow juries to function effectively in that role.

To begin,
only a precious few criminal cases in the modern American criminal justice
system ever involve a jury.  More than
nine of every ten federal and state convictions are the result of guilty pleas,
not jury trials.12 
Though a guilty plea may often be a sensible choice for a defendant, any
and every plea prevents a jury from having the power or opportunity to make any
sort of moral, retributivist judgments about the defendant and the charges
brought by the state.  In other words,
our modern criminal justice system’s
heavy reliance on pleas formally and functionally takes juries completely out
of the loop in more than ninety percent of all cases that proceed to
sentencing.  This reality makes it
especially problematic for Ball to seek a better understanding of Apprendi and the meaning of punishment
through the role of juries.

Moreover,
even in those rare criminal cases that go to trial, jurors are only asked and
are only permitted to find facts concerning whether the defendant has committed
certain alleged acts.  Though juries
retain a raw power to nullify through an acquittal in the face of clear factual
guilt, modern doctrines do not permit the litigants or the judge to inform
jurors that they have the authority to acquit against the evidence or to
consider legal and constitutional issues.13 
In fact, current law generally does not even permit the litigants or the
judge to inform jurors about the possible or likely sentencing consequences of
their factfinding.14 
Jurors can hardly serve as community representatives making
retributivist judgments when we treat them like moral mushrooms and keep them
in the dark about the true import and impact of the facts they find.

There is
an important and telling exception to the formal and functional limitations
placed on modern juries:  In the
administration of the death penalty, juries still have a profound and
profoundly important role that effectuates Ball’s
conceptual vision.  Modern death penalty
statutes, which legislators created in response to the Supreme Court’s Eighth Amendment jurisprudence, ensure that juries
will act as the conscience of the community, making moral, retributivist
judgments, in nearly every capital case. 
Because of special indictments and special jury selection procedures, capital
jurors know from the outset of their service that they will be asked to make a
moral judgment as to whether a particular offender deserves to die for his
alleged crimes.15 
Moreover, capital jurors are not merely asked to find whether a capital
defendant is factually guilty, they also decide whether a guilty defendant
should be sentenced to death for his crimes. 
And nearly every capital case involves a jury trial because, even if a
capital defendant admits guilt, he still can (and usually will) ask jurors to
impose a sentence other than death.

The fact
that Ball’s vision of juries finds ready
expression in modern capital punishment is not really surprising; as he
recognizes, his vision draws from Justice Stevens’s
discussion of the jury’s proper role in capital cases.16 
Moreover, long before Apprendi,
the Supreme Court’s capital jurisprudence focused on ensuring that jurors are able to give a “reasoned moral response” to mitigating evidence put forward by a capital
defendant.17 
And, tellingly, even Justices Stephen Breyer and Anthony Kennedy, both
of whom have actively and urgently argued against extending Apprendi in other settings, did not
object to its extension to death penalty proceedings in Ring v. Arizona.18 
Put simply, the concept of giving juries a prominent role and enabling
them to express moral judgment in capital punishment decisionmaking is not
conceptually controversial or functionally problematic.  Conceptually, few resist the idea that death
sentencing involves a moral, retributivist judgment that should reflect
community sentiments as expressed by a jury; functionally, because death is
pursued as a criminal punishment so rarely by the state, it is possible to
ensure that juries are central in the capital decisionmaking process.

But, outside
of the rare and high profile setting of capital cases, there is much conceptual
controversy over whether and how sentencing is to incorporate moral,
retributivist judgments and community sentiments.  As Ball notes, not long ago criminal justice
systems embraced a rehabilitative “therapeutic” model of punishment, in which sentencing judges and
parole boards sought to assess and predict which offenders were more or less
likely to commit future crimes.19 
And, with evidence-based, risk assessment sentencing models attracting
interest as prisons overflow, forward-looking utilitarian approaches to
sentencing may again soon eclipse any backward-looking retributivist approaches
in modern punishment systems.

Moreover,
beyond these conceptual issues, it is functionally unimaginable to have jurors
regularly involved in making moral, retributivist judgments in all or even most
noncapital criminal cases.  Dramatically
limiting the number of criminal cases resolved through pleas would likely be
unwise and would certainly be impracticable. 
Though some academics have recently urged greater jury involvement in
noncapital sentencing,20 nobody has seriously contended
that the vast array of constitutional rules that have come to ensure jury
involvement in capital cases should (or even could) be regularly incorporated
into noncapital criminal cases.  And yet,
if taken to its logical extreme, Ball’s
efforts to “locat[e] the Apprendi right in the jury’s
retributive role”21 suggests that procedurally
elaborate jury-centric proceedings used in capital cases can and should become
a Sixth Amendment mandate for all cases.

II.  Charting a
Different (and Sounder) Adventure in Apprendi-Land

Truth be
told, I am not convinced that Ball and other fans of Apprendi (of which I am one) are really so enamored with
juries.  I doubt Ball and other Apprendi fans would embrace and endorse
reforms designed specifically to ensure greater jury involvement in noncapital
criminal cases by, say, prohibiting all plea deals and requiring juries to make
sentencing decisions for all defendants convicted of the low level crimes, like
drug possession, burglary, and DUI, that
occupy the great bulk of courts’ criminal dockets.  Rather, I suspect what really motivates Ball
and other Apprendi fans—what certainly accounts for my affinity for Apprendi—is
the recognition that (1) sentencing
decisions are often far more consequential than basic guilt determinations, and
(2) it is highly problematic for defendants to enjoy so many procedural rights
and constitutional protections for guilt determinations, but so precious few
procedural rights and constitutional protections for sentencing decisions.  There is, I suspect, a strong inclination to
endorse and embrace broad Sixth Amendment jury trial rights because it is assumed
that when a defendant has a right to a jury he will also necessarily get all
the other procedural rights and constitutional protections that are associated
with a traditional criminal trial.

But, in
my view, it is legally sloppy and conceptually problematic to always rigidly
link and analyze the jury trial right with other procedural rights and
constitutional protections afforded criminal defendants.  Indeed, as I have stressed in other writings,22 the Supreme Court’s modern sentencing jurisprudence, which was
formalized in Apprendi and supersized
in Blakely v. Washington,23 has roots in constitutional
provisions and concerns beyond just the Sixth Amendment’s jury trial right. 
Specifically, the Due Process Clause and the notice provisions of the
Sixth Amendment initially played an important and foundational role in the
Supreme Court’s efforts in this line of
cases.  Unfortunately, the Supreme Court’s post-Apprendi rulings—as well as analyses and
criticisms of the Court’s jurisprudence like Ball’s article—have
been almost exclusively concerned with the reach and limits of jury trial
rights.

In the
often overlooked case of Jones v. United States—which first set out the key concepts that Apprendi turned into doctrine and that Blakely greatly expanded—the Supreme Court expressly drew on constitutional
provisions and principles beyond the Sixth Amendment’s jury trial right.24 
Decided in 1999, Jones was the first case in which five Justices
indicated that facts establishing higher penalties must be treated procedurally
as offense elements, and the Jones Court asserted the basis for its ruling in
expansive terms.  In a critical footnote,
the Court explained that “a
set of
constitutional concerns that have emerged through a series of our decisions
over the past quarter century”25 suggested the principle that, “under the Due Process Clause of the Fifth Amendment
and the notice and jury trial guarantees of the Sixth Amendment, any fact
(other than prior conviction) that increases the maximum penalty for a crime
must be charged in an indictment, submitted to a jury, and proven beyond a
reasonable doubt.”26 
Building upon Jones to establish a definitive constitutional
rule, the Supreme Court in Apprendi stressed the due process concepts set forth in In re Winship to formalize “beyond a reasonable doubt” as the standard of proof in criminal prosecutions.27 
The Apprendi Court explained that “we
have made clear beyond peradventure that Winship’s due process and associated jury protections
extend, to some degree, ‘to determinations that [go] not
to a defendant’s guilt or innocence, but simply
to the length of his sentence.’”28 
In short, when Apprendi-land
was first created, its terrain included not only a Sixth Amendment jury right
region, but also Fifth Amendment due process and Sixth Amendment notice
regions.

In nearly
every major post-Apprendi ruling,
however, the Supreme Court formally or implicitly restricted the importance and
impact of the nonjury aspects of Apprendi in order to preserve pre-Apprendi precedents that rejected defendants’
arguments for more procedural rights at sentencing.  In particular, in Harris v. United States, the Court reaffirmed its 1986 ruling in McMillan v. Pennsylvania that facts
triggering mandatory minimum sentences could be found by a judge based on a
preponderance standard of proof.29 
Similarly, in Blakely30 and United States v. Booker,31 the Court reaffirmed its 1949
ruling in Williams v. New York that judges could still find
facts through lax sentencing procedures in a discretionary or advisory sentencing
scheme.32 
In all of these critical post-Apprendi rulings, the Supreme Court discussed at great length the importance and reach
of the Sixth Amendment jury trial right, but failed to engage seriously with
the other procedural rights and principles discussed in Jones and Apprendi.  In short, as the Justices took later
adventures through Apprendi-land,
considerable time was spent exploring the Sixth Amendment jury right region,
while the Fifth Amendment due process and Sixth Amendment notice regions were largely
ignored and have now been all but forgotten.

To his
credit, Ball recognizes and discusses the fact that Apprendi concerns more than just the scope and application of the
Sixth Amendment jury right. 
Unfortunately, his analysis of other parts of Apprendi-land gets short shrift; it appears in Part IV at the very
end of his article and only after he has devoted extensive energy and many
words to “locating the Apprendi right
in the jury’s retributive role” and probing the “very jury power that Apprendi established.”33 
I believe the important goal of developing conceptually and functionally
sound procedural rules for modern sentencing decisionmaking would be much better
served if Ball and other commentators—as well as the
Justices themselves—would now give jury trial rights short
shrift and instead devote extensive energies toward adventuring into other now
forgotten regions of Apprendi-land.

Conclusion

For both
sentencing theorists and practitioners, Apprendi-land
remains an important part of the jurisprudential universe, and Ball’s article provides a helpful map and compass for
those still traveling into this mysterious domain.  Though Ball does extraordinary work
adventuring through Apprendi-land
guided by his lovely (and historically resonant) conceptual commitment to
juries, I think his article could ultimately be most beneficial if readers come
away with the realization that perhaps it is not conceptually useful or
constitutionally wise to have juries serve as the chief tour guides through Apprendi-land.

* William
B. Saxbe Designated Professor of Law, Moritz
College of Law at The Ohio State
University.

1 W. David Ball, Heinous,
Atrocious, and Cruel:  Apprendi, Indeterminate Sentencing, and
the Meaning of Punishment, 109 Colum. L. Rev. 893 (2009).

2 530 U.S. 466 (2000).

3 Ring v. Arizona,
536 U.S.
584, 613 (2001) (Scalia, J., concurring).

4 Ball, supra note 1, at 898–99.

5 Id. at 902–03.

6 Id. at 905–06.

7 Id. at 899.

8 Id. at 923–24.

9 Akhil Reed Amar, America’s Constitution: 
A Biography 234 (2005).

10 Id. at 238–39.

11 United States v. Khan, 325 F. Supp.
2d 218, 229–30 (E.D.N.Y. 2004) (internal
citations omitted).

12  See
Tracey Kyckelhahn & Thomas H.
Cohen, U.S. Dep’t of Justice, Bureau of Justice Statistics Bulletin:  Felony Defendants in Large Urban Counties, 2004, at 1 (2008), available at http://www.ojp.usdoj.gov/bjs/pub/pdf/fdluc04.pdf (on file with
the Columbia Law Review) (finding
that, in nation’s seventy-five most
populous counties, more than ninety-five percent of convictions occurred
through guilty plea); U.S. Sentencing Comm’n, 2008 Sourcebook of Federal Sentencing
Statistics fig. C (2009), available at http://www.ussc.gov/ANNRPT/2008/SBTOC08.htm (on file with the Columbia Law Review)
(reporting that 96.3% of all federal convictions in fiscal year 2008 were
result of guilty pleas).

13 See generally Joshua Dressler,
Understanding Criminal Law 5–8 (4th ed. 2006).

14 See Shannon v. United States, 512
U.S. 573, 579 (1994) (“The principle that juries are not
to consider the consequences of their verdicts is a reflection of the basic
division of labor in our legal system between judge and jury.”).

15 See generally Jeffrey Abramson,
Death-is-Different Jurisprudence and the Role of the Capital Jury, 2 Ohio St.
J. Crim. L. 117 (2004) (exploring
the role of the jury in making moral judgments about whether a particular
defendant deserves to die).

16 Ball, supra note 1, at 924–25 (describing Justice Stevens’s view of juries as developed in several death
penalty cases).

17 Perry v. Linaugh, 492 U.S. 302, 319 (1989) (quoting California
v. Brown, 479 U.S.
538, 545 (1987) (O’Connor, J., concurring)).

18 536 U.S. 584, 613 (2002) (Kennedy, J.,
concurring); id. at 613–14 (Breyer, J., concurring in
the judgment).

19 Ball, supra note 1, at 926–29.

20 E.g., Morris B. Hoffman, The Case
for Jury Sentencing, 52 Duke L.J. 951 (2003); Jenia Iontcheva, Jury Sentencing
as Democratic Practice, 89 Va.
L. Rev. 311 (2003); Adriaan Lanni, Note, Jury Sentencing in Noncapital Cases:  An Idea Whose Time Has Come (Again)?, 108
Yale L.J. 1775 (1999).

21 Ball, supra note 1, at 899.

22 See Douglas A. Berman, Beyond Blakely and Booker:  Pondering Modern
Sentencing Process, 95 J. Crim. L. & Criminology 653, 654–55 (2005).

23 542 U.S. 296, 303 (2004) (holding
that “[t]he ‘statutory maximum’
for Apprendi purposes is the maximum
sentence a judge may impose solely on the basis of the facts reflected in the
jury verdict or admitted by the defendant, not the maximum sentence a judge may
impose after finding additional facts”).

24 526 U.S. 227 (1999).

25 Id. at 251 n.11 (emphasis added).

26 Id. at 243 n.6.

27 Apprendi v. New
Jersey, 530 U.S.
466, 484 (2000).

28 Id.
(quoting Almendarez-Torres v. United States, 523 U.S. 224, 251 (1998) (Scalia, J.,
dissenting)).

29 536 U.S. 545, 568 (2002) (“Reaffirming McMillan and employing the approach outlined in that case, we conclude that . . . [b]asing a 2-year increase in
the defendant’s minimum sentence on a judicial
finding of brandishing does not evade the requirements of the Fifth and Sixth
Amendments.”).

30 Blakely v. Washington, 542 U.S. 296, 304–05 (2004).

31 543 U.S. 220, 233 (2005) (“We have never doubted the authority of a judge to
exercise broad discretion in imposing a sentence within a statutory range.”).

32 337 U.S. 241, 251–52 (1949).

33 Ball, supra note 1, at 899.

Preferred Citation:  Douglas A. Berman, Should Juries Be the Guide for Adventures Through Apprendi-Land?, 109 Colum. L. Rev. Sidebar 65 (2009), http://www.columbialawreview.org/Sidebar/volume/109/
65_Berman.pdf.