[EL] “Flaws in the Efficiency Gap”

Wed Oct 4 08:39:48 PDT 2017

Hi Eric,

Thanks for taking the time to provide us with your opinion. I want to
respond to your comments and clarify our argument to the extent that it may
be opaque.

To begin, I think that it is important to emphasize that we are economists
and are writing from the perspective of our discipline. We test measures
using thought experiments to verify that they are theoretically sound. A
failure of the measure in a setting that seems extreme but where the
“right” answer is clear provides us with a sufficient reason to disqualify
that measure from practical use. This methodology enables us to distinguish
defensible measures from ad hoc measures whose justification is essentially
limited to the claims of experts that they work well in a narrow set of
examples. I do not think our claims can be refuted without either (a)
arguing from within this methodology or (b) rejecting the applicability of
economic theory to the study of measurement. Obviously, the philosophy of
science is beyond the scope of this list, and it is not my goal at this
point to convince you of the correctness of the economic methodology. I'll
be happy to discuss it off-list, though, if you or others would like to
learn more.

With respect to your replies to our specific critiques, I will reply in
order.

1. Our claim that the efficiency gap contains an implicit form of
cost-benefit analysis is not based on the scenario where one party wins all
votes in the state. We do use that example to justify our claim that a
different measure of benefit would be more appropriate, but it is by no
means necessary to make this point. I have not seen the idea of separating
the efficiency gap into cost and benefit functions elsewhere in the
literature.

We do agree (as we wrote in the paper) that a proportionality standard
would not be desirable. However, it does not follow that all
non-proportional measures are necessarily better. Our argument was that the
logic of the cost-benefit analysis implicit in the efficiency gap leads to
proportionality. Hence, if we would not want a proportionality standard, we
should not use the efficiency gap.

One critique of proportionality is that Congress was given the power to
decide whether to use a proportional system, and therefore this decision
should not be made by the courts. This same critique can be easily raised
against any voting system in which the number of seats is purely a function
of the aggregate vote share in the election. My intuition tells me that the
form of “proportionality with a winners’ bonus” embodied by the efficiency
gap would be less popular, politically, then would be a purely proportional
system. But of course, others may not share my intuition.

I do not fully agree that the problem highlighted by the example is
entirely unrealistic. Regardless of whether any states currently fit this
profile, I can imagine the possibility of a 75% Republican state where the
25% minority of Democrats is geographically separated into an urban area.
(For example, consider a slightly more extreme version of Utah.) An
implication of the efficiency gap in this context is that the Republicans
would be justified in depriving Democrats of all seats in the legislature,
regardless of the population distribution. This seems problematic to me.

2. Why should the Equal Protection Clause not apply to partisans in small
states? Democrats in Utah, for example, feel as if Republicans have tried
to reduce their political influence through the gerrymandering of
Congressional district plans. I do agree, though, that the efficiency gap
can be separated from the guidance about how to apply it, and I'm glad to
see that you are flexible in this regard.

3. We provided a theory as to why political preferences (beyond strict
partisan affiliation) might matter, and gave an example as to how the
efficiency gap can make things worse off. The evidence that you cite does
not contradict this theory. And there is plenty of evidence that does not
fit the simple model where voters care only about party.

4. There are at least two concerns regarding randomness. The first is that
the efficiency gap is itself a random measure of redistricting. The second
is that elections are random events.

Regarding the first, I do not understand what you meant when you write that
the “volatility” of the efficiency gap is a desirable characteristic of a
measure because it “ensures that all the uncertainty is on the table before
a court intervenes.“ Why would we want the measure of gerrymandering to be
depend so heavily on chance?

My reading of your paper was that sensitivity testing is part of your
proposed test, and not part of the measure itself. I did not see a
description of sensitivity testing, how it works, and what assumptions are
made in the process. I hope, though, that while we may disagree with
respect to the soundness of the measure, we would agree on the more general
principle that a “good” test would not cure a defective measure.

Regarding the second, the efficiency gap, in its formulation and design,
seems to assume a deterministic model of voter behavior. It is presented in
your joint paper as being motivated by a model of packing and cracking in
which voter preferences (and turnout decisions) are fixed. As shown in the
literature we cited, it is not clear that politicians would pack and crack
in a setting of uncertainty.

I hope this clarifies our view. I'm happy to answer questions off of the
list as well.

Alan Miller

--

Alan D. Miller

Faculty of Law and Department of Economics
University of Haifa, Mount Carmel, Haifa, 31905, Israel
admiller at econ.haifa.ac.il; http://econ.haifa.ac.il/~admiller

On Tue, Oct 3, 2017 at 3:14 AM, Eric McGhee <mcghee at ppic.org> wrote:

> Engagement on the EG is always welcome, but these arguments either rehash
> points we’ve seen before or have little merit.  Taking them briefly in turn:
>
>
>
> 1.     The efficiency gap contains an implicit form of cost-benefit
> analysis; when made explicit, this form appears very peculiar and is hard
> to justify.
>
>
>
> This claim is based on a scenario where one party wins all the votes in a
> state.  Beyond the implausibility of the scenario, it touches on issues
> that were first raised in my original peer-reviewed political science
> article three years ago. The practical effects are nil:  once a party has
> all the seats, it cannot draw a plan that hands itself more districts so
> there is no gerrymandering question to discuss. The alternative the authors
> propose (proportionality) would be a non-starter for the courts.  Moreover,
> it has conceptual problems of its own, some of which are identified in my
> forthcoming Election Law Journal article: http://online.liebertpub.com/
> doi/abs/10.1089/elj.2017.0453.
>
>
>
> 2.     The suggested threshold for congressional districts is
> scale-dependent, making it much easier to find gerrymandering in large
> states than in small states.
>
>
>
> This is true, but has a very good justification: a seat is a seat from the
> perspective of control of the House of Representatives, so both the
> potential for and danger from gerrymandering are much greater in a state
> like California than in a state like Arkansas. At any rate, the threshold
> we propose is not central to the efficiency gap. If someone wanted to use a
> different threshold, or use percentages (or some combination), they would
> be free to do so. (And indeed Simon Jackman has done so in the ongoing
> North Carolina litigation.) The issue of percentages versus raw numbers is
> one that social scientists manage to work with all the time.
>
>
>
> 3.     The efficiency gap ignores whether voters are extreme or moderate,
> and instead identifies them only by their partisan labels. This can lead to
> polarization and can harm the supporters of the weaker party.
>
>
>
> This point is purely speculative and at odds with the realities of modern
> American politics.  There is a mountain of evidence from political science
> (some covered in this paper http://online.liebertpub.com/
> doi/abs/10.1089/elj.2017.0452?src=recsys) that polarization is largely
> independent of district composition. This is because Democratic
> (Republican) legislators tend to be liberal (conservative) whether they are
> elected from safe or swing districts. Similarly, voters today rarely split
> their tickets and exhibit highly consistent behavior from election to
> election.
>
>
>
> 4.     The efficiency gap ignores uncertainty. (This has been noticed, to
> some extent, in the literature.)
>
>
>
> This is simply false, as the authors themselves acknowledge.
> Stephanopoulos and I proposed sensitivity testing to deal with
> uncertainty.  If Chambers, et al. find that unsatisfying, Simon Jackman has
> spent many pages of trial court reports empirically addressing the
> uncertainty and volatility of the EG, both ex ante and ex post. Moreover,
> the “volatility” of the EG is a feature, not a bug, because it ensures that
> all the uncertainty is on the table before a court intervenes.
>
>
>
> In short, in my view this is largely old wine in new bottles.  To the
> extent that it is not, it is highly theoretical with only a limited
> understanding of the practicalities or relevant empirical literature.
>
>
> Eric
>
>
>
> Eric McGhee
> Research Fellow
>
> PUBLIC POLICY
> INSTITUTE OF CALIFORNIA
> 500 Washington Street, Suite 600
> San Francisco, CA 94111
> tel  415 291 4439
> fax  415 291 4401
> web  www.ppic.org<../../owa/redir.aspx?C=77b183a3bc734de5b9da280534391e
> 2a&URL=http%3a%2f%2fwww.ppic.org%2f>
>
> Any opinions expressed in this message are those of the author alone and
> do not necessarily reflect any position of the Public Policy Institute of
> California.
> ________________________________
> From: Law-election <law-election-bounces at department-lists.uci.edu> on
> behalf of Alan Miller <admiller at econ.haifa.ac.il>
> Sent: Monday, October 2, 2017 2:41 PM
> To: law-election at department-lists.uci.edu
> Subject: [EL] “Flaws in the Efficiency Gap”
>
> Hi everyone.
>
> I just want to bring to your attention a new paper that I've co-authored
> with Chris Chambers and Joel Sobel. The paper examines the efficiency gap
> and identifies several aspects that we consider to be significant flaws.
> For the most part, our criticisms do not appear elsewhere in the literature
> that I've seen.
>
> The paper is forthcoming in the Journal of Law and Politics, and the final
> version is available on SSRN: https://papers.ssrn.com/sol3/
> papers.cfm?abstract_id=3046797<https://papers.ssrn.com/sol3/
> papers.cfm?abstract_id=3046797>.
>
> To summarize our article in a few lines, we show that:
>
> 1. The efficiency gap contains an implicit form of cost-benefit analysis;
> when made explicit, this form appears very peculiar and is hard to justify.
>
> 2. The suggested threshold for congressional districts is scale-dependent,
> making it much easier to find gerrymandering in large states than in small
> states.
>
> 3. The efficiency gap ignores whether voters are extreme or moderate, and
> instead identifies them only by their partisan labels. This can lead to
> polarization and can harm the supporters of the weaker party.
>
> 4. The efficiency gap ignores uncertainty. (This has been noticed, to some
> extent, in the literature.)
>
> I hope that those of you following this topic find the paper interesting,
> and am happy to both hear feedback and answer questions.
>
> Best wishes,
>
> Alan D. Miller
>
> --
>
> Faculty of Law and Department of Economics
> University of Haifa, Mount Carmel, Haifa, 31905, Israel
> admiller at econ.haifa.ac.il<mailto:admiller at econ.haifa.ac.il>;
> http://econ.haifa.ac.il/~admiller<http://econ.haifa.ac.il/~admiller>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://webshare.law.ucla.edu/Listservs/law-election/attachments/20171004/fe436342/attachment.html>