Monkey Cage post

For those of you coming from the Monkey Cage: welcome!

This is a blog on my research and other topics of interest. I’m in the middle of a series on incorrect survey weighting, which is part of a larger series on reproduction in social science. I’m a proponent of research transparency, such as preregistration of experimental studies to reduce researcher degrees of freedom, third-party data collection to reduce fraud, and public online archiving of data and code to increase the likelihood that error is discovered.

My main research areas right now are race, law, and their intersection. I plan to blog on those and other topics: I am expecting to post on list experiments, abortion attitudes, the file drawer problem, Supreme Court nominations, and curiosities in the archives at the Time-Sharing Experiments for the Social Sciences. I hope that you find something of interest.

UPDATE (May 21, 2014)

Links to the Monkey Cage post have been made at SCOTUSBlog, Jonathan Bernstein, and the American Constitution Society.

UPDATE (May 21, 2014)

Jonathan Bernstein commented on my Monkey Cage guest post, expressing skepticism about a real distinction between delayed and hastened retirements. The first part of my response was as follows:

Hi Jonathan,

Let me expand on the distinction between delayed and hastened retirements.

Imagine that Clarence Thomas reveals that he wants to retire this summer, but conservatives pressure him to delay his retirement until a Republican is elected president. Compare that to liberals pressuring Ruth Bader Ginsburg to retire before the 2016 election.

Note the distinctions: liberals are trying to change Ginsburg’s mind about *whether* to retire, and conservatives are trying to change Thomas’s mind about *when* to retire; moreover, conservatives are asking Thomas to sacrifice *extra* *personal* time that he would have had in retirement, and liberals are asking Ginsburg to sacrifice *all* the rest of her years as *one of the most powerful persons in the United States.*

Orin Kerr of the Volokh Conspiracy also commented on the post, at the Monkey Cage itself, asking why a model is necessary when the sample of justices is small enough to ask justices or use past interviews. My response:

Hi Orin,

Artemus Ward has a valuable book, Deciding to Leave, that offers more richness than statistical models offer for investigating the often idiosyncratic reasons for Supreme Court retirements. But for addressing whether justices retire strategically and, if so, when and under what conditions — or for making quantitative predictions about whether a particular justice might retire at a given time — there is complementary value in a statistical model.

1. For one thing, there is sometimes reason to be skeptical of the reasons that political actors provide for their behavior: there is a line of research suggesting that personal policy preferences inform Supreme Court justice voting on cases, though many justices might not admit this in direct questioning. Regarding retirements, many justices have been forthcoming about their strategic retirement planning, but some justices have downplayed or denied strategic planning: for example, Ward described press skepticism of Potter Stewart’s assertion that he did not strategically delay retirement while Jimmy Carter was president (p. 194).

Statistical models permit us to test theories based on what Stewart and other justices *did* instead of what Stewart and other justices *said*, similar to the way that prosecutors might develop a theory of the crime based on forensic evidence instead of suspect statements.

2. But even if the justices were always honest and public about their reasons for retiring or not retiring, it is still necessary to apply some sort of statistical analysis to address our questions. By my count, from 1962 to 2010, 5 justices retired consistent with a delay strategy and 8 justices retired when the political environment was unfavorable. Observers using simple statistical tools might consider this evidence that justices are more likely to retire unstrategically than to delay retirement, but this overlooks the fact that justices have more opportunities to retire unstrategically than to delay retirement.

For example, assuming that no conservative retires during President Obama’s eight years in office, the five conservative justices as a group will each have had eight years to retire unstrategically, for a total of 40 opportunities; but liberal justices have had fewer opportunities to delay retirement: Breyer, Ginsburg, Souter, and Stevens each had one opportunity to retire consistent with a delay strategy in 2009, and — presuming that justices stay on another year to avoid a double summer vacancy — Breyer, Ginsburg, Sotomayor, and Stevens each had one opportunity to retire consistent with a delay strategy in 2010, for a total of 8 opportunities.

In this particular period, the proper comparison is not 2 delayed retirements to 0 unstrategic retirements, but instead is 2 delayed retirements out of 8 opportunities (25%) to 0 unstrategic retirements out of 40 opportunities (0%).

3. Sotomayor’s addition in the 2010 data highlights another value of statistical models: they permit us to control for other retirement pressures. Statistical models can help account — in a way that qualitative studies or direct questioning cannot — for the fact that the 2010 observation of Sotomayor is not equivalent to the 2010 observation of Ginsburg because these justices have different characteristics on other key variables, such as age. From 1962 to 2010, justices retired 14 percent of the time during delayed retirement opportunities, but retired only 4 percent of the time during unfavorable political environments. But these percentages should not be directly compared because there might be spurious correlations that have inflated or deflated the percentages: for example, perhaps older and infirm justices were more likely to experience a delayed opportunity and *that* is why the delayed percentage is relatively higher than the unstrategic percentage. Statistical models let us adjust summary statistics to address such spurious correlations.

Bill James is said to have said something to the effect that bad statistics are the alternative to good statistics. Relying only on justice statements instead of good statistics can introduce inferential error about justice retirement strategies in the aggregate in several ways: (1) justices might misrepresent their motives for retiring or not retiring; (2) we might not properly account for the fact that justices face more unstrategic opportunities than delayed opportunities or hasten opportunities; and (3) we might not properly account for variables such as age and illness that also influence decisions to retire.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s