An empirical analysis of life tenure: a response to professors Calabresi & Lindgren.

• Author: Stras, David R.
• Position: Response to Steven G. Calabresi and James Lindgren, Harvard Journal of Law & Public Policy. vol. 29, p. 769, 2006

INTRODUCTION I. PINNING DOWN THE EMPIRICAL CLAIM II. LIFE TENURE BY THE NUMBERS A. The Period-Selection Problem B. The Date-of-Observation Problem C. Regression Models for Length of Tenure 1. Calabresi and Lindgren's Reported Cubic Model 2. Calabresi and Lindgren's Unreported S-Curve Model 3. The Null Hypothesis 4. The Date-of-Observation Problem Revisited CONCLUSION INTRODUCTION

Opposition to life tenure has been steadily mounting in the legal academy, and Professors Steven Calabresi and James Lindgren are among those leading the charge. In an article published in the Harvard Journal of Law & Public Policy, (1) Calabresi and Lindgren denounce life tenure as "fundamentally flawed" (2) and "essentially a relic of pre-democratic times." (3) They deserve credit for assembling the most comprehensive critique of life tenure to date, carefully documenting what they see as its major drawbacks, and proposing a constitutional amendment providing for fixed, non-renewable eighteen-year terms for Supreme Court Justices. (4) Their article has attracted well-deserved attention among legal scholars and in the popular press. To its credit, it has found champions of all political persuasions. (5)

Calabresi and Lindgren direct three criticisms at life tenure. First, because it produces infrequent vacancies, life tenure affords the President and Senate too few opportunities to act as a democratic check on the Court through the appointment of new members. (6) Second, again because of infrequent vacancies, life tenure raises the stakes of each appointment to undesirable levels, thereby exacerbating the politicization of the appointment process. (7) Third, life tenure allows Justices to serve well into old age, increasing the risk of "mental decrepitude" on the Court. (8)

Crucial to these policy criticisms is an underlying empirical claim that "the real-world, practical meaning of life tenure" has changed dramatically since 1970. (9) Calabresi and Lindgren point to several trends as evidence of this transformation. The most frequently mentioned is the average length of tenure on the Court: Justices who left office between 1789 and 1970 served an average of 14.9 years on the bench, whereas Justices who left office between 1970 and 2006 averaged 26.1 years--an increase that Calabresi and Lindgren describe as "astonishing." (10) According to the authors, the trend marks a fundamental change in the operation of life tenure.

In the winter of 2005, we published an article taking a contrary view, defending life tenure as an institution worth preserving but proposing a package of retirement incentives--a "golden parachute" for Supreme Court Justices-to encourage mentally infirm Justices to retire in a timely manner. (11) We agreed with Calabresi and Lindgren on a number of issues, echoing their concerns about mental and physical infirmity on the Supreme Courty, (12) and agreeing that statutory efforts to abolish life tenure are unconstitutional. (13)

But we also dedicated less than two pages of our seventy-page article to a rebuttal of one aspect of Calabresi and Lindgren's empirical claim, arguing that the "astonishing" increase in average tenure "depends more on the chosen period lengths than a bona fide trend." (14) Our article included two simple charts demonstrating that, by choosing a shorter period length or using non-overlapping groups of Justices rather than periods of time, the data do not reveal a dramatic recent increase. (15) The revised version of Calabresi and Lindgren's article contains fully eleven pages of new material responding to our criticisms. (16)

This Article refines and elaborates upon our critique of Calabresi and Lindgren's empirical claim, arguing that changes in average tenure are not "astonishing" and "unprecedented." Instead, we defend the conventional view that, despite short-term fluctuations, length of tenure has increased slowly and steadily over the long term, and that we can expect more slow and steady growth in the future.

The Article proceeds in two Parts. We acknowledge here, as we have from the outset, (17) that the "practical meaning" of life tenure has changed over time, and it will continue to change as Americans enjoy longer life expectancies. The question, therefore, is not whether there has been any change, but whether the change has been so profound and so swift to warrant the abandonment of life tenure. In Part I, we pin down the precise empirical claim advanced by Calabresi and Lindgren: that changes in the practical operation of life tenure since 1970 are both (1) "dramatic" and (2) "unprecedented." We also briefly summarize the course of the debate thus far.

In Part II, we analyze and evaluate the data on life tenure. First, we defend our original contention that period selection affects the shape and apparent magnitude of the trend in length of tenure over time. Second, we demonstrate that making a simple change to Calabresi and Lindgren's assumption that each Justice's tenure should be treated as an observation at the date of departure, rather than date of swearing-in, undermines their model by calling into question the robustness of the trend they identify. Third, we critique the cubic regression model that Calabresi and Lindgren describe as an "elegant confirmation" of their theory of a post-1970 explosion in length of tenure. Instead, applying a different nonlinear regression model, alluded to by Calabresi and Lindgren but not reported in their article, better fits the data and supports our position that changes in term length have been gradual, not "astonishing."

1. PINNING DOWN THE EMPIRICAL CLAIM

   Shortly after Professors Calabresi and Lindgren published the first version of their article, we included a short critique of their empirical claim in an article proposing changes to retirement incentives for Supreme Court Justices. (18) Although we explicitly conceded that "average length [of tenure] may in crease over time due to advances in medical care," (19) we argued that Calabresi and Lindgren's "chosen period lengths" had the effect of "exaggerat[ing] the trend." (20) To illustrate the point, we presented two simple charts. The first rendered the data in periods of ten years, rather than thirty years. It showed that average tenure during the 1830s climbed approximately as high as it has in the decades since 1971. (21) The second rendered the data in groups of five Justices by date of appointment, to ensure that each group had an equal number of observations. It revealed only a modest recent increase in average tenure. (22)

   In the revised version of their article, Calabresi and Lindgren respond to our period-selection criticism in several ways. Calabresi and Lindgren criticize our two charts, finding both of them lacking for methodological reasons, (23) stating at one point that we "never bothered to check [our] factual claim" about period selection. (24) The authors include new charts showing a "lagging average," representing the average for overlapping groups of nine Justices for each of their metrics for life tenure. (25) Finally, they include the results of a cubic regression model that ostensibly "confirm[s]" their original rendering of the data. (26)

   Before responding to the revised version of Calabresi and Lindgren's article, we should clearly identify the point of disagreement. Calabresi and Lindgren's empirical claim about life tenure has two features. First, they contend that changes in the actual operation of life tenure since 1970 have been "dramatic." Second, they argue that the changes are historically "unprecedented." Both aspects of their claim deserve fuller explanation.

   First, Calabresi and Lindgren argue that the data on life tenure reveal a "dramatic" trend. They variously describe the change as "critical and significant," (27) "remarkable," (28) and "astounding." (29) They use the word "astonishing" twice, (30) and some form of the word "dramatic" no less than eight times. (31) The transformation in life tenure, we are assured, is nothing short of mind-boggling.

   Of course, these are verbal formulations, not mathematical ones, and reasonable people might disagree about whether a particular increase is "dramatic" or "astounding." Perhaps Calabresi and Lindgren mean that the difference in data since 1971 is statistically significant. At one point, for example, they take pains to demonstrate that the difference between the length of tenure for "the last twelve retirees as a group" and the length of tenure for "the typical Justice leaving the bench through 1970" is "more than large enough to be statistically significant (p=.0002)." (32) By that standard, however, the 13.4-year increase in average tenure during the period from 18211850 was equally dramatic (p=.003), (33) and in hindsight, so was the period from 1789-1820 (p=.00016). (34) Moreover, as Calabresi and Lindgren acknowledge, statistical significance shows only that "[t]he data do not appear to be random," (35) but not necessarily that the trend in the data is "astounding" or "dramatic."

   Thus, when Calabresi and Lindgren say that the recent trend is "dramatic," they must mean something other than a statistically significant change. Indeed, they appear to assert that the rate of increase in the recent trend is especially pronounced. Calabresi and Lindgren write, for example, that "one of the main contributions" of their work has been to show that the increase in Supreme Court tenure is "not a gradual, steady climb." (36) Although this aspect of the empirical claim defies exact quantification, the authors plainly have in mind an exponential or accelerating increase, as opposed to a linear or decelerating increase.

   Second, Calabresi and Lindgren argue that Supreme Court Justices now remain on the Court "for longer periods ... than ever before in American history." (37) They use the term "unprecedented" at least five times. (38) This aspect of the hypothesis is more susceptible to testing: if...