Ex-presidential approval: retrospective evaluations of presidential performance.

Panagopoulos, Costas

Political observers and presidents alike are preoccupied with the notion of presidential legacy. Speculation about how presidential actions will be viewed by future generations weighs heavily on the minds of chief executives as well as historians and analysts who seek to devise standards and techniques by which to measure the concept of legacy. To be sure, efforts to gauge presidential "greatness" not only vary significantly, but they often yield inconsistent results. One approach relies on expert judgments to evaluate presidential performance and to rate presidents against each other. Historians and other scholars will often rate past presidents based on a series of established criteria that may include the degree to which presidents pursue active legislative agendas, progress, and reform; possess wisdom, sagacity, administrative prowess; and are principled (Cronin and Genovese 2004). In 1996, the New York Times Magazine, aided by preeminent presidential historian Arthur Schlesigner Jr., commissioned a survey of historians about presidential performance. Respondents rated past presidents as "great," "near great," "average (high)", "average (low)," "below average," or "failure." The findings revealed that Abraham Lincoln, Franklin D. Roosevelt (or FDR), and George Washington were the only presidents considered to be "great," while Franklin Pierce, James Buchanan, Andrew Johnson, Ulysses Grant, Warren Harding, Herbert Hoover, and Richard Nixon were viewed as failures (Schlesinger, 1996, 48-49).

A similar survey of 58 presidential historians and political scientists conducted in 1995 by the Chicago Tribune shows that the 10 best presidents in U.S. history have been Lincoln, Washington, FDR, Thomas Jefferson, Theodore Roosevelt, Woodrow Wilson, Harry Truman, Andrew Jackson, Dwight Eisenhower, and James Polk, while the 10 worst are Harding (worst), Buchanan, Pierce, Grant, A. Johnson, Millard Filmore, Nixon, John Tyler, Calvin Coolidge, and Hoover (Neal 1995). By 2000, a Wall Street Journal survey of 73 scholars of history, politics, and law showed that Ronald

Reagan had entered the presidential pantheon of "top ten best" presidents, while Jimmy Carter was rated amongst the ten worst (Lindgren and Calabresi 2000).

Expert judgments about presidential greatness often do not coincide with the assessments of the public at large. Consider that in a survey conducted by the Gallup organization (national random sample) in 2003, 13% of Americans rated John E Kennedy as the greatest American president, even though Kennedy consistently registers as average or above average at best in surveys of experts (Cronin and Genovese 2004). In fact, Kennedy routinely receives high marks from Americans in opinion surveys relative to his White House predecessors and successors, despite the fact that experts are less enthusiastic about his presidency.

These observations suggests that the public's assessments of past presidential performance are worthy of study in their own right. How does the public evaluate the performance of past presidents? How do these assessments change over time? What factors help to explain the patterns we observe? Analysis of the available survey data on retrospective evaluations of past presidential performance help to advance answers to these questions.

Presidential and Past Presidential Approval

Americans are routinely surveyed about their impressions of incumbent presidential job performance. Scholars have shown that macrolevel presidential approval is dynamic and that it is influenced by social, political, and economic events (Erikson, MacKuen, and Stimson 2002). Polling organizations typically stop asking Americans to rate a president's job performance after presidential terms expire or as presidents otherwise depart the Oval Office.

In recent years, polling organizations have occasionally queried respondents to provide retrospective assessments about the performance of former presidents. These data enable us to examine developments over time in the public's perceptions of ex-presidential job approval. Despite the availability of these data, only one study of which I am aware (King 1999) has advanced a systematic evaluation of retrospective presidential evaluations. In this study, the author analyzed only two waves of retrospective approval ratings as gathered by Gallup in surveys conducted in 1990 and 1993 (King 1999). (1) The Gallup organization subsequently probed respondents about their assessments of former presidents in surveys conducted over seven different years, but analyses of these data have largely escaped scholarly inquiry. As a result, our understanding of ex-presidential approval and the factors that influence retrospective judgments of presidential performance remains limited. This study aims to fill this void and to address this question.

The traditional view commonly held by historians and political scientists alike suggests that popularity ratings for an incumbent president ultimately have no lasting impact on how the president's legacy, reputation, and performance in office are judged (Bishop, Mockabee, and Rademacher 2006; see also Ceaser 1988; Cronin and Genovese 2004; Neustadt 1980). This claim rests on tenuous theoretical grounds, however, and, as I discuss below, there are compelling theoretical reasons to expect stronger linkages between concurrent presidential approval and retrospective evaluations of presidential performance in office.

This article proceeds as follows. The next section describes the data used for the analyses. In the following section, I describe changes in presidential approval between surveys conducted during presidential terms and data available from polls conducted after presidential terms are over. I then develop a theoretical model to explain ex-presidential approval and estimate an empirical model to test the hypotheses I develop. The final section provides some concluding observations and remarks about the overall findings and their implications.

Data

Since the 1930s, polling organizations have asked respondents to rate the performance of the incumbent president. Respondents are typically asked, "Do you approve or disapprove of the way [president] is handling his job as president?" Edwards (2004) asserts this question is one of the most frequently asked survey items asked consistently since the advent of modern polling. Data on presidential approval is available for every president since Roosevelt.

Since 1990, the Gallup organization has also asked respondents to rate the job performance of former presidents. Respondents are asked the following question: "From what you have heard, read, or remember about some of our past presidents, please tell me if you approve or disapprove of the way they handled their job as president.... [President]." (2) Americans' retrospective evaluations of former presidents are available for every president since Roosevelt. These data are analyzed in the following section.

Concurrent vs. Retrospective Evaluations of Presidential Performance

In this section, I compare concurrent versus retrospective evaluations of presidential performance. Following Stimson (1976), all indicators of approval (for current as well as former presidents) used throughout this article are measured as relative approval, defined as 100 * [Percent Approve/(Percent Approve + Percent Disapprove)]. This operationalization corrects for bias that may result from systematic nonresponse and converts the approval ratio into a percentage. All surveys included in the analyses are conducted by the Gallup organization.

Figure 1 displays average approval ratings for incumbent presidents FDR to George W. Bush. Also depicted in the figure are average retrospective approval ratings for each president. Detailed data are presented in Table 1. Analysis of these data suggests that retrospective evaluations of former presidents tend to be more positive than assessments of presidential performance at the time. Table 1 indicates that postpresidency approval ratings exceed or are equivalent to mean levels of job approval for eight out of the 12 presidents included in the analysis. In fact, retrospective evaluations of presidential performance tend to be five percentage points higher on average compared to mean levels of approval throughout a president's term. In four cases (Lyndon Johnson, Nixon, Bush I and Bush II), ex-presidential approval appears to be lower than average approval during each respective presidency. It is important to note that mean job approval ratings for at least three of these presidents may be inflated by high approval ratings at the beginning of their terms. Lyndon Johnson's approval ratings were highest immediately following the Kennedy assassination and deteriorated over the course of his presidency; Nixon's approval ratings plummeted in the wake of Watergate; and George W. Bush's approval peaked during the 9/11 crisis that occurred relatively early-on in his first term.

[FIGURE 1 OMITTED]

Table 2 presents a comparison of average ex-presidential approval in postpresidency surveys with presidential approval in the last poll taken by Gallup during a president's term in office. Data presented in Table 2 indicate retrospective assessments of presidential performance can increase substantially long after the president leaves the White House. Retrospective evaluations of presidential job performance are 12 percentage points higher on average than approval in the final poll during a presidential term. Truman, who left the White House in 1953 with poor marks for job performance from the American public, appears redeemed in postpresidency polls, gaining 53 percentage points in approval. Similarly, Carter's average retrospective approval rating is--at 59 percent approval--21 percentage points higher than the final Gallup poll conducted before he left the office in 1981.

Figure 2 depicts the dynamics of ex-presidential approval in polls since 1990 for each president. (3) The solid line in each figure represents approval rating in each of the postpresidency polls that queried respondents about the corresponding president. The dashed line in each figure represents presidential approval in the last poll conducted by Gallup during each presidential term, and the dotted line in each diagram indicates the mean level of approval for each president over the duration of their administration. The evidence presented in Figure 2 confirms that former presidents often recover from public disapproval during their presidencies as citizens reflect on their administrations. The data in Figure 2 also indicate that retrospective evaluations of ex-presidential performance are dynamic. The perceptions of former presidential performance shift over time, experiencing ebbs and flows that are frequently sizeable and noteworthy. The following section offers insights about how these observed changes can be understood.

Explaining Ex-Presidential Approval

The task remains to explain levels of ex-presidential approval. In this section, I develop a theoretical model to predict ex-presidential approval as a function of a variety of covariates. I develop a series of hypotheses that I test by estimating the empirical models that follow.

An assertion commonly found in the scholarly literature suggests that retrospective evaluations of presidential performance should be unrelated to concurrent presidential job approval assessments. As Cronin and Genovese (2004, 84) argue, "[p]opularity in office is not a useful gauge for judging reputation." Similarly, Caesar posits, "[i]t is worth emphasizing how little approval rating have to do with any lasting judgment of presidential performance. A president's legacy derives from his accomplishments or failures, and no president will be long remembered for having an average approval rating of more than 60 percent, nor quickly forgotten for having an average lower than 45 percent" (quoted in Cronin and Genovese 2004, 84). Such claims are often advanced without empirical verification or theoretical justification. More plausible expectations can be derived from theories of opinion formation and information processing that suggest attitudes and opinions will be influenced by various factors including recollections of prior performance. In his memory-based model of information processing, Zaller (1992) argues that considerations about politics processed and stored in memory will be activated and called "to the top of the head" when respondents are asked for a response in an opinion poll. Zaller (1992) asserts that individuals will resist new information that is inconsistent with preexisting beliefs and that more accessible considerations carry more weight in formulating a response. This model suggests that memories of past assessments of job approval for former presidents may be activated when rending retrospective judgments of job performance. If true, this would imply that factors that explain contemporary presidential approval, including economic performance and the impact of unpopular wars (Edwards 2004), remain salient long after presidents leave office. Thus, we should expect presidential approval during a presidential term to impact post-presidency assessments of performance. The claims advanced by the Zaller (1992) model imply the following two hypotheses:

H1: Mean presidential approval during tenure in office will be positively associated with the level of ex-presidential approval, ceteris paribus.

H2: Presidential approval in the final poll conducted during a presidential term will be positively associated with the level of ex-presidential approval, ceteris paribus.

[FIGURE 2 OMITTED]

In contrast to Zaller (1992), Lodge, McGraw, and Stroh (1989) develop an online model of information processing that suggests opinions are updated when new information is encountered. This theory would suggest that new information about the job performance of former presidents would influence respondents' retrospective assessments of presidential approval. Moreover, it is conceivable that information about the performance of the incumbent president will influence respondents' retrospective approval ratings of former presidents, particularly for copartisans of the incumbent. As social psychologists argue, the stimuli present at the time of retrieval (retrieval cues) influences how encoded material is retrieved from memory (Richardson-Klavehn and Bjork 1988). From this, the following interaction hypothesis is derived:

H3: Incumbent presidential approval will be positively associated with ex-presidential approval, but only for former presidents of the same party as the incumbent, ceteris paribus.

Research by social psychologists has also asserted that individuals selectively retain information. People are naturally predisposed to resisting negative information. Moreover, studies of memory as a blend of cognitive and affective processes suggest that individuals are prone to nostalgia, a tendency to idealize the past (Stern 1992). As presidential scholar Richard Neustadt (1980, 202) puts it, "[r]etrospect can induce romance." Thus, it is plausible that individuals' retrospective evaluations of the performance of former presidents will be romanticized, retaining more favorable recollections and knowledge while discounting less favorable memories and information. As Stanley Renshon (2000, 41) puts it, "[p]residents are remembered for their accomplishments." The following two hypotheses emerge from these expectations:

H4: Ex-presidential approval will be positively related to the length of time the president has been out of office, ceteris paribus.

H5: Deceased presidents will be recalled more favorably than living former presidents, ceteris paribus.

It is also conceivable that there are partisan differences in the retrospective evaluations of former presidents. Americans' recollections of Democratic presidents may differ systematically from Republican executives. While I find little theoretical guidance to predict the direction of such an effect, I include an indicator of the former president's party to account for this possibility.

To test these hypotheses, the following multivariate model will be estimated using OLS regression analysis:

Y (Ex-Presidential [Approval.sub.i.t])

= a (Constant)

+ [b.sub.1] (Mean Presidential Approval during [Presidency.sub.i])

+ [b.sub.2] (Presidential Approval in the Final Poll Conducted During [Presidency.sub.i])

+ [b.sub.3] (Incumbent Presidential [Approval.sub.t])

+ [b.sub.4] (Ex-President-Incumbent President Same Party [Dummy.sub.t])

+ [b.sub.5] (Ex-President-Incumbent President Same Party Dummy

x Incumbent Presidential [Approval.sub.t])

+ [b.sub.6] (Total Years Elapsed since President Left [Office.sub.i.t])

+ [b.sub.7] (Ex-President [Democrat.sub.i])

+ e (error term)

The dependent variable in the analysis is the level of retrospective approval for former president i at time t of the postpresidency survey. To test the interaction effects hypothesis requires the inclusion of the ex-president-incumbent president of the same party dummy and incumbent presidential approval variables as well as their interaction term. Interaction effects exist when the interaction term is statistically significant after controlling for the variables that went into the construction of the interaction term.

Results

The results of the multivariate analyses are presented in Table 3. I estimate two specifications using ordinary least squares regression analysis. Model 1 does not include the interaction term, while model 2 includes the interaction term and its component variables. (4) The evidence presented in Table 3 reveals support for most of the hypotheses developed above, and the estimates remain robust across specifications. Retrospective approval of presidential performance increases the longer former chief executives are out of office. This finding is consistent with the general pattern observed by King (1999). Accordingly, the results I report lead me to concur that, "[w]hen it comes to assessing presidential performance, absence makes the heart grow fonder" (King 1999, 170). All else equal, I find postpresidency approval ratings grow by 0.8 percentage points for each year a president has been out of office.

The results also indicate ex-presidential approval is responsive both to the corresponding president's overall approval rating while in the White House and to the outgoing president's level of approval as measured by the final poll conducted during his tenure, all else equal. This evidence suggests the effects of factors that drive concurrent presidential evaluations linger, exerting long-term influence on how presidents are subsequently judged. Thus, even though presidents may be able to rehabilitate appraisals of their performance while in office over time, they cannot escape entirely from the conditions that affected levels of public approval during their presidencies.

Performance evaluations of the incumbent president at the time of the survey querying respondents about ex-presidential approval also appear to influence retrospective evaluations of former presidents, controlling for the other variables in the model. The effect of incumbent presidential approval exerts a positive effect on retrospective evaluations, however, regardless of whether the incumbent president and the former president share partisan identity. The interaction term in model 2 does not achieve significance at conventional levels.

Overall, retrospective appraisals of presidential performance do not seem to differ systematically by whether the former presidents were Democrats or Republicans, all else equal. Finally, and contrary to expectations, deceased presidents appear to be evaluated far less favorably than living former presidents in postpresidency approval polls. Retrospective appraisals of deceased presidents are about 18 percentage points lower on average compared to living presidents, ceteris paribus. Taken together, the variables included in the models explain about 70% of the variance in the dependent variable.

Conclusions

This study represents one of only a few efforts to examine the dynamics of retrospective evaluations of presidential performance. The findings reveal key insights that help to understand presidential legacies as captured in opinion polls conducted after presidents have left the White House. The analyses demonstrate that retrospective approval of presidential performance is generally higher than approval for performance during the president's administration and that retrospective assessments are dynamic, tending to strengthen as the presidential term falls deeper into the past. Moreover, retrospective judgments about presidential performance are linked to job approval while in office as well as to evaluations about the incumbent president at the time of postpresidency surveys.

The analyses described in this study help us to understand and to explain patterns in retrospective approval of past presidential performance. The results help to illuminate the forces driving the dynamics we observe and help to identify factors that impact the public's perceptions of presidential legacies.

References

Bishop, George E, Stephen Mockabee, and Eric Rademacher. 2006. "Presidential Approval Ratings: Meaning, Measurement and Validity." Paper presented at the Annual Meeting of the American Political Science Association, Philadephia, PA, September 1.

Ceaser, J. 1988. "The Reagan Presidency in American Public Opinion." In The Reagan Legacy: Promise and Performance, ed. Charles O. Jones. Chatham, NJ: Chatham House, 220.

Cronin, Thomas E., and Michael E. Genovese. 2004. The Paradoxes of the American Presidency. 2nd ed. New York: Oxford University Press.

Edwards, George C. 2004. On Deaf Ears: The Limits of the Bully Pulpit. New Haven, CT: Yale University Press.

Erikson, Robert S., Michael MacKuen, and James Stimson. 2002. The Macro Polity. New York: Cambridge University Press.

King, James D. 1999. "Looking Back at Chief Executives: Retrospective Presidential Approval." Presidential Studies Quarterly 29 (1): 166-74.

Lindgren, J., and S. Calahresi. 2000. "Ranking the Presidents," Wall Street Journal, November 16.

Lodge, Milton, Kathleen McGraw, and Pat Stroh. 1989. "An Impression-driven Model of Candidate Formation." American Political Science Review 83 (2): 399-420.

Neal, S. 1995. "Putting Presidents in Their Places." Chicago Sun-Times, November 19.

Neustadt, Richard. 1980. Presidential Power. 2nd ed. New York: John Wiley and Sons.

Renshon, Stanley. 2000. "After the Fall: The Clinton Presidency in Psychological Perspective." Political Science Quarterly 115 (1): 41-65.

Richardson-Klavehn, Alan, and Robert A. Bjork. 1988. "Measures of Memory." Annual Review of Psychology 39: 475-543.

Schlesinger, Arthur, Jr. 1996. "The Ultimate Approval Rating." New York Times Magazine, December 15.

Stimson, James A. 1976. "Public Support for American Presidents: A Cyclical Model." Public Opinion Quarterly 40 (1): 1-21.

Stern, B. B. 1992. "Historical and Personal Nostalgia in Advertising Text: The Fin-de-siecle Effect." Journal of Advertising 21 (4): 11-22.

Zaller, John. 1992. The Nature and Origins of Mass Opinion. New York: Cambridge University Press.

COSTAS PANAGOPOULOS

Fordham University

(1.) Gallup also asked a retrospective approval item in 1992, but this was excluded from the analyses reported in King (1999). These data are included in the current study.

(2.) The Gallup Poll has included this item in surveys conducted in November 1990, June 1992, November 1993, September 1994, January 1998, February 1999, February 2000, March 2002, June 2006, and November 2010. The 1993 and 1994 surveys were conducted by Gallup/CNN/USA Today.

(3.) Postpresidency approval for Roosevelt, Truman, and Eisenhower are only available in 1990, and for George W. Bush in the 2010 survey. Since ex-presidential approval data for these presidents are not available over time, they are excluded from Figure 2.

(4.) The variance inflation factor for model 1 (VIF = 1.88) suggests there is little reason to be concerned about multicollinearity. Collinearity diagnostics are less useful when interactions are included in models, but even so, the VIF for model 2 is below 10 (9.33), suggesting there is little cause for alarm.

Costas Panagopoulos is associate professor of political science and director of the Center for Electoral Politics and Democracy at Fordham University. He serves as coeditor of the "Polls and Elections" feature for Presidential Studies Quarterly.

TABLE 1
Average Presidential and Postpresidential Approval

President Average Average Difference
 Approval Approval
 during Postpresidency
 Presidency

Roosevelt 66 91 +25
Truman 55 89 +34
Eisenhower 76 88 +12
Kennedy 80 88 +8
Johnson 64 50 -14
Nixon 56 35 -21
Ford 56 69 +13
Carter 54 59 +5
Reagan 58 65 +7
Bush I 70 68 -2
Clinton 61 61 0
Bush II 56 48 -8
Average 63 68 +5

Source: Compiled by author from Gallup surveys.

TABLE 2
Average Postpresidential Approval and Presidential Approval
in the Last Poll Taken during a President's Term in Office

 Approval
 in Last Average
 Poll of Approval
President Presidency Postpresidency Difference

Roosevelt 73 91 +18
Truman 36 89 +53
Eisenhower 68 88 +20
Kennedy 66 88 +22
Johnson 56 50 -6
Nixon 27 35 +8
Ford 62 69 +7
Carter 38 59 +21
Reagan 69 65 -4
Bush I 60 68 +8
Clinton 70 61 -9
Bush II 36 48 +12
Average Change +12

Source: Compiled by author from Gallup surveys.

TABLE 3
Explaining Ex-Presidential Approval

Dependent Variable: Retrospective Approval of Former President

Independent (1) (2)
Variables

Total Years .81 *** .81 ***
Out of Office (.17) (.17)

Mean Approval .62 *** .64 ***
in Office (.19) (.20)

Approval in Final .63 *** .62 ***
Presidential Poll (.10) (.10)

Incumbent .25 ** .37 ***
Presidential (.11) (.16)
Approval

Deceased -18.25 *** -18.64 ***
 (4.12) (4.18)
Ex-President Democrat 2.66 4.46
 (2.74) (3.17)
Ex-President and Incumbent 13.00
President Same Party (13.46)

Interaction (Incumbent -.28
Presidential Approval x (.25)
Ex-President and Incumbent
President Same Party)

Constant -32.91 *** -41.02 ***
 (11.11) (13.64)
N 69 69
R-squared .68 .69

Notes: OLS regression. Standard errors in parentheses.

*** Estimated effect is p < .01, ** p < .05, using two-tailed tests.