17

Click here to load reader

Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

Embed Size (px)

Citation preview

Page 1: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

This article was downloaded by: [Newcastle University]On: 20 December 2014, At: 00:39Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

The Journal of SocialPsychologyPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/vsoc20

Predicting PresidentialPerformance in the UnitedStates: Equation Replication onRecent Survey ResultsDean Keith Simonton aa Department of Psychology , University ofCalifornia , DavisPublished online: 03 Apr 2010.

To cite this article: Dean Keith Simonton (2001) Predicting Presidential Performancein the United States: Equation Replication on Recent Survey Results, The Journal ofSocial Psychology, 141:3, 293-307, DOI: 10.1080/00224540109600552

To link to this article: http://dx.doi.org/10.1080/00224540109600552

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinions and viewsexpressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of theContent should not be relied upon and should be independently verified withprimary sources of information. Taylor and Francis shall not be liable for anylosses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly orindirectly in connection with, in relation to or arising out of the use of theContent.

Page 2: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan,sub-licensing, systematic supply, or distribution in any form to anyone isexpressly forbidden. Terms & Conditions of access and use can be found athttp://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 3: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

The Journal of Social Psvcholowv. 2001.141(3). 293-301

Predicting Presidential Performance in the United States: Equation Replication

on Recent Survey Results

DEAN KEITH SIMONTON Department of Psychology

University of California, Davis

ABSTRACT. For more than 2 decades, researchers have tried to identify the variables that predict the overall performance of U.S. presidents. In 1986, there emerged a 6-variable prediction equation (D. K. Simonton, 1986c, 1987b) that has been replicated repeatedly. The predictors are years in office, war years, scandal, assassination, heroism in war, and intellectual brilliance. The author again replicated the equation on recent rankings of all presidents from George Washington through William Jefferson Clinton according to a sur- vey of 719 experts (W. R. Ridings, Jr., & S. B. McIver, 1997). The original 6-variable equation successfully predicted both the overall rankings as well as the 5 core components of the rankings (leadership qualities, accomplishment, political skill, appointments, char- acter and integrity). The predictive value of the equation was illustrated for the presiden- cies of Ronald W. Reagan, George H. W. Bush, and Clinton.

Key words: equation replication, predictor variables, presidential performance rankings, United States

MANY CITIZENS OF THE UNITED STATES remember a stage in childhood in which they said that they wanted to become president of the United States.Yet, the pathway to the White House passes through minefields so treacherous that, in more than two centuries, only a few dozen American men have reached that supposed pinnacle of political success. Moreover, once Inauguration Day is past, the presidency itself may be replete with its own obstacles and traps for the elec- toral victor. Not everyone who secures a first term in office secures a second. Herbert Hoover entered ofice with a landslide victory and was forced out of

This article was based in part on a paper read at the symposium “Studying Political Lead- ers” for the International Congress of Applied Psychology held in San Francisco on August 11, 1998.

Address correspondence to Dean Keith Simonton. Department of Psychology, One Shields Avenue, Universio of California, Davis, CA 95616-8686; dksimonton@ ucdavis.edu (e-mail).

293

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 4: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

294 The Journal of Social Psychology

office by a landslide defeat-all because of his inability to cope with events that he could not have foreseen. Even under the best of circumstances, the incum- bent’s performance is under scrutiny by fellow politicians, the press, and the pub- lic at large (Brody, 1991). Nor is an ex-president free from judgment after his successor has taken office; he must face a retirement in which journalists, polit- ical scientists, and historians discuss whether he was a great leader.

During the first century and a half of the nation’s history, retrospective eval- uations of presidential performance were entirely informal, subjective, unsys- tematic, and contentious. But such assessments became more objective when Schlesinger (1948) asked 55 historians to use five grades-great, near great, average, below average, failure-to rate 29 past chief executives. Later he pub- lished a more extensive poll in which 75 historians rated 31 former presidents from George Washington through Harry S. Truman (Schlesinger, 1962). Many other presidential experts followed. Some conducted ever more extensive surveys (e.g., Maranell, 1970; Murray & Blessing, 1983). whereas others simply provid- ed their own opinions (e.g., Bailey, 1966; Sokolsky, 1964). As more recently retired presidents have been added to the historic roster, researchers have peri- odically updated such surveys (Murray BL Blessing, 1988, 1994).

According to multivariate analyses, the performance ratings just described represented a firm consensus about the best and worst presidents in U.S. history (Simonton, 1986b, 1987b). The various measures, no matter how diversely defined, can all be considered indicators of a single latent factor (Simonton, 1986b, 1991a). Equally important, this consensus seemed to transcend such potential contaminants as the rater’s age, gender, geographical region, specialty area, academic affiliation, professional status, and political ideology (Maranell & Dodder, 1970; Murray & Blessing, 1983). Moreover, because the consensus was based on the views of numerous knowledgeable judges, the evaluations enjoy face validity. Thus, there is a strong prima facie case that these greatness assess- ments reveal how U.S. presidents varied in their effectiveness as the nation’s highest political leader. Indeed, this global consensus is strongly correlated with many alternative indicators of presidential success, such as a sculpture on Mount Rushmore; a monument in Washington, DC; election to the American Hall of Fame; the early casting of a coin with the president’s profile; a low denomination of currency with the president’s portrait; and the amount of the president’s cov- erage in standard reference works (Simonton, 1987b).

Given the demonstrated reliability and apparent validity of the performance ratings just described, many researchers have tried to isolate the predictors of presidential greatness. Wendt and Light (1976) published the first such study, and many others followed suit, both psychologists (McCann, 1992; Simonton, 1981; Winter, 1987) and political scientists (Kenney & Rice, 1988; Nice, 1984). As might be expected, the various investigators usually examined different potential predictors and, almost as frequently, adopted divergent methodological ap- proaches (e.g., Deluga, 1997; Holmes & Elder, 1989; Spangler & House, 1991).

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 5: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

Sirnonton 295

As a consequence, the consensus about the predictors of performance is not as strong as the consensus about the criterion variable itself. Nonetheless, over the years, a single model has emerged that seems to have the greatest predictive validity (Simonton, 1986c, 1987b). The resulting equation contains six predictors of a president’s greatness: (a) the number of years in the nation’s highest office, (b) the number of years as a wartime commander in chief, (c) whether his admin- istration was tarnished with a major scandal, (d) whether he was assassinated in office, (e) whether he had entered office as a nationally recognized war hero, and (f) his overall level of intellectual brilliance. Five features of this six-variable model deserve special mention:

1. The correlation between the predicted and observed values of greatness is typically in the lower .90s; thus, the equation has usually accounted for at least 80% of the variation in rankings of presidential performance (e.g., Simonton, 1 9 8 6 ~ 1996). No other combination of variables has explained more variance than the cluster of six variables in the preceding paragraph (Simonton, 1991c, 1992; cf. McCann, 1992).

2. The predictive value of these six variables has not changed over time; rather, the effects have been transhistorically invariant (Simonton, 1987b).

3. The equation has been successfully replicated in an experimental simula- tion. Naive undergraduates, evaluating anonymous political leaders via bio- graphical sketches, reproduced the ratings assigned to the presidents by the experts, by using the same predictors and spontaneously allotting those predic- tors the same weights in their judgments (Simonton, 1986a, 1987b).

4. The equation corresponds closely to what has been found in several empirical investigations of leadership, political or otherwise (Cell, 1974; Simon- ton, 1984, 1987b, 1995; Sorokin, 1925, 1926). Especially significant is the role of intelligence, a variable that has been the most robust predictor of leader per- formance in a diversity of situations (Cox, 1926; Simonton, 1983, 1985a, 1991b; Walberg, Rasher, & Hase, 1978).

5. In earlier studies (Simonton, 1988, 1991c, 1992, 1996), I replicated the original equation by using additional predictors and control variables, alternative performance measures, and even enlarged samples of presidents. In all cases, the new variables did not replace any of the six predictors or produce a significant increment to the explained variance (e.g., Baltzell & Schneiderman, 1988; Delu- ga, 1997, 1998; Holmes & Elder, 1989; Kenney & Rice, 1988; Winter, 1987).

Nevertheless, Ridings and McIver (1997) have published a survey that pro- vides assessments of all presidents from George Washington through William Jef- ferson Clinton. Because almost all earlier surveys (Simonton, 1987b) omitted William Henry Harrison and James Garfield as well as Ronald W. Reagan and George H. W. Bush, I undertook the equation’s replication on a larger sample in the present study. If the prediction equation was truly as robust as claimed

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 6: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

296 The Journal of Social Psychology

(Simonton, 1987b), it would prove as effective with this enlarged sample. Fur- thermore, besides an overall ranking, previous researchers had their experts rank the presidents on what were considered the five most crucial components of per- formance (cf. Maranell, 1970): (a) leadership qualities, (b) accomplishment, (c) political skill, (d) appointments, and (e) character and integrity. Thus, I wanted to determine whether the equation would hold up across those separate components.

Method

Clinton was inaugurated as the 42nd president of the United States but was only the 41st person to serve (because Grover Cleveland served two nonconsec- utive terms as the 22nd and 24th president). Because Ridings and McIver (1997) had their 7 19 survey respondents rate all the presidents from Washington through Clinton, it is possible to have N = 41 as the sample size. However, I deleted Clin- ton from most analyses because he had not yet completed his term of office at the time of the present research. In addition, I omitted G. H. W. Bush from the main analyses, for lack of a suitable measure of intellectual ability. For most of the pre- sent analyses, therefore, the sample size was 39 rather than 41.

Performance Criteria

Ridings and McIver ( 1997) provided the methodological details behind the questionnaires and their sampling procedures. All of their data were ordinal measures that ranked the presidents (1, 2, 3, and so forth) on overall perfor- mance as well as on the following five components: (a) leadership qualities, (b) accomplishment, (c) political skill, (d) appointments, and (e) character and integrity. To make these six scores more comparable to the performance mea- sures used in previous studies (Simonton, 1987b), I inverted them all by sub- tracting each score from 42. The results are variables in which 41 represents the highest possible score and 1 the lowest possible score (M = 21.0, SD = 12.0 for all six measures).

I then took several steps to determine the reliability and validity of the six measures. The first was to examine how these measures were correlated with each other. The correlations were substantial. The correlations of the overall assessment with the five component assessments ranged from .72 to .94 (all ps c .001); the correlations of the five components among each other ranged from .37 to .92 (all ps c .05). The lowest correlations uniformly resulted from the charac- ter and integrity measure. When I deleted the measure of character and integrity, the foregoing correlations ranged from .87 to .96 (all ps < .001) and from .76 to .92 (all ps < .001), respectively. Thus, assessments of the president’s character and integrity appeared to be less related than the remaining four variables to leader performance (Simonton, 1981, 1983, 1984). The lowest correlation ( r = .37) was that of character and integrity with political skill, an attribute associat-

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 7: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

Simonton 297

ed with being a “wheeler-and-dealer.” In fact, political skill has been positively associated with a Machiavellian personality (Simonton, 1986~).

I further scrutinized the associations among the criterion variables by sub- jecting the five component assessments to a principal component analysis (Simonton, 1986b). Only one factor had an eigenvalue exceeding unity (4.04, whereas the second was only 0.71). Hence, the correlations among the five com- ponents were adequately explained by a single latent factor that accounted for approximately 81% of the total variance, a highly notable figure (Simonton, 1991a). The factor loadings ranged from .71 (character and integrity) to .97 (accomplishment). The lowest loading was .89 (political skill) when the charac- ter and integrity variable was ignored. When I combined the five measures into a single (equal-weight) additive-linear composite, the internal-consistency reliabil- ity (a) became .94, a notable figure (Simonton, 1990). Deleting the character and integrity component increased the reliability coefficient only to .96. As a final validation, I regressed the overall rating on all five components and obtained an equation that accounted for 98% of the total variance. Although all components emerged as statistically significant predictors 0, < .O l ) , their regression weights, both standardized (p) and unstandardized (B), ranged from .18 (character and integrity) to .26 (appointments). The second poorest predictor (political skill) had a regression coefficient of .20. When I combined these results with those of the principal component and reliability analyses, the five components all seemed to measure the same attribute, albeit the character and integrity component seemed less conspicuously associated with the global consensus than the other four com- ponents were. Even then, the concurrence of character and integrity with the other four components was substantial.

The next step was to determine how the Ridings-McIver (1997) ratings were correlated with comparable measures published earlier. I accomplished this com- parison most easily for the overall assessment measure, which exhibited strong positive relations with the following (all p s c .001; Ns = presidents evaluated): .94 (N = 29) with Schlesinger (1948), .80 (N = 30) with Rossiter (1956), .92 (N = 31) with Schlesinger (1962), .78 (N = 3 1) with Bailey (1966; as quantified by Kynerd, 1971), .96 ( N = 36) with a poll conducted for the Chicago Tribune Magazine (reported in Murray & Blessing, 1983), .96 ( N = 36) with Murray and Blessing (1983), and .96 (N = 39) with Kelly and Lonnstrom (1990). If one makes allowance for the contrasting measurement scales, rating sources, and other differences, the foregoing correlations are impressive (Simonton, 1987b). Clearly, the Ridings-McIver ratings did not depart from the previous evaluations of overall presidential performance.

Unfortunately, it was more difficult to find comparable measurements for the five component measures. Maranell ( 1970) published 571 American historians’ ratings of the presidents on a few relevant dimensions. The correlations of the Ridings-McIver ( 1997) assessment of accomplishments were .92 and .90 with Maranell’s assessments of accomplishments and presidential strength, respec-

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 8: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

298 The Journal of Social Psychology

tively (N = 33, all p s < .001). Still, even as the Ridings-McIver five component measures were highly intercorrelated, so were the relevant Maranell indicators; therefore, it was impossible to obtain any discrimination in the correlations. For the two sets of measures, the correlations ranged from .71 to .92 (all ps < .001). For the Ridings-McIver overall assessment, the correlations with the Maranell indicators of presidential prestige, strength, activity, and accomplishments ranged from .84 to .92. Therefore, all the intercorrelations probably reflected the underlying latent variable of global presidential performance (Simonton, 199 1 a).

Performance Predictors

The six predictors were the same as in previous inquiries (Simonton, 1986a, 1987b, 1991c, 1992). I updated them for the most recent former presidents by using various reference works (e.g., DeGregorio, 1993):

Years in ofice. This variable was defined as the number of years served, to the nearest tenth of a year (M = 5.09, SD = 2.54; range = 0.1 to 12.1). For most pres- idents, the resulting number was either 4.0 or 8.0, according to the number of terms served. However, some had noninteger values, such as those who died in office and those who became “accidental presidents” (Simonton, 1985b).

War years. This predictor was defined as the number of years during which the nation was at war, given as an integer value (A4 = 0.95, SD = 1.62, range = 0 to 5). G. H. W. Bush, for example, was granted 1 year’s credit for the Persian Gulf War.

Assassination. This 0-1 dummy variable equaled 1 if the president was assassi- nated and 0 if otherwise (M = 0.10, SD = 0.30). The assassinated presidents were Abraham Lincoln, Garfield, William McKinley, and John E Kennedy.

Scandal. This 0-1 dummy variable equaled 1 if the administration suffered a major scandal and 0 if otherwise (M = 0.10, SD = 0.30). A scandal entailed the commission of illegal acts by uppex-level officials within the executive branch. In addition to Ulysses S. Grant, Warren G. Harding, and Richard M. Nixon, this predictor applied to Reagan (the Iran-Contra scandal).

War hem. This 0-1 dummy variable equaled 1 if the president had previously attained a national reputation as a war hero and 0 if otherwise (M = 0.18, SD = 0.38). The following presidents were considered to qualify: Washington, Andrew Jackson, W. H. Harrison, Zachary Taylor, Grant, Theodore Roosevelt, and Dwight D. Eisenhower.

Intellectual brilliance. This standardized z score was based on a factor analysis of 110 assessments of all 39 presidents from Washington through Reagan (range = -2.00 to 3.10). Consequently, there was no analogous measure for G. H. W. Bush or Clinton in the present study. The 12-item composite has an internal-

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 9: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

consistency reliability (a) of .90. The assessments of the original items were made by eight independent raters on the basis of personality descriptions with all iden- tifying material removed (Simonton, 1986~). The highest scorer was Thomas Jefferson; the lowest, Harding.

Some of these measurements might be improved (McCann, 1992; Winter, 1987). For example, war years might be more accurately gauged to the first deci- mal fraction, and scandal might be weighted according to the magnitude of the president’s personal responsibility or knowledge (e.g., NixonNatergate vs. Rea- gankanxontra). However, I decided to retain the original operational definitions, to have the present results as comparable as possible to those in earlier studies.

The purpose of the present study was twofold: (a) to test whether these six variables would retain their predictive power with respect to the new evaluations of overall presidential performance and (b) to determine the predictive value of each variable with respect to each of the five performance components.

Results

I regressed the Ridings-McIver (1997) overall rating on the six predictors (full model; for results, see Table 1). All regression coefficients were in the expected direction, and all were statistically significant; hence, the results repli- cated the earlier findings, this time with W. H. Harrison, Garfield, and Reagan in the sample. When I deleted the intellectual brilliance measure (truncated model), I added G . H. W. Bush as well (for the outcome, see Table 1). Again, the results were as expected, and they may be compared with those of two earlier studies (Simonton, 1986b, 1986c) using the Murray and Blessing (1983) survey of 846

TABLE 1 Predictors of the Ridings and McIver (1997) Ratings

of Overall Presidential Performance

Variable Full model Truncated model

B P t B P t

Years in office 2.52 .53 5.36*** 2.68 .56 5.58*** War years 2.27 .30 3.00** 2.76 .37 3.68*** Assassination 8.54 .21 2.30* 10.53 .26 2.84** Scandal -15.09 -.38 -4.33*** -16.92 -.42 4.80*** War hero 6.92 .22 2.42* 6.34 .20 2.14* Intellectual brilliance 2.61 .21 2.17* - - -

Nore. I tested the full model on 39 presidents, yielding a constant of 5.48 (R2 = .77). In the truncated model, I deleted intellectual brilliance and, thus, tested it on 40 presidents, yielding a constant of 4.28 (R2 = .74). * p < .05. **p < .01. ***p < .001.

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 10: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

300 The Jouml of Social Psychology

experts. The full model accounted for 77% of the variance, whereas in the earli- er investigation (Simonton, 1986~). 79% of the variance was accounted for by the same predictors but in only 36 cases. The corresponding figures for the truncat- ed model and the results of an identical test in the earlier study were 74% and 78%, respectively (Simonton, 1986b). Thus, there was minimal shrinkage in the predictive validity of these equations. Moreover, the regression coefficients were highly similar as well, at least when one adjusts the standardized coefficients for the contrasting variance of the criterion and predictor variables (i.e., the Ps usu- ally differ by 5.05 or less). The only substantial discrepancy was that years in office moved from the third most important to most important predictor, proba- bly as a result of the addition of the 2 presidents with the shortest terms in office, W. €3. Harrison and Garfield.

The dependent variable in Table 1 is ordinal and, thus, rectilinearly distrib- uted. In comparison, the researcher assumes that the criterion is interval scaled and approximately normal in regression analysis (Darlington, 1990). According- ly, I defined a new variable consisting of the additive-linear composite of all five component variables, yielding a criterion that was a more reasonable approxima- tion to an interval scale and, at the same time, showed a much closer approxima- tion to a normal curve. Nevertheless, that new variable was so highly correlated ( r = .99) with the rank-order measure that the results of the two regression analyses were virtually identical, leaving all substantive conclusions completely unaltered. Moreover, the ordinal scale has the advantage that the unstandardized regression coefficients can be interpreted in terms of how the predictors change a president's rank. In particular, according to the full model, when a president enters office, he begins with a rating of 5.48 (the constant), unless he is also a former war hero, in which case he obtains an additional boost of 6.92 points. Each additional year that he serves increases his rating by 2.52 points, or a bit more than 10 points per full term. War years add yet another 2.27 points, and assassination in office augments his standing by 8.54 points. A major administration scandal, on the other hand, subtracts 15.09 points. Finally, for every standard deviation of his intellect above the presidential average, the chief executive's rating improves by 2.61 points. To convert these predicted ratings into the original rankings according to the Rid- ings-McIver (1997) survey, one needs only to subtract them from 42.

Because Ridings and McIver (1997) provided the ratings for the five core components, I also regressed them on the same six predictors (Table 2). Because all the dependent variables were on the same scale with identical means and stan- dard deviations and because the independent variables were always identical, only the unstandardized partial regression coefficients (Bs) are shown for the comparisons. Although these regression weights have the same sign across all five components, they are neither equal in magnitude nor uniformly statistically significant. It is also noteworthy that the amount of explained variance differed considerably. On the one hand, the ratings in leadership quality were better pre- dicted than the overall assessment, with 79% of the variance explained in con-

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 11: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

Simonton 301

TABLE 2 Predictors of the Five Performance Components:

Unstandardized Regression Coefficients (Bs)

Variable Leadership Political Character and

Accomplishment skill Appointments integrity

Years in office War years Assassination Scandal War hero Intellectual brilliance

Constant R*

2.61 *** 2.31**

1 I .4l**

10.65*** -6.06

3.06* 3.20 .79

2.47*** 2.37*** 2.07*** 1.58* 3.02*** 2.56** 1.96* 0.5 1 7.77 11.00* 6.95 1.05

-1 0.98** -5.74 -19.20*** -19.79*** 6.66* 4.82 3.99 4.34

2.05 2.50 2.97* 2.08 4.62 5.17 9.28** 1 4.1 0** .76 .67 .73 .46

Note. The results in the table are based on the 39 presidents from Washington through Reagan. * p < .05. * * p < .01. ***p < ,001.

trast with the previous 77% (cf. Table 1). Accomplishment and appointments were also quite well predicted, with 76% and 73% of the variance accounted for; those percentages, however, were slightly below that for the global assessment. On the other hand, the six predictors explained only 46% of the variation in the ratings of character and integrity. Even political skill had only 67% of its varia- tion accounted for. Nonetheless, these contrasts paralleled what was observed in the Methods section in the degree to which the five components represent the overall consensus. For example, the two variables with the lowest predictability were also the two variables with the lowest loadings on the first factor obtained by a principal component analysis of the five measures. Therefore, leadership, accomplishment, and appointments were more crucial to the assessment of the president’s performance than were the other two variables. The validity coeffi- cients reflect that difference.

For each of the five performance components, the following findings were noteworthy:

1. Leadership assessment was a positive function of the predictors years in office, war years, assassination, war hero, and intellectual brilliance. Only administration scandal was omitted as a predictor. Nonetheless, even scandal would have been statistically significant had we set the alpha level at .I0 (i.e., p = .080). Hence, the results were not that discrepant from what was found for the overall rating.

2. Accomplishment assessment was a positive function of the predictors years in office, war years, and war hero but a negative function of scandal. Although assassination did not satisfy the p < .05 criterion, it did come close

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 12: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

302 The Jouml of Social Psychology

(p = .051). Hence, the main discrepancy was the minimal impact of intellectual brilliance (p = .109).

3. Assessment of political skill was a positive function of the predictors years in ofice, war years, and assassination, albeit the intellectual brilliance vari- able could have been significant under a more liberal criterion (p = .095). Scan- dal and war-hero status, in contrast, were irrelevant (both ps > .17).

4. Appointments assessment was a positive function of the predictors years in ofice, war years, and intellectual brilliance but a negative function of scandal. Only assassination came close to inclusion (p = .093), whereas war-hero status was not relevant, even under a generous criterion (p = .209).

5. Assessment of character and integrity was a positive function of the pre- dictors years in office and a negative function of scandal, with no other predictor having the most remote value (all ps > .257). Again, the presidents’ moral caliber seemed to have much less to do with their performance as chief executive.

Only the predictor years in offce was consistent across all five components, a finding compatible with research on other forms of political leadership (Simon- ton, 1983, 1984; Sorokin, 1925, 1926). The other variables were more specifi- cally differentiated across the components. Thus, war years had no relevance for character. Furthermore, assassination was most useful for predicting leadership and political skill: scandal for predicting character, appointments, and accom- plishment; war-hero status for predicting leadership and accomplishment; and intellectual brilliance for predicting leadership and appointments.

Discussion

In the present study, I replicated the predictive utility of the six-variable equation, despite the addition of W. H. Harrison, Garfield, and Reagan to the sample and the use of an entirely new set of expert evaluations (Ridings & McIver, 1997). I also replicated the five-variable equation for the sample that included G. H. W. Bush. Because the multiple correlation was .88 for the first equation and .86 for the second, both were precise enough to use for the purpose of practical prediction. This conclusion is reinforced by the finding that there existed no substantial outliers-that is, the discrepancies between predicted and observed ratings were always within the expected range (e.g., as revealed by the standardized residuals). In addition, the residuals were homoscedastic in the sense that they displayed no systematic tendency to increase or decrease across the course of the American presidency. This outcome, which held for both the full and truncated models, indicates that the equations predicted the performance ratings of modern presidents with the same overall accuracy as they had predict- ed the ratings of the earliest presidents.

To illustrate how these equations might be used for prediction purposes, I examined 3 of the 4 most recent presidents.

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 13: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

1. Reagan remains a very controversial figure. Some people believe that he belongs on Mount Rushmore as one of the greatest presidents ever, whereas oth- ers maintain quite the contrary opinion. Reagan did not do very well in the Ridings-McIver (1997) poll, ranking only 26th of 41 presidents. In terms of the inverted score used here, that ranking amounts to an overall rating of 16. The pre- diction equation according to the full model tells much the same story. Beginning with the intercept of 5.48, Reagan gained 21.20 points for 8.0 years in office and for an intellectual brilliance score of 0.4, yielding a predicted rating of 26.68. Inverting this score (by subtracting it from 42) yielded a predicted rank of 15.3, which put him close to the upper third of the chief executives (between Kennedy and Cleveland). This rank was about the same level predicted in an earlier inves- tigation (Simonton, 1986~). But the subsequent Iran-Contra scandal evidently cost him 15.09 points, thus plunging him to a predicted rating of 11.59, or the rank equivalent of 30.41. That ranking is quite close to the Ridings- McIver overall ranking (and, in fact, the standardized residual was only 0.862). Whether Reagan should be held accountable for the Iran-Contra scandal is not the question here; the 7 19 survey respondents (Ridings & McIver) apparently held him accountable.

2. Because of the lack of a measure of intellectual brilliance for G. H. W. Bush, one might make a prediction based on the truncated model. Beginning with the constant of 4.28, one would then add 10.72 points for his 4.0 years in office and 2.76 points for his 1.0 war year, yielding a predicted rating of 17.76. Con- verting this figure into a predicted rank gives 23.24, which places G. H. W. Bush about where Clinton was ranked, in the bottom half of the presidents. This is very close to the 22nd position given in the survey (Ridings & McIver, 1997). Alter- natively, one might try to estimate Bush’s intellectual brilliance. Given his edu- cation and professional history, he must be at least one standard deviation above the average for U.S. chief executives. Now, by using the full model and includ- ing the additional 2.61 points, one would get a rating of 19.54, equivalent to a predicted ranking of 21.46. This prediction is even more accurate, only half of a rank off.

3. In Clinton’s case, a genuine prediction was possible, because he had not completed his term in office at the time of the present study. He was not a war hero and had not served as a wartime commander in chief. He had the opportu- nity to serve two full terms, and it seemed unlikely at the time that his adminis- tration would succumb to a major scandal. Under the truncated model, he would have a predicted rating of 25.72, or a reconstructed rank of 15.28-which is much better than the 23rd place that he received from the survey respondents (Ridings & McIver, 1997). The full model would give him some extra points, however. Clinton would probably receive at least as much credit for intellectual brilliance as would G. H. W. Bush; the extra credit would push him to a predict- ed rating of 28.25, or the equivalent of a 12.75 rank. This would place him in the same league as Lyndon B. Johnson and James Monroe. That the experts (Ridings

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 14: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

304 The J o u m l of Sociul Psychology

& McIver) ranked him 10 points lower may have reflected their suspicions about his character and integrity. On that component, Clinton was ranked 38th of 41 presidents (and that low rating occurred before the Monica Lewinsky episode).

Although the predictive utility of these equations should now be obvious, prediction is not tantamount to theoretical understanding. Among the various interpretations, two main possibilities deserve the most careful evaluation in future research (cf. Simonton, 1987b, 1994).

First, the predictor variables may reveal a great deal about the actual nature of presidential leadership. Certainly, one would expect great presidents to avoid scandal (Bailey, 1966). Moreover, chief executives who do not do so may actu- ally differ from other presidents in their personality profiles (Winter, 1987). The principal difficulty with accepting this premise as the sole interpretation is that not all predictors have an obvious connection with objective performance (e.g., assassination, which seems far more a situational than an individual attribute). There is simply no objective evidence that assassinated chief executives had per- formed better in the White House (for a review, see Simonton, 1987b). It is also curious that such variables as years in office are more important predictors than such variables as legislative success, diplomatic achievements, leadership style, popularity with contemporary voters, and the like. It seems that the ratings are influenced more by situational constraints than by the president’s actual leader- ship behaviors (Kenney & Rice, 1988; Nice, 1984).

Second, the predictors of presidential greatness may reflect largely how peo- ple tend to judge the politicians who occupy the nation’s highest office. The equations may reflect an attributional phenomenon, an interpretation put forth earlier (Simonton. 1987b). On the basis of research on attribution theory and leadership schema, it has been argued (Simonton, 1987b) that observers’judg- ments are shaped largely by salient events that seem to elicit a dispositional attri- bution of leadership. Saliency is more crucial than genuine relevance. To test this model, I conducted the experimental simulation mentioned earlier, the results of which confirmed the model’s predictions (Simonton, 1986a). According to this model, the findings reported in Table 2 would be interpreted as showing how the six predictors affected the hypothesized leadership schema via the five differen- tially weighted components.

The prediction equations may also reflect a reality that mixes the two possi- bilities described in the preceding two paragraphs. Global assessments of presi- dential leadership may be some combination of objective performance and sub- jective attribution. To gauge the relative significance of these two processes, more research is needed. For example, it is possible that some of the predictors serve as mere proxy variables for other variables that have more fundamental causal relations with presidential performance (Simonton, 198 1, 1986~). If so, the addition of the latter variables would force the supposed proxies to drop out of the equation. Thus, former war heroes may have certain dispositional charac-

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 15: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

Simonton 305

teristics that also contribute to their performance as chief executives. In addition, more research is needed regarding the possible predictors of these predictors (cf. Simonton, 1986c, 1987b). In searching for these indirect influences on assessed leadership, researchers must make all due allowance for the possibility of com- plex interactions between situational and individual variables (e.g., McCann, 1995; Simonton, 1987a; Winter, 1987). Only after construction of the complete structural model that supposedly underlies the six-variabIe equation will researchers be able to decipher its theoretical implications regarding the nature of political leadership. Such a comprehensive causal model may also enable researchers to make bona tide forecasts about the likely performance of presi- dential candidates (Simonton, 1993a, 199313).

REFERENCES Bailey, T. A. ( 1966). Presidential greatness. New York: Appleton-Century. Baltzell, E. D., & Schneiderman, H. G. (1988). Social class in the Oval Office. Society, 25,

Brody, R. A. (199 I ) . Assessing the president: The media, elite opinion, andpublic support.

Cell, C. P. (1974). Charismatic heads of state: The social context. Behavior Science

Cox, C. (1926). The early mental traits of three hundred geniuses. Stanford, CA: Stanford

Darlington, R. B. (1990). Regression and linear models. New York: McGraw-Hill. DeGregorio, W. A. (1993). The complete book of U.S. presidents (4th ed.). New York: Bar-

ricade Books. Deluga, R. J. (1997). Relationship among American presidential charismatic leadership,

narcissism, and related performance. Leadership Quarterly, 8, 5 1-65. Deluga, R. J. (1 998). American presidential proactivity, charismatic leadership, and rated

performance. Leadership Quarteriy, 9, 265-291. Holmes, J. E., & Elder, R. E. (1989). Our best and worst presidents: Some possible rea-

sons for perceived performance. Presidential Studies Quarterly, 19, 529-557. Kelly, T., & Lonnstrom, D. (1990). Presidents’ survey. Unpublished manuscript, Siena

College, Siena Research Institute at Loudonville, NY. Kenney, P. J., & Rice, T. W. (1988). The contextual determinants of presidential greatness.

Presidential Studies Quarterly, 18, 16 1-1 69. Kynerd, T. (197 1). An analysis of presidential greatness and “president rating.” Southern

Quarterly, 9, 309-329. Maranell, G. M. (1970). The evaluation of presidents: An extension of the Schlesinger

polls. Journal ofAmerican History. 57, 104-1 13. Maranell, G. M., & Dodder, R. (1970). Political orientation and the evaluation of presi-

dential prestige: A study of American historians. Social Science Quarterly, 51, 415421. McCann, S. J. H. (1992). Alternative formulas to predict the greatness of U.S. presidents:

Personological, situational, and zeitgeist factors. Journal of Personality and Social Psy- chology, 62, 469479.

McCann, S. J. H. (1995). Presidential candidate age and Schlesinger’s cycles of American history (1789-1992): When younger is better. Political Psychology, 16, 749-755.

Murray, R. K., & Blessing, T. H. (1983). The presidential performance study: A progress report. Journal of American History, 70, 535-555.

42-49.

Stanford, CA: Stanford University Press.

Research, 9, 255-305.

University Press.

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 16: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

306 The Jouml of Social Psychology

Murray, R. K., & Blessing, T. H. (1988). Greatness in the White House: Rating the pres- idents, Washington through Carter. University Park, PA: Pennsylvania State University Press.

Murray, R. K., & Blessing, T. H. (1994). Greatness in the White House: Rating the pres- idents, from George Washington through Ronald Reagan (2nd ed.). University Park, PA: Pennsylvania State University Press.

Nice, D. C. (1984). The influence of war and party system aging on the ranking of presi- dents. Western Political Quarterly, 37, 443455.

Ridings, W. J., Jr., & McIver, S. B. (1997). Rating the presidents: A ranking of U.S. lead- ers, from the great and honorable to the dishonest and incompetent. Secaucus, NJ: Citadel Press.

Rossiter, C. L. (1956). The American presidency. New York: Harcourt Brace. Schlesinger, A. M. (1948, November 1). Historians rate the U.S. presidents. Lfe, 25(18),

Schlesinger, A. M. (1962, July 29). Our presidents: A rating by 75 historians. New York Times Magazine, pp. 12-13,4041.43.

Simonton, D. K. (198 1). Presidential greatness and performance: Can we predict leader- ship in the White House? Journal of Personality, 49, 306-323.

Simonton, D. K. (1 983). Intergenerational transfer of individual differences in hereditary monarchs: Genes, role-modeling, cohort, or sociocultural effects? Journal of Personal- ity and Social Psychoiogy. 44, 354-364.

Simonton, D. K. (1984). Leaders as eponyms: Individual and situational determinants of monarchal eminence. Journal of Personality. 52, 1-2 1.

Simonton, D. K. (1985a). Intelligence and personal influence in groups: Four nonlinear models. Psychological Review, 92, 532-547.

Simonton, D. K. (1 985b). The vice-presidential succession effect: Individual or situation- al basis? Political Behavior; 7, 79-99.

Simonton, D. K. (1 986a). Dispositional attributions of (presidential) leadership: An exper- imental simulation of historiometric results. Journal of Experimental Social Psycholo- gy, 22, 389-418.

Simonton, D. K. (1986b). Presidential greatness: The historical consensus and its psycho- logical significance. Political Psychology, 7, 259-283.

Simonton, D. K. (1986~). Presidential personality: Biographical use of the Gough Adjec- tive Check List. Joumul of Personality and Social Psychology, 51, 149-160.

Simonton, D. K. (1987a). Presidential inflexibility and veto behavior: lbo individual- situational interactions. Journal of Personaliry, 55, 1-18.

Simonton, D. K. (1987b). Why presidents succeed: A political psychology of leadership. New Haven, C T Yale University Press.

Simonton, D. K. (1988). Residential style: Personality, biography, and performance. Journal of Personality and Social Psychology, 5.5, 928-936.

Simonton, D. K. (1990). Psychology, science, and history: An introduction to historiome- try. New Haven, C T Yale University Press.

Simonton, D. K. (1991a). Latent-variable models of posthumous reputation: A quest for Galton's G. Journal of Personality and Social Psychology, 60, 607-619.

Simonton, D. K. (1991b). Personality correlates of exceptional personal influence: A note on Thomdike's (1950) creators and leaders. Creativity Research Journal, 4, 67-78.

Simonton, D. K. (1991~). Predicting presidential greatness: An alternative to the Kenney and Rice Contextual Index. Presidential Studies Quarterly, 21, 301-305.

Simonton, D. K. (1992). Presidential greatness and personality: A response to McCann (1992). Journal of Personality and Social Psychology, 63, 676-679.

Simonton, D. K. (1993a). Further details on VOTER HELPERTM 1 .O: A response to the

65-66.68.73-74.

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014

Page 17: Predicting Presidential Performance in the United States: Equation Replication on Recent Survey Results

Simonton 307

editor’s comments. Political Psychology, 14, 555-558.

cy, and performance. Political Psychology, 14, 537-548. Simonton, D. K. (1993b). Putting the best leaders in the White House: Personality, poli-

Simonton, D. K. (1994). Greatness: Who makes history and why. New York: Guilford Press. Simonton. D. K. (1995). Personality and intellectual predictors of leadership. In D. H.

Saklofske & M. Zeidner (Eds.), International handbook of personality and intelligence (pp. 739-757). New York: Plenum.

Simonton, D. K. (1996). Presidents’ wives and first ladies: On achieving eminence with- in a traditional gender role. Sex Roles, 35, 309-336.

Sokolsky, E. (1 964). Our seven greatest presidents. New York: Exposition Press. Sorokin, P. A. (1925). Monarchs and rulers: A comparative statistical study (Pt. 1). Social

Forces, 4, 22-35. Sorokin, P. A. (1926). Monarchs and rulers: A comparative statistical study (Pt. 2). Social

Forces, 4, 523-533. Spangler, W. D., & House, R. J. (1991). Presidential effectiveness and the leadership

motive profile. Journal of Personality and Social Psychology, 60,439-455. Walberg, H. J., Rasher, S. P., & Hase, K. (1978). IQ correlates with high eminence. Gift-

ed Child Quarterly, 22, 196-200. Wendt, H. W., & Light, I? C. (1976). Measuring “greatness” in American presidents:

Model case for international research on political leadership? European Journal of Social Psychology, 6, 105-109.

Winter, D. G. (1987). Leader appeal, leader performance, and the motive profiles of lead- ers and followers: A study of American presidents and elections. Journal of Personali- ty and Social Psychology, 52, 196-202.

Received August 20, I998 Accepted September 8, 1999

Dow

nloa

ded

by [

New

cast

le U

nive

rsity

] at

00:

39 2

0 D

ecem

ber

2014