Last updated on April 17, 2014 at 1:21 EDT

The Contemporary Presidency: And We Will Know Their Greatness By the Trail of Controversy: Washington, Lincoln, and Franklin D. Roosevelt and Their Increasingly Contested Successors

September 2, 2008

By Grendstad, Gunnar

Presidential scholars hypothesize that presidential contestedness increases among recent presidents only because of unfinished research or within historical eras because of growing governing problems. Applying unused data from studies of scholarly assessments of presidential performance, this analysis disconfirms the recency hypothesis and confirms the era hypothesis. Presidential contestedness increases after each of the least contested presidents: George Washington, Abraham Lincoln, and Franklin D. Roosevelt. Uncontestedness is a hard act to follow. Presidential scholars argue that recent presidents are more contested than earlier presidents (Bose 2003; Lonnstrom and Kelly 2003; Murray and Blessing 1994). This is attributable to scholars’ limited information-research in progress, data embargos, and classified documents-which prevents contemporary presidents’ accomplishments from being revealed and settled. This position predicts that presidents since John F. Kennedy are more contested than earlier presidents. By contrast, a more encompassing view on presidential contestedness identifies national historical eras in which a few presidents are considered great because they contributed successfully to solve dire national problems and secured the survival of the government (Milkis and Nelson 2003). But because contemporary reconfigurations of the executive office cannot anticipate, or be fully geared toward, subsequent problems, later White House occupants become identified as sources of growing governing problems following the solution to the crisis (Skowronek 2002). This position predicts the existence of eras other than the most recent one in which scholars increasingly disagree on presidents’ accomplishments. The recency hypothesis can be true without the era hypothesis being true. But if the era hypothesis is true, it can subsume the recency hypothesis. The tests outlined here use data from presidential performance studies where measures of variance still have not been properly utilized.

Presidential Contestedness

Ever since ratings of presidential greatness began to be calculated in 1948 with the first study by Arthur M. Schlesinger, almost without exception scholarly ratershistorians, legal scholars, and political scientists-have considered George Washington, Abraham Lincoln, and Franklin D. Roosevelt “great” presidents and James Buchanan, Andrew Johnson, Ulysses S. Grant, Warren G. Harding, Herbert Hoover, and, later, Richard Nixon “failures.” Perhaps because of the repeatedly high correlations among more than a dozen expert studies rating presidential greatness (see Simonton 2006), no study has analyzed the disagreement over, or contestedness of, presidential greatness. Contestedness emerges when scholars rate presidents and disagree on an individual president’s degree of greatness or failure. The concept of presidential contestedness therefore refers to the reputational variance or disagreement that occurs when presidential scholars assess the presidents. This variance is located among the presidents, and not among the raters, and must be addressed accordingly.1 The absence of analysis of presidential contestedness is surprising, as so much of the debate in the presidential literature revolves around the achievements and debacles, triumphs and failures of the occupants of the White House.

As presidential scholars must consider a range of factors for each president, each scholar may assess some factors as more or less significant than other factors or even consider one factor to the exclusion of all others. The uneven performances of Harding, Lyndon B. Johnson, and Nixon, for example, are often addressed in the literature (e.g., Bose and Landis 2003; Greenstein 2004; Kessel 2001; Milkis and Nelson 2003; Schlesinger 1997). Whatever the choice of factors, the number of factors assessed, or the political bias of the scholars themselves, scholars are not likely to reach a complete consensus on any given president’s degree of greatness or failure for two additional reasons. First, institutional factors, or “ambivalence” (Mansfield 1993), ensures that the president is, at the same time, both the head of state and the head of the government’s executive branch. The duties of the head of state are expected to be far less partisan than the duties of the head of government. second, the temporal factor ensures that throughout his years in office, no president can be expected to perform equally well. Too many events and responses guarantee variation in accomplishments.

The recency hypothesis rests on three arguments. First, recent presidents are unable to escape the restricted temporal perspective of scholars. A longer time span permits scholars to place presidents within a broader and comparative frame, to have a proper historical context, and to allow assessments to “crystallize” (Schlesinger 1997, 183). second, the unavailability of sources and documents linked to a president’s term in office prevents a complete picture from emerging in the short run. Only when declassified documents and sources are made available and can be thoroughly analyzed can researchers assess the overall standing of individual presidents (see, however, Beschloss 2002). Third, scholars can be influenced by their own experience of recent presidents. Today’s media permits scholars to be present at presidential events and keeps those memories fresh with constant replay, such that events can “still spark strong reactions among reviewers” (Bose 2003, 4).2

Scholars seem to agree that the recency hypothesis covers presidents over the last 40 to 50 years (Bose 2003; Lonnstrom and Kelly 2003). Murray and Blessing (1994) opted for 25 to 30 years in a slightly earlier publication.3 When taking into consideration the reduction in contestedness of the Dwight D. Eisenhower presidency wrought by Greenstein’s (1994) study, scholars tend to propose the start of the Kennedy presidency as the cutoff point.

If such a cutoff point exists, and if recent presidents are found to be more contested than earlier presidents, one of three outcomes can be expected. First, on a general relationship of no correlation between presidents’ time in office and degree of contestedness, recent presidents represent a significant mesa only. This outcome will be the strongest evidence to support the recency hypothesis (see “mesa” line in Figure 1). Second, there is a positive linear relationship between time and contestedness in which differences between low and high contestedness would be observed continuously for any comparison between a group of earlier presidents with a group of later presidents. If this outcome is supported empirically, presidents since Kennedy are not contested simply because they are recent presidents but because they are at the tail end of the line of White House occupants. By implication, William Howard Taft and his immediate predecessors, for example, would be more contested than their predecessors if the assessment was done in 1913. This would confirm the form, but not the substance, of the recency hypothesis (see the “linear” line in Figure 1). Third, contestedness is a function of time and amplified by recency. If this outcome is supported empirically, one will observe a curvilinear relationship between presidents’ time of service and degree of contestedness. This outcome would also offer evidence for the recency hypothesis (see the “curvilinear” line in Figure 1).

The recency hypothesis-with a cutoff point at 1961-can be complemented, if not eclipsed, by other attempts at historical-cum- presidential periodization that offer “[t]hinking in terms of regime sequences rather than linear national development” (Skowronek 2006, 90). Although a number of smaller, larger, and cross-cutting periods exist (see, e.g., Gerring 1998; Mayhew 2002; Murray and Blessing 1994; Skowronek 2002), two cutoff points yielding three periods will be explored here. These three periods are generally recognized by scholars and therefore serve both as counterpoints to the recency hypothesis as well as an initial test of one likely version of the era hypothesis.

A number of scholars identify the Franklin Roosevelt presidency during the New Deal and World War II as the birth of the modern presidency (e.g., Lowi 1985; Milkis and Nelson 2003; Neustadt 1990; Pfiffner 2005). Shafer concludes that the New Deal is “the one place where the realignment synthesis got it right” (2005, 545). Roosevelt’s term in office brought about a “vast expansion of the scope and influence of the federal government,” offered a “benchmark” for later presidents, and turned the presidency into “the most visible landmark in the political landscape” (Greenstein 2004, 3, 22). The modern presidency also brought out the president as an individual to the extent that presidential polling shifted from the public’s policy attitudes to a “president’s personal image and appeal” (Jacobs and Burns 2004, 538).4

Scholars also propose Lincoln as a demarcation point in the development of the presidency, as he ensured the success of the “second American revolution” (McPherson 1992). Milkis and Nelson argue that “Lincoln’s presidency not only marked a critical moment in political history but also raised anew, under conditions of unprecedented urgency, questions about the appropriate place of executive power in the American system of government” (2003, 148). Technically, Washington’s presidency, being the first in a long sequence, does not constitute a cutoff point in the line of presidents. But he was a successful general during the War of Independence and, as the first president, he is precedent setting and ascribed the role of father of the country. Any first era involving presidents cannot move beyond Washington.5

Washington, Lincoln, and Franklin Roosevelt were successful presidents “when the survival of the government hung in the balance” (Blessing 2003, 47). The demarcations and changes that their presidencies represented can be interpreted first as “solution[s] to the problems of governing in America.” Given the reconfigurations of contemporary politics, presidents who succeeded them can be interpreted as representing the “source[s] of America’s governing problems” (Skowronek 2002, 743).6 If this configurationcum- reconfiguration cycle is correct, each of the Washington, Lincoln, and Roosevelt presidencies should not only inaugurate new periods of presidential contestedness-succeeding presidents should also be increasingly contested. Increasing presidential contestedness within these eras is represented by three separate positive relationships (see the three parallel lines in Figure 1).

In addition to the identification of time as a correlate of presidential contestedness, the literature offers four additional controls.7 First, the number of years a president serves can affect the degree of contestedness in two ways. On the one hand, “brevity in office” (Schlesinger 1997, 183) means fewer initiatives, less action, and less information through which unreliable assessments and contestedness are likely to increase. On the other hand, a longer period in office-more action, initiatives, and information- can bring about “discordance” and “disjunction” (183), on which contestedness will thrive. second, brevity may stem from the inability to complete a term in a normal manner. Absent Nixon, aborted presidencies stem from assassinations or death following illness, leading either to martyrdom and reduced contestedness or to incomplete information and increased contestedness. Third, assessments of presidents may be influenced by their post-White House years (Lonnstrom and Kelly 2003) because presidents may, deliberately or not, improve their image by continuing their political work or seeking the role of elder statesman (Updegrove 2006). Fourth, attribution may work backwards, so that outstanding and often moral achievements that are recognized but not easily measured may unduly influence the assessment of other actions or characteristics (Rosenthal and Rosnow 1991). Thus, Blessing (2003) suggests a “halo category” to control for any overstated uncontestedness of Washington, Lincoln, and Franklin Roosevelt, as their achievements in office and their status as personification of an era may repress the human and institutional mediocrities or errors for which they still would be held responsible.

Data, Variables, and Procedures

This is a case study of the contestedness of presidential greatness. A case study emphasizes the boundedness of the phenomenon within which analysts can define units so as to permit an increase of within-case observations (see, e.g., Gerring 2004; Stimson 1985). The study is spatially bounded, within the United States, and temporally bounded, within the years 1789 to 2000.

Data pertinent to the testing of the recency and era hypotheses are found in the 1996 study by Schlesinger (1997) and in the 2000 Federalist Society/Wall Street Journal (FS/WSJ) study (Taranto and Leo 2004). Each study included 39 presidential observations, and the studies were carried out in a similar manner. In each study, a number of scholars were asked to assess the men who had served as presidents. The 1996 Schlesinger study primarily drew on historians, whereas the FS/WSJ study drew on historians, law professors, and political scientists. The two studies’ presidential greatness ratings measure the same construct.8 Because the two studies’ presidential observations are repeated assessments of presidents, the observations were added, giving 78 units of observation. The added file allowed for the control for study effects.

Presidential contestedness is measured as the variance of a president’s greatness score. The variance stems from disagreements among scholars as to the degree of a president’s greatness or failure. The variance is measured as the coefficient of variation (Cv), which is the standard deviation divided by the mean, multiplied by 100 (see Appendix).9

Six independent variables were used in the study: (1) The president’s election year was receded into a full sequence of dummy variables to test the cutoff point for the recency hypothesis. If a cutoff point at approximately 1960 could not be confirmed, election year was used (or the year in which a president assumed office, if assuming office was not followed by winning an election; for example, Rutherford B. Hayes’s start year was set at 1877). Figures display years, whereas statistical analyses use a variable in which the year 1788 is set to zero. (2) Years in office was the successor’s year of assuming office subtracted by the election year. (3) A died in office dummy variable equaled unity for the six presidents who did not serve out their elected term. (4) Sway was year of death subtracted by successors’ election year (for presidents who were alive during the two studies, year of death was replaced by study years; the two Bill Clinton observations received a score of zero). (5) The halo effect was coded 1 for the Washington, Lincoln, and Franklin Roosevelt observations. (6) Type of data or a study’s choice of scholars may lead to biased assessments. Felzenberg (1997) and Piereson (1997) expected a liberal bias to influence the outcome in the 1996 Schlesinger study. A dummy variable was therefore included and coded 1 for the 2000 FS/ WSJ study.

The recency hypothesis can be true without the era hypothesis being true. If the era hypothesis is true, it can subsume the recency hypothesis. The first analysis therefore began with tests of the recency hypothesis that presidential contestedness increases significantly around 1960. To cover all outcomes, a full sequence of t-tests was carried out using a moving cutoff point that exhaustively compared all groups of earlier presidents with all groups of adjacent and later presidents. A t score exceeding the value of 2.0 meant that an earlier group of observations was more contested than a later group of observations. A t score below the value of-2.0 meant that a later group of observations was more contested than an earlier group of observations. If confirmed, the second analysis included in multiple regression models the dummy variable for the recency hypothesis. If not confirmed, election year was used. The third analysis tested presidential contestedness within historical eras. All multiple regression analyses included tests of multicollinearity, outliers, influential observations, and autocorrelation, as presidential scholars note that presidents may do better if they heed their predecessors (Burns 2003; Greenstein 2004; Pfiffner 2003).10

Trails of Contestedness

The solid line in Figure 2 connects the 38 t scores for mean differences between any group of earlier presidents with the adjacent group of subsequent presidents. The results do not lend support to the recency hypothesis. All groups of presidents following the entire sequence of presidents up to and including Franklin Pierce have been more contested than those preceding them. It is also the case that the groups of recent presidents that include Woodrow Wilson and Harding are more contested than previous groups of presidents. And it is also the case that the groups of recent presidents that include the sequence of Kennedy through Nixon are more contested than previous groups of presidents. Hence, from this test, the recency hypothesis received no empirical support as to convergent and discriminant validity, as a recent mesa of contestedness could not be identified.11 Figure 2 does not exclude the possibility of a linear trend of contestedness, but the elongated fluctuations suggest that periods of presidential contestedness may exist.

The linear increase version of the recency hypothesis was tested by way of multiple regression analyses using a base model, with election year and the study control as independent variables, and an expanded model, with the four additional controls (see Table 1, Recency Models A and B). In both models, election year was significant and positive: Contestedness increased linearly across all 39 presidents. There were no study effects in any model. In the expanded model, only election year and the halo effect were significant: Presidential contestedness increased by 1 percentage point every twenty-fifth year, and the three halo presidents (Washington, Lincoln, and Franklin Roosevelt) were, on average, 15 percentage points less contested than nonhalo presidents. Both models were statistically significant, had acceptable levels of collinearity and autocorrelation, and had adjusted R^sup 2^ of 6% and 25%, respectively. The result of the linear relationship test supports the form, but not the substance, of the recency hypothesis. Curvilinear models, the third version of the recency hypothesis, were also tested, but none contributed significantly beyond the two linear models in Table I.12

In the expanded linear recency model, Schlesinger’s Nixon observation was the only outlier, underpredicted by 31.5 percentage points. However, the most influential observations included Buchanan, Andrew Johnson, Harding, Franklin Roosevelt, and Nixon. Thus, absent the Nixon observations, model instability occurred around 1865-77 and 1921-33. These two dozen years correspond with the two center indentations in Figure 2. Common to these years is also the fact that historians have linked them to significant historic shirts and presidential scholars have linked them to significant reorganizations of executive powers. These years also converge on the cutoff points for the era hypothesis discussed earlier. The era hypothesis of presidential contestedness, therefore, was analyzed using regression analyses within three periods: 1789-1859, 1860-1931, and 1932-99. Only two of the base models were statistically significant and met the model criteria. These results also indicate the existence of an interaction effect between contestedness and election year within different periods. The first and third periods show distinct and increasing presidential contestedness. The second period shows a nonsignificant positive relationship (see Table I).13 The regression lines of the analyses of the three periods are superimposed on the scatter plot in Figure 3. Only in the first period was there a study effect, and the two regression lines show the additive effect: Scholars in the 2000 FS/WSJ study disagreed more on early presidents than did the historians in the 1996 Schlesinger study.

The first presidential period starts with the uncontested Washington and ends with the contested Buchanan presiding over the beginning of the Civil War. The third presidential period starts with the uncontested Franklin Roosevelt and ends with the contested Clinton. The outlier nature and influence that the Nixon observations exerted in the third period called for the two Nixon observations to be removed. This removal increased adjusted explained variance from 21% to 58%. Figure 3 clearly indicates Nixon, the only president to be driven out of the White House, as one of the most contested presidents. The “modern presidency,” starting with Franklin Roosevelt and excluding the Nixon observations, displays a distinct linear trend of increasing contestedness.

The analysis revealed that the second period, from Lincoln to Hoover, shows no significant linear relationship between election year and contestedness. The basic message is that the complete sequence of presidents in this period does not comply with the era hypothesis. Either the era hypothesis has limited validity applied to this period, or the era hypothesis may be applied to subperiods permitting smaller periods to be observed. But taking cues from a complementing approach that is “characteristically more aggressive in its manipulation of patterns and more radical in its departure from a chronological view of history” (Orren and Skowronek 2004, 80, the 1860-1931 period may be reorganized with Lincoln, Andrew Johnson, and Grant being keys to alternative interpretations. Compared with Washington and Franklin Roosevelt’s lines of succession, Lincoln’s abbreviated presidency failed to fully inaugurate a new era. This is because the degree of presidential contestedness that scholars have revealed in Lincoln’s two successors, Johnson and Grant, suggests that they were the runts of the Washington through Buchanan era (1789-1859). If this is the case, one may not only be able to observe a strong linear relationship that runs from Washington and bypasses Lincoln to include Johnson and Grant. One may also establish a linear relationship that starts with Lincoln, bypasses Johnson and Grant, and connects the rest of the presidents through Hoover.14

In sum, the analyses showed that the era hypothesis fared better than the recency hypothesis. As to the recency hypothesis, no consistent post-Eisenhower mesa of contestedness could be identified, and no curvilinear models performed better than the linear models. As to the era hypothesis, it subsumes the recency hypothesis and offers partly improved models of presidential contestedness.

Uncontestedness Is a Hard Act to Follow

The recency hypothesis of presidential contestedness claimed Eisenhower as the juncture after which presidential contestedness increased. This article concludes that the contestedness of recent presidents is not novel. Beginning with each of the three almost uncontested presidents-Washington, Lincoln, and Franklin Roosevelt- there exist at least two complete periods in which presidents are increasingly contested. The recency hypothesis was correct in identifying a trend. But the correlates of the recent contestedness- data embargos, classified documents, and research in progress-may be spurious, as these correlates can not explain the pattern of presidential contestedness in the previous periods. Alternatively, one can argue that each uncontested president contributed to forge a presidency that was an answer to a profound problem. In successfully solving these problems, these presidents offer benchmarks against which their successors are gauged. Subsequent presidents, therefore, come to occupy an office that turns out not to be geared toward new problems. Over time, therefore, the presidents and their presidencies become the “source of America’s governing problems” (Skowronek 2002, 743). Leuchtenburg (2001) was correct in stating that a new presidential period started with Franklin Roosevelt and that his formidable presidency cast such a “long shadow” that no recent president so far has been able to escape it. This analysis has shown that the uncontested greatness of Washington and Franklin Roosevelt, and, in part, Lincoln, has cast a series of long shadows over their successors. Uncontestedness is a hard act to follow.

1. Presidential raters’ demographic and scholastic properties explain between 0% and 5% of the variance in 33 of 35 presidents’ greatness, with Andrew Johnson and Graver Cleveland exceeding this range (Murray and Blessing 1994, 128).

2. “Thirteen presidents have occupied the White House in my lifetime, nearly one-third of all who have served,” Charles O. Jones, a presidential rater, reminisces. “I have strong, personal impressions of twelve of the thirteen (I still have my FDR button from the 1940 campaign)” (2002, 712).

3. The first edition of Murray and Blessing’s book was published in 1988.

4. The starting point of the “modern presidency” in 1932 is not uncontested (see, e.g., Gould 2003; Skowronek 1993; Tulis 1988).

5. Standard prediction models of presidential greatness significantly overestimate Washington (Simonton 1981, 1986).

6. Note that Skowronek’s (2002) chronology also backtracks to include contested presidents preceding great presidents.

7. Analysts consider time a proxy to be replaced by theoretical correlates.

8. Each study asked the scholars to rate 39 presidents on their merits. William Henry Harrison and James A. Garfield were omitted because of their brief and aborted terms. In the 1996 Schlesinger study, 32 scholars rated the presidents on a five-point scale: “great,”"near great,”"average,”"below average,” and “failure” (1997, 188f). The scholars were left to decide for themselves how presidential performance was to be judged, as it “was assumed that historians would recognize greatness-or failure-when they saw it, as Justice Potter Stewart once proposed to recognize pornography” (1997, 179). Means and standard deviations were recalculated for each president using 1 for “failure” and 5 for “great.” The 2000 FS/ WSJ study surveyed 78 scholars who were asked to “take into consideration the value of accomplishments of his presidency and the leadership he provided the nation, along with any other criteria you deem appropriate” (Lindgren 2004, 250 n. 3; emphasis added). The scale used “highly superior” (5), “above average,”"average,”"below average,” and “well below average” (1). Only these two studies were found to report, or permitted to be calculated, the means and standard deviations of presidential greatness. The correlation between the two studies’ greatness ratings was .95. The Murray- Blessing study (1994), which offered the recency hypothesis, was not included because of inconsistencies in its data.

9. The unit of measurement of presidential contestedness as the coefficient of variation is percent, and when measuring change, percentage points. The coefficient of variation is a dimensionless measure and very useful for comparing variances. Note that Cv is sensitive to errors in the mean and to the mean approaching zero (Kendall and Stuart 1958; Pearson 1897; Stuart and Ord 1994). Its relation to the mean also becomes inverse when measures are compared within the same class of things and/or when the variance is disproportionately large for smaller means (Lande 1977), both of which are the case here. The coefficient of variation is more valid than measures that simply compare presidents’ contestedness by observing a president’s change across the ordinal five-point (or the occasionally inferred six-point) failure-greatness scale in some studies. Initial approaches to presidential “controversiality,” using the standard deviation as absolute measures of variance, can be found in Murray and Blessing (1994) and Lindgren (2004).

10. All significance tests were two-tailed (where appropriate) at the .05 level. T-test scores were rounded to |2.0|, N = 78. The condition index should not exceed the value of 15 (Belsley 1991). Tolerance should not fall below .50. Outliers were identified when they exceeded |2.0|. Influential observations were identified through size-adjusted sdfbeta (2/[the square root of]N ) and sdffit (2[the square root of](p/N)) (Belsley, Kuh, and Welsch 1980). The Durbin-Watson test of autocorrelation took critical values from Savin and White (1977). The noninclusion of Harrison and Garfield limited more advance tests of autocorrelation.

11. Figure 2 also shows that it has never been the case that earlier groups of presidents have been more contested than later presidents.

12. The highest (absolute) values of correlation between the independent variables were between died in office and halo (r=.41) and between died in office and sway (r = -53)- The condition indexes did not exceed the value of 10 in either of the two models. The tolerance level did not fall below the value of .60 for any of the independent variables. None of the Durbin-Watson values fell in the “autocorrelation” regions of outcome. 13. The expanded models for the three periods failed to meet the acceptable criteria for condition index and tolerance. Moreover, all variables added in the expanded models were statistically insignificant, except for the halo effect in the 1860-1931 period. Curvilinearity was tested on all era models. Results are not reported but are referred to if significant improvements from linear models were detected.

14. If the second period is limited to the sequence of presidents from Hayes to Hoover only, a quadratic regression model is confirmed (p


Belsley, David A. 1991. Conditioning Diagnostics: Collinearity and Weak Data in Regression. New York: Wiley.

Belsley, David A., Edwin Kuh, and Roy E. Welsch. 1980. Regression Diagnostics: Identifying Influential Data and Sources of Collinearity. New York: Wiley.

Beschloss, Michael. 2002. “Knowing What Really Happened.” Presidential Studies Quarterly 32: 642-46.

Blessing, Tim H. 2003. “Presidents and Significance: Partisanship as a Source of Perceived Greatness.” White House Studies 3: 41-52.

Bose, Meena. 2003. “Presidential Ratings: Lessons and Liabilities.” In The Uses and Abuses of Presidential Ratings, edited by Meena Bose and Mark Landis. Hauppauge, NY: Nova, 3-23.

Bose, Meena, and Mark Landis, eds. 2003. The Uses and Abuses of Presidential Ratings. Hauppauge, NY: Nova.

Burns, James MacGregor. 2003. “Commentary.” In The Uses and Abuses of Presidential Ratings, edited by Meena Bose and Mark Landis. Hauppauge, NY: Nova, 93-103.

Felzenberg, Alvin S. 1997. “”There You Go Again”: Liberal Historians and the New York Times Deny Ronald Reagan His Due.” Policy Review 82: 51-53.

Gerring, John. 1998. Party Ideologies in America, 1828-1996. New York: Cambridge University Press.

_____. 2004. “What Is a Case Study and What Is It Good For?” American Political Science Review 98: 341-54.

Gould, Lewis L. 2003. The Modern American Presidency. Lawrence: University Press of Kansas.

Greenstein, Fred I. 1994. The Hidden-Hand Presidency: Eisenhower as Leader. Reprint ed. Baltimore: Johns Hopkins University Press.

_____. 2004. The Presidential Difference: Leadership Style from FDR to George W. Bush. Princeton, NJ: Princeton University Press.

Jacobs, Lawrence R., and Melanie Burns. 2004. “The Second Face of the Public Presidency: Presidential Polling and the Shift from Policy to Personality Polling.” Presidential Studies Quarterly 34: 536-56.

Jones, Charles O. 2002. “Knowing What We Want to Know about the Presidency.” Presidential Studies Quarterly 32: 710-19.

Kendall, Maurice G., and Alan Stuart. 1958. The Advanced Theory of Statistics. London: Griffin.

Kessel, John H. 2001. Presidents, the Presidency, and the Political Environment. Washington, DC: CQ Press.

Lande, Russell. 1977. “On Comparing Coefficients of Variation.” Systematic Zoology 26: 214-17.

Leuchtenburg, William E. 2001. In the Shadow of FDR: From Harry Truman to George W. Bush. Ithaca, NY: Cornell University Press.

Lindgren, James. 2004. “Methodology of Rankings.” In Presidential Leadership: Rating the Best and the Worst in the White House, edited by James Taranto and Leonard Leo. New York: Free Press, 249-66.

Lonnstrom, Douglas A., and Thomas O. Kelly. 2003. “Rating the Presidents: A Tracking Study.” Presidential Studies Quarterly 33: 625-34.

Lowi, Theodore J. 1985. The Personal President. Ithaca, NY: Cornell University Press.

Mansfield, Harvey C., Jr. 1993. Taming the Prince: The Ambivalence of Modern Executive. Reprint ed. Baltimore: Johns Hopkins University Press.

Mayhew, David R. 2002. Electoral Realignments: A Critique of an American Genre. New Haven, CT: Yale University Press.

McPherson, James M. 1992. Abraham Lincoln and the Second American Revolution. New York: Oxford University Press.

Milkis, Sidney M., and Michael Nelson. 2003. The American Presidency: Origins and Development, 1776-2002. 4th ed. Washington, DC: CQ Press.

Murray, Robert K., and Tim H. Blessing. 1994. Greatness in the White House: Rating the Presidents, from George Washington through Ronald Reagan. 2nd ed. University Park: Pennsylvania State University Press.

Neustadt, Richard E. 1990. Presidential Power and the Modern Presidents. The Politics of Leadership from Roosevelt to Reagan. 5th ed. New York: Free Press.

Orren, Karen, and Stephen Skowronek. 2004. The Search for American Political Development. New York: Cambridge University Press.

Pearson, Karl. 1897. “On the Scientific Measure of Variability.” Natural Science 11(66): 115-18.

Pfiffner, James P. 2003. “Ranking the Presidents: Continuity and Volatility.” In The Uses and Abuses of Presidential Ratings, edited by Meena Bose and Mark Landis. Hauppauge, NY: Nova.

_____. 2005. The Modern Presidency. 4th ed. Belmont, CA: Wadsworth.

Piereson, James. 1997. “Historians and the Reagan Legacy.” Weekly Standard, September 29, 22-4.

Rosenthal, Robert, and Ralph L. Rosnow. 1991. Essentials of Behavioral Research: Methods and Data Analysis. New York: McGraw- Hill.

Savin, N. E., and Kenneth J. White. 1977. “The Durbin-Watson Test for Serial Correlation with Extreme Sample Sizes or Many Regressors.” Econometrica 45: 1989-96.

Schlesinger, Arthur M. 1948. “Historians Rate U.S. Presidents.” Life, November 1, 65-66, 68, 73-74.

Schlesinger, Arthur M., Jr. 1997. “Rating the Presidents: Washington to Clinton.” Political Science Quarterly 12: 179-90.

Shafer, Byron E. 2005. “Orders and Eras in American Politics.” Polity 37: 541-47.

Simonton, Dean Keith. 1981. “Presidential Greatness and Performance: Can We Predict Leadership in the White House?” Journal of Personality 49: 306-22.

_____. 1986. “Presidential Greatness: The Historical Consensus and Its Psychological Significance.” Political Psychology 7: 259- 83.

_____. 2006. “Presidential IQ, Openness, Intellectual Brilliance, and Leadership: Estimates and Correlations for 42 U.S. Chief Executives.” Political Psychology 27: 511-26.

Skowronek, Stephen. 1993. The Politics Presidents Make: Leadership from John Adams to George Bush. Cambridge, MA: Belknap Press.

_____. 2002. “Presidency and American Political Development: A Third Look.” Presidential Studies Quarterly 32: 743-52.

_____. 2006. “Presidential Leadership in Political Time.” In The Presidency and the Political System, edited by M. Nelson. Washington, DC: CQ Press.

Stimson, James A. 1985. “Regression in Space and Time: A Statistical Essay.” American Journal of Political Science 29: 914- 47.

Stuart, Alan, and J. Keith Ord. 1994. Kendall’s Advanced Theory of Statistics. 6th ed. Vol. 1. London: Charles Griffin.

Taranto, James, and Leonard Leo, eds. 2004. Presidential Leadership: Rating the Best and the Worst in the White House. New York: Free Press.

Tulis, Jeffrey K. 1988. The Rhetorical Presidency. Princeton, NJ: Princeton University Press.

Updegrove, Mark K. 2006. Second Acts: Presidential Lives and Legacies after the White House. Guilford, CT: Lyons Press.


University of Bergen

Gunnar Grendstad is Associate Professor of Comparative Politics at the University of Bergen, Norway.

AUTHOR ‘S NOTE: Earlier versions of this paper received constructive comments at seminars at the Department of Comparative Politics, University of Bergen, and at the 2007 American Politics Group Annual Conference, University of Leicester. Thanks are due to Mike Alvarez, Tim H. Blessing, RichardJ. Ellis, Sandra Halverson, Tor Midtbe, Bill Staffer, Kristin Stremsnes, Lars Svasand, Eirik Vestrheim, and two anonymous reviewers for useful comments.

Copyright Center for the Study of the Presidency Sep 2008

(c) 2008 Presidential Studies Quarterly. Provided by ProQuest LLC. All rights Reserved.