Since Arthur Schlesinger’s 1948 survey, “Historians Rate U.S. Presidents,” published in Life, scholars—mostly historians—have been rating the presidents.1 In this first survey, the instant assessment of scholars was that Franklin D. Roosevelt was worthy of joining the pantheon of Great Presidents, following only Abraham Lincoln and George Washington in greatness. Contrariwise, in a survey immediately following Ronald Reagan’s presidency, the instant assessment was that Reagan was mediocre.2 As to why Reagan was reelected in a landslide and succeeded by a candidate of his party, well, as one survey put it, he ranked very high in luck.3
Of course, every scholarly ranking involves the implicit values of the rankers. To what extent, for example, do scholars value economic performance when rating presidents? The same kinds of questions can be asked of presidential elections and surveys of the general population: To what extent do voters value economic performance when considering a sitting president for reelection, or when considering the nominee of the president’s party when he is not running for reelection? These questions lead to a larger one: Should the judgments of scholars matter to voters, if the scholars’ values differ from those of voters? Or, to put this more pointedly, are scholarly rankings of presidents worthy contributions that the public should take seriously or are they a self-serving exercise by a relatively small, isolated, and privileged community?
From the mid- to late twentieth century, the alignment of the most highly ranked presidents with electoral success was mixed. As shown in table 1, in the 1996 survey conducted by Arthur Schlesinger Jr., the six highest-ranked presidents—Washington, Lincoln, Thomas Jefferson, Andrew Jackson, and both Theodore and Franklin D. Roosevelt—were all elected to second terms and succeeded by the candidates of their party.4 Nevertheless, three other presidents meeting this definition of electoral success—James Madison, William McKinley, and Reagan—were rated as average, and two more—Ulysses S. Grant and Calvin Coolidge—ranked as below average.
Rank and Electoral Success of U.S. Presidents: Schlesinger and Schlesinger Jr. Surveys
Elected to a second term and succeeded by the candidate of their party in the next following election
Rank in Schlesinger 1948
Rank in Schlesinger 1962
Rank in Schlesinger Jr. 1996
Source: Arthur M. Schlesinger, “Historians Rate U.S. Presidents,” Life, November 1, 1948; Arthur M. Schlesinger, “Our Presidents: A Rating by Seventy-Five Historians,” New York Times Magazine, July 29, 1962; Arthur M. Schlesinger Jr., “Rating the Presidents: Washington to Clinton,” Political Science Quarterly (Summer 1997).
It might be supposed that the cleavage between contemporary endorsement by the general public and scholarly ranking reflects the difference between expert judgment and popular opinion, and, with respect to early U.S. presidents, the effect on assessments of the prism of time. Consider the case of Madison. In real time, Federalists opposed the War of 1812, but the war’s conclusion in brilliant victories at Baltimore, Lake Champlain, and especially New Orleans, as well as the restoration of the status quo antebellum in the Treaty of Ghent, doomed the Federalists to extinction. Years later, there seems to be a consensus that that war should have been avoided.
The cleavage between expert judgment and popular opinion may also reflect differences in interests. Scholars, being drawn from the intellectual elite, may be disposed to those who are like them in terms of intellectual giftedness. Scholars may also be drawn to a progressive political agenda. And presidential scholars may, in particular, be disposed to executive power as distinct from legislative or judicial power, and to the exercise of government power at the national as opposed to the state level.
In Schlesinger’s 1948 survey of a small, carefully chosen panel of historians, panelists were asked simply to describe presidents as “Great,” “Near Great,” “Average,” “Below Average,” or “Failure.” Schlesinger specifically asked participants to consider performance as president only.5 The responses were then coded as 5, 4, 3, 2, 1, the numerical scores averaged, and the averages ranked, giving a complete ranking of the presidents from Washington to Franklin D. Roosevelt, except for William Henry Harrison and James Garfield, who died shortly after taking office. Schlesinger’s second survey (published by the New York Times in 1962) followed the same pattern,6 as did his son’s 1996 survey, except that Schlesinger Jr. coded “Failure” as −2 instead of as 1. The only consequence of this change was that it diminished Reagan’s ranking.7
Some subsequent surveys have followed the simple method developed by Schlesinger, while others have gotten more complicated. Two types can be distinguished: The first, as with Schlesinger’s 1948 survey, leaves the overall assessment to the scholar. The second queries scholars about certain aspects of presidents and forms an overall assessment by summing or averaging the assessments of these aspects. This type of survey imposes a weighting scheme onto detailed assessments that may differ from the weighting scheme the scholars themselves would choose.
To demonstrate that these two types of surveys are not the same, a 1982 Chicago Tribune survey asked scholars to rate presidents in five categories—leadership, accomplishments, political skill, appointments, and character—in addition to asking for an overall rating.8 Franklin D. Roosevelt ranked second according to the average of the scores in the five categories, but he ranked third in the overall assessment. On the other end of the spectrum, Richard Nixon ranked fourth from the bottom according to the average of the scores in the five categories, but he ranked second worst in the overall assessment.
More dramatically, in the 2002 Siena Research Institute survey, the published ranking based on twenty underlying categories placed Nixon as somewhat less than average.9 But an examination of the underlying categories reveals that he ranked near the bottom in the “your present overall view” category. In the underlying categories, Nixon’s scores range from relatively high in “foreign policy accomplishments” to the lowest in “integrity.” In forming an overall assessment, scholars clearly put more weight on “integrity” than the formula in the Siena survey allowed.
The first, scholar-centric, type of survey includes the three conducted by Schlesinger and by his son, the Chicago Tribune survey, and eight others: Organization of American Historians (OAH; Maranell, 1970), Murray-Blessing (1983), Ridings-McIver (1996), Intercollegiate Studies Institute (ISI; Gregg, 1996), the Wall Street Journal 2000 and 2005 surveys (Lindgren and Calabresi, 2001), and the category described as “your present overall view” in the 2002 and 2010 Siena Research Institute surveys.10
The ISI survey was targeted specifically to right-of-center scholars, and both Wall Street Journal surveys queried panels in which left- and right-of-center scholars were equally represented. The OAH survey asked three questions on prestige, strength of action, and accomplishments that could be interpreted as reflecting an overall assessment—the responses to which were very highly correlated.11 The Ridings-McIver survey used a hybrid method to obtain overall rankings, including a five-part assessment scale, a listing of the top and bottom ten presidents, and ratings based on categories with scholar-provided weights. OAH, Murray-Blessing, Ridings-McIver, and Siena polled large samples of scholars, while the other surveys queried relatively small, carefully chosen panels.
The second, survey-centric, type of survey includes the first few conducted by the Siena Research Institute (Lonnstrom and Kelly, 1997, 2003), and two conducted by C-Span.12 In the C-Span surveys, scholars assessed the presidents in ten categories: public persuasion, crisis leadership, economic management, moral authority, international relations, administrative skills, relations with Congress, setting the agenda, equal justice, and performance within the times. Assessments in these categories are summed, and the sum is used to rank presidents. This process is kind to Nixon, whose very low score in moral authority is partially offset by his high score in international relations.
Determinants of Greatness
Survey-takers appear to agree on the criteria. After his first survey, Schlesinger said that great presidents are associated with turning points in history, and are “idealistic but not doctrinaire,” strong moral leaders, and expanders of executive power.13 According to Murray and Blessing, all of the great and near-great presidents are action-oriented and “progressive,” meaning willing and able “to promote fundamental change.”14 Henry Steele Commager, who conducted a survey published in Parade in 1977, said great presidents are “on the side of the people” and push for “progress” through “reform.”15
This study seeks to determine the extent to which the values implicit in the ranking of presidents by scholars align with the values implicit in voting and other choices made by the general population. Accordingly, outcomes are examined, rather than personality and achievements before or after serving as president,16 using a four-factor model involving (1) real GDP growth, (2) federal non-defense expenditures, (3) war, and (4) scandal. Operational definitions and sources are given in a data appendix.
In terms of economic performance over presidential terms, real GDP growth has varied: 3.9 and 4.1 percent during the Reagan and the Bill Clinton administrations, for example, and 2.6 and 1.9 percent during the tenures of George H.W. Bush and George W. Bush.
With regard to the ratio of federal expenditure to GDP, it is dominated by spikes associated with war-time spending. To distinguish the growth of government from the conduct of war, this study focuses on non-defense expenditure. The ratio of non-defense expenditure to trend GDP rose from the 1950s to the 1970s, leveled off during the 1980s, and declined during the 1990s into the early 2000s.17
The scale of war, presented in table 2, is relative—taking into account the length, cost, and number of casualties in a war relative to the average length, cost, and number of casualties in wars in which the United States has been involved. The square root of each of the three ratios is calculated to account for diminishing marginal utility; then the three are multiplied, subject to a maximum of one for the product sum. Not surprisingly, the Civil War and World War II are indicated to be the largest two wars. Somewhat surprisingly, the Vietnam War joins World War I as almost as large (remember, this scale is nonlinear; its indicated differences are not proportional). The American Revolutionary War (not associated with any president), Korean War, and the wars in Iraq and Afghanistan constitute a third tier.
Scale of Wars
War of 1812
Kennedy, Johnson, Nixon
Persian Gulf War
Afghanistan and Iraq Wars
Bush, G.W., Obama
Scale = MIN(1,Cost^½ times Length^½ times Casualties^½) where Cost is the ratio of the Cost of a war (including an imputation for the economic loss due to those killed and wounded during the war) to the average cost of wars; Length is the ratio of the length of a war in years to the average length of wars; and Casualties is the ratio of the Americans killed and wounded during a war to the average number of casualties during wars.
This study’s scale of scandal, presented in table 3, is also relative, taking into account who, what, and the disposition of each scandal listed in the relevant Wikipedia article. A scandal involving the president and a high crime resulting in impeachment by the House and removal by the Senate or in a forced resignation is accorded the value 1. Scandals that involve persons a bit removed from the president and run-of-the-mill bribery, corruption, or a sex scandal that result in something less than a conviction or its equivalent get smaller numbers. Some of the listed scandals get a value of zero. The values of the scandals during each president’s tenure are totaled, subject to a maximum of 1. The worst presidents in terms of scandal are clearly Nixon, Grant, and Warren Harding. Andrew Johnson, Reagan, Clinton, and George W. Bush also wind up with large numbers (in the latter three cases, due almost exclusively to scandals during their second terms).
Scale of Scandals
Sec. of Treasury blackmailed over affair and “confesses”
Sec. of War John Henry Eaton has affair leading to suicide of cuckolded husband
Sec. of War resigns due to corruption charges
House impeaches president for violating
Tenure of Office Act, Senate does not remove
Congressman Oakes Ames censured by House of Representatives
kickbacks to Sec. of War
massive corruption during Grant administration
Sec. of War impeached and resigned for bribery
Commissioner of Indian Affairs forced to resign because of corruption
Director of Veterans Bureau found guilty of fraud and bribery
President involved with a mistress
Sec. of Navy resigns for his part in scandal
Sec. of Interior Albert B. Fall convicted and goes to jail
Chief of Staff resigns after not answering Congress about certain gifts
Aide to Pres. resigns after charges of “favoritism”
Aide to Pres. caught in gay liaison in YMCA bathroom
VP convicted of bribery while he was Gov. of Maryland, and resigns
President forced to resign
Dir. of OMB resigns amidst allegations of misuse of funds of a bank in Georgia
various members of administration resign and/or convicted
various members of administration resign and/or convicted
various members of administration resign and/or convicted
Treasurer of the US convicted of tax evasion
Assoc. Attorney General convicted of tax evasion
House impeaches president for lying under oath about a mistress, Senate does not remove
Sec of HUD resigns, convicted of false statements in conjunction with a mistress
various members of administration resign and/or convicted
Chief of Staff of VP guilty of false statements
Dep. Sec of State resigns after revelation he had frequented the “DC Madam”
Scale = “Who” times “What” times “Disposition,” where “Who” equals 1 if president, 2/3 if VP, cabinet officer, or chief of staff, and 1/3 if aide, subcabinet officer or member of Congress associated with the administration; “What” equals 1 if High Crime, 2/3 if bribery or other ordinary corruption, and 1/3 if sex scandal; and, “Disposition” equals 1 if impeached and removed or forced to resign, ¾ if convicted in a court of law, ½ if impeached and not removed, or if resigned or censured, and ¼ if indicted in a court of law but not convicted or if a public confession.
Scholars versus the General Public
The first column of table 4 presents a regression analysis of the 434 rankings of presidents by scholars in twelve surveys, relative to the above-discussed performance measures. The second and third columns present regressions of the popular votes and Electoral College votes received by the candidate of the incumbent party, relative to the performance measures of the prior term.18
Regression Analysis: Scholar Rankings versus Contemporary Popular Assessments (absolute values of t-statistics in parentheses)
Popular Vote 1824–2012
Electoral Vote 1796–2012
Average Rate of Real GDP Growth
Change in Federal non-defense Exp/Trend GDP
Conservatives times Change in Fed non-defense Exp/Trend GDP
Conservatives times Good versus Bad War
The results indicate that scholars are not impressed by presidents associated with a vibrant economy. Indeed, the coefficient on GDP growth is wrong-signed and significant. In contrast, GDP growth is positive and significant in both the popular and Electoral College vote regressions.
With regard to the change in federal non-defense expenditure, similar results are obtained in all three regressions, although significance varies. In one of only two instances in which conservative scholars appear to distinguish themselves from their fellow, predominantly liberal colleagues, they oppose increasing the size of government relative to other scholars.19
With regard to war, scholars appear to love wars—and liberal scholars appear to love war without discrimination. Conservative scholars show some discrimination in their taste for war, liking good wars a bit more and bad wars a bit less. In contrast, the people appear to dislike bad wars and react to bad wars differently than they react to good wars, although these findings fail to meet an acceptable level of significance.
With regard to scandal, scholars strongly penalize presidents who become associated with scandals. By contrast, the people do not appear to punish candidates of parties whose sitting presidents were associated with scandals.
The variables X and X2 are designed to reflect any tendency of the ranking of presidents by scholars to change during the twenty years following departure from office. It appears that presidents begin about ten percentiles below where they ultimately wind up, and that the path is nonlinear, first rising and then settling back a bit.
Surveys of the General Public
What about the values implicit in the choices made in surveys asking the general public for retrospective assessments of presidents? Scholars have generally dismissed such surveys; for example, Elmer Plischke wrote in 1985 that few participants in a 1956 Gallup Poll could even name three presidents.20 Mindful of this and other limitations in surveys of the general population, this study examined responses from two types of questions that pertain to approval ratings and four- or five-part assessments, while avoiding the data from a third, frequently-used type of question, which involves ranking by the percent naming which president is the greatest.
Table 5 presents regression analysis of retrospective surveys of the general population. Many of these surveys are posted on the Polling Report website (http://www.pollingreport.com/) and others were obtained from diverse places.21
Regression Analysis: Surveys of the General Public (absolute values of t-statistics in parentheses)
Retrospective Net Approval Rating: Past Presidents
Retrospective Net Approval Rating: Recent Presidents
Popular Greatness Assessment (on a scale 0 to 4)
Average Rate of Real GDP Growth
Change in Federal non-defense Exp/Trend GDP
Column 1 of table 5 presents retrospective approval ratings of presidents whose last full year of office was at least twenty years prior to the date of the survey. The approval rating is the percent saying “approve” minus the percent saying “disapprove” to a question such as, “Do you approve or disapprove of the way NAME handled his job as president?” Of the fifty-two retrospective assessments of long-past presidents, thirty-six are from a 2007 Rasmussen poll that (amazingly!) queried the public about the performance of every past president.22 Most of the other assessments are from Gallup polls.
Many participants in the 2007 Rasmussen Poll declined to offer an opinion regarding the lesser known presidents of the nineteenth century and early twentieth century. Based on the responses of those offering an opinion, it appears from the first column of table 5 that with regard to the performance of presidents of the distant past the general public—like scholars—does not much care about economic performance, loves good wars, does not dislike bad wars (which are loved by scholars), and deeply dislikes scandals. These results may reflect the impact of scholars on the general public in the teaching of history and other ways.23
The second column of table 5 examines the values implicit in retrospective approval ratings of recent presidents; the third column examines the values implicit in four- or five-part ratings of presidents no earlier than Franklin D. Roosevelt. Most of the data analyzed in the last column come from Zogby Polls conducted between 1997 and 2008.24 The first two such surveys, however, are Gallup Polls conducted in 1956 and 1957 that concerned Harry Truman.25 The third, also a Gallup Poll, was conducted in 1960 and concerned Dwight Eisenhower.26 And the fourth is a 1983 Harris Poll that concerned Kennedy, on the twentieth anniversary of his assassination.27 Beginning in the mid-1980s, Gallup began to inquire periodically about the greatness of Reagan and, later, of other current and recent presidents. This study uses the responses to surveys conducted in at least the seventh year of two-term presidents or following a president’s last full year in office.
In examining columns 2 and 3 of table 5 it is clear that the general public is positively impressed by a vibrant economy, hates bad wars, and distinguishes bad wars from good wars. Other findings are inconsistent or insignificant, perhaps due to small sample size. The consistent and significant results are consistent with those obtained with the popular and Electoral College vote (reported in table 4), and inconsistent with the findings obtained with scholars (reported in column 1 of table 4), and with approval ratings of long-past presidents (reported in column 1 of table 5).
Focusing on popular assessment of greatness (table 5, column 4), where assembling a large number of assessments was possible, allows for two additional observations: All coefficients are significant except the two that would reflect any tendency of ratings to evolve over time. Unlike scholars, the general public does not tend to underrate very recent presidents. Scandals, however, diminish the general public’s view of a president, even though voters do not penalize the candidates of parties whose sitting presidents were associated with scandals.
Summary and Discussion
Regression analysis reveals significant differences between the values implicit in the rankings of presidents by scholars and the values implicit in contemporary popular endorsement, with regard to voting and retrospective assessment of recent presidents. The people care a lot about the economy, while scholars do not. Scholars are drawn to war, whereas the people are repulsed by most wars. Scholars heavily discount presidents whose administrations are marred by scandal, while the people do not harshly judge the candidates nominated by the parties of these presidents. Although several differences are indicated between conservative and liberal scholars, the differences are only marginally significant. It seems fair to say that the concerns of the people are ordinary—and involve going about the business of life and not getting swept up in the affairs of state, whether this involves war-making, sexual intrigue, or petty corruption. By contrast, scholars seem fixated on drama.
The distinction between elite and popular opinion on war is well-known. In his speech at Moscow State University, Ronald Reagan put it this way, “People do not make wars, governments do.”28 The classical liberal supports democracy specifically because of the opportunity it provides to subordinate government to the interests of the ordinary person—and, these interests are, almost always, peace and prosperity.
Conservatives: 1 for ISI 1996 survey and ½ for Wall Street Journal 2000 and 2005 surveys.
Electoral College Vote: Percentage of the Electoral College votes won by the candidates of the two major parties, won by the candidate of the incumbent party, from David Leip’s Atlas of U.S Presidential Elections (http://uselectionatlas.org/).
Good War: Civil War and WWII.
Incumbent: 1 if the current president is the candidate of the incumbent party.
Popular Vote: Percentage of the votes received by the candidates of the two major parties, won by the candidate of the incumbent party, from David Leip’s Atlas.
Popular Greatness Assessment: Mean rating of presidents on a scale of 0 to 4 in surveys of the general population conducted during the seventh year of two-term presidents or at least the year following a president’s last full year in office.
Retrospective Net Approval Rating, Past Presidents: Percentage responding “Approve” minus percentage responding “Disapprove” of president in surveys of the general population conducted at least twenty years following a president’s last full year in office.
Retrospective Net Approval Rating, Recent Presidents: Percentage responding “Approve” minus percentage responding “Disapprove” of president in surveys of the general population conducted at least in the year following and not more than twenty years following a president’s last full year in office.
Scholar Rankings: Percentile rankings in Schlesinger 1948 and 1962, Schlesinger Jr. 1996, OAH 1970, Chicago Tribune 1982, Murray-Blessing 1982, Ridings-McIver 1996, ISI 1996, Wall Street Journal 2000 and 2005, and Siena 2002 and 2010. Rankings exclusive of Harrison, Garfield, and current president. Rankings in OAH survey based on average of prestige, strength of action, and accomplishments. Rankings in Siena 2002 and 2010 based on category “your present overall assessment.” Rankings in Schlesinger Jr. 1996 based on recoding “Failure” as 1 on a scale of 1 to 5.
War: Scale given in table 2 of War of 1812, Mexican-American War, American Civil War, Spanish-American War, Philippine-American War, WWI, WWII, Korean War, Vietnam War, Persian Gulf War, and wars in Iraq and Afghanistan.
Scandal: Scale given in table 3 of political scandals listed in Wikipedia.
Trend GDP: GDP Price Deflator times Trend Real GDP, where Ln(Trend Real GDP) = 1.4953 + 0.0385*(Year–1789)–0.0043*s1, and where s1 = Year–1886 if > 0, 1886 being determined by a grid search.
X = MAX(0, 1-(Year of survey–Last full year in office)/20