¹ Data for this report are primarily from the Center for Political Studies 1972 national election survey, the twelfth in a series, which was based on a standard University of Michigan Survey Research Center national stratified probability sample of 2705 cases. The data are derived from personal interviews with eligible voters (citizens 18 years and older) taken before the election (September 1 to November 6) and reinterviews with the same respondents taken immediately after the election. The data and relevant codebooks for the 1972 study are available to any interested scholars upon request through the Inter-university Consortium for Political Research, The University of Michigan, Ann Arbor, Michigan 48106.

² Party identification is based on the standard measure which asks the respondent: “Generally speaking, do you usually think of yourself as a Republican, a Democrat, an Independent, or what? Would you call yourself a strong Republican/Democrat or a not very strong Republican/Democrat?” If Independent—“Do you think of yourself as closer to the Republican or the Democratic party?” The question gives rise to a measure with seven categories that range from strong Democrat to strong Republican. For the purpose of this paper whenever Democrats, Independents and Republicans are referred to, the leaning Independents have been collapsed with the Independent-Independents.

³ The most extensive analysis of the concept of party identification is provided in Campbell, Angus, Converse, Philip E., Miller, Warren E., Stokes, Donald E., The American Voter (New York: John Wiley and Sons, Inc., 1960)Google Scholar. See especially p. 124 for the distribution of party identification in the 1950's.

⁴ In response to the question, “Generally speaking, would you say that you personally care a good deal which party wins the presidential election this fall, or that you don't care very much which party wins?”, the percentage personally caring about the outcome of the election in 1952 was 68%; in 1956, 65%; 1960, 67%; 1964, 69%; 1968, 64%; 1972, 62%.

⁵ Lipset, Seymour Martin and Raab, Earl, “The Election and the National Mood,” Commentary, 55 (01, 1973), 43Google Scholar. Emphasis in the original.

⁶ Kirkpatrick, Jeane, “The Revolt of the Masses,” Commentary, 55 (02, 1973), 60Google Scholar.

⁷ Rosenthal, Jack, “The Secret Key Issue: Study of Polls Shows Racial Attitudes to be Critical with Nixon Gainer,” The New York Times (11 7, 1972)Google Scholar.

⁸ Philips, Kevin, “How Nixon Will Win,” The New York Times Magazine (08 6, 1972), pp. 8, 34, 35–37Google Scholar.

⁹ The issue scales used in Table 1 ranged from 1 to 7, with the extremes indicating designated policy alternatives. The respondents were asked to place themselves on the issue scale. The scales were trichotomized for ease of tabular presentation: categories 1 to 3 were combined into the “left” category, number 4 formed the “center,” and categories 5 to 7 were designated as “right.” On all the issue scales the left position identifies the liberal position. This is most clearly identified on the last scale in the table which ranged from “liberal” to “conservative” without any specific issue content.

The five-issue policy orientation index is different in that it combined responses to five of the specific issue scales. To form this index an average score was obtained from self-placement on the issues of the Vietnam War, urban unrest, campus demonstrations, protecting the rights of the accused and government aid to minorities. Scores ranging from 1–3 were considered as pro-social change while scores of 5–7 were designated as pro-social control. Placement at 4 on the scale was labeled “center.” The distribution for this trichotomized index is 41% pro-social change, 29% at center, and 30% pro-social control. For a more complete discussion of this concept, see Miller, Arthur H., “Political Issues and Trust in Government,” The American Political Science Review, 68 (09, 1974), 951–972CrossRefGoogle Scholar.

¹⁰ It should be noted, though, that of these three issues only on the question of amnesty did a majority of McGovern voters take a liberal position; 70 per cent were against busing and 52 per cent favored increasing the penalties for the use of marijuana.

¹¹ Attitudes toward contemporary social issues were measured with 7-point scales in 1968, and again in 1970, as well as in 1972. In each of these studies the respondents not only placed themselves on these scales, but they also indicated where they perceived several political objects to be located relative to the policy alternatives represented by the scale extremities. Among the political objects placed on the scales were the presidential candidates and, after 1968, the parties, thus providing data for determining not only how the population has shifted on the issues but also how the perceived locations of various candidates and the parties compare over time.

It should be noted that the average values placed in Figure 1 de-emphasize the high degree of polarization existing in the attitudes toward these issues. While many of the values fall at the center of these scales, the data should not be interpreted to imply that the majority of the population prefers a centrist approach to these social problems. Only a brief glance back at the group distributions on these two issues in Table 1 is necessary to confirm this.

¹² The very large difference between the two candidates on the issue of Vietnam is highlighted by the finding that 89 per cent of the population perceived McGovern as left of center on this issue whereas only 22 per cent saw Nixon as favorably disposed to a policy of immediate withdrawal. Large differences were also perceived between the candidates on the issue of government assistance to minorities: 32 per cent thought Nixon liberal on this issue, compared to 73 per cent for McGovern. Minor differences were perceived between the candidates on only the issues of inflation and pollution.

¹³ A complete statement of the normal vote concept and computational methods can be found in Philip E. Converse, “The Concept of a Normal Vote” in Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., Elections and the Political Order (New York: John Wiley and Sons, 1966), pp. 9–39Google Scholar.

The normal vote parameters used here are not those originally calculated by Converse. The empirical observations employed by Converse to determine the partisanship values and expected turnout parameters used in the normal vote computations were from the national elections of 1952 through and including 1960. Since the recent decade has witnessed several electoral anomalies, including a massive victory for each of the major parties and a third party insurgency, the normal vote parameters (both turnout and partisanship) were recomputed on the basis of the six national elections during the period 1962 to 1972. The baseline used here is therefore more representative of the contemporary partisan situation than one computed from Converse's original estimates of the normal vote parameters would have been. A complete report of the new normal vote parameters is now in preparation.

Applying the normal vote computations to the 1972 total sample provides an expected Democratic vote of 54 per cent. When this value is compared to the 36 per cent of the two-party vote that McGovern actually received (after excluding the few per cent voting for minor party candidates), a deviation of 18 percentage points below the expected Democratic proportion of the vote for the total sample is observed.

¹⁴ See in particular Campbell, Angus et al. , The American Voter, pp. 120–167Google Scholar.

¹⁵ These normal vote indices were developed by Boyd, Richard W., “Popular Control of Public Policy: A Normal Vote Analysis of the 1968 Election,” The American Political Science Review, 66 (06, 1972), 429–449CrossRefGoogle Scholar.

A long-term index value (L) of zero indicates the lack of any relationship between traditional partisan divisions and an independent variable. Non-zero values indicate the average absolute deviation of the expected vote for each category of the independent variable from the expected vote for the total population. For example, the relationship between campus unrest and partisan difference indicates an average absolute deviation of 4.54 percentage points from the expected Democratic vote for the total population, a relationship that is slightly stronger than that for government help to minorities (L =4.24). For the short-term forces index (S), if there is no deviation from the baseline or if the deviation is in the same direction and equal for all categories of the independent variable, the variable would have no relationship to the defections from party and the short-term index value would be zero. Government help to minorities, for instance, has some relevance to the short-term forces in 1972(S = 8.63). But compared to Vietnam (S = 11.80), its short-term index is relatively moderate. The short-term index value for Vietnam indicates an average deviation of 11.80 percentage points from the baseline vote normally expected for each policy group (i.e., each category of the issue variable).

¹⁶ Boyd, , “Popular Control of Public Policy,” p. 448Google Scholar.

¹⁷ The issue of government aid to minority groups does not, however, elicit only racial attitudes. The primary component of this item is the degree to which the individual, rather than society, is to be blamed for the social and economic misfortunes of certain segments of society. It also refers to the extent of social change necessary to alleviate the adverse socioeconomic conditions of blacks and other minorities. While one aspect of this social attitude is definitely racial in nature, it is not as clearly racial in content as is the desegregation item of Figure 2f.

The Gallup Opinion Index shows that mention of racial problems as the most important issue facing the country today had dropped from 40 per cent in mid-October, 1967 to 5 per cent in early 1972. See Gallup Opinion Index, Report No. 78, December, 1971.

¹⁸ The rationality of the voter has been debated heatedly in the literature of late, and while we believe that the data presented here deal directly with this controversy, it is not the purpose of this report to review it.

¹⁹ These criteria are dealt with in Brody, Richard A. and Page, B. I., “Comment: The Assessment of Policy Voting,” The American Political Science Review, 66 (06 1972), 450–458CrossRefGoogle Scholar.

²⁰ Two features of the 7-point issue scales allow us to operationalize this linkage. First, all issues are screened for salience to the individual respondent. The survey questions utilizing the 7-point scales have a self-contained saliency screen which asks, “Or haven't you thought much about this?” This helps in meeting the saliency requirement of the proximity model of issue voting. Secondly, the respondent is asked to locate each candidate on each scale. Thus, for every issue a trichotomous proximity index parsimoniously reveals which candidate is seen as closer to the voter or if the two alternatives are equidistant.

The proximity measures are computed by simply comparing the absolute differences between the individual's self-location and the candidate position on the 7-point issue scale and assigning to the individual respondent a score of −1 if closer to McGovern, +1 if closer to Nixon, and zero if equidistant from both candidates.

²¹ It may be suggested, however, that the stronger relationship between the proximity measure and vote is simply a spurious relationship arising because individuals locating themselves on the left systematically placed McGovern closer to themselves and Nixon farther away, while those on the right did just the opposite when responding to the survey instrument. Relationships such as those between the Vietnam proximity and vote (presented in Figure 3a) persist even after controlling for left and right self-placement on the issue. This is clearly revealed by the partial correlations between vote and the proximity measures for each of 14 separate public policy issues. The partial correlations remain virtually identical to the zero-order correlation after controlling for the respondent's stated policy preference on each issue. However, when the proximity measure is used as the statistical control in computing the partial correlations, the resulting partial between vote and policy preference is always either very close to zero or at least much smaller than the zero-order relation between self-location and vote. These findings demonstrate that it is almost entirely the proximity model that explains the relationship between an individual's own policy preference and his vote decision. Regardless of an individual's own position on the issues, he is considerably more likely to vote for the candidate he believes to be closer to him on the salient issues than for the candidate he perceives as relatively further from his own policy preference.

Other data, however, suggest that rationalization and/or candidate persuasion may have inflated the correlation between the proximity measures and the vote. For instance, the long-term index values for the Vietnam and liberal-conservative proximity indices of Figure 3 were larger than those for the comparable self-location items of Figure 2. This indicates that party identification may have been inducing congruence between the vote and candidate proximity. That is, Democrats who were not entirely aware of McGovern's position on the issues may have been placing McGovern closer to themselves on the issues simply because he was a Democrat and they eventually voted for him because of that, not because of the issues.

Other analysis also suggested that there was some tendency for positive candidate affect to result in closer proximity to a candidate despite the respondent's own issue position. Thus, positive attitudes toward a candidate or his party may have caused respondents with unclear perception of the candidate's policy position to place their preferred candidate closer to themselves on the issues. This in turn would have increased, spuriously, the relationship between the proximity measures and vote. A similar result would have occurred if the respondent had not formulated an attitude on the issue but had simply placed himself at the issue position he believed to coincide with his preferred candidate or party. The rationalization and persuasion effects may have inflated the correlation between proximity measures and the vote, leading to an overestimate of the rational policy voting that did indeed occur. This problem is, however, greatly reduced when conducting multiple equation analyses that include measures of the factors presumed to cause the rationalization or persuasion effects. That is, if these effects are the result of affect toward the candidate or party loyalty, then their spurious relationship to the vote will be removed from the direct relationship between the issue proximity and the vote by including the assessment of candidate affect and party identification in the same explanatory equation containing the issue measures. This is exactly the strategy that was pursued when assessing the relative importance of the several factors affecting the election outcome.

²² A factor analysis of some twenty issues was employed to ascertain the dimensional structure of the issue space. The loadings of the items on their respective common factors were all above.6.

²³ The results of a multivariate analysis with the two-party vote as the dependent variable and the several individual issue proximities excluding the liberal-conservative scale but including party identification revealed the seven most important single issues. These were, in order of importance as determined by a step-wise regression: Vietnam, a government guarantee of a job and good standard of living, dealing with campus unrest, protecting the rights of individuals accused of crimes, legalization of marijuana, the use of busing to achieve racial integration of schools, and urban unrest. These seven issue proximities explained roughtly 60 per cent of the variance in the vote. The last four mentioned issues, however, each added only I per cent of explained variance and only 2 per cent more explained variance is achieved after adding eight other issues to the regression analysis.

²⁴ Converse, Philip E., “The Nature of Belief Systems in Mass Publics,” Chap. 6 in Ideology and Discontent, ed. Apter, David (New York: Free Press, 1964)Google Scholar.

²⁵ Converse, , “Nature of Belief Systems,” p. 227Google Scholar.

²⁶ Nie, Norman H. and Andersen, Kristi, “Mass Belief Systems Revisited: Political Change and Attitude Structure,” The Journal of Politics, 36 (08, 1974), 540–541CrossRefGoogle Scholar; Pomper, Gerald M., “From Confusion to Clarity: Issues and American Voters, 1956–1968,” The American Political Science Review, 66 (06, 1972), 415–428CrossRefGoogle Scholar.

²⁷ It might be suggested that if self-placement on the issues were used in place of the proximities, party identification might appear to be relatively more potent because these measures are more coordinate in the causal sense in which they are treated by Campbell, Angus et al. , The American Voter, p. 128Google Scholar. A multiple regression was therefore performed using the 1972 self-placement issue measures instead of proximities. The calculations show that some of the issues were still as important as party identification in predicting the vote. But as Kessel has pointed out, comparisons of the relative magnitude of relations using different types of question formats are fraught with methodological difficulties. The relationship between these different types of issue questions and party identification along with possible shifts in the relationship over time is quite difficult to discern. See Kessel, John H., “Comment: The Issues in Issue Voting,” The American Political Science Review, 66 (06 1972), 460–461CrossRefGoogle Scholar.

²⁸ The stage-wise regression employed here operated on a “case deletion” procedure, that is, a case was excluded from the analysis if missing data was encountered for any single variable used. The order of the variables entered and the beta coefficients for each variable in the final equation was as follows: party identification (.13), the liberal-conservative proximity (.11), war (.08), social (.16), economic (.03), and radical indices (.01), a measure of reaction to McGovern's campaign performance (.06), McGovern ( −.26), and Nixon (.25) thermometer ratings. The multiple R was .82 and R ² equaled .68.

²⁹ For a concise statement of path analysis see Stokes, Donald E., “Compound Paths in Political Analysis,” in Mathematical Applications in Political Science V, ed. Herndon, James F. and Bernd, Joseph L. (Charlottesville, Virginia: University Press of Virginia, 1971), pp. 70–92Google Scholar.

A four-variable, multiple equation, path analysis including party identification, a single composite issue proximity measure, a composite candidate thermometer and the vote confirms the results of the stage-wise regression and the limited causal analysis presented in the text. The path coefficients and total effects summed from the several equations for the three independent variables are presented below.

It should be noted that the total effects are only for themodel presented. This model assumes that issue attitudes are not determined by candidate ratings; thus it corresponds to the requirements for policy voting as stated in the text.

³⁰ A preliminary attempt to determine the effects of issues on the vote was made by D. M. Kovenock, P. L. Beardsley and J. W. Prothro, “Status, Party, Ideology Issues, and Candidate Choice: A Preliminary Theory-Relevant Analysis of the 1968 American Presidential Election,” paper delivered at Specialist Meeting B: XI (“New Approaches to the Study of Social Structure and Voting Behavior”), Eighth World Congress of the International Political Science Association, Munich, Germany, August 31–September 5, 1970. It should be pointed out that contrary to what we treat as issues Kovenock et al., included reactions to individuals such as the vice-presidential candidates as an issue.

³¹ Turnout in presidential elections from 1964 to 1972 was as follows:

These data are from the CPS national election surveys for the indicated years.

³² It should be noted that the Independents Category includes those individuals who lean toward one of the parties as well as those who have no tendency to identify with a party. When “leaning “ Independents are excluded, the per cent of Independents increased from 8 per cent in 1964 to 13 per cent in 1972. Most of this increase in Independents appeared, in the aggregate, to be at the expense of the Democrats since their proportion of the electorate declined from 52 per cent in 1964 to 41 per cent in 1972. At the same time the proportion of Republicans remained quite stable at about 25 per cent.

³³ The feeling thermometer is a measure that allows the respondent to indicate his feelings or affect toward the candidates. The measure runs from zero degrees as most unfavorable to 100 degrees as most favorable; 50 degrees is the neutral response.

³⁴ There were sharp racial differences underlying the overall candidate ratings. When whites are examined separately, Nixon's rating goes up while McGovern's drops to where only Chisholm's mean rating is lower and Wallace replaces Kennedy as the most popular Democratic political figure. This latter reversal basically reflects partisan differences in the assessment of Kennedy and the highly positive attitude of Republicans toward George Wallace. Among white Democrats, Wallace rated sixth, behind Kennedy, Humphrey, Shriver, McGovern, and Muskie, in that order. While Wallace's popularity increased dramatically among whites between 1968 and 1972, blacks remained steadfastly and strongly negative toward him, no doubt in reaction to the segregationist and strict law-and-order image popularly associated with Wallace. Black ratings of Kennedy, on the other hand, were exceedingly high. McGovern also rated relatively much higher among blacks, surpassing even civil rights champion Hubert Humphrey.

³⁵ The Eagleton racings appear to be aberrantly high and were apparently inflated by a sympathy factor; they tend to vary only slightly across otherwise analytically useful population subgroups and were therefore ignored in the various comparisons discussed.

³⁶ Converse, et al. , “Continuity and Change in American Politics,” Table 2 on p. 1088Google Scholar.

³⁷ DeVries, Walter and Tarrance, V. L., The Ticket-Splitter: A New Force in American Politics (Grand Rapids, Michigan: William B. Eerdmans Publishing Company, 1972)Google Scholar places primary importance on past voting behavior as an explanation of the individual electoral decision.

³⁸ The 1972 study did not have respondents place Humphrey on the issues; thus we cannot directly assess the difference between him and McGovern. Differences perceived between candidates can, however, be determined from analysis of the candidate thermometers. The general methodology of multidimensional scaling that provides this information, as well as a similar analysis of the 1968 presidential candidates, can be found in Weisberg, Herbert F. and Rusk, Jerrold, “Dimensions of Candidate Evaluation,” The American Political Science Review, 64 (12, 1970), 1167–1185CrossRefGoogle Scholar.

Weisberg and Rusk found that a two-dimensional space provided a “good” solution with a stress of .05 using Kruskal's formula one for stress. In 1972 the two-dimensional solution has a stress of .05 with formula two which gives larger stress values than does formula one for the same solution. The two-dimensional solution places the political figures at the following locations:

The solution indicates a clear differentiation between the Democratic cluster of leaders and the Republicans plus Wallace on the second dimension and very little difference between any of the political leaders on dimension one. This suggests what the stress value of .08 for a one-dimensional solution confirms, namely, that a one-dimensional solution fits the data nearly as well as a two-dimensional solution. This finding strongly supports the argument that the public was reacting to these political leaders in a highly ideological fashion. Furthermore, the solution demonstrates that while voters perceived a definite difference between McGovern and Humphrey, that difference is not strikingly great. Indeed, the only middle-of-the-road politician according to this solution was Henry (“Scoop”) Jackson. Finally, the configuration vividly displays the convergence between Wallace and Nixon that was discussed earlier with respect to the issues.

³⁹ McGovern, in fact, did better among some traditionally Democratic groups than was predicted prior to the election. Jews, for example, voted 3 per cent more Democratic than normally expected, a rather interesting finding since prior to the election there was much journalistic discussion of the Israeli urging that American Jews vote for Nixon because of his foreign policy.

For a more extensive discussion of how the voting in 1972 for various demographic groups compared to the expected and observed voting patterns of these groups since 1952 see Miller, Arthur H. and Miller, Warren E., “Issues, Candidates and Partisan Divisions in the 1972 U.S. Election,” The British Journal of Political Science, (09, 1975)Google Scholar.

⁴⁰ Ranney, Austin, “Turnout and Representation in Presidential Primary Elections,” The American Political Science Review, 66 (03, 1972), 27CrossRefGoogle Scholar. Ranney's findings indicate that those who typically vote in state primaries are usually older and have a higher status in terms of income and occupation categories than the rest of the voting population. They are also more likely to participate in civic, religious and political organizations.

⁴¹ That the liberal-conservative scale is a summary policy measure is evident from its relation to the issues in Table 2. Further evidence is derived from the finding that as more and more issues are added to a composite proximity measure, the correlation between the composite index and the liberal-conservative measure increases up to a value of about .7.

One reason why certain specific issues had an important impact on the vote in addition to the liberal-conservative effect is that about a quarter of the voters could not place themselves on that continuum. The bulk of these respondents could, however, place themselves as well as the candidates on the more specific issue questions.