¹ Among a growing set of articles that document the recent increase in issue voting are the following: Converse, Philip E., Miller, Warren E., Rusk, Jerrold G. and Wolfe, Arthur C., ‘Continuity and Change in American Politics: Parties and Issues in the 1968 Election’, American Political Science Review, LXIII (1969), 1083–105CrossRefGoogle Scholar; Repass, David E., ‘Issue Salience and Party Choice’, American Political Science Review, LXV (1971), 389–400CrossRefGoogle Scholar; Pomper, Gerald M., ‘From Confusion to Clarity: Issues and American Voters, 1956–1968’, American Political Science Review, LXVI (1972), 415–28.CrossRefGoogle Scholar

² For a specific discussion of issues and Wallace voting in 1968 see Converse et al., ‘Continuity and Change in American Polities’.

³ Issues that are uniformly approved or disapproved are called valence issues and they can be quite effectively used in the campaign by the candidates to obtain votes. For a discussion of these types of issues see Stokes, Donald E., ‘Some Dynamic Elements of Contests for the Presidency’, American Political Science Review, LX (1966), 19–28.CrossRefGoogle Scholar

⁴ A more detailed presentation of these requirements for issue voting can be found in Campbell, Angus, Converse, Philip E., Miller, Warren E. and Stokes, Donald E., The American Voter (New York: Wiley, 1960), pp. 168–87.Google Scholar

⁵ All of the issue scales used in the data analysis for this report contained a saliency screening question which asked ‘Or haven't you thought much about this?’ The percentage of respondents saying they had not thought about the issues ranged from a couple of per cent up to about 10 or 12 per cent except for a general liberal-conservative measure for which about one-third of the respondents said they had not thought about it. The generally low percentage of respondents unable to answer the issue items reflects how highly salient the issues were.It should be noted that the 1972 data are derived from personal interviews with 2,705 eligible voters (citizens 18 years and older) taken before the election (1 September to 6 November) and reinterviews with the same respondents taken immediately after the election. The sample was a standard University of Michigan Survey Research Center national stratified probability sample. The data and relevant codebooks for the 1972 study can be made available to any interested scholars upon request through the Inter-University Consortium for Political Research, The University of Michigan, Ann Arbor, Michigan 48106.

⁶ The issue scales ranged from 1 to 7 with the extremes indicating designated policy alternatives. The respondents were asked to place themselves and the presidential candidates on the scale. On all the issue scales the left position identifies the liberal position. This is most clearly identified on the last scale in Fig.1 which ranges from ‘liberal’ to ‘conservative’ without any specific issue content. 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 Independents.

⁷ Twenty different issue items were included in the factor analysis. The loadings of items on their respective factors were all above o.6.

⁸ The proximity measures are computed by simply comparing the absolute differences between the individual’s self-location and the position that the individual assigns to the candidates. The individual respondent was given a score of– I if closer to McGovern, +1 if closer to Nixon and a zero if equidistant from both candidates on each issue. These values were then combined for the issues in each of the major issue domains and collapsed into the five categories used in Fig. 2.

⁹ Despite the exclusion of some ‘party regulars’ from the presidential campaign because the candidates they supported did poorly in the primaries, the Democratic party apparatus was not any less successful in raising money and contacting people in 1972 than it had been in 1968. For example, in 1968 only 25 per cent of self-identified Democrats reported having been contacted by the party; in 1972, 32 per cent said they had been contacted. Also, in 1968, about 4 per cent of all Democrats said they worked for the party in some capacity during the campaign; in 1972, 5 per cent reported working for the party.

¹⁰ The impact that the absence of this traditional surge in support could have had on the vote can also be hypothetically reconstructed. If the data for those Democrats who made their vote decision at the time of the convention or later are mathematically adjusted to reflect such a ‘set back’ factor - which in this case is equivalent to the 8 per cent decline noted earlier - it is dis covered from the hypothetical recomputations that the convention could maximally account for 14 per cent of all Democratic defections. Stated another way, if the convention effect on Democrats had been hypothetically as positive as it appeared to be negative, then McGovern would’ have received 64 per cent of the vote from Democrats rather than the 58 per cent he did obtain, or, conversely, he would not have lost 6 per cent of the Democratic vote because of the convention. Translating this into a hypothetical two-party vote reveals that McGovern’s proportion of the vote could have increased at most by only 2 per cent if the convention had been a maximal ly positive factor in the election. In reality, the impact of the convention was probably less than is suggested by this hypothetical reconstruction. These data strongly suggest that the popular myth has over-estimated the importance of the Democratic convention as an explanation of the election outcome. But, given the nature of the data, these statements must remain far from conclusive.

¹¹ Some writers have argued that a large reserve of occasional voters with little emotional involvement in politics is useful to the political system. Such a group keeps political divisions from going too deep and, by entering the electorate from time to time when discontent runs high, provides the majorities for needed policy changes. This is essentially the view of Berelson, Bernard R., Lazarsfeld, Paul F. and Mcphee, William N., Voting (Chicago: University of Chicago Press, 1954)Google Scholar, Chap. 14, ‘Democratic Practice and Democratic Theory’.

¹² For a longitudinal analysis of the trends in political trust see Miller, Arthur H., ‘Political Issues and Trust in Government’, American Political Science Review, LXVIII (1974), 951–72.CrossRefGoogle Scholar See also Citrin, Jack, Mcclosky, Herbert, Shanks, J. Merrill and Sniderman, Paul M., ‘Personal and Political Sources of Political Alienation’, British Journal of Political Science, V (1975), 1–31.CrossRefGoogle Scholar

¹³ The question wording was, ‘How much of the time do you think you can trust the government in Washington to do what is right – just about always, most of the time or only some of the time?’

¹⁴ The actual question wording was, ‘Would you say the government is pretty much run by a few big interests looking out for themselves or that it is run for the benefit of all the people?’

¹⁵ The political trust measure is a Guttman scale, ranging from most to least, that uses the five survey items regularly used in the SRC-CPS election studies to measure trust in government. A factor analysis with more than twenty items indicated that as in previous years the five trust items form a single dimension with average loadings above 0.7. For a further discussion of this, see Miller, ‘Political Issues and Trust in Government’.

¹⁶ See Miller, ‘Political Issues and Trust in Government.’

¹⁷ See Miller, ‘Political Issues and Trust in Government’.

¹⁸ The five issues used to divide the voters into left, center and right groups dealt with the Vietnam War, urban unrest, campus demonstrations, protecting the rights of the accused and government aid to minorities. An index of average self-placement scores was computed and respondents with scores ranging from 1 to 3 were considered ‘left’; scores 5 to 7 were designated as ‘right’ and scores of 4 were labeled ‘center’. The distribution of this trichotomized index is 41 per cent ‘left’, 29 per cent ‘center’, and 30 per cent ‘right’.

¹⁹ Campbell, et al. , The American Voter, pp. 516–18.Google Scholar

²⁰ Converse, Philip E., ‘Change in the American Electorate’, in Campbell, Angus and Converse, Philip E., eds., The Human Meaning of Social Change (New York: Russell Sage Foundation, 1972), p. 327.Google Scholar

²¹ The political efficacy measure is a Guttman scale ranging from least to most. It uses six survey questions that have been regularly utilized in the SRC-CPS election studies. A factor analysis with more than twenty individual items indicated that the six political efficacy items form a single dimension with average loadings above o.6. For a recent discussion of political efficacy, see Converse, ‘Change in the American Electorate’, and Teresa E. Levitin and Warren E. Miller, ‘The New Politics and Partisan Realignment’, paper delivered at the 1972 meeting of the American Political Science Association in Washington, D.C.

²² The trust and efficacy scales have been dichotomized here by excluding the center category so as to clarify and simplify the cross-tabulation.

²³ The candidate affect measures used here are based on a series of questions asking the respondent what he likes and dislikes about each candidate. Three positive (likes) and three negative (dislikes) statements per respondent were utilized in computing the index. Scores for each respondent ranged from –3 to +3. The group index gives the average number of negative and positive remarks about the particular candidate for the specific groups. A similar version of these indices was originally employed by Stokes, Donald E., ‘Some Dynamic Elements of Contests for the Presidency’, American Political Science Review, LX (1966), 19–28.CrossRefGoogle Scholar

²⁴ 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.

²⁵ Nixon's 1968 thermometer rating is presented in Converse et al., ‘Continuity and Change in American Polities’, Table 2 on p. 1088

²⁶ Kirkpatrick, Jeane, ‘The Revolt of the Masses’, Commentary, LV (1973), 58–62, p. 50Google Scholar; and also ‘Representation in the American National Conventions: the Case of 1972’, British Journal of Political Science, v (1975), 265–322.Google Scholar

²⁷ The cultural orientation index utilizing the thermometer ratings of counter-culture and traditional groups is an average thermometer score computed for each individual after having reversed the thermometer ratings of the traditional groups. Low average scores indicate positive affect toward the counter culture and negative affect toward the traditional groups; high average scores indicate the opposite. The dichotomous version of the index used 60 degrees as the cutting point for forming the dichotomy. Levitin and Miller, ‘The New Politics and Partisan Realignment’, use a somewhat similar measure but point out that this is only one possible dimension for measuring cultural values.

²⁸ The liberal-conservative scale is the seven-point issue scale presented in Fig. I.

²⁹ The multiple regression technique is used to show the effect that certain independent variables have on the presidential vote decision. Readers not acquainted with this statistical technique should note that the beta coefficient refers to the relative importance and effect that a variable has on the vote, controlling for the effect of the other independent variables. In Table 6, the beta coefficients allow for a comparison of the relative importance of each independent variable, both within and between the selected demographic groups. R2, the coefficient of determination, indicates the amount of variance in the vote that is explained by all the independent variables, while R, the multiple correlation coefficient, indicates the total strength of the relationship between vote and the independent variables. For an excellent discussion of multiple regression analysis, see Wonnacott, Ronald J. and Wonnacott, Thomas H., Econometrics (New York: Wiley, 1970), pp. 15–80.Google ScholarEmploying the stepwise regression procedure enables us to determine, in addition to the single most important variable within and between groups, the per cent of additional variance explained added as independent variables are entered. The entry of the first most powerful predictor is determined from a simple correlation matrix. The variable selected is the independent variable most highly correlated with the vote. Then the stepwise procedure selects the next variable whose partial correlation coefficient with the vote is the highest. A partial F-test is made to discern the contribution of the first entered variable given the second variable, and, depending on the significance level (i.e., .001), it is retained or jettisoned. The stepwise procedure continues through each listed variable, and variation explained is added, if possible, from the residual variation left to be explained after the most important predictor variable has entered. See Draper, N. R. and Smith, H., Applied Regression Analysis (New York: Wiley, 1966), pp. 171–3,178–95.Google ScholarWhen using a dichotomous dependent variable such as the vote, users sometimes face the problem of heteroscedasticity, that is, changing variance in the error term. The disturbance, or error, term is assumed to have a constant variance. The normal procedure used for dealing with heteroscedasticity is weighted least squares. See Wonnacott, and Wonnacott, , Econometrics, pp. 133–5Google Scholar and Goldberger, Arthur S., Economic Theory (New York: Wiley, 1964), pp. 249–50.Google Scholar However, due to the slight marginal value of the weighted least squares adjustment, and the lack of heteroscedasticity, the ordinary least squares technique was used here.

³⁰ 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 party identification and the voters’ own issue position are more co-ordinate in the causal sense that they are treated in Campbell, et al. , The American Voter, p. 128.Google 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 John H. Kessel has pointed out, comparisons of the relative magnitude of relations when 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’, American Political Science Review, LXVI (1972), 460–1.Google Scholar

³¹ For a causal analysis of the relationship between the vote and party identification, the issues and candidate ratings see Miller, Arthur H., Miller, Warren E., Raine, Alden S. and Brown, Thad A., ‘A Majority Party in Disarray: Policy Polarization in the 1972 Election’, American Political Science Review (forthcoming, March 1976).Google Scholar

³² The statement of ‘variance added’ in the stepwise regression provides information about which variable had the largest zero-order correlation with the vote, and the relative importance of each added variable. The order in which the variables were added can be ascertained from the per cent of added variance, since that is the criterion used for adding a variable. Often, however, even the combination of information obtained from the stepwise regression and betas does not entirely clarify the relative importance of all the predictors. In these cases a better feel for the data can sometimes be realized from examining the correlation matrix itself. The differences found by level of education are a good example. An examination of the total correlation matrices, presented below, for the grade-school educated and college-educated group helps understanding of the entries in Table 3. The correlations (x 100) in the upper right triangle are for the college-educated and those in the lower left are for the grade-school educated. The differences in the correlations for the two groups clearly reveal more substantial relationships of the vote with the proximity measures for the better educated, stronger correlations of party identification with the issues, as well as smaller correlations (relative to the other correlations for the same group) between vote and the candidate ratings and McGovern’s campaign performance.

³³ The correlations between party identification and the two-party presidential vote from 1952 to 1972 are as follows:

³⁴ The candidate and party affect measures used in the regressions were based on the candidate and party ‘likes’/‘dislikes’ questions described above in fn. 23. The measures used here are ‘net’ affect measures, that is, they indicate the preponderance of positive or negative attitudes toward the Republican candidate or party over those toward the Democratic candidate or party.

⁵ Stokes, ‘Some Dynamic Elements of Contests for the Presidency’.

³⁶ The literature dealing with critical elections includes the following: Key, V. O. Jr,, ‘A Theory of Critical Elections’, Journal of Politics, XVII (1955), 3–18CrossRefGoogle Scholar; see also his ‘Secular Realignment and the Party System’, Journal of Politics (1959), 198–210; Burnham, Walter Dean, Critical Elections and the Mainsprings of American Politics (New York: W. W. Norton, 1970)Google Scholar; see also Schattschneider, E. E., The Semi-Sovereign People (New York: Holt, Rinehart and Winston, 1960)Google Scholar; and Pomper, Gerald, ‘A Classification of American Elections’, Journal of Politics, XXIX (1967), 1–40.Google Scholar

³⁷ 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: Wiley, 1966), pp. 9–39.Google 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.

³⁸ The high degree of Independence among younger individuals was expected because it has been demonstrated in Campbell, et al. , The American Voter, pp. 161–5Google Scholar, that attachment to party increases with age.

³⁹ Survey samples generally show inflated turnout figures since respondents often succumb to the social pressures of civic duty and report voting when in actuality there was none. The sample estimate of turnout in 1972 is 73 per cent while official turnout was roughly 55 per cent. A rough estimate of actual turnout can be obtained by subtracting 18 per cent (the difference between official turnout and the sample estimate) from the group turnout figures obtained in the sample. This correction would yield a turnout figure of 32 per cent voting among 18– to 20–year-olds and 48 per cent among 21– to 24–year–olds.

⁴⁰ Converse, Philip E., ‘On the Possibility of Major Political Realignment in the South’, in Campbell, Converse, Miller and Stokes, Elections and the Political Order, pp. 212–42.Google Scholar See also Converse, ‘Change in the American Electorate’.

⁴¹ Converse, , ‘Change in the American Electorate’, p. 312.Google Scholar

⁴² For a complete description of the calculations involved in estimating the defection rate see Converse, ‘The Concept of a Normal Vote’, p. 27.