¹ Estimates of turnout lack much meaning except as raw vote totals are stated as a proportion of the potential electorate. Whereas the number of adults over 21 in the nation is known with reasonable precision, the number of adult citizens over 21 who are “eligible” according to any of several possible definitions depends on cruder estimates, which can be quite diverse. For the 1952 figure we employ here Table No. 446 for the “Civilian Population 21 Years and Older,” Statistical Abstract of the United States (1958 edition). The 1960 figure rests on an estimated 107 million adults over 21.

² The South is defined throughout this article to include 15 border and deep Southern States. Texas, Oklahoma, Arkansas, Kentucky, West Virginia and Maryland (but not Delaware) are included and form the western and northern boundaries of the region.

³ Population growth in areas such as Texas and Florida includes the immigration of American citizens from outside the South who are more accustomed to voting in every election than are Southerners. It seems almost certain, however, that there was a real increase in motivation among long-term Southern residents as well. The factor of population change was quite insufficient to account for the 1960 increase in Southern turnout.

⁴ New York Times, 11 20, 1960, Section 4, p. E5.Google Scholar

⁵ Results of the 1956 survey, considered as a simple cross-section sample of the nation, are reported in Campbell, , Converse, , Miller, and Stokes, , The American Voter (New York, 1960)Google Scholar. There are natural difficulties in any attempt to retain contact with a farflung national sample over periods of two and four years, especially in a population as geographically mobile as that of the current United States. Of the original 1763 respondents interviewed twice in 1956, nearly 100 had died before the 1960 interview. Others had been effectively removed from the electorate by advanced senility or institutionalization. Of the remaining possible interviews, numbering somewhat over 1600 people, more than 1100 were successfully reinterviewed in the fall of 1960. The 1956 social, economic and political characteristics of the 1960 survivors show almost no sign of deviation from the characteristics of the larger pool of original 1956 respondents. Therefore, although attrition may seem substantial, there is no evidence of alarming bias.

⁶ The sequence of interviews in 1956, 1958 and 1960 was carried out under grants from the Rockefeller Foundation. The 1960 sample design provided not only contact with the 1956 panel which, due to aging, no longer gave an adequate representation of the 1960 electorate, but also a set of additional interviews filling out an up-to-date cross-section sample of all adult citizens living in private households in 1960. Analysis of the additional interview material is being carried out under a grant from the Social Science Research Council. Both the panel and cross-section bodies of data contribute, where appropriate, to materials in this article.

⁷ The longitudinal analysis of political change permitted by a panel design can only be poorly approximated in simple cross-section surveys, where deductions must rest on the respondent's recollection of his behavior in time past. Most analysts have justly felt uncomfortable with recall materials of this sort, since it has been clear that the accuracy of a vote report declines rapidly as time passes. In both 1958 and 1960, we asked our respondents to recall their 1956 vote. The results, as compared with actual reports collected just after the 1956 election, demonstrate forcefully the inaccuracies which accumulate with time.

⁸ Throughout this article, the “eligible electorate” is taken to consist of chose non-institutionalized citizens over 21. Negroes disqualified in many parts of the South, for example, are included in this bounding of the electorate, as well as those who had moved too recently to have established new voting residences in 1960.

⁹ The concept of party identification is treated in detail, Campbell et al., op. cit., p. 120 ff.

¹⁰ Our theoretical understanding of this net of relationships is suggested ibid., pp. 161–167.

¹¹ Participation rates by age in 1960 follow rather nicely the rates indicated ibid., Fig. 17-1, p. 494.

¹² The absence of any significant change in the distribution of party loyalties throughout the Eisenhower period is best illustrated by Table 6-1, ibid., p. 124. Distributions drawn in 1959 and 1960 continue the same pattern. Furthermore, there is no evidence that this surface stability of party identification is concealing a great flux of compensating changes beneath the surface. There have indeed been one or two slow, modest evolutions since 1952 which do involve compensating changes and hence are not visible in simple distributions. Nevertheless, our panel data show strikingly that among all the political orientations which we measure, partisan identification is by far the most stable for individuals over the periods which our data cover.

¹³ This figure should be taken to indicate a rough range. It would vary upward or downward slightly according to the assumptions made concerning the overall proportion of the electorate turning out. While the computations underlying this estimate are tedious, their rationale is entirely straightforward. Turnout rates and the two-party vote division within each of seven categories of party identifiers have shown remarkable regularities over the range of elections which we have studied. These rates are not constant from election to election, but do vary for each type of identifier quite dependably as a function of the net balance of short-term partisan forces characterizing the specific election. While we have observed no election which registered a perfect equilibrium of these forces, we have observed situations in which net forces were Democratic (e.g., 1958) as well as Republican (e.g., 1952 or 1956). It is therefore possible to compute the turnout rate and two-party vote division which could be expected for each type of party identifier in the intermediate case in which shortterm forces are balanced. The estimate employed above derives from a summation of these computations across categories of identifiers, weighted in a fashion appropriate to represent the entire electorate.

¹⁴ Re-examination of earlier data shows a faint residual relationship between Republican voting and church attendance among Democratic Protestants which is not statistically significant. In 1956, the rank-order correlations involved were about .05 both within and outside the South. On the other hand, the comparable coefficient for Independents in 1956 was negative, −.04. The text ignores these variations as probably inconsequential.

¹⁵ This is simply a special case of propositions concerning group identifications more generally, discussed in Campbell et al., op cit., ch. 12.

¹⁶ Two other motivational patterns associated with religion in 1960 deserve note. There were undoubtedly broad-minded Protestants who were drawn to a Kennedy vote out of a desire to see the religious precedent broken and hence buried; and there were undoubtedly Catholics who were drawn away from a Kennedy vote out of fear that the fact of a Catholic president would keep the religious issue uncomfortably prominent. It is hard to find instances of these viewpoints in our sample, however, and it is to be assumed that their incidence was slight.

¹⁷ Campbell et al., op. cit., ch. 19.

¹⁸ Campbell, Angus, “Surge and Decline: A Study of Electoral Change,” Public Opinion Quarterly, Vol. 24 (Fall 1960), pp. 397–418 CrossRefGoogle Scholar.