¹ The few exceptions include Campbell, Angus, Converse, Philip E., Miller, Warren E., and Stokes, Donald E., The American Voter (New York: John Wiley and Sons, 1960), pp. 266–289 Google Scholar; Kelley, Stanley Jr., Ayres, Richard E., and Bowen, William G., “Registration and Voting: Putting First Things First,” this Review, 61 (06 1967), 359–379 Google Scholar; Key, V. O. Jr., American State Politics (New York: Alfred A. Knopf, 1956)Google Scholar; Matthews, Donald R. and Prothro, James W., “Political Factors and Negro Voter Registration in the South,” this Review, 57 (06 1963), 355–367 Google Scholar; and Miller, Warren E., “Memorandum to the President's Commission on Registration and Voting Participation” (unpublished Michigan Survey Research Center paper, 1963)Google Scholar.

² See Burnham, Walter Dean, “The Changing Shape of the American Political Universe,” this Review, 59 (03 1965), 7–28 Google Scholar.

³ The history of the American ballot system can be found in various sources such as Albright, Spencer D., The American Ballot (Washington, D.C.: American Council on Public Affairs, 1942)Google Scholar; Evans, Eldon C., A History of the Australian Ballot System in the United States (Chicago: University of Chicago Press, 1917)Google Scholar; Harris, Joseph P., Election Administration in the United States (Washington, D.C.: The Brookings Institution, 1934)Google Scholar; Ludington, Arthur C., American Ballot Laws: 1888–1910 (Albany: New York State Library, 1911)Google Scholar; and Wigmore, J. H., The Australian Ballot System as Embodied in the Legislation of Various Countries (2nd ed.; Boston: Boston Book Co., 1889)Google Scholar. In general, these references give much better accounts of the Australian Ballot era than the voting systems in use prior to that time. An exception to this is the book by Evans.

⁴ Although the first Australian Ballots were adopted in 1888 and 1889, they did not become effective in federal elections until 1890.

⁵ The party strip or unofficial ballot came into use in most states during the first half of the nineteenth century. Before that time, the states (or colonies) had other forms of paper balloting, voice voting, or such unorthodox procedures as letting corn or beans designate one's vote. At one time, in the early 1800's, several states allowed the voter to make up his own ballot. The parties, noting this, were motivated to print their own ballots in order to thwart the use of handwritten ballots. This practice of party-prepared ballots was upheld in a Massachusetts Supreme Court decision in 1829, setting a precedent which was not challenged until the beginning days of the Australian Ballot reform.

⁶ These central hypotheses are based on a model of individual decision-making to be presented below. Testing such a model with aggregate election data is subject to the normal analytic cautions expressed by Robinson, W. S. in his article, “Ecological Correlations and the Behavior of Individuals,” American Sociological Review, 15 (06 1950), 351–357 CrossRefGoogle Scholar. For a discussion of the Robinson argument of “ecological fallacy” as it applies to this analysis, see Rusk, op, cit., especially pp. 67–70.

⁷ In contemporary research on survey data, Angus Campbell and Warren E. Miller show the voting effects of one of these partisan devices, the straight ticket provision. They find that there is a greater tendency toward party slate voting when a ballot format has a straight ticket device than when it does not. However, it is difficult to tell to what extent such effects are separate from ballot format considered as a variable since the correspondence between straight ticket provisions and party column formats (or multi-choice provisions and office bloc formats) is very close today. See Campbell, and Miller, , “The Motivational Basis of Straight and Split Ticket Voting,” this Review, 51 (06 1957), 293–312 Google Scholar.

⁸ Burnham, op. cit., p. 9.

⁹ The definition of a “statewide race” was expanded in this paper to include an aggregation of the district congressional vote. Aa a check on this operation, another measure was constructed excluding the congressional vote, and, when compared, both revealed similar behaviors and interpretations in the subsequent ballot analysis.

¹⁰ Burnham did not indicate what screening criteria he used to eliminate third party intrusions in his data. The rule established in this study was that no race would enter the computations if the third party vote contributed six or more percent to the total vote for that race. Admittedly, this figure was somewhat arbitrary and some might consider it too high; however, given the many and large third party movements in this period, the figure seems reasonable. To check this procedure, we also looked at those cases of states having third party vote figures greater than six percent for most of their races and yet where the percentage did not vary greatly across the set of contested races. These data, impure as they were, conformed to the same trendlines in the ballot analysis as the figures generated by the screening criterion selected for use with the Burnham measure. This gave increased confidence in the findings, and on two occasions in the ballot analysis below (Tables 5 and 8), we display the “impure” values of states with the “pure” values of other states, carefully distinguishing the two in such instances.

¹¹ For certain research purposes, the third party vote would be a useful statistic to keep in the split ticket picture. Third parties broaden the alternatives open to the voter and in this way contribute to split ticket voting in certain areas. However, when the concern is with fluctuations in the two party vote space, third party figures can only be viewed as creating needless disturbances in the data.

¹² Aggregate election analyses usually assume that voters crossover or split their tickets mainly in one direction—the direction of national trends. But even if this were not the case, the phenomenon of mutual crossovers would not contaminate our ballot analysis unless it occurred, for some reason, far more often in unofficial than in official ballot states, a proposition that seems unlikely.

¹³ By “race composition factor” is meant the types of races entering the Burnham index computations in any given state. Inspection of the data shows that this factor is fairly stable across elections though the index can in no way guarantee a complete control over this.

¹⁴ While the primary purpose of this test was to examine the effects of the race composition factor on across-state, within-year analyses, it was also used in the within-state, across-year analysis context, once again revealing an absence of correlation between the race composition factor and the predictions of ballot theory.

¹⁵ The three measures computed were based on the presidential-gubernatorial, presidential-congressional, and gubernatorial-congressional pairings of races. Ideally, they would have been the best measures to use in an analysis of ballot conditions across states within election years since they standardize race sets; however, there are several reasons why, in practice, they were not so useful. For one thing, they severely limit the number of cases entering the analysis. Even with the pairings mentioned above, one is assured of most states entering only the presidential-congressional computations since some of them do not contest the gubernatorial race in a presidential election year. With the addition of any third race to the index, more states would drop out of the computations for similar reasons. Second, the inclusion of but two races in an index for any given state does not allow much room for variation in the vote to occur. It certainly does not provide one with good estimates of a state's split ticket voting behavior. Third, such an index does not provide an adequate comparison with the results generated by the Burnham measure within states across election years. Comparability of the two modes of analysis is best achieved by using the same measure rather than two differently constructed ones. Also, it is more appropriate to use the Burnham measure consistently in order to compare our results with those produced by Burnham for his interpretation of the voting behavior of this period.

¹⁶ Burnham, of course, also observed a rising split ticket trend in these years. His evidence for this was the split ticket scores of five states. See Burnham, op. cit., pp. 13–14, 17–20.

¹⁷ In some election years, the split ticket distributions (as pictured on a standard low-to-high percentage value scale) were skewed to the right. Because of this, medians were computed to compare with the pattern of means. As one might guess, the results were similar in direction although some of the differences were smaller. Variance figures in this and other tables also behaved predictably in accordance with the pattern of mean values. Another way to look at this basic relationship is conveyed in Table 3 below, which presents the split ticket means for the ballot conditions for each state separately over all years, thereby holding constant the variation in the other states while viewing state X.

¹⁸ The voting estimates are “good” except for the fact that there is a rising split ticket curve in this period that makes sense substantively. Better estimates could be obtained by excluding the transition trends in the data.

¹⁹ There were two ways to vote a straight ticket in an unofficial ballot state—either deposit a given party strip in the ballot box unmarked or else mark it for all the candidates on it. Some states by law specified one way or the other, but some allowed both ways. In a party column state, one could vote a straight ticket by marking the entire ballot or by making a single mark if such an option was provided.

²⁰ The split ticket differences between the party column and party strip systems become even smaller if one contrasts the latter with only those party column states having emblem and “single mark” features. The effects of these two devices will be explored in later tables.

²¹ Two deviations occur in the data—for the years of 1890 and 1894. The deviation in 1890 is possibly due to the extremely small number of party column cases involved, 1890 being the first election in which this form of the Australian Ballot was used by any of the states. The deviation in 1894 is due to a separate party strip situation in one state. The differences between the two systems would have been larger and cleaner if one had contrasted pure party strip values with party column values; this tact was not taken in Figure 3 simply because two years had no values for the former condition.

²² Originally, such combinations were viewed within election years and the direction of mean split ticket values was basically as predicted. However, one quickly confronted cells with few cases since several ballot configurations appeared with rare frequency in reality. Given these conditions, a better strategy was to aggregate ballot cases across time to better observe their effects and the following analysis is based on this procedure.

²³ This interpretation of interactive effects in the party column—straight ticket case is suggested but not necessarily confirmed by the data at hand. As one can observe, the important offdiagonal cell entries in Table 6 have few cases and may not significantly depart from a value of 4.0, a condition indicating additive (separate) effects instead of interactive effects.

²⁴ A collapsing of some categories in Table 7 is needed to recover the points discussed for these two pairings of variables. However, this is a simple mental exercise and saves us from presenting two further bivariate tables at this time which would only convey the same information as Table 7.

²⁵ Comparison of the counterpart office bloc figures, 3.7 and 3.5, is unwarranted since the latter figure is based on so few cases. Also, the cells with 17 and 14 cases must be interpreted with caution, as footnote 23 earlier suggested.

²⁶ Burnham's fifth state, Oklahoma, is not relevant to this analysis since it entered the Union at the end of the time period covered in this study.

²⁷ The model used to compute these “expected” values is a simple additive one based on the formula, Y = X + Z. In the formula, X equals the state's pre-Australian split ticket effect, and Z refers to the net effect of other states with the same official ballot configuration as the state in question. The latter effect is computed by taking all such states for the election year (or year span) in question and subtracting their unofficial ballot mean from their official ballot mean. The type of election is held constant throughout the computations. An example for a state's first election year of official ballot use (year “A” in the table) will illustrate the procedure. The state of Ohio changed to a party column—straight ticket—emblem ballot in a presidential year. Therefore, its “X” value would be the mean split ticket value for presidential years in the unofficial ballot period. The “Z” value would find the same type of mean for all other states with that particular ballot configuration and subtract it from their mean value for the election year in which Ohio's ballot change occurred. Then, one merely adds the “X” and “Z” values together to get the “expected” split ticket value Y for Ohio in this particular election year.

²⁸ In an interesting piece of research, Jack Walker has shown that roll-off also varies according to the type of official ballot in use—the office bloc ballot contributing more to a roll-off effect than the party column ballot. Walker also views his findings as a possible modifying factor on Burnham's theory as the following quotation suggests: “The argument made by both Burnham and Stiefbold that increasing roll-off is an indication of increasing political alienation at least is weakened by our finding that changes in ballot forms directly affect the amount of roll-off” (p. 462). See Walker, Jack L., “Ballot Forms and Voter Fatigue: An Analysis of the Office Bloc and Party Column Ballots,” Midwest Journal of Political Science, 10 (11 1966), 448–463 CrossRefGoogle Scholar.