¹ Andrew Hacker suggests the “common sense” reasons in his article, “Does a ‘Divisive’ Primary Harm a Candidate's Election Chances?” American Political Science Review, 59, 1 (March, 1965), 105–110 CrossRefGoogle Scholar. The hypothesis that a divisive primary is likely to cause a loss of party workers in the Fall has been substantiated at the congressional level in Iowa by Johnson, Donald and Gibson, James, “The Divisive Primary Revisited: Party Activists in Iowa,” American Political Science Review, 68, 1 (March, 1974), 67–77 CrossRefGoogle Scholar.

² Hacker, p. 110.

³ Somewhat sadly, I say, because I am indebted to Professor Hacker for those hours in the classroom some years ago when he tried to instill in me the right mind-set to question his work. I half suspect he designed his article as a challenge and has waited for the past nine years for someone to take him up on it.

⁴ It seems clear that “divisive” (Hacker's quotation marks) is an unclear, emotion-laden term. Perhaps “close” would be better – certainly it would more clearly imply the operational definition used here. However, “divisive” is the prevalent term in the literature, and to tie in with that literature, “divisive” is used here as well.

⁵ I have not included in this analysis a few races in which a single candidate was unopposed; in which a third candidate received more than 20 per cent of the vote; or in which a candidate died after the primary, but before the general election. Also excluded are nonpartisan special elections.

⁶ This compares with the 35 per cent that Hacker found using the broader definition.

⁷ Hacker, p. 110.

⁸ The difference in winning percentage between incumbents who overcame divisive primaries and incumbents who faced challengers who won divisive primaries is statistically significant at the .05 level.

⁹ This difference is not statistically significant at the .05 level.

¹⁰ With some minor adjustments this is David's, Paul “composite B” in his Party Strength in the United States, 1872–1970 (Charlottesville: University Press of Virginia, 1972), Part IIGoogle Scholar. Biennial data for 1970 was calculated where missing, and estimated for 1972, where it would influence classification. David interpolated where necessary to obtain a biennial data-series for each type of race, then he simply averaged the three percentages.

¹¹ This classification differs somewhat from the Hacker classification, but the results of the later analysis are essentially the same using either classification scheme.

¹² This is essentially the position of Key, V. O. Jr. in his Politics, Parties and Pressure Groups, 5 th ed. (New York: Thomas Y. Crowell, 1964), pp. 437–439 Google Scholar. Hacker also reports an overall 2:1 ratio of primaries in “friendly” states to primaries in “hostile” states.

¹³ “Significant” here refers to statistically significant differences in winning percentage (at the .05 level) between those who had divisive primaries and those, of the same type, who did not.

¹⁴ This difference is not statistically significant.

¹⁵ The differences among the winning percentages in Table 5 are significant at the .05 level.

¹⁶ I am indebted to James L. Payne for this succinct statement of the flaw in the original analysis.

¹⁷ Hacker data based on Tables I, II, and VII, plus the discussion on p. 107. Because the Hacker study frequently combines gubernatorial and senatorial races, it is impossible to make other comparisons for just the Senate races. It would appear that divisive primaries have also hurt somewhat in gubernatorial races. For example, the combined figures show incumbents burdened with divisive primary winning just 60 per cent of the general elections; incumbents on an “equal footing” regarding primary experiences with their opponents winning 72 per cent of the time; and incumbents helped along by the challenger having a primary fight winning 82 per cent of the races.

¹⁸ The twelve are: Stevens, Fannin, Guiney, Cook, Beall, Jr., Brooke, McIntyre, Domenici, Helms, Thurmond (in his first race as a Republican), Baker, and Brock, Mathias and Weicker received help from Mahoney and Dodd, respectively, one-time Democrats running as independents.

¹⁹ Gravel, Dole, Metcalf, Hruska, and Abourezk. The difference in percentages between the weaker and stronger party representatives is statistically significant at the .01 level.