¹ Key, V. O. Jr., Politics, Parties, and Pressure Groups, 5th ed. (New York: Thomas Y. Crowell, 1964), pp. 567–568 Google Scholar.

² Stokes, Donald E. and Miller, Warren E., “Party Government and the Saliency of Congress,“ Public Opinion Quarterly, 26 (Winter, 1962), 531–546 CrossRefGoogle Scholar.

³ Campbell, Angus, “Voters and Elections: Past and Present,” Journal of Politics, 26 (11, 1964), 745–757 CrossRefGoogle Scholar; Campbell, Angus, “Surge and Decline: A Study of Electoral Change,” Public Opinion Quarterly, 24 (Fall, 1960), 397–418 CrossRefGoogle Scholar; and Key, pp. 568–569.

⁴ Hinckley, Barbara, “Interpreting House Midterm Elections: Toward a Measurement of the In-Party's ‘Expected’ Loss of Seats,” American Political Science Review, 61 (09, 1967), 700 CrossRefGoogle Scholar.

⁵ Franklin, Mark N., “A ‘Non-election’ in America? Predicting the Results of the 1970 Mid-term Election for the U.S. House of Representatives,” British Journal of Political Science, 1 (10, 1971), 508–513 CrossRefGoogle Scholar. See also King, Anthony, “Why All Governments Lose By-Elections,” New Society, 03 21, 1968, pp. 413–415 Google Scholar; Lawson, Nigel, “A New Theory of By-Elections,” Spectator, 11, 1968, pp. 651–652 Google Scholar; and Lees, John D., “Campaigns and Parties—The 1970 American Mid-Term Elections and Beyond,” Parliamentary Affairs, 24 (Autumn, 1971), 312–320 Google Scholar. The view that midterms represent, in large measure, the electoral swing of the pendulum (or electoral surge and decline) does not seem to be held by politicians. Sam Kernell has compiled convincing evidence that a central premise among American politicians is that “the president's popularity directly affects his congressional party candidates' chances for election” in midterms: Sam Kernell, “Presidential Popularity and Negative Voting: An Alternative Explanation of the Mid-Term Electoral Decline of the President's Party,” paper delivered at the 1974 annual meeting of the American Political Science Association. See also the variety of interpretations of the outcome of British by-elections in Cook, Chris and Ramsden, John, eds., By-Elections in British Politics (New York: St. Martin's Press, 1973)CrossRefGoogle Scholar. Alternative midterm models are discussed in Boyd, Richard W. and Murphy, James T., “How Many Seats Will the Republicans Lose? Changes in the House: A Prediction,” New Republic, 10 24, 1970, pp. 12–14 Google Scholar.

⁶ Key, p. 569.

⁷ This conventional description of the midterm electorate has been challenged by the evidence of Robert B. Arseneau and Raymond E. Wolfinger, “Voting Behavior in Congressional Elections,” paper de-livered at the 1973 annual meeting of the American Political Science Association.

⁸ Arseneau and Wolfinger, p. 3. Several other recent studies have greatly reinforced the evidence linking evaluations of the incumbent president to electoral choices in congressional races. See Kernell, “Presidential Popularity and Negative Voting,” for an extensive analysis of Gallup poll data in six midterm elections; and Piereson, James E., “Presidential Popularity and Midterm Voting at Different Electoral Levels,” American Journal of Political Science, forthcoming, 11, 1975 Google Scholar, which uses data from the 1970 election study conducted by the Center for Political Studies at the University of Michigan.

⁹ Kramer, Gerald H., “Short-Term Fluctuations in U.S. Voting Behavior, 1896–1964,” American Political Science Review, 65 (03, 1971), 131–143 CrossRefGoogle Scholar. A data error in this paper is corrected in its Bobbs-Merrill reprint (PS-498); see also Saul Goodman and Gerald H. Kramer, “Commentary on Arcelus and Meltzer, ‘The Effect of Aggregate Economic Conditions on Congressional Elections’,” American Political Science Review, forthcoming; Kramer, Gerald H. and Lepper, Susan J., “Congressional Elections,” in The Dimensions of Quantitative Research in History, ed. Aydelotte, William O., Bogue, Allan G., and Fogel, Robert William (Princeton: Princeton University Press, 1972), 256–284 Google Scholar; and Lepper, Susan J., “Voting Behavior and Aggregate Policy Targets,” Public Choice, 18 (Summer, 1974), 67–81 Google Scholar.

¹⁰ Stigler, George J., “General Economic Conditions and National Elections,” American Economic Review, 63 (05, 1973), 160–167 Google Scholar; and further discussion by Paul W. McCracken, Arthur M. Okun, and others, pp. 169–180. Stigler's method might best be described as a “most squares” technique: find the specification that maximizes the error variance. But the outstanding work in discovering the pessimum most squares model is Francisco Arcelus and Allen H. Meltzer, “The Effect of Aggregate Economic Conditions on Congressional Elections,” American Political Science Review, forthcoming; see the reply of Goodman and Kramer.

¹¹ On the translation of votes into seats, see Tufte, Edward R., “The Relationship Between Seats and Votes in Two-Party Systems,” American Political Science Review, 67 (06, 1973), 540–554 CrossRefGoogle Scholar.

¹² Hinckley, “Interpreting House Midterm Elections;” Zeidenstein, Harvey, “Measuring Congressional Seat Losses in Mid-Term Elections,” Journal of Politics, 34 (02, 1972), 272–276 CrossRefGoogle Scholar; and Taylor, A. H., “The Proportional Decline Hypothesis in English Elections,” Journal of the Royal Statistical Society, Series A, 135 (1972), 365–369 CrossRefGoogle Scholar.

¹³ See the normalizations in Flanigan, William H. and Zingale, Nancy H., “The Measurement of Electoral Change,” Political Methodology, 1 (Summer, 1974), 49–82 Google Scholar; also William H. Flanigan and Nancy H. Zingale, “Electoral Competition and Partisan Realignment,” paper delivered at the 1973 annual meeting of the American Political Science Association. The standard discussion is, of course, Converse, Philip E., “The Concept of a Normal Vote,” in Campbell, Angus et al., Elections and the Political Order (New York: Wiley, 1966), pp. 9–39 Google Scholar.

¹⁴ The Democratic loss in the 1938 midterm election is measured relative to the share of votes received by Democratic congressional candidates in the previous three elections (1932, 1934, and 1936), rather than the eight used for estimating the normal vote in the other midterms. The eight election normalization fails for 1938 because of the rapid and extensive realignment from 1930 to 1932; thus the averaged results of congressional elections from 1932 on gives a more reasonable estimate of the 1938 normal vote than the inclusion of several Republican dominated years prior to the realignment. The percentage share of the congressional vote received by the Democrats during that period shows the problem:

From this series, its seems clear that a reasonable estimate of the normal Democratic vote for the 1938 election should be based on the elections of 1932, 1934, and 1936, rather than earlier years.

¹⁵ See Kabaker, Harvey M., “Estimating the Normal Vote in Congressional Elections,” Midwest Journal of Political Science, 13 (02, 1969), 58–83 CrossRefGoogle Scholar.

¹⁶ The presidential approval ratings from 1946 to 1970 are from The Gallup Opinion Index, 64 (10, 1970), 16 Google Scholar; the 1938 approval rate, 57 per cent, was averaged (because of inconsistencies in question wording and survey dates) from two surveys: September, 1938—“Are you for or against Roosevelt today?” 55.2 per cent; and October, 1938—“In general do you approve or disapprove of Roosevelt as President?” 59.6 percent. The source for the 1938 data is Gallup, George, The Gallup Poll (New York: Random House, 1972), pp. 118, 122 Google ScholarPubMed. Similar, but not identical figures are reported in a fine study by Clark, Wesley C., “Economic Aspects of a President's Popularity” (Ph.D. dissertation, University of Pennsylvania, 1943), p. 47 Google Scholar, which also contains an extensive discussion of the early years of the series. A flawed analysis of the factors affecting the ratings is given in Mueller, John E., “Presidential Popularity from Truman to Johnson,” American Political Science Review, 64 (03, 1970), 18–34 CrossRefGoogle Scholar; and in Mueller, John E., War, Presidents and Public Opinion (New York: Wiley, 1973)Google Scholar. The substantive and statistical difficulties in Mueller's analysis are discussed in Richard A. Brody and Benjamin I. Page, “The Impact of Events on Presidential Popularity: The Johnson and Nixon Administrations,” paper delivered at the 1972 annual meeting of the American Political Science Association; and in Hibbs, Douglas A. Jr., “Problems of Statistical Estimation and Causal Inference in Time-Series Regression Models,” in Sociological Methodology, 1973–1974, ed. Costner, Herbert (San Francisco: Jossey-Bass, 1974), pp. 252–308 Google Scholar.

¹⁷ The details of the shifting wording are in Clark, , Economic Aspects of a President's Popularity, pp. 47, 55–60 Google Scholar.

¹⁸ Stigler's article itself, as well as the commentary, makes clear the lack of theoretical specificity found in current models correlating economic time series to electoral outcomes; see Stigler, , “General Economic Conditions and National Elections,” pp. 160–167 Google Scholar; and further comments, 168–180. Stigler uses a two-year difference to compute the change in economic conditions for his model, on the view that voters compare the change in the economy in the current election year with that at the time of the previous election two years before. That model seems doubtful, attributing an excessively long time perspective to voters, especially for moderate economic changes. Another way to look at the matter is to consider the voter's problem as the generation of an estimate of the rate of real economic change immediately prior to the election—in order to estimate what can be expected with respect to the performance of the economy under the incumbent administration. If those expectations are good, the party of the incumbent President receives a vote; otherwise the out-party gets the vote. If this is a reasonable description of the problem facing the voter, then the voter would probably not use Stigler's method—a two-year difference in economic changes—to estimate expected or immediately past economic performance. A one-year difference is at least slightly more realistic. Additional progress on this question can probably be made by examining survey evidence at the individual level rather than by still more specifications of aggregate models.

¹⁹ Kramer, p. 139; see also Goodman and Kramer.

²⁰ The yearly change in real disposable personal income per capita (in 1958 dollars) was computed from data in The Annual Report of the Council of Economic Advisers, 1973 (Washington, D.C.: U.S. Government Printing Office, 1973), p. 213 Google Scholar; and The Annual Report of the Council of Economic Advisers, 1971, p. 215 Google Scholar.

²¹ The midterm of 1942 is omitted from the analysis because of the special effect of wartime controls on the economy and of wartime conditions on evaluations of the incumbent President. Kramer also dropped wartime years; see Kramer, p. 137.

²² In regressions of this sort, involving such a small number of degrees of freedom, some prefer to use a corrected R ² that takes into account the loss in degrees of freedom as the coefficients are estimated. In our case, the corrected R ² is 0.88. See Christ, Carl F., Econometric Models and Methods (New York: Wiley, 1966), pp. 509–510 Google Scholar.

²³ Kramer, p. 141.

²⁴ Kernell, p. 32.

²⁵ The discarding of observations one at a time coupled with re-estimation is the first step in producing a “jackknife” estimate of a complex statistic along with a confidence interval; see Mosteller, Frederick and Tukey, John W., “Data Analysis, Including Statistics,” in Lindzey, Gardner and Aronson, Elliot, eds., The Handbook of Social Psychology, 2nd ed. (Reading, Massachusetts: Addison-Wesley, 1968), pp. 133–160 Google Scholar; and Miller, Rupert G. Jr., “The Jackknife—A Review,” Technical Report No. 50 (08 28, 1973), Department of Statistics, Stanford University, Stanford, California Google Scholar.

²⁶ Polls reported in Gallup, The Gallup Poll.

²⁷ “Gallup Says Poll Shows Ford Popularity on Rise,” The New York Times, October 24, 1974.

²⁸ “Poll Confirms GOP Fears,” The Washington Post, 11 4, 1974, p. A5 Google ScholarPubMed.

²⁹ Vote as reported in The Gallup Opinion Index, 118 (04, 1975), p. 27 Google Scholar.

³⁰ Data sources: Gallup approval rating for October, 1974 from “Gallup Says Poll Shows Ford Popularity on Rise,” The New York Times, October 24, 1974; other months from The Gallup Opinion Index, 103 (01, 1974), 3 Google Scholar; 108 (June, 1974), 1; and 111 (September, 1974), 12. The change in real disposable income per capita is available only by quarters and the monthly values are interpolated; it should also be noted that the quarterly figures are quite unstable, with provisional and final estimates often differing substantially. The computations are based on data in Bureau of Economic Analysis, Business Conditions Digest, 01, 1975 (Washington, D.C.: U.S. Government Printing Office, 1975), p. 69 Google Scholar. The reports of polls asking people how they intended to vote in the 1974 congressional election are from “Poll Confirms GOP Fears,” The Washington Post, 11 4, 1974, p. A–5 Google ScholarPubMed; “Poll: Democrats Will Sweep Into the House,” New York Post, 08 8, 1974, p. 24 Google ScholarPubMed; and The Gallup Opinion Index, 110 (08. 1974), 1–4 Google Scholar. Nonresponses have been divided equally between the two parties.

³¹ The model appears to over-respond to very large short-run improvements in economic conditions. Such improvements have been more typical of on-years than off-years. The model performs acceptably for ups and downs of less than $150 per capita in real disposable income, which has been the case for all midterms since 1938. In adition, the model might be examined comparatively—for example, in British by-elections. Finally, the model could be extended for on-year elections (although Gallup did not ask the presidential approval question for incumbent presidents in the six months prior to on-year elections until the 1960s). On-year elections are clearly a much more complex problem; and the midterm model is best seen as an alternative to “surge and decline” and the similar approaches cited in footnotes 1–5.

³² This technique is described in Tufte, , “Seats and Votes,” pp. 549–551 Google Scholar.

³³ The causes of recent increases in congressional tenure are not yet clear; see Mayhew, David R., “Congressional Elections: The Case of the Vanishing Marginals,” Polity, 6 (Spring, 1974), 295–317 CrossRefGoogle Scholar; Walter Dean Burnham, “Communication,” and Tufte, Edward R., “Communication,” American Political Science Review, 68 (03, 1974), 207–213 Google Scholar; and Erikson, Robert S., “Malapportionment, Gerrymandering, and Party Fortunes in Congressional Elections,” American Political Science Review, 66 (12, 1972), 1234–1255 CrossRefGoogle Scholar. Some consequences of seat changes are described in Brady, David W. and Lynn, Naomi B., “Switched-Seat Congressional Districts: Their Effect on Party Voting and Public Policy,” American Journal of Political Science, 17 (08, 1973), 528–543 CrossRefGoogle Scholar.

³⁴ Key, p. 568.

³⁵ Mueller, , “Presidential Popularity,” p. 25 Google Scholar.

³⁶ Tufte, Edward R., “The Political Manipulation of the Economy: Influence of the Electoral Cycle on Macroeconomic Performance and Policy,” manuscript, Princeton, 1974 Google Scholar.

³⁷ Stokes and Miller, pp. 544–546.

³⁸ See Kelley, Stanley Jr. and Mirer, Thad, “The Simple Act of Voting,” American Political Science Review, 68 (06, 1974), 572–591 CrossRefGoogle Scholar.