Economic geography and economic voting: Evidence from the US States

British Journal of Political Science

Volumn 36, Issue 3, 2006, Pages 527-547

  Ebeid, Michael ^a   Rodden, Jonathan ^b  

^a Boston University   (United States)

^b MASSACHUSETTS INSTITUTE OF TECHNOLOGY   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 33646857933     PISSN: 00071234     EISSN: 14692112     Source Type: Journal    
DOI: 10.1017/S0007123406000275     Document Type: Article

References (53)

1. For example, Gerald H. Kramer, 'Short-Term Fluctuations in U.S. Voting Behavior', American Political Science Review, 65 (1971), 131-43;
2. Ray Fair, 'The Effect of Economic Events on Votes for President', Review of Economics and Statistics, 60 (1978), 159-72;
3. Edward Tufte, Political Control of the Economy (Princeton, N.J.: Princeton University Press, 1978);
4. Steve J. Rosenstone, Forecasting Presidential Elections (New Haven, Conn.: Yale University Press, 1983);
5. Douglas A. Hibbs, The American Political Economy (Cambridge, Mass.: Harvard University Press, 1987);
6. Robert S. Erikson, 'Economic Conditions and the Presidential Vote', American Political Science Review, 83 (1989), 567-73;
7. Thomas M. Holbrook, 'Presidential Elections in Space and Time', American Journal of Political Science, 35 (1991), 91-109.
8. Patrick J. Kenney, 'The Effects of State Economic Conditions on the Vote for Governor', Social Science Quarterly, 64 (1983), 154-62;
9. Samuel Peltzman, 'Economic Conditions and Gubernatorial Elections', AEA Papers and Proceedings, 1 (1987), 293-7;
10. John E. Chubb, 'Institutions, the Economy, and the Dynamics of State Elections', American Political Science Review, 82 (1988), 133-54.
11. José Antonio Cheibub and Adam Przeworski, 'Democracy, Elections, and Accountability for Economic Outcomes', in Adam Przeworski, Susan Stokes and Bernard Manin, eds, Democracy, Accountability, and Representation (Cambridge: Cambridge University Press, 1999), p. 230.
12. G. Bingham Powell Jr and Guy D. Whitten, 'A Cross-National Analysis of Economic Voting: Taking Account of the Political Contest', American Journal of Political Science, 37 (1993), 391-414, p. 410;
13. Christopher Anderson, 'Economic Voting and Political Context: A Comparative Perspective', Electoral Studies, 19 (2000), 151-71, provides a review of similar findings, along with further evidence of the relationship between 'clarity of responsibility' and economic voting.
14. Kevin M. Leyden and Stephen A. Borrelli, 'The Effect of State Economic Conditions on Gubernatorial Elections: Does Unified Government Make A Difference?' Political Research Quarterly, 48 (1995), 275-90;
15. Michael Lewis-Beck and Richard Nadeau, 'French Electoral Institutions and the Economic Vote', Electoral Studies, 19 (2000), 171-3.
16. Business cycles can make the processing of such information more complicated, but voters might deal with this by simply expecting greater economic growth during good times and slower growth during bad times. Bernard Manin, Adam Przeworski and Susan Stokes, 'Elections and Representation', in Przeworski, Stokes, and Manin, eds, Democracy, Accountability, and Representation.
17. Jørn Rattsø, 'Vertical Fiscal Imbalance and Fiscal Behaviorin a Welfare State: Norway', in Jonathan Rodden, Gunnar Eskeland and Jennie Litvack, eds, Decentralization and the Challenge of Hard Constraints (Cambridge, Mass.: MIT Press, 2003).
18. Susanne Lohmann, David Brady and Douglas Rivers, 'Party Identification, Retrospective Voting, and Moderating Elections in a Federal System: West Germany, 1961-1989', Comparative Political Studies, 30 (1997), 420-49. German voters also have incentives to view Land elections through a federal lens because Land elections determine the make-up of the upper house of parliament.
19. Kenney, 'The Effects of State Economic Conditions on the Vote for Governor';
20. Peltzman, 'Economic Conditions and Gubernatorial Elections' ;
21. Chubb, 'Institutions, the Economy, and the Dynamics of State Elections'.
22. Using 1982 survey data on individual vote choice, Stein reaches the similar conclusion that voters generally hold the president (and by extension, the president's party) responsible for state economic conditions, and that governors are rewarded or penalized according to their partisan connection to the president; Robert M. Stein, 'Economic Voting for Governor and U.S. Senator: The Electoral Consequences of Federalism', Journal of Politics, 52 (1990), 29-53.
23. Peltzman, 'Economic Conditions and Gubernatorial Elections', p. 296.
24. For instance, Grady reports that in thirty-three of forty-six state-of-the-state addresses delivered in 1988, governors 'mentioned economic development as one of their top three priorities' (Dennis Grady, 'Managing the State Economy: The Governor's Role in Policymaking'. in Eric B. Herzik and Brent W. Brown, eds, Gubernatorial Leadership and State Policy (New York: Greenwood Press, 1990), pp. 105-20, at p. 106).
25. And while this type of credit claiming seems to have grown more prevalent in recent years, by no means is it an entirely new phenomenon. Southern economic historians have shown that aggressive investment promotion has been central to gubernatorial re-election strategies since at least the 1940s (James C. Cobb, The Selling of the South: The Southern Crusade for Industrial Development, 1936-1980 (Baton Rouge: Louisiana State University Press, 1982);
26. Gavin Wright, Old South, New South: Revolutions in the Southern Economy Since the Civil War (Baton Rouge: Louisiana State University Press, 1986)).
27. Leyden and Borrelli, 'The Effect of State Economic Conditions on Gubernatorial Elections'.
28. Timothy Besley and Ann Case, 'Incumbent Behavior: Vote-Seeking, Tax-Setting, and Yardstick Competition', American Economic Review, 85 (1995), 25-45.
29. Richard G. Niemi, Harold W. Stanley and Ronald J. Vogel, 'State Economies and State Taxes: Do Voters Hold Governors Accountable?' American Journal of Political Science, 39 (1995), 936-57;
30. Lonna Rae Atkeson and Randall W. Partin, 'Economic and Referendum Voting: A Comparison of Gubernatorial and Senatorial Elections', American Political Science Review, 89 (1995), 99-107.
31. See also Robert C. Lowry, James E. Alt and Karen E. Ferree, 'Fiscal Policy Outcomes and Electoral Accountability in American States', American Political Science Review, 92 (1998), 759-74.
32. Paul Krugman, Geography and Trade (Cambridge, Mass.: MIT Press, 1991);
33. Masahisa Fujita, Paul Krugman and Anthony Venables, The Spatial Economy: Cities, Regions, and International Trade (Cambridge, Mass.: MIT Press, 1999).
34. H.E. Perloff, E. Dunn and R. Muth, Regions, Resources, and Economic Growth (Baltimore, Md.: Johns Hopkins University Press, 1960).
35. State-level industry data is compiled by the Bureau of Economic Analysis (http://www.bea.doc.gov). The data we used come from the BEA's Series SA05.
36. For example, North Dakota's real per capita income fell by 15 per cent in 1980, a year in which the state suffered a severe drought.
37. Elizabeth Sanders, Roots of Reform: Farmers, Workers, and the American State 1877-1917 (Chicago: University of Chicago Press, 1999).
38. Both our model and the statistical technique we use to estimate it are similar to those used in Peltzman, 'Economic Conditions and Gubernatorial Elections'.
39. We do not claim to have included in our model every factor that might conceivably impact gubernatorial election outcomes. Among the possible factors that we do not evaluate are campaign spending, gubernatorial job performance ratings and presidential approval ratings. One reason to favour parsimony in model construction is that there remains much scholarly debate about whether - and to what extent - such factors do indeed make a difference in elections. For instance, Carsey and Wright find that presidential approval impacts gubernatorial election outcomes, while Atkeson and Partin do not (Thomas M. Carsey and Gerald C. Wright, 'State and National Factors in Senatorial and Gubernatorial Elections', American Journal of Political Science, 42 (1998), 994-1002;
40. Atkeson and Partin, 'Economic and Referendum Voting'). Practical data constraints also limit thorough consideration of certain factors as comprehensive data on such things as gubernatorial job performance simply are not available for all states in the complete time period covered in our study.
41. We define an election as 'competitive' if both the major parties received at least 20 per cent of the two-party vote. In practice, this requirement serves to eliminate from the dataset several elections of the mid-century 'solid south' period when Democratic candidates in many southern states routinely won by landslide margins.
42. We exclude Alaska and Hawaii. Also, the peculiarities of Louisiana's nonpartisan primaries and runoff elections make its electoral results difficult to compare to other states. The voting and incumbency data for 1950 to 1990 come from four ICPSR studies: State-Level Congressional, Gubernatorial and Senatorial Election Data for the United States, 1824-1972 (ICPSR 0075), State-Level Presidential Election Returns, 1824-1972 (ICPSR 0019), General Election Data for the United States, 1950-1990 (ICPSR 0013), and Candidate and Constituency Statistics of Elections in the United States, 1788-1990 (ICPSR 7757). Election and incumbency data for the 1990s were obtained from America Votes. Income data come from the US Department of Commerce's Bureau of Economic Analysis (available online at http://www.bea.doc.gov). Unemployment data are from the US Department of Labor's Bureau of Labor Statistics (available online at http://www.bls.gov for years since 1977, and for earlier years in the Statistical Abstract of the United States).
43. We are able to estimate the impact of presidential coattails more precisely than the midterm punishment effect. For the midterm punishment effect, we simply include an 'inverse' dummy variable to denote years in which a gubernatorial election occurred during the middle of a president's term (see Table 1 for the precise coding rule). For the coattails effect, we use the percentage of the two-party presidential vote received by the incumbent gubernatorial party's presidential candidate to measure the 'length' of the presidential candidate's coattails: the higher that percentage, the longer we expect the coattails to be - i.e., the more we expect the gubernatorial candidate to benefit from the strength of his party's presidential candidate. However, to assess the coattails effect accurately, we also must include a dummy variable that indicates whether a gubernatorial election was held in a presidential election year. Without this dummy variable we could not distinguish a zero value for the percentage of the two-party presidential vote received by the incumbent gubernatorial party's presidential candidate as meaning 'the incumbent gubernatorial party's presidential candidate received no votes' from meaning 'there was no presidential election that year'. Thus, presidential coattails are measured by the combined effects of the presidential election year dummy variable and the percentage of the two-party presidential vote received by the incumbent gubernatorial party's candidate.
44. There are a variety of ways to assess economic conditions using real per capita income or unemployment. One might consider, for instance, the absolute level of RPCI (or unemployment), the first difference, or the percentage change over a specified period. Perhaps the most common way - both in the relevant academic literature and the popular press - to operationalize these is with RPCI as a percentage change (one-year growth rate) and unemployment as an absolute level or first difference. However, for the sake of consistency and ease of interpretation, we report results using a percentage change operationalization (i.e., one-year growth rates) for both RPCI and unemployment. Note that we also estimated our models using levels and first differences and obtained similar results. Furthermore, we explored other economic indicators, such as the growth of real gross state product per capita, which is available since 1977. We do not report these results, which are very similar to the results reported below for real per capita income, because of the shorter time series and the fact that real gross state product per capita and real per capita income are correlated at above 0.9.
45. In the 1970s, the Bureau of Labor Studies changed the technique it uses to measure unemployment. Due to concerns about the comparability of unemployment rates before and after this change, we also estimated our model using only post-1977 data. The results, which are available upon request, are very similar to those reported below.
46. In all specifications, estimation of the model without an intercept (i.e., the regression line is forced through the origin) produces a coefficient of at least 0.90 for the normal vote. This suggests that a candidate for the incumbent party can expect to receive at least 90 per cent of the party's normal vote.
47. Note, however, that the coefficients for the presidential coattails variables are statistically significant only in the two version A specifications. The version B specifications are estimated with fewer observations over a shorter time period. This is a time period in which most states ceased holding gubernatorial elections in 'on-presidential' years. Thus, we can infer that, as a class of events, gubernatorial elections are now less susceptible to presidential coattails because most are held in non-presidential election years. Furthermore, we explored the possibility that presidential popularity might impact gubernatorial elections more generally than merely through the vote for president. To do so, we entered in our model a variable that interacts yearly national presidential approval data with an inverse dummy variable that indicates whether or not the incumbent governor was a co-partisan of the president. In no specification did the coefficient for this interaction variable approach statistical significance, so we did not include it in the final model specifications. This result provided little surprise considering that, as discussed above, previous researchers have reached markedly different conclusions about whether presidential popularity impacts gubernatorial elections.
48. As discussed above, we estimate presidential coattails through the combined effect of the presidential election year dummy variable and the presidential vote share variable. In this hypothetical example where the gubernatorial candidate's presidential candidate receives 55 per cent of the vote, the calculation from specification 1, version A of the impact on the gubernatorial candidate's vote is as follows: (55 × 0.18) + (1 × -8.61) = 1.29.
49. We also add to each specification a separate variable for the (un-interacted) primary product index. Although there is no reason to expect the primary product index to have a direct impact on incumbent vote share (i.e., there is no reason to expect voters in states where the primary product is low - or high - to be more or less likely to vote for an incumbent), it could nonetheless be correlated with incumbent vote share. That is to say, over time incumbent vote share and the primary product index might vary in unison for unknown or theoretically unclear reasons. We can control for this possibility - and ensure that our estimates of the interaction of the primary product index and state economic conditions are not contaminated by the possibility - by including an independent variable for the primary product index. However, because we do not expect the primary product index to have a direct impact on incumbent vote share, we also estimated all specifications reported in Table 3 without an independent variable for the primary product index. These results, which are available from the authors upon request, are nearly identical to those in Table 3 (i.e., including or excluding the independent variable for the primary product index does not have an appreciable impact on the estimates of the other variables).
50. Note that since the unemployment data are only available since 1968, the sample range displayed in Figure 3 for the primary product index has a different maximum from that seen in Figure 2.
51. As mentioned above the sample range for the primary product index is 0.78 to 54.15. The sample mean is 11.55 and the sample standard deviation is 10.65. Thus, a primary product index 1 standard deviation below its mean is 0.9, which is consistent with our intuitive sense that a primary product index of 1 or lower denotes very minimal reliance on primary products.
52. The sample range for the relative percentage increase in unemployment rate is -34.69 to 60.19, with mean of 0.52 and standard deviation of 13.86. For unemployment our definition of 'very good' economic performance is one standard deviation below the mean (because it is preferable, of course, that unemployment be lower rather than higher). Thus, the calculation for the hypothetical example is (-0.11 × - 13.34) + (0.01 × -13.34 × 1) = 1.33.
53. Wolfers presents a similar finding, which he asserts is evidence that voters rationally collect and process information about the economy. He goes on to show, however, that voters nonetheless make systematic attribution errors, for instance when they improperly blame or reward incumbents in the aftermath of oil shocks (Justin Wolfers, 'Are Voters Rational? Evidence from Gubernatorial Elections' (Stanford Graduate School of Business Research Paper No. 1730, 2002)).