The ‘Home Grown’ Presidency: empirical evidence on localism in presidential voting, 1972–2000

Abstract

This builds upon the conceptual framework of Lewis-Beck and Rice (American Journal of Political Science, 27, 548–56, 1983), in combination with the empirical design of Kjar and Laband (Public Choice, 112, 143–50, 2002), to investigate home grown-ness in US presidential elections from 1972–2000. It found that, ceteris paribus, home state vote shares for US Presidential election winners are 5.19–15.11 percentage points higher due to the home grown-ness effect. In the eight presidential elections analysed, this study confirms two aspects of prior work. First, the estimate of a home grown-ness effect in presidential elections of 5.19 percentage points (on average), supports the 4 percentage point average found by Lewis-Beck and Rice (1983). Second, that support for the winning president monotonically increases as moves are made away from the opponent's home territory confirms the cascading dummy variable series approach developed by Kjar and Laband (2002).

Acknowledgements

The authors thank an anonymous referee, David Laband, Kamal Upadhyaya and Troy Gibson for helpful comments. This work was supported through a BAC Faculty Scholars Grant. Any remaining errors are our own.

Notes

For an excellent discussion of the ‘nature of political services’ (i.e., as search or experience goods), see Crain and Goff (1988).

For an examination of the high costs of removing incumbent politicians from office, see Mixon (2000). Related literature on the Klein–Leffler performance bond includes Telser (1980) and Shapiro (1983).

Kjar and Laband (2002, p. 144–45) note that there are numerous studies examining home grown-ness in politics at the non-presidential level. These are Key (1950), Black and Black (1973), Johnston (1974), Tatalovich (1975), Rice and Macht (1987) and Thielmann (1993). See Kjar and Laband for a detailed discussion of each of these.

The dependent variable is modelled after that in Kjar and Laband (2002). They examine the race between two US House candidates (Turnham and Riley) using a dependent variable that measures the vote share of one of the two candidates (Turnham). Here, the winning candidate's vote share (out of the two major party candidates) for each presidential election from 1972 to 2000.

The log of PREZ was also regressed on the 12 regressors in version (2) of Table 2, given that a variable (PREZ) that is naturally in the unit interval (a share) is being explained. This model produces identical results to those in Table 2 in terms of parameter signs and significance levels.