Working Paper, Brown University, Department of Economics 2001-26
Contrary to widespread belief, voting machines of older types, such as lever and punchcard systems, are not used in counties with lower income - and newer machines, such as optical scanners and electronic machines, are not used in richer counties. We provide an economic explanation for this and other regularities of voting equipment usage in the U.S. We present a model in which, all other things being equal, a) the adoption of a new technology is more likely in richer and larger counties; b) the adoption of a new technology is less likely the more advanced is the technology already adopted in the county. We argue that the net benefits from adopting the more advanced optical or electronic machines after 1980 were not high enough to induce a technological upgrade in those (relatively richer and larger) counties that had adopted punchcard machines in previous decades. By contrast, net benefits from newer technologies were high enough to induce their adoption in relatively poorer and smaller counties that had not yet mechanized or computerized their voting system. Estimates of historical determinants of voting equipment choice support our hypothesis. In particular, the probability of using punchcard machines in the 1990s is positively related to a county's income in the 1960s, when punchcard machines were first introduced. When the effect of past income is controlled for, the effect of more recent levels of income on the probability of using punchcard machines becomes negative.