In the aftermath of the 2000 Presidential election and the disputed vote in Florida, controversy arose over the previously obscure issue of differences in voting equipment across jurisdictions. The American public became acquainted with the potential for punch card voting mechanisms to produce large numbers of invalidated ballots. A widespread perception emerged among politicians and in the news media that the use of punch cards, and of antiquated voting machinery more generally, was more common in counties with a greater percentage of minorities and poor people. A series of editorials and op-ed articles in the Washington Post claimed that "it is mainly affluent counties that have switched" from punch cards to more modern equipment while "poor and minority voters tend to be stuck with less accurate machines," that African Americans "were far more likely to be stuck with the lousy machines than were affluent whites," and that "the most error-prone machines tend to be in the poorest counties."
This conventional wisdom that emerged so rapidly in late 2000 was superficially plausible for two reasons. First, the proportion of ballots for which no valid presidential choice was registered was much higher in areas heavily populated by minorities and the poor than elsewhere. Second, income and ethnicity are often strongly related to the quality of other public services, such as education. It seems reasonable to assume that where incomes and local tax revenues are low, election administration would be less well funded, and inferior voting technology — namely, punch card equipment — would still be in use.
There are several flaws in this logic, however. First, although there were several previous local elections in which incompletely punched-out chad produced controversial recounts, there was no universal consensus among election administrators that they produced the highest rates of voter error. The drawbacks of punch cards are now well known: voters often punch the holes in the wrong places, or apply insufficient force so that incompletely-removed chad results in undervotes; because no candidate information is printed directly on the cards, it is difficult for voters to discern mistakes by examining their cards. However, technical studies of voting equipment (e.g., Saltman (1988), have also documented problems with lever machines, optical scan systems, and direct electronic recording (DRE) systems (which use touch screens, push buttons or keyboards). In fact, empirical analyses find virtually no difference overall in the rates of invalidated ballots produced by DRE and punch card systems (Caltech/MIT, 2001; Knack and Kropf, 2002a).
Second, minimizing invalidated ballots is only one of several criteria by which election administrators have typically assessed the performance of voting technology. Although the media now revile any technology that appears to produce voter error, until after the 2000 election the media’s interest was only in producing quick vote totals on election night — and by that criterion, punch card systems performed far better than optical scan systems, particularly for large jurisdictions such as Los Angeles County.
Third, election administration, and voting equipment more specifically, represents only a tiny fraction of any county’s budget, regardless of the type of voting equipment it uses. (In most states, elections are administered by county governments.) Moreover, it had never been a salient issue, or a visible item in county budgets. In counties with high income levels, with strong tastes for education, public parks and other services, election administration may receive as little public attention and as little funding as in a poorer county. For example, the only county in Maryland still using punch cards in 2000 (Montgomery) has the wealthiest, best-educated population in the state, and spends more per capita on education, parks and other services than almost any county in the nation.
Historical and idiosyncratic factors appear to be influential determinants of voting equipment in use. In the early 1960s, hand-counted paper ballots were used in most small counties, while lever machines were used in large counties. As many of the smaller counties have grown, they purchased optical scanning equipment, which use ballots very similar to the hand-counted ballots. Many lever machine counties switched to punch cards beginning in 1964; others retained lever machines until DREs became available more recently, while others still use the lever machines.
Because so few people within county government know much about voting equipment, election administrators often have a high level of discretion in choosing equipment type, and their choices are likely to reflect their own experiences, and those of their colleagues in other counties. Some election administrators are likely to be highly risk averse, and unwilling to adopt a system not already in wide use elsewhere. Some of them may have budgets allowing them to attend conferences and learn what has worked well and what has failed in other counties.
To the extent that cost does matter, income per capita may not predict quality as well as county size. Volume discounts from vendors, and economies of scale in setting up new systems, favor larger counties. As noted in FEC (1982: 11):
New voting systems are, typically, first adopted by large metropolitan jurisdictions where the complexity of the ballots and the volume of voters create pressures for improved vote recording and tabulating techniques. Such jurisdictions are also blessed with the fiscal, technical, and managerial resources equal to the challenge. Only when new devices are tested and debugged in this way are they normally then adopted by intermediate-sized jurisdictions.
Minorities tend to live in larger counties, so may benefit from more modern voting equipment, to the degree that cost matters. Tennessee is an illustrative case. In 1998, fewer than one fifth of all the state’s counties had electronic voting systems. However, these included the three largest counties of Shelby (Memphis), Davidson (Nashville), and Knox, which account for a disproportionate share of the state’s poor, minorities, and Democratic voters. Shelby County alone is home to nearly one half of the state’s African Americans, but just over one tenth of its whites.
Statistical analyses reported in Knack and Kropf (2002b) overwhelmingly reject the conventional wisdom that minorities and the poor disproportionately reside in areas using punch card and other inferior voting equipment. They combined county-level data on voting equipment in use in 1998 with U.S. Census data on minority and poor populations, county size, income levels, and property tax revenues per capita. For the U.S. overall, black-white differences in punch card use were negligible: 31.9% for whites and 31.4% of African Americans live in counties using this voting technology. Hispanics were much more likely to live in punch card counties than either whites or blacks. However, this difference was entirely attributable to Los Angeles County, where nearly one in seven Hispanics in the country reside. (About 1 in 10 voters using punch card equipment in 1998 lived in L.A. County.) Whites (27.7%) were more likely than blacks (21.8%) to live in optical scanning counties, but blacks (37.8%) were much more likely than whites (26%) to live in counties using either of the technologies for which "overvoting" is impossible if machines are programmed correctly: DRE and lever machines.
In Florida, African Americans were only slightly more likely than whites to live in punch card counties. The notable difference again was for Hispanics, 84% of whom live in punch card counties, compared to just over 60% for whites and African Americans. This difference is entirely attributable to the use of punch card voting in Miami-Dade County, home of more than one half of Florida’s Hispanics, but fewer than one in seven whites and fewer than one in five African Americans.
Findings are similar when looking at voting equipment used by the poor (persons living below the poverty level) and non-poor. The poor are slightly more likely (33.4% compared to 31.8%) to live in punch card counties than the non-poor nationwide, but they are also slightly more likely (9.8% to 8.6%) to use the most modern technology (DRE).
These comparisons were also conducted on a state-by-state basis, for the 29 states in which some but not all counties used punch card technology. In 18 of the 29 states, whites were more likely than African Americans to live in punch card counties. Whites were more likely than Hispanics to live in punch card counties in 21 of the 29 states. In 21 states, the non-poor are more likely than the poor to reside in counties using punch card voting equipment.
Other results in Knack and Kropf (2002b) provide little evidence for the importance of "affordability," as measured by county size, per capita income, or property tax revenues per capita. Punch card counties in Florida are much larger, wealthier, and more revenue-rich than counties using any other type of voting equipment. Similarly, for the U.S. as a whole, punch card counties are larger and wealthier on average than those using any other voting system. Counties using electronic voting constitute the group with the lowest incomes on average, and — by a wide margin — the lowest property tax revenues per capita.
Similar findings are produced by state-by-state comparisons across counties for the 28 states in which some counties use punch cards while others use modern (optical scanning or electronic voting) equipment. In 17 of the 28 states, punch card counties tend to be larger than counties with modern equipment. In 17 (but not the same 17 as in the case of county size) of the 28 states, punch card counties tend to have higher average incomes. Similarly, in 17 of the 28 states, punch card counties on average had higher property tax revenues per capita. Florida fits these general patterns. Population, income and tax revenues were all significantly higher in its 15 counties using punch cards in 1998 than in its 24 optical scan counties (electronic voting had not yet been approved for use in Florida).
Higher rates of invalidated ballots in heavily-minority areas cannot, therefore, be attributed to a greater likelihood that minorities must vote using punch card equipment. Controlling for the type of voting equipment in use, invalidated ballots remain higher in counties with more African Americans, Hispanics, and poor persons (Knack and Kropf, 2002b). However, although voting equipment is distributed in a race-neutral way, it may have effects that are not race-neutral. Several studies have found, using both precinct-level and county-level data, that the association between invalidated ballots and African American population is particularly strong where punch cards are used, but disappears where voting technology is used that can be programmed to prevent overvoting (Herron and Sekhon, 2001; Kimball et al., 2001; Tomz and Van Houweling, 2001; Knack and Kropf, 2002a).
The explanation for this pattern is not obvious. Survey data indicate that differences in deliberate undervoting between whites and blacks are fairly small, and there is little reason to expect these differences to vary much with the types of voting technology used. It is not the case that, within punch card counties, the oldest and most error-prone punch card devices are likely to be placed in the heavily-minority precincts: in most counties, the devices are stored in a central location between elections and are not earmarked for particular precincts, and they are assembled in a central location weeks prior to the election by a small number of elections administration staff (not by poll workers within the precinct). It is conceivable that African American voters are less likely than whites to request assistance from poll workers, or that poll workers in minority precincts are less able or willing to provide useful assistance. Although these studies control for educational level and voter experience to the extent allowed by the available data, they cannot control effectively for differences in education quality (as distinct from years of schooling completed), or for experience in undertaking unfamiliar administrative and bureaucratic tasks. In the absence of data to test such ideas, they must remain purely conjectural.
