Michael Lewis' 2003 book Moneyball: The Art of Winning an Unfair Game tells the story of how the Oakland A's general manager Billy Beane built successful baseball teams year after year, while spending far less money than his competitors. The secret to the team's success? Rather than relying on the gut instincts of old-time scouts, as was standard practice for decades, the A's adopted a research-driven approach that relied heavily on empirical analysis of player performance. This allowed them to look behind the conventional wisdom and assess what practices were most likely to lead to success on the field. When Beane and the A's carefully scrutinized the numbers, it turned out that many of the intuitions upon which their hard-bitten, tobacco-spitting scouts had relied for years were just plain wrong. By adopting a research-driven approach to baseball, Beane was able to build a franchise that has competed for pennants year after year, while spending a fraction of what other teams spend.
With the baseball playoffs in full swing, it seems like an especially appropriate time to compare recent developments in the world of elections to the phenomenon that Moneyball describes. To an unfortunate extent, efforts at election reform have been based an intuition-based approach resembling that of the old-time scouts in Lewis' book. That approach places too much weight on seat-of-the-pants assessments of what makes for good elections, neglecting serious empirical research into what works and what doesn't in the real world. In calling for a "Moneyball Approach to Election Reform," I mean to suggest a research-driven inquiry, in place of the anecdotal approach that has too often dominated election reform conversations. While anecdotes and intuition have their place, they're no substitute for hard data and rigorous analysis.
Some examples may help in illustrating these contrasting approaches. For a long time, many election observers have suspected that our election system does not treat all voters equally – more specifically, that racial minorities, non-English speaking citizens, people with disabilities, and low-income voters are denied equal access to the ballot. But for the most part, the evidence has been anecdotal, relying on stories such as that of those of four to five hour lines that some voters faced in Columbus, Ohio in the November 2004 election. Such anecdotes may be helpful in illuminating problems that demand further inquiry, but it is perilous to make policy based upon anecdotes alone.
Fortunately, there is a recently released report that extensively documents the functioning of our election system, and supports the conclusion that there really are grave inequalities in the administration of elections. Last month, the United States Election Assistance Commission released its 2004 Election Survey, prepared by Election Data Services. It reports the results of the most comprehensive survey of U.S. election officials ever conducted, covering such areas as voter registration, turnout, ballots counted, voting machine malfunctions, and polling place operations. Among the most striking results found is the existence of substantial disparities across demographic groups. For example, the report finds that:
Jurisdictions with low education and income, compared with other jurisdictions, tend to report more inactive voter registration, lower voter turnout, higher number of provisional ballots cast, higher drop-off and associated components of overvotes and undervotes, lower average number of poll workers per polling place, and greater percentage of inadequately staffed polling places.
The report also finds that jurisdictions with large black populations report a greater percentage of polling places with inadequate numbers of poll workers. Jurisdictions with large numbers of non-English proficient voters also show disparities in the registration rolls, turnout absentee ballots, and provisional ballots.
As for people with disabilities, the survey's findings are even more depressing. Only about half of the jurisdictions surveyed even bothered to answer the question whether their polling places were wheelchair accessible, and less than a quarter of polling places were reported to provide the means for blind citizens to vote unassisted. This provides strong evidence that, in most places, disability access ranks very low on the list of election priorities.
To be sure, the data reported in the 2004 Election Day Survey is only a starting point. Social scientists and other policy analysts will spend considerable time in the coming months parsing this data. There is reason to hope that a set of thoughtful policy recommendations will emerge from that analysis. This could ultimately lead to legal reforms that may move us closer to the promise of equal access for all voters, while promoting the integrity of the voting process. The Election Updates Blog recently started by political scientists Michael Alvarez and Thad Hall is an important contribution to this approach, providing up-to-date information on the latest developments on the election research front.
Unfortunately, the moneyball approach remains the exception rather than the rule. A regrettable example of the opposite approach is a recent report on voting fraud recently issued by a group calling itself the "American Center on Voting Rights." The report, entitled "Vote Fraud, Intimidation & Suppression in the 2004 Presidential Election" is based mainly on unverified newspaper articles. The report attempts to create the impression that fraud is rampant, especially in predominantly black communities like East St. Louis, Detroit, Milwaukee, Philadelphia, Cleveland, and Alabama's Black Belt. It places special emphasis on reports that registrations with phony names such as "Jive F. Turkey, Sr." were submitted, though there's scant evidence that any of those people actually attempted to vote. Based on these anecdotes, ACVR argues that a strict photo ID requirement should be implemented.
In short, the ACVR relies on anecdotal evidence in an attempt to create the impression that fraud is rampant, and on this basis seeks to justify photo ID laws. When one looks carefully, however, it turns out that there's much less there than meets the eye. For example, even according to ACVR's own estimates, there were only 100 alleged instances of double-voting in Milwaukee, which has a voting age population of some 425,000 people. Further reason for skepticism of ACVR's anecdotal report arises when one looks deeper. As it turns out, one of the leaders of the group is Mark P. "Thor" Hearne who served as National Election Counsel to Bush-Cheney 2004. This makes one wonder whether ACVR's report is driven more by partisan motivations than by objective factfinding. Completely lacking is even a cursory cost-benefit analysis, estimating how many fraudulent votes are prevented for every person who is discouraged from voting.
While there's been relatively little effort an empirical analysis on the issue of voter ID, there is one recent study that provides a helpful counterpoint to the ACVR report. John Pawansarat of the University of Wisconsin Milwaukee recently issued a report on The Driver License Status of the Voting Age Population in Wisconsin. The study looks at which demographic groups do and don't have drivers license. This information is clearly essential, in evaluating the likely impact of proposals for state-issued photo ID. The disparities found were striking. Overall, more than 80% of adults had a driver's license; but only 45% of black men and 51% of black women had one. Disparities in the possession of photo ID also exist with respect to Latinos, elderly, and people of low income. The bottom line is that we can expect requirements for state-issued photo ID to have a much greater impact on these demographic groups. This is particularly troublesome given the EAC study's findings, which suggest that racial minorities, language minorities, poor people and people with disabilities already face greater obstacles.
It's not only those on the right who have succumbed to the old-time scouts' approach when it comes to election reform. Some electronic voting critics have engaged in an approach similar to ACVR's faulty fraud report, in urging that a "voter verifiable paper audit trail" (VVPAT) be required for all electronic voting machines. There's no question that there have been problems with some electronic voting systems, just as there have occasionally problems with fraud by individual voters. Lost amid the criticism of electronic voting is the fact that they help reduce the rate of uncounted votes, and do better than other technologies in providing disability access. Social science research shows that the implementation of electronic voting can significantly reduce the racial gap in uncounted votes that exists with other systems – see for example this study by Michael Tomz and Robert P. Van Houweling. On the other hand, there's little evidence to support the conclusion that the most frequently proposed remedy (the VVPAT) would be an effective remedy. Nevertheless, several states have decided to move forward with this so-called "common sense" solution – one that, once the evidence is examined, may turn out to make not much sense at all.
A more constructive approach to the voting technology issue may be found in a recent report from the National Academies, entitled Asking the Right Questions of Electronic Voting. This approach urges that research on voting technology be made an ongoing national priority, wisely recognizing that the evaluation of systems requires the balancing of many considerations including security, speed, usability, and access to all voters. A similarly constructive approach is exemplified by the National Institute of Standards and Technologies attempt to identify and analyze the various threats to voting systems. This approach will likely generate a more realistic approach to voting technology, one that focuses on things might realistically go wrong and what preventative measures are most likely to be effective.
The "moneyball" approach to election reform demands that we engage in empirical research and rigorous analysis, before passing laws. The alternative is to make policy based on seat-of-the-pants judgments and so-called "common sense." While intuitively appealing, this approach may result in policy recommendations that, in the end, don't make much sense at all.