Is Voter Suppression Inflating Electoral Bias? Comparing Canada, India, US, UK

• In Event: Electoral Systems as a Source of Bias

Abstract

Using data gathered with the support of grants from the MIT Election Data Science Lab as well as the APSA Centennial Center (Alma Ostrom and Leah Hopkins Awan Civic Education Fund), in this paper I will explore the following question: Does voter suppression increase electoral bias in single member district (SMD) electoral systems?

Much of the literature on electoral bias assumes that its primary cause is gerrymandering, which is helping drive an avalanche of research on gerrymandering within the American politics field. But as I demonstrated in a recent journal article entitled “American Disproportionality”, a key characteristic of SMD is that its translation of votes into seats is heavily influenced by the distribution of party support across geographic space. This geographic influence leads to a situation in which electoral bias can be inflated for a wide range of reasons, many of which are effectively random, like the homogeneity of party support across districts, while others can be manipulated by political actors. For my APSA paper, I extend this conclusion by arguing that any type of voter underrepresentation, whether in the form of malapportionment, voter suppression, or even simply sagging turnout, can increase disproportionality against the party that represents the underrepresented group, especially if that underrepresentation is geographically concentrated. Effectively, this implies a “double whammy” for the underrepresented group, in which their representation in elected office is decreased both because they provide fewer votes and because this decrease in votes compounds itself through disproportionality that inflates the number of seats certain parties receive relative to their voter support. This argument implies that voter suppression is potentially a more effective, and therefore more dangerous, strategy in SMD than proportional systems, especially when that suppression is geographically concentrated, as it often is in the United States.

To complete this analysis, I created two new measures related to electoral bias. The first, DPIx, which I introduced in “American Disproportionality”, measures asymmetrical disproportionality, or when election rules inflate the number of seats one party or set of parties wins in relation to another party or set of parties. (Applying the DPIx to US House elections since 1900, I demonstrated that asymmetrical disproportionality spiked in favor of the Republican Party starting in 2012, especially in states like Pennsylvania and North Carolina, but that this current bias is actually relatively small by historical standards, strongly suggesting that electoral bias is a systemic problem in American politics that goes far beyond gerrymandering.) The second is a measure of voter underrepresentation. In terms of disproportionality in an SMP system, mathematically the impact of voter suppression is equivalent to the impact of malapportionment. (For example, in terms of proportionality, keeping African-Americans away from the polls via voter suppression techniques has the same impact as drawing districts so that rural, white districts have a lower population than urban and more African-American districts. In both cases, the reduction of the number of voters within a district increases the influence of those who actually vote in determining who wins the particular seat.) For this reason, geographically varied voter underrepresentation in SMD systems can be calculated by comparing the vote to an expected value of what the vote would have been had this underrepresentation not existed. This then produces an easy way to measure the impact of voter underrepresentation on asymmetrical disproportionality. The DPIx each election can be compared to the DPIx of that expected value, with the result indicating whether there was an increased impact of voter underrepresentation on election results.

To extend this analysis beyond just the American context, I expanded the district-level data to include elections to the lower house of the Canadian parliament since 1867, Indian Lok Sabha since 1951, UK House of Commons since 1868, and US House of Representatives since 1840. The data for this analysis, which I gathered using Windows software that I built in C# and other languages, currently includes approximately 300,000 candidates across 84,000 districts. To test the argument on the theoretical-level as well, I also built a Monte Carlo simulator that distributes support for parties across two dimensional space and then calculates electoral results across a large range of electoral system scenarios, including suppressing votes, redrawing district lines, and comparing results for SMP and PR systems. I believe that this combination of real and simulated election data increases the robustness of the findings, since they include both empirical results across four countries over a century and a half and a theoretical analysis containing many thousands of election simulations.

Author

• Bernard Tamas, Valdosta State University