Last month, I wrote an op-ed for Wired Magazine entitled “An Intriguing Link Between Police Shootings and Black Voter Registration”. Since then, I’ve received a number of questions and comments via email; I respond to some of the most popular themes* here.
Question: How might increasing black voter registration reduce over-representation of black Americans among fatal police shootings in the United States?
Answer: Let’s first get something straight – in the article, we found that higher rates of voter registration among eligible black Americans were correlated with lower over-representation ratios across the United States. This relationship is not necessarily causal.
In my opinion, increasing voter registration by any means necessary will not necessarily reduce fatal police shootings… But increasing minority voter registration by way of empowering minority communities might. As such, black voter registration acts as a measurable performance metric that can be used to gauge progress made with respect to underlying societal (i.e. enabling or disabling) conditions (e.g. systemic racism). (That said, it is important to acknowledge that voter registration is also an indicator of civic engagement and active citizenship, which may be protective within the context of fatal police shootings due to a more generalized form of familiarity bias, in which individuals prefer that which they know over that which they don’t.)
Question: Given that voter registration deadlines are known in advance, how are minority groups disproportionately suppressed when same-day registration isn’t an option?
Answer: On Election Day 2012, black Americans comprised 35% of those who engaged in same-day registration – despite accounting for only 22% of the electorate in the locations considered. So, why might this have been the case?
A few possible hypotheses come to mind:
- While online and mail-in registration has become increasingly available across the Untied States, it’s possible that education and awareness materials regarding these more convenient registration pathways are not readily available to minority groups. If this is the case, unregistered voters who live in states without same-day registration – and who believe that in-person registration is their only option – would have to take two days off of work: one day to register and one day to vote. Because poverty rates among minority groups are significantly higher than that of white Americans, taking two days off from work simply may not be financially feasible, thus resulting in registrant suppression.
- When registered voters change their place of residence, their registration is often immediately nullified, requiring them to re-register using their new address. Low-income individuals are not only more likely to belong to minority groups, but they’re also more likely to move. Thus, in the event that mobile voters are unaware that they must re-register or are unable to do so (for the reasons listed above), unavailability of same-day voter registration would amount to registrant suppression.
Question: How is voter eligibility defined?
Answer: According to Pew Research, eligible voters are defined as American citizens who are 18 years of age and older; as a result, this is the definition we used in the article. However, there are circumstances within which otherwise eligible voters may be temporarily barred from voting (or registering to vote). The most pertinent of these (within the context of the article) is felony conviction. Restoration of voting rights following incarceration and/or completion of sentence varies from state-to-state; however, all but two states revoke voting rights during incarceration.
Question: Are incarceration rates among black Americans accounted for in the model?
Answer: Black voter registration rates are – without a doubt – impacted by black incarceration rates. For this reason, a measure of black incarceration was included in the model (i.e. black-to-white incarceration ratio). By including this variable, we were able to control for the aforementioned relationship between black incarceration and black voter registration. (It is worth noting that black-to-white incarceration ratio is closely correlated with black incarceration rate (p = .02); we chose to use the former instead of the latter to partially capture state-varying judicial bias as well).
Question: What are some of the limitations of the data used?
Answer: All data – especially publicly available data! – have considerable limitations. Here are a few that pertain directly to the data set used for the article:
- The sample size for the multivariate analysis was quite small (N=39).
- Within this N=39 sample, there were 5 states for which the Census reported 0% black voter registration. However, this doesn’t necessarily mean that there are N=0 black American registrants in these states; rather, black Americans comprise such a small percentage of the total population in these states that the sampling method used by the Census yields denominators that are too small to estimate the true percentage that are registered to vote. This said, because the Census provides raw data for each state, we did manually calculate voter registration among black Americans for the 5 states with recorded values of 0%. When we ran this revised data set though our multivariate model, we found that – though the statistical significance associated with black voter registration dampened, the dampening effect was not significant in and of itself (p=.001 vs. p=.006). Ultimately, after considerable back-and-forth, we chose to use the original data set (with 0% values for the 5 states in question) because using the percentages we manually calculated from the raw data would have required considerable documentation (very appropriate for a scientific journal, but not for Wired).
- Because we used exclusively public data, there were variables that we wanted to include in our model that we couldn’t. (Race-stratified gun ownership is just one example.) It’s not to say that these data aren’t available somewhere; they most likely are, but they aren’t easy to find or readily accessible by public health research organizations.
…Which brings our Q&A full circle to one of the tenets of the article itself.
Thanks to the NRA-instigated ban on CDC gun violence research back in 1996, publicly available data on gun violence – data that would have typically been cleaned and curated by the CDC – are hard to come by. Though the ban was lifted two years ago, federal funding for gun violence research remains tight. In the meanwhile, public health researchers like myself do our best with whatever data we can scrape together… Which is exactly what we’ve tried to do here (and at Wired).
*Questions have been paraphrased to address multiple commenters simultaneously. Special thanks to those who took the time to reach out and get in touch!