Kayla Thompson ’19
Truly, the right to vote is the most sacred pillar of a democracy. The history of the United States is decorated with triumphs of civic liberties as the franchise has been expanded from land-owning white men to nearly all American citizens. Despite public sentiments, however, the history of the expansion of the right to vote has not been linear. In fact, de facto enfranchisement has ebbed and flowed with time and with degrees of domestic social tension. While social scientists have long posited that felon disenfranchisement benefits the GOP by removing large portions of minorities and low-income Americans from the polity, little definitive research has been performed to understand the impact of felon disenfranchisement. This analysis attempts to better understand the 2016 political landscape by analyzing the relationships between felon voting laws, racial inequities, and political outcomes. State-level trends from the 2016 election year reveal that states with large racial disparities in disenfranchisement tend to be red states with small minority populations and below average, overall rates of disenfranchisement. Moreover, Trump tended to win by a larger margin in states that had higher proportions of people of color. Voter information from the 2016 election was used to estimate the voting patterns of the disenfranchised population in each state and the election outcome was then recalculated with the projected votes of the disenfranchised included. The simulated change in vote margin revealed that eliminating disenfranchisement laws would benefit Democratic candidates and that the disenfranchised would have enough political power to influence elections with narrower margins than the presidential election in 2016.
In the 21st century, many elections the local and federal levels have been decided by razor-thin margins, reminding the public that every vote counts. No election was more central to this sense of urgency than the 2000 Presidential Election. The margin between Vice President Albert Gore and Governor George W. Bush was less than 7500 votes in five states (Berman, 2015). In 2016, Donald Trump’s victory over Hillary Clinton was essentially decided by Pennsylvania, Wisconsin, Florida, and Michigan, where Trump won by a margin of less than 2%. This series of tight elections shows that even slight changes to the polity can significantly alter the political direction of the United States. Such realizations have also brought about a renewed sense of urgency to investigate and dismantle infringements to the voting rights of citizens.
The most glaring obstruction to suffrage in the 21st century is felon disenfranchisement. According to an analysis performed by The Sentencing Project, there are an estimated 6.1 million disenfranchised persons in the US, accounting for nearly 2.5% of the total U.S. voting age population. For perspective, in the four aforementioned swing states Trump won by an average of 55,060 votes. Michelle Alexander’s The New Jim Crow helped bring awareness to the issue of mass incarceration and the disproportionate burden of felon disenfranchisement on the minorities communities of certain states. In particular, Alabama, Alaska, Arizona, Florida, Iowa, Kentucky, Mississippi, Nevada, Tennessee, Virginia, and Wyoming disenfranchise more than 10% of their black, voting age populations.
Many scholars and politicians believe that felon disenfranchisement is politically advantageous for Republicans because it disproportionately removes low-income persons and minorities from the polity (Behrens, Uggen, & Manza, 2003; Goldman, 2004; Wheelock, 2005). Preuhs (2001) performed a cross-sectional data analysis on felon disenfranchisement policies and concluded that that the size of the minority population, parity in incarceration rates, and the degree of legislative professionalism are primary explanatory factors of determining felon disenfranchisement policies. The logic of this argument is that states with large minority populations have a greater incentive to try to remove the voice of minority groups from the polity. Such allegations ought to be contextualized within the historical and contemporary struggle to overcome infringements to minority voting rights in the United States. At the turn of the 20th century, it was commonplace for local governments to implement nearly insurmountable barriers that prevented minority participation in elections. Most notably, poll taxes, literacy and property tests, and the white primary barred Americans of color from the ballot box. State’s with very large black populations had greater incentive to suppress the political power of their black population. In Louisiana and Mississippi, the black voting electorate was reduced from 44% to 1% and from 70% to 6% respectively, signaling an abrupt reversal of the enfranchisement efforts made during the Reconstruction period (Pinkard, 2013).
The passing of Amendment 4 in Florida via a ballot initiative in the November 2018 midterms is an emblem of the contemporary fight for increased suffrage and has been heralded by many media sources as a tremendous victory for Florida’s minority communities. The day after Amendment 4 passed, Vox ran an article entitled “Would ex-felon voting rights have flipped Florida for Democrats? It’s possible.” This likely could be the case – Florida disenfranchises a staggering 10.43% of its total population, and 21.3% of its black population (The Sentencing Project). Inspired by this same question, Burch (2010) sought to determine the effect of including ex-felons in the electorate on the 2000 Presidential election between Bush and Gore. Her analysis concluded that because Florida’s felon population is mostly white, the disenfranchised would have favored Bush and therefore the not have changed the election. On the other hand, Phillips and Deckard (2016) argue that felon disenfranchisement has politically marginalized the black population of Florida and in doing so helped Republicans maintain power in the state. Much of the discrepancy in opinion on the impact of disenfranchisement hinges prior belief regarding the voter turnout of ex-felon populations (Burch, 2010; Miles, 2004; Meredith and Morse, 2014). Haselswerdt (2009) showed that only 5% of legally-able ex-felons in Erie County, NY registered and voted in 2004 and 2005. Gerber et al., (2008) agrees that ex-felons have lower turnout rates but performed a field experiment showing that outreach efforts such as contacting ex-felons and providing them with information regarding reinstating and exercising voting rights can bring turnout amongst the disenfranchised up to levels comparable to that of non-felons of similar socioeconomic status. As such, a definitive conclusion on the voting behavior of ex-felons has not been reached.
While the literature has focused on voter turnout amongst the ex-felon population, relatively little attention has been paid to better understanding the political and social patterns that are correlated with political disenfranchisement. A better understanding of the relationship between felon disenfranchisement, race, and political outcomes at the time of the 2016 Presidential Election will allow the public to be better informed about the impact of policies on society and ultimately will allow for more informed policy decisions. To achieve this end, I test four hypotheses to reveal trends in felon disenfranchisements laws.
Research Question and Methodology
This analysis seeks to better understand the relationship between felon disenfranchisement laws, race, and political outcomes as it pertains to the 2016 election year. My analysis is based off of the longitudinal analysis of disenfranchisement performed by Burmila (2017). I focus on variables that are widely considered important indicators of voter choice: race, income, and home state.
Disparity in the disenfranchisement of Blacks and Latinos is correlated with the potential political power of the respective minority population.
Given that people of color tend to vote for Democrats, the removal of large portions of people of color from the polity should make it easier for Republicans to win elections. This hypothesis is supported by the de-aggregated 2016 voting data provided by the CNN Exit Poll and by Preuh’s research (2001) that suggests that race is a primary factor in determining the severity of voter disenfranchisement laws in a state.
I fit an ordinary least squares regression of the disparity in the disenfranchisement of black people in a given state using the proportion of the state’s population that is black as the explanatory variable. This process was repeated for Latino populations.
To perform this analysis, I calculated the disparity in Black and Latino disenfranchisement by state. The Sentencing Project provides data on the number of disenfranchised persons by state, and includes a breakdown of the number of Black and non-Black disenfranchised persons by state as well. I first verified that the estimates of the number of disenfranchised black persons in each state provided by The Sentencing Project are reasonable by comparing them with white-black arrest disparities by state and the share of the black population by state. After confirming that these statistics were reasonable, I used these figures in conjunction with state-level demographic information and arrest rates by race provided by the US Census and the Department of Justice to calculate the following statistics:
The disparity of black disenfranchisement in each state is calculated using the following model:
I then needed to calculate the number of disenfranchised persons who are White and Latino in each state. To do so, I derived the expected number of disenfranchised Latinos and Whites in each state based on the known number non-black disenfranchised persons, incarceration rates by race in each state, and state demographic information. Incarceration rates by race were provided by the Bureau of Justice Statistics. The number of disenfranchised Whites and Latinos in each state was calculated using the following model:
It is advantageous to note two limitations to this methodology. Firstly, this analysis only takes into account Black, Latino, and White persons as these statistics were consistently reported for all 50 states while Asian and Indigenous populations were often not included in datasets. We are comfortable not including Asian and Indigenous populations because Asian-Americans compose a nearly negligible proportion of the disenfranchised persons and the data on the incarceration of Indigenous persons in US prisons has not been reliable recorded. Secondly, this methodology assumes that the Sentencing Project grouped Afro-Latinos in the non-black disenfranchised population. I made this assumption because it is conventional in the US to measure Latinos are a separate group from either White or Black persons. It is worth noting that if this assumption is violated my methodology would double count Afro-Latinos, which would put upward pressure on the disparity of minority disenfranchisement in states with large Afro-Latino populations. Similar considerations can be made for the removal of mixed race individuals from the analysis.
Republican-controlled states have higher disparities in disenfranchisement and more severe disenfranchisement policies than Democrat-controlled states.
It is hypothesized that Republican-controlled governments are likely to support disenfranchisement laws in order to maintain their power. This hypothesis stems from explicit opinions expressed by Republican politicians such as Alabama Republican Party Chairman Marty Connors: “As frank as I can be,” he said, “we’re opposed to [restoring voting rights] because felons don’t tend to vote Republican” (Krajick, 2004). If it is in fact true that politicians use disenfranchisement laws as a political tool to alter the polity in their favor (or that Democrats support the enfranchisement of felons as a political tool), Red states should have more severe disenfranchisement laws and higher disparities in disenfranchisement as compared to Blue states.
To assess this claim, states were first grouped according to their party affiliation. The Cooks Partisan Voting Index (PVI) was used as the basis to make the following groups:
Then Wilcoxon rank sum tests and confidence intervals of the Disenfranchisement Law Severity Index and the disparities in disenfranchisement for Blacks and Latinos were assessed for each group. The Wilcoxon rank sum test was used because the underlying data is not normally distributed and the sample sizes are small.
The Disenfranchisement Law Severity Index was derived as follows:
1 – State only disenfranchises currently incarcerated individuals
2 – State disenfranchises ex-felons through the end of their parole and/or probation period
3 – State disenfranchises ex-felons for life or for a period of time after the end of their parole period
High proportions of people of color and low-income persons in a state are correlated with better political outcomes for Democrats.
Data from the CNN Exit Poll regarding the voting behavior of different demographics in the 2016 presidential election support the notion that people or color and low-income people tend to vote for Democrats. The polls revealed that 69% of people of color voted for Clinton as opposed to 37% of white Americans. Moreover, 53% of those with an annual income less under 30k voted for Clinton as compared to 47% of those making 100k or more. This demographic data by state suggests that states that have higher proportions of people of color and low-income people should have had a smaller vote margin in the 2016 election compared to states with more White and more affluent populations.
To test this hypothesis, OLS regressions were run on the vote gap by state using minority population share as the explanatory variable. Data from the most recent US Census was used to calculate the proportion of minorities in the Voting Age Population in each state. Demographic variables were updated for each state to remove the disenfranchised from the Census estimates.
The regression coefficients of each variable were then compared and the relationships between variables were visualized with scatterplots.
If all disenfranchised persons had been able to vote in the 2016 election, Clinton would have ultimately won.
Without question, adding all 6.1 million disenfranchised persons to the polity is large enough to sway a presidential election. As elections are determined by the electoral college, the simulated effect of the disenfranchised on the vote margin in each state must be considered.
I use a math-based model to estimate the vote margin of the disenfranchised by state using the assumption that the disenfranchised would vote according to the 2016 voting averages of their respective race by state. The new vote margin is calculated by adding the projected voting margin of the disenfranchised to the true margin in the 2016 election.
Although it would be ideal to factor in gender, age, and income, individual-level demographic information for ex-felons is not publically available and thus the voting behavior of the interactions between all of these variables cannot be determined. Race was selected as the predictor because it was the most influential variable in a regression of political affiliation against race, age, income, and gender. Not taking into account income will put upward pressure on the projected GOP vote count because race averages don’t take into account that the disenfranchised are disproportionately low-income. At the same time, not taking into account that the disenfranchised are disproportionately male will inflate the projected number of DNC votes. Not taking into account age will also put upward pressure on the projected DNC votes because the average age of the ex-felon population is likely to not have many 18-24 year olds and is thus slightly older than the average VAP population. As such, I suspect that the net effect of only using race as a predictor for the voting tendencies of the disenfranchised in each state will bias the projected vote margin in favor of Clinton.
First I sought to determine whether a positive linear relationship exists between the disparity in the disenfranchisement of Blacks and the potential political power of the respective minority population. As presented in Table 1, the results of the OLS regression models show that a statistically significant relationship exists between the minority share of the population, which is a proxy for the political power of a minority group, and disparity of disenfranchisement. Contrary to my hypothesis, however, the relationship is actually negative (see Figure 1). This means that states with large racial disparities in disenfranchisement have smaller minority population shares as compared to the minority population shares of states with small racial disparities in disenfranchisement.
Table 1: Results from Regression of Racial Disparity in Disenfranchisement and the Population of Latinos and Blacks.
Table 1: Results from Regression of Racial Disparity in Disenfranchisement as the Population Share of Latinos and Blacks
The results regarding the relationship between party control and the severity of disenfranchisement laws show that a statistically significant difference exists between the average severity of disenfranchisement laws for solidly blue states and that of red-leaning and of solidly red states. In keeping with my hypothesis, solidly blue states have less severe disenfranchisement laws. In particular, red states are more likely to disenfranchise ex-felons for life or require a Governor’s pardon in order for voting rights to be restored. Moreover, the average severity of disenfranchisement laws for solidly red states was slightly greater than that of red-leaning states (2.231 and 2.133 respectively). The 95% confidence intervals for the t-tests for the difference in average political polarization across states grouped according to the severity of their disenfranchisement laws is shown in Table 2.
In regards to racial disparities, there appears to be a negligible difference in the Black disparity of disenfranchisement between states that voted blue and red (3.512 and 3.539 respectively) while that gap was slightly larger for the Latino disparity in disenfranchisement between blue and red states (2.073 compared to 1.91). As shown in Table 3, however, the states with the largest disparities in black disenfranchisement were mostly red states and states with relatively small black populations. Similarly, Table 4 shows that disparity in the disenfranchisement of Latinos is large in states with relatively small Latin populations. Interestingly, states that had of the largest rates of disparity in the disenfranchisement of minorities also has amongst the lowest percentage of disenfranchisement of the 48 states with felon disenfranchisement laws (the average was 3.86%).
Table 2: 95% Confidence Intervals for comparison of the severity of disenfranchisement by states grouped by party affiliation
Table 3: States with a Disparity Index of greater than one standard deviation above the mean in terms of the disparity in the disenfranchisement of African Americans
Table 4: States with a Disparity index of greater than one standard deviation above the mean in terms of the disparity in the disenfranchisement of Latinos.
The scatterplots of state-level demographic information regarding income and race appear in Figure 2. The plots for mean income support the hypothesis that higher income states tend to have voted for Clinton. The plots for race, show a surprising incongruity between aggregate and de-aggregate voting tendencies. Even though the voting tendencies of people of color tend to favor Clinton, which aligns with the plots for Blue states, the trend is that in red states, Trump won by a larger margin when the minority share of the population was larger. This phenomenon can be explained by some combination of the voting tendencies of whites and males in Red states becoming more in favor of the GOP as the proportion of minorities in the state increases, and by a larger gap in voter turnout in Red states between minorities and whites. Although voter turnout information for the 2016 election by State is not available yet, the results of the CNN Exit Poll reveal that 64% whites voted for the GOP in Red states with above average proportions of minorities as compared to 58% of whites in states with below average proportions of minorities, accounting for some of the difference. It is reasonable to contribute that the rest of the difference in vote gap to difference in voter turnout, which could have resulted from minority voters turning out less and/or minorities being barred from voting due to a variety of obstacles such a felon disenfranchisement, voting ID laws, overcrowded polling stations, and limited voting windows.
Figure 2: Plots of state-level explanatory variables against vote margin for states grouped by whether they went Blue or Red during the 2016 Presidential Election
Finally, I address whether the inclusion of disenfranchised persons in the polity would have changed the results of the 2016 election. The mathematical approach to simulating the impact of including disenfranchised persons in the 2016 election did not result in a change in electoral college votes in any state. Florida, Michigan, Pennsylvania, and Wisconsin were regarded as the states that ultimately ensured Trump’s victory. Table 5 shows the projected change in vote margin of these four states and the observed vote margin. Although the disenfranchised would not have been able to flip these states in this election, they could have been the deciding votes in a closer election.
Table 5: Te simulated change in vote margin and the 2016 observed vote margin of four swing states that went to Trump
Nonetheless, the simulated results do show that felon disenfranchisement unequivocally hurts the DNC. Of the 48 states that disenfranchise at least some ex-felons, only 9 had a simulated change in vote margin that favored Trump. Shown in Table 6, these 9 states had relatively small margins in favor of Trump (with the exception of Kentucky). In comparison, of the 39 states whose change in vote margin would have favored Clinton, the average change in vote margin was 15,560. That being said, we expected our estimated to be slightly biased in favor of Clinton.
Table 6: States where the simulated vote margin favored Trump
Within each state, the racial composition of the disenfranchised population in absolute terms is a good proxy to understand the net effect of including the disenfranchised vote. In 26 states, the number of disenfranchised White people is larger than the number of disenfranchised people of color. This means that in 22 states, the number of disenfranchised persons is majority black and Latino.
Better understanding the nature of felon disenfranchisement should allow for more fruitful discussion and policy analysis. By assessing claims made in the literature regarding the interactions between racial disparities in disenfranchisement and voting outcomes in the context of the 2016 election, I uncover some surprising trends. Solidly blue states have less severe laws regarding felony disenfranchisement; however, less severe laws regarding felony disenfranchisement does not mean that a state is more likely to be blue. Also, at a state-level there exists a positive association between minority population share and preference for Trump, which raise questions regarding the ability of minority populations in red states to successfully participate in elections and the impact of multi-racial populations on state-level politics. Given that our analysis suggests that it is unlikely that disenfranchisement of minorities has led to Trump’s strong margins in these states, future research ought to focus on better understanding the casual relationships that underlie this trend.
A counterintuitive finding is that states with large minority population shares tend to have less racial disparity in the disenfranchised and also tend to disenfranchise a below average percentage of their total population. The average disparity in disenfranchisement for black persons across the 48 states with some disenfranchisement laws is 3.67 – meaning that there are 3.67 times the number of black disenfranchised American than would be equitable if disenfranchisement laws proportionally afflicted all Americans. The disparity in the disenfranchisement of Latinos is 1.7. Interestingly, this disparity is greater in states with relatively small minority populations. This phenomenon is a stark departure from the historical narrative of de facto disenfranchisement in the Post-Reconstruction South. A plausible explanation for this is that states with higher proportions of people of color develop more robust prison system infrastructure which results in higher incarceration rates amongst all racial groups. In contrast, states with small proportions of people of color may tend to have weaker prison infrastructure but disproportionately target people of color. This hypothesis is supported by the fact that these states with above average disparities in disenfranchisement amongst their below average sized minority populations also have below-average disenfranchisement of their entire populations. Further analysis is needed to bring to light exactly the mechanisms that cause this phenomenon.
By simulating the voting behavior of the disenfranchised for the 2016 election, this analysis shows that including the disenfranchised would not have altered the outcome of the 2016 election. The simulation does make clear that felon disenfranchisement hurts the DNC and is politically advantageous for the GOP. This finding, however, does not change the fact that disenfranchising large proportions of minorities has the potential to alter the political discourse in the state by weakening the incentive for politicians to cater to the needs of their Black and Latin populations. Moreover, the influence of disenfranchisement on local elections has very likely hurt the ability of Democrats to win. Yet, Democrat-controlled states have roughly the same minority disparities in disenfranchisement as Republican-controlled states, revealing that a motivation for disenfranchisement exists beyond the desire to manipulate partisan outcomes. This analysis makes clear that considerations about felon disenfranchisement should be seriously considered in all states, regardless of political tendencies or racial composition. This analysis leaves several questions unanswered. Future research should focus on better understanding the contemporary barriers to voting and the impact of multi-racial communities on the political leanings of citizens. A better understanding of the impact of disenfranchisement on the voting behavior of families and communities is needed to understand the long-term impacts of the 6.1 million citizens that are currently disenfranchised in the U.S.
Berman, Ari. Give Us the Ballot: The Modern Struggle for Voting Rights in America. New York: Picador/Farrar, Straus and Giroux, 2016.
Burch, Traci. “Did Disfranchisement Laws Help Elect President Bush? New Evidence on the Turnout Rates and Candidate Preferences of Florida’s Ex-Felons.” Political Behavior34, no. 1 (2010): 1-26. doi:10.1007/s11109-010-9150-9.
Burmila, Edward M. “Voter Turnout, Felon Disenfranchisement and Partisan Outcomes in Presidential Elections, 1988–2012.” Social Justice Research30, no. 1 (2017): 72-88. doi:10.1007/s11211-017-0277-2.
Gerber, Alan S., Donald P. Green, and Christopher W. Larimer. “Social Pressure and Voter Turnout: Evidence from a Large-Scale Field Experiment.” American Political Science Review102, no. 01 (2008): 33-48. doi:10.1017/s000305540808009x.
Haselswerdt, Michael V. “Con Job: An Estimate of Ex-Felon Voter Turnout Using Document-Based Data.” Social Science Quarterly90, no. 2 (2009): 262-73. doi:10.1111/j.1540-6237.2009.00616.x.
Krajick, Kevin. “Why Can’t Ex-Felons Vote?” The Washington Post. August 18, 2004. Accessed November 11, 2018. http://www.washingtonpost.com/wp-dyn/articles/A9785-2004Aug17.html.
Lopez. “Would Ex-felon Voting Rights Have Flipped Florida for Democrats? It’s Possible.” Vox.com. November 07, 2018. Accessed November 11, 2018. https://www.vox.com/midterm-elections/2018/11/7/18071404/florida-amendment-4-election-results-felon-voting-governor-senate.
Meredith, Marc, and Michael Morse. “Do Voting Rights Notification Laws Increase Ex-Felon Turnout?” The ANNALS of the American Academy of Political and Social Science651, no. 1 (2013): 220-49. doi:10.1177/0002716213502931.
Phillips, Anthony Jamal, and Natalie Deckard. “Felon Disenfranchisement Laws and the Feedback Loop of Political Exclusion: The Case of Florida.” Journal of African American Studies20, no. 1 (2015): 1-18. doi:10.1007/s12111-015-9314-0.
Pinkard, J. E. “African American Felon Disenfranchisement : Case Studies in Modern Racism and Political Exclusion.” LFB Scholarly Publishing LLC40, no. 1 (2013): 100-01.
Preuhs, Robert R. “State Felon Disenfranchisement Policy.” The Canadian Journal of Chemical Engineering. December 19, 2002. Accessed November 11, 2018. http://onlinelibrary.wiley.com/doi/10.1111/0038-4941.00056/abstract.
“State-by-State Data.” The Sentencing Project. Accessed November 11, 2018. https://www.sentencingproject.org/the-facts/#map.
Uggen, Christopher, and Jeff Manza. “Democratic Contraction? Political Consequences of Felon Disenfranchisement in the United States.” American Sociological Review67, no. 6 (2002): 777. doi:10.2307/3088970.
Uggen, Christopher, Jeff Manza, and Angela Behrans. “Felon Voting Rights and the Disenfranchisement of African Americans.” Racializing Justice, Disenfranchising Lives, 2007, 237-45. doi:10.1057/9780230607347_17.
Wheelock, Darren. “Collateral Consequences and Racial Inequality.” Journal of Contemporary Criminal Justice21, no. 1 (2005): 82-90. doi:10.1177/1043986204271702.
 Average Black population share across all 50 states is 10%. One standard deviation above the mean black disparity in disenfranchisement index is 5.16.
 Average Latino population share across all 50 states is 12%. One standard deviation above the mean black disparity in disenfranchisement index is 3.95.
 Noting that Maine and Vermont do not disenfranchise any citizens, regardless of their criminal record.