Prepared by Michael A. Smith, Associate Professor of Political Science for a forthcoming presentation at the 2015 meetings of the Northeastern Political Science Association

Emporia State University

DRAFT: Please do not circulate, reproduce, or forward without the author’s permission!
Data and Results

Two voter lists were obtained from the Kansas Secretary of State’s office. The first was a list of suspense voters—voters who had begun registering but could not vote because they did not show proof of citizenship. This list required an open records request made by the Lawrence Journal-World, and it listed suspense voters as of Fall, 2014, prior to the midterm election. The second list was a list of all registered voters in Kansas—active, inactive, and suspense. The second list was obtained after the November, 2014 midterms and included voting history.

The suspense voter list contained 23,691 names. The overall voter list contained 1,710,316 names, so suspense voters are about 1.4% of the combined list. The two datasets were merged using PASW (formerly SPSS) and then a search for duplicates was run using each voter’s unique ID number. Duplicates between the suspense and active voter lists indicate that a suspense voter had subsequently completed his or her registration. The result: 3705 voters who were on the suspense list in Fall, 2014 had subsequently completed their registrations by June, 2015 (see Chart 1). Of those, 1341 voted in the 2014 midterm election. Put differently, of all suspense voters, 16% completed their registrations and 5.6% voted in the midterm election that year. By contrast, among active voters, 45.4% voted in the 2014 midterm election (see Chart 2).

Another notable variable differentiating suspense voters from active voters is partisanship. Kansas is a closed-primary state where voters must choose a party affiliation when registering, in order to participate in the primary elections. There are striking differences between the suspense voters and the active voters in terms of party affiliation, summarized in Chart 3. The sharpest difference between the suspense voters and the active voters is the far larger percentage of suspense voters who are unaffiliated. This is particularly notable given another, recent legal change championed through the Kansas Legislature by Kobach: Kansas is no long a “semiclosed” primary state in which unaffiliated voters can vote in a party primary, then automatically be re-registered as members of that party. Instead, unaffiliated voters are now simply prohibited from primary voting (AP, 2014). Also noticeable is the smaller difference between percentages of Democrats and Republicans among the suspense voters, as opposed to the active voters. The suspense-and-completed-registration group includes a somewhat lower percentage of unaffiliated voters, and as somewhat higher percentage of Republicans, versus the overall suspense voters group. Still, both groups differ markedly from the active voters. Finally, a small but interesting finding: among the select group of suspense voters who later completed their registration, 2% registered as Libertarians, nearly three times the percentage of Libertarians among the active voters and far higher than the percentage for all suspense voters.

Chart 4 plots the three groups of voters by their average ages. Suspense voters who completed their registrations were the youngest group with an average age of 30, followed by all suspense voters at 38 and finally active voters at 51. Regarding gender, 50.6% of suspense voters are women while 48.7% of suspense voters who complete their registrations are women. Among active voters, 52.9% are women (See Chart 5).

The next project was to map the suspense voters by census tract. This was done by partitioning the suspense voter list into batches of 1000 voters each and running it through the U.S. Census online Geocoder tool to identify the census tracts where the voters lived. Map 1 lists nine of the ten census tracts with the highest percentages of suspense voters, with a tenth excluded due to its very small population. Of the remaining nine, five were located in the city of Wichita, in Sedgwick County. All were located in urban or suburban areas—quite remarkable for a state with a legendary rural reputation. In addition to the Wichita tracts, one was located in Topeka, two in suburban Kansas City, and one in college-town Lawrence. That census tract encompasses the heart of the University of Kansas campus. It had the highest percentage of suspense voters statewide, at 2.7%.

While Kansas is no longer as rural as its Wizard of Oz reputation may imply, these results are still remarkable. Johnson County, in suburban Kansas City, is the state’s most-populous county, yet Sedgwick County featured five tracts on the top-ten suspense voters list while Johnson County featured only two. Several of the tracts in Wichita, plus the one in Topeka, were located in inner-city areas. One of the two tracts in Johnson County was in an inner-ring suburban area more diverse, and with a far lower median income, than most others in that wealthy, suburban county. Also puzzling is the absence of Wyandotte County from this list. Just north of Johnson County, Wyandotte is the home of Kansas City, KS—a less-wealthy satellite city of Kansas City, MO, analogous to East St. Louis, IL. The census tract omitted from the map was located in Wyandotte county, but only had 38 residents—too few for a serious analysis.

In short, this analysis of the ten census tracts with the highest percentages of suspense voters produced:

  • Six in central cities, five of which were in Wichita
  • One in an inner ring suburban area
  • One in a wealthy suburban area—an unexplained outlier
  • One on a university campus, and
  • One excluded due to a very small population

Demographically, these nine census tracts varied, but many stood out for their high minority populations and low median incomes. The Wichita tracts ranged from 6.1% to 92.8% African-American, from 5.5% to 39.8% Hispanic, and 25.6% to 53% below the poverty line. The Topeka tract was also more diverse and had more poverty than statewide averages. One of the two Johnson County tracts was slightly less diverse and poor than statewide averages but still far more diverse and less wealthy than the neighboring tracts in this upscale area—an archtypical “changing” inner-ring suburb. The second Johnson County precinct was quite wealthy and not very diverse—an unexplained outlier. Finally, the University of Kansas tract was more diverse than the state as a whole, but the high poverty numbers are probably anomalous because the population would be made up of college students living on parents’ assets or student loans, at least in part. Note the very low median age of the KU tract: only 19.7 years old.

Map 2 is a “heat map” of all census tracts in the state by number of suspense voters—again, the pattern of concentration in Wichita and other urban areas is notable.

MAP 2: Kansas Suspense Voters By Census Tract

The next project was to run a multivariate regression on the percentage of suspense voters in each census tract, using the following independent variables, all from U.S. Census:

  • Median Age
  • % American Indian-Alaska Native (Native American)
  • % African-American
  • % Hispanic
  • % Below poverty level

Table 1: Model Summary
R / R Square / Adjusted R Square / Std. Error of the Estimate
.265 / .070 / .064 / .0035
Model / ANOVA: Sum of Squares / df / Mean Square / F / Sig
Regression / .001 / 5 / .000 / 11.541 / .000
Residual / .009 / 764 / .000
Total / .010 / 769
Model / Unstandardized Coefficients: B / Std. Error / Standardized Coefficients: Beta / t / Sig.
Constant / .007 / .001 / 9.808 / .000
Below Poverty Level % / .00006338 / .000 / .197 / 4.439 / .000
African-American % / .00003004 / .000 / .115 / 2.810 / .005
Hispanic % / -.00001952 / .000 / -0.75 / -1.903 / 0.57
Native American % / -.000002594 / .000 / -.002 / -.052 / .958
Median Age / -.00001275 / .000 / -.029 / -.758 / .449

*Dependent variable: % of suspense voters in each census tract.

The analysis was done with PASW/SPSS. A visual check of the residuals indicated that they are normally distributed.

While all coefficients are small, as is the R-square, the model still showed significant effects for the percent below poverty level and the percent African-American in the census tracts, both in the predicted direction. That is, poverty and African-American population were correlated with higher percentages of suspense voters. The coefficient for Hispanics was not significant, nor was that for Native Americans or median age.

In sum: suspense voters make up about 1.4 % of the total electorate in Kansas, not counting inactive voters. They tend to be younger and are concentrated in the state’s cities, particularly Wichita. Of those, several tracts with the highest percentages feature very high minority populations and poverty rates. The highest single percentage of suspense voters is on a university campus. The percentage tends to rise with the percentage of poverty and of African-Americans, however, this is not true for the other variables we tested: percent Hispanic, percent Native American, and median age. However, at an individual level, suspense voters tend to be younger, and those who complete their registrations tend to be younger still, than are active voters. Suspense voters are also less likely to be Republicans and far more likely to be unaffiliated. Suspense voters are also slightly more likely to be male, and those who later complete their registrations are even more likely to be male, compared to active voters. Finally, about 16% of suspense voters completed their registrations within the nine-month period we studied, and about 6% of those voted in the 2014 midterm election.

1