This document is a companion to Simon Jackman and Bradley Spahn’s working paper “Unlisted in America.” It provides additional information about respondents to the 2012 ANES, broken out by the citizenship categories discused in the paper.

Some of the tables and graphs below do not represent statistically significant differences between the respondent categories (voter, registered, etc.). Only comparisons with a p-value less than .035 can be classified as significant while controlling the false discovery rate at the .05 level.

### Gender

Distribution of citizen types (percentages), by gender, ANES 2012 face-to-face respondents (weighted). $$\chi^2 = 8.0$$, $$p = .042$$.
Male Female
2012 Voter 64 67
Registered 13 14
Unregistered 10 10
Unlisted 13 9
Total 48 52

### Marital Status - Women

Distribution of citizen types (percentages), by marrital status for women, ANES 2012 face-to-face respondents (weighted). $$\chi^2 = 85.8$$, $$p < .001$$.
Married Spouse Absent Widowed Divorced Separated Never Married
2012 Voter 74 82 72 59 71 49
Registered 10 6 15 19 14 20
Unregistered 10 10 7 14 8 10
Unlisted 6 2 5 7 7 20
Total 46 6 9 12 3 23

### Marital Status Breakdown - Men

Distribution of citizen types (percentages), by marrital status for men, ANES 2012 face-to-face respondents (weighted). $$\chi^2 = 154.7$$, $$p < .001$$.
Married Spouse Absent Widowed Divorced Separated Never Married
2012 Voter 79 77 61 53 45 40
Registered 7 18 14 19 27 19
Unregistered 9 1 16 10 8 13
Unlisted 5 4 10 18 21 27
Total 51 4 2 12 2 28

### Age

The median age of an unlisted person is 30 years, while the median age of 2012 voters is 50. Registered non-voters and unregistered persons are indistinguishable from one another with respect to age (median ages of 36 and 39, respectively).

Differences in the distribution of respondent ages by category were tested using a Kolmogorov–Smirnov test. All citizen types had statistically significant differences in their distribution from the other groups, except for the registered and unregistered populations, where the differences in the distribution of age could be due to sampling variability alone.

The plot shows the distribution of age by citizen type. The lines shown are densities, representing the distribution of respondents’ self-reported ages. The densities were normed to have the same in for each panel. The vertical dashed line show the median age for each group.

### Income

Voters are significantly richer than their non-voting counterparts, with the unlisted having the lowest income of the 4 groups. This reflects the the fact that people that aren’t registered to vote can still be listed if they appear on commercial lists. Unlisted people are on the periphery of this Venn-Diagram, appearing on neither voter or consumer files.

Differences in the distribution of respondents’ self-reported household income by category were tested using a Kolmogorov–Smirnov test. The 2012 voters group and each of the other citizen types had highly statistically significant differences in their distribution, whereas the differences in the distribution of income between the others groups could be due to sampling variability alone.

The plot shows the distribution of income by citizen type. The lines shown are densities, representing the distribution of respondents’ self-reported household income The densities were normed to have the same in for each panel. The vertical dashed line show the median income for each group.

### Voter ID

Many states have enacted voter ID laws in recent years, prompting concern that the laws might disproportionately disenfranchise minorities, who are less likely to have an acceptable form of ID.

Using our citizenship categories, we see that while 2012 voters all very high rates of having identification (either a non-expired passport, driver’s license, or other form of government-issued ID), the racial differences are quite stark in the non-voting categories. Whites are consistently as or more likely to have identification than non-white members of the same category. This poses a potential barrier to registration efforts in states that require identification, as not only will many of these minority voters need to be registered to vote, but they’ll also need to acquire a valid id.

Note that the fully interacted model of citizenship category and race is statisically significant (tested against the more restricted category-effects only, race-effects only, and non-interacted models) at the $$p < .001$$ level

### Ideology

Below are the respondent’s self-reported ideology on a 7-point scale, running from liberal to conservative. While we note in the paper that the non-voting groups, and the unlisted in particular, take liberal issue positions, these liberal positions are not reflected in their ideological self-placement. We think this is because of lower-levels of political information among these groups. Because they engage in less political discourse, they are much more likely to choose the politically neutral label of “moderate” to describe themselves.

Note that differences in the distribution between the registered, unregistered, and unlisted categories are not statistically significant.