Vote misreporting and black turnout studies in the U.S.

ABSTRACT

Vote misreporting is a major concern for studies of electoral participation. Concern over nonvoters in surveys who claim to vote is especially relevant for black turnout studies in the U.S., because blacks misreport voting more than others. This research tests theories that black Americans feel special pressure to vote that increases misreporting and causes turnout studies to overestimate the influence of participation in black churches, racial group consciousness, and other factors. Tests comparing results from self-reported and validated voting models indicate that studies of black turnout overemphasize the importance of participation in black churches and political efficacy, and underestimate the relative importance of income. White turnout models are not so affected, signifying that the consequences of misreporting for black turnout studies are unique.

KEYWORDS:

• Vote misreporting
• electoral participation
• black turnout
• black churches
• group consciousness

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Bernstein, Chadha, and Montjoy (2001) base their estimates on the 1980, 1984, 1986, and 1988 ANES surveys.

2 This study uses the terms “misreport” and “overreport” interchangeably, because almost all misreports are from validated nonvoters who report voting, rather than validated voters who claim not to vote.

3 The term “black church,” as typically used by scholars and the general public, refers to black Christian churches in the United States (Lincoln and Mamiya 1990).

4 Approximately 1% of blacks in the ANES surveys are Afro-Latinos.

5 A subpopulation analysis with Stata 14 subpop produced the same results but is not presented because the program does not provide an R² estimate. For an examination of subpopulation analysis of complex-sample data, see West, Berglund, and Herringa (2008).

6 Estimated from ANES data, in 1984–1988, 64% of blacks identify as Democrats and 4% identify as Republicans. In 2008, 73% of blacks identify as Democrats and 3% identify as Republicans. The correlation (r) between the 4-point partisan strength measure and 7-point partisan direction measure is .70, computed from 1984 ANES data.

7 Political empowerment through the election of black leaders, including U.S. House members, state legislators, and mayors, is an additional group resource whose influence may be overstated in black turnout models (Bobo and Gilliam 1990; Clark 2014; Gay 2001; Tate 1993). A preliminary test found that residing in a district with a black House member does not mobilize self-reported or validated presidential election voters. Political empowerment is not included in H₁ because only empowerment by black House members could be tested, as the ANES surveys do not identify respondents’ state legislative district or city of residence. In the preliminary test, black empowerment = 1 if an individual in the 1984 ANES has a black Representative in the 98th or 99th Congress, or an individual in the 1988 ANES has a black Representative in the 100th or 101st Congress; other = 0.

8 For a review of studies of church attendance overreporting, see Chaves and Stephens (2003).

9 For a review of measures of group consciousness, see McClain et al. (2009).

10 Polar affect is a second indicator of group consciousness available in the 1984–1988 ANES surveys (Miller et al. 1981). This study included polar affect in a preliminary test but removed it, as polar affect did not influence turnout and is a less prominent measure of group consciousness than group identification.

11 The linked fate measure asks whether: “What happens to black people in this country will have something to do with what happens in your life.”

12 The highest value for income is $100,000.

13 The political efficacy index for 1984–1988 combines question items that are dichotomies in 1984 and five-point scales in 1988 by coding all questions to range from 0 to 1. Dichotomies are coded “disagree” = 0, “agree” = 1; and five-point scales are coded “disagree strongly, disagree somewhat” = 0, “neither agree nor disagree” = .5, “agree strongly, agree somewhat” = 1. The items are then summed and the index is recalibrated to range from 0 to 1. The political efficacy scale for 1986–1990 combines a 0–1 dichotomy in 1986 and five-point scale ranging from 0 to 1 in 1990.

14 Zaller (1992, 335–336) concludes that neutral factual knowledge is the preferred measure of political awareness.

15 Group identification is coded 1 if respondents identify blacks as their only close group or closest group, .5 if respondents identify blacks as a group they feel close to although not the closest group and 0 if they do not name blacks as a group they feel close to.

16 By convention, admitted nonvoters are classified as validated nonvoters because self-reports of nonvoting are highly reliable, and ANES officials terminated efforts to validate self-reports of nonvoting in 1984.

17 An additional group consciousness indicator in the 1984 ANES that corresponds to a linked fate measure was tested:

Sometimes people think about other groups of people in society when they think about their own economic well-being, people who are being helped or hurt by economic conditions. When it comes to economic matters, what groups of people do you feel close to? (coded black = 1, other = 0)

This measure was not significant and was not included in the final model given its omission from the 1988 ANES.

18 Bernstein, Chadha, and Montjoy (2001) also find that white Southerners misreport more than their non-Southern counterparts and hypothesize that whites in the Deep South feel more pressure to vote and overreport voting due to racial appeals made to whites, the history of the civil rights struggle for voting in the region that heightened awareness of the importance of voting, and mobilization efforts by white evangelical leaders.

19 All studies using survey data cited in this paper examine turnout in presidential or “national” elections, including the studies of turnout in Miami-Dade County, Florida, and Georgia (Austin, Middleton, and Yon 2012; McKee, Hood, and Hill 2012). Bobo and Gilliam (1990) and Shingles (1981) examine voting participation indices that include voting in presidential and local elections, and Tate (1993) and Chong and Rogers (2005) examine voting in presidential elections and primaries.

20 Highton (2005) concludes that blacks misreport voting more than whites by comparing Current Population Survey (CPS) self- and proxy-reported turnout in 1992, 1996, and 2000. McKee, Hood, and Hill (2012) show that blacks misreport more by comparing CPS self-reported turnout and official turnout in Georgia in 2004 and 2008. Georgia is one of five states that record the race of registered voters. Bernstein, Chadha, and Montjoy (2003) examine state-level variation in vote misreporting in 1980–2000 CPS surveys and conclude that the explanation for state-level misreporting parallels the explanation at the individual level, as it includes concentrations of blacks and rates of church membership.

21 The CCES’s, first conducted in 2006, are opt-in, Internet surveys matched to random sampling with validated voting data from state voter-registration files. Ansolabehere and Hersh (2008) report that the sampling algorithm for the 2006 CCES did not garner a representative sample of blacks, raising serious concerns about using the CCES to examine black misreporting. A test of identical models of black turnout using 2008 CCES data and 2008 ANES data produced divergent results, raising additional questions about the representativeness of the CCES black sample. Church attendance did not increase self-reported (or validated) turnout using 2008 CCES data but significantly increased self-reported turnout using 2008 ANES data.