Efrén O. Pérez (Vanderbilt University, )

Online Appendix for “Ricochet: How Elite Discourse Politicizes Racial and Ethnic Identities” in Political Behavior.

Table A. Power Analysis Results

Anticipated effect size / Observed effect size / β / Power / Suggested n per condition / Observed n per condition
d = .60 / d = .75
(Pro-Latino)
d = .83
(Intent to register/vote) / .10 / .90 / 60 / ~ 63

Table B. Estimated Proportion of Disenfranchised Felons Among Latino Adults and Their Implications For The Survey Sample Under Analysis

Estimate 1:
Proportion of sentenced Latino prisoners in Latino adult population / Estimate 2:
Proportion of disenfranchised felons in 10 states with largest Latino populationsa
2011 number of Latino sentenced prisoners under state and federal jurisdiction (source: Carson and Sabol 2012) / 349,000 / ---
2011 number of adult Latinos in U.S. population (source: U.S. Census Bureau 2012) / 198,760,000 / ---
Proportion / .002 / ---
2010 number of disenfranchised felons in top 10 Latino U.S. states (source: Uggen, Shannon, and Manza 2012) / --- / 3,150,744
2010 number of adult Latinos in U.S. population (source: U.S. Census Bureau 2011) / --- / 197,395,000
Proportion / .016

Discussion: Given the sensitivity of the topic, I did not ask about criminal history in order to avoid affecting data quality through lower cooperation rates and/or attrition. In light of this decision, one might reason that the number of eligible unregistered Latinos in my sample is grossly inflated. While reliable data on disenfranchised Latino felons is sparse, I can use census data and national/state prison statistics to estimate the proportion of Latino adults that is ineligible to vote due to criminal history.

First, I took the number of Latino sentenced prisoners under state and federal jurisdiction for 2011 (Carson and Sabol 2012) and divided it by the 2011 estimated U.S. adult Latino population (U.S. Census Bureau 2012). This yields an estimated proportion = .002. Since my total sample is N=1,203, this figure suggests I may have missed about 2.4 individuals that are ineligible to vote.

Second, Uggen, Shannon, and Manza (2012) have compiled 2010 estimates of disenfranchised populations by U.S. states. While these data cannot be disaggregated by Latino status, I can use figures for the ten states with the largest Latino populations. This yields a crude but generous estimate of disenfranchised Latino felons in high Latino-density states. I divide this number by the 2010 estimated U.S. adult Latino population, which produces a proportion = .016. This obvious overestimate implies the true proportion of disenfranchised Latino felons is likely lower.

This latter estimate suggests that about 19 respondents in my representative sample of Latino adults (N=1,203) are ineligible to vote due to criminal history. Assuming that 19 respondents were inadvertently included in the sample under analysis (n = 192), their presence introduces systematic measurement error because they are asked to answer some questions (e.g., register to vote) that are inapplicable to them. This works against detecting any effects. Systematic measurement error decreases item variance that is attributable to the concept being assessed, which dampens the strength of associations between an affected variable and other variables (Brown 2006). If anything, then, the large effects I have unearthed are conservative estimates.

Table C. The Marginal Effect of Latino Identity in Light of Devaluing Rhetoric –

All Latinos versus Mexicans Only

Pro-Latino Attitude (A) / Pro-Latino Attitude (B) / Intent to Register and Vote (C) / Intent to Register and Vote (D)
ΔDV/
ΔLatino ID / .187
[.054, .319] / .161 / .228a
[.056, .400] / .288a

Table D. Ethnocentrism Among High Identifying Latinos: Pooled Sample vs. Unregistered Latinos

Latinos
(0-100) / Whites
(0-100) / Blacks
(0-100)
Δ Group rating/
Δ Devaluing rhetoric
(pooled sample) / 7.50
[2.23, 12.77] / .26
[-5.41, 5.93] / .88
[-4.52, 6.27]
Δ Group rating/
Δ Devaluing rhetoric
(unregistered Latinos) / 11.24 / 5.39 / .18

1