State Residency, State Laws, and Public Opinion

State Residency, State Laws, and Public Opinion

Barbara Norrander
Department of Political Science
University of Arizona
norrande@u.arizona.edu

and

Clyde Wilcox
Department of Government
Georgetown University
wilcoxc@georgetown.edu


Paper presented at the State Politics and Policy Conference, Michigan State, May 13-14, 2005.
_____________________________________________________________________________

Abstract




While considerable scholarly attention has been paid to the causes and consequences of
citizen ideology and partisanship in the 50 U.S. states, less attention has been given to attitudes
toward specific policies and issues. This paper uses the Senate National Election Surveys to
study the influence of state residency on three abortion issues, capital punishment, government
aid to African Americans, as well as partisanship and ideology. State residency can add 10 to 25
percent to the explained variance in individuals’ attitudes on these issues. Up to 25 percent of
the effects of state residency is due to state laws, which have a legitimation effect on public
opinion.

---

State Residency, State Laws, and Public Opinion


“Location, location, location,” is the mantra of real estate agents in describing the worth
of property. Echoing this theme, social scientists are increasingly readopting a geographic
perspective in analyzing public opinion and electoral outcomes. New measures of state ideology
and partisanship ignited a wealth of studies. Wright, Erikson and McIver (1985) developed their
influential measure of state ideology and partisanship by pooling national media surveys over a
number of years. Berry, Ringquist, Fording and Hanson (1998) created yearly indicators of
citizen ideology based on congressional ratings and election returns. Jones and Norrander (1996)
and Norrander (2001) demonstrate that the pooled 1988-92 Senate National Election Studies can
be used to produce measures of state ideology and partisanship as well.

State ideology has been successfully linked to a host of state policies, including laws
governing rape, alcohol, drugs, budgets, economic development, welfare, and the death penalty
(Berger, Neuman and Searles 1991; Berman and Martin 1992; Call, Nice, and Talarico 1991,
Johnson and Meier 1990; Meier 1992; Meier and Johnson 1990; Nice 1991; and Nice 1992)
State ideology also can be used to judge the behavior of state supreme court justices (Brace,
Langer, and Hall 2000). Furthermore, states are being studied for their contribution to national
outcomes, such as presidential elections (Campbell 1992; Erikson, McIver and Wright 1987;
Holbrook 1991; Jackson and Carsey 1999a, 1999b; Shelley and Archer 1994).

Because of data availability, most attention has been paid to state ideology or state
partisanship. Yet ideology and partisanship are not always successful surrogates for opinions on
a wider range of issues. Even with improvements in measuring issue positions (Sullivan,

1

---

Piereson and Marcus 1978) and a tighter connection between partisanship and ideology in recent
years (Abramowitz and Saunders 1998), the American public still cannot be described as holding
tightly constrained issue opinions that are shaped by an overarching ideology. Fortunately,
measures of state-level opinion on specific issues have been developed from the Senate National
Election Surveys (Jones and Norrander 1996; Norrander 2001 ) and the General Social Surveys
(Brace, Sims-Butler, Arceneaux and Johnson 2002). These aggregated issue attitudes have been
successfully linked to state policies on abortion (Norrander and Wilcox 1999) and capital
punishment (Mooney and Lee 2000; Norrander 2000) and to the number of women in state
legislatures (Arceneaux 2001). A complete understanding of public opinion in the states requires
the examination of a host of specific issues as well as ideology and partisanship.

State Variations in Public Opinion

As state public opinion is increasingly linked to state policies and national politics, a
better understanding of why public opinion varies by states is needed. The simplest explanation
for state-level variations in public opinion would be that compositional effects of state
electorates account for much of the variation in aggregate public opinion. In this vein, Cook,
Jelen and Wilcox (1993a) found in a study of abortion opinion in six states that interstate
differences could be accounted for by distributional differences in education, race, age and
religion.

State residency, per se, could have independent influences on public opinion in a number
of manners. State laws vary and laws have legitimizing effects on public opinion. For example,
Norrander (2000) notes approval of capital punishment is seven percentage points higher in

2

---

states with death penalty laws versus those that do not have such statutes. State laws on partisan
or nonpartisan voter registration and rules for primary participation influence levels of partisan
identification (Burden and Greene 2000; Finkel and Scarrow 1985; Norrander 1989).

Groups organize and promote their views, such that states with an actively organized
group could strengthen the opinions of their supporters and perhaps, cause countermobilization
among their opponents. In this vein, Cook, Jelen and Wilcox (1993b) in a study of abortion
opinion in 42 states found that the distribution of Catholic and Protestant residents mattered.
Protestants in heavily Catholic states became more supportive of abortion rights than Protestants
in other states.

Others attribute the influence of states on public opinion to variations across the states in
political culture. Elazar (1966) divided the states into three categories: traditionalistic,
individualistic, and moralistic. While these political cultures refer primarily to styles of
governing, they also have been linked to public policy (Lowery and Sigelman 1982) and state
opinion (Norrander 2000). Meanwhile, Erikson, McIver and Wright (1987) define “state
political culture as only that portion of state public opinion that cannot be accounted for by the
group characteristic of the state electorate” (p. 798, italic in original).

Data and Methods

State public opinion can be studied using the pooled Senate National Election Study
(Miller et al 1999). Unlike other NES surveys, the Senate study employs a state-based sampling
frame. Thus, the SNES provides representative samples of all 50 states. Jones and Norrander
(1996) demonstrate the reliability of many of the issue questions in the SNES as indicators of

3

---

aggregated opinion for the states. In this study, the concentration is in the other direction. How
much influence does state residency have on public opinion?

We adopt a common methodology for judging the influence of state residency on public
opinion (Cook, Jelen and Wilcox 1993a; Erikson, McIver and Wright 1987). We control for a
variety of demographic traits to eliminate, as much as possible, the influence of varying
population characteristics as reasons for state-level differences in opinion. Secondly, we add
dummy variables for each state (minus one) to judge the independent influence of state residency
on public opinion. We also designate as core political attitudes partisanship and ideology. These
core attitudes will be included as determinants of attitudes on specific issues (three abortion
measures, capital punishment, and government aid to blacks.)

While the SNES provides an ideal sampling frame, it provides less than ideal measures of
demographic characteristics. Basic measures of gender, race, job status, union membership, age
and income are available. The only indicator of religious view is denominational affiliation.
Still, including these basic demographic traits should account for much of the influence of an
individual’s background on public opinion.

Measures of demographic variables were constructed along the following lines.
Education is number of years of schooling (vPS0606). Income (vPS0629) and age (vPS0604)
were unchanged from their original format. Job status categories (vPS0613) were: homemaker
(codes 70 and 75), working (codes 10 though 18), unemployed (20, 40), retired (50), and student
(80). Disabled is the excluded category. Marital status designates those currently married (code
1 in vPS0605) from all others. With a large number of respondents in the SNES, racial
categories (vPS0631) were created for whites, blacks, Native Americans and Asian Americans.

4

---

The excluded category is “other” races. Latino designation includes all those with any Hispanic
background (vPS0632 codes 1 through 4). Four religion categories were created from the
denomination variable (vPS0630): no religion (codes 800 801 995), conservative Christians (i.e.,
evangelicals, Pentecostals, fundamentalists and Mormons, codes 20, 120 through 128, 133
through 148, 180 through 219, 221, 222, 250 through 269, and 301), Jews (500 through 503),
and Catholics (400). The excluded category represents mainly mainstream Protestants. Union
membership includes anyone in the family setting (vPS0617). All attitude questions were
recoded such that a high response would indicate a conservative opinion. To maintain more
respondents in the analysis, the summary ideology variable (vPS0550) is used. This measure
includes those who were reluctant to designate an ideological label in the first question but did
select a label in response to a second question.

Methodologically, OLS regression analyses were first run using only demographic
variables to explain attitudes. State residency was then added by including 49 dummy variables.
The excluded state in each analysis is the state whose citizens fall closest to the median opinion
on the issue nationwide. Thus, Oregon is the excluded category for the partisanship model,
Wisconsin for ideology, Texas for abortion attitudes, Pennsylvania for government funding of
abortions for poor women, Florida for parental consent for abortions for women under 18,
Kansas for the death penalty, and Iowa for government aid to blacks. The impact of state
residency is gauged by the change in R2 when the 49 dummy variables are added to the model
based on demographic traits. For the more specific issue questions (abortion, capital
punishment, aid to blacks), a change in R2 also is computed for the addition of the two core
attitudes (partisanship and ideology) to the demographic variable model. State residency is

5

---

added after the core attitudes. The significance of the change in R2 is indicated by the F-statistic
and its level of significance. The percent of explained variance due to demographic, core
attitudes, and state residency also is indicated in the tables.

Results

Table 1 examines the influence of state residency and demographic traits on the core
attitudes of party identification and ideology. Personal demographic traits explain 11.2 percent
of the variation in party identification. State residency adds an additional 1.4 percent to the
explanation of partisan preferences. For ideological self-identification, demographic traits
account for 5.5 percent of the total variance and state residency provides an additional 1.5
percent of explained variance. The amount of variance in individual attitudes attributable to state
residency at first appears to be quite small. The explained variances for demographic traits and
state residency reported here are similar in size to those described by Erikson, McIver and
Wright (1987) for their analysis of the pooled CBS News/New York Times polls. In addition, the
impact of state residency on party identification is more than such individual demographic traits
as income (change in R2 of .006**) and only slightly less than union membership (change in R2
of .018**).

State residency accounts for 11 (.014/.127) to 21 (.015/.070) percent of the total variance
explained. Regression coefficients for individual states (not shown) indicate that living in
Oklahoma, West Virginia or Hawaii increases attachment for the Democratic party at a
statistically significant rate (p #.05) while residing in Alaska increases Republican preferences.
For ideology, living in the South increases conservatism as the following states have statistically

6

---

significant coefficients: Alabama, Arkansas, Florida, Georgia, Louisiana, Oklahoma, South
Carolina and Virginia. Living in Nebraska also contributes a conservative influence on
ideological identification.

Public attitudes on three abortion questions are explored in Table 2. In this analysis, the
core attitudes of partisanship and ideology are included as predictive variables. Demographic
variables alone explain from 4.9 percent (parental consent) to 7 percent (abortion position) of the
overall variance in individual attitudes. Core attitudes contribute an additional 3.3 to 5.4 percent
of explained variance. State residency explains an additional 1.9 to 2.7 percent of the variation
in abortion attitudes. Living in Kentucky, Mississippi, South Dakota or West Virginia increases
the chances that an individual will express a pro-life opinion while living in Arizona,
Connecticut, Maine, New Hampshire, New Jersey, Nevada and New York increases the
likelihood that one will hold a pro-choice opinion. More conservative attitudes on government
funding of abortions for poor women are found in Louisiana and South Dakota while more
liberal attitudes are found in Nevada and Rhode Island. No state had a regression coefficient
significant at the .05 level in the parental consent model, but at the .10 level, residents in Indiana,
Nevada and Tennessee have more conservative opinions and residents of Maine, Oregon and
Vermont have more liberal opinions than their demographic traits or core attitudes would predict.
Demographic traits and core attitudes are more important than state residency in explaining
abortion attitudes, but state residency accounts for 18 to 20 percent of the total variation
explained.

Table 3 examines the same three sets of variables in explaining public opinion on capital
punishment and government aid to minorities. For attitudes on capital punishment, state

7

---

residency explains as much of the variation in individual-level opinions as do the two core
attitudes of party identification and ideology. (In fact, state residency explains a smidgeon
more.) State residency and these core attitudes each contribute a quarter to the total variation
explained. State residency accounts for 2.4 percent of the total variation in opinions on
government aid to blacks but accounts for 20 percent of the explained variance. Residents of
Florida and Oklahoma are more likely to express support of the death penalty than their
demographic and core attitude background would predict, and residents of Vermont are more
liberal. At the .10 level of significance, residency in Illinois, Missouri and Utah increases
support for the death penalty and living in Maine, North Dakota and Rhode Island decreases
support. Residency in many southern states (Alabama, Arkansas, Florida, Georgia, Louisiana,
Maryland, Mississippi, North Carolina, Oklahoma, South Carolina and Tennessee) and a variety
of other states (Idaho, Illinois, New Jersey, Nevada, and Wyoming) leads to more conservative
attitudes on government aid to minorities while only Vermont residency appears to increase
liberalism on this attitude.

Understanding the Influence of State Residency

Results from the analyses in Tables 1 through 3 indicate state residency has a significant
impact on individual-level public opinion. But what is the source of this influence? One
possibility is that state residency is simply picking up missed measures of demographic variation.
However, it is unlikely that untapped demographic variation would consistently match state
residency. State residency may impact public opinion because states have a specific political
milieu, whether it is political culture or political history. The concept of political culture,

8

---

however, is often used as a catchall for other unmeasured effects. Another possibility is that
state residency reflects concrete political conditions in a state such as variations in state laws.
Laws may provide a legitimation effect to one side of a public opinion issue. It is this last point
that can be most clearly tested.

State laws on and usage of capital punishment have been remarkably stable over time,
despite the disruption caused by the 1972 Supreme Court ruling in Furman v. Georgia. Most
states that employed capital punishment prior to the ruling simply rewrote their laws to conform
to the newer standards. After the 1970s, few states changed their position on capital punishment
and execution rates from the pre and post-Furman eras are highly correlated, with Pearson’s r =
.61, significance #.01 (Norrander 2000). Given the longevity of state policies on capital
punishment, such laws are more likely to have a legitimation effect.

To analyze the contribution of state laws to the explanation given by state residency to
public opinion, a dummy variable was developed for the presence or absence of a death penalty
law in each state in 1989. (The capital punishment question was first asked in the 1990 wave of
the SNES). The state law variable was added to the model containing demographic traits, core
attitudes and state residency variables. To avoid collinearity between the dummy variables for
state laws and state residency, the median state (in terms of public opinion) in each category of
the state law variable was excluded: Oregon for states having death penalty laws and Wisconsin
for states without such laws. The model was tested by sequentially adding demographic traits,
core attitudes, state laws and finally state residency. Changes in explained variances (R2) are
used to demonstrate the importance of each component to the model.

Results in Table 4 show that the presence of a state law contributed a small but

9

---