Representation and Local Policy:
Relating County-Level Public Opinion to Policy Outputs
Garrick Percival
Department of Political Science
University of California, Riverside
[email protected]
Martin Johnson
Department of Political Science
University of California, Riverside
[email protected]
Max Neiman
Department of Political Science
University of California, Riverside
[email protected]
Abstract
Scholars of comparative state politics have successfully used national survey data to create
aggregate measures of state-level ideology. Consequently, they have shown that state-level
political ideology has a profound affect on policy outputs and outcomes across a wide range of
public policy issues. In this paper, we suggest the political orientations of even more localized
mass publics affect sub-national public policy outputs. Specifically, we argue that state politics
research would be advanced by taking better account of the significant ideological variation
found within individual states.
These ideological dispositions are likely to affect policy
implemented at the local level. Using methodology similar to that used by Erikson, Wright, and
McIver (1993) at the state level, we develop measures of county-level ideology by
disaggregating statewide California Field Poll surveys, 1990-1999. We examine the relationship
between this more localized measure of political ideology and a number of policies at the local
level, including criminal justice, health care, educational services, welfare, and transportation.
Paper prepared for presentation at the annual meeting of the Midwest Political Science
Association, Palmer House Hilton, Chicago, IL, April 15-18, 2004.
Scholars of comparative state politics have long been interested in the extent to which
policy outputs of state governments reflect mass public opinion in the states. A well-established
line of scholarship investigates public opinion-policy linkages by re-aggregating to the state level
survey data collected using a national sampling frame (e.g., Wright, Erikson, and McIver 1985).
There is a clear association between states’ general political attitudes and the policy choices of
state elected policy makers (Erikson, Wright, and McIver, 1993). Increasingly, researchers are
investigating the influence of more specific policy attitudes regarding abortion rights,
environmental protection, and the death penalty (among others) on state policy outputs
(Arceneaux, 2002; Brace, Sims-Butler, Arceneaux, and Johnson, 2002; Johnson, Brace, and
Arceneaux, 2003; Norrander, 2001).
In this paper, we investigate an additional refinement to the expectations of students of
representation. We are interested in potential ideological variation within individual states –
localized ideological dispositions. First, we ask whether state politics scholars can use data
collected at the state level to estimate valid, reliable, and stable measures of mass political
orientations at even lower-levels of political geography. Second, we ask to what extent do
localized measures of political ideology influence local-level policy-making? That is, do mass
ideological beliefs at the local level constrain policy choices in ways similar to those forces
operating at the state level? Our results suggest that data from a well-established, multi-year
survey research enterprise, the California Field Poll, can be used to create reliable, valid, and
stable measures of local ideology. Further, we show that local ideology has a significant
influence on policy outputs at the local level.
Below, we review literature focused on explaining policy adoption at the local level,
develop our arguments as to why we expect local measures of ideology to impact local policy,
1
and highlight recent work that guides our methods used in this research. We then present the
specific hypotheses tested, and conclude with our empirical results and discussion.
Representation and Local Politics
We are interested in the potential influence of public opinion on local policy making.
Students of representation have demonstrated the influence of public opinion on national policy
outputs (Wlezien, 1995, 1996; Stimson, MacKuen, and Erikson, 1996). As noted above, scholars
have also found clear associations between state policy decisions and general political
orientations (Erikson, Wright, and McIver, 1993; Hill and Hinton-Andersson, 1995; Hill and
Leighley, 1996) and more specific public demands. Early research on representation focused on
smaller, localized constituencies, identifying the conditional influence of public opinion on the
voting behavior of members of Congress (Achen 1978; Erikson 1978; Miller and Stokes 1963).
However, little work has systematically investigated policy responsiveness to public opinion on
local politics (but see Berkman, Plutzer, and Boydstum 2003).
Local ideology may have a significant impact on local policy processes, and affect policy
in a way similar to the influence of state political orientations on state policy outputs. This
expectation is largely driven by the effect of increased urbanization among counties, changing
institutional relationships between state and county governments, and the changing nature of
policies adopted and implemented by county governments. More specifically, as counties have
become increasingly urbanized, the demand for services has also increased. In response, the
breadth of services offered by county governments has increased since the early 1980s (Benton
and Rigos, 1985; Martin, 1990; Cigler, 1990; Ciglar, Mezel, and Benton, 1990, 1991). Overall,
research shows counties have expanded their services from the more traditional services like
property tax assessment, law enforcement, and elections to other additional services including
2
health care, educational services, pollution control, and mass transit among others (Lawrence and
Degrove, 1976; Ciglar, Mezel and Bention, 1990). Ideological divisions often characterize many
of these “new” policies being adopted and implemented at the county level. Moreover, in order
to carry out increasingly complex functions, state governments have tended to increase the
amount of policy discretion and hence decision making authority to county governments
(Bowman and Kearney, 1986; Martin, 1990).
Significant discretion in policy decision-making may allow counties’ ideological
disposition to affect policy in a couple of ways. Street-level bureaucrats in more
conservative/liberal counties may be more likely to be ideological conservatives/liberals and
therefore work to implement policies in ways that closely reflect their ideological preferences
(see Lipsky 1980 for a discussion of the significance of street-level bureaucrats in policy
implementation). In addition, local elected officials (e.g. County Boards of Supervisors)
operating in diverse political environments are likely to try to adopt policies that reflect the
ideological preferences of their constituents which works to increase the chance of reelection.
Thus, the added likelihood that county governments adopt and implement ideological divisive
public policies, and the added decision making authority given to county governments in the
policy making process, will likely increase the chances that local ideological dispositions will
significantly influence local policy adoption and implementation.
However, there is also a strong case to be made for the null model – that public opinion
has at best a muted effect on local policy making. U.S. states have legal Constitutional authority
over their own policy spheres and thus have significant policy discretion across a broad range of
public policy domains. This policy discretion allows state-level political values to play a
significant role in the state policy making process. Conversely, local governments, (or more
specifically county governments that are the focus here) are mere creatures of the state where
3
theoretically, following “Dillon’s Rule,” states could simply terminate local governments. Local
governments then, possess only those powers expressly given to them by their states’
governments. Indeed historically, county governments were created as mere administrative arms
of the state, whereby it was believed county governments could best manage state concerns at the
local level (Bowman and Kearney, 1986). This perception that county governments, being
merely administrative arms of the state implementing states’ policy mandates in Weberian-like
fashion, lends support to the expectation that local politics, or local ideological dispositions more
generally should have little impact on policy, given the relatively little amount of policy
discretion given to county governments in particular.
Additionally, Peterson’s (1981) seminal study of urban policy making implies politics has
only a small role in shaping city policy outputs. This does not suggest local officials are
apolitical, but rather the policy domains in which local governments have authority are
inherently characterized by little political conflict. Peterson’s central point rests on the idea that
local governments are constrained by the American federal structure in what they can do, and
because of these constraints, local governments are primarily involved with simple
“housekeeping” duties and policy-making that stimulates economic growth. These policies are
often non-controversial and widely popular, bringing little political conflict.
More specifically, Peterson (1981) divides public policy into three primary categories—
developmental, allocational, and redistributive, where some policy areas more than others are
characterized by a high degree of political conflict. Developmental policies work to enhance the
local tax base and generate additional resources that can be used to help the welfare of the city.
These policies, according to Peterson, are widely popular, and because of this, political conflict
tends to be low. Another primary policy function of local government is associated with what
Peterson calls allocational policies—or simple housekeeping functions like street sweeping,
4
garbage collection, community policing and fire protection, among others. Political conflict
associated with allocational policies also tends to be low as policy adoption/change is typically
incremental, determined largely by standardized formulas (Waste, 1988). Finally, redistributive
policies refer to those policies that redistribute wealth from those who are better off to those who
are worse off. These policies have a relatively high degree of political conflict because of their
harmful nature to the local economy. For rational economic reasons, Peterson shows local
governments are less likely to adopt redistributive policies (as it hurts economic growth and may
attract a greater amount of impoverished people), where political conflict is high, but rather are
more likely to pursue developmental or allocation policies—both areas with low levels of
political conflict. Extending this argument suggests that local governments are less likely to play
an active role in areas of policy where politics matters most, and more likely to adopt and
implement policies where politics matters less. Because of this, it might be expected that
ideological values play an insignificant part in the local policy adoption.
Measuring Political Orientations at the Local Level in California
It has often noted that local communities tend to share distinct sets of beliefs, ideas, and
values that are likely a result of shared historical experiences in a given locale. Scholars of local
politics have long been interested in capturing the impact of these shared beliefs on the policy
making process. Attempts at measuring shared beliefs among communities have led to concepts
such as local “political culture” or “political ethos” (see Banfield and Wilson, 1963). Such terms
have been helpful in distinguishing between different communities and their respective values,
but the concepts suffer from inherent vagueness and are not reliably measured. Other scholars
have used innovative simulation techniques to create measures of school board ideology across
the states, but these measures too, are not ideal, given they rely on simulations (Cite needed)
5
To create a measure of local political ideology, we advance the argument that scholars of
local politics can use the methods similar to those of Erikson, Wright and McIver (1993).
These scholars significantly advanced the understanding of state-level public opinion by creating
political ideology and partisanship measures by pooling 1976-1988 nationally sampled
CBS/New York Times polls and re-aggregating them to the state level. Importantly, these
measures were shown to be reliable, stable, and valid.
Here, we use similar methodology to create measures of county-level ideology by
disaggregating to the county-level statewide California Field Poll surveys conducted between
1990-1999. The California Field Poll offers researchers an excellent tool that can be used to
create measures of county-level ideology. Established in 1947, and continuing every year since,
the Field Poll routinely fields surveys questions to California residents on a wide range of public
policy issues and questions regarding their support for various political candidates and the
national, state and local levels of government.1 Data were gathered from 48 Field Poll surveys
totaling 51,930 individual respondents. Fortunately for our purposes here, the Field Poll
consistently asks respondents to place themselves along a 3-point political ideology continuum.
Specifically, respondents were asked, “do you consider yourself to be politically conservative,
liberal, middle-of-the road, or don’t you think of yourself in this way.” Conservatives were
coded 100, middle-of-the-road 0, and liberals –100. In addition, the Field Poll asks each
respondent his or her county of residence, allowing us to link each response to a given county.
Individual responses were then aggregated to create ideological scores for California’s 58
counties. The number of cases in each county ranged from 13,873 in Los Angeles county to 17 in
1
The Field Poll uses samples of the California telephone household population drawn from random digit dial (RDD)
samples of Survey Sampling Incorporated. The sample is a stratified sample of California counties where samples
are systematically stratified to all counties in proportion to each county’s share of telephone households in the
survey area. Further sampling information can be referenced from the California Field Poll Code Books 1990-1999.
6
Alpine county (mean=659.01). Ideology scores ranged from the most conservative Madera
county (41.26) to the most liberal San Francisco county (-25.35) with a mean=21.28.
Auditing the county-level measure of ideology
Individual responses are treated here as aggregate data, and therefore it is not appropriate
to use standard measures of individual-level reliability like Cronbach’s alpha (Brace et al., 2002).
Because of this, Jones and Norrander (1996) recommend testing reliability analysis on the basis
of aggregate units, and not individuals. To first test the generalizability of the ideology measure,
we use the O’Brien coefficient (O’Brien, 1990). Presented by Jones and Norrander (1996), the
O’Brien generalizability test seeks compares within-unit variance to the across-unit variance2
while taking into account sample size (Norrander, 2001: 113). Using this method, measures of
ideology will be more generalizable across units with less intra-county variation and more
variation in ideological dispositions inter-county. An O’Brien generalizability coefficient that
exceeds .70 is considered to be highly generalizable, and values between .60 and .70 are
considered to be moderately generalizable. The O’Brien coefficient for the county-level
ideology measure is .96.
An additional test of reliability is the split-half approach used by Erikson, Wright, and
McIver (1993). The split-half approach involves splitting the Field Poll sample into two subsets
by assigning odd-year surveys to one subset and even years to the other. Mean scores for county
ideology were calculated for each subset and correlated using Pearson’s r coefficients. The
Spearman-Brown prophesy formula was used to assess the reliability of each measure:
2
O’Brien’s (1990) generalizability coefficient for the R:A Design contemplates the mean square, an estimate of the
population variance between aggregate units, MS(a), and the mean square for individual-level scores within the
aggregated units, MS(r:a), using the formula:
Eρˆ 2 =
[ MS (a) − MS (r : a )]
MS (a)
7
2r12
,
1 + r12
where r12= the Pearson’s r correlation between the split-halves. Reliability scores of .70 and
above are considered reliable and those between .60 and .70 are considered moderately reliable,
and those below .60 are considered unreliable (Jones and Norrander, 1996). The SpearmanBrown coefficient for the reliability of the county-level ideology measure equals .60.
To test the stability of the measure, the Field Poll sample was divided into “early” and
“late” subsets. The early subset included survey years between 1990-1995 and the late subset
between 1996-1999. Mean scores for county ideology were calculated and correlated. The
Spearman-Brown coefficient for the stability of county-level ideology was .62. In sum, the
assessment of the reliability of the ideology measure is mixed. The O’Brien measure is highly
reliable although the Spearman-Bowman coefficients using the split-half approach is at the low
end of scores considered to be “moderately” reliable. We have chosen to use the ideology
measure here, but at the same time make note of its possible deficiencies.
How well do the re-aggregated Field Poll data represent county-level California demographics?
California counties are not the population of interest for the Field Poll, and thus we
cannot assume that the sampling frame employed will produce representative estimates of county
population. As noted by Hill and Hurley (1984), a sample bias may be introduced when creating
a nonrandom sample from state residents, and therefore caution must be used. To test the validity
of the Field Poll sample a series of demographic characteristics were derived from the Field Poll
sample and correlated with county demographic characteristics collected by the U.S. Census (see
Brace et al., 2002). Results presented in Table 1 show that county samples obtained from the
Field Poll are remarkably representative. Specifically, we find a strong correlation between the
MS(a) and MS(r:a) were estimated using the One-Way ANOVA
procedure in SPSS.
8
educational attainment of the sample and educational attainment reported by the U.S. Census in
1990 and 2000. A similarly strong relationship is found among between the income of Field Poll
respondents and U.S. Census statistics. Racial characteristics of respondents, although showing
a slightly weaker correlation to U.S. Census figures than do the education and income figures,
are moderately strong nonetheless. Importantly, the strong correlations for education and
income, and the moderately strong correlations for the race variables suggest that the Field Poll
samples reflect county populations.
[Table 1 about here]
Ideology and Local Policy Outputs in California
In order to examine the relationship between policy and the ideological disposition of
California counties, we collect data on a series of policy outputs implemented at the county level
which serve as dependent variables. These are drawn from a number of different policy areas
including those associated with criminal justice, educational services, public health, welfare
payments, and transportation. We chose these policy categories for two primary purposes. First,
these are all policy areas where county governments are actively involved in the policy making
process. Second, following the work of Peterson (1981), these policy areas are likely associated
with different degrees of political conflict, and hence, the extent to which ideological
dispositions influence policy making at the county-level should also vary across policy issues.
Policies associated with criminal justice, educational spending, and welfare payments are often
ideologically divisive and therefore more likely to be influenced by local ideological
dispositions. In contrast, allocational policies, like road construction and repair are less likely
bring political conflict and therefore we might expect ideology to matter less on these and other
9
“allocational” policies. In short, incorporating a wide range of policy issues into our analysis
allows us to test for differentiating effects of ideology across issues.
Policy Indicators
Three dependent variables are included as criminal justice variables. The first is a
measure of incarcerations per 1,000 persons in each county for drug-related offenses averaged
between 1996-1999. This measure has been used elsewhere to investigate the extent to which
local contextual characteristics influence the implementation of California’s Substance Abuse
and Crime Prevention Act (Percival, 2004). Drug offenses in this measure include those
incarcerations for drug sales, manufacturing, possession in quantities large enough to presume
intent to sell, and low-level drug possession. Data is gathered from the Data Analysis Unit,
California Department of Corrections. A second measure includes counties’ average per-capita
spending for public defenders offices, 1995-1999. These expenditures are used to pay for
expenses related to clients’ use of legal services provided by counties’ public attorneys. Finally,
we include a county per-capita measure of the number of individuals incarcerated (in 1999) on
second or third strikes related to California’s three strikes law (Proposition 184) enacted by the
voters via ballot initiative. This measure is created by summing the total number individuals in
each county incarcerated for second or third strike related offenses in that year and dividing by
the total population of each county. Data is collected from the California Department of
Corrections.
We also use a series of indicators for social service policies and transportation. Counties
average per-capita welfare cash grants between 1995-1999 serves as a good measure of
redistributive policy adopted and implemented at the county-level. Data is collected from the
California Institute of County Governments. Dependent variables related to public health include
the total per-capita expenditures on health care and per-capita county expenditures on more
10
specific drug treatment programs between 1995-1999. Data is collected from the California
Institute of County Governments and California’s Department of Alcohol and Drug Treatment
Programs. Counties total per-capita expenditures on road construction and repair between 19951999 is incorporated as a dependent variable that Peterson (1981) would consider an allocational
policy. Finally, we include a measure of average county per-capita expenditures on library
services between 1995-1999. This variable serves as a measure of educational services provided
by county governments.
Hypotheses
Our central hypothesis is that counties’ ideological make-up will influence policy outputs
at the county level where we expect policy to closely reflect ideological preferences of counties’
residents. With respect to the specific dependent variables under consideration, we expect more
ideological conservative counties to incarcerate more individuals for drug-related crimes and
more likely to instill tough second and third strike penalties on criminal behavior, leading to a
greater number of incarcerations. In general, since the early 1980s, California’s approach to
fighting crime has followed a deterrence and incapacitation theory, which promotes increased
arrests, stricter probation and parole monitoring, mandatory sentences, and higher rates of
incarceration to dissuade street crime by removing offenders from the larger community
(Maxwell, 1999; Tonry, 1999). Given this however, we expect approaches to fighting crime to
vary within counties with some being “tougher” on crime than others. Because traditionally,
ideological conservatives have been more likely to support these punitive approaches, while
liberals have tended to favor more rehabilitative models (see Beckett, 1999), we expect drugrelated and two and three strike related incarcerations to be higher in more conservative counties.
11
It is also expected that more liberal counties are likely to expend greater amounts of
funds toward public defense than more conservative counties. This follows traditionally,
ideological liberals have pushed for greater spending to assist the impoverished. Increased
spending to aid the impoverished in the legal system is one such example, where it is often
argued in popular political debate that poor citizens have historically received bad legal council
as they move through legal processes.
Traditionally, ideologically liberals have been more likely to believe that solving drug
abuse and drug-related crime is best done via the treatment process, whereas ideological
conservatives have tended to support tougher punitive measures (Beckett, 1999). Following
this, we expect ideological liberal counties to be more likely to expend greater funds toward drug
treatment than conservative counties. With respect to public health expenditures, we expect
more liberal counties to have greater per-capita funding as this follows liberals traditional
support for a more active governmental role in the health care process.
Welfare politics has long been an ideological and partisan issue. Traditionally,
conservatives push for more restrictive welfare policies whereas liberals push for greater benefits
with fewer restrictions (Rom, 1999). Empirical research at the state level has found greater
welfare spending among states with more liberal publics and political elites (Erikson, Wright,
McIver, 1993; Hill, Leighly, and Hinton-Andersson, 1995). Following this, we expect more
liberal counties to have greater welfare spending.
We also expect more liberal counties to have greater spending on library services. This
follows traditional support among ideological liberals for higher rates of education-related
spending, and prior research at the state level which has shown greater education spending in
more ideologically liberal states (Norrander, 2001).
12
Finally, we expect ideology to have little or no effect on counties’ road construction and
repair expenditures. Road construction and repair can be considered in Peterson’s (1981; 1995)
terms – an allocational policy. Accordingly, allocational policies are generally associated with
lower levels of political conflict and therefore we expect ideology to have little impact.
Table 2 presents bivariate regression models assessing the relationship between the
county ideology measure and each dependent variable. Ideology has a significant relationship
with five of the eight dependent variables and the direction of the association in each case is in
the predicted direction. Conservative counties have a larger number of drug-related and threestrikes incarcerations, and fewer expenditures per capita on public defenders, library services,
and health care. As expected, there is no significant association between ideology and
expenditures for road construction and repair. Surprisingly, no significant correlation is found
between ideology and welfare spending and drug treatment expenditures.
[Table 2 about here]
Although the bivariate models presented in Table 2 are enlightening, in order to properly
test the ideology-policy linkage we must control for other possible confounding variables.
Literature focused on the local policy making process suggests several other possible factors that
may influence outputs.
Socioeconomic Characteristics. At the state level, fiscal capacity is a strong predictor of
policy outputs (Dye, 1966; 1979). Studies of subnational governments show that economic
factors like per-capita income of certain jurisdictions or locales are strong predictors of public
sector expansion (Feiock and West, 1993). Education levels are also important, where more
educated publics are more likely to be politically involved, and better able to finance public
programs. In the models presented below, we incorporate for each county, measures of median
household income, and the percentage of residents who have earned a high school diploma or
13
higher. Data is drawn from the 2000 U.S. Census. It might be expected that counties with
greater median incomes and educational attainment levels would have a greater capacity to
increase spending on drug treatment programs, library services, and public health. We expect
counties with higher socioeconomic status to be less likely to incarcerate drug offenders and less
likely to apply second and third strikes. These expectations are largely driven by the assumption
that counties with higher socioeconomic have lower violent crime rates and lower rates of drug
abuse.
Community Needs. In its simplest form, the needs based approach to explaining policy
adoption and outputs suggests that local governments will adopt or implement policies in
response to an objective need for the policy. Indeed, a small body of research shows clear links
between need and adoption of economic development incentives, groundwater protection
programs, drug treatment expenditures (Rubin and Rubin, 1987; Lester and Kepter, 1984; Meier,
1990). Following this, counties where the drug abuse problem is more severe may be more
likely to expend greater funds toward drug treatment. Policy needs are also likely to influence
counties welfare spending. Counties with lower economic status may have a greater need for
welfare expenditures, motivating elected political leaders to respond to these demands. Yet, as
Peterson, 1981; 1995) argues, policy makers operating in environments where the need for
redistributive policies may be greatest, may actually be less inclined to provide benefits for fear
of attracting more impoverished people and dampening opportunities for economic growth. To
account for measures of policy need to predict welfare spending and drug treatment expenditures
across counties, we incorporate measures of county median income levels and the average
number of drug deaths caused by drug overdose in 1999 and 2000.
Race. Increasingly, political scholars point to racial diversity within any given
environment as a significant predictor of public policy, and how public programs are distributed
14
(Hero, 1998). Racial diversity may impact policy in a couple ways. Geographic proximity to
large number of racial minorities may increase sentiments of racial threat among whites (Key,
1949; Stein, Post, and Rinden, 2000). Perceptions of racial threat tends to decrease support of
policies perceived to help minority members (Stein, Post, and Rinden, 2000). Moreover, when
public programs are perceived to target minority groups, program allocations tend to become less
generous (Katz, 1989). Research of most relevance here shows racial minorities are more likely
to be incarcerated than whites, and local governments are tend to impose tougher sanctions and
fewer benefits to welfare recipients living in racially diverse contextual environments (Meier,
1990; Keiser, Mueser, and Seung-Whan Choi, 2004). To control for the possible influence of
race on the dependent variables, we include a measure of the percentage of black residents
residing in each county. Data is drawn from the 2000 U.S. Census.
Multivariate Results
Table 3 provides OLS regression results of eight dependent variables regressed on
selected predictors. Overall, the results are clear--county-level ideology appears to influence
local policy outputs across a wide range of policy issues, even after controlling for other possible
predictors. Ideologically conservative counties are more likely to incarcerate individuals for
drug-related offenses, in addition to having higher rates of incarceration for second and third
“strike” related criminal acts. These results closely reflect conservatives traditional support for
more punitive crime fighting policies. Counties with higher poverty rates and a higher
percentage of black residents are also associated with higher rates of second and third strike
incarcerations. This matches other work done on the implementation of the California’s “three
strikes” initiative that shows tougher sanctions are disproportionately imposed on the poor and
racial minorities (Macalliar and Males, 2003).
15
Results in Table 3 also indicate that ceteris paribus, liberal counties are more likely to
expend greater funds for public defenders, library services, drug treatment, and general health
care than those counties that are more ideological conservative.
[Table 3 about here]
The relationship between the ideology measure and all four of these dependent variables
are in the expected direction with each model explaining a substantial amount of variation in the
dependent variables. The debate over assistance to the poor (e.g. public defense expenditures),
education-related expenditures such as library services, drug treatment, and the scope of
government involvement in health care more generally are often structured by ideological beliefs
and attitudes. Given this, it is not surprising to find that local ideological dispositions impact
outputs across these policy areas. Several additional controls have a significant influence on the
dependent variables as well. Surprisingly, counties where the drug problem is more severe spend
less on drug treatment. It may be that counties where residents have greater drug problems are
more likely to take a tough punitive approach to drug use rather than a more rehabilitative one
associated with treatment, although more research is needed to better understand this
relationship. Higher educational levels are associated with increased spending on library
services, health care, and public defense, while income is negatively associated with drug
treatment, and health care. The percentage of black residents is associated with decreased health
care spending.
After controlling for additional variables we find ideology to be significantly related to
counties’ welfare spending but not to road construction and repair. The negative coefficient on
the ideology measure and welfare spending indicates more conservative counties, on average
provide lower welfare payments than do liberal counties. Similar to results found among the
U.S. states, where states’ ideological dispositions are found to influence welfare payments,
16
county-level ideology significantly influences payments at the local level. Further, counties’
income is negatively associated with welfare payments while the percentage of black residents
increases spending. The former result lends support to Peterson’s argument that locales with
higher socioeconomic status have little incentive to adopt and implement redistributive policies.
The positive coefficient on the race variable conflicts with prior research on race and welfare,
although the results presented here should be taken with caution given a more complex diversity
index (see Hero, 1998) might produce different results.
The insignificant relationship between ideology and counties’ road construction and
repair expenditures also matches our expectations. This finding lends support to Peterson’s
(1981) assertion that allocational policies like road repair, garbage collection, and street
sweeping cause little political conflict and as we advance here, less likely to be influenced by
local ideological dispositions. Significantly, these results suggest that the importance of
ideology on local level policy making will be dependent, to some extent, on the policy issues
under consideration.
As noted above, the split-halves coefficients derived from the Spearman-Bowman
coefficients were found to be moderately reliable. To further test the reliability of our findings in
Table 3, we conduct separate regressions, using identical independent and dependent variables,
but including in the sample only those counties where ideological scores are created with a reaggregated sample size of 50 or greater.3 A larger sample size within counties can ensure more
reliable measures, although depending on the distribution of demographic characteristics within
geographic units, reliable and valid measures can be drawn with a relatively small N (see Brace
et al., 2002). If regression results that incorporate only those counties where ideology scores
were calculated with an N greater than 50 are similar to those using all 58 counties, we can be
17
more certain our findings (presented in Table 3) accurately reflects the influence of county
ideology on the local policy making process.
Table 4 presents regressions of county policy as a function of ideology and additional
control variables, but includes only those counties with a sample size greater than 50.
[Table 4 about here]
Overall, the relationship between ideology and the dependent variables are similar to those
presented in Table 3. Importantly, county ideology remains a significant predictor of both
incarceration measures, public defense spending, library services, drug treatment, health care,
and welfare, although these latter two are significant at the p<.10 level.
Conclusion
The results presented here suggest students of state politics would benefit from paying
additional attention to the ideological variation within individual states as well as variation
between states. Some important implications of our work is that state politics scholars can use a
well-established, multi-year state survey instrument like the California Field Poll to create
reliable, valid, and stable measures of local ideology. Using similar methodological tools
outlined here, scholars can advance our understanding of the way local public opinion influences
local policy making and the extent to which local policy represents the political interests of
diverse populations. We further show that local policy outputs are constrained by counties’
ideological dispositions where more liberal/conservative counties produce more
liberal/conservative outputs across a wide range of policy areas including criminal justice, public
health, and educational services. Moreover, the influence of counties’ ideology on outputs varies
across different policy issues with ideology playing a more important role on those issues
3
Ten counties are excluded from the sample with an N<50. Excluded counties include Alpine, Inyo, Del Norte,
18
characterized by a higher degree of political conflict, and a less important role on issues where
little conflict occurs. Scholars may also profitably use other statewide survey data to construct
local measures of ideology or perhaps even more policy-specific attitudes that can be used to
further our understanding the public opinion-policy linkage at the local-level.
Glenn, Mariposa, Modoc, Mono, Plumas, Sierra, and Trinity.
19
References
Achen, Christopher H. 1978 “Measuring Representation.” American Journal of Political Science
22:475-510.
Arceneaux, Kevin. 2001. “The ‘Gender Gap’ in State Legislative Representation: New Data to
Tackle and Old Question.” Political Research Quarterly 54:143-160.
Banfield, Edward, and James Q. Wilson. 1963. City Politics. Cambridge: MA, Harvard and
MIT Press.
Beckett, Katherine. 1997. Making Crime Pay: Law and Order in Contemporary American
Politics. New York, NY: Oxford University Press.
Benton, Edwin J., and Platon Rigos. 1985. “Patterns of Metropolitan Service Dominance:
Central City and Central County Service Roles Compared.” Urban Affairs Quarterly
20:285-302.
Berkman, Michael B., Eric Plutzer, and Amber Boydstum. 2003. “The Consequences of State
Institutions for Local Policy Responsiveness: Public Preferences and School Budget
Referendum.” Presented at the Third Annual State Politics Conference, Westward Look
Resort, Tucson, Arizona, March 14-15, 2003.
Brace, Paul, Kellie N. Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Measuring Public
Opinion in the American States: An Expanded Range of Aggregated Measures, 19741998.” American Journal of Political Science 46:173-189.
Bowman, Ann O’m, and Richard C. Kearney. 1986. The Resurgence of the States. Engelwood
Cliffs, NJ: Prentice Hall.
Cigler, Beverly A. 1990. “County Government: A Century of Change. In the Municipal
Yearbook 1989. Washington D.C.: International City Management Association, 55-65.
Dye, Thomas R. 1966. Politics, Economics, and Public Policy: Outcomes in the American
20
States. Chicago, IL: Rand McNally.
____. 1979. "Politics and Economics: The Development of the Literature on Policy
Determination. Policy Studies Journal June, 652-662.
Erikson, Robert S. 1978. “Constituency Opinion and Congressional Behavior: A Reexamination
of the Miller-Stokes Representation Data.” American Journal of Political Science
22:511-35.
Erikson, Robert S., Gerald C. Wright, and John P. McIver. 1993. Statehouse Democracy: Public
Opinion and Policy in the American States. Cambridge: Cambridge University Press.
Feiock, Richard C., and Jonathan West 1993. “Testing Competing Explanations for Local Policy
Adoption: Municipal Solid Waste Recycling Programs. Political Research Quarterly
46,2: 399-419.
Hero, Rodney E. 1998. Faces of Inequality: Social Diversity in American Politics. New York:
Oxford University Press.
Hill, Kim Quaile, and Angela Hinton-Andersson. 1995. “Pathways of Representation: A Causal
Analysis of Public Opinion-Policy Linkages.” American Journal of Political Science 39:
924-35.
Hill, Kim Quaile, and Jan E. Leighley. 1996. “Political Parties and Class Mobilization in
Contemporary United States Elections.” American Journal of Political Science 40:787804.
Johnson, Martin, Paul Brace, and Kevin Arceneaux. 2003. “Environmental Regulation,
Conditions, and Consumer Attitudes in the American States.” Presented at the Third
Annual State Politics Conference, Westward Look Resort, Tucson, Arizona, March 1415, 2003.
21
Jones, Bradford S., and Barbara Norrander. 1996. "The Reliability of Aggregated Public
Opinion Measures." American Journal of Political Science 40: 295-309.
Katz, Micael R. 1989. The Undeserving Poor: From the War on Poverty to the War on Welfare.
New York. Pantheon.
Keiser, Lael., R., Peter R. Mueser, and Seung-Whan Choi. 2004. “Race, Bureaucratic
Discretion, and the Implementation of Welfare Reform.” American Journal of Political
Science 48: 314-328.
Key, V.O. 1949. Southern Politics in State and Nation. New York. Alfred A. Knopf.
Lester, James P., and Patrick Keptner. 1984. “State Budgetary Commitments to Environmental
Quality Under Austerity.” In Western Public Lands, eds. John G. Francis and Richard
Ganzel. Totawa, NJ: Rowman and Allanheld.
Lipsky, Michael. 1980. Street-level Bureaucracy: Dilemmas of the Individual in Public
Services. New York: Russell Sage Foundation.
Martin, Lawrence L. 1990. “States and Counties: Adversaries or Partners? The Florida
Perspective.” Paper presented at the Annual Conference of the American Society for
Public Administration. Los Angeles 7-11.
Maxwell, Sheila R. 1999. "Conservative Sanctioning and Correctional Innovations in the United
States: An Examination of Recent Trends." International Journal of the Sociology of
Law. 27: 401-412.
Meier, Kenneth J. 1994. The Politics of Sin: Drugs, Alcohol and Public Policy. Armonk, NY:
M.E. Sharpe, Inc.
Miller, Warren E., and Donald E. Stokes. 1963. “Constituency Influence in Congress.”
American Political Science Review 57: 45-56.
22
Norrander, Barbara. 2000. “The Multi-Layered Impact of Public Opinion on Capital Punishment
Implementation in the American States.” Political Research Quarterly 53: 771-94.
_____. 2001. “Measuring State Public Opinion with the Senate National Election Study.” State
Politics and Policy Quarterly 1:111-25.
O'Brien, Robert M. 1990. "Estimating the Reliability of Aggregate-Level Variables on
Individual-Level Characteristics." Sociological Methods and Research 18:473-504.
Rom, Mark C. 1999. “Transforming State Health and Welfare Programs.” In Politics in the
American States, ed. Virginia Gray and Herbert Jacobs. Washington D.C.: CQ Press.
Page, Benjamin I., and Robert Y. Shapiro. 1983. “Effects of Public Opinion on Policy.”
American Political Science Review 77:175-90.
Percival, Garrick L. 2004. “The Influence of Local Contextual Characteristics on the
Implementation of a Statewide Voter Initiative: The Case of California’s Substance
Abuse and Crime Prevention Act.” A paper presented at the Annual Meeting of the
Southern Political Science Association. January 2004.
Peterson, Paul A. 1981. City Limits. Chicago, IL: University of Chicago Press.
Rubin, Irene, and Harold Rubin. 1987. “Economic Development Incentives: The Poor (Cities)
Pay More. Urban Affairs Quarterly 23:39-57.
Stimson, James A., Michael B. MacKuen, and Robert S. Erikson. 1995. “Dynamic
Representation.” American Political Science Review 89:543-65.
Tonry, Michael. 1999. "Why Are U.S. Incarceration Rates So High?" Crime and Delinquency.
45: 419-437.
Waste, Robert J. 1988. The Ecology of City Policymaking. New York: Oxford University
Press.
23
Wlezien, Christopher. 1995. “The Public as Thermostat: Dynamics of Preferences for
Spending.” American Journal of Political Science 39: 981-1000.
_____. 1996. “Dynamics of Representation: The Case of U.S. Spending on Defence.” British
Journal of Political Science 26:81-103.
Wright, Gerald C., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship
and Ideology with Survey Data.” Journal of Politics 47: 469-89.
_____.1987. “Public Opinion and Policy Liberalism in the American States.” American Journal
of Political Science 31: 980-1001.
24
Table 1. Representativeness of Field Poll County Samples 1
Education
Income
White
Black
Asian
Democratic Party
Registration
2000
.91**
.90**
.75**
.87**
.86**
.76**
1990
.91**
.92**
.76**
.88**
.87**
.78**
Average 1990/2000
.91**
.92**
.76**
.88**
.87**
.77**
1. ** p<.01 Education is measured by the percentage of county residents who have earned a
bachelors degree or higher. Income is measured by correlating the percentage of Field Poll
respondents who mentioned their total household income was between $20,000-$40,000 dollars
and the median household income of the respondent’s county reported by the U.S. Census.
Racial characteristics are based on sample estimates drawn from self-reported information from
the Field Poll and are correlated with U.S. Census data. Democratic Party registration is based
on the percentage of Field Poll respondents who identified themselves as members of the
Democratic Party and correlated with voter registration data housed by the California Secretary
of State.
25
Table 2. Modeling County-level Policy Outputs as a Function of County Ideology
Drug-related
Incarcerations
All Counties
Pearson’s
b
r
(s.e.)
.32*
.01*
(.003)
R2=.11
Counties with N>50
Pearson’s
b
r
(s.e.)
.40**
.01**
(.003)
R2=.16
“Three Strikes”
Incarcerations
.28*
.09*
(.00)
R2=.08
.31*
.09*
(.00)
R2=.10
Public Defense
-.27*
-.16*
(.08)
R2=.08
-.40**
-.24**
(.08)
R2=.16
Drug Treatment
Expenditures
-.17
-.13
(.108)
R2=.02
.14
-.05
(.06)
R2=.02
Library Services
-.48***
-.61***
(.159)
R2=.23
.37**
-.16**
(.06)
2
R =.14
Welfare
.07
.27
(.760)
R2=.01
.16
.95
(.85)
R2=.02
Road
Construction
-.15
-2.1
(1.92)
R2=.02
.24
.82
(.50)
R2=.05
Health Care
Spending
-.34**
-1.00**
(.380)
R2=.12
-.29*
-.39*
(.19)
R2=.08
***p<.001, **p<.01, *p<.05
26
Table 3. County Policy as a Function of County Ideology, Control Variables (All Counties)
Drug
Jailings
b
(s.e.)
.01**
Three
Strikes
b
(s.e.)
.01**
Public
Defense
b
(s.e.)
-.184**
Library
Services
b
(s.e.)
-.66***
Drug
Treat
b
(s.e.)
-.98**
Health
Roads
Welfare
b
(s.e.)
-.96*
b
(s.e.)
1.59
b
(s.e.)
-1.57**
(.002)
(.00)
(.088)
(.000)
(.331)
(.387)
(1.33)
(.508)
Income
-.00*
(.000)
--
-.00
(.00)
.01*
(.000)
.-001**
(.000)
-.001*
(.001)
-.004*
(.002)
-.004***
(.001)
Education
.00
(.007)
.00
(.002)
.36**
(.188)
.64*
(.313)
.94
(.652)
1.58*
(.773)
3.99
(2.420
-2.97**
(1.09)
1.29
.004**
-17.59
-83.88
40.54
-289.6*
-927.5*
548.58*
(1.04)
(.00)
(35.91)
(60.191)
(121.96)
(146.82
(460.2)
(209.59)
-.002*
(.01)
.00
(.00)
--
--
--
--
--
--
--
.02**
(.000)
--
--
--
--
--
364.31**
(99.38)
--
--
County
Ideology
% Black
Age
Poverty
Drugs
Constant
N
R2
F(d.f.)
--
1.75***
(.390)
57
.45
8.07**
(55)
--
-.00
(.001)
58
.32
4.92**
(56)
--
1.01
(13.54)
57
.18
2.75**
(55)
--
5.37
(22.74)
52
.39
7.52**
(51)
57.80
(49.74)
54
.33
4.82**
(53)
27.96
(55.98)
58
.28
5.03**
(56)
-30.29
(174.5)
57
.28
4.93**
(53)
--
--
643.52**
(78.661)
58
.64
23.42***
(57)
Reported above are OLS b coefficients with standard errors in parentheses. ***p<.001, **p<.01,
*p<.05
27
Table 4. County Policy as a Function of County Ideology, Control Variables (N>50)
Drug
Jailings
b
(s.e.)
.01*
Three
Strikes
b
(s.e.)
.01*
Public
Defense
b
(s.e.)
-.28**
Library
Services
b
(s.e.)
-.16**
Drug
Treat
b
(s.e.)
-.57**
Health
Roads
Welfare
b
(s.e.)
-.41+
b
(s.e.)
.32
b
(s.e.)
-1.1+
(.003)
(.000)
(.089)
(.069)
(.278)
(.226)
(.537)
(.656)
Income
-.02
(.013)
--
-.00
(.000)
-.01+
(.00)
-.00
(.00)
-.001
(.001)
-.001*
(.001)
-.004**
(.001)
Education
.01
(.01)
.01+
(.000)
.17
(.193)
.26**
(.120)
.04
(.45)
.22
(.624)
.74
(.959)
-3.0**
(1.30)
% Black
1.04
(1.14)
.00
(.002)
-15.17
(34.70)
-18.09
(22.09)
197.9**
(83.40)
-45.77
(76.61)
-418.7*
(170.2)
577.0*
(234.9)
Age
-.02+
(.012)
.02+
(.000)
--
--
--
--
--
--
County
Ideology
Poverty
--
.00**
(.000)
--
--
Drugs
--
--
--
--
Constant
N
R2
F(d.f.)
.313
(.874)
46
.40
5.62 **
(42)
-.001
(.001)
47
.38
5.10**
(43)
14.99
(13.44)
46
.21
2.818
(42)
-.169
(8.67)
43
.26
3.43**
(39)
--
19.20
(96.58)
22.36
(37.10)
48
.26
2.69**
(43)
--
--
--
--
--
--
45.76
(61.02)
46
.16
1.61
(42)
68.45
(69.51)
46
.26
3.61**
(42)
618.70
(93.02)
47
.60
16.70
Reported above are OLS b coefficients with standard errors in parentheses. ***p<.001, **p<.01,
*p<.05, +p<.10
28