THEORETICAL EXPECTATIONS

Research on “policy feedback” has shown that direct policy benefits from social programs—such as Medicare and the GI bill—may galvanize constituencies to protect those policies (e.g., Campbell Reference Campbell2005; Mettler Reference Mettler2005). Similarly, direct policy costs may provoke opposition due to financial self-interest, such as wealthy homeowners opposing high property taxes (Sears and Citrin Reference Sears and Citrin1982). Such costs may be not only financial, but also spatial, provoking self-interest typified by NIMBY (“Not In My Backyard”) opposition (e.g., Hankinson Reference Hankinson2018; Stokes Reference Stokes2016). Public policy to confront the opioid crisis may activate an especially powerful sense of self-interest because of the policies’ concentrated burdens. Specifically, self-interest may motivate people to support policies in which they will pay less relative to policies where they will pay more, and support treatment infrastructure less when it is closer to them rather than far.

However, partisan polarization may limit the influence of self-interest. Preferences for drug treatment policy—as with many other policies—have historically cleaved along partisan lines: Democrats have ordinarily been more supportive of treatment programs than Republicans (Meier Reference Meier1994). At national, state, and local levels, evidence abounds that partisanship shapes policy outputs (Caughey, Warshaw, and Xu Reference Caughey, Warshaw and Xu2017; de Benedictis-Kessner and Warshaw Reference de Benedictis-Kessner and Warshaw2016). Growing polarization among elites and the nationalization of politics may reinforce the degree to which this partisanship affects public opinion (Druckman, Peterson, and Slothuus Reference Druckman, Peterson and Slothuus2013; Hopkins Reference Hopkins2018). Yet in some policy areas, partisan polarization has led to few policy differences (Anzia and Moe Reference Anzia and Moe2017; Grumbach Reference Grumbach2018). Especially in the area of health policy, personal experiences and self-interest may attenuate the role of partisanship (Campbell Reference Campbell, Wailoo, Cohen, Zelizer and Colby2015; Lerman and McCabe Reference Lerman and McCabe2017).

Because of the geography of the opioid crisis, the communities most affected by today’s crisis—those who are also currently paying the cost of treatment policy—are much more likely to be conservative than in past drug crises. Therefore, the voters whose self-interest—based on their income, context, and location—might influence their policy preferences may also have differing partisan predispositions toward health policy. In other words, voters may be cross-pressured (Chong, Citrin, and Conley Reference Chong, Citrin and Conley2001; Klar Reference Klar2013) by their partisanship and self-interest in the formation of policy preferences concerning the opioid crisis. We expect that financial and spatial self-interest may influence individuals to form opinions that decrease attitudinal polarization between parties.

RESEARCH DESIGN AND DATA

To test these theoretical questions, we surveyed a representative sample of 2,000 United States residents in 2018.^{Footnote 2} In this survey, we asked respondents their support for a policy funding opioid treatment and a policy that would open a new treatment clinic nearby. We measured support for each using a five-point scale from “strongly support” to “strongly oppose.”^{Footnote 3}

The first policy proposal involved a $100 million state bill funding medication-assisted opioid addiction treatment. Using a split-sample design, we described a funding model designed in one of two ways that varied the allocation of the policy’s financial burden.^{Footnote 4} In the first of the two split-sample options, the income-based option, we describe the policy as funded redistributively, with those people who have a household income higher than their state’s median income paying more than those with a lower household income. In the second split-sample option, the overdose rate-based option, we describe the policy as funded instead according to a community’s overdose rate, with those living in areas with a higher opioid overdose rate relative to their state’s median rate paying more than those in areas with a lower overdose rate.^{Footnote 5} This second option—allocating financial burden based on the local overdose burden—matches the costs of treatment to those who would benefit from the funding, a framework championed for other policies with spatially concentrated benefits (e.g., Mullin, Smith, and McNamara Reference Mullin, Smith and McNamara2019). Furthermore, this design of policy matches the status quo, in which cities and counties with high rates of opioid use have few outside resources to which they can turn and must shoulder the financial burden for these policies.^{Footnote 6} In short, our first policy option allocates costs redistributively while the second option mirrors the status quo, allocating financial costs according to need.

The second policy proposal involved the construction of a new opioid treatment clinic. We experimentally varied the clinic’s proposed location to manipulate its spatial cost and therefore assess the causal effect of respondents’ spatial self-interest. We described the proposed location as either two miles away (a 40-minute walk), in the far treatment condition, or 1/4 mile away (a 5-minute walk), in the near treatment condition, from the respondent’s home. The difference in average levels of support for the treatment clinic between the two experimental conditions represents the treatment effect of the proximity of policy infrastructure.

To assess moderators of self-interest, we measured respondent exposure to the opioid crisis, as contact with societal ills may affect political participation and policy attitudes (e.g., Bateson Reference Bateson2012). First, we asked if respondents personally knew anyone who had been addicted to opioids, including heroin and prescription pain killers. Second, we measured respondents’ contextual exposure to overdose deaths by connecting their ZIP code to county-level data on opioid death rates in 2015 from the National Vital Statistics System (NVSS) provided by the Centers for Disease Control’s National Center for Health Statistics. Finally, we recorded respondents’ self-reported partisan identification, ideology, income, and ZIP code.

FINANCIAL SELF-INTEREST

We next assess the role of self-interest in support for our first policy proposal, the state bill funding opioid treatment. In both funding models, we operationalize financial self-interest as the difference in support between those who would pay more and those who would pay less. For those who viewed the proposal funded redistributively, this means the difference in support between above-median income respondents, who would pay $55, and below-median income respondents, who would pay $5. In the overdose rate-based funding model, we operationalize self-interest as the difference in support between those who live in areas with higher rates of overdose and those in areas with lower rates of overdose.

To capture the relative importance of self-interest, we measured the difference in support for each funding condition across the dimension determining self-interest and present these results in Figure 2, with raw differences plotted as circles and covariate-adjusted model coefficients as triangles.^{Footnote 9} For the redistributive funding model, on the left, higher-income respondents were nine percentage points less supportive of the policy than lower-income respondents. Similar results hold for the needs-based funding model, shown on the right. Respondents living in high-overdose areas were eight percentage points less in favor of needs-based treatment funding than those in low-overdose areas.^{Footnote 10}

FIGURE 2.

Effect of Self-Interest on Policy Support Across Both Funding Models With 95% (Thin Lines) and 90% (Thick Lines) Confidence Intervals

To put these effects in context, we compare these differences between groups with opposing self-interest to the polarization between partisan groups. In Figure 3, we plot the effect of an indicator for whether or not the respondent was a Republican (rather than a Democrat) for both funding models, with raw differences again plotted as circles and covariate-adjusted model coefficients as triangles. For the redistributive funding model, Republicans were 28 percentage points less supportive of the redistributively funded policy than Democrats and 15 percentage points less supportive of the overdose rate-based funding model. These sharp partisan divides show the power of partisanship in policy preferences. In comparison, the effect of partisanship is more than double that of self-interest on support for these policy proposals.

FIGURE 3.

Effect of Partisanship on Policy Support Across Both Funding Models With 95% (Thin Lines) and 90% (Thick Lines) Confidence Intervals

We next assess cross-pressuring within each party by interacting the relevant self-interest variable with an indicator for party identification. We plot these differential effects in Figure 4, for the redistributive funding model in the left panel and the overdose rate-based funding model in the right panel. For the redistributive funding model, self-interest—as measured by income group—has an effect close to zero among Democrats, whereas it has a 19 percentage point effect among Republicans. This interaction of 21 percentage points between self-interest and partisanship is statistically significant (p < 0.01), demonstrating that Republicans were cross-pressured when faced with a redistributive policy proposal. This effect manifests as a boost in support among low-income Republicans and a decrease in the partisan gap in support by more than half, as we show in Online Appendix Figure A.2. In contrast, for the overdose rate-based model, self-interest—as measured by local overdose rates—had a similar effect among both Democrats and Republicans.

FIGURE 4.

Effect of Self-Interest on Policy Support by Partisanship Across Both Funding Models With 95% (Thin Lines) and 90% (Thick Lines) Confidence Intervals

SPATIAL SELF-INTEREST AND THE IMPLEMENTATION OF INFRASTRUCTURE

Next, we test the causal effect of spatial self-interest by experimentally varying the proximity of a proposed treatment clinic. We operationalize the effect of self-interest as the difference in support between those respondents in the “near” condition, who evaluated a clinic proposed 1/4 mile away, and those in the “far” condition, who evaluated a clinic proposed two miles away. We plot this treatment effect in Figure 5 for the full sample, on the left, and partisan subgroups, on the right.

FIGURE 5.

Effect of Spatial Self-Interest on Policy Support by Partisanship With 95% (Thin Lines) and 90% (Thick Lines) Confidence Intervals

Among our full sample, 53% who were in the far condition supported the new treatment clinic, whereas respondents who were in the near condition were 14 percentage points less supportive of clinic construction (p < 0.01). This treatment effect demonstrates how the proximity of infrastructure and corresponding spatial self-interest shapes opinion on implementing opioid treatment policy.

To put this effect in context, we can compare the influence of spatial self-interest to that of partisanship for this policy proposal. As discussed earlier, Republicans were 25 percentage points less supportive of clinic construction than Democrats. While the magnitude of self-interest’s influence is less than that of partisanship, it is still substantial.

To test whether spatial self-interest cross-pressures partisans as it did for financial self-interest, we separately estimate this treatment effect by respondents’ partisanship and test for an interaction between spatial self-interest and partisanship. As shown on the right side of Figure 5, the effect of spatial self-interest among Republicans and Democrats is largely similar, whether or not covariates are used in the models. This stability demonstrates the enduring power of spatial self-interest across partisan subgroups.^{Footnote 11}

SUPPLEMENTARY MATERIAL

To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055419000443.

Replication materials can be found on Dataverse at: https://doi.org/10.7910/DVN/OLTXFC.