A Third Mechanism of Policy Diffusion - (SSRN) Papers

A Third Mechanism of Policy Diffusion: The Social Contagion Model * Julianna Pacheco [email protected] 8/9/2010 Scholars have offered two explanations for the large role that neighboring states play in the diffusion process: (1) state officials learn by observing the outcomes of neighboring policies and (2) states seek an economic advantage over neighboring states. Both models contend that decision-making occurs laterally as state legislators learn from or react to the policy decisions of other elites; the public is but a minor facet of policy diffusion. I offer a third explanation, the social contagion model, which makes public opinion the driving force behind elite decision making. The social contagion model suggests that state residents learn about neighboring policies and react by changing their aggregate opinions on that policy. If state opinion becomes supportive, state officials respond by enacting similar policies or risk being ousted from office. Using a dataset on anti-smoking legislation from 1990-2008, I find empirical support for the social contagion model.

*Paper prepared for the Annual Meetings of the American Political Science Association, Sept. 14, 2010

Policies spread across the American states—a process called policy diffusion (e.g., Berry and Berry 1990; Mooney and Lee 1995; Gray 1973; Walker 1969) and diffusion is more likely to occur across neighboring states (e.g., Berry and Berry 1990; Walker 1969). In fact, the positive influence of a neighbor’s policy has been found across a range of policy issues including Indian gaming (Boehmke & Witmer 2004), the lottery (Berry & Berry 1990), anti-smoking legislation (Shipan & Volden 2006), school charters (Mintrom 1997; Mintrom & Vergari 1998), healthcare (Balla 2001), taxes (Berry & Berry 1992), enterprise zone programs (Turner & Cassell 2007), and welfare (Volden 2007), among others. Scholars have offered two explanations for the large role that neighboring states play in the diffusion process: the social learning model and the economic competition model. The social learning model suggests that state officials use information from neighboring states to learn about the problems and successes of new policies prior to adoption (Gray 1973; Volden 2006; Berry and Baybeck 2005). The economic competition model suggests that states make policy choices in order to gain an economic advantage over proximate states (Tiebolt 1956; Berry & Berry 1990; Volden 2002). In both of these models, however, the primary actors are state officials (e.g., governors or state legislators) who respond to neighboring states when making policy decisions. Largely ignored is the fact that state officials are constrained by the preferences of state residents. This is a large omission considering that the primary goal of officials is reelection (Mayhew 1974). Motivated by reelection, state officials are likely to exhibit political expediency in which they enact policies that the public supports (Erikson, Wright, and McIver 1993; Erikson, MacKuen, and Stimson 2002) or risk being ousted from office. If officials are constrained by public preferences then what role does public opinion play in the diffusion of policies across neighboring states?

Pacheco 1

I answer this question by looking at how public preferences influence the diffusion of anti-smoking legislation across the states. In so doing, I offer a third explanation for the positive role of neighboring policies. I term this model the social contagion model, which suggests that state residents learn about policies in neighboring states. As state residents learn about these new policies, their aggregate public opinion has the potential to change and, if in support of neighboring policies, encourages state officials to adopt a similar program. State officials, who are motivated by re-election, then respond to the changing public opinion of their residents by adopting neighboring policies. I empirically test hypotheses that emerge from the social contagion model using data on the diffusion of smoking bans in restaurants from 1990 to 2008. A necessary component of the social contagion model is the measurement of state preferences on a particular policy, in this case, anti-smoking legislation in restaurants. The challenges of measuring state public opinion are well documented (e.g., Erikson, Wright, and McIver 1993), however, recent advances in small area estimation have made it possible to measure state public opinion accurately and reliably (Lax and Phillips 2009a; Park, Bafumi, and Gelman 2005) and across time (Pacheco 2009). I employ these techniques—multilevel regression, imputation, and poststratification—to create a unique dataset that includes state preferences on smoking bans in restaurants from 19912008. Using this dataset, I find empirical support for the social contagion model. Public support for anti-smoking legislation is influenced by the policies of neighboring states suggesting that state residents learn about neighboring policies and then react in the aggregate to those policies. Moreover, as state opinion becomes more supportive of anti-smoking legislation, states are more likely to adopt those policies. I find no empirical support, however, that certain

Pacheco 2

institutional factors, such as the presence of initiatives or legislative professionalism, mediate the influence of public opinion on policy adoption. The results presented here suggest that ordinary state citizens play a major role in the diffusion of policies across the states. On the other hand, state officials play a less critical role as they simply respond to the changing attitudes of their constituents. The implications for this research are twofold. First, policy adoptions have the potential to influence public opinion in neighboring states. State residents do not live in a vacuum; they are often aware of and react to policies in neighboring states that are similar in ideology and demography. While past research has hinted that state residents may react to neighboring policies by “voting with their feet” (e.g., Tiebolt 1956), this paper finds direct evidence that aggregate public opinion is influenced by the policies of other states. Citizens may not move to neighboring states, but instead, change their opinion and then pressure state officials to adopt similar policies in the home state. Second, policy adoptions occur in response to state opinion, which provides additional evidence that the impact of public opinion on policy is causal at the sub-national level. While it may seem obvious that public opinion should positively influence policy adoptions, past research has not been able to empirically support this conclusion. Most research has focused on the political and economic determinants of policy adoption, largely ignoring the expressed preferences of state residents. Studies that control for public opinion often use a static measure of ideology (e.g., Berry & Berry 1990) or a proxy for public opinion, such as the percentage of fundamentalists ( Berry & Berry 1990) or the percentage of adult smokers (Shipan and Volden 2006). In fact, I know of no study on policy adoption that explicitly includes a time-varying measure of public opinion that is specific to the policy innovation, no doubt because of the methodological challenges to measuring state opinion. The results presented here confirm that

Pacheco 3

state opinion plays an important, positive, and causal role on the adoption of new policies across the states. Mechanisms of Policy Diffusion: Social Learning vs. Economic Competition According to Walker (1969), an innovation is a “program or policy which is new to the states adopting it, no matter how old the program may be or how many other states may have adopted it” (881). Once a state innovates, the probability that another state will innovate is great; the process by which an innovation spreads across states is called diffusion (Gray 1973). Although diffusion occurs both across time and space, this article concentrates on the distinct geographic patterns of diffusion. Diffusion is more likely to occur across neighboring or nearby states; for instance, as the number of neighbors that have adopted a policy increases so does the probability of adoption (e.g., Berry and Berry 1990). The positive influence of neighboring states is well documented in past research (e.g., Mooney and Lee 1995, among numerous others); the challenge, however, has been to identify the mechanisms of diffusion. Traditionally, political scientists have attributed policy diffusion across neighbors as arising from a process of social learning (Walker 1969; Gray 1973; Glick & Hays 1991; Mooney & Lee 1995; Mooney 2001). The social learning model contends that state politicians use other states to learn about public policies prior to adoption. Political officials, in their search for answers to complex problems, engage in a form of “satisficing” (Simon 1955), whereby they wait and see how a policy works out before adopting it in their own state (Volden 2006). Officials look first to neighboring or nearby states to learn about policies for reasons of political and demographic similarity (Walker 1969; Berry & Berry 1990) and political networking (Mintrom & Vergari 1998). Neighboring states and those that are in the same region tend to share ideology and demographics (Erikson, Wright, & McIver 1993) and, thus, are more

Pacheco 4

influential than distant states. In the area of anti-smoking bans, the social learning model contends that state officials, who may want to improve the public health of its residents, watch how smoking bans in neighboring states fare before deciding on policy adoption in their home state. In contrast, the process of diffusion through economic competition is built from the idea that states are constantly competing for business and tax dollars. States make policy choices in order to gain an economic advantage over proximate states. This economic advantage helps states attract the “best residents” (Tiebolt 1956) who in turn will contribute to the financial success of the state. Because of the mobility constraints of residents, however, states are more likely to compete with nearby states than those far away on economic policies. Moreover, the pattern of diffusion may differ depending on the policy. In the case of welfare, for instance, states who are fearful of attracting lower income residents from neighboring states may adjust their welfare policies by offering fewer benefits when neighboring states adopt new policies (Peterson and Rom 1990; Berry, Fording, and Hansen 2003). On the other hand, competition over policies that provide revenue, such as the lottery, may increase the extent of the policy as neighboring states compete over financial resources (Berry and Berry 1990, 1992; Boehmke and Witmer 2004; Berry and Baybeck 2005). In the case of anti-smoking legislation, the economic competition model suggests that diffusion occurs because states enjoy an economic advantage to offering non-smoking establishments, similar to the economic advantage that arises from the lottery. By offering smoke-free restaurants, states attract business from neighboring states whose residents prefer to dine-out in smoke-free restaurants. States that offer smoke-free establishments may also attract “better” residents who are non-smokers and, therefore, healthier;

Pacheco 5

this is particularly advantageous since the economic costs from lost productivity and health care expenditures of smoking are exceedingly high (The American Lung Association 2009). While both models lead to the presence of regional diffusion, neither does very well at explaining the role that public opinion may play in this process. This is likely due to the fact that measuring state policy preferences reliably and accurately is challenging (e.g., Erikson, Wright, & McIver 1993). And, these challenges are multiplied when measuring state opinion over time, which is necessary to incorporate into models of policy diffusion. Putting these methodological difficulties aside, there is reason to suspect that public opinion plays at least some role in the process of policy diffusion. If nothing else, we know that politicians are largely constrained by the public preferences of state residents (e.g., Erikson, Wright, and McIver 1993). Whether public opinion and policy are measured as global indicators along a left/right dimension (Erikson, Wright, and McIver 1993, 2007; Brace et al. 2002) or on specific policies such as welfare (Hill, Leighley and Hinton-Andersson 1995; Johnson 2003), abortion (Norrander and Wilcox 1999, Arcaneaux 2002; Camobreco and Barnello 2008), the death penalty (Nice 1992; Norrander 2000; Mooney and Lee 2000), healthcare (Grogan 1994; Gray, Lowery and Godwin 2007), homosexual rights (Haider-Markel and Kaufman 2006; Lax and Phillips 2009b), and the environment (Johnson, Brace, and Arceneaux 2005) the conclusion is the same: state policies reflect the preferences of state citizens. Hence, there is little reason to suspect that officials would adopt neighboring policies that were not supported by the public, regardless of the success or economic benefits of those policies. But, it may be that the public plays a more active role in the process of policy diffusion as state residents learn about neighboring policies. We know, for instance, that citizens are aware of and often react to the policies of neighboring states. When cigarette taxes increase, price

Pacheco 6

sensitive smokers purchase cigarettes from lower or non-taxed venues, including out-of-state vendors (Hyland et al. 2005). People living in a state without a lottery often go to other states with lotteries in order to purchase tickets (Fink and Rork 2003; Mikesell 1987; Berry & Baybeck 2005). And, women travel to less restrictive states (e.g., those that do not have mandatory waiting periods or counseling) to obtain abortions (Althaus & Henshaw 1994; Joyce et al. 2009). The fact that residents are aware of neighboring policies suggests that aggregate state preferences have the potential to change in response to neighboring policies, which in turn may influence the decision making of elected officials in the home state. In other words, state residents exhibit a form of social learning about neighboring policies. As citizens gain more information about neighboring policies, their aggregate opinion can become more supportive of these policies. State officials who are interested in reelection then enact policies that the public supports and that neighboring states have already adopted. I explain this third mechanism of diffusion, which focuses on the social learning process of neighboring policies at the citizen level and the policy responsiveness of state officials, in more detail below. A Third Mechanism: Social Contagion To describe the third model of policy diffusion, the social contagion model, let us first imagine two neighboring states, State A and State B. Let us assume for simplicity that these are the only neighboring states that matter.1 We observe at time t that State A, for whatever reason, adopts a new policy, Policy 1. At a later time, t+n, we observe that State B similarly adopts Policy 1; in other words, we observe the policy diffusion of Policy 1 from State A to State B. What is the mechanism that influenced State B to adopt State A’s policy?

1

In reality, states have more than one neighbor with potentially divergent policies. The study of whether one neighbor has a larger effect than others or what happens when neighboring states exhibit different policies is beyond the scope of this paper.

Pacheco 7

According to the social contagion model, residents in State B are aware of Policy 1 after State A adopts it. There are various ways in which residents in State B learn about Policy 1 after adoption in State A. First, residents in State B may experience Policy 1 directly as they travel to State A. Of course, not everyone residing in State B has the same propensity to learn about Policy 1 from State A; State B residents living close to the border of State A are at a higher risk of learning about Policy 1 compared to residents living further from State A (Berry & Baybeck 2005). While identifying the population of concern (e.g., the individuals at risk of learning about Policy 1) requires the use of geographic information systems (Berry & Baybeck 2005) and is beyond the scope of this paper, it is important to note that direct learning of neighboring policies is possible. A second way in which State B residents may learn about Policy 1 is via the media if news markets from State A and State B overlap. Individuals in multi-state markets have considerably more opportunity to learn about neighboring policies. Research has shown, for instance, that New Jersey residents living in the New York media market knew much more about New York politics compared to those living in other parts of New Jersey (Zukin & Snyder 1984). Finally, the spread of information about Policy 1 may occur through social networks. Residents in State B may learn about Policy 1 directly from residents in State A that are within their social networks. Knowledge about Policy 1 may also spread from resident to resident within State B. For instance, State B residents may hear about Policy 1 via the media and then discuss Policy 1 with other State B residents. Social networks have been shown to be a strong transmitter of political information (Huckfeldt et al. 1993; Huckfeldt 1986) and can influence political preferences (Huckfeldt & Sprague 1995; Mutz & Martin 2001; Kenny 1992) and participation (Huckfeldt 1979; Huckfeldt & Sprague 1995; McClurg 2003). Regardless of how residents learn about neighboring policies, the larger point is that as State B residents learn about Policy 1,

Pacheco 8

aggregate public opinion in State B about Policy 1 has the potential to change in support of that policy. 2 Social Contagion Hypothesis: Adoption of neighboring policies increases the public support of those policies in the home state. As support for Policy 1 in State B increases, the probability that State B adopts Policy 1 also increases. There are two ways in which public support increases the chance that State B adopts Policy 1. The first way is through political expediency ( Erikson, MacKuen, & Stimson 2002). Because state officials are interested in re-election, they have incentives to gauge and respond to changing public opinion in their home state. The second way that public support influences the adoption of Policy 1 in State B is through electoral turnover (Erikson, MacKuen, & Stimson 2002). If public support for Policy 1 is strong enough and sitting state officials do not adopt Policy 1, then residents in State B may vote for new officials in the next election cycle who then adopt Policy 1. Via both of these mechanisms—political expediency and electoral turnover—policies reflect the changing political preferences of State B in regards to Policy 1. In other words, policy responsiveness occurs. Policy Responsiveness Hypothesis: As public support for neighboring policies increases, the probability of adoption in the home state also increases. To better illustrate the way in which the social contagion model differs from past models, Figure 1 depicts the process of policy diffusion of Policy 1 from State A to State B via the traditional models (e.g., the social learning model and the economic competition model), shown in the upper panel, and the social contagion model, shown in the lower panel. As shown in the upper panel, both the social learning and economic competition models theorize that diffusion 2

Of course, there are probably instances in which residents may learn about a neighboring policy and support for that policy declines. There may also be mediating factors that account for how the public reacts to neighboring policies. Identifying these other circumstances is beyond the scope of this paper.

Pacheco 9

occurs via state officials. State A adopts Policy 1, state officials in State B respond to the adoption of Policy 1, either because of social learning or economic competition, by adopting similar policies in their home state. By contrast, the social contagion model, depicted in the lower panel of Figure 1, suggests that diffusion occurs via the public. State A adopts Policy 1, residents in State B learn about Policy 1, state officials in State B respond to changing public opinion and then adopt Policy 1 in their home state. Figure 1, thus, illustrates the main theoretical distinction between the two traditional models of policy diffusion and the social contagion model. The social learning and economic competition models focus on the decisionmakers of state officials absent from public opinion while the social contagion model suggests that the public plays a large role—if not the main role—in the diffusion of policies across neighboring states. It is important to note that Figure 1 depicts theoretical models of policy diffusion. Empirically, it has been difficult for scholars to model the preferences of state officials directly, for instance, by using roll call votes or elite interviews. Instead, it has been assumed that legislative voting is the mechanism through which policy diffusion occurs across neighboring states. Similar to past research, I do not model the preferences of state officials directly and, instead, assume that any effect public opinion has on policy adoption is through legislative voting or electoral turnover and, therefore, policy responsiveness. Moreover, the social contagion model depicted in Figure 1 is not absolute. We know that, regardless of the process of policy diffusion, some states will never pass particular innovations. What conditions make policy diffusion via the social contagion model more likely? As depicted in Figure 1, one component of the model that may mediate the likelihood of adoption is institutions. Because the social contagion model introduces a component of policy

Pacheco 10

responsiveness into the policy diffusion process, we can theorize about institutional factors that makes states more or less responsive to changing public opinion. Scholars have identified a number of factors that encourage or inhibit policy responsiveness; I focus on two conditional factors: the presence of initiatives and the level of professionalism within the state legislatures. Institutional Mediating Factors Two important institutional characteristics that have been found to condition policy responsiveness include the presence of initiatives and legislative professionalism. Initiatives are a form of direct democracy where citizens can directly propose laws and policies to the legislature. Several studies have shown that initiatives heighten policy responsiveness to public opinion on a variety of issues including abortion (Areceneaux 2002; Bowler & Donovan 2004; Gerber 1999), government spending (Matsusaka 2004), the death penalty (Gerber 1996), minority rights policies (Gerber & Hug 2002), and gay rights (Gerber & Hug 2001). Why does the presence of initiatives increase policy responsiveness? One theory is that legislators take into account public opinion when drafting legislation in anticipation of future initiatives (Gerber 1996). Hence, the mere “threat” of an initiative is enough for elected officials to respond to changing levels of public opinion. Another theory is that initiatives give legislators more accurate information about voter preferences (Romer & Rosenthal 1979; Matsusaka 2004). The presence of initiatives, thus, gives legislators an extra information source from which to gauge changing public opinion. Both of these theories imply that initiatives will influence the political expediency of state legislators. Initiatives Hypothesis: The effect of public opinion on policy adoption will be stronger in states that have initiatives.

Pacheco 11

Legislative professionalism also has been shown to increase policy responsiveness. Professional legislatures are those in which legislators meet in unlimited session, are paid well, and are provided with superior staff resources and facilities (Squire 1992). These increased resources allow professional legislators to have more contact with their constituents (Squire 1993), monitor changing preferences better, and, therefore, be more attentive to constituent concerns (Maestas 2000). Legislative professionalism also creates an environment that attracts highly skilled politicians who vie for elected office; individuals seeking office in professional legislators are more likely to be “career” or professional politicians (Berkman 1994; Squire 1992; Thompson & Moncrief 1992) who have ambitious long term goals of higher office (Maestas 2000; 2003). All of these characteristics make professional legislatures more responsive to changes in public opinion (Maestas 2000). Professionalism Hypothesis: The effect of public opinion on policy adoption will be stronger in more professional legislatures. Testing the Social Contagion Model using Anti-Smoking Legislation While the above hypotheses are likely to hold in numerous policy areas, I focus on the role that public opinion plays in the diffusion of anti-smoking legislation from 1990-2008. Over this time period, 27 states enacted comprehensive smoking bans in restaurants. California and Utah lead the way by enacting smoking bans in restaurants in 1994, while other states like South Dakota and Massachusetts enacted smoking bans in restaurants in 2008. Studies of policy innovation and diffusion using public health issues and anti-smoking legislation, in particular, are well suited to explore the process of policy diffusion. First, several studies in both the political science and public health fields have conducted analyses on the innovation and diffusion of anti-smoking legislations; this allows me to build on existing

Pacheco 12

literature when thinking about other important factors that influence the adoption of smoking bans across the states. Second, there is large variation in the timing of adoption across the states. Of the 27 states who enacted smoking bans in restaurants from 1990-2008, 2 enacted bans in 1994, 1 in 2002, 2 in 2003, 6 in 2005, 5 in 2006, 6 in 2007, and 5 in 2008. Hence, unlike other policy areas in which states enact policies over a relatively short time frame (e.g., the reinstatement of the death penalty) there is large longitudinal variation in the enactment of smoking bans in restaurants. Finally, it is important to know what influences states to enact smoking bans in restaurants since the policy implications of smoke-free laws are so great. Every year an estimated 438,000 Americans die from tobacco-related diseases (American Lung Association 2008) and health officials have declared secondhand smoke dangerous, suggesting that comprehensive smoke-free legislation can improve public health. Understanding why some states are quicker to enact comprehensive smoke-free laws, such as smoking bans in restaurants, thus has important implications for the health of our society. Previous studies on the policy adoption of anti-smoking laws have identified important causal influences, which I incorporate in the form of control variables (more detail below). The most comprehensive analysis on the diffusion of anti-smoking legislation has been conducted by Shipan and Volden (2006). Looking at three types of anti-smoking policies (government building restrictions, restaurant restrictions, and out of package sales restrictions) from 19752000, Shipan and Volden (2006) find that states are more likely to adopt anti-smoking restrictions in all three policy domains if neighboring states pass such policies, if health organizations play a prominent role in the state, and if government officials are liberal. Conversely, one of the most important factors that decreased the probability that a state will enact smoking restrictions was whether or not the state was a tobacco-producing state. These

Pacheco 13

findings are congruent with analyses from the public health literature that finds that legislative voting on anti-smoking legislation is influenced by political ideology (Cohen et al. 2000) and the amount of involvement by the health community (Jacobson, Wasserman,and Raube 1993). Little is known about the role that public opinion plays in the adoption and diffusion of anti-smoking legislation. Shipan and Volden (2006) conclude that preferences towards smoking do not significantly influence the adoption of anti-smoking legislation. However, they infer preferences based on the percentage of smokers in each state. Over time, however, there is little evidence that the percentage of smokers correlate with public preferences on anti-smoking legislation. Using the dataset described below, I find that the correlation between the changes in the percentage of adult smokers and in public preferences towards smoking bans in restaurants is a mere -.08. Hence, inferring public preferences via the percentage of adult smokers may have lead to incorrect inferences in the past regarding the influence of public opinion on anti-smoking legislation. Shipan and Volden (2006) also include various measures of political ideology; however, none of these proxies significantly influenced the adoption of anti-smoking legislation. By including direct measures of public opinion on smoking legislation, I not only explore the role that public opinion plays on policy innovation and diffusion generally, but also look at how public opinion influences the adoption of a particularly important health policy. Data Analysis The data analysis proceeds as follows. First, I empirically test for the social contagion hypothesis by modeling the effect that neighboring policies have on changing public support for smoking bans in restaurants. In this analysis, I employ traditional time series analyses whereby the dependent variable is changing state opinion towards smoking bans in restaurants. The main independent variable is the policy adoptions of neighboring states; that is, the changing

Pacheco 14

proportion of neighboring states that have already adopted a smoking ban in restaurants. In the second set of analyses, I test for the policy responsiveness hypothesis by modeling the effect that state public support for smoking bans has on the probability that a state will adopt a smoking ban in restaurants. For this analysis (and the two subsequent analyses), I employ traditional event history analysis, where the dependent variable is whether or not a state has adopted a smoking ban in restaurants. Finally, I test for the mediating effects that the presence of initiatives and legislative professionalism may have on the influence of public opinion on policy adoption by employing interaction terms. In all of these analyses, it is critical to have dynamic a measure of state opinion towards smoking bans, which I describe in further detail below. Measuring State Public Opinion on Smoking Bans in Restaurants The challenges involved with measuring dynamic state public opinion, particularly from national surveys, are well documented (Erikson, Wright, & McIver 1993, 2006, 2007; Brace et al. 2002, 2007; Berry et al. 1998, 2007; Cohen 2006; Park, Bafumi, & Gelman 2004, 2006; Lax and Phillips 2009a, 2009b; Pacheco 2009). Recent statistical innovations in small area measurement, however, have made it possible to obtain reliable and valid measures of state public opinion over time. Pacheco (2009) shows that scholars can reliably estimate yearly measures of state public opinion by using multilevel regression, imputation, and poststratification (hereafter referred to as MRP; Park et al. 2004) coupled with a small-window moving average, such as a three or five year pooled time frame. In systematic assessments, the MRP technique has been found to be superior to the aggregation method in terms of error and precision (Lax and Phillips 2009a), particularly when measuring state public opinion over time (Pacheco 2009). Pacheco (2009) suggests that data availability is the driving force behind scholars’ decisions about whether a three or five year pooled time frame is appropriate. She then

Pacheco 15

demonstrates dynamic validity using the small pooled time frames with measures of state partisanship and state ideology (Pacheco 2009). I follow the guidelines outlined by Pacheco (2009) and combine responses across two survey organizations to measure dynamic public opinion on smoking bans using the MRP approach on a three year pooled time frame. 3 Specifically, I measure the percentage who favored a smoking ban in restaurants using the Current Population Survey Tobacco Use Supplement (CPS-TUS) and Gallup polls. Details on the statistical technique used to create these measures can be found in Appendix A and the exact question wording is included in Table A1. The aggregate public opinion data cover the 50 states as well as DC from 1991-2006 for anti-smoking opinion. All states are missing on the anti-smoking public opinion measure in 1997, 2004, 2007, and 2008 because questions were not asked about anti-smoking legislation in these years. The time gaps limit our ability to make inferences about dynamic relationships between public opinion and policy, which require continuous time series. There are a number of different options to impute data for the missing values including mean interpolation (for instance, to get a value for time t simply take the mean of the values at t-1 and t+1), pooled time series cross sectional multiple imputation, and individual level multiple imputation estimated prior to MRP. The models reported here use the estimates obtained from performing multiple imputation at the pooled time series cross sectional state level. 4 More detail on how multiple imputation was performed at the state level via this approach can be found in Appendix B. In

3

I use the full model specification with gender, age, race, education, and state as covariates as suggested by Pacheco (2009). Detailed explanation about this method is provided in Appendix A. 4 . I find little difference in the conclusions using each of these techniques; this information is available by request.

Pacheco 16

addition, the models presented in this paper are nearly identical regardless of which technique is used to recover missing data (results available from the author by request). Figure 2 shows maps of the percentage of residents who favor a smoking ban in restaurants and 1991 compared to 2008. As shown in Figure 2, residents have increased support for smoking bans in restaurants. In 1991, public support for smoking bans ranged from 26% to 52% while in 2008, support ranged from 42% to 94%. More important for this paper, Figure 2 shows that public opinion towards smoking bans in restaurants is geographically clustered. The least supportive states in 1991 include many that rely on the tobacco industry for their revenue including North Carolina and South Carolina, but also include states that are traditionally against government intervention such as Kentucky, Tennessee, and many Midwest states including South Dakota, Nebraska, Montana, and Wyoming. The most supportive states are those on the West coast (e.g, California and Washington) and in the Northeast (e.g., Connecticut). By 2008, the geographic pattern has not changed much; the least supportive states still reside in the Deep South and Midwest while the most supportive states are on the West coast and in the Northeast. As further evidence that opinion towards smoking bans in restaurants is geographically clustered, I correlated state opinion on smoking bans in restaurants with the average level of support for smoking bans in neighboring states. In other words, if State A is surrounded by State B, State C, and State D, I correlated public support in State A with the average level of support in State B, State C, and State D. I find that the correlation is .57, suggesting that state residents are similar to neighboring residents in their level of public support for anti-smoking legislation. Testing the Social Contagion Hypothesis To test for the social contagion hypothesis, I employ traditional time series methods to explore how the proportion of neighboring policies influences changes in state public opinion on

Pacheco 17

those same policies. More specifically, I use an Error Correction Model (ECM) shown in Equation 1. Eq. 1 An ECM allows for the estimation of both short and long term effects of independent variables and tells us how quickly or slowly the system returns to equilibrium or the overall mean after being disrupted. The dependent variable captures the changes in public opinion over time and is measured by differencing public support for smoking bans in restaurants using the public opinion measures described above. A lagged dependent variable (LDV) is included to account for time dependence. For all time varying covariates (described below), I include both short and long term effects; that is, for each variable, I include both the differenced independent variable ( and the lagged independent variable (

.

The main independent variable is the proportion of neighboring states that have already adopted a smoking ban in restaurants. The majority of the anti-smoking legislation data come from the Centers of Disease Control and Prevention’s State Tobacco Activities Tracking and Evaluation (STATE) System. Where there are gaps in the data (e.g., prior to 1995), I used the National Cancer Institute’s State Cancer Legislation Database Program to determine if and when states enacted certain restrictions. Of course, policy adoption is not static. Instead, the proportion of neighboring states that have already adopted a smoking ban can change over time. Because this variable is time varying, I include both the differenced ( (

and lagged version

in the empirical model. 5 According to the social contagion hypothesis, as the number of

neighboring states adopts a smoking ban in restaurants, public support for smoking bans in

5

Alaska and Hawaii have no neighbors, consequently, this variable is measured as 0 for those two states.

Pacheco 18

restaurants should also increase. In other words, the coefficients on the proportion of neighboring states variables should be positive and statistically significant. I control for a number of other factors that may influence changes in state opinion on smoking bans over time. First, I control for certain state demographic characteristics. The most important determinant of public support for smoking restrictions at the individual level is smoker status with the expectation that smokers are less supportive of anti-smoking legislation compared to non-smokers. Hence, I include a measure of the percentage of adult smokers in each state obtained from the Center for Disease Control’s STATE system. But other demographic characteristics may also matter. State ideology and partisanship are included with the usual expectation that more liberal and democratic states will be more likely to endorse anti-smoking legislation. These variables are measured via the MRP approach as described by Pacheco (2009). Historical state culture towards the production of tobacco may also influence state opinion towards smoking restrictions. The expectation is that states with a long history of tobacco production may be less likely to support smoking restrictions compared to other states, even after accounting for other important determinants. Hence, I include a dummy variable equal to 1 in all states where tobacco is produced and 0 otherwise. Results for the Social Contagion Hypothesis Table 1 displays the results of the ECM, which tests for the social contagion hypothesis. With an ECM, the coefficient on the lagged dependent variable (

gives the error correction

rate with a value closer to zero indicating a slow return to equilibrium. The coefficient on the lagged dependent variable for state attitudes towards smoking bans is -.08 suggesting that state public opinion is slow to return to equilibrium when disrupted.

Pacheco 19

Consistent with the social contagion hypothesis, changes in the proportion of neighbors with a smoking ban in restaurants increase support for smoking restrictions in the home state. More specifically, the model suggests that as the proportion of neighboring states adopts smoking bans in restaurants, public support for such policies increases in both the short and long term. The coefficient on the differenced proportion of neighboring states variable (

gives us the short term effect of policy adoption on state public opinion. To

get the estimated effect of a unit change in X, we simply multiply this effect with the coefficient (

For instance, a .20 increase in the proportion of neighboring states with anti-smoking

legislation (which is roughly 2 standard deviations above the mean change) increases public support for smoking bans in restaurants in the next year by about .75% (e.g., .20*3.79). The influence that neighboring policies has on public opinion does not stop there. The long term effect of policy adoptions on public opinion can be calculated using the coefficient of the lagged public opinion variable (e.g.,

by (De Boef & Keele 2008) Eq. 2

Using Equation 2, that same increase in neighboring policies increases public support for smoking restrictions in restaurants by an additional 5% over the long run. Additionally, antismoking legislation has a significant influence on public opinion towards smoking bans even after controlling for important demographic variables (e.g., the percentage of adult smokers) and state culture (e.g., tobacco producing state). This is strong support for the social contagion hypothesis. The control variables behave according to expectations. Tobacco producing states have significantly lower support for smoking bans in restaurants compared to non-tobacco producing states. Moreover, the percentage of adult smokers has negative effects in both the short and long Pacheco 20

term on support for anti-smoking legislation. Finally, liberal states are more likely to endorse support for anti-smoking legislation. Testing the Policy Responsiveness Hypothesis The first component of the social contagion model of policy diffusion has empirical support. As shown in Table 1, public support for a particular policy adoption is influenced by policy adoptions in neighboring states. The next component of the theoretical model is the policy responsiveness hypothesis, which posits that as public support in the home state increases, so too does the probability of adoption. The dependent variable is whether a state adopts a smoking ban in restaurants. Consequently, this variable is coded as having a value of 0 for the years in which the state has not yet adopted a smoking ban in restaurants and a 1 in the year of adoption using the anti-smoking legislation data described above. In subsequent years, the state is dropped from the dataset since it is no longer “at risk” of innovating; this is the conventional coding scheme for event history analysis (Berry & Berry 1990). This yields one observation per state per year, for a total of 50 x 18 = 900 observations. Excluding observations not in the riskset for adoption (those states after the policy has already been adopted), leaves 742 observations suitable for analysis. Since the dependent variable is dichotomous, I employ logistic regression. To account for potential problems of non-independence of observations and of heteroskedasticity, I rely on the cluster procedure whereby observations are clustered by stateyear. The main independent variable needed in order to test the policy responsiveness hypothesis is state public opinion on smoking bans in restaurants, as described above. Because I am interested in how changing public opinion influences the probability that a state will adopt a policy, this variable is differenced. According to the policy responsiveness hypothesis, as public

Pacheco 21

support for smoking bans in restaurants increases, the probability for adoption should also increase; the coefficient on the public opinion measure should be positive. Differencing the public opinion measure not only more accurately reflects the policy responsiveness hypothesis, but also solves a problem of multi-collinearity. As we saw in the previous analysis, state opinion towards smoking bans is highly correlated to the proportion of neighbors with a smoking ban in restaurants. Via differencing, the correlation with the proportion of neighbors decreases (r=.14) and allows me to include the proportion of neighbors variable into the analysis. The proportion of neighbors with a smoking ban in restaurants is included, as described in the previous section, as a control variable. As stated, the most important factor found in previous research that influences the probability that a state innovates is whether neighbors have already innovated (e.g., Berry & Berry 1990; Shipan & Volden 2006). The expectation, however, is that once public opinion is accounted for, neighboring policies will no longer be significantly related to policy adoptions. In other words, the social contagion model suggests that policy diffusion across neighbors occurs because of the influence that neighboring policies has on public opinion; methodologically, the influence of neighboring policies should not be significant because public opinion is included in the model. I also control for a number of other important determinants of policy adoption. As previous research has shown, organized interests can play a large role in whether a state passes certain policies. States with a high number of health organization lobbyists have more antismoking legislation while those with a high number of tobacco industry lobbyists have less antismoking legislation (Shipan & Volden 2006). Similar to past research (e.g., Shipan & Volden 2006), I capture the influence of state organized interests via four variables. 6 The first pair of

6

I am grateful to Shipan and Volden for allowing me to use their dataset from which the interest group variables come from.

Pacheco 22

variables is a ratio of the number of health (or tobacco) lobbyists in the state to the total number of registered lobbyists (Goldstein & Bearman 1996; Shipan & Volden 2006). These variables capture the presence of health and tobacco lobbyists compared to other organized interest groups. The second pair of variables captures perceived power. Specifically, I measure whether health (or tobacco) interests were listed as one of the ten most effective lobbies within a state (coded as 2), one of the top 20 groups (coded as 1) or not mentioned (coded as 0). This variable comes from a survey of public officials and political observers in each state as conducted by Thomas and Hrebenar (1999). All four of these variables come directly from Shipan and Volden’s (2006) analysis of anti-smoking legislation. I also include a measure of the percentage of adult smokers in each state. In the past, this variable has been used to capture state preferences towards anti-smoking legislation (e.g., Shipan & Volden 2006). For these analyses, it is used to test whether public opinion captures something different from a variable that measures the percentage of adult smokers in each state. In effect, I am testing whether using the percentage of state smokers is sufficient to capture important state differences in anti-smoking preferences or whether the measure developed is are better. The contemporaneous measure of the percentage of adult smokers is correlated with the proportion of neighboring states that have already adopted smoking bans in restaurants (r= -.41). In order to reduce multi-collinearity, I measure the percentage of smokers as a differenced covariate; indeed the correlation between the differenced percentage of smokers and the proportion of neighboring states that have already adopted smoking bans in restaurants is small (r= -.10). Including the differenced percentage of smokers variable also allows me to test whether changes in the usage of cigarettes influences the probability that a state will innovate, which is more directly comparable to the public opinion measure.

Pacheco 23

Political variables also matter for policy innovations with the typical expectation that states under Democratic control will be more likely to adopt anti-smoking legislation. Democratic strength is measured as a sum of percentages of state house and senate that are Democrats plus 100 if the governor is a Democrat (Bailey & Rom 2004). 7 I control for the ideological and partisan preferences of state residents by including the percentage of residents who support the Democratic Party or who are liberal as used in the previous analyses. Finally, I include a dummy variable to indicate whether a state is a tobacco producer. This is an important variable to include since Shipan and Volden (2006) find that tobacco producing states are significantly less likely to pass anti-smoking legislation. Results for the Policy Responsiveness Hypothesis As can be shown in Table 2, public opinion plays a key role in the probability that a state innovates. More specifically, the model predicts that a state that experienced the most positive change in preferences towards smoking bans has a probability of enacting smoking bans in restaurants that is 27% higher than a state that experienced a decline in support for smoking restrictions. Consistent with the policy responsiveness hypothesis, public opinion has a positive influence over whether a state innovates. This is an important finding since previous research has generally concluded that public opinion plays no or a very little role in the probability that a state innovates (e.g., Berry & Berry 1990). Interestingly, once public opinion towards smoking bans in restaurants is controlled, the influence of neighboring state policies becomes insignificant. This is especially noteworthy since previous research has found that neighboring states have a large influence over whether a state innovates. Moreover, the results in Table 2 coupled with the results from Table 1 provide

7

The democratic strength variable is not available for 2008. Consequently, I use the estimates from 2007 to account for differing levels of democratic strength across the states in 2008

Pacheco 24

strong empirical support for the social contagion model. There is evidence that state residents learn about neighboring policies, change their opinion in response to neighboring policies, and then state officials enact those policies in order to be electorally successful. The results in Table 2 suggest that the positive influence of neighboring states is due to social learning about neighboring policies by state residents. The model also predicts that states under Democratic control are more likely to adopt anti-smoking legislation. Specifically, the states under the strongest Democratic control have a 8% higher chance of adopting a smoking ban in restaurants compared to the least Democratically controlled states. The states that house liberal residents are also more likely to adopt smoking bans in restaurants (by 5%) compared to more conservative states. Other control variables are not significantly related to the adoption of smoking bans across the states. Testing the Initiative and Professionalism Hypotheses In this final section, I test for the mediating effects that the presence of the initiative and legislative professionalism may have on the relationship between state opinion and policy adoption. To test the initiative hypothesis, I interact a dummy variable that measures whether a state has a presence of initiatives with the public opinion measure. 8 According to the initiative hypothesis, states with initiatives should be more responsive to changing public opinion; the coefficient on the interaction variable should be positive and significant. To test for the professionalism hypothesis, I interact a variable that measures state legislative professionalism with the public opinion measure. Specifically, I use the Squire’s (2007) 2003 measure of legislative professionalism which includes indicators of pay, session length, and staff resources in its calculation. Squire (2007) shows this score to be reliable, valid, and stable over time. For both of the interactive models, I include the same control variables in the previous analysis. 8

I am grateful to Daniel A. Smith for providing the data on initiative states.

Pacheco 25

Results for the Interactive Effects of Institutions Results for the initiative and professionalism hypotheses are shown in Table 3. In find no empirical support for either the initiative hypothesis or the professionalism hypothesis. In the first column, the coefficient on the interaction term between the presence of initiatives and changing public support for smoking bans is insignificant. Because the presence of initiatives variable is dichotomized, it may not accurately capture variations in the use of initiatives across states (Bowler and Donovan 2004). I re-estimated this model by interacting the public opinion measure with a continuous variable that measured the annual number of initiatives used by states, as well as two indexes created by Bowler and Donovan (2004) that captures the process by which ballots are submitted. The interaction terms using these alternatives failed to achieve statistical significance. Similarly, in the second column of Table 3, the coefficient on the interaction term between the legislative professionalism measure and changing public support for smoking bans, while in the right direction, is not significant. The results presented in Table 3 provide no support for the mediating effect that initiatives or legislative professionalism has on the influence of public opinion on policy adoption. Conclusions The goal of this paper was to look at the role that public opinion plays in the diffusion of policies across the neighboring states using anti-smoking legislation as a case study. Past research on policy adoption and diffusion has largely ignored the role that public opinion plays in this process, often because of the methodological challenges in measuring state opinion over time, which has lead to inconclusive results in the past. I sought to overcome this by exploring how public preferences towards smoking bans in restaurants, measured via small estimation techniques, influence the probability that a state enacts a smoking ban in restaurants.

Pacheco 26

Through a variety of models, I find that state preferences play a key role in the the diffusion of policies across neighboring states. State citizens react to policies in neighboring states, in this case, by becoming more supportive of anti-smoking legislation after neighboring states adopt smoking bans in restaurants. As public opinion becomes more supportive of smoking bans in restaurants, a state is then more likely to enact similar policies. This finding is consistent with traditional notions of democracy and is not unexpected. What is novel about these findings, however, is that public opinion plays an important role in the diffusion of policies across proximate states. The majority of past research has found that neighboring states have a large effect over the adoption of innovations (e.g., Berry and Berry 1990). And, scholars have hypothesized that the mechanism through which neighboring states influence innovation is economic competition or social learning (e.g., Baybeck and Berry 2006). The results presented in this paper suggest a third mechanism, which I term the social contagion model. According to the social contagion model, state residents learn about neighboring policies and change their opinion in support of those policies. State officials then catch wind of changing public opinion and adopt similar policies or risk being ousted from office. Whereas traditional models of policy diffusion concentrate on the actions of state officials, the social contagion model of policy diffusion suggests that state residents play a critical role in the diffusion of policies across neighboring states. More research is needed, however, to confirm these conclusions. Are the results presented here unique to anti-smoking legislation or do they characterize the role that public opinion plays in the adoption and diffusion of innovations more generally? Perhaps state residents respond differently to issues that are less salient, therefore, less understood. Moreover, do some policies elicit negative reactions from neighboring residents? In the case study provided

Pacheco 27

here, the public responded to neighboring policies positively, although some issues may elicit public backlash after adoption. What is the mechanism through which state residents learn about neighboring policies? I’ve suggested various ways in which residents learn about policies outside their state, such as direct experience, the media, and social networks, but more research is needed to identify these factors. Are there other factors that mediate the impact that public opinion has on policy adoption? The analyses presented here suggest that policy responsiveness occurs equally across states that vary on initiatives and legislative professionalism, yet other institutional factors may be more important. Regardless, the results presented here suggest that the impact of public opinion on the diffusion of policies across states is more complex than originally thought.

Pacheco 28

Appendix A: Construction of Attitudes towards Smoking Bans in Restaurants The challenges to using national surveys to obtain valid and reliable estimates of state public opinion are well documented (Erikson, Wright, and McIver 1993, 2006, 2007; Brace et al. 2002, 2007; Berry et al. 1998, 2007; Cohen 2006; Park, Bafumi, and Gelman 2004, 2006; Lax and Phillips 2009a, 2009b). These challenges arise primarily because national surveys employ multistage area probability designs, which present two problems for estimating state opinion. First, there is no guarantee that the first stage selection of PSUs will be representative in each state. The crucial point is that while the design may be unbiased in terms of expected values, any particular implementation of the sampling design could produce a non-representative selection of PSUs for a particular state. Second, the amount of information per state is directly proportional to state population. Less populous states tend to have inadequate sample sizes. In addition, some years (e.g., 2005) have less information than others, leading to very small samples for the less populous states in certain years. Less information and inadequate samples sizes lead to imprecise estimates of state public opinion. Imprecision leads to larger variances and attenuated correlations and/or coefficients. One technique which has been shown to overcome these challenges is multilevel modeling, imputation, and post-stratification (referred to as MRP) developed by Gelman and Little (1997) and extended by Park et al. (2004; 2006), Lax and Phillips (2009), and Pacheco (2009). We begin with a multilevel model to estimate state public opinion for individuals given demographics and state. The MRP approach includes various predictors to estimate state public opinion. Following Pacheco (2009), I use gender (0=male, 1=female), race (0=non-black, 1=black), age (four categories: 18-29, 30-44, 45-64, and 65+) and education (four categories: no high school degree, high school degree, some college, and college+) for estimating public

Pacheco 29

opinion towards both education and welfare. I write the model below using indexes j,k, and l for state, age category, and education category, respectively; the subscript i refers to individual respondents.9

(2)

Level 1: Pr(yi=1) = logit-¹( β0 + β1Femalei + β2Blacki + αj[i] + αk[i] + αl[i])

(3)

Level 2: αj ~ N (0, σ²state) for j=1,…,51 αk ~ N (0,σ²age) for k=1,…,4 αl ~ N (0,σ²education) for l=1,…,4

The next step is imputation. I define each combination of demographic characteristics and state (for instance, a non-black, female, aged 18-29, with a high school degree from Connecticut) as a “person type.” Each of the 3,264 person types has an associated probability of supporting a particular policy, which is modeled in the multilevel regression as a function of state, gender, age, race and education. Imputation is conducted on each person type even if absent from the sample. After imputation, we have θc, which is the inverse logistic given the relevant predictors and their estimated coefficients (θc, is an average based on 1,000 simulations with c indexing the 3,264 unique combinations). The final stage is post-stratification. Post-stratification corrects for differences between state samples and state populations by weighting the predicted values of each person type in each state by actual Census counts of that person type in a state. For example, the 2000 Census reports that there were 581 people who were white, male, age 18-29, no high school degree, and living in Alabama: 1.7% of Alabama’s population. The imputed opinion of each person type, θc, 9

Following Park et al. (2004; 2006) and Gelman and Hill (2007), I fit the model using the Bayesian software WinBugs (Spiegelhalter et al. 1999) as called from R (R Development Core Team 2003) using Gelman’s (2003) Bugs.R. Bayesian multilevel models are especially useful for more complicated multilevel models, for example those with non-nested components, and also allow the estimation of uncertainty by using prior distributions, which are given to all parameters (Gelman & Hill 2007 345). Parameters can then be drawn from these distributions over a number of simulations. I assign normal distributions to the coefficients with means of 0 and standard deviations σ²state, σ²age, σ²educ, estimated from the data given non-informative uniform prior densities (Park et al. 2004 378).

Pacheco 30

is then weighted by the corresponding population frequencies. In the final step, we calculate the average response over each person type in each state and summarize to get point predictions and uncertainty intervals. Adding a Time Component As can be seen in the notation above, the MRP technique does not take into account time. However, we can increase the amount of information, but still preserve a time component, by pooling across a small number of years prior to executing MRP. I employ three year moving averages, pooling individual responds on surveys from the specified time. For instance, to get point estimates for 1992 using a three year pooled window, I combine estimates from 1991, 1992, and 1993 and then perform the MRP technique on this pooled dataset. The MRP process is repeated for each year after moving the time frame up a year at a time. Pacheco (2009) shows that while there is a tradeoff between the reliability of estimates and sensitivity to very short-term shocks, the efficiency benefits of pooling over a small time period outweighs the costs of biasedness.

Pacheco 31

Appendix B: Multiple Imputation at State-Year Level of Missing Years The public opinion measure towards smoking restrictions in restaurants spans from 1991 to 2006 for all 50 states and DC. All states, however, have missing data in 1994, 2004, 2007, and 2008 since questions were not asked in these years. I use multiple imputation (MI) at the state-year level to impute public opinion towards smoking restrictions in restaurants. When performing MI, the first step is to decide what to include in the models. As Honaker et al. (2009) suggest, it is crucial to include at least as much information as in the analysis model. Again, because the model is predictive and not causal it is defensible to use as many variables as possible as well as lag and lead variables in the pooled TSCS case. The following variables are included to impute preferences towards smoking restrictions in restaurants at the state-year level for 1994 and 2004: lag and lead values of % Democrat, of % Liberal, % college educated, % black population, population density logged, population logged, per capita income in constant 2000 dollars, policy restrictions in restaurants, policy restrictions in hotels/motels, policy restrictions in government workplaces, state taxes on cigarettes, % smokers, % support abortions in any circumstances, % of residents who support smoking bans in restaurants, and % of residents who support smoking bans in workplaces. And, because states may exhibit different dynamic patterns, each of these variables is interacted with the state variable for both education and welfare. Mean estimates are then calculated by taking the mean estimate for each state across the 5 imputed datasets.

Pacheco 32

Works Cited Althaus, FA and Stanley Henshaw. 1994. “The Effects of Mandatory Delay Laws on Abortion Patients and Providers.” Family Planning Perspectives 26(5): 228-231 & 233. American Lung Assocation. 2009. “2009 State Legislated Actions on Tobacco Issues Mid-Term Report.” Arceneaux, Kevin. 2002. “Direct Democracy and the Link Between Public Opinion and State Abortion Policy.” State Politics and Policy Quarterly, Vol. 2 (4): 372-387. Balla, Steven J. 2001. “Interstate Professional Associations and the Diffusion of Policy Innovations.” American Politics Research. Vol. 29 (3): 221-245. Bailey, Michael A. and Mark Carl Rom. 2004. “A Wider Race? Interstate Competition across Health and Welfare Programs.” Journal of Politics, 66(2): 326-347. Berry, Frances Stokes, and William Berry. 1990. “State Lottery Adoptions as Policy Innovations: An Event History Analysis.” APSR. Vol. 84: 395-415. Berry, Frances Stokes and William D. Berry. 1992. “Tax Innovation in the States: Capitalizing on Political Opportunity.” American Journal of Political Science, Vol. 36: 715-742. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 1960-93.” American Journal of Political Science. 42.1:337-348. Berry, William, Richard Fording, and Russell Hanson. 2003. "Reassessing the "Race to the Bottom" in State Welfare Policy." The Journal of Politics. 65(2): 327-349 Berry, William D. and Brady Baybeck. 2005. “Using Geographic Information Systems to Study Interstate Competition.” American Political Science Review, Vol. 99: 505-519. Berkman, Michael. 1994. “State Legislators in Congress: Strategic Politicians, Professional Legislatures, and the Party Nexus.” American Journal of Political Science, 38: 10251055. Boehmke, Frederick J. and Richard Witmer. 2004. “Disentangling Diffusion: The Effects of Social Learning and Economic Competition on State Policy Innovation and Expansion. Political Research Quarterly. Vol. 57 (1): 39-51. Bowler, Shawn and Todd Donovan. 2004. “Measuring the Effects of Direct Democracy on State Policy: Not All Initiatives are Created Equal.” State Politics and Policy Quarterly, 4(3): 345-363. Brace, Paul, Kellie Sims-Butler, K Arceneaux, and M Johnson. 2002. “Public Opinion In The American States: New Perspectives Using National Survey Data.” American Journal Of Political Science 46 (January): 173-189. Camobreco, John F. and Michelle A. Barnello. 2008. “Democratic Responsiveness and Policy Shock: The Case of State Abortion Policy.” State Politics and Policy Quarterly, 8 (1): 48-65. Cohen, Jeffrey E. 2006. “Conclusions: Where We Have Been, Where We Should Go.” In Public Opinion in State Politics, Jeffrey Cohen, ed. Stanford University Press. Stanford, California: Stanford University Press, 254-269. De Boef, Suzanna and Luke Keele. 2008. "Taking Time Seriously." American Journal of Political Science, 52 (1): 184-200. Erikson, Robert S., Gerald C. Wright, and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. Cambridge University Press.

Pacheco 33

Erikson, Robert S., Gerald C. Wright, and John P. McIver. 2007. “Measuring the Public’s Ideological Preferences in the 50 States: Survey Responses versus Roll Call Data.” State Politics and Policy Quarterly, 7(2): 141-151. Erikson, Robert S., Michael B. MacKuen, and James A. Stimson. 2002. The Macro Polity. Cambridge: Cambridge University Press. Fink, Stephen and Jonathan Rork. 2003. “The Importance of Self Selection in Casino Cannibalization of State Lotteries.” Economics Bulletin 8(10): 1-8. Gerber, Elisabeth R. 1996. “Legislative Response to the Threat of Popular Initiatives.” American Journal of Political Science, Vol. 40: 99-128. Gerber, Elisabeth R. and Simon Hug. 2001. “Legislative Responses to Referendum.” In Referendum Democracy: Citizens, Elites, and Deliberation in Referendum Campaigns. Eds. M. Mendelsohn and A. Parkin. Toronto: Palgrave, 191-210. Glick, Henry R. and Scott P. Hays. 1991. “Innovation and Reinvention in State Policymaking: Theory and the Evolution of Living Will Laws.” The Journal of Politics. Vol. 53 (3): 835-850. Goldstein, Adam O. and Nathan S. Bearman. 1996. "State Tobacco Lobbyists and Organizations in the United States: Crossed Lines." American Journal of Public Health, 86(8): 11371142. Gray, Virginia. 1973. “Innovation in the States: A Diffusion Study.” The American Political Science Review. Vol. 67 (4): 1174-1185. Gray, Virginia, David Lowery, and Erik K. Godwin. 2007. “Public Preferences and Organized Interests in Health Policy: State Pharmacy Assistance Programs as Innovations.” Journal of Health Politics, Policy, and Law. Vol. 32 (1): 89-127. Haider-Markel, Donald P. and Matthew S. Kaufman. 2006. “Public Opinion and Policy Making in the Culture Wars: Is There a Connection Between Opinion and State Policy on Gay & Lesbian Issues?” In Public Opinion in State Politics, Jeffrey Cohen, ed. Stanford University Press. Hill, Kim Quaile, Jan E. Leighley, and Angela Hinton-Andersson. 1995. "Lower-Class Mobilization and Policy Linkage in the US States." American Journal of Political Science, 39 (1): 75-86. Huckfeldt, Robert. 1979. “Political Participation and the Neighborhood Social Context.” American Journal of Political Science, 23: 579-621. Huckfeldt, Robert. 1986. Politics in Context: Assimilation and Conflict in Urban Neighborhoods. New York: Agathon. Huckfeldt, Robert, and John Sprague. 1995. Citizens, Politics, and Social Communication. Cambridge University Press. Huckfeldt, Robert, Eric Plutzer, and John Sprague. 1993. “Alternative Contexts of Political Behavior: Churches, Neighborhoods, and Individuals.” Journal of Politics 55: 365-81. Hyland, A., JE Bauer, Q Li, SM Abrams, C Higbee, L Peppone, KM Cummings. 2005. “Higher Cigarette Prices Influence Cigarette Purchase Patterns.” Tobacco Control 14: 86-92. Johnson, Martin, Paul Brace, and Kevin Arceneaux. 2005. “Public Opinion and Dynamic Representation in the American States: The Case of Environmental Attitude.” Social Science Quarterly. 86 (1): 87-108. Johnson, Martin. 2003. "Racial Context, Public Attitudes, and Welfare Effort in the American States." In Joe Soss, Sanford F. Schram, and Richard C. Fording (eds.) Race and the Politics of Welfare Reform. Ann Arbor, University of Michigan Press. Pacheco 34

Joyce, Theodore, Stanley K. Henshaw, Amanda Dennis, Lawrence B. Finer, and Kelly Blanchard. 2009. “The Impact of State Mandatory Counseling and Waiting Period Laws on Abortion: A Literature Review.” Guttmacher Institute. April. Kenny, Christopher B. 1992. “Political Participation and Effects from the Social Environment.” American Journal of Political Science 36 (1): 259-267. Lax, Jeffrey R. and Justin H. Phillips. 2009a. "How Should We Estimate Public Opinion in the States?" American Journal of Political Science, 53 (1): 107-121. Lax, Jeffrey R. and Justin H. Phillips. 2009b. "Gay Rights in the States: Public Opinion and Policy Responsiveness." American Political Science Review: Forthcoming. Maestas, Cherie. 2000. “Professional Legislatures and Ambitious Politicians: Policy Responsiveness of State Institutions.” Legislative Studies Quarterly. 25 (4): 663-690. Maestas, Cherie. 2003. “The Incentive to Listen: Progressive Ambition, Resources, and Opinion Monitoring among State Legislators.” Journal of Politics, 65: 439-456. Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale University Press. Matsusaka, John G. 2004. The Initiative Process, Public Policy, and American Democracy. Chicago, IL: The University of Chicago Press. McClurg, Scott. 2003. “Social Networks and Political Participation: The Role of Social Interaction in Explaining Political Participation.” Political Research Quarterly, 56 (4): 449-464. Mikesell, John L. 1987. “The Effect of Maturity and Competition on State Lottery Markets.” Journal of Policy Analysis and Management 6 (Winter): 251-253. Mintrom, Michael. 1997. “The State-Local Nexus in Policy Innovation Diffusion: The Case of School Choice.” Publius. Vol. 27 (3): 41-59. Mintrom, Michael and Sandra Vergari. 1998. “Policy Networks and Innovation Diffusion: The Case of State Education Reforms.” The Journal of Politics. Vol. 60 (1): 126-148. Mooney, Christopher Z. and Mei-Hsein Lee. 1995. “Legislative Morality in the American States: The Case of Pre-Roe Abortion Regulation Reform.” American Journal of Political Science. Vol. 39 (3): 599-627. Mooney, Christopher Z. and Mei-Hsein Lee. 2000. “The Influence of Values on Consensus and Contentious Morality Policy: US Death Penalty Reform, 1956-1982.” The Journal of Politics. Vol. 62 (1): 223-239. Mooney, Christopher Z. 2001. “Modeling Regional Effects on State Policy Diffusion.” Political Research Quarterly 54:103-124. Mutz, Diana C. and Paul S. Martin. 2001. "Facilitating Communication across Lines of Political Difference: The Role of Mass Media." American Political Science Review, 95(1): 97-114. National Cancer Institute’s State Cancer Legislation Database Program. http://www.scldnci.net/index.cfml. Accessed March 8, 2010. Norrander, Barbara. 2000. “The Multi-Layered Impact of Public Opinion on Capital Punishment Implementation in the American States.” Political Research Quarterly. Vol. 53 (4): 771793. Norrander, Barbara and Clyde Wilcox. 1999. “Public Opinion and Policymaking in the States: The Case of Post-Roe Abortion Policy.” Policy Studies Journal. Vol. 27 (4): 707-722. Nice, David C. 1992. “The States and the Death Penalty.” The Western Political Quarterly, 45(4): 1037-1048.

Pacheco 35

Pacheco, Julianna. 2009. "Using National Surveys to Measure Dynamic State Public Opinion: A Guideline for Scholars and an Application." Presented at State Politics and Policy Conference, May 22-23 in Chapel Hill, NC. Park, David K., Andrew Gelman, and Joseph Bafumi. 2005. “Bayesian Multilevel Estimation with Poststratification: State-Level Estimates from National Polls.” Political Analysis. Vol. 12 (4): 375-385. Romer, Thomas and Howard Rosenthal. 1979. “Bureaucrats Versus Voters: On the Political Economy of Resource Allocation by Direct Democracy.” The Quarterly Journal of Economics, 93(4): 563-587. Shipan, Charles, and Craig Volden. 2006. “Bottom-Up Federalism: The Diffusion of Antismoking Policies from US Cities to States.” AJPS. Vol. 50 (4): 825-843. Simon, HA. 1955. “A Behavioral Model of Rational Choice.” Quarterly Journal of Economics, Vol. 69: 99-118. Squire, Peverill. 1992. “Legislative Professionalization and Membership Diversity in State Legislatures.” Legislative Studies Quarterly. Vol. 17 (1): 69-79. Squire, Peverill. 1993. “Professionalism and Public Opinion of State Legislatures.” Journal of Politics, 55: 479-491. Squire, Peverill. 2007. “Measuring State Legislative Professionalism: The Squire Index Revisted.” State Politics and Policy Quarterly, Vol. 7 (2): 211-227. State Tobacco Activities Tracking and Evaluation (STATE) System. http://apps.nccd.cdc.gov/statesystem/statesystem.aspx?selectedTopic=100&Measure=9& dir=epi_report&ucName=UCTimeTrend. Accessed March 8, 2010. Thomas, Clive and Ronald J. Hrebenar. 1999. "Interest Groups in the States." in Politics in the American States: A Comparative Analysis, 7th ed., ed. Virginia Gray, Russell L. Hanson, and Herbert Jacob. Washington, DC: CQ Press, 113-143. Tiebolt, Charles M. 1956. "A Pure Theory of Local Expenditures." Journal of Political Economy 63: 416-24. Turner, Robert C. and Mark K. Cassell. 2007. “When Do States Pursue Targeted Economic Development Policies? The Adoption and Expansion of State Enterprise Zone Programs.” Social Science Quarterly. Vol. 88 (1): 86-103. Volden, Craig. 2002. “The Politics of Competitive Federalism: A Race to the Bottom in Welfare Benefits?” American Journal of Political Science, Vol. 46: 352-363. Volden, Craig. 2006a. “States as Policy Laboratories: Emulating Success in the Children’s Health Insurance Program.” American Journal of Political Science.. Vol. 50 (2): 294312. Volden, Craig. 2006b. “Experimenting with Welfare Reform: Emulating Success, Cutting Costs, or Racing to the Bottom?” Social Science Quarterly, Vol. 87 (4): 791-797. Walker, Jack L. 1969. “The Diffusion of Innovations among the American States.” The American Political Science Review. Vol. 63 (3): 880-889. Zukin, Cliff and Robin Snyder. 1984. “Passive Learning: When the Median Environment is the Message.” Public Opinion Quarterly 48: 629-638.

Pacheco 36

Table 1. Testing the Social Contagion Hypothesis: Error Correction Model Predicting Changes in Public Support for Smoking Bans in Restaurants (N=823) Percentage Favor Smoking Bans in Restaurants (t-1)

-.08 *** (.02)

Proportion Neighbors with Smoking Ban in Restaurants (t-1)

2.16 ** (.99)

Δ Proportion Neighbors with Smoking Ban in Restaurants

3.79 ** (1.68)

Tobacco Producer

-.60 ** (.19)

Percentage Adult Smokers (t-1)

-.21 *** (.05)

Δ Percentage Adult Smokers

-.26 *** (.07)

Percentage Liberal (t-1)

.12 *** (.03)

Δ Percentage Liberal

.06 (.07)

Percentage Democrat (t-1)

.01 (.02)

Δ Percentage Democrat

.02 (.06)

Constant

7.60 *** (1.89)

Note: Newey-West Standard Errors in parentheses. Significance levels: **.05, ***.01 with a twotailed test

Pacheco 37

Table 2. Testing the Policy Responsiveness Hypothesis: Logistic Model Predicting Policy Adoption of Smoking Bans in Restaurants (N=742) ΔPredicted Probabilities min->max Δ Percentage Favor Smoking Bans in Restaurants .19 *** 27% (.07)

Proportion of Neighboring States with Smoking Ban

.94

2%

(.95)

Ratio Tobacco Lobbyists

-5.50

0%

(20.90)

Ratio Health Lobbyists

-5.26

-2%

(4.42)

Power Tobacco Lobbyists

.16

1%

(.47)

Power Health Lobbyists

-.03

0%

(.31)

Δ Percentage Adult Smokers

-.09

-1%

(.12)

Democratic Strength

.01 ***

8%

(.003)

Percentage Liberal

.15 **

5%

(.06)

Percentage Democrat

-.04

-2%

(.04)

Tobacco Producing State

-.86

-1%

(.76)

Constant

-6.42 *** (1.99)

Note: Robust Standard Errors in parentheses clustered by state-year. Significance levels: **.05, ***.01 with a two-tailed test. To caculate predicted probabilities, all other variables were kept at their mean levels.

Pacheco 38

Table 3. Testing the Initiative and Professionalism Hypotheses: Logistic Model Predicting Policy Adoption of Smoking Bans in Restaurants (N=742) Initiative Hypothesis Δ Percentage Favor Smoking Bans in Restaurants Presence of an Initiative

Professionalism Hypothesis

.19

.15

(.10)

(.15)

1.28 (.73)

Initiative*Percentage Favor Smoking Bans in Restaurants

-.04 (.11)

State Legislative Professionalism

2.86 (3.69)

Professionalism*Percentage Favor Smoking Bans in Restaurants

.17 (.57)

Proportion of Neighboring States with Smoking Ban Ratio Tobacco Lobbyists Ratio Health Lobbyists Power Tobacco Lobbyists Power Health Lobbyists Δ Percentage Adult Smokers Democratic Strength

1.36

1.16

(.94)

(.93)

10.11

1.86

(21.30)

(21.72)

-6.17

-8.44

(4.97)

(5.79)

.16

-.23

(.52)

(.59)

.06

-.12

(.32)

(.33)

-.10

-.11

(.12)

(.12)

.01 *** (.003)

Percentage Liberal Percentage Democrat Tobacco Producing State Constant

.01 *** (.003)

.15 ***

.12

(.05)

(.06)

-.01

-.04

(.04)

(.04)

-.78

-.79

(.82)

(.77)

-8.63 *** (2.22)

-6.01 *** (2.12)

Note: Robust Standard Errors in parentheses clustered by state-year. Significance levels: **.05, ***.01 with a two-tailed test.

Pacheco 39

Table A1. Survey Organization, Question Wording, Coding of State Public Opinion for Smoking Bans in Restaurants Question Wording

Current Population Survey

Gallup

In restaurants, do you think that smoking should be allowed in all areas, in some areas, or not allowed at all?

What is your opinion regarding smoking in public places? First, in restaurants should they set aside certain areas, should they totally ban smoking, or should there be no restrictions on smoking?

Years of Survey

Coding

1=not allowed at 1992, 1993, 1995, 1996, all, 0=in some 1998, 1999, 2001, 2002, areas, allowed in 2006, 2007 all areas 1=totally ban, 1990, 1991, 1994, 1999, 0=set aside 2000, 2001, 2003, 2005, areas, no 2007 restrictions

Pacheco 40

Figure 1. Theoretical Models of Policy Diffusion of Policy 1 from State A to State B

The Social Learning Model and Economic Competition Model State A Adopts Policy 1

State B Adopts Policy 1

State B State Officials

The Social Contagion Model

State A Adopts Policy 1

State B Adopts Policy 1

State B State Officials

Institutions State B Public Opinion on Policy 1

Pacheco 41

Figure 2. Maps of Percentage of State Residents Favoring a Smoking Ban in Restaurants

Pacheco 42