How should we wage the war on drugs? Determinants of public preferences for drug control...

When citizens are presented with alternative policy solutions to a given social problem, why do they choose to support one over the other(s)? Much scholarship has investigated the strength and determinants of public support for government spending on such issues as education, crime, welfare, space exploration, and national defense (Hansen, 1998; Inniss & Sittig, 1996; Jacoby, 1994). Other studies have investigated the sources of public support for different components of welfare policy (Hasenfeld & Rafferty, 1989; Pereira & Van Ryzin, 1998). However, few studies have examined determinants of public choice among competing strategies to combat the same social problem. In light of growing evidence that public opinion has effects on public policy outcomes (Burstein, 1998; Monroe, 1998), it is important to understand how the public chooses to support one policy option versus another, since many policy debates revolve around not whether but, rather, how government should allocate resources.

A clear example of such a debate is that surrounding American drug control policy. During the past 2 decades, six consecutive presidential administrations have waged a "war on drugs," emphasizing domestic criminal justice and international interdiction programs. Although for several years, the public has doubted the success of these strategies (Cintron & Johnson, 1996; Pew Research Center for the People and the Press [Pew], 2001), citizens have continued to offer overwhelming support for government efforts to combat the drug problem (Blendon & Young, 1998; Rasinski, Timberlake, & Lock, 2000). Yet studies have found no such public consensus on the means by which government should intervene (Cintron & Johnson, 1996; Lock, Timberlake, & Rasinski, 2002). In short, everyone wants to do something about the drug problem. What that something is, however, is still a matter of considerable public debate.

In this article, we analyze a survey of residents of the five largest U.S. metropolitan areas to understand public support for spending on three major components of American drug control policy: law enforcement programs, rehabilitative services for addicts, and school-based prevention programs. Our goals are twofold: First, we compare the explanatory power of three models of public support for social policy-self-interest, political socialization, and policy attitudes. Second, we investigate whether and how the effects of variables derived from these models change across two dimensions of support for spending-total support for drug control spending and preferences for one policy alternative versus the others.

This article contributes to the literature on public support for social policy in at least two ways. First, by utilizing three prominent models of public support for social policy, we provide a more theoretically grounded and comprehensive description of public attitudes toward American drug control policy than is currently available. A broader contribution derives from our analysis of determinants of policy preferences. We know of no other study that investigates determinants of both overall support for spending on a social problem and choices among competing policy alternatives. Our results suggest that a more comprehensive understanding of the determinants of public support for social policy can be achieved by conceptualizing support along both dimensions.

Public Attitudes Toward the "War on Drugs"

The federal drug control budget has exploded over the past 2 decades, with expenditures increasing more than 600% in real terms (2000 dollars) from 1981 to 2000. As Figure 1 demonstrates, the lion's share of this growth was devoted to domestic criminal justice programs. Expenditures in this area increased from $0.84 billion (29% of the budget) in 1981 to $8.83 billion (49%) in 2000. During this period, the share of the total drug control budget allocated to treatment declined from 36% to 16%, although absolute expenditures did increase in real terms (from $1.05 billion to $2.92 billion) (Office of National Drug Control Policy [ONDCP], 1995, 2002). Expenditures on prevention and interdiction also increased, with the share of the total drug control budget increasing for prevention (from 7.6% to 13.0%) and decreasing for interdiction (from 27.2% to 21.5%).

The dominant and increasing role of criminal justice programs in American drug control policy has received harsh criticism in the academic and popular press (e.g., Gray, 2001; Jonas, 1994; Nadelmann, 1992; Soros, 1997). Many critics claim that current drug control strategy is not only unnecessarily punitive but also largely ineffective. Recent research has shown that the public agrees that the drug war has not been successful; however, in contrast to much anti-drug war rhetoric, the public has consistently and strongly supported government efforts at combating the drug problem. Research has also found much public disagreement about the strategies used to wage the drug war.

For example, Koven & Shelley (1993) analyzed a 1989 CBS News/New York Times poll and found strong approval (69%) for the Bush administration's "proposals for fighting the drug problem." However, they also found that although 44% of those surveyed thought the administration had struck the right balance between spending on law enforcement and drug treatment and education, 32% thought the budget overemphasized law enforcement, whereas 8% thought it focused too much on treatment and education. Cintron & Johnson (1996) analyzed the 1995 National Opinion Survey on Crime and Justice. They found that respondents were dubious about the past success of the war on drugs-64% reported no effect of the drug war on drug use, and only 18% said the war on drugs had reduced the amount of drug use in their communities. Respondents were divided equally on the question of whether supply reduction ("military border controls" and "police street-level efforts") or demand reduction ("drug education programs" and "drug treatment programs") policies were superior. Blendon & Young (1998) found that although Americans generally do not think the war on drugs has succeeded, 87% of those surveyed supported increased funding for the police. A smaller percentage (though still a majority-77%) supported increased spending on drug treatment programs. Timberlake, Rasinski, & Lock (2001) found that between 1984 and 1998, 64% of respondents to the General Social Survey supported increased spending on "dealing with drug addiction," with a lower percentage (59%) supporting increased spending on "drug rehabilitation." Finally, 74% of respondents to a February 2001 survey (Pew, 2001) believed that "we are losing the drug war" and that "demand is so high we will never stop drug use." Respondents rated interdiction (48%) and arresting drug dealers (19%) as the "most effective" drug control policies, followed by drug education (15%), treatment (10%), and arresting drug users (4%).

Models of Public Support for Social Policy

The analyses described above have contributed important information about aggregate public support for various drug control policies. However, we still know little about the sources of public preferences. What factors are most important in leading citizens to support drug control spending? How does the public decide among competing policy alternatives? Many scholarly investigations of such questions have employed one or more of three explanations: self-interest, political socialization, and attitudes about the individuals targeted by policy or characteristics of the policies themselves.

Self-Interest and Political Socialization

The oldest theory of public support for social policy is based on rational choice models of economic self-interest. Classical self-interest theory contends that citizens calculate the expected costs and benefits of policy alternatives, choosing to support the policy that maximizes some utility function. Scholars have suggested that citizens make utility maximizing decisions at three levels of aggregation. At the individual level, self-interest theory proposes that citizens prefer policies that they believe will advance their own interests or the interests of close friends and family members (Citrin & Green, 1990). A broader version of self-interest theory holds that individual preferences are based in part on judgments about how a policy would affect the well-being of the social group with which one strongly identifies (Bobo & Kluegel, 1993). Finally, scholars have proposed that citizens sacrifice individual and group self-interest in public policy opinions in favor of "societal interest" or the interests of the nation as a whole (Funk, 2000).

The primary alternative to self-interest theory contends that policy attitudes are based primarily on the early socialization of individuals with fairly stable sets of political ideologies and values (Sears & Funk, 1999). In this view, policy attitudes are not derived ad hoc from the rational calculation of individual, group, or societal interest but, rather, are generated by more or less stable predilections to support certain types of policies. Both self-interest and political socialization emphasize the effects of general and stable (at least in the short-term) characteristics of individuals, such as political party or socioeconomic status. Cook & Barrett (1992) refer to such characteristics as "dispositions," which tend to affect support for similar social policies in similar ways. For example, political liberals are expected to favor policies that provide generous social supports for individuals, whereas conservatives are expected to favor policies that place restrictions or requirements on recipients (Pereira & Van Ryzin, 1998).

Policy Attitudes

A third strain of scholarship stresses the effects on social policy support of citizens' attitudes about the groups targeted for help by policy, and about characteristics of policies themselves. This model acknowledges that dispositional attributes are important causes of social policy attitudes. The policy attitudes model complements self-interest and political socialization theory by suggesting that individuals within a political party or socioeconomic class may hold heterogeneous, unstable attitudes about social polices and their beneficiaries (Cook & Barrett, 1992).

Causal Attributions. Scholars have suggested that citizens are more likely to support social spending if they believe the targeted recipients of spending are not to blame for their plight and therefore deserve help. Effects of these "causal attributions" have been examined in studies of public support for a wide array of social groups, such as African Americans (Bobo & Kluegel, 1993), the poor (Cook & Barrett, 1992), the unemployed (Feather, 1985), and AIDS patients and applicants for organ transplants (Skitka & Tetlock, 1993). In the context of drug policy, Covington (1987) and Littrell and Diwan (1995) found that beliefs in environmental causes of addiction were associated with greater support for heroin legalization. Littrell and Diwan (1995) also found that beliefs that heroin is intrinsically addictive were associated with reduced support for legalization.

Policy Effectiveness. Scholars have also suggested that citizens base support for social policy in part on assessments of the effectiveness of that policy. Hasenfeld & Rafferty (1989) found greater support for welfare spending among individuals who believed welfare programs were efficiently run. Cook & Barrett (1992) found positive effects of perceived policy effectiveness on support for social welfare spending. Littrell & Diwan (1995) found no effects of assessments of the effectiveness of drug policies on support for heroin legalization. Finally, Pereira & Van Ryzin (1998) found that beliefs that welfare programs were ineffective were negatively associated with support for spending on the poor and positively associated with restrictive welfare reforms, such as fingerprinting welfare applicants and imposing time limits on the receipt of benefits.

Data and Measures

The data for this study come from the first wave of the Multi-City Study of Attitudes about Addiction (MCSAA). The MCSAA was a two-wave panel survey designed to assess the impact of viewing a nationally broadcast, five-episode television program about drug addiction on attitudes toward addiction and drug control policy. In March and April 1998, the National Opinion Research Center (NORC) drew a simple random sample of individuals with listed telephone numbers in each of the five largest Designated Market Areas (DMAs) in the United States: New York, Los Angeles, Chicago, Houston, and Philadelphia. DMAs were provided to NORC by the Nielsen Company and correspond to the broadcast areas of major network affiliates in the central cities of these five metropolitan areas. NORC interviewers completed a total of 1,360 phone interviews. Because the MCSAA was designed to evaluate the effects of watching a television program with a fixed broadcast schedule, time constraints limited follow-up calls to three to reach sample members who did not answer the telephone or to convert people who initially refused the interview. As a result, the unconditional response rate was about 30%, and the cooperation rate was 46% (American Association for Public Opinion Research, 1998, p. 20). These rates are virtually identical to those obtained by other polling organizations based on the same number of follow-up calls (Gallup, 2001).'

Dependent Variables

Respondents to the MCSAA were asked whether we are spending "too much" (scored -1), "about the right amount" (scored 0), or "too little" (scored +1) on six drug control policies: (1) law enforcement costs related to drug use, (2) programs aimed at preventing illegal drugs from entering the country, (3) drug treatment programs, (4) health care services for addicts, (5) employment and other services for addicts, and (6) school-based drug prevention programs. In exploratory factor analyses, we found that items (1) and (2) loaded strongly on one varimax-rotated factor, and items (3) through (5) loaded on another. Item (6) occupied an intermediate position, loading strongly on both factors. We therefore created indexes of support for law enforcement by averaging answers to items (1) and (2) (Cronbach's (alpha = 0.64), and for rehabilitation spending by averaging items (3) through (5) (alpha = 0.70). We measured support for prevention spending with item (6). We then constructed an index of "overall support for drug control spending" by summing these three measures.2 We constructed three additional dependent variables by subtracting each of the three policy support measures from the two others. Table 1 presents descriptive statistics for all variables used in the analysis.

Independent Variables

Self-interest. We measured individual self-interest with two variables. Because parents have strong interests in protecting children from the problems associated with drugs, we accounted for the presence of minor children in the household. We also included a measure of whether or not the respondent is close to a current or recovering addict, reasoning that individuals would have interests in seeing close friends or family members in treatment facilities rather than in prison.4 Following Bobo & Kluegel (1993), we operationalized group self-interest with a four-- category measure of ethnicity. Finally, we measured societal interest by summing respondents' scores on assessments of the importance to society of the problems of drug and alcohol addiction (alpha = 0.68).

Political Socialization. We operationalized political socialization as self-reported political ideology, political party affiliation, and religious affiliation (conservative and mainline Protestant versus other religious affiliations). Prior research has shown that law enforcement drug control policies tend to appeal to individuals on the conservative end of the political spectrum, with the opposite being true for drug rehabilitation programs (Cintron & Johnson, 1996).

Policy Attitudes. We constructed a measure of individuals' beliefs that addicts are to blame for their plight by summing responses to items concerning the role of individual willpower in causing addiction (alpha = 0.59). We measured beliefs in a physiological addiction etiology by summing responses to items concerning the role of brain chemistry in causing addiction (alpha = 0.45).5 We then subtracted the latter index from the former to construct a measure of beliefs in willpower versus physiological causes of addiction. We measured policy effectiveness attitudes by summing scales of respondents' beliefs about the effectiveness of law enforcement (alpha. = 0.72), rehabilitation (alpha = 0.60), and prevention (alpha = 0.63) approaches to drug control. We created an index of overall belief in drug control policy effectiveness by summing scores on the three indexes (alpha = 0.70). We also created three additional independent variables by subtracting each effectiveness index from the two others.

Control Variables. We included measures of age, gender, education, family income, urban residence, and DMA of residence in our analysis to control for spuriousness associations between the other independent variables and support for drug spending. Finally, because we imputed missing values of income by taking the mean income of respondents' race x education x marital status category, we also included a dummy variable for imputed income.

Results

We present results from ordinary least squares (OLS) regression models in Tables 2 and 3. The models in Table 2 estimate effects of self-interest, political socialization, and policy attitudes on support for overall drug control spending, with positive coefficients indicating greater spending support. In Table 3, positive coefficients indicate more support for the policy listed first in the column heading, and negative coefficients indicate the reverse. In both tables, Model 1 includes self-interest, political socialization, and control variables, whereas Model 2 adds measures of policy attitudes. We present standardized coefficients for continuous variables and semistandardized coefficients for dummy variables. Coefficients on continuous variables should be interpreted as the number of standard deviations of change in the dependent variable associated with a one standard deviation change in the independent variable. Coefficients on dummy variables should be interpreted as the difference (in standard deviations of the dependent variable) between the category of the variable listed in the table and the omitted category.6 Regression diagnostics indicate that collinearity among the independent variables is not a concern in this analysis.' We use alpha levels of .01 and .05 for all statistical tests.

Support for Drug Control Spending Overall

The results in Table 2 show that two self-interest measures were significantly associated with total support for drug control spending. In Model 2, African Americans showed greater support for spending than Whites and Hispanics. Second, beliefs that addiction is an important social problem were strongly associated with support for increased spending. Neither knowledge of a current or former addict nor the presence of minor children in the household was significantly related to spending support.

In Model 2, none of our measures of political socialization was associated with overall support for drug spending at conventional significance levels. Beliefs that lack of willpower causes addiction were negatively associated with support for drug control spending, whereas beliefs that drug policy is effective were positively associated with spending support. The addition of the three policy attitudes increases the proportion of explained variation in Model 3 by 87% over Model 1. The results in Model 2 suggest that the overall level of drug control spending is not a partisan issue, nor one that triggers individual self-interest. Rather, support for drug control spending appears to depend on racial group membership, perceptions of how important the problem is to society, and attitudes about the causes of drug addiction and the effectiveness of drug control policy.

Support for Drug Control Policy Alternatives

A somewhat different picture emerges when we conceptualize (in Table 3) support for drug spending as preferences for policy alternatives. Whereas Whites showed less support than African Americans for drug spending overall, Whites showed greater support for law enforcement and prevention versus rehabilitation. The effect of societal interest was not significant in two of the comparisons, and substantively small (beta = 0.067) in the third. This suggests that beliefs in the importance of addiction for society do not lead to preferences for one method of combating the drug problem versus another. We also find that net of the effects of our measures of self-interest, political conservatives, Republicans, and conservative Protestants all favored law enforcement to rehabilitation spending. These findings echo those in Timberlake et al. (2001), who found less support among conservative sociopolitical groups for drug rehabilitation relative to a generally worded version of drug policy. In addition, Republicans and mainline Protestants showed greater support for prevention than for rehabilitation, and political conservatives, Republicans, and conservative Protestants showed greater support for law enforcement than for prevention.

Thus, whereas support for drug control spending overall was dominated by racial group membership and perceptions of the importance of the problem of addiction for society, choices between policy alternatives hinged much more on political socialization. Whites, Republicans and mainline Protestants diverged sharply from African Americans, Democrats, and members of other religions over spending on drug rehabilitation in particular but were as differentiated in terms of support for law enforcement and prevention (note the nonsignificant "White" and "mainline Protestant" coefficients and the smaller, marginally significant "Republican" coefficient in Model 2 of the right-hand column). Political conservatives and conservative Protestants appear to be strongly in favor of law enforcement spending and did not distinguish between prevention and rehabilitation (note the nonsignificant coefficients in Model 2 of the middle column).

Each of our measures of policy attitudes was significant in the full models in Table 3. That is, net of effects of self-interest and political socialization, beliefs in a willpower versus physiological addiction etiology were associated with support for law enforcement spending over both prevention and rehabilitation and with prevention over rehabilitation. Beliefs in the effectiveness of each policy over the others were also associated with support for that policy. Note that effects of political socialization remain fairly stable from Model 1 to Model 2, indicating that the effects of political socialization are not simply mediated by policy attitudes.

Conclusions

Our analysis leads to two conclusions about the nature of public support for drug policy in particular and social policy in general. First, this analysis demonstrates that scholars should draw sharp distinctions between public support for general actions by government, such as "helping the poor" or "solving the drug problem," and support for particular policy proposals. We found that the effects of many conventionally used variables on support for drug control policy depended deeply on the conceptualization of "support." With the notable exception of race, described above, support for overall spending on the drug problem was not determined by traditional social cleavages, such as income, political ideology, party, or religion. On the other hand, support for each policy approach versus the others did cleave along traditional sociopolitical lines. Put differently, we found that when support is conceptualized as the total level of support for drug control spending, conservatives and liberals are strikingly in agreement. However, given the choice among policy alternatives, these groups diverge sharply. Drug policy may be somewhat unusual in that the two primary sets of policy alternatives-law enforcement and rehabilitation-have traditionally appealed to polarized ideological camps. Nevertheless, virtually all policy debates concern some choice between alternatives; thus, a more comprehensive understanding of the determinants of public support for social policy can be gained by taking into account choices between competing alternatives.

A second conclusion from this research is that scholars must continue to pay close attention to policy attitudes in their attempts to understand not just public opinion but the entire policy formation process. Classical models of public opinion propounded that citizens support candidates and issues largely on the basis of self-interest and party loyalty (Campbell, Converse, Miller, & Stokes, 1960). However, the percentage of respondents to the General Social Survey reporting an independent party affiliation increased from 27% in 1972 to 38% in 1998 (Davis & Smith, 1998). Other sources of group identification have also declined in the past 4 decades. For example, union membership has declined from over 35% in 1965 to about 10% in 1996 (Wilson, 1999, p. 58). Thus, politicians and policy strategists have been forced to appeal to public attitudes and sentiments in attracting public support for or opposition to social policy. The 1988 Bush campaign's "Willie Horton" campaign is perhaps the most notorious example of this strategy; however, there are countless others. For example, much of the political rhetoric surrounding welfare reform in the early 1990s was focused on convincing the public that the poor were responsible for their plight and therefore did not deserve help (Bane & Ellwood, 1994). Republicans derailed President Clinton's proposed Health Security Act in 1993 in large part by portraying the bureaucracy required to implement the act as hopelessly complex and inefficient (Koch, 1998). The Gore campaign in 2000 attacked George W. Bush's tax cut proposal not only by appealing to the self-interest of lower- and middle-class voters but also by attempting to convince the public that it was unfair to devote such a large proportion of the tax cut to a small stratum of citizens.

We conclude by stressing the importance of policy attitudes for understanding the past and future course of American drug control policy. Nearly a decade of research has found strong support among the public for government intervention in the drug problem. Although the public has been wavering in its certainty that the war on drugs is succeeding, it seems unlikely that citizens will support a massive de-escalation on all fronts. As we have argued, the central question for research on public attitudes toward drug control policy is not whether but, rather, how citizens want government to address the drug problem. We suggest that one reason the massive growth in the drug budget has been disproportionately devoted to law enforcement spending is that the public has supported an individual willpower model of addiction over the past few decades. For example, the "Just Say No" campaign of the Reagan administration and the "This is Your Brain on Drugs" public service announcements stress the importance of individual choices in causing or preventing addiction. However, in recent years critics have castigated the drug war for its ineffectiveness and unnecessary punitiveness, based in part on the argument that addiction results from physiological factors rather than moral failings. If the public responds to this shift in rhetoric, we should expect to see increasing support for spending on drug prevention and rehabilitation and waning support for the current course of the war on drugs.

Acknowledgments

This research was supported by the Robert Wood Johnson Foundation, Substance Abuse Policy Research Program Grant 031608, and by a predoctoral fellowship from the Population Research Center, NORC and the University of Chicago, NICHD Grant 2 T32-HD073-02. The data file and computer programs used in the analyses for this article are available at www.norc.uchicago.edu or by contacting Kenneth A. Rasinski. Principal Research Scientist, NORC and the University of Chicago, 1155 E. 60th St., Chicago, IL 60637.