Printer Friendly

What's legal and what's not: the regulation of opiates in 1912.

This paper develops a model to explain state-level opiate regulation in 1912. The personal choice of whether to engage in mainstream or deviant activity is determined by consumption technology and market prices, and voting determines the legality of deviant behavior. Voting outcomes depend on population characteristics including diversity, tolerance, visiblilty of deviance, and the distribution of consumption efficiencies. A logit equation whose dependent variable is the presence of a state opiate prescription law is estimated. Results broadly support the collective choice model and disconfirm the role of interest groups, particularly physicians, in determining prescription regulation.


Alcohol and nicotine may be the most dangerous drugs sold today. Considering their effects on health and accidents, legal restrictions on the sale and private use of tobacco and alcohol are minor. Some currently illegal substances, including marijuana and opiates, might in fact be less harmful than alcohol and tobacco ff used in moderation. Why are some commodities legal and others not? Economists should be interested in the question for two reasons. The first is that the work of non-economists often uses economic metaphors. A number of historians have described the prohibition of alcohol and opiates in America in terms of interest-group politics.(1) Economists, however, have yet to explain these episodes using their own models of regulation and collective choice. The second reason is that the historical facts are dramatic enough to tax the explanatory power of economics. Between 1890 and 1920, private transactions in alcohol and opiates went from being virtually unregulated by the federal government to complete illegality.(2) This change in legal status was not a response to new scientific knowledge about the substances.

Becker's model of household production |1981, chs. 1 and 2~ formalizes the questions at issue. lannaccone |1988~ has recently used Becker's model to analyze the sociological distinction between mainstream churches and religious sects. Assuming individuals have differing consumption technologies, he derives propositions about the characteristics of religious sects and their members. Iannaccone |1988, $242-45~ notes that churches are characterized by inclusivity, and sects by exclusivity. Church members have more heterogeneous beliefs, socialize mainly with people from outside their church, and place few demands on their church for non-religious support services. Churches are more tolerant of diverse behavior and place fewer restrictions on the conduct of members. Church members span a wider range of demographics than do sect members. They are on average economically better off than sect members, but sect members contribute more funds to their institutions.

The distinction between churches and sects is analogous to that between licit and illicit drug use. Drug users are demographically more homogeneous than nonusers, more poorly socialized, and more likely to spend a major portion of their time in drug-related activity.(3) They have lower productivity than alcohol or tobacco users, both at work and in household consumption,(4) and are more likely to restrict themselves to companions with similar consumption patterns. Kandel and Maloff |1983~ suggest that learning to use drugs in the subculture resembles an initiation into the rites and beliefs of a deviant sect. According to Winick |1974~, the exit from drug-taking, like that from a sect, is often a total "burnout" to zero use which occurs at a predictable time. The church-sect model yields such predictions about drug use and generalizes individualistic models of that activity.(5)

In the next section, I model an individual who receives utility from consumption itself and from the attitude or state of mind in which consumption takes place. Consumption and attitude are the outputs of household production functions. After deriving the optimum in a two-activity model I analyze the choice between mainstream and deviant attitudinal activities. These production functions are sometimes called activities. Those attracted to deviant activities are less productive at mainstream consumption, and more likely to become obsessively attached to deviance. In section III the individual choice model is combined with a political model of regulation. This step leads to tests which will discriminate between majority-voting and interest-group models of the decision to regulate deviant activity. Section IV compares the predictions of the two models with regard to the attributes of state opiate laws in the years before federal law preempted state regulation. Prior to the Harrison Act of 1914, individual states regulated the labeling, prescription requirements, and allowable concentrations of drugs which could be sold without prescription. According to the U.S. Public Health Service |1912, 34-41~, in 1912 opiates could be sold without prescription in twenty-two of the forty-eight states and the District of Columbia. To anticipate the results, a majority-voting model is a better predictor of the likelihood of a prescription law than an interest-group model. In some ways the results seriously conflict with the predictions of an interest-group model.


Maximization by the Individual

An individual i is assumed to value two primary commodities, consumption, |C.sup.i~, and attitude, |A.sup.i~. Utility is a quasi-concave function of their quantities.

(1) |U.sup.i~ = |U.sup.i~(|C.sup.i~, |A.sup.i~).

Consumption and attitude are analogues of Iannaccone's secular and religious goods. Attitude is a state of mind which both increases the marginal utility of the consumption good and is valued by itself. Like lannaccone's religious good, attitude is in reality multidimensionaL "including assurance of personal salvation |or a drug user's analogue~, fellowship with others, spiritual enlightenment, and so forth" |1988, $245~. A full treatment of drug use and socialization would require an intertemporal model such as Becker and Murphy's |1988~ or Chaloupka's |1991~. The propositions of importance here, however, require only a sing~e-period formulation.

Each good is the output of a household production function. For consumption,

(2) |C.sup.i~ = f(||T.sub.i~.sup.C~, ||X.sub.i~.sup.C~~, ||S.sub.i~.sup.C~, |D.sup.i~).

One method of producing attitude uses X, the same market commodity as the consumption activity,

(3a) ||A.sub.i~.sup.x~ = |g.sup.x~)(||T.sub.i~.sup.Ax~,||X.sub.i~.supp.A~,||S.sub.i~.sup .Ax~,|D.sup.i~).

A second method is described below. ||T.sub.i~.sup.C~ and ||T.sub.i~.sup.Ax~ are time spent in each activity. A unit of time can be allocated to only one activity, and time spent working can be allocated to neither. || and ||X.sub.i~.sup.A~ are market goods, whose use is also exclusive between activities. Marginal products of X and T are positive and decreasing in both activities. Individuals differ in the productivities with which they combine these inputs to obtain C and A.

||S.sub.i~.sup.C~ and ||S.sub.i~.sup.Ax~ are stocks of experience capital accumulated from past indulgence in each activity. They too are assumed to have positive and diminishing marginal products in each.(6) Whether one's attitudinal habit is mainstream (drinking) or deviant (illegal drugs), it requires learning such things as how to enjoy the substance and how to administer the proper dose to oneself. Such experience capital might be efficiently obtained by socializing with one's peers.(7) According to the behavioral literature, both the setting of the drug experience and the biochemistry of the drug contribute to the user's enjoyment. Interaction with others during the drug experience also enhances its quality.(8)

|D.sup.i~ is the individual's demeanor, a measure of lifestyle. To choose |D.sup.i~ is to choose such things as an ethic and a personal appearance. Such choices are only incidental~y connected with time or market goods.(9) Examples include self-imposed standards such as diet, abstention from alcohol, items of dress, and respect for the property of others. I assume that implementing a demeanor requires negligible direct expenditure of time or money. A standard of demeanor will affect the attainable levels of both |C.sup.i~ and |A.sup.i~. The chosen level of Di is assumed to be unalterable across the consumption and attitudinal activities, i.e., one cannot credibly display differing demeanors in the two of them.

The individual maximizes utility subject to a time constraint, which can be transformed into a full-income constraint:

(4) |W.sup.i~(T* - ||T.sub.i~.sup.C~ - ||T.sup.i~.sub.A~ = |P.sub.x~(||X.sup.i~.sub.C~ + ||X.sup.i~.sub.A~).

The individual's market wage is Wi, and the price of market goods is |P.sub.x~. Hours of work are the difference between total time available, |T*~, and time spent in the two consumption activities. For simplicity, I assume that the individual does not value leisure. Maximizing a Lagrangian, L, with objective (1) subject to (2a), (2b), and (3) yields the following conditions (subscripts indicate partial derivatives, superscripts relating to agent i are suppressed):(10)

(5a) |Delta~L/|Delta~|X.sub.C~ = |U.sub.1~|f.sub.2~ - |Lambda~|P.sub.X~ = 0

(5b) |Delta~L/|Delta~|X.sub.A~ = |U.sub.2~||g.sub.2~.sup.X~ - |Lambda~|P.sub.X~ = 0

(5c) |Delta~L/|Delta~|T.sub.C~ = |U.sub.1~|f.sub.1~ - |Lambda~W = 0

(5d) |Delta~L/|Delta~|T.sub.AX~ = |U.sub.2~||g.sub.1~.sup.X~ - |Lambda~W = 0

(5e) |Delta~L/|Delta~ = |U.sub.1~|f.sub.4~ + |U.sub.2~||g.sub.4~.sup.X~ = 0

Together with the full-income constraint (4) (5a) through (5e) can be solved for optimal values of ||X.sub.i~.sup.C~, ||X.sub.i~.sup.A~, ||T.sub.i~.sup.C~, ||T.sub.i~.sup.A~, |D.sup.i~, and the Lagrange multiplier |Lambda~. Since |U.sub.c~ and |U.sub.A~ are positive, (5e) states that unless the same value of D is coincidentally optimal for both consumption and attitude, its marginal products in the two activities will be oppositely signed.

To introduce deviance, assume an alternative attitudinal activity. If chosen, it must be to the exclusion of (3a). This activity uses a different market good, Y, which is only of value for the production of attitude.

||A.sub.y~.sup.1~ = |g.sup.y~(||T.sub.Ay~.sup.i~, |Y.sup.i~, ||S.sub.Ay~.sup.i~, |D.sup.i~

The marginal products of |g.sup.x~ and |g.sup.y~ have the same properties. As before, Di is assumed unalterable between (3b) and the consumption activity. In what follows, the attitudinal activity which uses X will be called mainstream, and that using Y will be called deviant. For example, X might be a composite market good which includes alcohol, and Y might be opiates. Maximizing utility with (3b) substituted for (3a) and Y substituted for X in (4) yields optimum conditions analogous to (5a)-(5e). The choice of the mainstream or the deviant attitudinal activity is then determined by which of the two maximizations yields the higher utility level.

Geometry of the Model

To derive the feasible set of consumption-attitude combinations geometrically, assume that the amounts of time and market goods in the consumption and attitudinal activities have already been chosen and are fixed.(11) Assume first that one attitudinal activity has been chosen, to the exclusion of the other. In Figure 1a, C and A then depend only on D. The maxima of C and A, denoted C* and A*, are assumed to require similar but not identical standards of demeanor, denoted |D.sub.1~ and |D.sub.2~. A single choice of D must apply to both C and A. If D |is less than~ |D.sub.1~, increasing it will improve productivity in both activities. If D |is greater than~ |D.sub.2~, increasing it will degrade productivity in both. Between |D.sub.1~ and |D.sub.2~, a change in D will improve productivity in one and degrade it in the other.

Figure 1b shows the resulting feasible set. If D |is less than~ |D.sub.1~ or D |is greater than~ |D.sub.2~, the corresponding boundary of the feasible set is upward sloping, since changing D changes both A and C in the same direction. The maximum attainable C (at C*) corresponds to demeanor |D.sub.1~, and the maximum attainable A (at A*) corresponds to demeanor |D.sub.2~. The downward-sloping range corresponds to demeanors between |D.sub.1~ and |D.sub.2~. It is continuous if A(D) and C(D) are continuous, and is locally concave if (but not only if) the two functions are locally concave.(12) The tangency between the highest attainable indifference curve I* and the frontier determines optimal demeanor and the corresponding mix of activities. If A(D) and C(D) have the same maximum, the relevant frontier is a single point.

If the maxima of the two functions are far apart, the boundary of the feasible set can become convex. In Figure 2a, let C(D) and |A.sub.2~(D) be the only possible activities. The feasible set then takes a shape such as F2 in Figure 2b. For D |is less than~ |D.sub.1~ or D |is greater than~ |D.sub.3~, its boundary slopes upward. Even if indifference curves are convex, the optimum choice of demeanor might be close to that which maximizes A or C alone, leaving the other activity at its residual level. It is also possible that the chosen D will lie between |D.sub.1~ and |D.sub.2~ because the curvature of the frontier is less than that of the indifference curves.(13)

Now let a second attitudinal activity such as |A.sub.1~(D) be available. If |A.sub.1~(D) is a leftward lateral translation of |A.sub.2~(D), the associated frontier will be |F.sub.1~ which lies entirely above |F.sub.2~ except at A* and C*. If |A.sub.1~(D) is an option, a utility maximizer will never choose to engage in |A.sub.2~(D). If |A.sub.1~(D) is shifted further leftward toward C(D), the frontier eventually becomes convex. If |A.sub.1~(D) is both a lateral and a downward translation of |A.sub.2~(D), |F.sub.1~ and |F.sub.2~ can cross. If a person has a choice between such an |A.sub.1~(D) and |A.sub.2~(D) but cannot engage in both, he may choose the less productive attitudinal activity |A.sub.1~(D) if its optimal demeanor is close to that which maximizes C(D).

Productivity Change and the Choice of Activities

Now return to a single consumption activity C(D) and a single attitudinal activity A(D). Let C(D) grow in productivity for all possible D, while its maximum remains at the same D as before (e.g., an equiproportional upward shift). If the productivity of A(D) does not change (or changes more slowly in a way which is not overly biased), the optimal level of demeanor eventually moves closer to the one which corresponds to the maximum of C(D). If the productivity of A(D) is not changing, this movement cuts attitudinal output.

Now let there be two mutually exclusive attitudinal activities. In Figure 3a, |A.sub.1~(D) is the mainstream activity, whose output is assumed to be maximized at D*, the same demeanor which maximizes the output of consumption activity C(D). |A.sub.2~(D), for which more distant demeanors are productive, is the deviant activity. Assume that any learning which enhances the productivity of |A.sub.2~(D) has already occurred and that no further improvements are possible. Also let C(D) and |A.sub.1~(D) both become more productive with cumulated experience, but not necessarily at the same rates.

If C(D) and |A.sub.2~(D) are the choices, the resulting feasible set is |F.sub.2~ in Figure 3b. If indifference curve |I.sub.1~ is optimal, demeanor close to |D.sub.2~ will be chosen. If alternatively C(D) and |A.sub.1~(D) are the possible activities, the only relevant portion of the feasible set is the single point (|A.sub.1~*, C*), with associated demeanor D*. (All other demeanors produce less of both goods.) The individual in Figure 3b optimizes by choosing deviant activity |A.sub.2~(D) along with a relatively low level of C(D). As cumulated consumption raises the productivity of C(D), C* moves upward. Eventually, (|A.sub.1~*, C*) comes to lie above |I.sub.1~, and mainstream attitudinal activity |A.sub.1~(D) replaces the deviant activity. Deviance ends not because it ceases to be pleasurable, but because as efficiency in consumption increases, the demeanor necessary for deviance exacts too high a cost in lost consumption. The deviant activity is abandoned in full, rather than being slowly cut at the margin. This is a plausible interpretation of the well-documented "burnout" of illegal drug use as individuals go through their twenties.(14)


A Majority-Voting Model

If people have different household technologies and market wages, they will choose different consumption bundles and demeanors. To motivate a collective-choice model, assume that deviant demeanor produces disutility for a mainstream consumer, and that the disutility increases with the distance from one's own demeanor.(15) At any moment, there will be a distribution of individual demeanors, such as h(D) in Figure 4, with deviant demeanor in a tail.(16) Assume that the question is whether to discourage demeanor above D* by outlawing or regulating the market good associated with it. As one possible formalization, assume that a person will favor the measure if his chosen demeanor lies more than K units away from D*. One whose demeanor is closer to D* will vote against regulation.

An individual can only vote for regulation if it is on the ballot. There must be a coalition which can force the issue to a vote, and it must be large enough to win a majority. In Figure 4, over 50 percent of the mass of the distribution lies below D* - K, so a majority coalition is feasible. Let the likelihood of regulation be increasing in the number of winning coalitions which can conceivably be formed; i.e., the larger the number of such coalitions, the more likely at least one of them will cover the cost of forming it. There are N voters, M |is less than~ N/2 of whom have D |is greater than~ D* - K. The number of winning coalitions is given by

|Mathematical Expression Omitted~

The larger is K, the smaller is N/2 - M, and the less likely a vote takes place.

Next, consider a mean-preserving spread of the D distribution where a majority favors regulation. As an example, move one individual downward from the mean, and move another upward from the mean by the same number of units. This increases the variance of the distribution. Referring to Figure 4, the person moved upward may change his vote from regulation to tolerance, and will definitely not change it in the opposite direction. The person moved downward will remain in favor of regulation. Under a mean-preserving spread, the number of potential winning coalitions for regulation will either shrink or not increase.

As a first approximation assume that K is uniform among voters.(17) Its effective value, however, depends on the visibility of deviance. A person who is out of sight and out of mind may in fact have a demeanor which I would find repulsive on my doorstep. If such a person were on my doorstep, I would be more interested in discouraging that demeanor. The effective size of K shrinks if deviant demeanor is more visible to mainstream members. Deviance and the associated market goods are more likely to be regulated in such a society.

Finally, a population of voters who are more productive at mainstream activities will be an easier pool from which to draw a winning coalition than an unproductive one. The more productive population, for example, will be more literate and therefore easier for a political entrepreneur to cull a winning coalition from. Such a population will also have a higher opportunity cost of political participation. In light of Wolfinger and Rosenstone's |1980~ findings, I assume that the net effect is an increase in the likelihood of voting.

An Interest-Group Model

In contrast to a majority coalition, interest groups may dominate the decision to regulate. Legislation and regulation will reflect group pressures to transfer wealth to themselves. Small, organized interests whose individual members have appreciable wealth at stake will carry an advantage over diffuse, unorganized ones. The cost of political organization will be lower for a more concentrated interest which is already in existence, for example as a professional association. Free rider problems will be easier to overcome in a small group, particularly if the group's sanctions can impose substantial costs, for example by suspending a non-cooperator's license to practice.

In an interest-group model, the likelihood of a certain regulation will increase with the strength of those who gain from it, and decrease with the strength of those who lose from it. The characteristics of those unorganized individuals whose personal stakes are small will be of little importance in determining its likelihood, even if the aggregate of these stakes is large. Regression analyses of regulation and legislation, such as those of Michaels |1980~ and Stigler |1971~, often proxy the strength of interests by measures of their size.

If a majority-coalition model holds, the likelihood of a regulation will be determined by factors which differ from those suggested by the interest-group model. There are four determinants of a collective choice to regulate in the majority-coalition model. First, the more productive the average voter is in mainstream activities, the more likely that deviance will be regulated. Second, the greater the variance of educational attainment or workplace productivity, the larger the fraction of the population who will be relatively inefficient at the mainstream activity and the less likely the deviant activity will be regulated. Third, the higher the variance of individual demeanors, the less likely that deviance will be regulated, since the population from which a pro-regulatory coalition can be drawn becomes relatively smaller. Fourth, the more visible the deviant activity, the more likely it will be regulated. These four factors should have minimal effects in an interest-group model. There, the differential strengths of organized interests will be the determinants of regulation.


The Harrison Act

In 1914 the Harrison Act became law.(18) It replaced a patchwork of state narcotics and poison legislation with a federal registration and record-keeping requirement for all transactions in opiates. Only medical practitioners and pharmacists who held federal tax stamps (sold for $1 per year) were allowed to prescribe and dispense opiates. As written and originally interpreted, the law did not explicitly limit volumes which could be transacted. Instead it imposed a "good faith" standard on prescriptions. Patent medicines with low concentrations of opiates were intended to remain beyond the law's reach as enacted. It seems clear from the record that support for the act by physicians and pharmacists was economically motivated.(19) The Supreme Court originally held that it allowed prescriptions to maintain the habits of addicts, which appears consistent with legislative intent.(20) Within five years of enactment, however, the Court reversed itself and held maintenance prescriptions illegal.(21)

An analysis of Congressional voting on the Harrison Act immediately suggests itself, e.g., as in Kalt and Zupan |1984~. Unfortunately, all votes taken were voice votes. As an alternative, the U.S. Public Health Service compiled all state laws regarding opiates and poisons in effect immediately before passage of the act. The compilation included data on states in which opium and opium compounds could only be dispensed by prescription.(22) The dependent variable in the reduced-form equations below equals one in states requiring prescriptions and zero in states not requiring them. Its use entails a timing problem: thirty-one of the forty-eight states and the District of Columbia instituted such laws between 1907 and 1915.(23) It also entails a measurement problem: There is no available indication of the stringency of enforcement or the sizes of penalties.

Independent Variables

Given the available state-level data, the independent variables corresponding to the theoretical constructs are fairly crude aggregates:

Productivity in Mainstream Consumption. Wages measure both workplace productivity and the time cost of consumption. Since average hourly earnings of all workers in a state are unavailable, I use average annual compensation of wage workers in manufacturing in 1909.(24) It should be positively correlated with the likelihood of a prescription law. Efficiency in mainstream consumption activity also increases with education. Since data on years of schooling were not collected until the 1930s, I use the percentage of individuals ten years old or over reported as literate.(25)

Variance of Efficiency in Mainstream Consumption. If proxies for mean earnings or education are hard to find, data on their distributions pose an even larger problem. As the best of a bad lot, I estimated the standard deviation and coefficient of variation of 1919 average annual full-time earnings by industry.(26) Greater variance of earnings, as noted above, would decrease the likelihood of regulation. The theory does not, however, indicate whether absolute variability or variability relative to the mean is the better measure. Available data do not suggest an obvious proxy for the variance of the educational distribution.

The Range of Tolerated Mainstream Activities. I use two measures of the diversity of tastes in consumption and attitudinal activities. The first is a Herfindahl index of the concentration of religions in a state, constructed from the U.S. Bureau of the Census's |1910~ Special Census of Religious Bodies, taken in 1906. The index is the sum of squares of shares of the population claiming to be Catholics, Protestants, members of other churches, and members of no church.(27) One expects greater tolerance (less likelihood of a prescription requirement) in states with lower values of the index.(28) The second measure of heterogeneity of tastes is the proportion of a state's population born in foreign countries.(29) The presumption is that individuals born elsewhere would be more likely to engage in attitudinal activities which differ from those of the native-born. The more numerous the foreign-born, the less able the native-born majority will be to control deviant activities.(30)

Visibility of Deviant Activity. A first measure of the visibility of deviance is the percentage of the state's population resident in urban areas.(31)

Interest Groups. The interest groups gaining from a prescription law are easy to identify, but not to count. Physicians, taken as a fraction of a state's population, are the only relevant occupation counted in the 1910 Census.(32) With data on pharmacists unavailable, I measure their professional strength by the number of years for which the State Pharmaceutical Association had been existence prior to 1912.(33) Liquor dealers are likely to gain marginally from restricting opiates and should thus favor prescription laws.(34)


Census Physicians

Table I presents three logit regressions which include the independent variables described above.(35) The dependent variable in each equals one if a prescription is required for opiates, and zero if it is not. The equation in the first column uses the standard deviation of the earnings distribution as a measure of variation in consumption efficiency. The second column uses its coefficient of variation. Consistent with the model, in both equations the fraction of the population which is urbanized significantly increases the likelihood of a prescription requirement. The first measure of heterogeneity of tastes, the religious Herfindahl index, enters both equations with the correct sign, but a relatively large standard error. The second, the fraction of the population foreign born, enters as predicted in both, with a negative coefficient over twice its standard error.

The first measure of productivity in mainstream consumption, the average wage in manufacturing, enters with the expected positive sign and a coefficient over twice its standard error. The second measure, the percentage of the adult population which is literate, is insignificant in both equations. Both measures of the variability of efficiency in consumption, the TABULAR DATA OMITTED standard deviation and coefficient of variation of earnings, enter with the correct signs but with coefficients identical in size to their standard errors.

Regarding interest-group measures, liquor outlets per capita enter the equations as expected, positively signed and with coefficients twice their standard errors. The age of the state pharmacy association never enters an equation significantly. The most problematic variable, however, is physicians per capita. It invariably enters the equation with a negative coefficient three times its standard error, in clear violation of the interest group model. The third column of Table I provides further indication that physicians are important in a reduced-form specification: exclusion of this variable markedly worsens the fit and lowers the ratios of all other coefficients to their standard errors. The negative coefficient on physicians is robust to different specifications of the regression and is not the result of outlier observations.(36) The coefficient is also not a reflection of state size. When population is added as an independent variable, its coefficient is insignificant while that of physicians remains significantly negative.

A possible explanation for the negative coefficient is that there are free rider problems in organizing large numbers of physicians to support prescription legislation. One argument against this reasoning is historical. According to Burrow |1963, chs. 1-3~, physicians were among the first occupations to succeed in forming adjuncts of their professional group, the American Medical Association, in every state. By 1912 they had already succeeded in the major task of imposing licensing requirements to practice in nearly all states. Another argument against free riding is statistical. Reestimate the logit equation using the total number of physicians in a state rather than physicians per capita. If the free rider hypothesis is correct, the likelihood of a prescription requirement should decline with the total number. In the reestimation, the coefficient of this variable is insignificant.

The 1919 Treasury Report

One possible reason for the negative coefficient on physicians is that some of those in practice at the time disapproved of opiate addiction and would not write prescriptions to maintain addicts. I thus attempted to estimate the number of physicians with an active interest in prescribing to addicts. My estimate uses a unique 1918 survey conducted by the U.S. Treasury Department |1919~. Charged with enforcement of the Harrison Act, the Department sent questionnaires to all registered physicians, requesting information on the number of addicts they were treating. The date is fortuitous, since it is after the 1916 Supreme Court decision which ruled that maintenance prescriptions were allowed by the Act and before the 1919 decisions which effectively reversed this holding.

The Treasury report supplied state-level figures which included the response rate of physicians and the number of addicts under treatment. Addicts were defined by the judgement of the respondent. Viewing it as politics, Musto |1987, 134-140~ argues that the survey was intended to produce an inflated estimate of addicts. In Table II, physician interest is given by the number of physicians responding to the survey as a fraction of the state population. While some physicians treating addicts undoubtedly failed to respond, it seems reasonable to assume that a larger proportion of nonresponders were in fact not treating any addicts.(37) The survey also provides a potentially better proxy for pharmacist interest, total reported opiate prescriptions, again converted to a per capita basis.(38) In 1918, four years into the Harrison Act, prescriptions were required for all opiates, but the Jin Fuey Moy decision then in effect held that prescriptions for addict maintenance were legal. Assuming that demand conditions did not change greatly between 1912 and 1918, the prescription figure is thus a measure of market size.

Table II presents regressions in which physician pressure is measured by the per capita number responding to the Treasury survey. The regressions also include opiate prescriptions per capita as a measure of pharmacist pressure. Other variables are TABULAR DATA OMITTED those of Table I. Because of unavailable data, four states must be dropped from the sample.(39) The results are not always similar to those of Table I, but the coefficient of the new measure of physician interest remains unambiguously negative. In the left-hand column of Table II, the coefficients of the urban and foreign-born fractions of the population remain over twice their standard errors, and that of the religious Herfindahl becomes likewise. The coefficient of the wage, however, falls to insignificance, and that of literacy remains insignificant. Among the interest-group variables, liquor outlets per capita becomes insignificant as does opiate prescriptions per capita. As in Table I, this measure of physicians carries a significantly negative coefficient. The potentially sharper measure of physician interest also disconfirms an interest-group theory.

The second equation in Table II contains an additional variable, reported prescriptions per physician responding to the Treasury survey. It is an attempt to measure the intensity of the typical physician's interest in a prescription law. As noted above, prescriptions are from the survey of 1918, when a federal reporting requirement was in effect and prescriptions for addict maintenance were allowed. The likelihood of a prescription requirement also varies inversely with this variable. When it is included in the equation, the coefficient of physicians per capita responding to the survey remains significantly negative. The coefficient of reported prescriptions per capita, however, becomes positive and 1.5 times its standard error, consistent with an interest group theory of regulation.


Past authors have successfully applied the household production model to both market and nonmarket activities. Here I have combined it with a collective-choice model to deal with the origin of laws which regulate deviance. lannaccone |1988~ has used household production theory to operationalize the sociological distinction between churches and sects. His method is applicable to formally similar questions about the choice between mainstream and deviant drug use. Incorporated into a collective choice paradigm, that model yields predictions about the effects of economic and demographic variables on the choice of regulations. Such a formulation also allows a reduced-form test to evaluate whether the control of deviance is better explained by collective choice or interest-group theory.

American policy toward opiates at the turn of the century provides a hitherto neglected data set with which to test the model. The states were almost evenly divided in 1912 between those requiring and not requiring prescriptions. Regressions to explain the policy choice work surprisingly well, particularly in light of the quality of available data. The equations, however, leave about as many new puzzles in their wake as existed prior to their estimation. In particular, the unqualifiedly significant and incorrectly signed coefficients on both measures of physician pressure are strong evidence against an interest-group model.

Historians have interpreted federal opiate legislation as stemming in part from pressure by physician and pharmacy interests. Although the Harrison Act was originally interpreted by the courts as allowing maintenance prescriptions for addicts, a historical accident led to the reversal of this holding a short time afterwards. With overwhelming public and medical opinion against opiates today, the roots of the laws we currently live with bear further examination. Economists have successfully studied the origins of much legislation and regulation. It appears that they can also produce useful insights into narcotics law.

1. See Kerr |1985~, Kyvig |1979~, Musto |1987~, Odegard |1928~, and Sinclair |1962~.

2. According to the U.S, Public Health Service |1912~, the only material federal restrictions on opium prior to 1914 were a prohibition against its importation by Chinese persons and restrictions on its sale by Americans in China and the Pacific. Szasz |1974~ describes how opiate laws came to concentrate on the Chinese.

3. Recent waves of the National Longitudinal Survey of Youth have included questions on illegal drug use. Mensch and Kandel |1988~ and Gill and Michaels |1991~ examine the characteristics of drug users and non-users in this population.

4. Recent work questions this view of the drug earnings relationship. Using data from the National Longitudinal Survey of Youth, Gill and Michaels |1992~ find that drug users earn higher wages than similar non-users, after accounting for the self-selection of users. Using similar methods, Berger and Leigh |1988~ find that the use of alcohol increases earnings.

5. Such models include those of Barthold and Hochman |1988~, Becker and Murphy |1988~, Michaels |1988~, and Winston |1980~.

6. Cumulated use of a substance may in fact produce tolerance which leads to a need for higher doses. In an intertemporal model, this can lead to "burnout" and compulsive restarring of a once-abandoned habit, as in Michaels |1988~. Since the fact of drug use rather than the mechanism of addiction is of interest in the text, I do not further consider this explicitly dynamic complication. Becker and Murphy |1988~ and Chaloupka |1991~ model this aspect of addiction.

7. See Feldman |1968~ and Kandel and Maloff |1983~. Such socialization exists for both legal and illegal substances. Most smokers report that their first cigarette was in fact an unpleasant experience, and many drinkers only learn about proper dosages the hard way. Kaplan |1983, ch. 1~ notes that many individuals find their first sample of heroin so unpleasant that they never use it again.

8. See Zinberg |1984, chs. 5 and 6~. Non-social drug use, particularly of alcohol, is an important phenomenon. The limits of toleration, however, seem to be set by its social aspects. As an example, Kerr |1985~ notes that the most important pro-prohibition lobby was the Anti-Saloon League, whose efforts largely dealt with bars as social institutions rather than as retail liquor outlets. Most people view heavy drinkers who stay at home as having a medical problem, to be dealt with through voluntary treatment. Only those who socialize on Skid Row or drive drunk in traffic are considered apt subjects for law enforcement. The medical nature of alcohol "abuse" is debatable. See Fingarette |1988~.

9. For applications to religious activity, see lannaccone |1988, S247~.

10. I assume all constraints are binding in ways which ensure an interior solution, for reasons which will become clear below. Hence the text does not show the full set of Kuhn-Tucker conditions.

11. To verify the graphics, numerical experiments have been run with explicit functional forms for the activities. Using the notation of the text and suppressing subscripts, functions of the form

C = |X.sup.|Alpha~~|T.sup.|Beta~~ ||sigma~.sup.-1~ exp { - ||(D - |D.sub.1~)/||omicron~~.sup.2~}

for both C and A were optimized to derive the feasible set. The optimal values of X and T in each activity vary with the point chosen on the feasible set. If they are held fixed between activities, the constrained feasible set lies within the unconstrained feasible set, except possibly for a single tangency point.

12. Iannaccone |1988, S2625-66~ derives a more precise set of conditions under which concavity holds.

13. Michaels |1988~ shows that the interior equilibrium is unlikey to persist if there is productivity change in the household technology.

14. See Winick |1974~ for a summary of such studies.

15. Mainstream demeanor may also create disutility for deviants. In what follows, I assume that deviants are sufficiently rare or incompetent that they cannot mobilize politically. I also do not consider the possibility that deviants may form a coalition with near-deviants in order to counter a mainstream movement. Drug users have been oddly ineffective in forming such alliances.

16. For simplicity I restrict the discussion to the case where deviants are in one tail of the distribution. The model can be extended to one in which the critical demeanors are in both tails.

17. This is not innocuous, since it is assuming that one's tolerance is independent of his choice of mainstream or deviant demeanor. As casual counterexamples, members of religious sects may have lower levels of tolerance than members of mainstream churches, and drug users may have more of a live-and-let-live attitude toward mainstream consumers than mainstream consumers have toward them.

18. Internal Revenue Code of 1954, 4701-36.

19. Musto |1987, ch. 3~ describes the pressure on Congress by organized medicine and pharmaceutical trade associations. The original draft of the Act was extensively rewritten to satisfy both interests.

20. U.S. v. Jin Fuey Moy, 241 U.S. 394 (1916).

21. U.S. v. Doremus, 249 U.S. 86 (1919); Webb et al. v. U.S., 249 U.S. 96 (1919).

22. U.S. Public Health Service |1912, 34-31~.

23. An alternative dependent variable is the year a prescription requirement became law, as compiled by Hamowy |1987, 10-11~ This measure is also subject to the timing problem. Regressions using it were generally inferior in performance to those of the text.

24. U.S. Bureau of the Census |1910, abstract 525-27~. Total wages in the source are divided by the number of persons employed. No data are available on hours.

25. See U.S. Bureau of the Census |1913, abstract 248~.

26. Earnings data are in many cases estimates from secondary sources, compiled by Leven |1925, 100-4~ Because of missing data, earnings in Trade were assumed equal to those in Domestic Services, and earnings of those in Transport, Professional, Public Service, and Clerical Work were assumed equal to those in "Miscellaneous Industries." Employment by industry and state in 1910 is from U.S. Bureau of the Census |1913, vol. IV 96-151~.

27. Data are from U.S. Bureau of the Census |1910, 46-49~. Splitting Protestants into Baptists and others does not noticeably affect the results. Beyond this distinction, there is no obvious ranking of tolerance among Protestant denominations. (I lumped Mormons with Protestants.) The index ranges from .396 (Louisiana) to .766 (Wyoming, most of whose population were not churchgoers).

28. As an alternative measure of diversity, I constructed a Herfindahl index of the 1910 concentration of males in a state in nine major occupational categories, taken from U.S. Bureau of the Census |1910, vol. IV 96-151~. Its value ranges from .166 (Nevada) to .596 (Mississippi) It never entered a regression significantly.

29. There is a large literature on ethnic variations in attitudinal activity. Bennett and Ames |1985~ show that the appropriate occasions for alcohol consumption, the proper amounts to consume, and attitudes toward overindulgence vary greatly among religious and ethnic groups in the U.S. Data are not available to correct for citizenship, which would be a more reasonable measure of political power. There is no appreciable difference if males (who alone could vote at the time) are used instead of a total.

30. The foreign-born figures are from U.S. Bureau of the Census |1913, abstract 83~. Despite the frequently-alleged (and often nonexistent) link between narcotics and foreigners discussed by Szasz |1974~, the opiate-using population was at the time dominated by rural and southern adults. They frequently acquired and maintained the habit by consuming patent medicines which contained opium. Blacks were underrepresented in that population, and were thought to be at greater risk of using cocaine. See Musto |1987, 6 and 98~ and Terry and Pellens |1928~.

31. Data are from U.S. Bureau of the Census |1913, abstract 56~.

32. Data are from U.S. Bureau of the Census |1913, vol. IV, 96-151~.

33. See Sonnedecker |1976, 379-80~. I count reorganizations of a previously existing association as giving birth to a new one. This only affects five states.

34. Liquor dealer data are in North |1911, 408~. According to Pekkanen |1980, 92~, other drugs, including alcohol, are frequently substituted for opiates when supplies are short. Because alcohol tax rates vary widely among states, tax revenues are of little use in determining the volume of profit at stake. If prescriptions are not required, alcohol retailers might also sell opiates, but there is no evidence that this is of any importance.

35. Coefficients of probit regressions on the same independent variables invariably display the same patterns of sign and significance.

36. Most prescription laws in effect at the time allowed physicians to dispense drugs if they wished to do so instead of sending patients to a pharmacist. A prescription requirement is thus valuable to physicians in both their roles as therapists and as drug dispensers.

37. The response pattern is broadly consistent with known facts about the addict population at the time. For example, response rates in southern states, known to contain more addicts, were higher than those in the mountain states, known to contain fewer. See Ball and Chambers |1970~.

38. An estimate of the number of pharmacists cannot be constructed from data in the Treasury report.

39. Excluding those state from a recomputation of Table I does not alter its results.


Barthold, Thomas A., and Harold M. Hochman. "Addiction as Extreme-Seeking." Economic Inquiry, January 1988, 89-106.

Ball, John C., and Carl D. Chambers. The Epidemiology of Opiate Addiction in the United States. Springfield, Ill.: Charles C. Thomas, 1970.

Becker, Gary S. A Treatise on the Family. Cambridge: Harvard University Press, 1981.

Becker, Gary S., and Kevin M. Murphy. "A Theory of Rational Addiction. Journal of Political Economy, August 1988, 675-700.

Bennett, Linda A., and Genevieve M. Ames. The American Experience with Alcohol: Contrasting Cultural Perspectives. New York: Plenum Press, 1985.

Berger, Mark C., and J. Paul Leigh. "The Effect of Alcohol Use on Wages." Applied Economics, December 1988, 1343-51.

Burrow, James G. AMA, Voice of American Medicine. Baltimore: Johns Hopkins Press, 1963.

Chaloupka, Frank. "Rational Addictive Behavior and Cigarette Smoking." Journal of Political Economy, August 1991, 722-42.

Feldman, Harvey W. "Ideological Supports to Becoming and Remaining a Heroin Addict." Journal of Health and Social Behavior 9(1), 1968, 131-39.

Fingarette, Herbert. "Alcoholism: The Mythical Disease." The Public Interest, Spring 1988, 3-22.

Gill, Andrew M., and Robert J. Michaels. "The Determinants of Illegal Drug Use." Contemporary Policy Issues, July 1991, 93-103.

-----. "Does Drug Use Lower Wages?" Industrial and Labor Relations Review, April 1992, 419-34.

Hamowy, Ronald. "Introduction: Illicit Drugs and Government Control," in Dealing with Drugs, edited by R. Hamowy. Lexington: Lexington Books, 1987, 1-34.

Iannaccone, Laurence R. "A Formal Model of Church and Sect." American Journal of Sociology, Supplement 1988, S241-S268.

Kalt, Joseph P., and M. A. Zupan. "Capture and Ideology in the Economic Theory of Politics." American Economic Review, June 1984, 279-300.

Kandel, Denise B., and Deborah R. Maloff. "Commonalities in Drug Use: A Sociological Perspective," in Commonalities in Substance Abuse and Habitual Behavior, edited by P. Levison, D. Gerstein, and D. Maloff. Lexington: Heath Lexington, 1983, 3-27.

Kaplan, John B. The Hardest Drug: Heroin and Public Policy. Chicago: University of Chicago Press, 1983.

Kerr, K. Austin. Organized for Prohibition. New Haven: Yale University Press, 1985.

Kyvig, David E. Repealing National Prohibition. Chicago: University of Chicago Press, 1979.

Leven, Maurice. Income in the Various States. New York: National Bureau of Economic Research, 1925.

Mensch, Barbara S., and Denise B. Kandel. "Do Job Conditions Influence the Use of Drugs?" Journal of Health and Social Behavior, June 1988, 169-84.

Michaels, Robert J. "Bureaucrats, Legislators, and the Decline of the State Mental Hospital." Journal of Economics and Business, Spring 1980, 198-205.

-----. "Addiction, Compulsion, and the Technology of Consumption." Economic Inquiry, January 1988, 75-88.

Musto, David F. The American Disease, expanded ed. New York: Oxford University Press, 1987.

North, S.N.D. The American Year Book 1910. New York: Appleton, 1911.

Odegard, Peter. Pressure Politics: The Story of the Anti-Saloon League. New York: Columbia University Press, 1928.

Pekkanen, John R. "Drug Law Enforcement Efforts," in The Facts about Drug Abuse, edited by Patricia Wald for the Drug Abuse Council. New York: The Free Press, 1980, 63-94.

Sinclair, Andrew, The Era of Excess. Boston: Little, Brown, 1962.

Sonnedecker, Glenn. Kremers and Urdang's History of Pharmacy, 4th ed. Philadelphia: Lippincott, 1976.

Stigler, George J. "The Theory of Economic Regulation." Bell Journal of Economics and Management Science, Spring 1971, 3-21.

Szasz, Thomas S. Ceremonial Chemistry: The Ritual Persecutions of Drugs, Addicts, and Pushers. New York: Doubleday, 1974.

Terry, Charles E., and Mildred Pellens. The Opium Problem. New York: Bureau of Social Hygiene, 1928.

U.S.Bureau of the Census. Special Reports: Religious Bodies 1906, Pt. 1. Washington: Government Printing Office, 1910.

-----. Thirteenth Census of the United States 1910. Washington: U.S. Government Printing Office, 1913.

U.S. Public Health Service. Digest of Laws and Regulations in Force in the United States Relating to the Possession, Use, Sale, and Manufacture of Poisons and Habit-Forming Drugs. Public Health Bulletin No. 56. Washington: Government Printing Office, 1912.

U.S. Treasury Department. Traffic in Narcotic Drugs. Report of the Special Committee of Investigation, Appointed March 25, 1918, by the Secretary of the Treasury. Washington: Government Printing Office, 1919.

Winick, Charles. "Some Aspects of Careers of Chronic Heroin Users," in Drug Use: Epidemiological and Sociological Approaches, edited by E. Josephson and E. Carroll. Washington: Hemisphere, 1974, 51-73.

Winston, Gordon C. "Addiction and Backsliding: A Theory of Compulsive Consumption." Journal of Economic Behavior and Organization 1(4), 1980, 295-324.

Wolfinger, Raymond E., and Steven J. Rosenstone. Who Votes? New Haven: Yale University Press, 1980.

Zinberg, Norman E. Drug, Set, and Setting. New Haven: Yale University Press, 1984.

Professor of Economics, California State University, Fullerton, Calif. 92634. The author thanks Andrew Gill, Laurence Iannaccone, Sam Peltzman, Rodney Smith, participants at a Claremont Graduate School seminar, and an anonymous referee for helpful comments. An earlier version of this paper was presented at the 1989 Annual Meeting of the Public Choice Society.
COPYRIGHT 1992 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Michaels, Robert J.
Publication:Economic Inquiry
Date:Oct 1, 1992
Previous Article:Writing as a responsibility of science: a reply to Laband and Taylor.
Next Article:Enforcement costs and the choice of policy instruments for controlling pollution.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters