FROM ISSUE NUMBER 12 ~ SUMMER 2012 GO TO TABLE OF CONTENTS

TEXT SIZE A A A

The Drug-Policy Roulette

JONATHAN P. CAULKINS and MICHAEL A. C. LEE

In June 2011, the Global Commission on Drug Policy soberly proclaimed: "The global war on drugs has failed, with devastating consequences for individuals and societies around the world." In its report, the commission — members of which include economists, policy experts, and several former world leaders — argued that, 40 years after President Richard Nixon launched the U.S. War on Drugs, circumstances today demand a new approach. Among the commissioners' recommendations, two stand out: to "end the criminalization, marginalization and stigmatization of people who use drugs but who do no harm to others," and to "encourage experimentation by governments with models of legal regulation of drugs," particularly cannabis, "to undermine the power of organized crime and safeguard the health and security of their citizens."

The commission's advice echoes four decades of arguments by advocates of legalization, who have long promoted their cause as a simple solution to the violence, disease, incarceration, social decay, and other ills fostered by the drug trade. Recently, the case for legalization has appeared to gain steam, drawing support from groups ranging from doctors who favor legitimizing medical marijuana to political libertarians to African-American leaders who believe that legalization can alleviate many problems facing the black community. And that case is not wholly without merit: Allowing legal businesses to produce and distribute drugs would in fact drive criminal organizations out of the market and, at the same time, eliminate a great deal of the harm they do to society.

But the arguments for legalization often overlook its considerable downsides and risks. They serially underplay, for instance, the possibility of substantially increased use of and dependence on drugs. Though no one really knows precisely how much drug use would go up if it were legalized, advocates tend to disingenuously offer exact estimates favorable to their cause — suggesting that they can know with confidence that increased use would be limited and controllable. This false certitude neglects the fact that no nation in the modern era has legalized the production of any of the major illegal drugs for unsupervised use. (Even the Netherlands allows only retail sales of marijuana, not production or wholesale distribution.) Legalization is thus a leap into uncharted and potentially dangerous waters.

Given these concerns, it has become fashionable for legalization's proponents to moderate their requests, asking only for "experiments" with legalization. The implicit (if not explicit) promise is that if legalization turns out badly, we can always re-impose prohibition — ending up no worse off, and at least a bit wiser.

Before we consider "experimenting" with legalization, though, we must first educate ourselves about its likely consequences. What effect would such experiments have on prices, use rates, and criminal activity? The evidence suggests that only by legalizing the expensive drugs — cocaine and crack, heroin, and meth — could policymakers bring about the social and economic benefits proponents claim, as legalizing marijuana alone would not alter the basic character of the drug problem. But legalizing those more expensive drugs would also likely yield enormous drawbacks. Moreover, the notion that societies can legalize any currently illicit drug only temporarily — incurring no real risks or irreversible damage along the way — appears to be false. Policymakers should think long and hard about some key facts before opening the door to drug legalization, including the "experimentation" approach — which is at best a misnomer, and at worst a perilous deceit.

THE COSTS OF PRODUCTION

To understand the consequences of legalization, it is essential to start with a crucial factor too often ignored: prohibition's effects on production costs.

Today, illegal drugs are shockingly expensive. Even run-of-the-mill "commercial grade" marijuana sells for $100 per ounce. Cocaine and crack, heroin, and methamphetamine all sell for $100 or more per pure gram, making them more valuable than gold. (If cigarettes cost that much, a standard pack of 20 would carry a price tag of roughly $2,000.) There is no physical reason why drugs need to be so expensive: Drug "labs" are rudimentary affairs; with a few exceptions (such as lysergic acid diethylamide, or LSD), they employ at most undergraduate-level chemistry. And the associated crops — coca, cannabis, poppy — are neither difficult to grow nor resource-intensive; they are in fact often produced in some of the least-developed nations of the world.

What makes illegal drugs so expensive is precisely the fact that their production is prohibited, and that this prohibition is often strictly enforced. One factor is what economists call "compensating wage differentials," or compensation for taking risks. Suppliers of illegal drugs court real dangers, including arrest, imprisonment, physical injury, even death. Thus, in addition to seeking wages that compensate for their time and allow for normal profits, people employed in drug distribution also seek compensation for assuming these risks — much as coal miners and deep-sea divers typically earn higher wages than people performing similar jobs under less hazardous circumstances.

Another factor is the inefficiency that stems from having to operate covertly. The precautions required to evade detection make the production of drugs very labor intensive. Grocery-store cashiers, for instance, are more than 100 times as productive as retail drug sellers in terms of items sold per labor hour. Similarly, hired hands working for crack dealers can fill about 100 vials per hour, whereas even older-model sugar-packing machines can fill between 500 and 1,000 sugar packets per minute. This labor intensity of drug production, combined with the high wages demanded for that labor, are what drive up the costs of drugs; by comparison, materials and supplies — glassine bags, gram balances, and even guns — are relatively cheap.

How would these dynamics — and, with them, drug prices — be altered if drugs were legalized? To begin, legalization would cut production costs in source countries. Cultivators and preparers would no longer fear crop eradication; the need to employ only manual labor would disappear. Production could expand and mechanize, becoming more efficient and therefore less expensive.

The real reductions, however, would come in the area of transport. To estimate transport costs most conservatively, one would assume no decrease in in-country production costs, and instead use current illegal prices. In its 2010 World Drug Report, the United Nations Office on Drugs and Crime put the price of cocaine in Colombia at roughly $2,300 per kilogram and that of heroin in Afghanistan at roughly $2,400 per kilogram. Current wholesale prices in the United States, meanwhile, can range from $10,000 to $43,000 per kilogram for cocaine and from $40,000 to $100,000 per kilogram for heroin.

Even at the low ends of these ranges, the markup that occurs between the site of production and the site of wholesale — in other words, the cost of international transport under the current prohibition regime — is enormous. If cocaine and heroin were legal, distributors would be able to ship them by any of the various means now used to transport other legal goods. FedEx, for instance, charges about $65 to send a one-kilogram package from Colombia to the U.S. and about $200 to send a one-kilogram package from Afghanistan. Those rates are less than 1% of the lowest figures for current transportation costs for cocaine and heroin.

As substantial as the price decline would be at wholesale, retail price declines would be even greater. To understand why, one must recall that the free-market retail price for a good equals the good's wholesale price plus a markup to account for distribution and retail costs (including taxes). In the case of drugs, there is every reason to believe that this markup would be small.

Some have argued that if drugs were legal, the distribution markup would, as a percentage, parallel the markups for other legal agricultural goods. But distribution costs depend mainly on a good's weight and hence its value-to-weight ratio (how much the good is worth per unit of weight). No major legal agricultural product comes close to having a value-to-weight ratio of more than $2,000 per kilogram; cocaine and heroin, then, are not likely to have distribution markups comparable to those of legal agricultural products. Moreover, the large markups for legal crops emerge from processing: For instance, the wholesale cost of wheat accounts for a small proportion of the price of bread. But the prices quoted above for drugs in Colombia and Afghanistan are for finished products that are ready to be consumed without further processing. So the percentage markups for legal agricultural products are not appropriate parallels.

To anticipate what might happen to retail prices of drugs following legalization, a more useful comparison would be with a product that has a similar value-to-weight ratio, and one that requires very little processing. Here, the best analogue may be silver, which has a value-to-weight ratio within a factor of two of those of wholesale cocaine and heroin. Typically, the difference between the commodity-exchange price for an ounce of silver and the retail price (for, say, a one-ounce silver round at a coin shop) is a few dollars. If one takes into account that wholesale-to-retail markup, makes adjustments for the fact that drugs (unlike silver) are diluted between the wholesale and retail purchase points, and considers the effects of sales taxes, one might expect free-market retail prices for cocaine and heroin (at current retail purities) of roughly $2.50 and $2.00 per gram, respectively. According to the U.N. , illegal cocaine currently retails for between $10 and $350 per gram and heroin between $55 and $150 per gram; relative to the midpoints of those ranges, the price declines one might expect to see as a result of legalization are in the range of 95%. Calculations for methamphetamine are broadly similar.

As for marijuana, the 2010 RAND Corporation study Altered State? — which examined California's Proposition 19, an initiative to legalize recreational marijuana use and cultivation — included the authors' calculations of marijuana's likely legal price. Even with ongoing federal prohibition of outright farming, the study estimated that grow-house based production, distribution, and retailing that avoided federal law enforcement's traditional targeting levels could profitably produce at a price of $38 per ounce of sinsemilla (high-potency) marijuana — a decline of more than 80% from today's prices, which range from $300 to $450 per ounce.

Moreover, were nationwide legalization to allow for the large-scale farming (instead of mere gardening) of marijuana, yields would be on the order of 500 to 2,000 pounds per acre, suggesting production costs ranging from a few dollars to a few tens of dollars per pound (not per ounce). And it would take fewer than 10,000 acres — or 0.0025% of the 400 million acres of cropland in the United States — to supply the entire current U.S. market for marijuana. It is difficult to imagine such mass cultivation exerting anything but dramatic downward pressure on marijuana prices. And it is impossible to imagine that such steep price declines for marijuana — as well as for cocaine, heroin, and methamphetamines — would not significantly increase the rates of drug use.

SUPPLY AND DEMAND

When the price of a good or service declines, consumers usually purchase more of it; the question is how much more. Economists quantify how consumption of a good responds to price changes by measuring its "price elasticity of demand," which is the percentage change in consumption induced by a 1% increase in price. When consumption responds to price fluctuations more than proportionately, the demand is called "elastic"; when consumption responds, but less than proportionately, demand is said to be "inelastic." If consumers continue to purchase the same amount of a good as they did before the price changed, then the good's demand is "perfectly inelastic."

In the past, many believed that demand for addictive drugs was perfectly inelastic: The presumption was that an addict would do whatever it took to get a "fix," regardless of the drug's price. People who had actually observed drug users' behavior, however, always knew better. For example, even people who are dependent on heroin — the substance John Kaplan famously called "the hardest drug" in his 1983 book by that title — have spells of abstinence mixed with "runs" of daily use. And a wide range of indicators of drug use — including treatment admissions, emergency-department mentions, ambulance call-outs, and urine tests of arrestees — are negatively correlated over time with price movements, meaning they decline when prices go up.

Observing historical correlation is of course insufficient to establish a causal relationship. But in a pioneering econometric study conducted in the late 1970s, Lester Silverman and Nancy Spruill estimated that a 50% increase in the price of heroin would reduce heroin consumption by 13%. And over the past 35 years, a large and increasingly international body of literature has replicated Silverman and Spruill's general finding.

Other studies have expanded upon their research. Assessing more than just total elasticity, scholars have examined the "participation elasticity" of drugs, which measures the change in the number of consumers (in the case of drugs, users). They have also looked at the "conditional elasticity" of drugs, which measures the change in the number of incidents of usage. For example, in 1999, Frank Chaloupka and his co-authors estimated a past-year participation elasticity for cocaine among youth of -0.89, a conditional elasticity of -0.40, and an overall elasticity of -1.28. In other words, a 10% decrease in the price of cocaine would lead to an 8.9% increase in the number of young people using, a 4% increase in the occasions of usage, and a 12.8% increase in the overall quantity consumed.

Economists apply various measurement methodologies to assess how different populations of drug users respond to fluctuations in price. For instance, to measure price responsiveness among typical users who are, by and large, not dependent on their drugs, researchers draw from surveys of young people and the general population. To measure heavier or more compulsive use, researchers look at the patterns of overdoses, treatment admissions, and proportions of arrestees testing positive for drugs. But even within a single measurement methodology, price responsiveness can vary across groups. For example, economist Jenny Williams has looked at marijuana price responsiveness in Australia, and has documented differences across age and gender groups. This means there is no single, magic elasticity number that can determine exactly how drug use would be affected by a dramatic decline in prices.

The best that scholars can offer is a series of approximations and estimates. After taking account of the various values that the research literature has put forth, RAND's Altered State? report considered elasticity values for marijuana ranging from -0.4 to -1.26, deciding on -0.54 as the best single estimate of marijuana's total price elasticity. The best guess for cocaine's elasticity is a bit greater, perhaps -0.75, with an uncertainty band at least as broad.

While it may seem counterintuitive that demand for cocaine would be more elastic than demand for marijuana, it is important to realize that, even though cocaine is consumed primarily by compulsive users, those users are people who devote a large share of their disposable income to purchasing the drug, and are therefore more sensitive to fluctuations in its price. Marijuana, by contrast, is relatively cheaper, and a smaller proportion of its users are people who live under severe financial constraints. (For similar reasons, it is not always the case that dependent drug users are much less responsive to prices than recreational users are.)

There is clearly considerable uncertainty about how much consumption of and dependence on drugs would increase in response to a modest decline in price. That uncertainty is multiplied many times, however, when contemplating the much larger price declines that would result from legalization. The reason for this great uncertainty is simple: All of the empirical evidence underpinning the elasticity studies comes from the relatively narrow range of prices that we have observed historically — that is, under a regime in which these drugs have always been illegal. Legalizing them would reduce prices far below any levels that have been observed in any industrialized country in modern times.

Projecting legalization's effects on use and dependence thus requires extrapolating far beyond any available empirical evidence, and no economic analysis or modeling — regardless of its sophistication — can erase that fundamental limitation. As Altered State? shows, were drug-price declines to reach or exceed 80%, alternative models of different demand curves — all equally plausible and equally consistent with historical data — give radically different projections of the consequences for marijuana use and dependency. For example, under one particular scenario (a $50-per-ounce tax, evaded 25% of the time), a 50% confidence interval for the consumption increase with a linear demand model runs from 75% to 98%, whereas the comparable range with a bowed ("constant elasticity") demand curve is a 167-289% increase. Best estimates put the number of Americans who currently meet clinical definitions for marijuana abuse or dependence at about 4.5 million, so the difference between an 80% increase in consumption and a 280% increase is enormous.

For obvious reasons, spikes in drug dependence would have enormous societal and economic consequences. Those who would "experiment" with legalization must acknowledge that they have no way of knowing to what degree they would unleash these consequences. Anyone who claims to know with certainty that experimenting with legalization would have only limited effects on use simply cannot be telling the truth — an important point to keep in mind when weighing the arguments for and against legalization.

THE TAX DELUSION

Some legalization advocates suggest that the likely price collapses could be counteracted, and the subsequent increases in drug use avoided, by the imposition of excise taxes. (The prospect of revenues from such taxes is also a deal-sweetener used to attract political support for legalization.) What these advocates neglect, however, is the fact that taxes capable of achieving these aims would be exceedingly difficult to collect.

The reasons why are straightforward. The first is a matter of amount: As we have seen, the price collapse that would be induced by legalization is enormous. Any tax that aimed to offset it, even in part, would therefore have to be staggeringly high.

Consider the case of marijuana, which is the drug most conducive to taxation. This is partly because marijuana is the bulkiest of the major illegal drugs: A bulky product is more difficult to produce, ship, and sell without being detected, thus making tax evasion more difficult. Moreover, the legalization-induced price decline for marijuana would be the least dramatic among the major drugs, meaning that it would require the lowest offsetting taxes. Nevertheless, despite these favorable circumstances, even the numbers for marijuana are discouraging.

In California, for instance, Democratic assemblyman Tom Ammiano proposed legalizing marijuana and imposing an excise tax of $50 per ounce. That is a steep levy, to be sure. And yet it would offset only about one-fifth of the price decline that Altered State? suggests would come from marijuana legalization in California (and less than one-fifth of the decline that would be caused by legalization nationwide). Cocaine, heroin, and methamphetamine present cases that are even more challenging: These drugs are much more compact, and their prices would fall further in percentage terms as a result of legalization.

To understand how difficult it would be to collect excise taxes on these drugs, it is useful to compare them to tobacco, the excise tax on which is the highest per unit weight among consumer goods in the United States. Were tobacco to be taxed at a per-weight rate comparable to Ammiano's proposal for marijuana, it would amount to a $35 levy on each standard (20-gram) pack of cigarettes — more than five times the highest state tobacco tax in the United States today. For cocaine, heroin, and methamphetamine, offsetting the legalization-induced price declines could require taxes 500 to 1,000 times current tobacco taxes in terms of value per unit weight.

These tax-per-unit-weight (or volume) figures are crucial, because they govern the incentive to sell untaxed products in a tax-evading "gray market." To see why, consider the amount of profit a person could earn by evading taxes on a good that fits in the trunk of a car — a trunk capable of holding, say, 250 pounds of material without making the car sag visibly (thereby drawing the attention of law enforcement). This would correspond to about 40 gallons of gasoline, which, in the U.S. , carries an average excise tax of about $0.40 per gallon. The potential windfall from using the car trunk to smuggle gray-market gasoline is thus only about $15. By contrast, the windfall from evading a $3-per-pack excise tax on the same weight of cigarettes would be more than $17,000. The corresponding windfall for evading Ammiano's $50-per-ounce marijuana tax would be $200,000. The figures for cocaine, heroin, and methamphetamine would be truly astronomical — as would the incentives to reap such profits by smuggling gray-market drugs.

The case of tobacco also illustrates the effects of those incentives. There is abundant empirical evidence that tax-evading contraband cigarettes capture a substantial portion of the market in many countries. In 2008, the Center for Public Integrity released a report titled Made to be Smuggled, which examined an underground Russian cigarette brand called "Jin Ling." As the report says, "the brand is never advertised and cannot be bought in shops. It is only sold illegally — smuggled by gangs who hope to pocket immense profits by selling unlicensed, untaxed cigarettes on black markets across Europe." Jin Ling became so successful that it rivaled Marlboro as the most smuggled cigarette brand in Europe.

Canada, too, has been engaged in a well-documented struggle with the gray market for untaxed cigarettes. Canada was forced to repeal excise-tax hikes enacted in the early 1990s after consumption of contraband cigarettes increased quickly, in some areas accounting for the majority of the market. Lawmakers tried again in the early 2000s, this time raising tobacco excise taxes more modestly. According to Nachum Gabler and Diane Katz of the Fraser Institute (a Canadian think tank), contraband cigarettes satisfied 27% of Canadian demand after the measure passed.

The U.S. is hardly immune to gray-market tax evasion. In their report Cigarette Taxes and Smuggling, Michael LaFaive, Patrick Fleenor, and Todd Nesbit collected data for smuggling rates and excise taxes in 47 states. They found a positive relationship between a state's excise tax and the market share of smuggled cigarettes: The higher the excise tax, the more the market is taken over by contraband tobacco. If we use this relationship to project the effects of Ammiano's $50-per-ounce excise tax on Californian marijuana, while conservatively accepting the low rates of current tax evasion claimed in public-health studies, the expected tax-evasion rate on marijuana would still be 50%. If we use the same considerations but assume tax-evasion rates closer to those reported in public-finance studies, the numbers suggest that contraband marijuana would meet essentially all marijuana demand in the United States.

To be sure, using the tax-evasion data from tobacco to try to determine the implications for marijuana has limitations. Even so, the numbers are sobering — and indicate fairly clearly that marijuana-tax evasion, or evasion of taxes on any legalized drug, could be a serious problem. This suggests that collecting taxes capable of offsetting price declines, even partially, would be very difficult.

Some have argued that tax evasion could be deterred with sanctions, but this claim is very likely naïve. With a $50-per-ounce tax on marijuana, tax evaders could reap profits of as much as $800 per pound. That is roughly double the reward traffickers now earn for transporting fully illegal marijuana across the Mexico-U.S. border. Inasmuch as Mexico now supplies the majority of marijuana consumed in the United States, the implication is clear: Deterring evasion of a $50-per-ounce tax would require imposing sanction risks greater than those now applied to international drug trafficking. The public is unlikely to support such tough sentencing of people who are merely evading taxes on a fully legal commodity; moreover, replacing drug arrests with tax-evasion arrests would defeat a primary purpose of legalization.

A policy "solution" that eliminates one class of criminals only to replace them with another — wreaking economic and social havoc along the way — should, naturally, be a non-starter. And yet legalization advocates continue pushing their agenda anyway, falling back on the claim that, if "experiments" with legalization do produce failure, we can always go back to the way things were without inflicting any long-term harm. Unfortunately, this view, too, is naïve — and could prove disastrously so.

NO GOING BACK

The concept of reversibility (as employed in the drug-legalization debate) comes from the physical sciences, and has to do with whether an object can be returned to its original condition after being changed by some process. The difference between ice and bread is a good example: Applying heat to ice will melt it into water; removing the same amount of heat energy by putting the water in the freezer allows it to refreeze into the original state. By contrast, when one applies heat to bread, the bread becomes toast. Removing that heat energy by putting the toast into the freezer creates cold toast, not fresh bread. Melting ice is a reversible process; toasting bread is not. The related social-science concept is called "path dependence," meaning that outcomes depend not only on current inputs, but also on the past history of the system.

Legalizing drugs is like toasting bread: Not all of the resulting changes can be undone by re-imposing prohibition. Pre-legalization conditions will be gone forever. This stems from a few of legalization's consequences in particular. For instance, the plasticity of the human brain — particularly during adolescence and young adulthood — means that the increased drug consumption caused by legalization would result in more users' becoming dependent. Social norms, too, would change, making drug use a more accepted part of daily life (what sociologists call "banalization"). Lobbyists for a drug industry that manufactured cocaine, heroin, marijuana, and so on would likely become an important influence over our politics. Drug markets themselves would reach a high-use equilibrium.

None of these outcomes would be fully reversible, for reasons that are subtle and bear elaboration. The first major outcome to examine is increased drug use and dependency. The argument that both would be reversible is rooted in classic economic models, which often assume that people's preferences remain stable over time. Even for conventional goods, that notion is oversimplified: Ballet and opera, for instance, can be acquired tastes, meaning that past consumption increases current demand. Addictive drugs, however, take history-dependent preferences to an extreme.

Among some users, past drug use leads to very long-lasting increases in demand, even to craving. The hallmark of dependence is in fact the alteration of the brain that persists long after the chemical has been expunged from the body. This is why "detox" is not treatment: Even expensive 28-day residential detox programs have extremely high relapse rates if they are not followed by interventions that deal with the chemical dependence. It is why people who quit smoking years ago, for instance, may still have intense cravings when they smell cigarettes. There is also the effect of conditioning: After using a substance often enough in one physical environment, the user can involuntarily associate that environment with the reward the drug gives. These psychological cues can become ingrained in the brain's neural pathways, compounding the strength of dependent behavior. Simply stopping drug use is therefore not enough: Once a chemical has been consumed, purging it from the body does not restore a dependent person to a pre-consumption state.

The second major outcome to consider is legalization's irreversible effects on the drug market, particularly the market-clearing equilibria at which supply and demand fall into balance. Freshman economics teaches that the supply and demand curves within a market can intersect at only one point; in the case of drugs, this would suggest a great deal of reversibility, as re-criminalizing drugs would drive prices back up and, at the same time, push down demand. But the real world does not always fit the models of freshman economics, and in fact there are economic systems capable of containing multiple stable equilibria. This means that once the market has balanced at a point of high drug use, it may not necessarily move back down to the lower equilibrium of its own accord.

The best illustration of this concept is Nobel laureate Thomas Schelling's classic example about the prevalence of corruption in a society of self-interested, amoral individuals. In such a society, if initially no one is corrupt, a newcomer will be honest; being the only corrupt person invites swift sanction. But if an identical person, with the same moral character, joins a society in which everyone is corrupt, that person too may choose to be corrupt. Not doing so might make it difficult to get things done. Moreover, when everyone is corrupt, the risk of sanction for any particular person is low.

Now imagine a society in which, initially, no one is corrupt, and in which the punishments for corruption have been eliminated — such that being corrupt brings benefits but no costs. Then, Schelling observes, the people in that society will become corrupt. If later on, after everyone has become corrupt, the original sanctions are re-imposed, society will not necessarily return to being clean. With so many people breaking the law, the likelihood of any one violator's being punished will be small; the occasional sanction may be perceived as simply a necessary cost of doing business. Schelling shows that a society can function at a level of low corruption and also at a level of high corruption, but that a transition from one to the other may be very difficult.

Markets for illegal drugs are like Schelling's corruption market. They can persist in either a low-use or a high-use equilibrium, and once the transition has been made, there may be no going back.

What is the difference between these equilibria? A drug market settled at a low-use equilibrium has few users and dealers. Additional users do not enter the market because supply is either not readily available or available only at high prices, and supply does not increase because there aren't enough users. It is relatively easy, through only modest levels of enforcement, to prevent a market in low-use equilibrium from tipping to a high-use equilibrium.

High-use-equilibrium markets are another matter. Greater supply makes it easy for potential users to join the market, and more users support a larger number of suppliers. A large number of suppliers also helps the illegal distribution chain operate more efficiently by diluting or "swamping" enforcement efforts, allowing any sellers who are incarcerated to be rapidly replaced. Ratcheting up enforcement may thus have only modest effects on reducing supply or demand. And once a drug market tips to a high-use equilibrium, it can be nearly impossible to shrink back to its low-use state.

One might suppose that tipping point to be midway between the low-use and high-use equilibria. But most models place the point of no return much closer to the low-use end. This means that even small increases in use can carry the market beyond its tipping point — that is, beyond the point at which re-imposing prohibition would be capable only of slowing, not reversing, the progression toward a high-use equilibrium. Once that point is passed, the market remains perpetually in the high-use state, and additional enforcement becomes ineffective.

This reality threatens the feasibility of "experimental" legalization. Once drugs are legalized — even if the prohibition is meant to be lifted only temporarily — prices will drop so greatly as to produce usage spikes beyond our capacity to predict. And if use increases enough to tip a drug market to its high-use equilibrium, there will be no going back. To legalize temporarily, then, is to invite unforeseeable consequences that may be permanent.

MARIJUANA'S EMPTY PROMISE

Given these enormous risks, policymakers must consider what rewards, if any, legalization might provide in return. Advocates claim that legalizing drugs would bring a host of benefits — erasing the problems of prison overcrowding, drug-related crime and violence, corruption, overdose deaths, and the spread of blood-borne diseases (principally HIV / AIDS and Hepatitis C). But there is an irony in this agenda that should not be overlooked: The only major illegal drug with any realistic prospect of being legalized in the United States is marijuana — and legalizing marijuana would not materially reduce any of these problems.

This is an important point to consider, especially as momentum builds for marijuana legalization. In an October 2011 Gallup poll, 50% of Americans expressed support for legalizing marijuana. California came close to legalizing marijuana in 2010 when Proposition 19 gained 46% support, and voters in Colorado, Washington, and perhaps Oregon will vote this fall on whether to make production and distribution of marijuana for non-medical use legal in their states.

To be sure, legalizing marijuana would sharply reduce the number of drug arrests (although other categories of arrest might rise, to the extent that police sometimes use marijuana arrests as a pretext for looking into other suspected criminal activity). Marijuana in fact accounts for roughly half of all drug arrests, and about two-thirds of them involve adults. (None of the current legalization proposals would permit use by minors.) Marijuana arrests of adults would largely disappear in a state that legalized, more than offsetting any increase in enforcement by the federal government (which is likely to retain its prohibitions on cannabis) and in the state's enforcement of regulatory statutes on production and sale included in the legalization proposal. Ultimately, legalizing marijuana would put about 10 million Americans "on the right side of the law." (The other two-thirds of the 29 million in total who use marijuana each year are minors, use other drugs, drive under the influence, or break other laws.)

In considering marijuana legalization, though, it is just as important to note what would not change. Marijuana-related arrests have little effect on the size of the prison population: Prosecutors know the judicial and prison systems are full of criminals more deserving of their attention, so prison terms are typically reserved for those trafficking large amounts or for those also convicted of other offenses. Only about one in eight of the 350,000 people in state and federal prison for drug-law violations are there for marijuana violations, and half of those have concurrent convictions for other drugs or weapons violations. Legalizing marijuana could make a big difference to individuals in the other half, but freeing those 20,000 would hardly empty the nation's prisons, which house about 1.6 million inmates in total.

Drug use and distribution do drive a great deal of non-drug crime, but the culprits are the "expensive majors" — heroin, cocaine and crack, and meth — not marijuana. One rarely hears of drive-by shootings between rival marijuana gangs in the United States. Fifty years ago, scare-mongering propaganda claimed that marijuana intoxication made people violent, but such claims have long since been discredited. Likewise, marijuana is cheap enough that even heavy users have much less incentive to turn to property crime than do, say, dependent heroin users.

Similarly, the Office of National Drug Control Policy reported in 2002 — the most recent official data available — that marijuana constituted only one-sixth of illicit drug-use spending and only 16% of treatment admissions. Marijuana consumption has grown by about 40% since then (mostly as a result of more high-frequency users rather than more users overall), so those proportions have risen. Yet they are still less than one-third of the total for all illegal drugs. And marijuana is almost irrelevant to two of the most serious drivers of drug-related social costs: Since it is not injected, marijuana is not associated with HIV / AIDS transmission, and there is essentially no risk of fatal overdose.

Eliminating these problems would require legalizing the "expensive majors" — a proposal very few people are willing to entertain. Meanwhile, legalizing only marijuana would have only modest effects on the downsides to prohibition discussed above.

THE LEGALIZATION GAMBLE

Both sophisticated social science and old-fashioned common sense warn that even well-intentioned policies can yield unanticipated, and often harmful, consequences. Drug prohibition is no different. It has succeeded in discouraging widespread drug use: None of the illegal drugs — not even marijuana — is used nearly as widely as their two legal counterparts, alcohol and nicotine; illegal heroin, meanwhile, causes far fewer overdoses than do legal prescription opioids. But prohibition has certainly brought its share of undesired side effects: violence and organized crime, corruption, crowded prisons, and heightened racial tensions. These are the effects that have prompted the Global Commission on Drug Policy to so forcefully announce that "the global war on drugs has failed."

But there is every reason to believe that legalization would invite unintended consequences just as harmful, if not more so. Indeed, recognition of this may be what prompts the Global Commission and other advocates to promote what they see as middle paths.

The first such path, decriminalization — meaning eliminating criminal penalties for users but not suppliers — combines aspects of the worst of both worlds. While holding obvious appeal to current users, the crime, violence, and high-level corruption that exist under today's prohibition regime could continue, and potentially be exacerbated by somewhat increased use and dependence. (Decriminalization's effects on use would be relatively modest precisely because it keeps production illegal, and so avoids the price collapse that would accompany full legalization.)

The second — legalizing only marijuana — limits the downside risk, because dependence on marijuana is less debilitating than is dependence on some of the other illegal drugs. But legalizing marijuana would not alter the basic character of the drug problem; had marijuana been legalized a decade ago, all of the same ills that now prompt interest in legalization would still be with us.

The third middle path is encouraging "experiments" with legalization. The phrasing is clever; experiments are scientific, objective, and avoid pre-judging. Who could oppose them? Yet the evidence suggests that "experimenting" with legalization is no middle path at all; as we have seen, to open the door through experiments is to invite all of the problems of full legalization.

Experimenting with legalization is like playing drug-policy roulette. If you have been consistently putting money on black at the roulette table with only mixed success, would it make sense for you to place all of your remaining money on red the next turn and expect a certain win? Of course not. Expecting legalization to rectify prohibition's unintended consequences without creating any of its own is similarly unwise. Worse, while a roulette player is free to alternate from black to red and red to black, betting on legalization could be an irreversible mistake.

Jonathan P. Caulkins is the Stever Professor of Operations Research at Carnegie Mellon University's Heinz College and a consultant with RAND's Drug Policy Research Center. Michael A. C. Lee has served as a drug-policy researcher at Carnegie Mellon University's Heinz College.