How Much Do Labor Unions Affect Wages?

Below I have some excerpts from Deirdre McCloskey’s book Bourgeois Dignity, just in time for Labor Day. I find it obnoxious the way that some commentators attribute far too much explanatory power to labor unions. There’s a bumper sticker that says, “Enjoy weekends? Thank your local union.” Sorry, but this is just implausible. The workers got richer because they became more productive. Their greater earnings led them to purchase, among other things, more leisure time. That is why wages have been rising and the average work-week has been shortening for the past two centuries. It’s plausible that labor unions, in some select cases at certain times, made these advances happen faster than they would have otherwise, but it’s completely unrealistic to credit the labor movement with all the gains since the mid-19^th century as some glib writers do.

The struggle by unions for higher wages and better working conditions has inspired many good songs and many good people. Clearly, say the songs, we gotta' go down and join the union; we need to fight for Harry Bridges and build the CIO; and we need to be a union maid when she fought for higher pay. "She'd show her card / To the National Guard, / And this is what she'd say:/ `Oh, you cant scare me: / I'm stickin to the union."' But the fight to get a higher share of the pie for union members leaves out nonunion members, who were always a high percentage of the workforce in the United States. If the size of the pie is pretty much fixed-the underlying assumption of classical economics and of union logic-then higher pay for auto workers implies lower pay for auto mechanics, not something to be celebrated in songs by Woody Guthrie and history books by progressive historians. The economist H. Gregg Lewis painstakingly estimated the effect of unions 1967-1979 on wages of their members relative to comparable nonmembers and found it to be, as an upper bound, rather small: some 14 percent. My uncle, who was himself a union electrician in Michigan, paid union wages for his employees gladly, he said, because he passed the expense on to the companies and hospitals and school buildings that he wired. Oh, good. But who pays that? Other workers. There's no one else to pay it. As Pogo famously said, "We has met the enemy, and he is us."

So, essentially unions succeeded in raising the pay of their own member by at most 14%, and that came at the expense of other workers. Meanwhile, the average wage in the developed world rose by something more like 1600% to 10,000% (a factor of 16 to 100, depending on how you account for changes in quality and other considerations).

And if, as was on the contrary the case, the pie was in fact exploding, then a shift in the size of the slice going to workers as a whole (setting aside that it would in fact go only to the minority of union members) would anyway play a small role in the betterment of even those workers. Workers in the United States and elsewhere grew radically better off from 1800 on, and 1900 on, and even from 1970 on because of the Bourgeois Deal, not because they went down and joined the union. How do I know? Answer: the nonunion people shared nearly equally in the gain, even though they were paying as consumers for the fancier wages for union electricians in Michigan. The maximum of 14 percent of "concessions" extracted by bargaining or strike doesn't come close to accounting for the great magnitude of rise in the real wage. There's not enough profit-usually 10 or 15 percent of national income-to raise the level of the rest of national income by a factor of even two, that is, 100 percent, much less the factor of eighteen it in fact rose after 18oo. Expropriating 15 percent of national income claimed by the wretched profiteers teers and transferring it to the 85 percent earned by us workers raises our income only 18 percent. That's a long, long way from 1,700 percent.

Deirdre McCloskey is making the point that labor unions can only rearrange portions of a relatively fixed pie; nothing they do is likely to grow the pie. They can negotiate with the capitalists to redistribute some of the profits from capitalists to laborers, but they can’t increase the total productivity available for distribution. People love to romanticize the 19th century labor struggle. So much blood was shed, it's irresistible to conclude that some tremendous benefit must have come from it. Alas, a little bit of numeracy and some back-of-the-envelope math implies otherwise. 

Casting More Doubt on the "Prescription Painkiller Epidemic"

When a doctor “certifies” a cause of death, his certification is based upon his evaluation of the evidence available to him, but it is still just his opinion and does not set a precedent for similar cases.

It turns out assigning a cause of death is really hard. 

I’m reading through Karch’s Pathology of Drug Abuse. This is the standard textbook on the topic. It repeatedly makes the point that toxicology results can’t be used to determine the cause of death. I’ve only read a fraction of it so far, but it’s definitely strengthened my belief that the recent rise in overdose deaths is, at least in part, a spurious trend. The book’s authors go into some detail on the mechanisms that make postmortem toxicology unreliable. This is important because there is much ink spilt about the recent supposed rise in prescription painkiller deaths, and much more recently the rise in heroin overdose deaths. It’s true that there are many more prescriptions today than there were 15 years ago; the sheer tonnage of opioids prescribed is easy enough to track, and the government does track these figures. It is far less clear that the rise in opiod-related deaths is real. It could simply be the case that, with more people walking around with opioids in their system, more deaths are being labeled “drug poisonings” even though the drugs are incidental and the real cause of death was actually something else. When a medical examiner labels a death a “drug poisoning,” they are expressing an opinion.

When you see numbers such as “There were 47,000 drug overdose deaths last year” or “There were 18,000 deaths from prescription painkillers last year”, these figures are basically a guess. For each such death, somebody (a coroner or medical examiner) looked over the body and determined it was a drug overdose, and that death went into a big database of all 2.6 million deaths that happened that year. Then later somebody ran a query against that database that counted up all the deaths that matched criteria X, Y, and Z. I sincerely wish that the people and organizations reporting these figures would be a bit more careful about issuing caveats and admitting to uncertainties in their data and methods. I hear a news story once a week or so about the prescription painkiller epidemic, and I'm getting increasingly annoyed that the news outlets aren't doing the slightest bit of vetting on the underlying numbers. Did not a single one of these reporters bother to crack a textbook on the topic? Did nobody bother to ask "How do you know if a death is a drug overdose?" The database and query are actually pretty straightforward, but the whole thing falls apart if you don’t have much confidence that the data reflect the true causes of death. It’s very likely that many of these deaths were due to some other cause, and the drug use was an incidental finding.

From the section on methadone:

Because methadone has a very high volume of distribution, concentrations can be expected to rise after death. Indeed, fourfold increases have been reported (Levine et al., 1995; Milroy and Forrest, 2000). Postmortem methadone concentrations are also site dependent and, for unexplained reasons, the increases appear to be greater in men than women (Caplehorn and Drummer, 2002). In every published series methadone blood concentrations in fatal cases completely overlap those found in methadone maintenance program participants.

So basically the blood concentrations are useless in terms of determining the cause of death. For one thing, drug users develop a tolerance, so what would be a fatal dose to a naïve user is safe to an experienced user. This is particularly true of opioids, although all drugs build some degree of tolerance in their users. The problem of “postmortem redistribution”, diffusion between blood and tissues, exacerbates this problem.

Not one single control study, even in animals, has ever shown that postmortem drug concentrations accurately reflect drug concentrations at the time of death, but a goodly number have shown quite the opposite to be true, chiefly because of the problem of postmortem redistribution (Pounder et al., 1996; Hilberg et al., 1999; Moriya and Hashimoto, 1999; Drummer and Gerostamoulos, 2002; Flanagan et al., 2003; Ferner, 2008). Postmortem redistribution is defined as the movement of a drug down a concentration gradient after death.

In other words, blood concentrations don’t follow a smooth decay curve after death.

The postmortem movement of free morphine out of the tissue, back into the plasma, makes it impossible to estimate the time of ingestion by calculating the ratio of free to conjugated morphine.

[T]he Vss [steady-state blood concentration] for methamphetamine, in healthy volunteers given fixed doses of drug and then living in a locked ward, was found to range from 2 to 11 L/kg (Schepers et al., 2003). This enormous and unpredictable degree of variability, even in the living, makes it impossible to establish any sort of relationship between postmortem drug concentrations and reported dosage, though it would seem reasonable to conclude that if low levels of drug are detected, the drug was consumed.

In other words, people given the same dose may show wildly varying blood levels of the drug, even in healthy living test subjects. This complicates setting any threshold for a toxicology screening, as in “Anything above X is a fatal dose, anything below X is nonfatal.” All the stuff about redistribution after death complicates matters even further.

This next part borders on a philosophy of causation lecture:

Could the death of a cocaine and heroin abuser, with cardiomyopathy, be related to the fact that his cardiologist had just begun treatment with carvedilol, or did the doctor at the methadone maintenance clinic administer the wrong dose of methadone? It may be impossible to say, but these complex issues exist and cannot be ignored. 

The authors even discuss how gut bacteria cause drugs and their metabolites to diffuse through the body after death (if I’m reading them correctly):

It is not uncommon to hear an “expert” offer an opinion on how much time had elapsed between the last dose of heroin/morphine and death. Such speculation is unscientific, not just because of the differences in Vss between parent drug and metabolite, because bacteria translocate. That is, they migrate through the walls of the bowel, invading the rest of the body within a few hours of death (Kellerman et al., 1976).

Two short sentences about how genetic differences between individuals can contribute to different drug metabolism:

Thus, even in the living, without knowing an individual’s ability to conjugate morphine (which would require DNA resequencing), comparing the ratio of bound morphine to free morphine is a meaningless exercise.

Genetic polymorphism is, no doubt, responsible for more than a few cases of drug toxicity, but the true incidence of this problem is not really known.

In other words, people metabolize drugs differently. A fatal dose to one person may be nonfatal to another, even before considering things like acquired tolerance. The authors go on to describe a specific genetic trait and its effect on codeine metabolism:

Individuals entirely lacking CYP2D6 activity (7%–10% of Caucasians) are called PMs and are not likely to get much pain relief from codeine because they cannot convert enough of it into morphine. But it is also possible to have multiple duplications of CYP2D6, making the individual into an ultrametabolizer who can transform abnormally large amounts of codeine into morphine. Thus it is quite conceivable that a person with a gene duplication might experience an overdose even though only a modest dose of codeine had been taken.

The author gives two examples of deaths that might be labeled an “overdose” and explains why the true cause of death isn’t so clear:

If toxicology results cannot be considered in a vacuum, neither can autopsy findings. When a doctor “certifies” a cause of death, his certification is based upon his evaluation of the evidence available to him, but it is still just his opinion and does not set a precedent for similar cases. If the decedent was a known cocaine user, the body was noted to have frothy pulmonary edema, blood cocaine concentration is 1 mg/L, and a crack pipe was found at his side, the decision is not difficult. But what if the blood cocaine concentration had been 0.050 mg/L, there was minimal pulmonary edema, and no other apparent stigmata of drug abuse or anatomic abnormalities were observed, but concentric cardiac hypertrophy was evident? How would the pathologist determine which abnormality was causal—the one he could see and weigh or the hERG channel–cocaine interaction he knows might exist, but cannot measure?

Another example that illustrates how difficult it is to determine the cause of death:

In the same fashion, suppose the pathologist was confronted with a decedent who had a remote history of heroin abuse and only mild pulmonary edema with a blood morphine concentration of 20 ng/mL evident at autopsy. In the absence of other identifiable anatomic changes, the pathologist would, no doubt, determine the manner of death as heroin overdose and an accidental death. But that would be incorrect for at least two reasons. First, unless hair testing had been performed, the pathologist would have no idea whether the decedent was a regular user and therefore tolerant or whether the decedent was naïve and intolerant. Second, the cocaine that was present could have unmasked one of many genetic defects that can interfere with the metabolism of morphine or exaggerate its effects (Oertel et al., 2009). Whether such a death would be classified as natural or accidental could be a matter for debate. What is not debatable is that the autopsy pathologist has no way to look at the body and determine whether or not the decedent suffered from a P-gp variant or an SCN9A mutation, at least not without performing genetic resequencing first.

There is much, much more. In fact, everything above is from the chapter on opioids. Other chapters are full of similar warnings about assigning a cause of death or interpreting blood concentrations of a drug. I’ve written several posts about how the recent rise in drug poisoning deaths could be spurious, here, here, here, here, and here . I still think that part of the increase in real, but some large fraction of it is spurious. Those news outlets who are fanning the flames of a moral panic need to cool their jets until we really understand what’s going on. 

I'm still reading this book. More to come if I find more interesting topics.

Income Inequality is Not a Problem

The problem with identifying “income inequality” as a social problem is that it misses all the real problems and it implicates non-problems as problems. The desire to alleviate poverty is legitimate, but the blaming of rich people isn’t.

Some people have a hard time in life, for a variety of reasons. Some people have serious emotional issues that make them difficult to employ. Some people have serious absentee/tardiness issues, or just general conscientiousness problems, that make them unreliable as employees. Some people, for whatever reason, failed to acquire any kind of relevant education (college or vocational) and thus have no useful job skills. Some people have a relevant education, but for whatever reason refuse to work at a meaningfully remunerative job (I have seen so much of this, and it is almost always baffling to me). Some people are government workers in cozy jobs whose cost-of-living-adjustment suddenly stops because the government goes broke. All of this is merely to observe and explain, not to blame. Sometimes it really *is* the person’s fault, and sometimes it isn’t. Either way, it’s wrong-headed to tell these people, “You’d be doing a lot better if these rich people weren’t hoarding society’s resources!” Or “If only the political system weren’t dominated by rich interests, it would work better for you, and you’d be sharing more of the wealth.” That’s not what these people need to hear. This kind of scapegoating is shameless pandering when a politician does it and self-congratulating escapism when the “underpaid” workers themselves do it. Either way it’s grossly irresponsible. There are usually actions the person can take to improve their circumstances, and we should encourage that. We shouldn’t be telling them, in effect, “Your problem is due to someone else’s greed, and only a massive change in government policy will fix it.” Without casting any moral judgment, I want to point out the causal mechanism proposed within the income inequality framing is simply mistaken. It just isn’t true that the rich are richer *because* the poor are poorer.

If you aren’t careful, it’s easy to take on a zero-sum worldview (inadvertently or explicitly). It’s easy to think “If he had less, we’d all have more.” Or “Society’s income is $16 trillion per year. We just need to dole it out more equitably.” We don’t live in a zero-sum world. “Society’s income” isn’t something that belongs to all of us; it’s not something that is wrongly and arbitrarily doled out unequally to make a few random winners rich. There really are people whose annual productivity is in the millions of dollars. If large companies didn’t shell out to place the most talented possible individual in their executive roles, they could lose out on millions or even billions of dollars. And without an appropriate incentive to perform, that talented executive might make avoidable million dollar errors. It’s easily possible that the *very* best person for the job of, say, running Microsoft will contribute $10 million more a year to the expected bottom line than the second best person. Getting that person for $5 million would be a steal for the shareholders and for the software users who are using the improved product and the employees of Microsoft who are made more productive by the CEO’s actions. Getting that very best person to put in their very best effort could easily be worth another $10 million. If the difference between a motivated and an unmotivated executive is $10 million to the company’s bottom line, then a generous bonus in the millions of dollars is economically justified. Perhaps the very best executive has a 1% chance of totally destroying the value of a large company over their tenure (say Microsoft, with a market cap of $450 billion), but the second best has a 2% chance of totally destroying the company. This consideration alone makes that top executive worth ([2%-1%] * $450 B = )  $4.5 billion. If Microsoft can acquire this person by making them a billionaire once or twice or even thrice over, it’s worth it. Sorry to burst anyone’s populist bubble, but those very large executive compensation packages make economic sense. You’re *richer* because of them, not poorer. I think it’s not hard to understand that an executive’s productivity should scale up with the total value of the company, and that their take-home pay should reflect this productivity. People react with outrage to these big compensation packages, but economically speaking they are justified.

Consider now a more modest form of inequality: those salary differentials between the professions. These are a feature, not a bug. They are a market signal directing people to where work is most dearly needed. Very high salaries for doctors, lawyers and actuaries are signals that “The world needs one more doctor/lawyer/actuary more than it needs one more teacher/secretary/cashier.” If the pay differential were too small, you wouldn’t have enough people who are willing to undergo the long years of training necessary to enter these professions, and you wouldn’t find enough people to do the often thankless, boring, or even dirty tasks that these jobs entail (and while working longer hours than an average worker).  More to the point, people aren’t being assigned these professions at birth; they choose their profession *knowing* that these salary differentials exist. It’s extremely churlish to choose a low-paying profession while complaining that a higher-paying profession earns more. People don’t just fall into these positions. They end up there after having followed a long and arduous path. Your current profession is a sum of many past choices and your past effort. 

Some people make the mistake of interpreting salary differentials in terms of value judgments, as in “Isn’t education just as important as medicine?” It’s the wrong question, and answering it doesn’t tell you anything useful about how much doctors should be paid vs how much educators should be paid. The actual question is more like, “What is the social value of *one additional doctor* vs *one additional teacher*?” And more to the point, “What salary will attract the extra doctor we need, and what salary will attract the extra teacher we need?” It’s not as though there are a fixed number of slots to fill, and we can just fill them all with any arbitrarily chosen salary figures. The actual number of people willing to enter the profession changes, moves up and down, in response to the salaries offered, and in turn those salaries adjust up or down with the number of people willing to enter the profession. (This is the standard intersecting-supply-and-demand-curves stuff from econ 101.) Inserting your own moralistic value judgment into this consideration would be arbitrary and wouldn’t help you answer the question “What *should* these relative salaries be?”

I have no doubt that there are some people who make an undeserved fortune by adjusting the rules to favor themselves. Perhaps an auto producer or textile company profits from an import quota or tariff. Or perhaps a mega-farm gets paid by the government to leave thousands of useful acres of farmland fallow. Or onerous regulations make it hard for small start-ups to challenge incumbents. It’s often the case that the companies benefiting from these market restrictions are the same companies that actively lobbied for those policies. These gains are ill-gotten. But it’s the “ill-gotten” part that’s the problem, *not* the “gains” part. By fixating wrongly on inequality, we would falsely impugn the honest business owners who earn their fortunes honorably through voluntary transactions. And we might wrongly overlook these kinds of ill-gotten gains if the thief fails to enrich himself compared to his competitors. In both cases, “inequality” is a terrible proxy for the real problem. For both kinds of errors, the inequality framing *completely misses* the problem. 

Countervailing Forces

I think it’s very likely that under drug legalization, the mix of drugs people are taking could change, but total drug-related harms wouldn’t increase. It’s far more likely these harms would decrease substantially. There are countervailing forces here, so we have to be clear about what these forces are and somehow estimate their magnitudes. (I’ll give an incomplete list what those forces are below, but this won’t be a thorough, rigorous cost-benefit analysis.)

Consider the forces that would decrease drug-related harms:

1)      Some currently illegal drugs are extremely non-toxic, contrary to popular belief. Their legal status, while not exactly making them “unavailable”, does raise their relative cost. Legalization, at least for these drugs, would shift consumption toward these and away from the current legal drugs (alcohol, tobacco, and pharmaceuticals), which are more toxic and more prone to addiction problems.

2)      Related to 1), but the same point for drugs that are potentially harmful. Someone might satisfy their need to get intoxicated on a low dose of cocaine or heroin, while needing a very much larger (and overall more harmful) dose of alcohol to achieve the same satisfaction. Don’t scoff at the notion that substituting from alcohol to heroin or from nicotine to methamphetamine could have net positive health and social effects. These drugs are not harmful if properly administered, and some of them (particularly opioids) don’t cause cumulative organ damage.

3)      Currently illegal drugs are made *far more* dangerous by prohibition than they would be in a legal market. Drug interdiction causes the dosage to fluctuate wildly, which is the cause of a lot of overdoses. Often the active ingredient is a completely different chemical than the user realizes (fentanyl sold as heroin, or experimental designer drugs sold as MDMA). Impurities and adulterants are hard to control in a black market. Dealers have a preference for very concentrated doses of the drugs, because they are easier to transfer and smuggle, so these higher-concentration versions of drugs dominate the market.

4)      It’s easier to advise people against self-harming behaviors in a legal market. Warning labels against specific harmful behaviors can be put on the packaging and purchasing the substance puts the buyer in front of a pharmacist who can advise them. Someone who buys two drugs that interact can be told “Don’t mix those.” (Considering that most drug poisoningsare multi-substance interactions, this would be a very important behavioral tweak.) Relatedly, it becomes easier to usher someone into drug rehab if their habit doesn’t make them a criminal. More people are likely to seek necessary help if doing so doesn’t risk legal sanctions.

5)      The legal sanctions are themselves harmful to people’s lives, and these would obviously disappear under full legalization. These include decades in prison, low-level harassment of millions of innocent people, and the occasional beating or shooting of a drug suspect. We’d do away with this class of social harms entirely.

6)      Black market violence is a huge driver of our overall crime rates. Estimates vary, but by all accounts are big. Read “Drug War Crimes” by Jeffrey Miron for the full story, but something like ¼ to ¾ of our murders can be blamed on the drug war, and that’s not even the entire class of black-market social costs.

Now consider the force that would lead to an increase in drug-related harms:

1)      Someone who otherwise wouldn’t have tried cocaine (or some other illegal addictive substance) now tries it and develops a self-harming habit. This is the opposite of 2) above. These people require a large, harmful quantity of illegal drugs to become satisfied. In other words, they are willing to undergo the self-harm for cocaine (and heroin, etc.), but not for the effects of alcohol or something else that is relatively inexpensive under our current regime.

For drug legalization to be a bad idea, the one harm-increasing force would have to dominate all the harm-reducing forces. I find this implausible. Don’t interpret this as a “This list is longer than that list, therefore…” argument. It’s possible in principle for the one harm to overwhelm all the benefits, even though it’s one item. I could arbitrarily express it as four items with some clever manipulation, or compress my list of benefits. We have to weight these things somehow, not simply count them. Still, I think drug legalization comes out ahead. The one “harm-increasing” force is almost surely small. There just aren’t that many people who 1) are willing to endure enormous amounts of self-harm and 2) are deterred significantly by weak legal sanctions. (Here is a more rigorous treatment of the problem, stated in terms of demand curves and total costs.)

Prohibition probably causes people to substitute some drugs in place of others, but it doesn’t alter their underlying willingness to undergo self-harm. If you look at evidence from alcohol and tobacco, it appears that the self-harm dominates the dollar price. No doubt this result generalizes to other addictive substances.  The goal of drug policy should be to reduce the social harms of drugs, not to reduce drug use *per se*. This goal can be achieved by allowing people to achieve their desired state of intoxication in the least harmful way possible. If we had a wider array of intoxicants (some of them almost entirely non-toxic) available to us, this would be a lot easier.

The fear is that light dabbling in these other kinds of drugs will inevitably lead to full-fledged addiction, with all the overdoses and other health problems. But this “try it once and you’re hooked” story is a myth. It doesn’t describe the vast majority of illegal drug use. Normal, well-adjusted people have nothing to worry about. It’s the impulsive people with self-control problems that we need to worry about, and these people are mostly already hurting themselves. Drugs don’t dominate the will and turn the former type of person into the latter. There is just an enormous amount to gain here and very little to lose. Legalizing drugs is the lowest of the low-hanging fruit. It's long past due that we pluck it already.

Alcohol and Tobacco: The Purchase Price vs the Full Price

I want to spell out in this post why it might be exceedingly difficult to deter the use of harmful drugs. Spotty but useful information about the market prices of illegal drugs exists, along with estimates of total consumption and social costs. But there are two legal drugs that cause enormous social harms, so I will fixate on them. Alcohol and tobacco are legal, so there is good public information about the amounts sold and the social harms attributable to them. I’ll give away the punchline: the purchase price of these substances is a fraction of the full price, considering that people pay in damaged health and lost productivity in addition to forking over the literal dollar amount on the price tag.

Tobacco

 I’m not doing a thorough cost-benefit analysis, just a quick Googling and some back-of-the-envelope order-of-magnitude estimates. I’m perfectly happy to revise any of the figures I use in this post or redo a calculation with more conservative or favorable assumptions. Everything I have about tobacco comes from this short post by the CDC:

http://www.cdc.gov/tobacco/data_statistics/fact_sheets/economics/econ_facts/

About 264 billion cigarettes are sold each year in the US. At 20 per pack and about $7 per pack, that comes to a total of $92.4 billion in revenues from cigarettes. There are other forms of tobacco use, but as cigarette smoking is by far the most common and the most likely to lead to heavy sustained use and serious health issues, I’ll ignore the others. Tobacco related health costs are about $170 billion and productivity loss amounts to $156 billion. These are mere accounting numbers. It’s hard to say what health costs are faced by the smokers themselves. Perhaps the government pays for half of that $170 billion, in which case you might be tempted to say “So the tobacco users are only saddled with $85 billion in health costs!” However, who pays the actual bills isn’t relevant to this discussion. I’m interested in the question: What are the costs *to the smokers* of their bad habit? Even if that $170 billion is paid in full by the government, those medical bills still represent a lot of suffering and disease faced by the smokers. It’s not as if the expenditure magically cures the smokers’ illnesses.

Similarly for the productivity loss of $156 B. It might seem like this is something that happens to someone else. The employer misses out on the work his sick employee would have done. Or society as a whole misses out on the work not done by sick or dead employees. But that’s a mistake. If you have a bad habit that injures your own productivity, you will see that taken out of your own earnings. Think of people who are too sick to work *at all*, or people whose constant absenteeism catches up to them and they get fired. The CDC is presenting these figures as if they are costs imposed by smokers upon the rest of us, but mostly these costs fall on the smokers themselves. Yes, in the very *near* term your boss suffers if you skip a day of work, but a chronic illness will hit the smoker’s personal earnings by making him less employable.

Anyway, I don’t know how to adjust these cost figures to come up with the cost faced by the smokers themselves, so I’ll use the raw numbers. We have $92.4 billion in cigarette sales compared to $326 billion in smoking-related costs (both annual figures). The non-price costs of smoking trump the price costs, and there is a big enough margin that I don’t think any amount of adjustment will flip that result.

Alcohol

Alcohol also has large social costs, but unlike in the case of tobacco many of these costs fall on third parties. So I can’t just add up the costs and say “These fall on the alcohol users” like I can for tobacco. (Yes, I acknowledge that second hand smoke is a potential hazard to third parties, but according to the CDC article above they are a small proportion of the total costs.) I found an interesting breakdown of the costs in this study, “Economic Costs of Excessive Alcohol Consumption in the United States, 2006.” Unfortunately, this also follows a “strict accounting” approach. Where it says Alcohol Abuse and Dependence costs $10.668 billion per year, it’s literally adding up bills for counseling and rehabilitation. It’s not telling me what I’d *like* to know, which is “the cost of the agony of knowing you’re bad habit has damaged your life.” Notice how the vast majority of the costs are productivity losses. I’ll continue with my argument that the bulk of these costs fall on the drinker.

Once again, many of these costs look at first blush like things that happen to other people, but the $20.9 billion spent on “the criminal justice system” isn’t just a drain on taxpayers. It also represents a person languishing in prison or having his life ruined by a (perhaps deserved) encounter with the justice system. That’s not to say none of that $20.9 billion falls on other people, just that the alcohol abuser feels some portion of that.

Once again, I don’t have a clean way to convert these raw numbers into the non-price cost faced by alcohol users, but let’s just say it’s in the $150 billion per year range, including health, lost productivity, and everything else.

Total revenue for alcohol sales is $211.57 billion according to this article. Now, I’m willing to believe that the vast majority of cigarette sales are to smokers with a serious habit, most of whom are *actually* exposed to all those health and productivity loss costs mentioned above. But that’s not going to be true of alcohol. There are a great many casual drinkers who never engage in reckless behavior, and for whom it would be unreasonable to say, “That beer you’re drinking is contributing $1.70 worth of social costs!” It wouldn’t be fair to blandly state that any and all alcohol consumption is contributing to the total social cost. The costs come from a subset of drinkers who are *problem* drinkers. These are the people who binge drink frequently enough to damage their livers, or whose drunk driving occasionally kills another motorist. I’m happy to use someone else’s estimate, but let’s generously say half of the revenue is coming from heavy problem drinkers, the other from wine and craft-beer sippers. Just as an order-of-magnitude estimate, it looks like the money price of alcohol is trumped by the various health and productivity-related costs. Above I’m getting about $150 billion in non-price costs and about $100 billion in monetary costs (again, specifically for those people who are at risk for alcohol-related health and productivity problems). If I drop $5 on a 40 ounce bottle of malt liquor, I’m spending more like $12 when I factor in those other costs. (On the other hand, if I drop $10 on a 4-pack of craft beers, which I sip at the rate of one-per-sitting, those additional health and productivity costs certainly aren’t relevant to me.)

General Lesson

I don’t think my general point is terribly controversial. My above figures are hand-wavy, but if you think in terms of potential harms and opportunity cost then clearly for many kinds of drug consumption the non-price costs trump the actual price paid at the register (or to the dealer). Some psychedelics, for example, essentially put you out of commission for a full day (though most of these are surprisingly non-toxic, also cheap). And many seriously addicted heroin users spend all day scrounging up enough money to get their fix. (I’m remembering from David Simon’s “The Corner” a recovering junkie giving a speech to a group of other recovering junkies: “We all know that nobody works harder than a junkie!”) It would be silly to talk about the dollar price of these substances when the *actual* price is “a day out of my life” or “a full decade of my life” or perhaps even simply “my life.” 

Given all this, it is ludicrous to think we can change people’s behavior much via drug prohibition. Drug prohibition probably has a significant effect on the market price of the drugs, and adds significant non-monetary search costs in terms of finding a dealer. But the actual money that changes hands is a small proportion of the full price of drug use. People who are willing to pay the enormous health, productivity-loss, and opportunity costs detailed above are unlikely to be deterred by a slightly higher price or a legal slap on the wrist for a possession charge. And the data bears this out. While there have been some economic studies that attempt to determine the responsiveness of drug demand to price changes (these are trying to measure the elasticity of demand for various drugs), the overall pattern of drug use over the decades doesn’t seem to match any simple “price elasticity” story. There are wildly fluctuating patters of use since we’ve collected data on them, where one drug becomes popular then fades away to be replaced by another. Prices for a pure gram of heroin and cocaine have dropped significantly since the 70s, without a corresponding monotonic rise in use rates. (It appears that the demand curve itself is shifting due to cultural reasons and momentary trends, and this shift is more important than momentary movements along the demand curve. I suspect all these econometric studies measuring the price elasticity of drugs are an exercise of mathematical masturbation, given that the dollar price is often a small piece of the full price.) If drug prohibition has any significant effect on drug use, it’s probably because the illegal market makes the drugs so much deadlier, though adulterants and wildly varying dosage. I’m willing to believe that you could deter the purchase of something by attaching a death penalty to it for a few random buyers. Adding harm to drug use raises the cost, so you get less total drug use. However, if you’re careful about accounting for how much drug use is deterred per unit of added harm, you quickly see that this is a losing game (here and here). The total harm is always going up with increasing drug penalties, and it’s because demand is inelastic.

Most people observe these large social costs and they get the lesson completely backwards. They say, “Look at all these enormous social costs! We need to get people to stop doing this!” They should be taking a step back and saying, “Look at the enormous price people are willing to pay to consume these substances. It’s going to be very difficult to deter such an activity.”

The False Positives / False Negatives Trade-off

I’ll try to illustrate the False-Positive False-Negative Trade-off in this post with a basic example. Suppose you have a test that tells you with what probability you have a disease. Such a test is some composite of your medical history, blood and tissue tests, genetic screening, etc. All this information goes into an algorithm, and out comes a single number: “The chance that you have Disease X is 20.6%.” The test is well-calibrated. Of those people who are told they have a 20.6% chance of having the disease, 20.6% have it. (And likewise for 1, 2, 3, …100%.) Given that you have this test, you now have to decide how to act on it. Basically, you have to set an arbitrary threshold: “Everyone with a probability above Y% should be treated as though they have the disease; everyone with a probability below Y% should be treated as though they are healthy.” Even though the test is very well calibrated, any policy of how to act on the test results will invariably worry some healthy people (false positives) and leave some diseased people untreated (false negatives). All you can do is set the value of Y; your task is essentially to manage the number of false positives and false negatives as wisely as you can.

You might respond, “Well, come up with a better test! Come up with something that neatly segregates the population into “100% healthy” and “100% diseased”!” That would be wonderful, but pretend for the sake of argument that we are using all possible information, and the algorithm that gives you your “diseased probability” is optimized to be as well-calibrated as possible. Often in life, we have to make due with limited information in situations when total omniscience would solve our problem. Maybe a better test for this disease will come along some day, but for the time being, we’ll have to make the best we can of this algorithm.

I’ve made up some data. I created a random set of 1001 people whose “diseased probability” ranges from 0 to 1 in increments of 0.001. I then randomly assign each of these people to be “diseased” or “healthy” based on their probability. The disease probability is known (it’s the output of an algorithm we’ve created), but the actual disease status of the person is not. I need to set a rule such as “treat everyone with a disease probability of Y% or greater as diseased, and everyone else is healthy.” So here’s what happens when you vary Y from 0 to 100%:

On the vertical axis we have false negatives. These are the sick people that our selection criterion is missing. On the horizontal axis we have false positives: these are the healthy people we are misidentifying as diseased. There is fundamentally an inverse relationship between the two types of errors. Following the curve down and to the right, you are decreasing the probability threshold (“Y” above) to identify someone as “diseased.” Following the curve up and to the left, you are increasing the threshold.

The problem arises because uncertainty is inherent in all decision-making. How *should* you treat someone who has a 50% chance of having a disease? Or how about a 40 or 30 or 10% chance? What general policy should apply to this very heterogeneous population, with a full range of disease probabilities? Surely it depends on the relative cost of treating a healthy person vs the cost of ignoring a sick one. Some discussions single-mindedly focus on one kind of error and ignore the possibility of the other kind. This is foolish because both kinds of errors are costly. A mature discussion would acknowledge the trade-off and specify where on the curve is the optimal place to be.

Discussions of public policy often miss this point. Suppose you are deciding what the threshold of “reasonable suspicion” is for a police officer to search a car, a home, or a pedestrian’s pockets. Suppose you are deciding which chronic pain sufferers should get strong prescription opioids and which ones shouldn’t, based perhaps on the severity of their condition or the chance that they are “abusers”. Suppose you are setting the standard for when a police officer can use deadly force. Suppose you are deciding which parolees are likely to re-offend and which ones are likely to stay clean, for the sake of sending them back to prison or keeping them on parole. Suppose we have to dole out society’s limited welfare resources to the needy. What’s the threshold for “needy” and based on what criteria? How many guilty men are you willing to leave unpunished to ensure that an innocent man isn’t imprisoned? Whether you have a fuzzy, judgment-based criterion or a hard-coded algorithmic criterion, all these decisions suffer from the false-positive/false-negative trade-off.  With given resources and given information, “make fewer mistakes” isn’t an option. You have to choose which kind of mistakes you are more comfortable with.

Take opioid prescriptions, for example. Maybe you think a single chronic pain sufferer is worth a thousand opioid addictions; in this case you’d choose a permissive standard. The “false positives” (giving opioids to someone who doesn’t need them) are relatively costless compared to the “false negatives” (denying opioids to a real pain sufferer).

For another example that's in recent news, take police shootings. Perhaps you want police officers to be able to defend themselves from an obvious threat, but you also think that police officers assume certain risks by taking the job. There are cases where innocent people are understandably confused about who exactly is bursting into their home in the middle of the night. These can lead to tense stand-offs where an innocent home-owner doesn’t know who’s threatening him and the police don’t know who’s shooting at them. It makes sense that the police should assume *some* portion of that risk, given that sometimes these are completely innocent people (often *not* the target of the raid at all) who the police are charged with protecting. If you force the police to be more restrained in their use of force, it means more cops get shot at but it also means there is more opportunity for the innocent homeowner to realize what’s happening and disarm voluntarily. Less restraint, and you get the opposite. The task here is to set the standard for the use of deadly force, and set it such that you are comfortable with the proportions of both kinds of mistakes. You want to set it high enough that police don’t instantly gun down innocent people at the first sign of a plausible threat, but low enough that people are still willing to endure the risk of becoming police officers. This is a hard problem, and I think many people (probably including my younger self) blow through it without much thought. Of course, “Stop doing no-knock raids” would put an end to most of these volatile encounters, and that’s a perfectly sensible policy solution. Even so, these volatile situations *could* arise during legitimate, routine police work. “Make fewer mistakes” isn’t a policy lever. *Given* the institutions we have, you can only shift the proportion of false positives vs false negatives. Given that both kinds of errors are inevitable, the task, once again, is to figure out which kinds of mistakes you’re more comfortable with (and by how much). All this assumes that we have a policy lever that controls when police fire their weapons during tense encounters. Maybe this sounds implausible, but a liability rule or standardized protocol might do the trick. In the recent discussion of police shootings, much of the commentary proceeds as if all police shootings neatly separate into “plainly 100% justified” and “plainly 100% unnecessary.” Surely no mortal perceives this world with such god-like clarity. Surely borderline cases exist, and the border is probably much broader than a line.

My task here is not to adjudicate any of the policy questions mentioned in the above paragraphs. If it sounds like I’m venturing too strong an opinion, I apologize for the distraction. My much more modest goal with this post is to articulate a trade-off. It’s a trade-off that is crucial to many policy discussions, but often gets completely glossed over. 

Drop Drug Prohibition, Issue Specific Warnings Instead

I’ve looked into the drug poisoning data and the drug use surveys in enormous detail. What jumps out at me is that there are certain risk factors that make an overdose death much more likely. Given this, our policy of general drug prohibition is a mistake. We should instead issue *specific* warnings to people with risk factors, and we should warn against specific risky behaviors.

Age and Health

Age and health are both huge risk-factors in an overdose death. As I mentioned in this previous post (here), older people appear to be dying of drug overdoses *way* out of proportion to their use rates. And the opposite is true of young people; the young, while certainly not immune to drug poisonings, are far more resistant than the older users. There is an alternative explanation for these data, which I explore in the link. It’s possible that many of these “overdose deaths” are misclassified, where the decedent dies of a different cause. It’s just that people with drugs in their system are overwhelmingly labeled “drug poisonings,” even when another cause is responsible. The apparent age-disparity is driven by the higher mortality of older age groups. I think there is something to this alternative explanation, but 1) it implies that there are far fewer drug poisonings than we thought and 2) it stands to reason that as people get older and more infirm they’d be more susceptible to something that would not have killed their younger self. General warnings against drug use (“Just say no!”) are not effective and not informative; specific warnings (“Just because your heart could take it at 20 doesn’t mean your heart can take it at 40!”) might actually prevent some deaths.

In looking at the individual records on these poisoning deaths, it really jumps out at you how unhealthy these people are. It seems like every other record has some non-drug related health issue attached. (See my analysis here.) Sleep apnea and obesity are very commonly represented among these drug poisoning deaths. It’s easy to imagine that someone who has trouble breathing is more susceptible to an overdose from drugs like opioids, benzodiazepines, and alcohol, which all can make you sleepy and suppress respiration. Also, heart conditions are fairly commonly listed on the death records. It’s fair to ask if these people would have died if they’d been in better physical health. Once again, a general warning against drug use misses the point. A specific warning to people in poor health might be more helpful. Very specific drug-illness interactions would be even more helpful. “Don’t use cocaine if you have a heart condition.” “Don’t use more opioids than prescribed if you have sleep apnea.” Etcetera. We could still even issue the general warning, so long as it’s followed up with one of these “especially if…” provisions.

Multi-Drug Interactions

Overwhelmingly, these drug poisonings are multi-drug interactions. I thoroughly explored the interaction issue in this post. For every major category of potentially lethal drug, the vast majority of deaths involve multiple substances. Some of these interactions are well understood. Benzodiazepines, opioids, and alcohol interact to suppress respiration. If someone takes a non-toxic dose of several of these substances, the combination can be deadly even though each substance alone would not have been. Once again, a specific warning not to combine certain drugs would be extremely helpful and could save a lot of these lives. A ham-fisted “Drugs are bad. Don’t do drugs” doesn’t accomplish much. The problem is that people start ignoring such warnings as soon as they survive an encounter with drugs. They might start assuming that everything the government tells them is a lie, and then they won’t believe anything else they hear. Well, here’s a quantifiable risk, with easy-to-cite data sources. We don’t even have to wait for a big policy change to implement this warning. People can unilaterally change their behavior once they know the *specific* risks of drug use. Such a behavior change doesn’t require a big congressional battle over drug policy, just accurate information about risks.

One caveat here is that sometimes there are combinations of drugs on the death record that don’t seem to have a dangerous pharmacological interaction. To name a couple of dubious combinations: Does cocaine really negatively interact with benzodiazepines? Do opioids interact dangerously with cannabis? (For that matter, did cannabis actually *cause* any of the deaths in which it’s listed on the certificate?) Or are the medical examiners promiscuously listing everything they find on the toxicology screening? It’s possible that some of these truly are single-drug poisonings and irrelevant factors are listed on the death certificate. But if we admit that irrelevant factors are listed on the death certificate, that once again opens up the possibility that many of these deaths aren’t really drug poisonings in the first place. I suspect that the opioid-benzodiazepine-alcohol interactions are real and really are driving some of these death numbers. A specific warning against using any of these in combination would spare us a large fraction of the total drug poisonings.

The really sad thing here is that we’re missing an opportunity to meaningfully inform people of risks. We’re missing a chance to save lives. And it’s all because the US government has insisted on a puritanical “thou shalt not” policy regarding all questionable substances.

Someone else came to a similar conclusion after doing their own research. See this excellent article, including the comments section (Less Wrong tends to have excellent comments):

http://lesswrong.com/lw/45u/a_rationalists_guide_to_psychoactive_drugs/

The very first comment makes the following point:

“Even the worst drugs that legitimately screw people up do so because those people are taking them in ridiculously stupid ways. These drugs can often be useful tools to have when used responsibly.”

This is consistent with what I’ve seen in the CDC data. Infirm people, often using multiple substances, are driving the death counts.