Saturday, December 30, 2017

Drug Poisoning Data 2016 Update - Detailed Data

Here is my annual update. For the last two years I've dissected and analyzed the CDC's drug poisoning data. You can pull summarized data from the CDC's Wonder database. I've already done that in this post, so you can go there to see very rough trends. See also my analyses from last year and from 2014. You can find the raw data and documentation for them here.

There were 63,938 drug poisoning deaths in 2016, up from 52,623 in 2015. That's a 21% increase. Out of this total, 55,071 were accidental deaths (86.1%), 5,102 were suicides (8%), 3,655 were "unknown intent" (5.7%), and 110 were "murder" (0.2%). In my analysis I mostly fixate on the accidental deaths. Suicides are tragic, but I don't think they count as "the drug problem." If we are trying to quantify the harm caused by drug use or misguided drug policy, I don't think it makes sense to count intentional deaths. Doing so is incredibly speculative ("But for his drug problem, this person would never have fallen into despair and ended his life.") This doesn't matter much, because the majority of drug poisoning deaths are accidental anyway. But I'm always a little bit put off by articles that cite the raw number without a breakdown by intent, or don't justify their methodology of adding together different things to get a bigger total. That's a warning sign. If you read anything that says there were "63,938 drug overdose deaths in 2016" without a breakdown, read the rest of it skeptically. This is similar to when people write about "gun violence" and include suicides when discussing the number of "gun deaths", except in the case of drug poisonings it's not quite as big a deal (suicides do make up a very large proportion of "gun deaths" on the other hand).

Some notable changes from 2015. Most significantly, "Other synthetic narcotics", a category that includes fentanyl and other very powerful opioids, continued its dramatic upward trend; it increased from 9,681 to 19,542. Heroin also increased, from 13,069 to 15,575. Benzodiazepines increased slightly from 9,002 to 10,898. Cocaine increased significantly from 6,945 to 10,588. "Other opioids", a category that includes most prescription opioid pills (Oxycontin, Vicodin, etc.) rose from 12,927 to 14,713. "Psychostimulants with abuse potential", a category including methamphetamine and prescription ADHD treatments like Adderall and Ritalin, increased from 6,082 to 8,036.

Most drug poisoning deaths are multi-drug interactions, so looking at the raw totals by substance will miss most of the story. That's the value of these raw data files. We can look at each individual death record and see what combinations of substances are killing people. (Don't worry, they are anonymized. When you die, the CDC doesn't publish your name and a list of the vices that killed you. So chill out.)

Take a look at the table below to see how important the "multi-drug poisoning" phenomenon is. For most classes of substances, a single-drug overdose is the exception rather than the rule. (Click to enlarge the data table. Sorry, I can't think of a more elegant way of sharing this in a blog post.)

Note that the totals don't match the text above; presumably the WONDER database is using a different filter. I am filtering the full set of death records from 2016 for deaths with a drug poisoning as the underlying cause of death. Some deaths may have a drug category listed as a contributing factor without being categorized as a drug poisoning; I think that explains part of the difference. Also, I am filtering for accidental deaths; the totals listed in the text above are showing suicides, undetermined intent, etc. All my tables and figures in this post are based on accidental deaths, excluding suicides, undetermined intent, and murder. (BTW, there were 2.75 million deaths from all causes in the US in 2016.) [12/31/2017: I edited this paragraph after posting. My original post neglected to mention filtering for accidents. Apologies for any confusion.]

I defined some of my own categories. "Any Opioids" includes heroin, "other opioids", "other synthetic opioids", and/or methadone. Most media accounts of "the opioid epidemic" count deaths this way, but I think this is wrong. It sums together two very different things. Heroin users are a very small number of people taking a very large risk, and prescription opioid users are a very large number of people doing something with very little risk. I included it because everyone else is doing it, but I should state for the record that this conflation is wrong and I don't care for it. "Prescription Opioids" is "other opioids" and/or methadone. I'm trying to capture "opioids commonly prescribed for pain." This might not be right, because a lot of the methadone deaths might be from heroin addicts on maintenance treatment. "Heroin or fentanyl" means the death had either heroin or "other synthetic narcotics" or both. I'm trying to capture the fact that a lot of "heroin overdoses" are due to fentanyl, and the coroner/medical examiner might write either one on the death certificate. But these are really part of the same problem. "Cocaine or Heroin" is exactly what it sounds like. I'm trying to capture the fact that someone who dies with one of these drugs in their system probably had a serious drug habit. So finding one of these in combination with something more innocent (like prescription opioids or benzodiazepines) probably means it wasn't an innocent slip-up by someone who is a legitimate prescription drug users. "Any alcohol" includes "ethanol", "Accidental poisoning by exposure to alcohol", and "alcohol, unspecified." (Why three categories for ostensibly the same thing? Are there specified uses for all of these categories? Or does the medical examiner just say "screw it" and pick one?)

Now take a look at what drugs tend to occur in combination with each other. (Again, click to enlarge; sorry for the terrible formatting.)

It's a lot to digest. What is so striking is the increase in the co-occurrence of cocaine and other synthetic narcotics. In the 2015 data 23% of cocaine-related deaths also included very powerful synthetic opioids like fentanyl. In the 2016 data this is up to 40%. Really, this co-occurrence is up across the board in 2016, for all substances. Someone without access to the detailed data (in other words, someone who is only looking at the summarized data from the WONDER database) might suspect that drug use across the board has increased, so we're seeing more overdoses from every category. That doesn't appear to be what's happening. Use rates haven't shot up in the most recent year, so a more likely story is that fentanyl and its analogues are infesting other drug habits. Fentanyl is very powerful, so it is easy for drug dealers to concentrate, conceal, and transport. But someone has to mix it into street drugs for retail. There is no guarantee that the person doing this understands what they are doing or knows how to dose it. The final consumer of the drugs often has no idea what they are getting. (Mike Riggs at Reason magazine totally called this one.) This is bad. Drug use is becoming deadlier without the total number of users growing.

Consider also "psychostimulants with abuse potential", where deaths increased from ~6000 to ~8000 in 2015 and 2016. Well, in that same one-year span the overlap with "other synthetic narcotics increased from 461 to 991. (As in, 461 deaths involved both in 2015 and 991 deaths involved both in 2016.) The overlap with "heroin" increased from 1058 to 1509. This is a point we have to be very careful about. One substance can drive up the mortality rates of other substances, even if use patterns for those other substances aren't changing. The multi-drug poisoning effect is an underappreciated thing. Benzodiazepines are a more extreme example. A pure "overdose" of benzodiazepines is incredibly rare, but if you look at the aggregated data from the CDC's WONDER database it looks like one of the most dangerous substances. (Again, see my post from a few days ago.) If any substance is driving up the overall mortality from drug poisonings, it will probably drive up benzodaizapine-related deaths. But it would be foolish to describe this as a "benzodiazepines epidemic." Just so for an "opioid epidemic."

I also think it's worth looking at the age distribution of these overdose deaths. Here are the average ages of the decedent for the most deadly substances or categories:

For most substances, the average age is around 40. This should leave a lot of readers scratching their heads, because we're used to sad stories about young people overdosing. For every 19-year-old heroin overdose victim, there's a corresponding 59-year-old. (I know, I know, distributions don't have to be perfectly symmetric. See the graph below for the details of the distribution.) In reality, drug use skews young, even though poisoning deaths skew old. I've written about that here and here. There are two possible interpretations of this pattern. One is that age is a huge risk factor in overdose death. That's perfectly believable. The more infirm you are and the more health problems you have, presumably the more susceptible you are to that one last straw. Another interpretation is that many of these deaths have nothing to do with drugs. They are mislabeled as "drug overdoses", perhaps because there was no other obvious cause of death. It's easy to imagine that some of these older individuals might die of a sudden cardiac arrhythmia, which may leave no obvious trace. If a family member mentions that the decedent was a drug user or if a toxicology screening turns up evidence of cocaine or heroin or prescription painkillers (which are used by ~1/3 of the population in any given year!), that might be a handy explanation for an otherwise inexplicable death. I've written about this here and here, with both posts drawing heavily on Karch's Pathology of Drug Abuse. It is incredibly difficult to establish the cause of anything, and it's important to remember that the "cause of death" on a death certificate is just some guy's educated guess. It may be the best possible guess under limited information, and in some cases it may be more obvious than in others. But don't dismiss the possibility that some of the recent increase in "drug overdose deaths" is due to miscoding and misattributing the cause of death.

Here are the age distributions for some of the most deadly substances:

Kind of surprising. At least it was to me. The distribution doesn't taper off until about the mid 60s or early 70s. (Age 999 is the code for "We didn't know the age of the decedent." Of course it's pretty rare. )

Here's my overall takeaway from the 2016 data. I feel like I called this. I didn't predict a sustained increase in poisoning deaths, but I did see the importance of fentanyl early on. The sustained increase in poisoning deaths is due to fentanyl and other even more powerful opioids. It's worth pointing out that this trend is entirely a result of the illegal status of recreational drugs. There is simply no natural demand for fentanyl (recreationally speaking anyway; chronic pain patients are often prescribed the fentanyl gel patch and some of them find it a godsend). Heroin users are afraid of it. The increase in cocaine-related poisoning deaths is sort of a smoking gun. I seriously doubt that cocaine users were telling their dealers to "Get me some of those hot shots!" I think it's far more likely that a roughly unchanged population of cocaine users were clandestinely given something they didn't ask for. It didn't have a label listing the various active ingredients and their dosages. It wasn't subjected to the kind of quality checks you see in legal markets. And people died.

In my previous work-ups of this data (for 2014 and 2015, linked to above), I stressed my skepticism of the raw numbers. Are things coded correctly? Are "heroin" deaths mislabeled? Should some of them be "other synthetic opioid" deaths? Are substances  that were irrelevant to the death listed on the death record anyway? Were some of these actually drug poisonings at all? I think my skepticism holds up pretty well, too, but I should concede that there is little doubt that drug poisoning deaths really have been rising in the past few years. I don't want to adopt a stance of extreme skepticism, where we can't know anything about anything. But I think there is a lot to be skeptical of in the CDC's data. And even if we are to interpret it literally, there are extremely important details that typical media stories overlook.

Talking Dog Jokes

Trying something a little different. On this blog I have some serious posts and some attempts at humor. This post is one of the latter. If you don't care for my style of humor, I hope you can simply skip over my humor posts and we can still be friends.

Here is the original joke, and below are my own variations:

I'm walking my dog, and I run into a neighbor.
"Cute dog!" my neighbor says.
"Yeah, she's actually a talking dog."
"A talking dog? Nonsense. There's no such thing."
"Oh yeah? Hey, Lucy, what's the thing on top of a house?"
"What's the stuff on the outside of a tree?"
"Okay. How does singer/songwriter Gotye feel about his recent breakup?"

(Neighbor shakes his head, mutters "Bullshit!" and walks away. I just kind of look at Lucy and shrug.)
Lucy: (mustering my best Scooby Doo voice) "I mean, he wrote an entire fucking song about it. I'd say he feels downright shitty."

And here's a different ending starting from the third question:
"When the neighborhood stud dog breaks into the yard, what kind of sex do you guys have?"
(Neighbor shakes his head and walks away.)
Lucy: "I would have said 'doggy style with lots of neck biting', but I thought that description would be a little too graphic for him."

A World Full of Sexually Aggressive Gorillas Smiling and Winking At You

Or as something I read recently put it,

Guys, imagine the world is full of 300 pound NFL linebackers who are sexually interested in you. Most are polite and even likable, but some noticeable fraction of them are total creeps. They give you a knowing smile and wink, maybe even a kissy-face. Some fraction of these are true sexual predators, who would overpower you if they got you alone. Suppose that your only chance at having a successful love life is to navigate this sea of potential predators, who on average have twice as much upper body strength as you do. Think about how uncomfortable it would be to have so much unwanted attention, happening suddenly and unexpectedly from any member of this brutishly-strong half of the human species.

(Paraphrasing from the book Mate by Geoffrey Miller and Tucker Max, which is sort of an evolutionary psychology view of dating and coupling.)

Interestingly enough, Louis CK said something very similar.
If you're a guy, try to imagine that you could only date a half-bear, half-lion. "I hope this one's nice. I hope he doesn't do [raises eyebrows] what he's going to do."
I'm thinking back to when I was younger and a lot more socially clueless than I am today. I probably did not understand this dynamic. "Gee, just decline the unwanted advances. What's the big deal?" I definitely think it's possible to overdo the stigma and to categorize normal-but-unwanted sexual advances as "misconduct." But, Jesus, what a dangerous minefield to have to navigate. Even if there is an initial mutual interest, differing expectations can lead to a dangerous conflict with someone who can overpower you. I can't even imagine. Not even with the vivid scenarios spelled out above.

BTW, I think Louis CK's sex scandal thing was "kind of gross," but not some kind of terrible sex crime.  He claimed it was consensual, while acknowledging that the women would have had a hard time saying "No". If that's accurate, I'd put it in the category of "Gross, but I wish I didn't know those kinds of embarrassing details of someone's awkward sex life." I don't think he deserves the "celebrity career death penalty" for it. If it turns out that he actually barred the door to block the escape route and made repeated advances toward these women, like Harvey Weinstein apparently did, then I'll change my tune. I think it's unfortunate that these things happened at the same time, because they are categorically different. I also don't quite buy that Louis CK had "power" in any meaningful sense in the early 2000s, when those events happened. Certainly not the "I can make or break your career" kind of power that Weinstein had, and even if  he did it isn't clear that he threatened or implied he would use it. I wish the news-reading public had enough bandwidth to process the nuance in these kinds of stories, rather than lumping everything superficially similar into the same category.

Friday, December 29, 2017

Stop Enforcing Bad Social Norms

I recall hearing about someone I knew who didn't get a job because he failed a drug test. Very smart kid, perfectly qualified, but for the test he would have been hired. I thought this was unfair, and the practice was counter-productive for the company following it. I may have even used the word "bullshit."

Someone else defended the idea of using a drug test as relevant information for a new hire. (Neither of us were in a capacity of making hiring decisions, for this individual or anyone else.) He agreed that casual drug use doesn't impair your ability to work. At any rate, if you can directly monitor someone's work performance, you don't need to know an irrelevant detail that might or might not affect someone's work performance. He even agreed that the correlation between drug use and performance is probably poor. So it's not even useful as a signal that someone is "the kind of person who generally makes a bad employee," much less actually causing poor performance. So what was his defense of the practice? Something like, "You're just supposed to know that it's the kind of thing you don't do before a job interview."

There may be some truth to this. It's pretty dumb to finish your college career and enter the field as a job candidate knowing that a drug test will come back dirty. So maybe it is a signal of poor judgment after all? My response to this argument is that it's only a signal of poor judgment if everyone agrees that it's a signal of poor judgment. Many rules of etiquette work this way. Some are perfectly innocuous. Some are downright useful: respect people's personal space, don't make people feel uncomfortable with overly-crude language, don't look distracted during a meeting with a colleague. But some are out-of-date and need to go. Before enforcing or endorsing a social norm, think hard about whether it's actually helpful. If Bob is using too much crude language too loudly and is annoying his neighbors, Bob should probably be reprimanded for creating an unpleasant environment. That's a fairly obvious case where enforcing the norm makes everyone else's lives a little better. "Take your hat off when you get to the office" is more neutral. What to do with this one? Unless there are die-hard hat enthusiasts, enforcing it doesn't really do much harm. But the drug test thing seems like it's in a different category. The company that drug tests is imposing much larger restrictions the lives of its potential employees. If you didn't snoop into someone's blood or urine for traces of their hobby, nobody would know. And even if you do snoop, most people won't care.

I think it's worse than that. Enforcing this norm is socially costly. We'd all be better off if everyone dropped the expectation and stopped drug testing. But maybe you can continue to benefit at everyone else's expense if others are dropping the expectation. Job candidates who fastidiously cohere to all conceivable expectations are probably at least slightly better than job candidates who flout norms. So maybe you can maintain a very slight edge in snatching up the top candidates, but it comes at the cost of keeping a bad social norm in play. We should all strive to stop doing this. Let's leave the world better than we found it.

I can think of other things. Mild swearing (in an interview or at work)? Let it slide, unless it gets truly vulgar or obnoxious. Someone has a single beer at a work lunch? So what? Again, let it slide, unless they are actually getting intoxicated. Someone isn't dressing to the "business casual" standards of your workplace? Again, if it's truly distracting say something, but if it's a violation of routinely-flouted Regulation 17.B.III.d of the official workplace dress guidelines, let it go.

Excerpts from Seeing Like a State

Below are some passages I highlighted in James C. Scott's Seeing Like a State. Don't look to me for a full book review. Scott Alexander and samzdat did the definitive book reviews. I consider that well-trodden grounds. So I'll just share some excerpts that I found interesting. It's hard to summarize, but Scott's overall thesis is that life is messy, interesting, and un-"legible". Governments attempt to shoehorn their citizens into neat categories and standardized methods of production to make taxation and conscription easier.

Strewn throughout the book are statements of incredibly naive vaguely-leftist anti-market sentiment, such as this:
A market necessarily reduces quality to quantity via the price mechanism and promotes standardization ; in markets , money talks , not people.
It's offputting. If I make a big deal out of this, it's because these are so jarring and distracting from the flow of the book. It's like how David Friedman describes the book: there is a lot of great anarchist philosophy, but once in a while the author goes out of his way to remind the reader that he isn't one of those icky pro-market libertarians. Moving on.

A couple of passages on how units of measurement might not agree in different regions:

The pinte in eighteenth - century Paris , for example , was equivalent to . 93 liters , whereas in Seine - Montagne it was 1.99 liters and in Precy - sous - Thil , an astounding 3.33 liters .
Even when the unit of measurement — say , the bushel — was apparently agreed upon by all , the fun had just begun . Virtually everywhere in early modern Europe were endless micropolitics about how baskets might be adjusted through wear , bulging , tricks of weaving , moisture , the thickness of the rim , and so on . In some areas the local standards for the bushel and other units of measurement were kept in metallic form and placed in the care of a trusted official or else literally carved into the stone of a church or the town hall. Nor did it end there . How the grain was to be poured ( from shoulder height , which packed it somewhat , or from waist height ? ) , how damp it could be , whether the container could be shaken down , and , finally , if and how it was to be leveled off when full were subjects of long and bitter controversy .

The following reminded me of how producers will respond to price controls, by varying (usually lowering) the quality of their goods if they can't make a profit or break even at the dictated price:
Kula shows in remarkable detail how bakers , afraid to provoke a riot by directly violating the “ just price , ” managed nevertheless to manipulate the size and weight of the loaf to compensate to some degree for changes in the price of wheat and rye flour . 
Maybe this is something libertarians should be more attuned to. We might be stuck with some of the nasty consequences of bad economic policy even in the absence of government. Social norms can potentially be as oppressive and misguided as populist economic policy.

Some interesting details of property rights in medieval villages:
Trees that have been planted and any fruit they may bear are the property of the family who planted them , no matter where they are now growing . Fruit fallen from such trees , however , is the property of anyone who gathers it . When a family fells one of its trees or a tree is felled by a storm , the trunk belongs to the family , the branches to the immediate neighbors , and the “ tops ” ( leaves and twigs ) to any poorer villager who carries them off . Land is set aside for use or leasing out by widows with children and dependents of conscripted males . Usufruct rights to land and trees may be let to anyone in the village ; the only time they may be let to someone outside the village is if no one in the community wishes to claim them .
A comparison of state-crafted law versus village common law:
Modern freehold tenure is tenure that is mediated through the state and therefore readily decipherable only to those who have sufficient training and a grasp of the state statutes. Its relative simplicity is lost on those who cannot break the code , just as the relative clarity of customary tenure is lost on those who live outside the village.
On the difficulty of valuing a parcel of land:
Surely many things about a parcel of land are far more important than its surface area and the location of its boundaries . What kind of soil it has , what crops can be grown on it , how hard it is to work , and how close it is to a market are the first questions a potential buyer might ask . These are questions a tax assessor would also want to ask. From a capitalist perspective , the physical dimensions of land are beside the point.
On the unintended consequences of government attempts to "efficiently" measure and monitor taxable properties:
The door - and - window tax established in France under the Directory and abolished only in 1917 is a striking case in point . 91 Its originator must have reasoned that the number of windows and doors in a dwelling was proportional to the dwelling’s size . Thus a tax assessor need not enter the house or measure it but merely count the doors and windows . As a simple , workable formula , it was a brilliant stroke , but it was not without consequences . Peasant dwellings were subsequently designed or renovated with the formula in mind so as to have as few openings as possible . While the fiscal losses could be recouped by raising the tax per opening , the long - term effects on the health of the rural population lasted for more than a century.
Fewer windows means less ventilation, pretty awful when your main source of heat is a smokey fire and everything smells like horse poop.

On the central planner's obsession with carving all of reality into neat squares with perfect right triangles:
With a T - square and a triangle , finally , the municipal engineer could , without the slightest training as either an architect or a sociologist , ‘ plan ’ a metropolis , with its standard lots , its standard blocks , its standard width streets . . . . The very absence of more specific adaptation to landscape or to human purpose only increased , by its very indefiniteness , its general usefulness for exchange ”.
Scott does occasionally permit that some of these government-sponsored simplifications and categorizations might be useful:
The Center for Disease Control in Atlanta is a striking case in point . Its network of sample hospitals allowed it to first “ discover ” — in the epidemiological sense — such hitherto unknown diseases as toxic shock syndrome , Legionnaire’s disease , and AIDS . Stylized facts of this kind are a powerful form of state knowledge , making it possible for officials to intervene early in epidemics , to understand economic trends that greatly affect public welfare , to gauge whether their policies are having the desired effect , and to make policy with many of the crucial facts at hand. These facts permit discriminating interventions , some of which are literally lifesaving .
The description of statecraft as "internal colonization" rings true to me:
This caricature of society as a military parade ground is overdrawn , but the grain of truth that it embodies may help us understand the grandiose plans we will examine later. The aspiration to such uniformity and order alerts us to the fact that modern statecraft is largely a project of internal colonization , often glossed , as it is in imperial rhetoric , as a “ civilizing mission . ” The builders of the modern nation - state do not merely describe , observe , and map ; they strive to shape a people and landscape that will fit their techniques of observation .
On the obvious utility of simplification:
To complain that a map lacks nuance and detail makes no sense unless it omits information necessary to its function . A city map that aspired to represent every traffic light , every pothole , every building , and every bush and tree in every park would threaten to become as large and as complex as the city that it depicted. And it certainly would defeat the purpose of mapping , which is to abstract and summarize . A map is an instrument designed for a purpose . We may judge that purpose noble or morally offensive , but the map itself either serves or fails to serve its intended use .
On Taylorism:
An American contribution came from the influential work of Frederick Taylor , whose minute decomposition of factory labor into isolable , precise , repetitive motions had begun to revolutionize the organization of factory work .
Something I learned from a The Great Courses series by Timothy Taylor (no relation, I'm sure), specifically the History of the U.S. Economy in the 20th Century. Lenin was obsessed with Taylorism. Just as communist regimes were generally obsessed with "efficiency" and economies of scale (even though they were unable to achieve said efficiencies or economies):
Lenin was not slow to appreciate how Taylorism on the factory floor offered advantages for the socialist control of production . Although he had earlier denounced such techniques , calling them the “ scientific extortion of sweat , ” by the time of the revolution he had become an enthusiastic advocate of systematic control as practiced in Germany .
On zoning:
The logic of this rigid segregation of functions is perfectly clear . It is far easier to plan an urban zone if it has just one purpose . It is far easier to plan the circulation of pedestrians if they do not have to compete with automobiles and trains . It is far easier to plan a forest if its sole purpose is to maximize the yield of furniture - grade timber . When two purposes must be served by a single facility or plan , the trade - offs become nettlesome .
It's hard to explain without the context of the book (or at least one of the two excellent reviews linked above), but central planning really killed the natural "street life" that grows organically in non-planned cities. 
A large part of the answer can be inferred from Le Corbusier’s second principle of the new urbanism : “ the death of the street . ” Brasília was designed to eliminate the street and the square as places for public life . Although the elimination of local barrio loyalties and rivalries may not have been planned , they were also a casualty of the new city .
There is a square . But what a square ! The vast , monumental Plaza of the Three Powers , flanked by the Esplanade of the Ministries , is of such a scale as to dwarf even a military parade. In comparison , Tiananmen Square and the Red Square are positively cozy and intimate . The plaza is best seen , as are many of Le Corbusier’s plans , from the air . If one were to arrange to meet a friend there , it would be rather like trying to meet someone in the middle of the Gobi desert . And if one did meet up with one’s friend , there would be nothing to do . Functional simplification demands that the rationale for the square as a public visiting room be designed out of Brasília . This plaza is a symbolic center for the state ; the only activity that goes on around it is the work of the ministries . Whereas the vitality of the older square depended on the mix of residence , commerce , and administration in its catchment area , those who work in the ministries must drive to their residences and then again to the separate commercial centers of each residential area .
So maybe it's not so bad to have an un-zoned city. You could have an apartment building, with a pub on the corner to serve all its residents, next to an office building where many of the residents work, and so on, and an interesting social dynamic can develop. All the elements of a normal person's life are all there in one place. People can bump into each other and interact spontaneously. Central planning potentially kills off this dynamic.

Most of those who have moved to Brasília from other cities are amazed to discover “ that it is a city without crowds . ” People complain that Brasília lacks the bustle of street life , that it has none of the busy street corners and long stretches of storefront facades that animate a sidewalk for pedestrians.
A description of a thriving "street-life":
Its streets were thronged with pedestrians throughout the day owing to the density of convenience and grocery stores , bars , restaurants , bakeries , and other shops . It was a place where people came to shop and stroll and to watch others shop and stroll . The shopkeepers had the most direct interest in watching the sidewalk : they knew many people by name , they were there all day , and their businesses depended on the neighborhood traffic . Those who came and went on errands or to eat or drink also provided eyes on the street , as did the elderly who watched the passing scene from their apartment windows .
I highlighted the following passage with the annotation "Private law enforcement!":
Jacobs recounts a revealing incident that occurred on her mixeduse street in Manhattan when an older man seemed to be trying to cajole an eight - or nine - year - old girl to go with him . As Jacobs watched this from her second - floor window , wondering if she should intervene , the butcher’s wife appeared on the sidewalk , as did the owner of the deli , two patrons of a bar , a fruit vendor , and a laundryman , and several other people watched openly from their tenement windows , ready to frustrate a possible abduction . No “ peace officer ” appeared or was necessary. 
More in the same vein:
Jacobs explains that when a friend used their apartment while she and her husband were away or when they didn’t want to wait up for a late - arriving visitor , they would leave the key to their apartment with the deli owner , who had a special drawer for such keys and who held them for the friends. She noted that every nearby mixeduse street had someone who played the same role : a grocer , candy - store owner , barber , butcher , dry cleaner , or bookshop owner . This is one of the many public functions of private business .
Is this what bad zoning is depriving us of? Geez, it's worse than I thought. I read an article or blog post every week or so about how bad zoning is "destroying property values." Maybe that's a very dry and dehumanized euphemism for "destroying bonds between human beings."
What’s more , Jacobs argues , the formal public institutions of order function successfully only when they are undergirded by this rich , informal public life . An urban space where the police are the sole agents of order is a very dangerous place .
Emphasis mine. Attaboy, Scott!
In addition , a high volume of foot traffic stimulated by an animated and colorful neighborhood has economic effects on commerce and property values , which are hardly trivial . The popularity of a district and its economic success go hand in hand . Once created , such places will attract activities that most planners would have specially sequestered elsewhere .
On a zoned restaurant versus a mixeduse district restaurant:
Such a restaurant must make virtually all its profit between 10 A.M . and 3 P.M . , the hours when office workers take their midmorning coffee breaks and lunch breaks before commuting home at the end of the day , leaving the street silent . The restaurant in a mixeduse district , on the other hand , has potential clients passing by throughout the day and into the night . It may therefore stay open for more hours , benefiting not only its own business...
Like monocropped forests , single - purpose districts , although they may initially catch a boom , are especially susceptible to stress . The diverse neighborhood is more sustainable. ...
As a scientific matter it reduced the number of unknowns for which the planner had to find a solution . Like simultaneous equations in algebra , too many unknowns in urban planning rendered any solution problematic or else required heroic assumptions .
Loved this:
Just as it saves a prison trouble and money if all prisoners wear uniforms of the same material , color , and size , every concession to diversity is likely to entail a corresponding increase in administrative time and budgetary cost .
Another intrusion of Scott's market-bashing, once again in the middle of celebrating the emergent order that comes with a little freedom:
[Jacobs] is no free - market libertarian , however ; she understands clearly that capitalists and speculators are , willy - nilly , transforming the city with their commercial muscle and political influence .
Scott doesn't seem to understand that "capitalism" is a series of consensual transactions. The capitalists can't do anything without the consent of the city dwellers.

Then he recites a brilliant quote on the folly of central planning:
To illustrate the diversity of urban life , Jacobs lists more than a dozen uses which have been served over the years by the center for the arts in Louisville : stable , school , theater , bar , athletic club , blacksmith’s forge , factory , warehouse , artists studio . She then asks , rhetorically , “ Who could anticipate or provide for such a succession of hopes and services ? ” Her answer is simple : “ Only an
unimaginative man would think he could ; only an arrogant man would want to . ”
But then he retreats again to naive anti-capitalism:
The analogy to the division of labor in modern capitalist production has implications roughly parallel to those of the military metaphor . Both , for example , require authoritarian methods and central control .
My annotation says "Scott really misses the mark here." He just doesn't seem to understand that these "capitalist power structures" are consensual. Sure, someone working at a big company is part of a hierarchy, and his bosses and his bosses bosses have "power" over him in a vague sense. But his continued employment is voluntary. Almost every organization we join entails a surrender of rights and an agreement to abide by certain norms of behavior, be it an employer, church, social club, gym, or what have you. What makes these institutions useful is that everyone else also agrees to those restrictions. That is fundamentally what makes them so attractive and why people end up joining them.

On revolution and the prospect of counter-revolution (emphasis mine):
After seizing state power , the victors have a powerful interest in moving the revolution out of the streets and into the museums and schoolbooks as quickly as possible , lest the people decide to repeat the experience.  A schematic account highlighting the decisiveness of a handful of leaders reinforces their legitimacy ; its emphasis on cohesion , uniformity , and central purpose makes it seem inevitable and therefore , it is to be hoped , permanent .
On the terrible incentives of communism:
Actually meeting a quota , they knew from bitter experience , only raised the ante for the next round of procurements . In this respect collective farmers were in a more precarious situation than industrial workers , who still received their wages and ration cards whether or not the factory met its quota . For the kolkhozniki , however , meeting the quota might mean starvation . Indeed , the great famine of 1933 - 34 can only be called a collectivization and procurement famine . Those who were tempted to make trouble risked running afoul of a more grisly quota : the one for kulaks and enemies of the state .
For all his capitalism-bashing, the man sees through communism. But then as he starts making sense again he says stuff like this:
The logic was not unlike the management scheme at McDonald’s : modular , similarly designed units producing similar products , according to a common formula and work routine . Units can easily be duplicated across the landscape , and the inspectors coming to assess their operations enter legible domains which they can evaluate with a single checklist .
The comparison totally misses the mark. If McDonald's fails to attract customers and make a decent rate of return for all its standardizing and organizing, it goes out of business. It dies a dignified death. Communism limped along for over half a century, after using brutal methods to force people into compliance. Maybe you can kind of squint and see a similarity between McDonald's and Soviet Russia's grand designs, but one is subject to a very strong selection pressure and the other is not. One requires the willing and continued participation of millions of human beings to persist, and the other does not.

This part reminded me of a recent article by Bryan Caplan:
The state managed to get its hands on enough grain to push rapid industrialization , even while contending with staggering inefficiencies , stagnant yields , and ecological devastation . 
See the graphs of "production frontiers" in the Caplan piece.

Finally Scott offers a few nice words about centralized planning:
And it is apparent that centralized high - modernist solutions can be the most efficient , equitable , and satisfactory for many tasks . Space exploration , the planning of transportation networks , flood control , airplane manufacturing , and other endeavors may require huge organizations minutely coordinated by a few experts . The control of epidemics or of pollution requires a center staffed by experts receiving and digesting standard information from hundreds of reporting units . On the other hand , these methods seem singularly maladroit at such tasks as putting a really good meal on the table or performing surgery .

He identifies true public goods in his list of things that might be well achieved by central planning, and private goods (food and medicine) in his list of things that aren't suited to central planning. Somewhere inside James Scott, there is a decent economist.

What happens when people object to having their lives centrally planned?
Force and brutality was used . The police were the ones empowered together with some government officials . For example at Katanazuza in Kalinzi , … the police had to take charge physically . In some areas where peasants refused to pack their belongings and board the Operation lorries and trucks , their houses were destroyed through burning or pulling them down . House destruction was witnessed in Nyange village . It became a routine order of the day .
Oh, yeah.

I thought this following passage was beautiful. Parisian taxi drivers have spontaneously learned "efficient rule-breaking." If one were to literally follow all the rules and regulations, nothing would be done. They can exploit this with a "work-to-rule" strike:
The premise behind what are tellingly called work - to - rule strikes is a case in point . When Parisian taxi drivers want to press a point on the municipal authorities about regulations or fees , they sometimes launch a work - to - rule strike . It consists merely in following meticulously all the regulations in the Code routier and thereby bringing traffic throughout central Paris.
Here is a case where Scott blames capitalism for something that is really the fault of bad public policy,  namely farm subsidies:
One of the basic sources of increasing uniformity in crops arises from the intense commercial pressures to maximize profits in a competitive mass market . Thus the effort to increase planting densities...
To his credit, Scott is very critical of government policy leading to the same uniformity throughout the book.

Throughout the long section on agriculture and the folly of large, industrialized "monocrops," I was left thinking, "Aren't these crops feeding the world? Didn't large megafarms with row-crops lead to massive labor savings? Didn't it allow us to feed a growing world population on a shrinking acreage?" He offers a rare concession:
This is not the place to attempt to demonstrate the superiority of polyculture over monoculture , nor am I qualified to do so .
My annotation reads "There's the hedge I was looking for." It's a simple one-liner, but in my opinion it undercuts an entire chapter that is far too critical of monoculture. 
A closely related question , which we will address in the next chapter , is why so many successful changes in agricultural practices and production have been pioneered , not by the state , but by the autonomous initiative of cultivators themselves .
I kind of wish he would acknowledge that large organizations can be innovative and efficient and pro-social, just as autonomous cultivators can. 

There are a few great excerpts on the hard-won knowledge gained through experience:
In the days when a case of diphtheria in town was still an occasion for quarantining the patient at home , a doctor was taking a young medical student along with him on his rounds . When they had been admitted to the front hall of a quarantined house but before they had seen the patient , the older man paused and said , “ Stop . Smell the odor ! Never forget this smell ; this is the smell of a house with diphtheria . ”
I laughed out loud at the following:
Chinese recipes , it has always amused me , often contain the following instruction : “ Heat the oil until it is almost smoking . ” The recipes assume that the cook has made enough mistakes to know what oil looks like just before it begins smoking . The rule of thumb for maple syrup and for oil are , by definition , the rules of experience .
The "almost smoking" thing amused me so much I re-read it several times. 

There is a wonderful section spanning two or three pages about an ant infesting a prized mango-tree, and the locals trying to introduce a rival ant species to kill off the infestation:
He then placed the egg - infested fronds against the mango tree and observed the ensuing week - long Armageddon . Several neighbors , many of them skeptical , and their children followed the fortunes of the ant war closely . Although smaller by half or more , the black ants finally had the weight of numbers to prevail against the red ants and gain possession of the ground at the base of the mango tree . As the black ants were not interested in the mango leaves or fruits while the fruits were still on the tree , the crop was saved .
Mat Isa [a village elder] made it clear that such skill in practical entomology was quite widespread , at least among his older neighbors , and that people remembered something like this strategy having worked once or twice in the past .
So we have an improvised solution to a rare problem, requiring knowledge from several different disciplines. Not necessarily the kind of thing you can "centrally plan" for.

Another atrocious case of conflating capitalist profit-making with government taxation:
The reduction or , more utopian still , the elimination of mētis and the local control it entails are preconditions , in the case of the state , of administrative order and fiscal appropriation and , in the case of the large capitalist firm , of worker discipline and profit. 
Really, the capitalism-bashing doesn't ruin the otherwise excellent book, but passages like this one were kind of jarring and out-of-place.

I think in this passage he is trying to paint the picture of an anarchosocialist utopia. Independent artisans perfecting their craft and selling their wares.
To the degree that efficient production could still be organized on an artisanal basis ( such as early woolen manufacturing and silk ribbon weaving , according to Marglin ) , to that degree was it difficult for the capitalist to appropriate the profits of a dispersed craft population .
It's funny. I had just written this post, where I said:
Suppose we started out in a socialist utopia. Independent, self-employed artisans selling their wares to willing customers, who are themselves all self-employed artisans. This wouldn't last long. Some of them would realize they'd be more prosperous if they banded together and realized the efficiencies of division of labor. Before long, they'd realize they could hire some highly skilled managers to organize this division of labor. A highly skilled boss with a grand vision for enhanced productivity, plus a track record of delivering on his promises, would make them all much richer. They'd have to pay him a cut of their sales, but with the added revenue gained through organizational efficiency it would be easy to finance this payment.
 At the time I thought some readers might scoff thinking my description of socialist utopia is a straw-man. Then a few days later I saw the above line in Seeing Like a State. Maybe I pass an Ideological Turing Test on this one?

So that's a taste. Read the book! It's worth you're time, if you're into this kind of thing. Even if you've already read Scott Alexander's lengthy review, there are themes and details you will have missed without the actual book.

Thursday, December 28, 2017

Falsification Exercise: Econometrics and the Minimum Wage

I love the concept of a "falsification exercise." It basically means, take an analysis that supposedly shows a causal effect of some cause (say, a minimum wage increase) on some effect (say, a change in the income share of the lower class). Apply the same analysis to measure the causal effect on some other effect, where there's no conceivable theoretical justification for a causal relationship. This is an incredibly useful concept. Maybe your statistical Rube Goldberg machine shows that x causes y, but it also shows that x causes a, b, c, z, and triple-z, none of which any conceivable theory would have predicted. So maybe we should doubt that your clever economic analysis is telling us what you think its telling us.

Here's an excerpt from the excellent book Minimum Wages by Neumark and Wascher:
[T]he upper-tail evidence constitutes what is often referred to as a falsification exercise. That is, if theory predicts an effect of x (the minimum wage) on y (lower-tail inequality), and such evidence is found, researchers often also explore whether there is an effect of x on another variable, z (upper-tail inequality), which is conceptually related to y but for which theory does not predict an effect on z. If no evidence suggesting an effect of x on z is found, the evidence of an effect of x on y is viewed as more convincing, and vice versa. The point of the Autor, Katz, and Kearney analysis is that, in this case, the falsification exercise fails.

More here:

Autor, Katz, and Kearney also cast serious doubt on previous research that emphasized the importance of minimum wages for changes in wage inequality. The most striking evidence they present is that the minimum wage is strongly correlated with upper-tail wage inequality as well as with lower-tail wage inequality. Indeed, in simple regressions of the 90/50 or 50/10 wage gaps on the real minimum wage, the estimated coefficient on the real minimum wage is larger for the 90/50 gap than for the 50/10 gap (-0.44 vs. -0.27, with both significant). In more complete regression models that account for a time trend, the relative supply of more- and less-educated workers, and aggregate economic conditions, a significant relationship between these gaps and the real minimum wage persists, although the estimated coefficient on the minimum wage is about two-thirds lower for the 90/50 gap than for the 50/10 gap. Autor, Katz, and Kearney conclude that correlations between tween the minimum wage and both upper- and lower-tail inequality measures suggest that the "time series correlation between minimum wages and inequality is unlikely to provide an accurate account of the causal effect of the minimum wage on earnings inequality. Indeed, we view the relationship between the minimum wage and upper tail inequality as potential evidence of spurious causation".
This is good econometrics. Really, it borders on philosophy of science.

Wednesday, December 27, 2017

Ridiculous 60 Minutes Piece on the Opioid Epidemic

Here is the original piece. See McKesson's response. HT Jeff for the story.

The 60 Minutes piece basically rehashes the standard narrative of the opioid epidemic. The CBS correspondent breathlessly and unskeptically interviews DEA agent David Schiller. I am struck by how self-serving and self-congratulatory Schiller comes off. 60 Minutes obviously made no attempt to get a contrary point of view. It leaves the viewer with the mistaken impression that McKesson was unwilling to talk about the story. ("We wanted to speak to a McKesson representative on camera but they declined. But in a statement, McKesson said, "In the interest of moving beyond disagreements… The company agreed to settle with the DEA and DOJ." Ellipses in original; see the link to the 60 Minutes story.) McKesson's very public response seems to contradict that impression. Here is an excerpt:
MYTH: McKesson fueled the opioid abuse problem in this country.
This is an unfounded accusation. We are committed to doing our part to help solve this epidemic, but we are only one part of the pharmaceutical supply chain. McKesson’s role is to distribute medications provided by pharmaceutical manufacturers – including opioid medications – that are ordered by DEA-registered pharmacies and prescribed by DEA-registered and state licensed health care providers.
Each participant in the supply chain can play an important role in combatting this crisis, including:
-The Drug Enforcement Administration (DEA), which sets yearly quotas for the volume of opioids that can be manufactured
-Drug manufacturers that design, develop and promote the medication
-Doctors who identify and diagnose the need for the medication, in addition to prescribing the medication
-Pharmacists who dispense the medication
-Private and public health insurance groups that determine what they will pay for
-Distributors like McKesson that deliver medications ordered by pharmacists to fill prescriptions written by doctors
 MYTH: McKesson knowingly supplies controlled substances, like opioids, to rogue pharmacies who in turn divert the drugs to bad actors.
-This is false. McKesson only distributes controlled substances, including opioids, to DEA-registered and state licensed pharmacies. McKesson maintains – and continuously enhances – our programs that detect and prevent opioid diversion at these DEA-registered pharmacies. Further, every controlled substance ordered by a pharmacy in the U.S. – including both orders that are shipped and those that are deemed suspicious and blocked – is reported to the DEA for their internal database.
-It’s important to note that distributors like McKesson have no line of sight into whether a pharmacy is ordering from multiple distributors. The DEA is the only entity that knows the total amount of controlled substances being dispensed, pharmacy-by-pharmacy across the country.
Self-serving? Sure, maybe, but their description of the supply chain is pretty accurate. It seems absurd to go after the distributor when there are so many other people in the supply chain closer to any wrongdoing. It seems like doctors, pharmacists, and the DEA (who are being fed a steady data-stream on opioid shipments!) are better positioned to monitor whether or not there is misuse of prescription opioids. Given that McKesson reports its shipments to the DEA, it's pretty despicable for the DEA to turn around and say, "Do our job for us. Police America's drug problem." I think this is what's really going on here: It's just easier to blame a large company at the center of production than to monitor all the various distributors, doctors, pharmacies, and patients who might actually be breaking the law or doing something wrong. This is government attempting to be "efficient", meaning lazy.

I see two separate strands of puritanism coming together here in an unholy alliance. Right-wing puritanism, which prohibits anything fun (well...anything that isn't on the "approved" list of vices). From this viewpoint, drug use is "just wrong". The inherent wrong-ness of drug use (err...certain drug use) justifies any amount of violence to suppress it and shields any anti-drug policy from cost-benefit analysis. And there is also left-wing puritanism, which stigmatizes making money. From this viewpoint, large companies can do no right. Large companies are responsible for the foolish behavior of individuals who misuse their products. These separate strands of puritanism come together into a truly revolting blend of populism.

Slick operators like David Schiller fan the flames of populist outrage. They call for the heads of unsympathetic targets, e.g. a faceless corporation like McKesson. Don't fall for it. These guys talk a good talk. They are good at manipulating the media. They want you to believe they are acting in the public interest. But they are careerist shills. Schiller is hunting big game, and he wants to mount a big head on his wall because it will raise his status. The pliant interviewer utterly failed to challenge his assertions.

It's pretty good policy to believe almost nothing you read about the so-called "opioid epidemic." You need to do your own digging to have an informed opinion. Assume that whoever wrote what you are reading got lazy with details, got basic facts wrong, and omitted some relevant information. Yes, you should apply that lesson to this blog, too. Check references. The media just aren't doing their job here. They know that the public eats up a sexy drug panic story, and they are exploiting this weakness to the hilt. This is irresponsible journalism. We're going to end up stuck with bad drug policy because of it. People with intractable chronic pain won't be able to get the only medicine that works for them, and it's all because of sleazy drug cops and an all-too-credulous media.

Drug Poisoning Data from 2016

I have the aggregated data from the CDC's Wonder database, not the individual death record level data. So we'll at least see what that shows. (I just checked just now. The individual death record data isn't out yet. Not sure why it takes over a year to publish this stuff.)

Below are the raw death totals for the most lethal substances (excluding alcohol; tobacco is also lethal but from cumulative harm, not from acute poisoning). Careful interpreting this. Most drug poisoning deaths are multi-drug interactions, not single drug poisonings. So if you try to add these together you will vastly overstate the number of total deaths. I have seen articles that make this embarrassing mistake. (Click directly on the graph for a cleaner display.)

Here are the death rates per 100k population.

Here's what I see.

Heroin (thick yellow line) is continuing a trend started in ~2011.

"Other Synthetic Narcotics" (dark blue) includes fentanyl and other substances that are many times stronger than heroin. It is continuing a dramatic increase that began in 2014, following a steady increase from 1999 to 2005 and a roughly flat trend from 2006 to 2013. Like I've said, this is a product of prohibition. There would basically be no demand for recreational fentanyl if there were a legal market for recreational opioids. This very high death toll is totally unnecessary.

Cocaine (orange) deaths are up above their 2006 peak. For a long time, cocaine was the most lethal drug, but use rates and death rates both started declining after 2006. Cocaine use is flat; it didn't spike up in recent years and is basically unchanged in 2016. (See page 17 here.) So cocaine use is getting deadlier. Mike Riggs at Reason Magazine has written about reports of cocaine laced with fentanyl. This spike in cocaine deaths, along with the continuing rise of synthetic narcotics deaths, seems to corroborate those reports. When the CDC finally publishes the individual death records data, I can look at those and see if the co-incidence of cocaine and "other synthetic narcotics" to see if it rose in 2016. I'll just go on record here predicting that is has.

"Psychostimulants with abuse potential" (pink) includes things like methamphetamine and prescription drugs like Adderall. Meth is chemically almost identical to ADHD drugs, and in fact the government tracks both deaths and use rates for these substances in the same category. (So if you hear someone comparing meth and Adderall, don't assume it's a hyperbolic libertarian drug legalizer shooting from the hip. They're basically in line with official government statistics on drug abuse and the ICD-10 coding system.) People don't usually overdose from these drugs. The same is true of cocaine. I believe there is the occasional fatal arrhythmia from using too much of a stimulant; the same is possible with caffeine. I suspect this is another "tainted with fentanyl" story. Again, I'll report back when I look at the individual death records. I predict I'll see a large uptick in this category overlapping with the "other synthetic narcotics" category, meaning a lot of drug poisonings involved both classes of substances. Use rates of stimulants have basically been flat for a decade and a half.
See page 9 of the 2014 SAMHSA report here; in the 2016 version linked to above, use rates are still at the 0.6 percent of the population, right were it left off in 2014.

(Incidentally I don't know why the people at SAMHSA decline to give time series when they are 1) readily available and 2) relevant to policy discussions. Methodology change? Fine, so show all the data and disclose the methodology change. For example: "The spike/decrease in 2015 is most likely due to a change in the survey question." It's not that hard. I'm thinking someone who is trying to interpret trends in drug poisonings would want to know something about trends in use rates. In my experience, the only people who give this much careful attention to detail are the skeptics of the official standard opioid epidemic narrative.)

Benzodiazepines (purple) and "other opioids" (light blue) are both up, after flattening in 2010. Once again, there is a lot of overlap. Again, I suspect the rising heroin/fentanyl overdoses are driving up the numbers for these drugs. I'll know for sure when I see the detailed 2016 data. Notice how closely these lines follow each other. I've written about this before. Something like 1/3 of prescription opioid poisonings involve benzodiazepines, and something like 85% of benzodiazepine-related deaths involve some kind of opioid (based on the 2015 data I have at my fingertips). The "benzodiazepines" story is really an "opioid interaction" story, and I wish people writing about this emphasized the multi-drug nature of these poisonings. We'd save a lot of lives if we just adopted harm reduction. It's pointless to try to get people to stop indulging entirely, but we can at least educate them on the safest ways to imbibe.

Methadone (green) is continuing its downward trend. I'm not really sure what's going on. In about 2006 people realized that methadone was kind of dangerous for pain management because it has such a long half-life and low cross-tolerance with other opioids. Doctors stopped prescribing it for pain and instead prescribed other opioids like vicodin and oxycontin.

I wish people who wrote about the "opioid epidemic" would be clearer about their timelines. I'm reading a book called Dreamland by Sam Quinones. He lays out the basic standard narrative of the current epidemic:
Doctors became loose with the prescription pad because of a new philosophy of treating chronic pain. Abuse of prescription opioids exploded. Heroin got cheaper, so people with a medically-acquired taste for opioids switched to heroin. 
It all sounds very neat and tidy, but the timeline is all wrong. The revolution in doctors' attitudes about pain started in the 80s. The spike in heroin deaths didn't happen until 2010. Dreamland follows the story of a cartel that sold Mexican black tar heroin in the 1990s and 2000s, but that can't possibly explain the sudden recent increase. The steady rise in prescription opioid deaths from 1999 to 2010 follows a tripling of legal prescriptions over that same time period, but as I've said before illicit use of prescription narcotics was flat, possibly even declining, over that time period. Did prescription narcotics abuse skyrocket up until 2002 (when the SAMHSA use data starts), then level off? Why is this out of sync with the overdose data? Where is the drug use survey data prior to this period, and what does it say? What do drug overdose trends look like in the 1980-1999 era? These were under a different coding system (ICD-9 as opposed to ICD-10), but if the data exist let's at least take a look. It's much more plausible that the very recent spike in heroin and fentanyl overdoses is a product of prohibition. If this very successful Mexican cartel was disrupted by law enforcement, that would bring new players into the market. Maybe that's where all this fentanyl is coming from. It could very well be that the successful Mexican cartel, by maintaining a consistent quality, decreased heroin overdose deaths. Anyway, it's hard to blame what you see in the figures above on a revolution in doctors' attitudes toward pain management that happened two or three decades ago.

The Elusive Stellar Performance Review

Here’s an office-world dynamic that annoys many managers and employees. Employees in most companies get a “performance review”, an annual review of how well they’re doing. Of course, they want their reviews to say they absolutely sparkle. This expectation is bound to disappoint. The average person is…well…average. Even though this person is of course very qualified and was selected out of a pool of many qualified employees, there is bound to be a distribution of “employee quality” with most aggregating around some average value. The “selected from a pool of candidates” filter isn’t going to change the fundamental fact that average is average.

Still, performance reviews tend to be average more often than the statistics of distributions would predict. There’s a selection effect at work. Bad employees get fired or moved to a role more suited to their abilities. Good employees tend to get promoted, and with the promotion comes an inflation of expectations. An employee with stellar performance reviews year after year is an employee who is being held back. Even a pay increase with no nominal change in job description could inflate the expectations of that employee, thus holding down his/her performance reviews. "Yes, you kick ass at your job. If we paid you what we paid your co-workers, you'd get a stellar review this year! With higher pay comes higher expectations." 

I was talking to someone about this recently who gave a mediocre performance review, and the recipient was very upset by it. I was reminded of someone at work who explained the dynamic (described above) to me, years ago. Beginning employees tend to get average reviews, because they haven’t had a chance to shine yet. Employees who shine get placed in a position where the expectations are higher, so once again can expect an average performance review given the expectations. Sorry, but we can’t all live in Lake Wobegone where “all the children are above average.”

A contrary thought occurs to me. A company could probably make their employees happier by making this simple concession of “grade inflation.” But then perhaps that would be costly in the long run, because it might be hard to fire employees with consistent “above average” performance? (According to the official record!) It might even open up the company to lawsuits if it starts massive layoffs of its “above  average” work force. Have some company’s bitten this bullet and adopted a policy of giving above average performance reviews? I'm not sure, but I suspect my experience is somewhat typical. 

The Rapid Decay of Your Education

Throughout my education, I had the repeated feeling that everything I had done before was pointless. It was as though there was no accumulation, no building on prior knowledge. Cross the finish line and graduate high school, and all those silly extracurriculars are forgotten and nobody cares what you learned. Graduate college, and then you can go into any graduate studies program. It helps to have majored in the subject you're going to grad school for, but only a little. There is plenty of time to make up for your deficiencies, anyway.

This has been a constant theme in my life. I graduated high school with a good start. I'd had AP chemistry, calculus, and biology, so I probably started college with about one college semester's worth of course work completed. (If not the actual credit; I declined to take the AP chemistry or biology exams for actual credit.) But I was quickly learning that there wasn't much accumulation or overlap. Taking upper-level physics, it probably helped me a little to have taken linear algebra (for solving problems that collapsed into simultaneous linear equations) and calculus and some differential equations (for equations of motion in classical mechanics and solving for the wave function in quantum mechanics). But in reality I simply re-learned that stuff ad hoc when I needed it. I don't know to what extent I actually "applied" the math I had learned. If my education is any kind of guide, there is very little "transfer of learning" in the world. Almost everything is ad hoc, with an extremely limited application.

My humanities and social science courses in high school left me poorly prepared for my college courses. Again, there was almost no building on past learnings. I re-learned world history, having little recollection of my high school courses on the same topic. After college, I began to read tremendous amounts of non-fiction. In grad school I read everything I could get my hands on from Steven Pinker, Dan Dennett, and Richard Dawkins. And I thought, "How was I missing all of this? How did my college education fail to give me a solid grounding on these basic topics of 'philosophy of science'? How did my 'general education' fail to synthesize all of these separate topics of philosophy, biology, history, and psychology into a coherent whole?" But there I was, learning it on my own.

It gets better. Around the end of my graduate school career, I started studying for the actuarial exams. My friend and I printed off some practice problems, which were really basic probability and statistics questions. I found that most of the questions were fairly simple, and the more difficult questions had some simple concept behind them that I just hadn't learned yet. But I'd never taken a stats course in my life. Even the limited amount of "probability theory" from my physics classes, implicit in statistical mechanics and quantum mechanics, was never taught to me at a basic level. If anything prepared me for this exam, it was my philosophy course on logic and reasoning which did cover some basic probability theory. I studied for the first actuarial exam with a cheap study guide and a large sample of free practice questions I got online. I did read a couple of books on probability theory and distributions, which I borrowed from the university's library. By now I've finished all of the actuarial exams (all nine, or ten depending on how you count the multi-part exam). Plus two online "modules", which are shorter exams that cover a lot of simple insurance concepts. Plus a class each on corporate finance, linear regression, time series analysis, and microeconomics.

After taking my microeconomics class, I had something of a viewquake. It was another "How in the hell was I missing these concepts?!" kind of moment. I consider it an utter failure of my general education, at the high school and college level, that I never had to learn economics. I had a very good textbook. It was Steven Landsburg's price theory text. It lays out everything very nicely and contains some of the most clear-headed reasoning I have ever seen. I have since read anything and everything by Landsburg, and I read tons of books about economics by other authors. One such book was Jeffrey Miron's Drug War Crimes. It was the first place where I encountered the distinction between a "positive statement" and a "normative statement." A positive statement is a statement that's either true or false, something that is fundamentally an empirical question. A normative statement is a "should" question, requiring some kind of value judgment. So many people blather on, egged on by counter-blatherers, because they can't be clear about which kind of statement they are making. Suddenly I had a framework for thinking about the news, analyzing government policy, sometimes even reaching conclusions. There is no coherent way to do this without economics, so I don't know how so many people can finish a formal education without learning any. I keep hearing the argument that "education is about having informed citizens." I don't know what the advocates for this position have in mind. Are students supposed to agitate for the sake of agitating? Are they supposed to learn a canonical set of conclusions from their liberal arts education, without really knowing how to defend those conclusions? Are they supposed to feel more confident in their political conclusions (because now they are "educated") without actually being given the tools to analyze policy? Help me out here. I'm genuinely stumped.

All this brings me up through my actuarial education. Surely with those exams completed, I had learned all the skills required to do my job, right? Nope. My day-to-day job occasionally uses an equation or concept from the formal actuarial syllabus. But I'm more of a data scientist than an actuary. There is some material on generalized linear models and clustering and principal component analysis on one of the upper-level actuarial exams. But to do my job I had to learn to learn the R language, some SQL, the hundreds (thousands?) of little tricks and pitfalls of data manipulation, and various statistical concepts underlying predictive modeling. It's something I've picked up on the job as a result of having to do this or that project. When you need a skill to complete a task, you can generally acquire it fairly quickly. And if you work a modern job, you will most likely be given a problem that nobody has figured out before. It may be something fairly trivial even, like analyzing a novel data set that is specific to your company. But some tasks will stretch you beyond what you've learned in your formal education. If you're lucky and your job is the appropriate level of "challenging", you will constantly be stretched beyond your on-the-job experience, too. I expect this to keep happening to me. My R skills will become irrelevant and I'll have to learn Python instead, or whatever replaces Python. SQL will become obsolete and I'll have to learn MDX. And then someone will invent a clever graphical user interface for all these tools, rendering my hard-won programming skills obsolete once again. I'll have to constantly retool to keep up with emerging trends. I feel like I've gone through at least one cycle like this already. Early on at my first actuarial job, I got pretty good at Excel and VBA. I never use VBA anymore, and Excel just won't cut it for most of the analyses I do.

If your model for education is that you go to school, learn to assemble widgets, then get a job making widgets for the rest of your life, your model is wrong. It is at least a half century out of date. I still hear stunningly naive statements by adults who are several years or decades into their careers.
"How did you pass those actuarial exams without taking a bunch of statistics courses?"
"I think I'll use my irrelevant, 20-year-old foreign language skills to get a different job."
"I don't like my job, so I'll go back to school for another masters degree."
And sure, there's nothing wrong with taking some formal stats courses to help you with statistics-heavy exams, or reviving an old and rusty skill set, or studying a new topic in school. But in each of these cases I sensed a hidden assumption that one needs formal education to actually do stuff, or that stuff from your formal education is inherently useful for landing a job. Nope. Nope on both counts.

If you want a good liberal arts education, pick a topic that you enjoy and read deeply about it. Follow discussions, relevant blogs, academic arguments, lectures on Youtube, internet shouting matches even, track down citations and underlying data. Have a range of interests beyond your core topic. That kind of hobby will round you out as a person, even if it doesn't necessarily build any job-relevant skills. If you want a good career, get a college degree in something that's as technical as you can tolerate. Can't hack math or engineering? Do chemistry. Or biology, or economics, or philosophy, or whatever, and take some courses in other stuff that interests you. Look into industry exams, like the actuarial exams, or the CPCU exams (there are several exam series and certifications geared toward the insurance industry), or Six Sigma, or a  nursing degree. Look for good supplemental education. For me that was Datacamp, plus a bunch of textbooks on machine learning, plus Andrew Ng's Coursera course, plus any interesting webinars I could find, plus playing with every interesting new R library I found. The certifications are great for signalling "I'm a competent person," and you'll need this to land the right job. But you'll learn more practical skills just doing your job and reading supplemental materials on your own.

Two things prompted me to lay this all out. Bryan Caplan's book The Case Against Education is coming out soon (February 2nd in fact!). The book will argue that most of formal education is about signalling your existing talent, not building human capital. (You can check out some of his posts at Econlog about his book.) Given my career trajectory, that all rings true. David Henderson's book The Joy of Freedom (a joy to read, by the way) has a long chapter about the uselessness of education. It's a pretty radical indictment of formal schooling, but it rings true. He starts out by asking you to write down the 10 most important things you've learned in your life. Most of them are probably practical skills unrelated to schooling (how to drive, dress, tie shoes, etc.). You may even write down a few items that you actually did learn about in school, but Henderson makes the point that your formal education probably just barely got you started. Learned to type in school? Sure, but you probably learned to type well by practicing on your own time. Learned math in school? Sure, but you probably learned to apply it to real-world problems by figuring out something at work or by running your own business. So maybe school gives you a bare-bones foundation that you then build upon, but it's hard to believe you wouldn't acquire that on your own as needed. 

Tuesday, December 26, 2017

Admissions I Wish People Would Make

“While I have some confidence in my political conclusions, I must admit that there is sufficient doubt and uncertainty about those conclusions, such that it would be irresponsible to actually implement my preferred policies. I acknowledge the lack of consensus among experts who have studied this topic and broad literature of studies coming to opposing conclusions. I recognize that a policy misstep could do harm, for which I would be responsible. Even for a ‘harmless’ but ineffective and expensive project, the enormous public investment with zero offsetting benefit would be on my conscience. I therefore recognize that the responsible course of action is to do nothing.”

-said nobody. Ever. Unfortunately.

Lifting this from an old Facebook post. Several others emerged in the comments. Most are some form of "I acknowledge there is an empirical literature casting doubt on my favorite policy." 

"Individuals acting privately have not, in general, embraced my ideas, even though they have had plenty of opportunities to do so. I recognize that this outcome represents a verdict against those ideas."

"A higher density of guns will tend to deter crime, as rational criminals recognize the larger risk of victimizing a random person. A higher density of guns will also increase the number of irrational shootings, as a more heavily armed population can more easily commit impulsive gun crimes. I propose disarming peaceful gun enthusiasts on my personal hunch that the second effect is larger than the first."

“I personally believe that a large government expenditure, on the order of a trillion dollars, would stimulate the economy and result in a public benefit larger than the expenditure. However, I recognize that many credible critics doubt the Keynesian framework, and many attempts to empirically measure the size of the multiplier have found it to be smaller than one, implying a negative return to government expenditures. I recognize that it would be irresponsible to spend hundreds of billions of dollars of other people’s money based on disputed and discredited economic theories. I also concede that decisions about how to spend the stimulus money will be made by a political process rife with corruption, which will necessarily degrade the value of any stimulus.”

“I acknowledge the vast and credible literature showing that subsidized medicine doesn’t make people any healthier, and plausibly does the opposite. I nonetheless recommend spending hundreds of billions of dollars every year on subsidized medicine, based on my hunch that the studies are all wrong.”

History versus Data: How Much Do We Know About Past Drug Epidemics?

More to the point, how much is it possible to know about an "epidemic" that happened prior to modern record-keeping? How much do we trust accounts that were assembled prior to the existence of a modern public health bureaucracy?

I recently saw a snarky comment on Facebook linked to a story on the Chinese Opium Wars. The comment mocked the standard drug legalization stance, that most of the problems of drug use are a result of prohibition. To the poster, this was an obvious case of a drug epidemic wreaking havoc without the assistance of bad government policy.

But I read his comment and thought, "The modern United States has a pretty decent public health bureaucracy, collects data on every single death and it's causes, and collects data on drug use rates, and we still don't know what the hell is going on!" I've argued repeatedly that we don't really know how many "drug overdose deaths" are actually drug overdoses versus some unrelated cause. I think plainly deaths, from opioids and particularly heroin and fenatnyl sold has heroin, have increased in recent years. But nobody know the exact counts because nobody actually knows for sure how many of these are real overdoses versus deaths from other causes miscoded as overdoses.

Am I just trying to shade the issue with bland skepticism? Not really. Read a book called Drug War Heresies by MacCoun and Reuter. There is a long section about how the tracking of drug abuse rates and overdose deaths is not very good, even in modern times in European countries. Italy saw a regime change, from a relaxed policy to a harsh one, then back to a relaxed one, based on rhetorical flourishes by politicians about the nation's "drug problem." But there was no good data underlying any of these claims about Italy's drug policy working or not working. No country really had data that was detailed or high-quality enough to assess their drug problem, much less measure the effects of various drug policies, except in the very roughest sense.

I anticipate a reaction to this post that goes something like, "Data schmeta! We have testimonials and descriptions of the problem by contemporaries who witnessed it. They knew a problem when they saw one." We should certainly acknowledge these kinds of testimonials and take them seriously, but with a grain of salt. Remember the guy who got high on bath salts and ate someone's face? And then it turned out he wasn't actually high on bath salts? Someone who watched a lot of news in the 80s might have gotten the mistaken impression that there was a cocaine epidemic destroying our country. But this "epidemic" was grossly exaggerated, like so many drug scares. Most people who used cocaine never had a problem. People tend to exaggerate scary new social problems, and new drugs are especially ripe for this treatment. Read Smoke and Mirrors by Dan Baum for a detailed account of the media totally flubbing its coverage of drug stories and politicians crafting policy based on these misleading accounts.

I'm also tempted to dismiss the accounts of someone who routinely sees a lot of dysfunctional adults and thinks they have a special insight into the cause of the problem. Someone can have a career that allows them to rack up a lot of "plural anecdotes", but that person still lacks the data to analyze the problem. What's missing from their sample is the very large number of people who don't ever have problems with their drug habits.

See a critical take on the Opium Wars by Jeffrey Miron here.
Thus, the evidence suggests China's opium prohibition had a minimal impact on opium consumption.
Here is the full text of their conclusions section:
 This paper has demonstrated that Chinaís legalization of opium in 1858 was not associated with a perceptible increase in opium consumption. This conclusion is subject to several caveats, most importantly that it rests on data for opium exports from India to China, not on direct observation of Chinese opium consumption. Nevertheless, the export data fail to provide even a hint that prohibition had reduced consumption.
The other main caveat is that this conclusion may not apply to other prohibitions. Beyond the obvious differences in time and place, there is little evidence the Chinese government expended substantial resources attempting to enforce opium prohibition, despite its ample rhetoric. Thus, it is not surprising the prohibition had minimal impact. This episode nevertheless raises a cautionary note about the impact of weakly enforced prohibitions, and the abundant evasion and corruption spawned by Chinaís opium prohibition are reminders of the constraints faced by any prohibition, even one with significant enforcement.
There is much more of interest in the paper.