Author Archives: Robin Hanson

On Our Own

I didn’t expect this result, and it seems so terribly sad. Perhaps the saddest thing I have ever heard.

While other animals have feelings on fairness, and inclinations to retaliate against unfairness done directly to them, as far as I know only humans having norms requiring generic third parties help fix such complaints. So while other animals assume that they will only get as much justice as they (and their immediate allies) can force the world to give them, we humans are led to see our larger society as responsible for creating a justice world around us, and even led to expect that it will in fact typically provide such justice.

It must then be a crushing blow to realize that this is just not so. Even for pretty big injustices, most all of us see ourselves as better off to just suffer them, instead of publicly complaining about them, the vast majority of the time. Our larger societies do not in fact provide much justice; the justice we get is in fact mostly whatever we (and our immediate allies) can force the world to give us. In terms of justice, human societies today only produce a minor correction to the basic animal situation. That is, we are on our own. Our friends and family may help, but the rest of the world will not.

Oh, some of us probably do mostly get justice from the world. The rich, the pretty, the popular, the well-connected. When they complain, enough people care, and actually do something. For them, maybe it makes sense to complain about most big injustices. But not for the vast majority of us.

Was it ever any different in human history? I suspect not, alas. Maybe, someday in the future, the human thirst for justice will lead us to create societies that actually do stop most injustice, so that people who are are treated unjustly will usually think it worth their bother to publicly complain. So that such injustices are stopped. That is, someday, we may find a way to slake our thirst for justice. But so far, we remain incurably thirsty. That is the human condition.

Added 9a 12June: Some claimed that my poll wording was oft misinterpreted, and some claimed that restricting to “bothers you a lot” did not sufficiently distinguish minor from major injustices. So I did three more polls like the above, reworded a bit to avoid the misinterpretation, and distinguishing three levels of injustice: would have paid <$100, $100-10K, and >$10K to avoid them.

Looking at median of lognormal fits to % of cases where complaining is a net win, I find 7.1% for “bothers you a lot”, 1.9% for <$100, 4.1% for $100-$10K, and 7.7% for >$10K injustices. Thus the threshold for “bothers you a lot” seems to be near $10K, and while we do find it in our interest to complain more often for larger injustices, cases where complaining wins remain rare exceptions even for large injustices. Even then, we are mostly on our own.

GD Star Rating
loading...
Tagged as: ,

No Recent Automation Revolution

Unless you’ve been living under a rock, you know that for many years the media has been almost screaming that we entering a big automation revolution, with huge associated job losses, due to new AI tech, especially deep learning. The media has cited many “experts” making such claims, most every management consulting firm has felt compelled to issue a related report, and the subject came up in the Democratic US presidential debates.

Last December, Keller Scholl and I posted a working paper suggesting that this whole narrative is bullshit, at least so far. An automation revolution driven by a new kind of automation tech should induce changes in the total amount and rate of automation, and in which kinds of jobs get more automated. But looking at all U.S. jobs 1999-2019, we find no change whatsoever in the kinds of jobs more likely to be automated. We don’t even see a net change in overall level of automation, though language habits may be masking such changes. And having a job get more automated is not correlated at all with changes in its pay or employment. (There may be effects in narrow categories, like jobs that use robots, but nothing visible at the overall level of all automation.)

Two metrics created by groups trying to predict which jobs will get automated soon did predict past automaton, but not after we included 25 mundane job features like Pace Determined By Speed Of Equipment and Importance of Repeating Same Tasks, which together predict over half of the variance in job automation. The main change over the last two decades may be that job tasks have gradually become more suitable for automation, because nearby tasks have become automated.

Our paper has so far received zero media attention, even though it contradicts a lot of quite high visibility media hype, which continues on at the same rate. It has now been officially published in a respected peer reviewed journal: Economics Letters. Will that induce more media coverage? Probably not, as most of those other papers got media attention before they were peer reviewed. The patterns seems to be that hype gets covered, contradictory deflations of hype do not. Unless of course the deflation comes from someone prestigious enough.

For Economics Letters we had to greatly compress the paper. Here is the new 40 word abstract:

Wages and employment predict automation in 832 U.S. jobs, 1999 to 2019, but add little to top 25 O*NET job features, whose best predictive model did not change over this period. Automation changes predict changes in neither wages nor employment.

And Highlights:

  • 25 simple job features explain over half the variance in which jobs are how automated.
  • The strongest job automation predictor is: Pace Determined By Speed Of Equipment.
  • Which job features predict job automation how did not change from 1999 to 2019.
  • Jobs that get more automated do not on average change in pay or employment.
  • Labor markets change more often due to changes in demand, relative to supply.
GD Star Rating
loading...
Tagged as: , ,

Philosopher Kings in Blue?

When things go wrong in our lives, we are often tempted to invoke governments to fix them. So we add more systems wherein governments do things, and we make more laws to influence what other people do. However, in the messy process of translating our general purposes into particular system and rules, we often allow various groups to control important details, and turn them more to their purposes. We also get random outcomes due to randomness in which political factions happen have more control when we turn our attention to changing each particular system or rule. In addition, we often leave out details because we are hypocritical, and unwilling to fully admit our real purposes. For example, we often want to appear to oppose things more than we do, like say drug use, prostitution, or adultery.

The net effect of these many messy processes is that our government systems and rules are poorly integrated, clumsy, and vague. We don’t bother to work out many details, and we don’t decide how to make key tradeoffs between different systems and rules. For such elaboration, the public and their politicians often punt to judges and government agencies. And for details where agencies don’t set policies, they punt to individual civil servants.

To influence these agencies and their civil servants, we set bosses who can give them orders, and perhaps promote or fire them. Bosses who have other bosses all the way up to the politicians we elect. But we are afraid of new politicians taking too much hidden control over these agencies, say by firing everyone and hiring all their friends. So we often limit politicians’ powers to direct and fire civil servants. This gives agencies and civil servants more discretion, to do what they choose.

Of course in any one social equilibrium, an individual civil servant may not feel they have great discretion. But that doesn’t contradict the claim that collectively they have a lot. That is, there can be many possible government equilibria consistent with the overall government rules and larger political and social worlds. Some of this government discretion may be captured by the schools and other systems that train people to become civil servants.

To enforce rules on both civil servants, and on ordinary people, we threaten to punish people for violating rules. The civil servants we put in charge of this enforcement process are “police” (in which I include prosecutors, judges, and other civil servants with rule-enforcing discretion). And to help police in these roles, we give them various budgets and powers.

The above description so far is pretty generic, applying nearly as well to a quite minimal state as to a strong “police state”, wherein police have strong powers to punish most anyone they choose. Where any one state sits on this spectrum is determined by many factors, including (1) police monetary budgets, (2) police direct powers to invade spaces, demand info, etc., (3) police negotiating powers regarding court proceedings, and (4) the frequency and severity of rules that people frequently violate.

While once upon a time (say two centuries ago) the U.S. system looked more like a minimal state, today it looks more like a police state. Maybe not as bad a police state as the old Soviet Union, but still, a police state. This transformation is detailed in William Stuntz’ excellent book The Collapse of American Criminal Justice. Some key changes:

  1. We’ve added a lot more laws, so many that we don’t understand most, and regularly violate many.
  2. We’ve cut the use of juries and also many legal defenses, which previously helped evade guilty verdicts.
  3. Rise of big cities means county-set laws are set by folks different from those suffer, cause most crime.
  4. States, who set prison budgets but don’t control conviction rates, greatly increased prison budgets.
  5. Legal trial complexity & cost has risen greatly, and is now beyond what most can afford.
  6. Plea bargaining is now allowed, which strongly pushes people to plead guilty, even when they aren’t.
  7. The new doctrine of qualified immunity protects government officials from many lawsuits.
  8. Most complaints about police have long been investigated by the same agency that employs them.
  9. The rise in civil servant unions, especially police unions.
  10. Surveillance, tracking, and info collection has in many ways become much cheaper.

(Some of these changes resulted from courts seeking to encourage big moral movements, such as those against slavery, alcohol, drugs, prostitution, polygamy, and gambling.)

The net effect of all this is that police can, if they so choose, target most anyone for punishment. That is, for most any target, police can relatively cheaply find a rule the target violated, pressure others to testify against the target, and then finally pressure the target to plead guilty. And police collectively have a lot of discretion in how they use this power. (The rich and politically well-connected may of course be able to discourage such use of power against themselves.)

Of course, the fact that police are powerful hardly implies that they use such powers badly. It remains quite possible that, like the proverbial super-hero, they use their super-powers for good. Many people have long claimed that the best form of government is one run by good-hearted but unconstrained philosopher kings.

This is the context in which I’d like you to consider current complaints about police mistreatment of detainees. Police must make difficult and context-dependent tradeoffs between how carefully to avoid hurting detainees, and how aggressively to discourage them from defiance or escape.

These are the sort of areas where, in our system, local civil servants and their agencies have great discretion, and where the basic nature of our government and legal systems makes it hard to pull back such discretion. I’m not saying that nothing can be done; things can and should be done. But I’m pretty sure that the sort of modest changes being now considered (more training, more record keeping, “requiring” body cams, etc.) can’t greatly change what is in essence a police state. (In contrast, changing to a bounty system might do a lot more.)

Look, imagine that while interacting with police you started to insult them and call them terrible ugly names. In many places, this is probably perfectly legal. However, you’d be rightly reluctant to do this, as you’d know they have a many ways to retaliate. If their local people and culture are inclined to retaliate, and to build a “blue wall of silence” around it, there is little most people can do to protect themselves.

This is why you can’t really count on laws that say you have the right to film police, etc. We basically live in a police state, and in such a state its hard for mere rules to greatly change police behavior. We may well be gaining some benefits from such a police state, but being able to exert detailed control over police and how they use their great discretion is just not one of them.

GD Star Rating
loading...
Tagged as: , ,

Missing Model: Too Much Do-Gooding

Grim view of human nature … is mistaken, a persistent and counterproductive myth. … the evidence for mass selfishness is extremely thin. … The surprising truth is that people tend to be­have decently in a crisis. To the British, the all-too-familiar example is the cheerful demeanour of Londoners during the Blitz. … New Orleans after Hurricane Katrina … rumours ran wild about the murder and rape of children inside the Louisiana Superdome; but when the national guard showed up, … met instead by a nurse asking for medical supplies. (more)

Friday I asked the author of a pandemic novel what he thought went most wrong in his fictional world. He said selfishness: blaming others, and not sacrificing enough to protect others from infection. He also said he was surprised to see people acting less selfishly than he predicted in our real pandemic.

As the above quote indicates, that’s a common mistake. In this pandemic I estimate that the bigger problem is people pushing for too much “helping”, rather than too little. That’s a common problem in health and medicine, and this poll says 2-1 that it is the more common problem:

Of course my Twitter followers are probably unusual by this metric; I’d bet most think selfishness is the bigger problem. One reason is that it can look suspiciously selfish to say there’s too much do-gooding, as if you were trying to excuse your selfish behavior. Another reason is that the theory of selfishness is simpler. In economics, for example, we teach many quite simple game theory models of temptations to selfishness. In contrast, it seems harder to explain the core theory of why there might be too much do-gooding.

This seems to suggest a good and feasible project: generate or identify some good simple game theory models that predict too much do-gooding. Not just personal signaling acts that do too much, but acts that push collective norms and decisions toward too much do-gooding. I’d be happy to help with such a project. Of course it would make only a small contribution to the problem, but still I’d guess one worth the trouble.

GD Star Rating
loading...
Tagged as: , , ,

Science 2.0

Skepticism … is generally a questioning attitude or doubt towards one or more items of putative knowledge or belief or dogma. It is often directed at domains, such as the supernatural, morality (moral skepticism), theism (skepticism about the existence of God), or knowledge (skepticism about the possibility of knowledge, or of certainty). (More)

Humans have long had many possible sources for our beliefs about the physical world. These include intuitive folk physics, sacred scriptures, inherited traditions, traveler stories, drug-induced experiences, gadget sales pitches, and expert beliefs within various professions. And for a very long time, we paid the most attention to the highest status sources, even if they were less reliable. This encouraged gullibility; we often believed pretty crazy stuff, endorsed by the high status.

One ancient high status group was astronomers, whose status was high because their topic was high – the sky above. It so happened that astronomers naturally focused on a small number of very standard parameters of wide interest: the sky positions of planets and comets (anything that moved relative to the stars). Astronomers often gained status by being better able to predict these positions, and for this purpose they found it useful to: (1) collect and share careful records on past positions, (2) master sufficient math to precisely describe past patterns, and (3) use those patterns to predict future parameter values.

For a long time astronomy seemed quite exceptional. Most other domains of interest seemed to have too much fuzziness, change, and variety to support a similar approach. What can you usefully measure while walking through a jungle? What useful general patterns can simple math describe there? But slowly and painfully, humans learned to identify a few relatively stable focal parameters of wide interest in other domains as well. First in physics: velocity, weight, density, temperature, pressure, toughness, heat of reaction, etc. Then in dozens of practical domains.

With such standard focal parameters in hand, domain experts also gained status by being able to predict future parameter values. As a result, they also learned that it helped to carefully collect shared systematic data, and to master sufficient math to capture their patterns.

And thus beget the scientific revolution, which helped beget the industrial revolution. A measurement revolution starting in astronomy, moving to physics, and then invading dozens of industrial domains. As domains acquired better stable focal parameters to observe, and better predictions, many such domains acquired industrial power. That is, those who had mastered such things could create devices and plans of greater social value. This raised the status of such domain experts, so that eventually this “scientific” process acquired high status: carefully collecting stable focal parameters, systematically collecting and sharing data on them, and making math models to describe their patterns. “Science” was high status.

One way to think about all this is in terms of the rise of skepticism. If you allow yourself to doubt if you can believe what your sources tell you about the physical world, your main doubt will be “who can I trust?” To overcome such doubt, you’ll want to focus on a small number of focal parameters, and for those seek shared data and explicit math models. That is, data where everyone can check how the data is collected, or collect it themselves, with redundant records to protect against tampering, and explicit shared math models describing their patterns. That is, you will turn to the methods to which those astronomers first turned.

Which is all to say that the skeptics turned out to be right. Not the extreme skeptics who doubted their own eyes, but the more moderate ones, who doubted holy scriptures and inherited traditions. Our distant ancestors were wrong (factually, if not strategically) to too eagerly trust their high status sources, and skeptics were right to focus on the few sources that they could most trust, when inclined toward great doubt. Slow methodical collection and study of the sort of data of which skeptics could most approve turned out to be a big key to enabling humanity’s current levels of wealth and power.

For a while now, I’ve been exploring the following thesis: this same sort of skepticism, if extended to our social relations, can similarly allow a great extension of our “scientific” and “industrial” revolutions, making our social systems far more effective and efficient. Today, we mainly use prestige markers to select and reward the many agents who serve us, instead of more directly paying for results or following track records. If asked, many say we do this because we can’t measure results well. But as with the first scientific revolution, with work we can find ways to coordinate to measure more stable focal parameters, sufficient to let us pay for results. Let me explain.

In civilization, we don’t do everything for ourselves. We instead rely on a great many expert agents to advise us and act for us. Plumbers, cooks, bankers, fund managers, manufacturers, politicians, contractors, reporters, teachers, researchers, police, regulators, priests, doctors, lawyers, therapists, and so on. They all claim to work on our behalf. But if you will allow yourself to doubt such claims, you will find plenty of room for skepticism. Instead of being as useful as they can, why don’t they just do what is easy, or what benefits them?

We don’t pay experts like doctors or lawyers directly for results in improving our cases, and we don’t even know their track records in previous cases. But aside from a few “bad apples”, we are told that we can trust them. They are loyal to us, coming from our nation, city, neighborhood, ethnicity, gender, or political faction. Or they follow proper procedures, required by authorities.

Or, most important, they are prestigious. They went to respected schools, are affiliated with respected institutions, and satisfied demanding licensing criteria. Gossip shows us that others choose and respect them. If they misbehave then we can sue them, or regulators may punish them. (Though such events are rare.) What more could we want?

But of course prestige doesn’t obviously induce a lawyer to win our case or promote justice, nor a doctor to make us well. Or a reporter to tell us the truth. Yes, it is logically possible that selecting them on prestige happens to also max gains for us. But we rarely hear any supporting argument for such common but remarkable claims; we are just supposed to accept them because, well, prestigious people say so.

Just as our distant ancestors were too gullible (factually, if not strategically) about their sources of knowledge on the physical world around them, we today are too gullible on how much we can trust the many experts on which we rely. Oh we are quite capable of skepticism about our rivals, such as rival governments and their laws and officials. Or rival professions and their experts. Or rival suppliers within our profession. But without such rivalry, we revert to gullibility, at least regarding “our” prestigious experts who follow proper procedures.

Yes, it will take work to develop better ways to measure results, and to collect track records. (And supporting math.) But progress here also requires removing many legal obstacles. For example, trial lawyers all win or lose in public proceedings, records of which are public. Yet it is very hard to actually collect such records into a shared database; many sit in filing cabinets in dusty county courthouse basements.

Contingency fees are a way to pay lawyers for results, but they are illegal in many places. Bounty hunters are paid for results in catching fugitives, but are illegal in many places. Bail bonds give results incentives to those who choose jail versus freedom, but they are being made illegal now. And so on. Similarly, medical records are more often stored electronically, but medical ethics rules make it very hard to aggregate them, and also to use creative ways to pay doctors based on results.

I’ve written many posts on how we could work to pay more for results, and choose more based on track records. And I plan to write more. But in this post I wanted to make the key point that what should drive us in this direction is skepticism about how well we can trust our usual experts, chosen mainly for their prestige (and loyalty and procedures) and using weak payment incentives. You might feel embarrassed by such skepticism, thinking it shows you to be low status and anti-social. After all, don’t all the friendly high status popular people trust their experts?

But the ancient skeptics were right about distrusting their sources on the physical world, and following their inclination helped to create science and industry, and our vast wealth today. Continuing to follow skeptical intuitions, this time regarding our expert agents, may allow us to create and maintain far better systems of law, medicine, governance, and much more. Onward, to Science 2.0!

GD Star Rating
loading...
Tagged as: ,

Max Policing Is Disproportionate

NYT: Minneapolis Police Use Force Against Black People at 7 Times the Rate of Whites … most use of force happens in areas where more black people live. Although crime rates are higher in those areas, black people are also subject to police force more often than white people in some mostly white and wealthy neighborhoods, though the total number of episodes in those areas is small. (More)

Star Slate Codex: When restricted to index crimes, dozens of individual-level studies have shown that a simple direct influence of race on pretrial release, plea bargaining, conviction, sentence length, and the death penalty among adults is small to nonexistent once legally relevant variables (e.g. prior record) are controlled.

in 62% of studies, police are not searching blacks disproportionately to the amount of crimes committed or presumed “indicators of suspiciousness”. In 38% of studies, they are. … there are two possible hypotheses here: either police are biased, or black people actually commit these crimes at higher rates than other groups. The second hypothesis has been strongly supported by crime victimization surveys,

Blacks appear to be arrested for drug use at a rate four times that of whites. Adjusting for known confounds reduces their rate to twice that of whites. However, other theorized confounders could mean that the real relative risk is anywhere between two and parity. … the people shot by police are less black than the people shooting police or the violent shooters police are presumably worried about. … 24% of blacks charged with drug dealing are acquitted, compared to only 14% of whites. … There seems to be a strong racial bias in capital punishment and a moderate racial bias in sentence length and decision to jail. (More)

The race and policing literature usually asks if the relative rates of stopping, arrest, conviction, sentences, etc. are disproportionate to race, after controlling for relevant factors like crime rates. I know this not just from the above Star Slate Codex review, but also because I’ve supervised many related research papers while teach grad law & economics.

In that context, the NYT article is crazy biased, as it just reports raw race correlations without controlling for anything. Surely NYT crime reporters know better; they should have at least noted relative black/white crime rates in Minneapolis.

The point of my post here, however, is just to point out that we shouldn’t expect proportionate stops, arrests, etc. in the context of policing designed to max crime reduction.

Imagine you are a police charged with cutting crime in your area. Or more realistically, as you don’t control courts or prosecution, you seek to max the number of crimes you charge that are convicted, weighted by their severity (perhaps proxied by sentence severity). Given this goal, you survey a world of possible and reported crimes, and for each one you estimate the chances that a crime happened there, and if so your chance of finding enough evidence to achieve a conviction. In such a situation, your efforts are often disproportionate.

For example, consider two possible murders, one with 10 and the other with 100 plausible suspects. If you put more effort into investigating that first murder, then each suspect there will get a disproportionate fraction of your attention, relative to their chance of guilt.

As another example, imagine a stream of cars passes you, and you pick which of them to stop. You put each car into a class with which you associate a prior rate of success in finding evidence of crime, weighted by crime importance. As you have limited time, you will limit yourself to stopping the car classes with the highest prior rate of success. This can result in a disproportionate relation between the rate of crime and stopping.

Policing that maxes crime detection and prosecution will not in general effect all people, groups, and situations in linear proportion to their rate or chances of committing crimes, perhaps weighted by severity. It will instead focus on the people, groups, and situations with the very highest rates of crime, and lowest costs of finding and collecting sufficient evidence of those crimes. And often that will produce rates of arrest, etc. that are disproportionate to crime rates.

We do know of one situation where maximizing police effort is proportionate to crime rates: when the chance to find sufficient evidence of crime uncovered goes as the logarithm of police effort re each possible crime. This is when every time you double your effort into a possible crime, say putting in two hours instead of one, you increase your chance of finding sufficient evidence on that crime by the same small amount (say 0.1%).

But while there may be situations where this applies, logarithmic dependence hardly seems a general feature of policing. For example, we we know that it fails at the high end, as the chance of conviction can’t exceed 100%. And it fails at the low end when there are fixed costs to start any investigation whatsoever.

We thus expect disproportionate crime-fighting efforts aimed at groups with higher crime rates, and especially at groups with the highest crime rates for identifiable subgroups. The variance, not just the average, matters a lot; the high tails of these distributions should dominate.

People who are poor, unskilled, and impulsive, and who mix more closely with other such folks, are likely to put in less effort, end less expertise, into hiding their crimes. As a result, police effort is likely to detect such crimes more easily. So police who are tasked with finding crimes are likely to focus their efforts more on these people. If you think that is a problem, an obvious solution would be to prioritize policing of the rich, skilled, and self-controlled. If police seek to max fines or sentence severity, why then increase the fines or sentences of such convictions.

But before you do that, ask yourself honestly: are the crimes of the rich actually more socially harmful? It makes more sense to me to just make punishments proportional to our best estimates of their social harm, regardless of who does the crime. And it makes sense to have punishments compensate for lower probabilities of conviction, to maintain the same expected punishment for someone who commits a crime.

But otherwise, if the poor are actually causing more harm, or making it easier to catch their crimes, well it makes sense to me to pursue those crimes more. Even disproportionately.

GD Star Rating
loading...
Tagged as: ,

The High Cost of Masks

In my debate last night with Tomas Pueyo, after I emphasized the high costs of lockdowns he said lockdowns are over. From now on we’ll let people go out, as long as they (M) wear masks and keep 6 feet apart when not at home. Which let me to wonder: just how much cheaper is M than lockdowns?

So I tried some Twitter polls. I asked:

  1. [1.14%] If you by law had to M unless you paid, what % of income would you pay?
  2. [4.6%] If you didn’t have to, but were offered % of income to M, what would it take?
  3. [19.6%] Like (2), except it is 2019 and there is no pandemic.
  4. [8.0%] Like (3), except half of the folks in your area are already M-ing
  5. [5.2%] Like (4), except only masks, no 6ft distancing.
  6. [15.8%] Like (4), except both at home and away.

In brackets before each question is the median answer (using lognormal fits to % responses re theses 4 options: <3%, 3-6%, 6-12%, >12%). Note how widely these estimates vary!

The following factors plausibly influence these responses: (A) personal pain and trouble of wearing masks and keeping apart, (B) endowment effect of preferring to stick with what current law seems to endorse, (C) not wanting to look or act too differently from others nearby, (D) wanting to be and seem pro-social and helpful in slowing the pandemic, and (E) wanting to support your side of current culture wars.

The big variation in median answers suggests that non-A effects are big! And the big variation within each poll also suggests that these costs vary greatly across individuals. We might gain lots from policies that let some pay to avoid M-like policies.

Question (4) seems to me to offer the best estimate of the real social cost of doing M. It has the best chance of avoiding effects C & D, and I don’t see a way to avoid E. Regarding B, I think we do want the endowment effect to go this direction, because in fact M was not the legal default a year ago. Yes we enjoy helping our community in a crisis, but we wouldn’t endorse creating crises just to enjoy such effects. So we shouldn’t include then when calculating how much to avoid or reduce crises.

Now this 8.0% median (27% mean) of income cost of masks and public distancing doesn’t include all costs; businesses and other places must also pay to accommodate your distancing. But it is probably a big fraction of costs. And it is quite a bit lower than the 32% of GDP estimate for recent strong lockdowns.

However, my estimate of the total cost of having 50% of the population infected at 0.5% IFR was 3 weeks of income, or 6% of one year’s income. So if we wear masks for 9 months, that single cost equals the entire cost of failing to contain. So as with lockdowns, we should be wary of spending more to prevent infections than the infections would cause. Yes, we should be willing to overspend for very effective preventions, with high elasticity. But masks aren’t plausibly such a very effective method.

Added 30May: I added new polls, included on the above list as #5,6. Seems more than half the cost is from the masks alone. It also seems that masks at home would give similar benefits, and the above suggests that it costs about the same too. So why isn’t there more of push to require that?

GD Star Rating
loading...
Tagged as: ,

Time to Mitigate, Not Contain

In a few hours, I debate “Covid-19: Contain or Mitigate?” with Tomas “Hammer & Dance” Pueyo, author of many popular pandemic posts (e.g. 1 2 3 4 5 6 7). Let me try to summarize my position.

We have long needed a Plan B for the scenario where a big fraction of everyone gets exposed to Covid19, and for this plan I’ve explored variolation and other forms of deliberate exposure. To be ready, variolation just needs a small (~100) short (1-2mo.) trial to verify the quite likely (>75%) case that it works (cuts harm by 3-30x), but alas, while funding and volunteers can be found, med ethics panels have consistently disapproved. (Months later, they haven’t even allowed the widely praised human challenge trials for vaccines, some of which would include variolation trials.)

One big reason we aren’t preparing enough for Plan B is that many of us are mentally stuck in a Plan A “monkey trap.” Like a monkey who gets caught because it won’t let go of a tasty nut held in its fist within a gourd, we are so focused on containing this pandemic that we won’t explore options premised on failing to contain.

Containment seeks to keep infections permanently low, via keeping most from being exposed, for at least several years until a strong lasting vaccine is widely applied. Mitigation, in contrast, accepts that most will be exposed, and seeks only to limit the rate or exposure to keep medical systems from being overwhelmed, and to maintain enough critical workers in key roles.

Succeeding at containment is of course a much bigger win, which is why containment is the usual focus early in a pandemic. Catch it fast enough, and hit it hard enough with testing, tracing, and isolation, and the world is saved. But eventually, if you fail at that Plan A, and it grows big enough across across a wide enough area, you may need to admit failure and switch to a Plan B.

And, alas, that’s where we seem to be now with Covid-19. Over the last 7 weeks since the cases peak the official worldwide case count has been rising slowly, while official deaths are down ~25%. In the US, deaths are down ~1/2 since he peak 6 weeks ago. You might think, “yay, declines, we are winning!” But no, these declines are just too slow, as well as too uncertain.

Most of the US decline has been in New York, which has just now reached bottom, with no more room to decline. And even if US could maintain the rate of declining by 1/3 every 6 weeks, and repeat that 5 times over 30 weeks (= 7 months), not at all a sure thing, that would only bring daily US cases from 31K at the peak down to 4.2K. Which isn’t clearly low enough for test and trace to keep it under control without lockdown. And, more important, we just can’t afford to lockdown for that long.

You see, lockdown is very expensive. On average, around the world, lockdowns seems to cost about one third of local average income. Yet I estimate the cost of failing to contain, and letting half the population get infected with Covid-19 (at IFR ~0.5%), to be about 3 weeks of income. (Each death loses ~8 QALY.) So 9 weeks of strong lockdown produces about the same total harm as failing to contain! And where I live, we have almost had 10 weeks of lockdown.

If without lockdown the death rate would double due to an overloaded medical system, then paying for less than 9 weeks of added lockdown to prevent that is a good deal. But at that point, paying more than an additional 9 weeks of strong lockdown to prevent all deaths would not be a good deal. So our willingness to pay for lockdowns to cut deaths should really be quite limited. Sure, if we were at a tipping point where spending just a bit more would make all the difference between success and failure, then sure we should spend that bit more. But that’s just not plausibly where we are now.

Yes, sometimes we pay more to prevent harm than we suffer on average from such harms. For example, we pay more for door locks, and security guards, than we lose on average from theft. But those are very effective ways to prevent harm; paying 10% more there cuts harms by much more than 10%. Yet according to my Twitters polls, most see 10% more spent on lockdown as producing much less than 10% fewer deaths. If so, we should spend much less on lockdowns than we suffer from pandemic deaths.

Now if Pueyo is reading this, I suspect he’s screaming “But we’ve been doing it all wrong! Other possible policies exist that are far more effective, and if we use them containment becomes cost-effective. See Taiwan or South Korea.” And yes, other places have achieved better outcomes via better policies. We might not be able to do as well as them now that we’ve lost so much time, but we might well do much better than currently. Pueyo has sketched out plans, and they even seem to be good sketches.

So if we suddenly made Tomas Pueyo into a policy czar tomorrow, with an unlimited budget and able to run roughshod across most related laws or policies, we’d probably get much better Covid-19 outcomes, perhaps even cost-effective containment. But once such a precedent was set, I’d fear for the effectiveness of future czars. Ambitious politicians and rent-seekers would seek to manufacture crises and pseudo-Pueyo-czar candidates, all to get access to those unlimited budgets and legal powers.

Which is a big part of why we have the political systems we do. All around the world, we have created public health agencies tasked with pandemic policy, and political systems that oversee them. These agencies are staffed with experts trained in various schools of thought, who consult with academic experts of varying degrees of prestige. And all are constrained by local legal precedent, and by public perceptions, distrust, and axes of political conflict. These are the people and systems that have produced the varying policies we have now, all around the world.

Yes, those of us in places which have seen worse outcomes should ask ourselves how strong a critique that fact offers of our existing institutions and political cultures, and what we might do to reform them. But there is no easy and fast answer there; good reforms will have to be carefully considered, tested, and debated. We can hope to eventually improve via reforms, but, and this is the key point, we have no good reason to expect much better pandemic policy in the near future than we have seen in the near past. Even when policy makers have access to well-considered policy analyses by folks like Pueyo.

Now it might be possible to get faster political action if Pueyo and many other elites would coordinate and publicly back one specific standard plan, say the “John Hopkins Plan”, that specifies many details on how to do testing, tracing, isolation, etc. Especially if this plan pretty directly copied a particular successful policy package from, say, Taiwan. If enough people yelled in unison “We must do this or millions will die!”, why then politicians might well cave and make it happen.

But that is just not what is happening. Instead, we have dozens of essays and white papers pushing for dozens of related but different proposals. So there’s no clear political threat for politicians to fear defying. Whatever they do, come re-election time politicians can point to some who pushed for some of what they did. So all these divergent essays have mainly limited short term political effects, though they may do much more to raise the status of their authors.

So if political complexity argues against containment now in many places, why doesn’t that same argument apply equally well to mitigation? After all, mitigation can also be done well or badly, and it must be overseen by the same agencies and politicians that would oversee containment. As there is no escaping the fact that many detailed policy choices must be made, why not push for the best detailed packages of choices that we know?

Imagine that you were driving from A to B, and your first instinct was to take a simple route via two interstate freeways, both in big valleys. Your friend instead suggests that you take a backroad mountain short cut, using eight different road segments, many of them with only one lane, and some very wiggly. (Assume no phone or GPS.) That plan might look like it would take less total time, but you should worry about your competence to follow it. If you are very tired, bad at following road directions, or bad at sticking to wiggly roads, you might prefer to instead take the interstates. Especially if it was your tired 16 year old teen son who will do the driving.

Like the wiggly backroad short cut, containment is a more fragile plan, more sensitive to details; it has to be done more exactly right to work. To contain well, we need the right combination of the right rules about who can work and shop, and with what masks, gloves, and distance, tests going to the right people run by the right testing orgs, the right tracing done the right way by the right orgs with the right supporting apps, and the right rules requiring who gets isolated where upon what indications of possible infection. All run by the right sort of people using the right sort of local orgs and legal authority. And coordinated to right degree with neighboring jurisdictions, to avoid “peeing section of the pool” problems.

Yes, we might relax lockdown badly, but we are relaxing toward a known standard policy: no lockdown. So there are fewer ways to go wrong there. In contrast, there are just more ways to go wrong in trying to lockdown even more strictly. And that’s why it can make sense for the public to say to the government, “you guys haven’t been doing so well at containment, so let’s quit that and relax lockdown faster, shooting only for mitigation.” Yes, that might go badly, but it can’t go quite as badly as the worse possible scenario, where we trash the economy with long painful lockdowns, and yet still fail and most everyone gets exposed.

And that’s my argument for mitigation, relative to containment.

GD Star Rating
loading...
Tagged as: , ,

Pandemic Futarchy Design

Researchers … use … feeds from a global network of students, staff and alumni to construct a “stringency index” that boils down to a single number how strictly governments in 160 countries are locking down their economies and societies to contain the spread of the virus. … plans to include state-by-state measures of stringency in the U.S. … latest version … draws on 17 indicators to determine the stringency of the government response. (More)

Not that I hear anyone eagerly clamoring to try, but let me sketch out how one would use decision markets to set pandemic policy. Just to plant another flag on how widely they could be usefully applied, if only enough folks cared about effective policy.

As you may recall, a decision market is a speculative (i.e., betting) market on a key outcome of a decision, conditional on which discrete decision is made. To apply these to the current pandemic, we need to pick

  1. key ex-post-measurable outcome(s) of interest,
  2. likely-enough key decisions which could substantially influence those outcomes,
  3. participants capable of learning a bit about how decisions related to outcomes,
  4. sponsors who care enough about informing these decisions, and
  5. legal jurisdictions that may allow such markets.

Regarding participants, sponsors, and permission, it makes sense to be opportunistic. Seek any sponsors interested in relevant questions, any participants you can get to trade on them, and any jurisdiction that let you want to do. Alas I have no sponsor leads.

For key decisions, we could consider using bills before legislatures, administrative rulings, or election results. But there are a great many of these, we don’t get much warnings about many, and most have little overall impact. So I’d prefer to aggregate decisions, and summarize policy via three key choice metrics per region:
Lockdown Strictness. As described in the quote above, some have created metrics on lockdown strictness across jurisdictions. Such metrics could be supplemented by cell-phone based data on trips outside the home.
Testing Volume. The number of tests per unit time, perhaps separated into the main test types, and perhaps also into accuracy classes.
Tracing Volume. The number of full-time equivalent tracers working to trace who infected whom. Perhaps supplemented by the % of local folks use apps that report their travels to tracing authorities.

Yes, worse pandemic outcomes will likely cause more lockdown, tests, and tracing. But one could look at outcomes that happen after decisions. Such as how average future outcomes depend on the decisions made this month or quarter.

For key outcomes, the obvious options are deaths and economic growth.

For deaths, we can avoid testing problems by looking at total deaths, or equivalently “excess” deaths relative to prior years. It helps to note the ages of deaths, which can be combined with local mortality tables to estimate life-years lost. Even better, if possible, note the co-morbidities of those who died, to better estimate life-years lost. And even more better, have estimates of the relative quality of those life-years.

For economic growth, just take standard measures of regional income or GDP, and integrate them many years into the future, using an appropriate discount factor. Assuming that the temporary disruption from a pandemic is over within say 10 years, one could end the bets after say ten years, projecting the last few years of regional income out into the indefinite future.

As usual, there will be a tradeoff here re how far to go in accounting for these many complexities. I’d be happy to just see measures of life years lost related to lockdown strictness, perhaps broken into three discrete categories of strictness. But I’d of course be even happier to include economic growth as an outcome, and tests and tracing as decisions. Either aggregate all outcomes into one overall measure (using values of life years), or have different markets estimate different outcomes. For decisions, either separate markets for each type of decision. Or, ideally, combinatorial markets looking at all possible combinations of outcomes, decisions, and regions.

GD Star Rating
loading...
Tagged as: ,

Five Ways to Rate

When we have people and orgs do things for us, we need ways to rate them. So we can pick who to have do what, and how much to have them do. And how we evaluate suppliers matters a lot, as they put great effort into looking good according to the metrics we use.

Results – When we buy a pound of roast chicken, or have someone to mow our lawn, we can pretty directly pay for the things we want. The more aspects of what we want that we can articulate and verifiably measure, the more of them we can specify in a contract. When action is risky, not always reliably producing desired results, paying for results means those who do things face payment risk, which they don’t like. And competitors might find a way to produce better results and displace them, a risk they also don’t like.

Record – If we get similar things over and over in a relatively stable context, and also stick with one provider for a while, then we can see a track record of how well they do for us. So we can pay them more of a steady fee not as closely tied to what we get, and drop them when their record seems unsatisfactory. We might switch between providers to sample quality, or hear gossip from associates about their experiences. Groups like Consumer Reports can collect stats on overall customer results. Suppliers face lower payment risks here, though still substantial competition risks.

Prestige – When data on customer results isn’t available, we may rely on a general opinion based on many weak clues about the quality of relevant people and orgs. For people, such clues include wealth, attractiveness, intelligence, social savvy, well-connectedness, etc. For orgs, there is also name-recognition, sponsorships, prestigious projects, and many other elements. Early education and training varies in prestige and adds to individual prestige. When individuals affiliate with orgs, the prestige of each adds to the prestige of the other. I count network effects under prestige; you use the system everyone else respects, As prestige is usually pretty stable over time, suppliers chosen by prestige tend to have a secure position.

Loyalty – While we less often admit it, we often choose suppliers to show our loyalty to “our sides”. Those we choose make sure to signal which sides they are on, and we help to ensure that our associates see those signals, so they can credit us for loyalty. It matters less to us that these signals actually correlate with the things we claim that our side seeks to achieve, as long as they are widely seen as clearly marking folks as on our side relative to other sides. The more stable are sides and signals, the more security a supplier can gain by clearly picking a side.

Procedure – Often specialists create official procedures re how to do something, and rules saying what not to do along the way. Then suppliers can brag to customers that they follow good procedures, they may be required by regulators to do so, and may be punished by courts as negligent if they deviate. Civil servants, for example, are typically paid and promoted based on following official procedures, and on internal politics, not on rates or results. Divisions within private orgs also try to become silos evaluated via rules and procedures, not results. Suppliers tend to like being evaluated by stable rules and procedures that can be achieved with limited effort, as this ensures high job/supplier stability.

While all these methods have their place, the first one, results, seems the most solid and hardest to corrupt from the customers’ point of view. Yes there are obstacles to applying it widely, but such problems are often exaggerated to excuse the other methods. Suppliers would generally rather be evaluated via prestige, loyalty, and procedure, as once these are established they can usually look forward to long stable lucrative relationships, unthreatened by upstart competitors.

I want to find ways to tell people that we could pay for results far more than we now do. Yes, suppliers would resist such a switch, but we customers would get more of what we wanted.

GD Star Rating
loading...
Tagged as: