derifatives

Launched: Music Recommendations for Google Music Store

2011-11-16T20:32:00.001-08:00

Today, after much hard work, Google finally launched its music store. My team does all the automated recommendations for this. Hope you like them, and if not, they'll be even better soon.

Recently Read Fantastical Fictions.

2011-08-28T21:21:00.000-07:00

I've managed to make time to read a bit more fiction. I partially blame the iPad and its Kindle functionality. Some spoilers on older books follow.

"Consider Phlebas", Iain M. Banks. What a weird book. I finished and can't decide how much I liked it. Which I guess means it was OK? You starts off thinkng you're in for a grand epic space opera a la Vernor Vinge's "Fire Upon The Deep", but then midway through you realize it's more of a fun romp with the intergalactic war set as a background. Then it gets to the end, and you die, the girl dies, everybody dies, for no good reason. You are wondering if perhaps the whole book is a joke, or a meditation on the futility of war. You read the epilog, where it is revealed that any surviving characters have lost their zest for life and kill themselves or put themselves into cryosleep and wake up millions of years later and THEN kill themselves, that everyone who won a battle went on to lose a war, and you learn in passing that the protagonist's entire race goes extinct later in the conflict. I wish I knew what Banks was trying to do so I could understand whether he'd done it. Three stars, with a big standard deviation.

"Revelation Space", Alastair Reynolds. OK, sure, I enjoyed. Ancient galactic civilizations, threats to the existence of the human race, mysteries, bravado, etc. I'm not great at analyzing plots, but this one suffered from one flaw I know how to look for since my friend Laura tipped me off to it back when we both read "Harry Potter and the Goblet of Fire." "OK, let's get this straight. We're the forces of evil. We are trying to kidnap Harry Potter. We've got a professor at Hogwarts who's on our side, and he can magic up any item such that when Harry Potter touches it, he'll teleport wherever we want. Now, what item should we magic? How about we magic up a pencil and then leave it on his desk? No, wait, I've got a better idea! We'll magic up the cup he'll find at the center of the maze when he WINS THE ENTIRE OLYMPICS." Ummmm... really? Really? Anyways, this book is the sci-fi equivalent, a ridiculously complicated plot to achieve a straightforward result. Still, I couldn't put it down, so three and a half stars.

"What Ho, Automaton", Chris Dolley. Short and sweet. It's steampunk Wodehouse. If you think you want to read that, you most certainly do. The narrator is pretty much a complete Rif character [the sort of character I nearly invariably play in role-playing or assassin games], going off on random tangents, holding on way too tight to his obviously ludicrous ideas. I could not stop laughing. FIve stars.

"Zoo City", Lauren Beukes. It's South African noir with animal familiars. It ended up feeling like maybe a little less than the sum of its parts, and the end was a little more violent than I care for, but solid. Three and half stars.

PoliSci Round Up

2010-06-15T16:30:00.000-07:00

I've been reading a lot of political books recently, and I'll briefly review the most recent three.

"Capitalism and Freedom", by Milton Friedman. A classic. Written 50 years ago, it's amazing how relevant it still feels. Most of the major policy ideas of libertarians are laid out clearly here. [Friedman called himself a "liberal" rather than a "libertarian", but the meanings have clearly drifted over time.] By libertarian standards, Friedman is fairly moderate; he is willing to stand behind using taxation to alleviate poverty and to guarantee primary education, on the grounds that these [at least to him] have substantial externalities [which he charmingly calls "neighborhood effects"]. He is strongly in favor of vouchers rather than privileging government run schools, and he is strongly in favor of a negative income tax or cash transfers to the poor rather than provision of housing projects. He defends each of these clearly. The writing throughout is excellent; I appreciate how Friedman comes off as very humble about the things he doesn't know. I especially enjoyed the chapter on occupational licensure, where the relative merits of registration ["If you want to be a doctor, you have to tell the government"], certification ["The government certifies that you have passed some test"], and licensure ["You cannot practice as a doctor without the government's OK"] are discussed. The book is short and easy to read. Highly recommended. 5/5.

"Government's End", by Jon Rauch. Discusses how the United States government has become paralyzed by special interests, and how the special interests are really everybody. A little bit like Sowell's "Knowledge and Decisions", but not as smart. Offers no real policy prescriptions, other than "Expect government to be big and accomplish little, but keep fighting." For me, the most interesting parts were the discussions of the failed [or at best partially successful] reform efforts of Reagan, Gingrich, and Clinton, and his analysis about why they failed. The book was written in 1998, and I think Rauch would've been shocked by the government passing the PPAC. Meh. 3/5.

"The Battle: How The Fight Between Free Enterprise and Big Government Will Shape America's Future," by Arthur Brooks. Endorsed by Newt Gingrich and Karl Rove, I am apparently not enough of a conservative, or not the right kind of conservative, to appreciate this book. There were many things in the book I agreed with, notably its defense of free markets as one of humanity's greatest discoveries, but the writing was terrible. The book paid occasional lip service to the idea that Bush was as much a big government president as Obama, but it was really an anti-Democrat book in a not very convincing disguise. [I think the Democrats are doing a terrible job, but the Republicans did a terrible job too.] The book's basic thesis, that 70% of the populace wants small government and free markets and equality of opportunity while the other 30% wants big government and equality of outcomes, simply didn't ring true to me at all. Many of my friends support far more redistributionist policies than I do, but that doesn't mean it's black and white. This book ends up demonizing people unnecessarily, making the world appear far more polarized than it is. I guess that's what you need to do to sell books. 1/5.

The French Health Care System

2010-06-12T17:47:00.000-07:00

When I started this series, I planned for the bulk of the posts to be discussions of the health care systems in individual countries. I was sidetracked by a number of other issues, but I'm hoping to start mixing in some individual country posts. I figure France is a good place to start, since it was ranked #1 in the world by the WHO study. The Reid book tends to be a pretty good source book for facts [although I disagree strongly with what I consider its viewpoint]. Another good source for facts is The Grass Is Not Always Greener: A Look at National Health Care Systems Around the World, by Michael D. Tanner of the Cato Foundation. I have found no real disagreements between the Reid book and the Cato report on matters of fact, although the Reid book omits mention of many of the problems described by the Cato report.

The basics of the French system:

Everyone in France is required to belong to a health insurance fund. Which fund you're in is basically a function of your occupation. The largest fund covers about 83% of French residents. Premiums are funded through payroll taxes. Premiums, benefits, and provider reimbursement rates are all set by the government.
The French system is the world's third most expensive as a percentage of GDP, at around 11%.
You're generally free to see whatever doctor you want. Recently, there's been something of a move to using primary care physicians as gatekeepers for specialists, although this is more of a gentle nudge than a hard constraint.
The system in theory is supported by payroll taxes [generally about 20% of income], but in practice, the system has been running large deficits for years.
The average French doctor makes about $55,000 per year, less than half what a US doctor makes. On the other hand, French doctors do not pay for medical school, and their malpractice insurance is vastly cheaper.

Compared to many of the other systems we'll look at, the French system has substantial exposure to market mechanisms:

Most services require substantial copayments [10-40%] at time of service. The percentage of health care paid out of pocket in France is about 13%, which is roughly the same as in the US. Compared to the US, the actual payments for the services do tend to be much lower. Additionally, extremely poor patients and patients with expensive chronic diseases end up being exempted from copays; in this regard, the system can be viewed as having "catastrophic illness" coverage built in.
The French government sets reimbursement rates. It does not set what physicians are allowed to charge. About one-third of French doctors charge rates higher than the government reimburses; in Paris it's about 80%. [This is for private physicians; physicians employed by public hospitals can't bill more.]
Because not all desired services are provided by the mandatory insurance, and because many doctors charge higher rates, approximately 92% of French citizens purchase supplemental private health insurance. Compared to the US, this private insurance market is relatively unregulated: the private plans can decide what benefits to offer, can exclude preexisting conditions, and there is no guaranteed issue.

On the whole, these market mechanisms help the French avoid many of the worst problems we'll see in other countries. On the other hand, there are certainly problems:

Costs in France have been growing very rapidly. The system is viewed as being in financial crisis. The system is about 2/3 as expensive as in the United States, and it seems plausible that the only reason it's not more expensive is that the doctors are paid much less.
Setting doctors salary by decree doesn't really work in the long term; in the last decade, there have been doctor strikes in France.
At least some French experts view their own system as "unsustainable", and feel that private hospitals in the US have better incentives to provide high quality care.
The US "fee for service" system has been criticized because it incentivizes hospitals and physicians to perform more procedures than are necessary. On the other hand, the French system of lump-sum payments to state-run hospitals is being phased out in favor of fee-for-service, in the hopes that this will make it easier for the government to track if the money is being spent efficiently.
As is the case in many countries with national healthcare systems, there's been a lack of capital investment, and patients in the US now have much better access to MRI and CT scanners than in France. [This is actually the case in basically every country with national healthcare except Japan, which is very technology heavy in its approach.]
A French government-commissioned 2004 report described the system as "badly regulated" and "badly governed", and at least partially blamed the deaths of 15,000 elderly French during a 2003 heatwave on the system.
The French system has large inequalities in both access to care and health outcomes based on income and class [see the section "Inequalities in health"], likely because of differing abilities to afford supplemental private insurance. Note that the extremely poor who are not making co-pays will end up having to go to public hospitals rather than seeing private physicians. The average French person seems not to be aware of this, and Reid certainly doesn't mention it.

Overall, the French system is not a terrible one. It is arguably better than the system we have in the US right now, and [in my opinion] almost certainly better than what we're going to get after the provisions of the PPAC come into effect. The system has substantial exposure to market mechanisms, and while it is officially based on notions of equality and universal access, like the US, it is a two-tier system where the rich get better care and have better outcomes. The parts of the system that seem to be working best are the free market-based parts; the government controlled parts seem to be suffering from rapidly rising costs, occasional strikes, and lack of capital. It will be interesting to see how the French system evolves over the next decade.

As a final note, Reid has some fascinating quotes about his interactions with doctors. One doctor says that he manages to improve his small income slightly by "billing his Audi to the medical office, even when he uses the car for personal trips." [Reid, p. 61] When discussing a 2004 law that changed the copay for seeing a specialist without a GP referral from 30% to 40%, another doctor says that none of his patients pay this extra fee because he routinely certifies that there's a GP referral whether or not it's true. [p. 56] From his tone, Reid seems to be in favor or at least neutral on both of these behaviors, but from the point of view of the system, these are both what is generally referred to as corruption [see also "fraud, waste and abuse."] I personally find this absolutely fascinating as a direct example of how when government regulations restrict behavior, there is always a temptation to cheat. This is one of the costs of a regulation-heavy system, and it is one that is easy to underestimate.

Medicare, efficiency, and medical loss ratios.

2010-06-09T19:26:00.000-07:00

A common argument by those who favor government involvement or provision of healthcare is that private insurance companies spend too much money on non-medical uses. The so-called medical loss ratio, which measures the fraction of revenue that insurance companies pay out as medical expenses, is around 80 percent for many private insurance companies, meaning they spend about 20 cents of every dollar on non-medical expenses [although there is also dispute about what's counted as a medical expense, see below]. On the other hand, Medicare supposedly spends only 3% of its income on administration. From this, we can see that "Medicare provides health care much more efficiently than private companies."

I want to make clear that while this is by no means the only argument for government provision of health care, it is a top line argument. Reid makes the point multiple times in his book [pp. 37, 174, for example]. Paul Krugman, a Nobel-prize winning economist who is widely revered on the left, makes this argument.

Does this argument stack up? In my opinion, it does not. This Heritage foundation report by Robert Book discusses the issue in plenty of detail, but I'll summarize the key points here:

The administrative costs for Medicare are underestimated, because various support functions, notably the collection of money, are handled by other agencies. With these figured added in, the correct number for Medicare looks more like 6%.
The administrative costs for private insurance include all the thing you normally think of &mdash marketing, the CEO's exorbitant pay, denying care to the sick &mdash but it also includes things like nurse hotlines and other wellness programs that do not directly take the form of paying claims to outside doctors. [You may have heard about this in the news recently; with the new law requiring certain medical loss ratios, insurers will move to reclassify things like this outside of administrative costs.]
Most importantly by far, the basic concept of the percent of money spent on administrative costs is deeply deeply flawed. A percentage is a ratio of two things. On top you have the administrative costs, and on bottom you have the total costs. Medicare patients on average are much older and sicker than customers of private insurance, so for medicare, the denominator is much bigger. But do we really care about the ratio, or the numerator? According to some reasonable measurements, medicare actually spends more dollars per patient in administrative costs than private insurance. Certainly the numbers are at least close. [Note that if the administrative costs naturally scaled linearly with the amount of care consumed or the number of doctors visits, Medicare might still look efficient. However, Medicare spends less than half a percent of its money processing claims.]

Overall, I find this report pretty much demolishes the claim that "We need nationalized health insurance because Medicare is so much more efficient than private insurance."

It's interesting to look at a dialog related to this report:

Paul Krugman bashes the study in a blog post.
Robert Book, the author of the study, replies to Krugman in the comments section.
Krugman elaborates on his argument, although he doesn't directly acknowledge Book.
Book responds to Krugman again in another comment.

Summarizing some takeaways from the exchange:

If you insure a healthy 25 year old who never goes to the doctor, the medical loss ratio for that person will be 100%, no matter how little you charge them. Private insurance has a lot of these people compared to Medicare. This is the "ratio argument" made vivid.
Krugman considers Medicare and Medicare Advantage to be a good apples to apples comparison, because they serve "similar populations." His argument is that [government-administered] Medicare has a much lower expense ratio than [private] Medicare Advantage plans. Government efficient, private inefficient.
Book points out that Medicare Advantage patients are much healthier than the rest of Medicare, so the populations are not comparable.
Book also points out that the high administrative ratio that the CBO reports for Medicare Advantage includes administrative expenses by doctors, whereas the low number for Medicare includes only plan-level expenses, so the expense ratios are not even an apple-to-apples comparison.

After reading the report and the exchange between Krugman and Book, I am left thinking that Medicare cannot be held up as a shining example of the government delivering care much more efficiently than the private sector and that the ratio of income to medical expenses is a surprisingly non-meaningful measure. But I'm curious what you think.

That Mood Indigo

2010-06-08T17:09:00.001-07:00

I wanted to share a transcription I made as part of my music work this semester.

Here is a transcription of the piano and bass parts to the intro of a recording of "Mood Indigo" by Nina Simone. This is off her very first album, originally released as "Jazz as Played in an Exclusive Side Street Club," now generally known as "Little Girl Blue." As far as I can tell, this album was recorded on Nina's 24th birthday, and she already displays an astonishing, fully formed talent; in some ways she never made a better recording. Already we see her trademark combination of jazz and classical music, as the second half of the intro includes a complex counterpoint that was in my opinion most likely precomposed. The whole track is amazing, not just the intro. Especially noteworthy is her vocal entrance, immediately after the transcription ends, which always sends a shiver down my spine.

This was the first transcription I made at this level of detail, although hopefully not the last. It probably took me 15-20 hours to do the transcription. I didn't use any software to slow the music down, although I did make use of a recorder that allowed me to easily loop just a few seconds over and over again. I also spent about the same amount of time learning Finale, which honestly made me stabby; I use computers for a living, and I found it terribly non-intuitive.

One way to hear the track is to listen to it on Mixapp, which you can join in 15 seconds if you have a Facebook account. [Nina did a couple different versions of Mood Indigo; you want the one that's 4:04 rather than 2:17.]

Enjoy!

Economic implications of the PPAC.

2010-05-31T10:34:00.000-07:00

Some of the economic problems with the Patient Protection and Affordable Care Act [also known as the Obama Health Plan, or perhaps Obamacare a little further to the right] are described beautifully in this short paper. It's about a five minute read, so not a huge commitment.

Questions for thought:

Assuming you support the PPAC in general, do you consider this design, under which the accounting for the bill is done under the presumptions that employers will not drop their health plans but the economic incentives strongly urge them to do so, to be a misfeature?
Assuming you support the PPAC in general, how would you modify the bill to avoid these problems?
For everyone, why or how did the bill get written this way?

The WHO Study

2010-05-21T04:45:00.000-07:00

I have frequently heard supporters of more United States government involvement in healthcare quote the World Health Organization's Health Report 2000. This study ranks the healthcare systems of the world, and the US comes out 37th, just below Costa Rica and just above Slovenia [pp. 152-156; all page numbers are from the WHO study]. Reid, in his book "The Healing of America," refers favorably to the study many times [it appears 12 times in the index]. Many people seem to think the WHO measurements are a straightforward, objective way of ranking healthcare systems, but there are numerous difficulties:

There are actually two different rankings of interest, the Overall Attainment (OA) ranking and the Overall Performance (OP) ranking. The OA ranking attempts to measure actual outcomes, while the OP attempts to adjust it relative to costs and literacy. On the OA measure, the US is 15, Slovenia 29, and Costa Rica is 45. The US is right between Germany at 14 and France at 16. So in terms of the WHO's own attempt to measure the quality of the system, rather than quality vs. cost, we're already in with the socialized medicine countries.
The OA measure itself is deeply flawed. It consists of a combination of four pieces:
1. Overall population health. This is at least a plausible thing to measure. However, they measure it using disability-adjusted life expectancy, which I've previously demonstrated has a tenuous relationship to the healthcare system. The US ranks 24th on this metric.
2. Health inequality. This is measured based on child mortality data, which I've previously shown is complicated, difficult to compare, and not especially indicative. They want to measure this by distribution of DALE, but they don't have good data for enough countries [p. 30]; so this metric is really infant mortality and life expectancy again. But that's not the real issue here, which is the weirdness of using health inequality as a metric at all, after you've used overall health level. If England made a change to their system that caused half the population to live five years longer, and no change in the other half, the country would look worse by this metric. I find that bizarre. A metric like "What is the absolute performance of the bottom tenth percentile" might be reasonable, but this metric is not that at all. The US is 32nd by this measure, which is rephrased as "equality of child survival" in the tables [p. 177].
3. Responsiveness. This is an aggregate of "respect for dignity, confidentiality, autonomy, prompt attention, quality of amenities, access to social support networks, and choice of provider" [p. 32]. I think this is at least a reasonable metric; it attempts to measure the quality of people's interaction with the system, independently of outcomes. It's a little tricky to find details, but it seems this was measured by survey data from a "network of informants" [p. 33]. The US ranks 1st on this metric [p. 184].
4. Fair financing. This is a metric of the distribution of percentage of income spent on health by the poor and the rich, with less variance being better. Whether this should be included or not is a political concern. However, I will point out that if you want to argue that the US system will make people healthier and have better outcomes if we make it distributionally more fair, using this metric to rank the systems in advance is circular reasoning. The US is 54th on this metric [p. 189].
The data is noisy. The WHO published 80% confidence intervals, and at this level of confidence, we can only say that the US's OA rank is between 7 and 24 [p. 196].
The adjustment from OA to OP is an attempt to correct for how much we pay rather than how much we get. It is statistically challenging [p. 56], and this Science policy forum article argues that the WHO's methodology of basing this metric off of only life expectancy and literacy is problematic. The WHO themselves state that "The object is not to explain what each country or health system has attained, so much as to form an estimate as to what should be possible." I find the WHO's approach quite suspect.

So what does the WHO study really say? It says the US isn't doing that well in DALE; that makes sense given our high homicide and accident rate and genetic and cultural makeup. It says the US isn't doing that well in terms of infant mortality; that makes sense given our high preterm birthrate. The US is number 1 in terms of responsiveness. The US ranks only 54th in fairness, but unless you care about fairness for its own sake rather than for its relation to health outcomes, this doesn't tell you anything. Additionally, the confidence intervals are large enough to affect the rankings substantially.

If you feel the WHO study is a good argument for massive federal intervention in healthcare delivery, why? Is improving our rankings by these metrics important? Do you think you can improve outcomes? Our outcomes are already pretty good. Do you think you can reduce costs? The bill we've just passed costs about a trillion dollars in the first ten years, and more after that. In Massachusetts, which is the closest state-level example, costs have been much higher than expected. What's going on here? I just don't get it.

Izmirian's first proposal.

2010-05-16T08:01:00.001-07:00

Izmirian commented with a simple version of his own health care plan.

Everyone would pay for health-care expenses out of pocket. This would generate much more cost awareness of the health care provided and generally promote all the efficiency benefits that you get from a free-market. Once your annual costs had reached 10% of your income the government would pay the rest of your health care for that year. To retain some of the cost awareness even after that point had been reached, a certain fraction of your costs past the 10% threshold would accumulate in a bill that you would have to pay back in subsequent years.

My initial reaction is that this system is a lot better than what we have now in the US, which in turn is better still than what we're going to have in ten years. There are still substantial problems.

It gets some things very right. For most people, most of the time, most healthcare expenses would be out of pocket. This would encourage people to shop on price and service, and would encourage providers to compete. The taxation limit of 10% is distortionary, but not heavily so; the lack of a sharp cutoff means that the incentive to not work and earn is fairly low. The requirement that money beyond the 10% threshold accumulates in a bill removes some of the incentive to strongly manage your income across years.

On the other hand, I think there are difficulties. The idea of a "10% of income" requirement seems designed to have the rich pay more than the poor before government expenses kick in, but this leads to tricky tradeoffs between income and wealth. These issues will be especially acute when they are related to expensive care at the end of life, where there is no prospect of future income to pay off the bill. If the bill is not passed on to your estate, then this will strongly shift incentives towards expensive end of life treatments, paid for by the taxpayer. If the bill is passed on to your estate, we are now taxing wealth instead of income, which leads to a large and expensive monitoring regime.

Your proposal doesn't talk about what treatments are covered, how treatment is provided, or spending limits. Based on the idea of a bill that will in theory be paid back at least in part, I think the proposal is that people could decide for themselves what treatments they wanted, and that treatment would be provided without government price controls. So far so good. But spending limits are a problem. I think we have to recognize that it is impossible to guarantee the most expensive and newest and best treatments to everyone, and that the best bang for our buck comes from subsidizing access to moderate cost treatments that have been proven effective.

Thinking about it more, a variant of the proposal with an overall spending cap per year that isn't way out in the tails of spending [say $25K-50K/year, although we'd have to look at stats to argue about the number more], and with the provision that people could buy whatever supplementary insurance they wanted to in a purely voluntary market, would not be a horrible system. Medical expenses would be somewhat higher in total than in a purely open system, taxes would be higher, there might be other unexpected consequences, but at first glance, I think this would work at least OK.

I will note in passing that the European systems that work relatively well [e.g. France] do feature substantial cost-sharing; I will discuss in detail in later posts. I will also note that the bill we are getting in the US is moving us much farther from the market rather than closer to it.

Adverse Selection

2010-05-14T06:00:00.001-07:00

I've been working a bit on a long post on "How many Americans are dying each year because they lack of health insurance?" but it's turned out to be quite complicated, and I'm bogged down reading original research papers. I'm also very busy with a couple of other things, so it might be a while before I finish that post.

In the meantime, I'd like to write about adverse selection. Adverse selection is the idea that in the presence of asymmetric information, markets can fail. One common example is the market for used cars. Suppose the person selling you a used car knows if it's a lemon, but you don't; because of this, you will pay less for the car than you would otherwise. This will cause people with good cars do be less willing to sell their cars, which in turn will mean more cars in the market will be lemons, which in turn means you will pay even less, and so the market fails in a vicious cycle.

In health insurance, the analogous idea is that you know if you're a good risk, but the insurance company doesn't. Therefore, the insurance company must charge you a higher price than they would otherwise, to make up for the fact that you might already know that you're going to a need a lot of healthcare. In turn, because the price is higher, you will be more likely to wait until you know you'll need a lot of healthcare before buying insurance, and so on. This problem is greatly exacerbated if the insurance company is forced to sell a policy to anyone who wants one, and it's even made even worse if the company is forced to sell to anyone who wants one at a regulated price. The problem of adverse selection is frequently used to justify a very heavily regulated market, where insurance companies are forced to sell to all comers at prices that don't depend strongly on their current health.

Is this solution necessary? Well, the lynchpin of the whole argument is that the insurance company has to not be able to know if you're a good risk or not. Is this an essential component of the situation? In some cases, market forces will rewards those who come up with solutions. For instance, for used cars, it is now easy to buy used cars that are dealer certified and come with warrantees; you are able to know you are not buying a lemon, you are willing to pay more for it, the market functions, problem solved. For health insurance, there are many ways that health insurance companies can reduce information asymmetry &mdash notably requiring health tests before they sell you insurance and rescinding your policy if it becomes clear you've lied to them about your health.

Therefore, I don't believe that adverse selection is necessarily a first-order problem in the health insurance market. Instead, heavy government regulation of health insurance is about government control and massive subsidies. [The subsidies are massive: under the recently passed health bill, a family of 4 making up to $88,000 is eligible for subsidies. We are not just subsidizing the very poor here.] It is about massive subsidies from the young to the old, even though the old are on average much wealthier.

As usual, I believe that to whatever extent we believe redistribution is a critical element of healthcare in this country, the right way to do it is via direct cash transfers. Although I believe those transfers should be small or ideally nonexistant. I'm not strongly opposed to transfers to the poor to pay for basic care, but I think we're fooling ourselves if we think that a system that offers massive subsidies to the middle class is a good idea; we're taxing ourselves to pay ourselves, inefficiently pushing the money through a government bureaucracy.

Here's a post by Bryan Caplan with a better written version of some of this same material. And here's another.

Insurance Innovations

2010-05-10T05:08:00.000-07:00

In my previous post, I argued that insurance should be purchased privately in a free market. Many of the comments seemed to rely on arguments that basically assumed that the insurance purchased would have essentially the same characteristics and properties as today's insurance, and many of the rest were based on a "folk economics" view of insurance companies that assumes that under a free market system, insurance companies would behave the same way they do today.

First, I agree that insurance companies want to make money. I think that is a good thing. But business models are not fixed, as we'll see.

To me, the primary issue with insurance right now is that you are not a customer of your insurance company in a very meaningful sense. You get your insurance [in enough cases that this basically determines how the market works] through your employer, and you are not choosing an insurance company. Your employer has some interest in choosing an insurance company that satisfies you, to the extent that they would not like you to choose a different job because that other job offers better health insurance, but it is extremely tenuous for at least two reasons. The first is that the employer is trying to choose a plan that is good for its employees on average, rather than a plan that is good for you. The second, more important reason is that health insurance is one of a large number of factors in choosing a job, and your employer knows that the specific health insurance company will not be a decisive factor for most employees. So your relationship to your employer-sponsored insurance company is not as a customer, but as a cost. [I will actually try to talk in detail about the numerous problems of the current insurance system in a separate post; there are additional regulations that make things even worse.]

This serves as a natural explanation for many of the behaviors that people object to, most notably a seemingly random denial of claims, massive paperwork requirements for appeals, and high rates of recision. I agree that the insurance company has a strong incentive to avoid paying large health bills once you get really sick. But what I disagree with is that this is fundamentally different from an airline's strong incentive to avoid flying you somewhere after you've bought the ticket, or a car maker's strong incentive to avoid honoring their warrantees. If we are to move beyond the level of discussion of "greedy companies want to make money so they'll screw their customers whenever possible," we have to ask why carmakers frequently do honor their warrantees, and why airlines nearly always do actually fly you somewhere when you pay, and if we think those arguments don't apply to health insurance we must think clearly about why health insurance is so fundamentally different, rather than just saying "companies are greedy." Insurance companies are fundamentally no more or less greedy than other businesses.

Companies are trying to make money. Absolutely. In a free market, where customers have options and are making real choices, companies have to compete by offering value to customers. In the case of insurance, companies can compete not just on price and benefit, but also on customer service, reputation, and trust. You don't want to be hassled when you need to make claims, and you definitely don't want to face recision when you get very sick. If you had a choice, you would go with the company that offered these things.

Health insurance is admittedly a complex product. The challenges are that you pay for it for a very long time, it has negative expected value, and once you are asking the insurance company to honor "their end" of the contract, they face a large cost to do so. Under the current system, where you are not a customer, the cheapest and therefore most profitable way to deal with that is for them to try to avoid paying. In a free market system, a company needs to convince you that they're not going to screw you if you get sick, compared to that other company that also wants your business. Which brings me to the real topic of my post, which is innovation in insurance contracts.

A great way to convince me that you're not going to screw me is to sign a contract detailing how you're not going to screw me. So I brainstormed some things we might see in health insurance contracts under a free market system. I'm not saying we would see any one of these provisions in particular; certainly they wouldn't be required, and the market will likely dream up other better ones. But these are some ideas:

Late payment protection. Afraid of losing your insurance if a payment is one day late? Our contract says you can be up to three months behind, in case times get tough. Of course, you'll pay a small late free, but no need to worry about being one day late.

Hate paperwork? Under our contract, if a claim is denied for any reason, there'll be a 24-hour number where you can talk to a supervisor immediately. These supervisors will speak your language and have enough authority to resolve claims.

Afraid of recision? If you've been a customer of ours for more than 1 year, we agree not to cancel your contract for any reason other than discovery that you directly lied to us about the condition you're applying for treatment for [e.g., you already had diabetes and didn't tell us.] After 5 years we agree that we are not allowed to cancel your contract for any reason.

Afraid of huge payment increases later? This contract binds us to a schedule of maximum premium increases. We may have to renegotiate a rider later if technology makes expensive new treatments available and you want those covered as well.

Want to be able to switch companies later? We will hold X% of your premiums in an "escrow account". When we pay claims, we will take them out of this escrow account first. If you cancel your insurance at any time, we'll give you back the contents of the account. If you wish, you may transfer this money to your new insurance company, which should make them more interested in accepting you, even though you'll be older when asking for insurance.

And so on and so on. I'm not saying any particular company would offer any particular provision. I'm arguing that in a world where companies had to compete for your business, there are other incentives in play that strongly counterbalance the simple incentive to not deliver a service that you think you've bought.

[In fact, you do see some of these effects in some European countries with "nationalized" healthcare. There are some countries where private insurance companies compete on things like customer service. They make claims like "All claims processed and paid within 5 business days!" One of the things I'll be attempting to show in later posts is that the aspects of the European systems that work best are those that allow for the most competition.]

Why don't these things exist now in the US? Because right now, your insurance company doesn't have to convince you they won't screw you. You're not a customer, you're a cost.

Separation of Issues, And A First Proposal

2010-05-07T17:19:00.001-07:00

From thinking things through, it seems there are at least a couple of major issues here, and I'd like to separate them. People who are in favor of national systems of health care are really in favor of one or both of two different propositions.

The government should ensure that the poor can afford health care through some form of forced redistribution.

The government should take an active role in the provision and delivery of health care.

It is worth thinking about how you feel about these two things separately. For instance, I am opposed to both, but I think that [1] would be only a little bit bad, while I think that [2] is likely to have severe negative consequences for overall human welfare.

To give a bit more detail, under [1], everyone with income below a certain level would get money placed in a special account. They could use this account for any medical expense. They could buy primary preventive care, they could be catastrophic insurance in order to pool risk, etc. Just like the rest of us. The government would not mandate what procedures were and were not covered for who, and would not set prices for anyone.

If I were going to a propose a system, it would be some amount of [1], along with the government getting out of direct provision of care [Medicare, Medicaid, VA] and heavy regulation of insurance [no more corporate tax deductibility of insurance premiums, so your insurance would not be tied to your job, insurance could be sold across state lines, high deductible or "pay for performance" or any kind of plan the customer wanted would be allowed, insurance companies could charge different amounts to different people and don't have to take all comers, people are not required to buy insurance].

Questions for thought:

Would you find a system like this acceptable? Better or worse than the current system in the US?
Does your opinion change if the poor are ultimately able to decide for themselves whether something is or not a medical expense?

What would be a better system than this?

Life Expectancy, and Its Cousins

2010-05-06T04:44:00.000-07:00

We've all heard it. "The US spends twice as much money per capita as most other countries on healthcare, yet our life expectancy is lower." Here's a graph from 538.com showing the result. Reid, on p. 241, says:

"The world's richest country ranks forty-seventh... The United States is among the worst of the industrialized nations on this score... the quality of, and access to, national health care plays a large role in any country's average life expectancy; the fact that tens of millions of us don't have access to a doctor means Americans are dying of health problems that would probably be cured in any other developed country."

My goal in this post is to argue that the final sentence of this quote is false, and that in fact, the quality of and access to national health care plays a very small role in any country's average life expectancy.

An article from the National Center for Public Policy Research makes the point that for a statistic to be good for comparing healthcare systems across nations, it must (a) assume an actual interaction with the health care system, and (b) measure a phenomenon that the health care system can actually affect. For example, a statistic measuring rates of cancer survival satisfies both (a) and (b), a statistic measuring rates of cancer incidence only satisfies (a), and a statistic measuring homicide rates satisfies neither. How well does life expectancy do?

Well, the US has extremely high rates of both homicides and traffic accidents, and according to at least some studies, correcting for just these factors moves us up substantially in the rankings. More importantly, the weight of evidence supports the idea that a number of other issues including genetics, diet, lifestyle and culture play a much larger role in life expectancy than the healthcare system does. This study from Harvard examines life expectancy by race and demographics in some detail. While many people know that black Americans have much worse life expectancy than whites [even after controlling for the much higher accident and homicide rate], this can easily be explained by either genetic and cultural differences or the argument that blacks have worse access to healthcare. On the other hand, the results for Asian Americans are pretty striking. Comparing Asian Americans as a group to "middle America" [white Americans with a couple of identifiable large poor rural populations in the north and Appalachia removed], the middle America group has a much higher average income, a somewhat higher high school completion rate, better access to healthcare, and a much lower life expectancy. In fact, the Asian population in the US has a higher life expectancy than the Japanese population, which has the highest country-level life expectancy in the world. Comparing Mormon and non-Mormon whites in Utah, the Mormons had much higher life expectancies, even after controlling for tobacco usage; the study suggests that "may be due to factors associated with religious activity in general, such as better physical health, better social support, and healthier lifestyle behaviors. Religious activity may also have an independent protective effect against mortality." In this paper, researchers at the University of Pennsylvania's Population Studies Center examine these issues in detail, stating that

"We consider in greater depth mortality from prostate cancer and breast cancer, diseases for which effective methods of identification and treatment have been developed and where behavioral factors do not play a dominant role. We show that the US has had significantly faster declines in mortality from these two diseases than comparison countries. We conclude that the low longevity ranking of the United States is not likely to be a result of a poorly functioning health care system."

Reid himself mentions that life expectancy is a problematic metric, although he ignores the issues I mentioned above. Instead, he concentrates on the idea that quality of life is important, and suggests the DALE, or Disability-Adjusted Life Expectancy, which is a complicated formula developed by the World Health Organization that ranks people with various conditions. Reid states:

"Still, the U.S. ranking in DALE terms, twenty-fourth, is considerably higher than the forty-seventh place we scored on the simpler ranking of life expectancy at birth. That difference means that American health care is making people healthier &mdash at least, for those who have access to it. It's because we fail to provide access to regular health care for 45 million Americans that our overall rank for healthy life expectancy trails the rest of the developed world."

Although Reid frequently cites statistics and studies in his book, there is no reference or support for his simple assertion about why our DALE score falls where it is. Yet DALE is subject to all the same problems I discussed above. In fact, given the high rate of accidents and non-health mortality in the US, it is highly plausible that the US currently looks great in DALE at the country level. Other metrics, such as QALY and DALY, suffer from exactly the same problem.

If government provided healthcare strongly improved health outcomes, we should expect that the introduction of Medicare, offering government provided healthcare to the entire elderly population of the United States, should have had a huge effect. But NBER economists Finkelstein and McKnight studied the impact of the introduction of Medicare and found that

"the introduction of Medicare had no discernible impact on elderly mortality in its first ten years in operation. They present evidence suggesting instead that, prior to Medicare, elderly individuals with life- threatening, treatable health conditions (such as pneumonia) sought care even if they lacked insurance, as long as they had legal access to hospitals."

[Finkelstein is now at MIT; the paper is available free at her website.]

I welcome pointers to studies that take the other position here, but I have not yet found any studies that convincingly argue [rather than merely stating] that changes to our healthcare system could on their own have a large effect on country level life expectancy. The US is a diverse country, much more diverse than most of the European countries we're comparing against.

National health care is a large intervention in the economy. If you're going to use differences in statistics like life expectancy or infant mortality to argue for national health insurance, you should be able to present a compelling case that these statistics are strongly under the control of the healthcare system, and that a national healthcare system has a good chance of greatly improving these statistics. I believe that supporters of national health care use these statistics, but in a superficial and unjustified way. I haven't proved that national health care is "wrong", or that we shouldn't do it. But I hope I've shown that arguments talking about differences in life expectancy [or DALE's or QALY's] do not justify an enormous costly intervention in the economy, because there is little evidence that these measurements are strongly under the control of the healthcare system.

[Note: In the developed world, life expectancy is still increasing noticeably relative to the scale of differences between countries. Results from 2000 which are still floating around the web look different from results from 2006 or 2008. So if you want to use absolute numbers, be very careful about using the same dataset.]

My Healthcare Thoughts: Infant Mortality

2010-05-03T14:48:00.001-07:00

Most of the arguments I've seen for more government involvement in healthcare are outcome based. "Poor people without health insurance are dying because of preventable diseases, and government provided healthcare will stop this." Most seem willing to agree that if an expanded government system weren't going to improve health outcomes, there wouldn't be a reason to have it. And many even agree that if government involvement weren't going to improve health outcomes reasonably strongly relative to how much it cost and relative to other ways to improve human welfare, it wouldn't be a great idea. This post will be the first of several exploring to what extent people are dying in this country because of a lack of access to healthcare.

One of the primary metrics people like to mention is infant mortality. Quoting Reid [pp. 33-34]:

Perhaps the most tragic indicator of America's troubled health care system is the number of newborns who die each year. This statistic is called infant mortality, or neonatal death. It generally refers to babies who die within one year of birth. To me, that seems the most painful thing that could ever happen: The expectant parents go through the anticipation and sheer joy of watching their baby develop in the womb, enter the world, join the family — and then have to bury a tiny corpse a few weeks later. Surely any decent health care system would develop effective mechanisms to avoid neonatal death. But out of twenty-three wealthy countries, the American health care system ranks dead last when it comes to keeping newborns alive. Our rate of infant mortality is more than twice as high as the rate in the top-ranked countries, Sweden and Japan. A key reason, as we'll see in later chapters, is that other rich countries offer free prenatal and neonatal care for every mother and every baby.

According to wikipedia, the baseline infant mortality rate is 6.3 [out of 1000] in the United States, making us 33rd in the world in infant mortality. Switzerland, Germany, France and Belgium all fall in the range of 4.1-4.3, with the UK coming in at 4.8. Japan, Sweden and Norway are at 3.2-3.3. [Throughout these posts, I will tend to ignore statistics from very small countries, such as Iceland or Singapore.]

It is often stated or implied that the high US infant mortality rate is a consequence of our unequal healthcare system, and that many of these deaths could be prevented with better prenatal and postnatal care. If premature births or infant deaths could be easily prevented by better access to moderately priced government-provided healthcare or nutrition, we might expect the infant mortality rate for people on Medicaid and WIC to be roughly in line with those of the European countries we seek to emulate. Medicaid has been in operation since the 1970's, and approximately 40% of US births are financed by Medicaid. While some studies do show a positive effect of Medicaid and WIC on infant mortality, I have not found any data indicating that people receiving prenatal and postnatal care via Medicaid and WIC experience Western European-style infant mortality rates. I would welcome such data if you have it, but I would think if it existed it would be widely trumpeted. In fact, this Minnesota Department of HHS study seems to indicate that the overall infant mortality rate for Medicaid births is substantially higher than for non-Medicaid.

If you believe that expansion of government provided or financed healthcare programs would reduce infant mortality rates to Western European levels, it seems you must believe one or more of the following:

People on Medicaid have much lower infant mortality rates than the average US rate. I have seen no data that supports this.

The current Medicaid system somehow does not constitute "access to care," and an expanded system would somehow deliver care to everyone that is much better than what Medicaid delivers now. This would essentially be an admission that Medicaid doesn't do a good job of providing care; while this may be possible, if true, it hardly argues for further government involvement. Medicaid has had 30 years to work on this.

A large fraction of total births are currently to people who are uninsured AND these people have very high infant mortality rates AND these people, when given government coverage, would have infant mortality rates substantially lower than the current Medicaid population. I cannot rule this out based on the numbers alone, since I cannot find data on the fraction of uninsured births or their infant mortality rates, but the conjunction seems highly unlikely to me.

If lack of access to healthcare is not the major cause of our high infant mortality rate, what is the cause? For starters, comparing infant mortality rates can be nontrivial, as some countries [including France] "do not report all live births of babies under 500 g and/or 22 weeks of gestation." This could lead to substantial differences in reported infant mortality rates. Even more important is the confounding issue of premature births. The US has a very high rate of premature births, and premature babies are much more likely to die. According to the CDC, "If the US had Sweden's distribution of births by gestational age ... the US infant mortality rate would be one-third lower." One third lower would put us smack in the middle of a pack of large European countries with national health care. So if you really want to get the infant mortality rate down, you have to either greatly increase the survival rate for premature babies [which is already relatively good in the US] or greatly decrease the fraction of premature babies.

The causes of preterm birth are not well-understood. There seems to be a strong correlation with factors such as lifestyle and socioeconomic status well before pregnancy, such as very young or very old mothers, unmarried mothers, and mothers who have health problems . Fertility drugs can also play a role. Note that if access to healthcare by itself prevented premature births, we would see fewer premature births and better infant mortality for Medicaid births, which we do not.

Summary: While infant mortality is a tragedy, and the US infant mortality rate is in fact very high, the discrepancy between the US and European countries can be largely explained by a difference in premature birth rates, which are likely in large part caused by demographic and behavioral factors that are not under the direct control of the healthcare system. Given that we have a large population that already has access to healthcare via Medicaid and WIC, and that this population does not have European-style infant mortality rates, it is difficult to conclude that an expansion of government healthcare to cover the rest of the population would on its own lead to a large drop in infant mortality.

[Side note: In the Reid quote, he says that we'll see "in later chapters" that a key reason for other countries' lower infant mortality rates are that they "offer free prenatal and neonatal care for every mother and every baby." I read the whole book, but I somehow missed this. If you have the book and see anything that looks like an actual coherent argument for this, I'd be happy to hear about it.]

My Healthcare Thoughts: An Introduction

2010-05-02T12:53:00.000-07:00

Prelude: A few weeks back, I went to a walk-in clinic at MGH, saw a doctor for about ten minutes, had some basic tests run, and was astonished to later learn that the "official" bill for this service was $700 [although my copay was only $10, thanks to my gold-plated insurance plan]. I somewhat naively suggested that under a more free-market system where consumers were exposed to real prices, cost would drop and quality would increase. A firestorm of comments ensued, including some lengthy discussions. At a friend's urging, I obtained and read T. R. Reid's The Healing of America: A Global Quest for Better, Cheaper, and Fairer Healthcare. I took extensive notes, and researched various topics further on my own. I believe I am somewhat better informed than I was a few weeks ago, and I would like to share my thoughts.

In this post, I'll give a brief overview of my current position, which I'll be attempting to support from various angles throughout this series. My posts will touch on philosophy, economics, data analysis, statistics, and who knows what else. If you think my arguments are weak, call me on them. If you think my data's bad, point it out. Hopefully we can all learn something.

My current thoughts on health care policy:

I am a strong supporter of free market reforms. I believe a free market approach would be morally just, and would also lead to high quality, low costs, and wide availability of care.

The current US system is not a free market system. It is heavily regulated in some very ugly ways. In many ways, our regulations are worse than those in many European countries.

There are some European countries, notably France, Germany, and Switzerland, about which I feel a reasonable case can be made that their systems are better than the current US system. I don't actually think they are better, but I do think a case could be made. There are a number of other countries with nationalized universal health care, notably Britain, Canada, Italy, Spain and Portugal, whose systems seem obviously worse to me. In particular, I strongly disagree with statements by Reid and others that "Every other rich industrialized country obviously does health care better and cheaper than us."

My ideal system would be one in which all health care was paid for privately and in which costs were set by the providers. Individuals would tend to pay for routine care out of pocket, and for large unexpected expenses via high-deductible [and therefore low-cost] insurance. If we as a society wanted to ensure healthcare for the poor, we would do it via tax credits. I feel strongly that this system would be better than any currently existing system on the planet.

My thoughts on Reid's book are that are there a few bright spots, but in general it is a very bad book. It is filled with arguments that are specious or incoherent or innumerate, and frequently selectively interprets or omits data in a way that supports its overall thesis that nationalizing healthcare is a good idea. However, it is an interesting book, in that it states in one place most of the traditional liberal arguments I've heard for nationalizing healthcare, as well as the data backing up these arguments. I will therefore address many of its arguments specifically, not because I care about attacking Reid in particular, but because I want to point what I view as the flaws in these lines of reasoning.

Day of the Dead Costume, Take One

2010-01-09T14:40:00.001-08:00

This was meant to be my Halloween costume, but I didn't get it together in time. The idea was inspired by Mexican Day of the Dead imagery, and also by Grim Fandango. I wore it to an early December cabaret party, and had a great time. Compared to my original conception, it had fewer, larger flowers, but I think it works well. Costume work by Jenn Martinez, makeup by Anna. I'll definitely be wearing this again, possibly with some modifications.

An Open Letter To JetBlue: How To Improve Your Customer Experience At Little To No Cost

2009-07-25T09:25:00.000-07:00

Dear JetBlue,

I am a moderately frequent flyer, taking 4-7 cross country trips a year plus an international trip or two. I frequently fly your airline. Yesterday I flew from San Francisco to Boston, by way of New York. It was a hard day. Some of the problems obviously had nothing to do with you. I do not blame you for the fog in Long Beach that delayed my plane arriving in San Francisco, which in turn delayed the departure. And I do not blame you that I (just) missed my JFK connection, requiring me to take a later flight. And I do not even blame you that the later JFK flight was also delayed, causing me to get home nearly 3 hours later on an already long day.

However, there are a number of things you could have done better. These are not the expensive things people would admittedly love, like bigger seat pitch, more direct flights on the schedule, and weather control satellites. These are instead fairly simple things that would cost you nearly nothing to implement, yet would have made my and many other people's experience much better.

The first point contains your electronic display of (expected) departure times. I believe that keeping this display somewhat more updated and more accurate would go a long way to making things better. Yesterday, I experienced two different delayed flights --- one in SF, and another at JFK. In both cases, the display was updated to reflect an initial delay, but this initial update was both substantially too optimistic and was left on the board even when it was obviously not true. You may not know for certain what time my flight is going to leave, but if it's 12:45 now and I'm still sitting at the gate and the plane I'm going to get on hasn't arrived yet, surely you can come up with a better guess than 12:15? Yes, it is frustrating when a predicted delay turns into even more delay, but you do not make the situation better by pretending it's not happening. Just update the board!

The second point contains knowledge of flight times by people working at the check-in counter. My original plan for yesterday was to catch an 11:00 a.m. flight out of SFO, "officially" arriving at JFK at 8:10, and then a 9:15 from JFK to Boston. My breakdown of this is that we're going to spend five hours in the air, twenty minutes on each end taxiing, and if all goes well we'll actually get to JFK around 7:40. At 12:45 and still sitting at the gate, I asked the gate worker whether there was any information about being able to make my connection. She said that I would definitely make it, that the flight was *still* going to land at 8:10 because "It's much shorter going west to east than east to west and they can make up time in the air." This is simply not plausible. If I am in SFO and my plane has not started boarding at 12:45, there is no chance I'm going to be deplaned and on the ground at JFK in four hours and twenty-five minutes unless they can swap in an experimental fighter jet. I fly ten times a year, and I know this! Why would a professional gate attendant not know this? It might have been possible at that point that I would still make my connection, but it was not possible that I was going to be at JFK at 8:10.

The third point contains inconsistent handling of delay-related information and connection information by the flight crew. The flight crew was certainly aware that there were many people on this flight attempting to make tight connections. I know this because at least one of the gate agents, who had been talking to us, was on the flight. People trying to make tight connections have a couple of special needs, beyond everyone's need to get where they're going as safely and quickly as possible. One is to have accurate information about when the flight is going to land. In this case, I feel that you failed. At about 8 pm EST, the pilot stated that we would be on the ground at 9 pm. As it turned out, we landed at 9:15 and it was nearly 9:30 by the time we got to the gate. The second and even more important thing is information about and help with connections. In this case, everyone connecting to Boston and Buffalo missed their connections. The appropriate thing to do would have been for a flight attendant to say "Passengers travelling to Boston and Buffalo, we're sorry but you've missed your connection. There is a later flight, and we are automatically rebooking you. When we arrive, please go to LOCATION to receive your new boarding pass. We apologize for this, but we didn't feel we could further inconvenience the passengers on your connecting flight by holding it." Instead, absolute silence from the flight attendants (and we can't get up to go ask them while taxiing, of course), while we sit there stewing, wondering what's going on. This is really not good enough!

I have an iPhone, so I attempted to use that while taxiing to check the status of my flight. The status was unavailable. All the other previous JFK to Boston flights were there, and mine was simply missing from the page. It might have been better if I'd had the flight number for the next flight handy, but your mobile website did not make my life easy, and as far as I can tell was actively broken, missing the most important piece of information. Please make a better mobile website.

When I arrived at the gate, I had to go to the information screens to figure out if my flight had left. Since it wasn't being shown on the screen at all, I figured it had. I couldn't immediately figure out where to go, but was eventually pointed to a "Just Ask" desk. This desk was staffed by a single woman who was in all honesty somewhat hostile. I think she knew she was in for a hard time and so immediately switched into "I'm trying to help you, I'm yelling at you to stop yelling at me" mode even though I wasn't yelling. In fairness, the woman right next to me who showed up a couple minutes later was yelling. Overall, I don't blame this woman, but there was a real problem here. You had enough information to know there were 20 people on that plane who had missed connections. Why did you not have some simple plan in place for dealing with us? What do you think people who have been flying delayed all day and are trying to get somewhere are going to do? People are going to get frustrated, and start panicking and getting angry. You can avoid this. I note in passing that I later walked down a different arm of the terminal and saw another JetBlue desk with three staffers sitting at it doing nothing, so it's not clear it's a total manpower problem. I have had this experience go much better on other airlines.

I think the central themes of these ideas are that people would rather hear the truth about delays, presented politely, then be told something that is obviously not true or that will clearly be revealed to be untrue in a short amount of time, and that we would like you to help us manage our connecting flights in a reasonable way.

I recognize that JetBlue is a relatively small airline. I recognize that I will sometimes have to take a connecting flight to get where I want, because you have fewer direct flights than the competition, and that weather will cause unavoidable delays. There's a lot I like about your airline, and I hope to keep flying with you. However, today's handling of the connection experience will make me strongly consider using a competing airline with direct service in preference to taking your connecting flight again. Your poor handling of the situation added substantial additional uncertainty and frustration to an already difficult experience.

Sincerely,

rif

ps. The Map channel. I love that you have a map channel. However, there is no good reason why your map channel should show 30+ seconds of ads for every 10 seconds of maps. Particularly given that you have only about six ads in rotation. I don't frequently see the map channel used, but to a small segment of your customers (at least me), it represents a major value add. I am a nervous flyer, and seeing where we are calms me down. However, seeing the same ads over and over again for most of the time I want to see the map is incredibly irritating. What other television station could survive if 3/4 of the content were ads? Do you really think I'm going to order a glass of wine because I've seen the 100th ad about your brand of Zinfandel? Do you really think I'm going to acquire a pet and use your jetpaws service, or that I had no other plausible to find out about it? To see this service done right, please take any Virgin America flight. They not only have a map with no ads, they let you control the scale. Please improve your map channel. I know not many passengers care about it, but those ads you're showing now can't be worth much.

2008 Themes and Reflections

2009-01-05T17:26:00.000-08:00

2008 was on the whole a positive year for me, although perhaps not as positive as most.

Themes for 2008:

I'm not really a "software engineer." Software engineer is my current job title. I think Google is a great place to work, and I might make it as a software engineer at Google, but I'm much more of a mixed bag. In general, I'm happiest and most productive when I'm either working on a hard, chewy problem with some research-y aspects, or working at the strategic level, trying to figure out the right plan. When it comes time to grind out lots of fairly straightforward, bulletproof, heavily-tested code that implements the design, I'm not nearly as strong. I'm still pretty good, it's just out of my sweet spot. I wasn't sure about this before I came, and now that I've worked as a software engineer for 1 1/2 years, I have a lot more clarity on it.

Trying and failing is not bad. This is really huge for me. In some relatively trivial areas of my life, I have no trouble taking risks: I'll get a crazy haircut or a piercing without thinking too hard about it. But I think that a big problem with my career has been my unwillingness to take risks, to say "I'm going to spend 6 months to a year trying this, even though it really might not work." When I was in research, I tended to favor projects that were more analysis, or pointing out flaws in previous work, or very small incremental improvements using known tools. At Google, I've let myself fall into projects that I knew were doable, that were "just work" and didn't really excite me, because they were "safe." I dream big, and to achieve big I have to take more risks. Many of those risks will fail. That's OK.

Meditation and Buddhist ideas. I've spent a lot of time this year working with mindfulness, meditation, trying not to get too attached to my thoughts and emotions, learning not to get upset by the actions of others. I'll have a lot more to say about this in future posts, but I feel like this year has been the beginning of something important for me, a sort of recognizing of something I'd always known but never been aware of.

Things I have made progress on:

Being open and non-judgemental about people. If someone does something that hurts me, at work or personally, it is much better to pragmatically figure out how to improve the situation then to decide they are bad people who need to be punished. Getting angry or upset doesn't help anything.

Making new friends. I've been more open to meeting and spending time with new people this year, more willing to put myself out there. It hasn't always worked perfectly, I've been rejected some and hurt some, but on the whole I'm happy with what I've done and the choices I've made. I think that being honest and open, even though it makes me a little more vulnerable, is definitely the right stance.

Focussing. When I worked at Honda, there was noone else in the office most of the time, and I often slacked off. At Google, I am surrounded by smart friendly reasonably hard-working people, getting good stuff done. This has done wonders for me. For most of this year I worked on a project I didn't much care for, and while I can't say I enjoyed it, I can say I was able to work my way through it and I'm glad. Community works wonders.

Non-theme for 2008: Worrying about money. I feel pretty good about both how much money I have and my attitude towards it. This is a real improvement from some previous years.

Things I am most grateful for in 2008: My ridiculously awesome wife and best friend Anna, who continues to amaze me. My friends and family. Music, dancing, yoga, math, computer programming, good food and wine, dark chocolate. Oh, everything really.

Resolution for 2009: Spend more time playing music and deliberately practicing to improve. I have other goals, like spending more time dancing, losing weight, exercising, getting some good stuff done at work, making new friends and keeping the old, but if I do just OK on those and make much better music, I'll be more than satisfied.

Gladwell on Teaching: Some Good Ideas, Some Silly Analogies

2008-12-09T16:33:00.000-08:00

Malcolm Gladwell has a piece in the latest New Yorker on what's wrong with our school system, and how to fix it.

Gladwell starts by discussing college scouting, and how difficult it is to predict how well the top college quarterbacks will do in the NFL. For quarterbacks, the NFL game is qualitatively different from the college game: much faster, much more dynamic, much more athletic. The only way to tell for certain that they'll be great NFL quarterbacks is to see how they play in the NFL. Gladwell goes on to suggest that selecting teachers faces similar problems — that we cannot hope to predict in advance how good teachers are going to be. As a secondary example, he uses a financial advising firm, who interviewed 1,000 candidates, offered internships to 49, jobs to 23 of those, and expected roughly half of them to end up making it. In the case of the advising company, it took three to four years of sustained observation to find the actual performers.

Gladwell also makes some points about dynamic range in teaching. If a median teacher covers one year of material in one year (by definition, I suppose), there exist many bad teachers who cover only half a year of material and many good ones who cover a year and a half of material. Teacher effects are a very strong predictor: on average, moving from a 50th to 85th percentile quality teacher is about as good as cutting class size in half. According to one study, if we replaced the bottom 6-10 percent of teachers with median quality teachers, we could close the education gap we have with a bunch of other better performing countries.

Gladwell's suggestions are that we make it much easier to get into the teaching business, eliminating things like required masters degrees which are expensive and have little correlation with performance. He suggests eliminating easy and automatic tenure, rigorously evaluating performers, and paying good performers a lot more.

These suggestions are all fine as far as they go, but Gladwell kind of goes off the rails because he fails to understand some key differences between teaching, football and financial advising. Football has very strong zero-sum and winner-take-all properties: there are a a total of 267 victories each season to divide among 32 teams, and only one team can win the Superbowl. The difference between a world-best performer and an almost-but-not-quite-world-best performer is therefore enormous. It's critical to constantly evaluate whether your quarterback is the best possible rather than merely very very good — the merely very good ones are actually "failures." Financial advising isn't quite as zero-sum, but I posit that there are many more people who want to be financial advisers than there is a need for financial advisers.

Teaching, on the other hand, is a very different situation. Gladwell himself notes that merely replacing the worst teachers with median quality teachers would be an enormous win. And we don't need just a few good teachers, we need a giant mob of them.

I am a firm believer that in nearly all cases, practice and effort can improve skills. It's true for football, it's true for financial advising, and it's true for teaching. In the case of teaching, Gladwell makes a good argument that we have identified some of the skills involved: how good is the teacher at giving rapidly iterated useful feedback, how often is the teacher interacting with the group. Everyone who wants to get better at anything needs to practice and put in effort.

But Gladwell thinks we need to hire four teachers for everyone one we "end up" with, weeding them out and washing them out like financial advisers or football players. I disagree. We need more focus on the skills that make teachers successful, and more incentives for teachers to learn and practice those skills. But we're not trying to find the 30 best teachers, we're trying to teach a whole nation of kids. We're trying to create a situation where every teacher is as good as the 85th percentile teacher is today. Instead of taking a stance that teaching is something that only a chosen lucky few could ever do, let's assume that nearly all teachers can get better with appropriate practice, and that many of them can probably become good enough with the right incentives.

Analogy from my personal life: I am an amateur musician. If my goal is to become a professional classical concert pianist, the appropriate course of action (other than, at my age, seeking psychotherapy) is to spend every waking moment practicing. There is very limited demand for my skills, and I have to be not only very very good, but actually better than almost anyone else. On the other hand, if I want to become a very good piano player, it is sufficient to practice for a couple hours a day, especially if I practice in good ways. I think becoming a teacher is a lot more like becoming a good piano player than a professional concert pianist, and I think our policies should reflect this. So while I agree with Gladwell about relaxing masters degree requirements, rigorously evaluating performance, and increasing pay for performance, I don't think the analogies with quarterbacks or financial advising are very strong.

(Aside: I am no football expert, but Gladwell could check his facts better. One of the lynchpins of his football argument is that colleges generally run the shotgun spread and the NFL doesn't. However, Tuesday Morning Quarterback suggests otherwise.)

(Update: Anna says that attacking Malcolm Gladwell is kind of beating up a strawman.)

Book Review: Talent is Overrated

2008-11-25T16:55:00.000-08:00

"Talent Is Overrated: What Really Separates World-Class Performers from Everybody Else" by Geoff Colvin

A short, fast, enjoyable read. The basic arguments of the book, all of which I found more or less compelling, if not entirely groundbreaking:

There is very little evidence of "innate talent" in most fields, such as chess, music, mathematics, or business. (In sports, there is obviously some influence of body type; the book argues that this is less profound than you might think, although it's obviously important.)

The way to get good at something is to practice it. The way to get very good is to practice it a lot.

"General intelligence" is not well correlated with being world-class.

The kind of practice that actually helps is "deliberate practice", which is practice that focusses on specifics, has short direct feedback cycles, and is at the boundaries of current performance. Deliberate practice is difficult and mentally demanding. Teachers are often crucial in designing deliberate practice regimes, especially in the early stages.

Many so-called "geniuses" (Mozart and Tiger Woods are studied in some detail) who have exhibited astonishing skill at what seem to be astonishingly young ages are often touted as having innate skills, but they are in fact the product of precisely this sort of deliberate practice.

It takes about ten years and 10,000 hours of deliberate practice to become world-class. This is a rule of thumb. There is evidence that in the current world, it takes longer than that in the sciences.

In some fields such as music or athletics, it is very important to start young, because the effect of practice compounds over time and once adolesence ends, other demands have a very strong tendency to interfere if the practice isn't central to someone's life. A study examined music students in three different departments: one in which the students could go on to become international soloists, a second in which they went on to become orchestra players, and a third in which they went on to become teachers. The only variables that had any real explanatory power for dividing these groups were how many hours they practiced, and how many hours they had practiced before turning 18.

In business it is less important to start young, primarily because very few people are trained to be business leaders from an early age. The book briefly examines the wisdom of this, and suggests that this is likely to start happening more.

At any stage of life and for nearly any skill, deliberate practice is a key to increasing performance.

Most organizations and businesses are very bad at encouraging deliberate practice. (In general, I found the section on organizations to be the weakest, although I will have to think about it some more).

To become world-class, achievement has to be primarily intrinsically rather than extrinsically motivated. However, extrinsic motivation is frequently a good way to start children on a path --- they'll either get into it or they won't. Additionally, the line between extrinsic and intrinsic motivations is blurry: specific, constructive feedback is both at once, and many child prodigies start out of a desire to spend time with or imitate their parents.

To become world-class takes a huge amount of sacrifice, and there's nothing intrinsically right or wrong about making this sacrifice. It's a cost/benefit tradeoff. In general, the older you are, the greater the costs relative to the benefits.

I'm still digesting the implications for my own life. I think my practice often tends to by nature by somewhat deliberate, although I could (and will try to) apply this principle more consistently.

Going Nondark

2008-11-18T21:49:00.000-08:00

As many of you know, I joined Google as a software engineer in May 2007. I have worked on two major projects at Google, both of which have recently been announced.

When I first arrived at Google, I was assigned to the Book Search team. One of their major efforts was a top-secret plan to settle the class-action lawsuit that the book. This settlement was made public on October 28th 2008. One piece of the settlement is that Google will begin selling online access to a large collection of books --- basically books that are out-of-print but still in copyright --- and sharing revenue on those books with publishers. These books are in most cases long out of print, and will have to be priced somehow. My work was a first pass at an algorithmic pricing model for the books. In a nutshell, I took a bunch of features of the book and attempted to predict what the price of that book would be were the book currently being sold in bookstores. We presented this model to the publishers in December 2007, and while it was definitely a first pass, it made a convincing case that algorithmic publishing was feasible, that by and large we could figure out which books we could charge more and less for.

Since December, I have been spending most of my time working on something that turned into The LIFE photo archive hosted by Google. My responsibilities were mainly on the backend, receiving and processing data from TimeLife and getting it ready for serving. I also had a pretty big hand in product definition.

It's nice to be able to talk a little more about what I'm doing, and to be able to see the fruits of my labors.

Not Entirely Unlike Obama

2008-10-13T21:08:00.001-07:00

Done from a photo Anna had from a few months ago.

The good: I feel like I got the proportions on the clothes pretty OK. The jacket and tie are reasonably effective, although the sleeve of the jacket is definitely too dark.

The bad: The face is too long, not round enough. My shading sucks. I have no idea what an ear looks like.

Notes: It actually looks a little better on screen than it does on the page. Also, there was a time before I started shading it when the features looked better than they do now.

2008-10-12T19:36:00.000-07:00

Done with a uni-ball Vision Exact. This is taken off an image from the photobook "Night Vision: The Art of Urban Exploration" by Troy Paiva, which is a pretty awesome book. My sketch doesn't do any justice to the haunting nature of the original.

A General Theory of Love, by Lewis, Amini, and Lannon

2008-10-10T08:54:00.000-07:00

I've been thinking about love recently and wanted to learn more. I read this book last night. Well, actually, Anna told me it had to go back to the library today, so I decided to read the first chapter to see if I wanted to order it again, and then I got a bit sucked in, and then I decided it didn't require full reading but only skimming, and then I was done skimming it.

I was hoping that the book would be a "General Theory" a la Einstein's "General Theory", but it ended up feeling more like "Unspecific, Vague Theories of Love." I was also pretty familiar with a lot of the material. I can't really recommend this book, but I'll quickly summarize the key points:

In some sense, we have three brains: the reptile brain, the limbic brain, and the neocortex. All mammals have at least some limbic brain. Higher mammals have more limbic brain and at least some neocortex. Abstract reasoning and logical thought are all about the neocortex. Love and emotions come more from the limbic brain. Thus they cannot be (fully) controlled rationally.
Neural networks are rad. I don't necessarily agree with this point or understand why the book was making it.
Traditional psychoanalyst theories (Freud etc.) are not valuable. The authors clearly had an axe to grind here. I have no dog in it.
We fall in love because we're mammals, and mammals need affection and close connection. This was really the main point.

The authors were trying to argue that new insights in brain theory tell us a lot about love, but I dind't find it especially compelling. It's always fun to read about chicken wire mother experiments, and how emotional expressions cross culture and also species barriers to a large extent, but it didn't really make me think about how to live my life. At least the book was short. A fun fact:

Echidna's are pretty much as close to the borderline with reptiles you can get and still be mammals. They are solitary and meet only to mate. They lay eggs and carry the eggs around in a kind of open-air uterus. They also have the smallest limbic brain of any mammal.

First Post

2008-10-10T06:01:00.001-07:00

Welcome to derifatives.

This is my "public" blog, as opposed to my "private" blog over on livejournal. Of course, with me the public and private are never too far apart, so I'm not 100% sure yet what's going to go here and what's going to go over on livejournal.

Things you're likely to read about here at derifatives:

My musical projects
My attempts to learn to draw better than a five-year old
Technical issues related to machine learning or computer science
Books I've been reading
Attempts to balance traditional professional, personal, and artistic goals in a world of limited time and resources

I'm sure other topics will pop up as I go along.