Saturday, August 31, 2013

more limericks

To contribute to some poetry reading with friends last night, I wrote another (extremely dumb, but hey, I was vacuuming) limerick:

An insecure cephalopod
felt his limbs were unfortunately flawed.
Their number was nine,
which would've been fine,
but they stood out for being so odd.

And here's Matt's from a long time ago actually. Controls theory background required...

When designing adaptive control
And desiring a proof of your goal
Use Lyapunov, not Routh
Or, with sine waves at "out"
You'll end up with an infinite pole.

Thursday, August 15, 2013

signaling is powerful

Person2: He is giving a talk on behavioral economics here in a month.
Me: Oh cool. It's hard to tell what he works on judging by google results, but I'd bet he doesn't stay in academia, in case you're interested in hiring him.
Person2: He's starting part time in a week :)

These kinds of inferences aren't unjustified prejudice because people signal intentionally. When an academic has a lot of google hits from public talks they've given on general interest topics, doesn't have a professional website but does have a linked-in page, and has few publications/working papers despite being out of grad school for at least a year or two, it's pretty clear they don't want to stay in academia. It's so clear because it's a signal that's been carefully crafted to send that message.

(And in case it's not clear from my google results, maybe because I have this silly unbreakable blogging habit, I am almost exclusively interested in staying in academia. Please hire me :)

Tuesday, August 13, 2013

the tragedy of ignoring marginal incentives

A heartbreaking example of the perils of ignoring marginal benefit analysis, the unintended consequences of perverse incentives, and the danger of letting good intentions overrule good economic sense. All such programs need to be vetted/sanity-checked by economists to avoid these kinds of outcomes.

This well-illustrates a common problem with aid programs more generally speaking, that occurs when benefits are contingent on some threshold level of poverty. Implicit marginal tax rates (that is, tax rates including both taxes and benefits) can be substantially higher than 100%.

(Any economists reading can skip the next two paragraphs. But I'm including it because you'd be amazed how many people, even very smart and well-educated people, don't realize how marginal tax rates work. The misconception is even published as a "tip" in places like USA Today. This is a serious failure of the public educational system. It's insane that you're required to read Romeo and Juliet to graduate from high school, but you can get by without having any clue how interest rates or tax rates or tax-deferred retirement funds or social security or credit reports work. And that list doesn't even include anything about more complicated investment vehicles, like stocks and bonds and CDs and mutual funds and index funds and all that fun stuff that middle-or-higher-class Americans lose millions on due to lack of information, just the stuff anyone needs to know about...)

In terms of income tax rates alone, the system is designed so that your marginal tax rate is never above 100%. That is, when you make one more dollar, you can never lose more than that due to paying higher taxes. Getting bumped up to a higher tax bracket can never hurt you.

Say there are two tax brackets. People who make at most $20,000 have a marginal tax rate of 10%, and everyone else has a marginal tax rate of 20%. That means that even rich people pay only 10% of the first $20,000 they make. Then they pay 20% of any additional income they make beyond that. Going from making $20,000 to $20,001 in income raises your taxes from $2,000 to $2,000.20. Income taxes diminish the incentive to make more money, but never reverse it.

Unfortunately, this breaks down when you include benefits. If you're so poor that you're hardly paying any taxes to start with, and you're receiving benefits that only kick in if you stay poor, the implicit marginal tax rate can be much higher than 100%. That means that earning more money can make you poorer, once you subtract the value of the child credits and subsidized health care and other benefits that you'll lose. These two charts shows strikingly how little incentive there is for a very poor family to earn more money, unless they can earn much more money, over $40,000, well above the poverty line. The first chart shows the family's real income (net of taxes and benefits). See how it drops down in several places? That means that making more money hurt the hypothetical family. The second chart shows the implicit marginal tax rate, which shows why this happens. It's well over 100% for a significant range of income.

This isn't an abstract problem. There is ample evidence, in the context of countless specific programs, that people respond to these perverse incentives by trying not to lose their benefits by doing a bit better on their own and failing to qualify for them. We need to be more careful about designing these aid programs. People need help, but first and foremost they should be able to help themselves without that effort turning out to be counterproductive.

The AIDS-contingent benefits are just a particularly heartbreaking example of this wider problem.

Wednesday, August 7, 2013

more SMBC disrespect

Ok the re-do doesn't really portray economists in a nice light either but at least I got to express my frustration at the "economics isn't a science" club :)

re-SMBC part 2, 3037:

Edit: to clarify, since I have zero artistic skills, that's supposed to be the economics-as-science skeptic being cooked at the end...

Tuesday, August 6, 2013


I am anal enough about digital organization not to do this myself, but when collaborating it's inevitable. After the last 16 hours of emailing who-knows-how-many versions of this paper draft between three coauthors, merging in all our edits, making more edits, and ending up with filenames as ridiculous as the title of this post (because it makes me sad to have to start adding hours and minutes to the datestamp), this comic feels very relevant. And I'm starting to listen to Matt about the virtues of git*...

*To clarify, I have nothing against git, I'm sure it's wonderful, it's just that I mostly work by myself, so what's the point. And for it to be useful with collaborators you have to work with people who use it, which seems unlikely given that in my field it's hard enough to get people to use miraculous timesaving things like LaTeX, let alone programming languages more real than Stata or Matlab, let alone miraculous timesaving things like Linux, let alone git, so what's the point in starting. Although I could at least force my future RAs to use it with me... and hopefully start a trend...