A week ago sunday I participated in the Tussey Mountainback 50-mile race outside state college (PA). I had gone to visit a friend who teaches at penn state and who (what a coincidence!) invited me to give a seminar at the business school on that friday. I did actually have a great time meeting folks there, since they have an interesting and diverse mix of topics that they focus on. And my friend (Tony) and his family were kind enough to let me stay with them, which was also a lot of fun.

I say participated in the race rather than ran it, not because I didn’t finish, but because Tony (not an ultrarunner, but a triathlete and slightly faster marathoner than I) and I signed up as a two-man relay team. Team name: “Forward Induction”!  The race is divided into 12 segments, ranging from under 3 miles to over 6, and one can either run solo or as a relay of two to six (or more) runners. This was the 50-mile road national championship for the solo ultrarunners, and a new course record was set by the prolific and versatile Michael Wardian; Connie Gardner won for the women.

Wardian’s time of 5:33 is amazing in part because the course is not flat (they estimate 5000′ of gain overall) and in part because the previous weekend he had run not one but two marathons. Warm-up, you ask? Hardly! The first was a 2:22, and the second was 2:32 on a somewhat hilly course. Wow. And yet this is typical for him, week in and week out.

Tony and I had the easier task of alternating six legs each, which was not something I’d ever tried before — and hence was curious about. He took the odds and I took the evens, which left me with slightly more miles overall (27.x to 22.y). It’s essentially a big loop along mostly gravel roads, with some pavement, so we and the other teams drove a vehicle from checkpoint to checkpoint. I and the other drivers for our wave (made up of various team sizes) left a few mins before the runners, and drove the 3.2 uphill miles to the first station. Toward the top we started passing slower runners from the previous wave, which had started 30mins earlier. We parked near the transition zone, made finals preparations, and then got out to wait.

In no time at all he was there: about 25mins elapsed, or just under 8:00 pace (almost all uphill; ~700′ total gain). I took the ‘baton’ (a wrist-snap-bracelet doohicky) and headed out for my first leg: 4 miles downhill. Given the adrenaline, and the fact that I had forgotten to bring my watch (so was a bit nervous about time, although not very), and not knowing how much the format would get to me, I ended up flying down at well under 7:00 pace. Tony passed me in the car a couple of miles in, which was of course the pattern for the rest of the day. One of the strange aspects of doing this as a two-person team is that you basically never see or talk to the other person during the entire event, other than quick questions or a few words of encouragement either during the transition or while passing.

So on we went: Tony did leg 3, just under 4 miles with a lot of downhill, in just over 7:00/mi. Leg 4 was the longest of the race, at 6.2mi or 10k, mostly flat with one big hill of about 500′. I ran it at…

Continue reading

Apparently the best bet is to hang around famous people (so the media will also be there) and then say things about topics that readers will care about (so the media will want to tell them about it). In this case I was at a Federal Reserve research conference on the long-term effects of the great recession. Chairman Bernanke was also there ( = the famous person, at least in certain circles), and he gave a keynote speech just after lunch. I didn’t find it too enlightening, although apparently if you parse his words carefully enough, it was a big deal.

But let’s return to me. My colleague Anat Bracha and I had prepared a paper for the conference, about the effect of the recent real estate crash on consumer attitudes toward homeownership. Short story: people in hard-hit areas are less confident (relative to those in less hard-hit areas) about buying homes if they are young, but are more confident if they are old. In both cases, we only see an effect if the person has had direct personal experience of loss (e.g. someone close to them was foreclosed upon). The working paper can be found on the conference website linked to above, but we’re revising it now in light of comments so I won’t push that here.

I think it was fairly well-received, and everyone seemed to be interested at least. Hence the conclusion that the topic is one that average readers can relate to… and have an opinion about. Bernanke happened to be there for most of our session (his schedule didn’t allow him to attend the remainder of the conference), and at the end of his prepared remarks he added something along the lines of: “Economists are apparently getting into social psychology, and this is probably a good thing.” We take this as a compliment, but perhaps that’s an example of interpreting all evidence to confirm our priors!

We got a lot less media attention than he did, but a few reports did appear online that summarized and discussed our results, including in the Wall Street Journal econ blog, the New York Times econ blog, and Business Insider. Several of my friends and fellow economists are used to this sort of thing, but it’s new and kind of fun for me. Some of the comments on the web are very amusing, although (honestly!) I didn’t read them all. To be frank I find it somewhat astonishing that there are almost 50 comments (as of now) across the WSJ and NYT, along with people tweeting it and liking it on facebook: e.g. 183 likes so far on the WSJ post!  Thank you all, but even I don’t find it that exciting. Also amusing that one headline focuses on the young and one focuses on the old; I couldn’t have guessed which would be more newsworthy.

I’ll finish with a few other anecdotes from the conference:

• The largest financial institutions in France and the UK have assets greater than their respective GDPs (and the ratio in Switzerland is 2.5), which came as a big surprise to me.
• An attendee from the banking industry said that in their world “extremely long-run” = “3-6 months” and that essentially all profits come from short-term trades and deals.
• Someone quoted Kissinger: “The illegal we do immediately. The unconstitutional takes a little longer.” But don’t jump to any conclusions about the context…

Every year the super bowl has a loser, and every year pre-printed t-shirts celebrating that team as the winner are rendered useless. So every year those t-shirts are donated to people in developing countries, and every year (at least recently) this leads to claims by the aid community that this is a horrible program: they don’t need our t-shirts; it hurts the local textile industry; cash would be better; etc etc. This argument applies to many other in-kind aid programs, including most notably US food aid worth billions of dollars per year. See long list here of relevant posts from one of the  flare-ups.

My colleague Dean Karlan discussed the issue rationally earlier this year, but let’s take a closer look at the tradeoffs involved. As usual, it all comes down to picking the correct counterfactual. I can imagine roughly five possibilities, taking the t-shirts as an example (but with obvious analogues for food or anything else):

• A: do nothing
• B: donate the t-shirts as currently happens
• C: instead buy equivalent t-shirts locally and donate them
• D: give the cash equivalent of what it would cost to buy the shirts locally
• E: give as cash the costs of administering the t-shirt aid program

Note that E would take include all the managerial and shipping costs, as well as perhaps any meager revenue that could be generated by selling the shirts in the US (possibly as rags as Dean suggests, or pillow filler, or whatever). Rough calculations by others suggests that D and E are approximately the same amount in this case, so for simplicity I will treat them as a single possibility. But note that this needn’t be the case in general (and either could be larger); presumably E is the more appropriate counterfactual, but many people seem to imagine D as the natural alternative.

My first claim is that B is clearly better than A, which not everyone appears to agree with. Basically we are adding positive net value to society X, so we must be making society X better off. Now it’s quite possible that some members of that society (e.g. textile manufacturers and t-shirt vendors) will be made worse off, just as getting rid of farm subsidies in the US will make farmers worse off. But in both cases the net gain is positive, and frankly in both cases the losers almost certainly have higher average income than the winners (poor recipients of free t-shirts in one case; typical taxpayers in the other), so inequality will be reduced as well.

The only argument I can see on the other side is that in-kind aid not only hurts current local industry, it depresses the long-term industrial capacity of the society, making them ‘dependent’ on aid and unable to move forward on their own. This is possibly true for some goods, but very unlikely for clothing or food. There is basically no chance that local economies will actually regress in their ability to produce such staples, in part because there is no chance that aid will actually displace all local production in either area. So if they can spend more productive capacity in the next decade or two on other things (education, health, etc) while receiving e.g. food aid, I have no doubt they will be better placed to return to own production in the future.

Next we compare B and C, which most people see as a no-contest win for C, since it yields the same t-shirts for the poor while also…

Continue reading

Richard Feynman was basically the equivalent of a deity at caltech, which is an ironic twist given some of his views on religion. But that’s irrelevant for today, when we celebrate his enthusiasm for life and nature and beauty — both aesthetic and scientific. The photography in this video is fabulous, and his words are a nice complement. Reminded me of one of my favorite Feynman quotes:

What men are poets who can speak of Jupiter if he were like a man, but if he is an immense spinning sphere of methane and ammonia must be silent?

Here’s the video…  Feynman Series: Beauty

Now I’m off to watch #3, on Curiosity.

Last year I predicted the winners of the Nobel prize on my friend Chris Blattman’s blog, before TFB existed. Well, to be slightly more precise, I inaccurately predicted the winners, although I was happy anyway because Peter Diamond — one of my advisors in grad school — was a co-winner. Fall and chill air have come again to boston, meaning it is time to consult the magic 8-ball once more; the prize will be announced in a few days (on 10/10).

I’m going to be boring (or consistent and principled, as the case may be) and choose the same top prediction as last year*: Hansen, Hausman, and [Hal] White for econometrics, making it ha-ha-ha. Second choice, just to diversify my chances, is Jean Tirole for industrial organization. For the record, I did take a class in grad school from Hausman, and Tirole visited every summer. Go MIT!  [And hey - you're only six places behind Caltech among the world's top universities... better luck next time!]

*We had a betting pool in grad school on who would win the prize. Every year for four years I picked Amartya Sen, and every year he failed to win (that’s $8 I’ll never get back). Then I graduated, and the next year he won. Hope this doesn’t jinx Hausman et al similarly.

[Update: wrong again. At least Sargent was one of my top choices from last year, as linked above...]

Apparently F9 exists in a now-derelict rock climbing rating scale for face climbs. In this case it refers to the Hair Raiser Buttress, an unlikely-looking 5-starred three-pitch classic route in the granite basin formation 30mins east of highway 395 in the northern owens valley. It is rated hard 5.9 (a typo for 5.10a?) in the modern system, and is slightly run out (i.e. risky to lead, given the distance between protection) even with the ‘extra’ bolts that were added after the first ascent… and chopped… and added again.

But stepping back for a moment: I had just finished two solid days of moderate technical climbing (including two very specific sequences of more than moderate climbing) in the high sierras, as described in too much detail here. My partner Peter and I had a morning to kill, and he had read about the granite basin area in his guidebook, but neither of us had ever been there. I wasn’t sure I was up for the HRB itself (since 5.6 – 5.7 is usually more my thing), but the guidebook authors raved about it, and there were easier climbs on the crag if necessary. We took off in my rented Kia, and after passing the unmarked turn once (“It can’t be that small dirt road, can it?”), we came to the same realization as Sherlock Holmes: Wherever all other contingency fails, whatever remains, no matter how improbably small a dirt road it is, must be the truth.

After many minutes to go a few miles, we saw the formation in front of us. The air smelled pleasantly of something sweet, which we eventually hypothesized was burnt sagebrush from continually rubbing against the hot underside of the car as we drove over it. We parked a few hundred feet from the base, since the road got even worse at the end and we figured we could probably walk that far, even on sore legs. The buttress route itself was obvious from a distance, as the prominent prow near the middle of the granite outcropping, with strange pockmarks becoming visible as we drew closer. I decided I had to give it a shot.

We found the actual start of the route after a bit of searching: you have to step off a large flat block onto the lower part of the face itself, perhaps 15 feet off the ground itself at that point. There was a bolt visible up and to the right, requiring some nontrivial unprotected climbing to reach. The surface is an amazing scalloped melange of cusps and depressions at various scales: lots of texture and lots of features, but very few really solid holds for the hands or feet. No cracks or anywhere else for trad pro, making the bolts mandatory (currently six on the first pitch vs three on the first ascent: hard core!).

Peter was leading, of course, and I belayed him as he went more slowly than usual up to the first hanging belay anchor. His caution, along with the grade and the unusual rock, was making me more than a little anxious. I had no idea how far I would be able to get, but as the follower there was really no harm in trying (he could lower me to the ground if necessary, and we were planning to rappel the route anyway so it wouldn’t change anything for him). He reached the anchor safely and put me on belay (with a bit of tension until I got safely going!) – I…

Continue reading

California is home to fifteen mountains that are at least 14,000′ high, ignoring niceties of prominence, including the highest in the continental US: Mt. Whitney, at 14,505′. Colorado has over fifty, although there’s even more dispute there about exactly what counts as a separate peak. There are speed records in both states for who can do them all the fastest (measured in days; pretty spectacular!), and there are people who have done them all in the winter, and surely the youngest, and so on. Needless to say, I will not be breaking any records, but I would like to climb all fifteen of the california ones. It’s not so much to accomplish a tick-list (although maybe that enters on the margin) as a way to organize and to choose and to commit to getting out there.

I believe my first one was Mt. Tyndall in june or july of 2001, followed closely by its neighbor Mt. Williamson (second highest in the state). Since then I had done six more, along with a number of failed attempts, placing me just past halfway. None of these require technical roped climbing, although there was a lot of scrambling (e.g. some short fourth class on Williamson, and the spectacularly exposed third-class east ridge of Russell), and some snow travel, and I had chosen one technical route (the swiss arete on Mt. Sill). But I knew that both Thunderbolt and Starlight had summit blocks that required serious fifth class climbing, even if only for a few moves. Some people don’t go to the very top (the summit registers are at the bases of the blocks for obvious reasons), and some people get basically hauled up, but I wanted to free climb both of them if I could do so safely. This was the goal of my most recent trip to the sierras, along with North Palisade, another fourteener along the same ridge.

After finishing a conference in palo alto on sunday almost two weeks ago, I drove west on 580 and then 108 and 120, bringing back pleasant memories of the many previous times I had made the same pilgrimage. I still enjoy both the name and the contours of Old Priest Grade Road, every time I head through there. And then yosemite itself, 20 wonderful degrees cooler than the central valley strip-mall-opolis: views down into the valley and toward half dome; lake tenaya; the stately pleasure domes of tuolumne; and finally my favorite view, back toward the aptly named cathedral peak on the skyline (where I had once solo’d to the summit in John Muir’s footsteps, with nary a soul in sight). Then down tioga pass and out of the park, driving south on 395 to mammoth lakes, where I could sleep at the highest-altitude motels in the region.

On monday I did some work and then met up with my partner Peter, a far better climber who had been in the area for a couple of weeks already, to warm up on the 5.7 north arete of crystal crag near mammoth. He led all the technical pitches, and I simply enjoyed the approach hike and the unusual white rock and the views of the lake. Tuesday morning we did a gear sort and packed up for the approach to the palisades. An early storm had come through the week before, bringing snow and cold temps up high, but we were hopeful that more moderate temps would prevail and would melt off most of the snow along the route. The weather forecast…

Continue reading

Advance warning: this is going to be a non-pc post!

A recent ruling by the international federation that governs running has engendered much discussion by aficionados of the sport. The ruling body decided that women’s records would only count if they were run in races that only had female competitors, since otherwise some women might have an unfair advantage through being paced by (faster) men. Pacing can be a direct advantage if you draft off someone, or an indirect advantage if you get pulled forward psychologically (see fascinating recent research re tricking people into going faster). Indeed, there have been multiple occasions when men specifically paced women for exactly these purposes, so it is a real concern to that extent. Of course, all of this only applies to road races with common starts, not to track races that tend to have much smaller and gender-segregated fields already.

Specifically, many marathons had already switched to starting the elite women separately from the elite men, so that in effect they are running separate races. The new ruling says that anything else doesn’t officially count, but they went a step further and applied this retroactively… which had the effect of erasing the women’s marathon world record!  It had been run with a common start and therefore is no longer considered legitimate. Fortunately, and perhaps not coincidentally, the new [slower] record will be held by the same woman, Britain’s Paula Radcliffe, although this is probably imperfect compensation for her. For the record, whatever I think of the basic rationale, the retroactive component strikes me as ridiculous and unjust.

Some folks have asked what this means for age-group records, since obviously anyone not at the very front can be paced in the same way. It seems to me that the natural solution is instead to create one main open category for all comers, for which official records would be kept, and then to semi-officially keep track of any other times that one wants: fastest woman; fastest local runner; fastest over 40; etc. This is fair to everyone, since everyone is treated equally when it matters. In fact, the official rules needn’t mention gender at all (which would also solve the Caster Semenya problem).

A talented female ultrarunner who shall remain nameless was recently complaining about ‘biased’ media coverage of a large race, in which far more attention was paid to the male winner (and indeed several top male runners) than to the female winner. She seemed to miss the point that all those men had run the race significantly faster than the woman. Speed isn’t everything, but it is the main point of a race, and it is the only reason to focus on the winners at all. Maybe the woman had trained harder and given literally her all and maybe not; maybe the men had and maybe not; maybe the guy or gal who came in last had worked harder than all of them. That would certainly be worth a newspaper article. All that we can judge directly, however, is the time from the start to the finish, and as the recent ruling shows any other divisions are likely to be more arbitrary than we like to imagine, unfair to some, and subject to even more dispute than usual.

When I was younger I followed tennis closely, and I remember the period of time when complaints about unfair pay led to e.g. the US Open (one of the few tournaments at which both men and women competed, albeit not with each other) giving the same monetary amount to…

Continue reading

In college, my friend and I used to laugh at the ubiquitous media pollsters who gravely announced that candidate X was ahead of candidate Y by 3 percentage points, but that the poll’s margin of error was 4.5, so in fact they were in a “statistical dead heat”. [Yes, we went to caltech and were are nerds.]  All it means is that we can’t, with suitable confidence, statistically reject the possibility that they are running even. But the poll still tells us that X is ahead! In particular, if that pollster and his/her statisticians were to offer me 50-50 odds on the candidates, I would pick X in a heartbeat. The statistics would be behind me on that, and indeed I would win money on average.

Unfortunately, many scientists (perhaps especially social scientists? I’m not sure) seem to likewise focus on statistical significance and miss the real picture. Suppose I run a regression and find a coefficient of 900 and a p-value of 0.09 (which is considered barely significant for economists; substitute 0.04 if you’re a lab scientist). Suppose further that 900 is the estimated effect of whatever (education on wages? drug packaging on adherence rates?) and is large enough that, if true, it would be meaningful and might well affect policy proposals. Looking good so far – nice result!

Now suppose I run 99 more regressions trying to estimate the same effect. Some of them use the same data but different specifications (e.g. more controls); some of them use data from another state or country or year; etc. All of my coefficients are between 800 and 1000, but all of the p-values are between 0.1 and 0.15: statistically insignificant! One way to phrase this is that 99 out of 100 specifications were insignificant, so my result went away (and I might think I should have just stopped after the first one, so that I could publish this… or that I should only report the first one in my paper). But a better way to look at it is that I tried 100 quite distinct ways to estimate the same effect, and they all gave me pretty much the same answer. In other words, I would have more confidence in the result after the 99 extra regressions, not less as it would appear if you framed this as “1 significant; 99 insignificant” and completely ignored the magnitude of the coefficient… which is presumably what you were actually interested in originally.

I realize there are often ways to combine data and do meta-analyses, but I think my point stands: using the usual metric, many scientists would conclude that my result had gotten worse, when in fact (I believe) it got better.

One more hypothetical example: suppose I have to pick between two policies (one of which might be the status quo). I have exactly one study to refer to for each policy. For policy A, the research estimates a large coefficient (translating into large benefits, let’s say) but has a p-value of 0.11 (insignificant). For policy B, the research estimates small (but positive) benefits, with a p-value of 0.09 (significant). I have to pick one now, without the benefit of any future research, so which should I choose? Obviously A.

I think most (but not all) social scientists would get that one right in practice, but they should then realize that it means they are implicitly making some sort of tradeoff (as far as determining what makes a good result) between the coefficient and the p-value. This is what they ought to do, but when they talk…

Continue reading

My friend Austin Frakt at The Incidental Economist has discussed and mostly defended welfare economics previously, in the context of health policy as well as more generally (the latter partly in response to a post I co-wrote with him). Apparently as a result of a recent twitter exchange, Paul Kelleher (a philosopher interested in health issues) provides a different take on the matter. Both Frakt and Kelleher reference Uwe Reinhardt, a highly-respected health economist at princeton, who writes intelligently on policy.

However I think both Kelleher and Reinhardt are either slightly confused or slightly misleading in their conclusion that welfare economics is taken too far normatively speaking. In Reinhardt’s amusingly titled “How Economists Bastardized Benthamite Utilitarianism” (which Kelleher links to and seems to build on), he claims that statements such as “on the efficiency criterion any change in policy that makes George $2 richer and Martha only $1 poorer is a good thing” depend on the dubious assumption that “an additional, say, $100 will yield the same pleasure to a billionaire as it would to a pauper, and that a $100 loss will visit on the billionaire the same degree and intensity of pain as it would on the pauper.” I fully agree that the latter is dubious (indeed wrong), but the first statement does not depend on anything nearly that strong.

Indeed suppose that a policy makes G $2 richer and makes M $1 poorer, but that the extra $2 only increases G’s welfare by 2 utils (WLOG), while the loss of $1 decreases M’s welfare by x utils. If M is poorer than G, x will be greater than 1 (diminishing marginal utility) and indeed the interesting case that Reinhardt has in mind is when x>2, which is quite plausible if their relative wealths are sufficiently disparate. Then the policy has decreased total utility: gain of 2 utils vs loss of more than 2 utils.

But what will happen next? G, feeling rich, may want to purchase something with his new cash… after all, money is only useful for what it can buy. And M, feeling poor, may decide that she has to sell something. They do not operate in autarky, and that’s the whole point of economics. Let’s suppose that M has a nice ripe apple, which is worth 2+ɛ utils to both G and M: recall that the underlying idea is exactly that they’re both human, differing only in wealth, and should be treated equally. We assume that ɛ is small enough that 2+ɛ<x, and since the alleged problem only occurs because x>2, we know that we can find something in that range.

Now M sells her nice apple to G for a price $p falling between $1 and $2 (it doesn’t matter exactly where). Clearly this trade benefits both of them: for M, $p>$1~x>2+ɛ (where I use ~ to denote the equivalence for her between $1 and x utils, which was our assumption given her wealth); while for G, 2+ɛ>2~$2>$p. Hence they each prefer their situation post-trade to their situation pre-trade.

Naturally the interesting question is where they now stand relative to before this ‘silly’ policy was implemented in the first place. Well, M has more money than originally (lost $1 but gained $p) and has lost 2+ɛ utils from her apple. Meanwhile, G also has more money (gained $2 and only lost $p) and has additionally gained 2+ɛ utils from the apple. So in total the utility dimension cancels out and they are overall better off, even in utility terms since they…

Continue reading