« What I've been reading | Main | Every meal counts »
Does trade spread AIDS?
Emily Oster tackles this question:
I generate new data on HIV incidence and prevalence in Africa based on inference from mortality rates. I use these data to relate economic activity (specifically, exports) to new HIV infections in Africa and argue there is a significant and large positive relationship between the two: a doubling of exports leads to as much as a quadrupling in new HIV infections. This relationship is consistent with a model of the epidemic in which truckers and other migrants have higher rates of risky behavior, and their numbers increase in periods with greater exports. I present evidence suggesting that the relationship between exports and HIV is causal and works, at least in part, through increased transit. The result has important policy implications, suggesting (for example) that there is significant value in prevention focused on these transit-oriented groups. I apply this result to study the case of Uganda, and argue that a decline in exports in the early 1990s in that country appears to explain between 30% and 60% of the decline in HIV infections. This suggests that the success of the Ugandan education campaign against HIV...has been overstated.
Since I used to believe Samuel Brittan when he argued that trade spreads sex, this result accords with my intuitions.
I thank Scott for the pointer. There should be an algorithm informing me every time there is a new Emily Oster paper. If Scott is indeed such an algorithm, I am pleased. And of course I am that algorithm for you.
Posted by Tyler Cowen on October 27, 2007 at 07:05 AM in Economics | Permalink
Comments
"Since I used to believe Samuel Brittan when he argued that trade spreads sex, this result accords with my intuitions."
What changed your mind on this?
Posted by: Andrew at Oct 27, 2007 8:30:42 AM
Something like http://www.changedetection.com/ or http://www.changedetect.com/ on http://home.uchicago.edu/~eoster/papers.html would probably by a good algorithm to inform you when she has a new paper out.
Posted by: Daniel at Oct 27, 2007 9:33:38 AM
Migrants are NOT exports and there are significant exports that do not involve people. This instrument has serious problems. Why not just connect migration and HIV? Better yet, how about people having sex and HIV? Oh wait.
Posted by: David Zetland at Oct 27, 2007 12:17:21 PM
why not connect foreign soldiers invading countries with HIV? In the 70s soldiers that were fighting in Africa came to the Americas
Posted by: Jules at Oct 27, 2007 1:58:35 PM
that relationship is not causal.
It's not exports or trade per se, it's migration, especially seasonal migration.
Increase of export in ethiopian coffee doesnt increase migration as much as increase of export of copper in DRC and Zambia.
In South Africa, the case is even more complicated as migrant work was made larger by apartheid policies that prevented workers to permanently move (with their families and most importantly their wives) to the region where they worked.
Posted by: nu at Oct 27, 2007 3:16:57 PM
also:
everyone knows that sailors spread diseases. but not all means of transportation are equal.
At the same volume of trade, you don't have the same number of potential diffusion agent or even of potential victims depending on the means of transportation used.
Boats who tend to go on longer missions and certainly in deeper isolation have a big volume per crew.
Planes have small volume per crew but also shorter missions.
Trains have a bigger volume bigger per crew than planes but smaller than boats.
Road transportation is the worse. Small very little volume per crew member and the possibility of stopping anywhere.
Now think about variation the number of transmission agents generated by the export of say a million tons of weat if it's done by road, train, boats or planes.
Posted by: nu at Oct 27, 2007 3:48:12 PM
I have a problem with any comparison of African AIDS to AIDS in the US and Europe. They are really 2 different diseases.
In Europe & US AIDs requires the presence of the HIV virus plus diagnosis of one of 28(or so) diseases. The one I can always remember offhand is tuberculosis. Thus, if you have TB AND HIV you have AIDS. If you have TB and no HIV you just have TB.
In Africa, they use the WHO definition of AIDs. This does not require testing for HIV. It looks only at symptoms including:
Weight loss
protracted asthenia
Continuous or repeated attacks of fever for more than a month
Diarrhoea lasting for more than a month
Other signs
Cough 2
Pneumopathy 2
Oropharyngeal candidiasis
And so on.
Absent a common definition of AIDs, lumping US incidences and African incidences is comparing apples and oranges and is a fool's errand.
John Henry
Chronic or relapsing cutaneous herpes 4
Generalized pruritic dermatosis 4
Herpes zoster (relapsing) 4
Generalized adenopathy 2
Neurological signs 2
Generalized Kaposi's sarcoma 12
The def is:
Posted by: John Henry at Oct 27, 2007 5:05:12 PM
John - in the US, AIDS is defined as the presence of an accompanying disease to HIV? So, it'd be possible to have a very high viral count and a very low T-cell but not have AIDS, or have a low viral count and a high T-cell plus one of the opportunistic infections and have AIDS?
Posted by: jason voorhees at Oct 27, 2007 7:12:19 PM
Since many diseases in Africa are spread by truck drivers who, bedding down for the night, often make use of local prostitutes, it could conceivably be argued that the trade that requires these trucks is also spreading the diseases. But it's more accurate to say that ignorance and risky behaviour spreads disease.
Posted by: Geoff Hamilton at Oct 28, 2007 2:25:02 AM
The New York Review of Books had a good essay on a more relevant cause of the high rates of HIV infection in Africa -- different African family structures:
"Epstein's view is that the cause of the AIDS crisis in Africa is what has now become known in AIDS jargon as "concurrent" relationships. Africans have about the same number of sexual partners as anyone else; they are just more likely to have more than one long-term partner at a time. Crucially, both men and women have multiple partners, in contrast to other poor societies where men may often stray but women's monogamy is jealously guarded. Western men and women are more likely to practice serial monogamy or engage in one-night stands. To oversimplify a little, Africa's AIDS tragedy is that it combines greater Western-style sexual equality for women with social norms that permit simultaneous long-term sexual relationships for both partners."
http://isteve.blogspot.com/2007/10/what-causes-high-rates-of-aids-in.html
I'd suggest you read my long excerpt from the New York Review of Books, but in a nutshell, the issue is that African husbands tend to be less actively jealous about making sure that their wife's or wives' babeies are really his. That ties in to the traditional economic structure of much of SubSaharan Africa, where women do most of the labor (e.g., hoeing the fields). In turn, husband's don't invest as much in their nominal children, so they aren't as obsessive about keeping their wives away from other men.
James Q. Wilson's 2003 book The Marriage Problem has a couple of good chapters on family structure in Africa, which would be an easy-to-read introduction for economists. The subject of sub-Saharan African family structure was studied intensively in the 1960s by anthropologists, but has disappeared down the memory hole since. It's truly the key to understanding Africa, and economists would do the world a lot of good if they'd study up on it before issuing pronouncements about Africa based on ignorance of the basics.
Posted by: Steve Sailer at Oct 28, 2007 2:44:14 AM
Look, it's an argument against trade that Tyler is willing to accept! :-)
Posted by: David Wright at Oct 28, 2007 3:53:53 AM
This is a surprise?
This differs from any other communicable disease?
What's the point?
Posted by: Shakespeare's Fool at Oct 28, 2007 10:50:53 AM
Shakespeare - one of the more interesting things in this article is Emily's finding that the ABC campaign may be overstated in terms of its overall success. I'd be interested in reading more articles that carefully examine the impact of ABC, since the Bush Administration believes its efficacy to be very high. I also wonder if AIDS mortality might also be the cause of falls in AIDS incidence during the period where Uganda's AIDS rates fell. Thomas Dee and Harrell Chesson have a 2003 article entitled "AIDS Mortality May Have Contributed to the Decline in Syphilis Rates in the United States in the 1990s" that finds the rapid fall in syphilis can be explained by the explosion of the AIDS epidemic here. They do not isolate the mechanism itself (AIDS mortality can cause syphilis to fall by killing individuals with syphilis, by making the sexual network less efficient, and/or through deterrence effects), but find nonetheless causality running in the direction of AIDS mortality on syphilis rates. Since Uganda had very high AIDS mortality, how much of the fall was due to simply the at-risk population becoming smaller? I'm not sure if Emily attempts to address that in her paper or not.
Posted by: jason voorhees at Oct 28, 2007 1:01:51 PM
Question - is Emily's methodology in this paper basically just to (1) provide new estimates of past HIV incidence based on AIDS mortality and assumptions about the incubation spells of AIDS and (2) regress those new estimates of AIDS mortality onto lagged exports?
Posted by: jason voorhees at Oct 28, 2007 1:06:20 PM
I think she also has to come up with AIDS mortality statistics by comparing death rates in various age groups across different countries that are either affected or not affected by AIDS.
Posted by: LN at Oct 28, 2007 3:59:52 PM
Tyler, have you seen this:
Undercover Economist: Emily Oster Is Probably Wrong
It's on a different paper (from 2005) but worth reading nonetheless.
Oster is also very well known for her paper on witch trials and weather. A history grad student friend of mine looked at the main graph of this paper and was very puzzled by the fact that it showed high trial numbers (and correspondingly low temperatures) for the 18th century. She was surprised because the witch trials in Europe go into serious decline around 1650. It turned out that Oster gets this effect by basically dropping out countries other than Hungary for this period (because they had 0 trials!) My friend was also rather surprised that Oster did not include any data from Germany (where most trials took place) and realized upon closer inspection that she had double-counted Essex trials (which were also included under "Home Circuit"). Furthermore, a couple of the regions had an average of 1 trial per year and a terrible survival rate for legal records. In short, it seems that the data used was both unrepresentative and not very credible.
I can see why a paper about witchcraft and weather might not get much scrutiny, but shouldn't papers about AIDS and Africa get *very* close attention from epidemiologists? It's not like this is an obscure field without any policy applications.
Posted by: LN at Oct 28, 2007 8:43:57 PM
Jason said:
John - in the US, AIDS is defined as the presence of an accompanying disease to HIV? So, it'd be possible to have a very high viral count and a very low T-cell but not have AIDS, or have a low viral count and a high T-cell plus one of the opportunistic infections and have AIDS?
It had been a while since I looked at the definitions. At one time, it was the presence of the 28(?) diseases PLUS HIV. If HIV was not present, it was not AIDS.
I just looked at the CDC def in Wikipedia and it has changed some from what I remember but still requires positively testing for HIV.
The WHO/Bangui def used in Africa now recommends but does not require testing for HIV. AIDS can be diagnosed in Africa based solely on external symptoms. These symptoms can also be caused by non-AIDS diseases.
John Henry
Posted by: john henry at Oct 28, 2007 10:07:12 PM
There seems to be a general tendency for young freakonomists to announce some Amazing Discovery based on their use of correlation analysis, only to have it eventually emerge that they botched it up because, up in their ivory towers, they didn't really understand how the part of the world they were working on actually worked.
Posted by: Steve Sailer at Oct 28, 2007 10:31:16 PM
The forthcoming QJE correction that Hartford mentions in that article... Is it this correction? It does not appear that her main results are affected much by noting the incorrect transmission rates she'd been using.
Posted by: jason voorhees at Oct 28, 2007 11:51:55 PM
John Henry - were you referring to the 1980s, early 1990s definition? I think in 1993, CDC reclassified the definition of AIDS. Yes, I figured you had to test positive for HIV to be testing positive for AIDS, since my understanding is that AIDS is full-blown HIV, usually accompanied by extremely low T-cell counts and high viral counts of the HIV virus. I just wasn't aware that the definition itself depended on the presence of other opportunistic infections.
I'm not sure that Steve Sailer is really living by what he preaches. He seems to be (as always) on his Levitt bandwagon, and painting with a very broad brush. I'd be willing to be $100 he has neither read the original Oster paper or the rejoinder just noted, yet he seems pretty confident the studies are complete crap. Asshole.
Posted by: jason voorhees at Oct 28, 2007 11:54:37 PM
Harford's article is very light-handed, but he seems to be saying:
(1) There was an obvious data error in Oster's paper that epidemiologists could very quickly spot
(2) Epidemiologists have considered and rejected her conclusion
As for (1), it doesn't seem to be a very big deal in terms of changing her results (they still look almost exactly the same), but I imagine it's a bit telling in that it reveals that no epidemiologist looked at her paper.
(2) is actually important, no?
Posted by: LN at Oct 29, 2007 5:57:22 AM
In Harford's own words:
But Oster’s conclusion is probably wrong. Epidemiologists embraced the idea of treating other sexually transmitted diseases a long time ago, but it has been discredited (to their deep disappointment) by a series of rigorous clinical trials. Oster says that the most convincing evidence came out after her paper was written; still, she has repeated her recommendations more recently in Esquire magazine.
Oster also made a mistake in handling her data. The error – which she has acknowledged, and which makes a modest but noticeable difference to her calculations – was quickly spotted when I asked two epidemiologists to review her research. The QJE will be publishing a correction.
Oster quite reasonably says that her article has other merits.
Posted by: LN at Oct 29, 2007 6:13:04 AM
LN - I saw that quote in the article. The error appears to be Oster's use of two sets of transmission rates from other papers, right? But, the updated paper with the corrected data does not appear to change her results substantially. So then why does Hartford argue she shouldn't be repeating her findings? Just because epidemiologists, in work before Oster's work, had not find evidence for this effect?
As for (2). Hartford's article notes that the QJE and Oster had provided earlier drafts to epidemiologists. I'm not sure if that means the QJE had it peer-reviewed by epidemiologists or not, but it does mention that effort was made to enter into public health and epidemiological community to get comments and discussion. Personally, in my experience, I've found a lot of turf mentality among the people in public health when economists begin working in that area. Or, just outright dismissal. Without more information about what Hartford is talking about, it's hard for me to believe ex ante that Oster screwed up royally by hiding her work from the public health community, especially since Hartford indicates they had the work but never bothered to comment on it.
Plus, nothing that I've seen so far appears to justify Hartford's dismissal of the paper. The corrections he notes don't appear to undo the paper's findings. And simply referencing two anonymous epidemiologists is just an fallacious appeal to authority without information on exactly what they did to investigate her findings. The chasm separating economics from other disciplines is partly methodological, partly ego, and while it is easy to blame the imperialist coming in, in my experience, I've seen close-mindedness and insecurities on the other end that contributes significantly to the lack of communication.
(I'm actualy amazed at the new transmission rates on a different note - M-to-F transmission rates that are slightly lower than F-to-M? I had no idea in Africa that that was the case, as under "normal" circumstances, the risk of infection is much higher for uninfected women matched with infected men than the other way around. Does this suggest that heterosexual coital practices in Africa increase the transmission rate of F-to-M? If so, what is so different that would make the transmission rate so much stronger for F-to-M?)
Posted by: jason voorhees at Oct 29, 2007 10:04:31 AM
Hartford's article notes that the QJE and Oster had provided earlier drafts to epidemiologists.
Really? It says that Oster could not name an epidemiologist familiar with her article, and that "epidemiologists would not typically agree to review papers for the QJE" -- which is awfully vague as to whether or not any attempt was made to have any epidemiologist take a look.
It is true that the reference to the unnamed epimediologists is an appeal to authority. Of course most of this discussion is occurring on the basis of appeal to authority (Larry Summers, Steve Levitt, and Tyler Cowen say Emily Oster is great, she's a professor at UChicago, she has a PhD from Harvard, economists are smart people and not cranks). Most of us do not have the time or the training necessary to delve into her papers and discuss the nuances, but we are able to say "according to superstar young economist Emily Oster, witch trials were partly driven by colder weather, hepatitis B is responsible for half of the missing women in Asia, we can best approach AIDS by treating other STIs, and a good part of the decline in the AIDS rate in Uganda is due to falling exports."
My history grad student friend tried showing the witch trials paper to some colleagues, and was met with indifference -- why would historians care about some paper in the Journal of Economic Perspectives? Essentially, academics don't see why they should waste time with cranks. I'm not too interested here in debating Emily Oster in particular (although if someone wants to dig through the witchcraft paper I am fairly familiar with it) -- the more general point is that economists are boldly entering other fields and often coming up with results that are flatly dismissed by experts in those fields. This doesn't mean that they are wrong of course -- hardly -- but anyone interested in truth should be concerned with what those experts have to say. If a historian or epidemiologist can spot errors in just a few minutes, this is a problem.
Tyler has done his share of Oster promotion; will he see it worthwhile to point out that it's very debatable that Oster has discovered anything *true* so far in her brief career? Or does this just get in the way of the "gee whiz economics is cool" promotion?
Posted by: LN at Oct 29, 2007 10:50:13 AM
I was thinking of Oster's claim that people in health had seen the paper. I doubt she's lying about that. It could be referring to a conference where she presented the paper, as my recollection from speaking with Oster is that she ends up on a circuit that does put her around other people working in development, which is a diverse group of scientists and policy-makers. This point seems rather minor, especially in today's environment where the Internet makes such work accessible.
The real point of Hartford's skepticism is this QJE correction. But, as I said, I don't understand how his skepticism is warranted by that correction. The correction doesn't appear to be that big of a deal unless I'm missing something (or maybe Hartford is talking about something else).
As for your point that no one in history will bother to read her witchcraft paper because no one wants to interact with cranks. That pretty much sums up the arrogance and turf mentality I was talking about. Basically, they remain insulated from people outside their circle, and somehow that's Oster's fault? Your friend should consider writing a comment on that paper, in all seriousness. If it is such a badly written paper, then point out the problems and send the comment to JEP. Or send it to some journal in history. It's one thing to say "I don't believe the result" but it's another to say "this paper has methodological flaws because data selection criteria bias the resutls in favor of Oster's hypothesis." The latter seems to merit a response. The former (which honestly is more of what I encounter when people object to economists working in this area - it's more of an aesthetic response than a scientific one) does warrant a response.
Posted by: jason voorhees at Oct 29, 2007 11:19:36 AM
