Forensic analysts look for abnormal data patterns that allow them to catch bad guys doing bad things, with many economics applications. One of their recent non-economics triumphs has been to catch Victoria’s Secret’s blatant photo- shopping of their ads, notably the example below (HT to Tyler Cowen as usual).
The most obvious giveaway is that they snatched the young lady’s handbag out of her right hand, leaving her holding – nothing. This made the forensic photo expert suspicious and he also caught Victoria’s Secret in more subtle photo shopping. Most predictably, they increased the young lady’s bust size. (This is documented in way more expert detail than you really want.) Not only does Victoria’s Secret objectify women to be like their gorgeous models, but even the models have to be objectified to be their concept of a fantasy woman.
I’m not a marketing expert, but I’m not sure “wear our stuff and you might look good enough to be photo-shopped” is the best ad campaign.
To Victoria’s Secret’s credit, after they got caught, they undid some of the photo-shopping and reposted the picture on their web site. They gave the young lady back her handbag. However, they did not undo the fake bust.
I realize that this is all pretty tame compared with the expectations raised by the headline. But Aid Watch NEVER exploits supermodels! Even here, I refrained from giving a far more sexy, hyper-objectified female example of photo-shopping.
Forensic economics does similar things with patterns in data rather than photos. Ray Fisman at Columbia famously caught some Indonesian companies as corruptly linked to Suharto, because their stock prices would fall whenever Suharto got sick. Ray has made a specialty of this – he also caught some countries smuggling art and antiques, using discrepancies between their exports of these items to country X and the country X data on imports of these items from these same countries.
So here’s the challenge – can we use forensic economics to keep tabs on aid agencies? Oops, I forgot, there’s a lot fewer people who care about aid than Victoria’s Secret models. Can both of you please forward your suggestions on forensic aid evaluation?
12 Comments
True that there are more people who are interested in Victoria’s secret than forensic economics, but they probably aren’t surfing on Aid Watch. I know you love to use humor to engage, but i would have preferred more on forensic economics and less on VS.
Jeff, sorry to distract you. How about addressing the challenge in the final para. What data patterns would you expect to see in a well-functioning aid agency or project, as opposed to a bad one?
Oh no we need lots more on VS. It shows that economists are normal, well, sort of…
You’re late to the photoshopping game, Easterly. Google Ralph Lauren’s recent faux pas making their models waists skinnier than their heads…and getting caught for it!
That said, very interesting idea to use forensic economics on aid.
Great post. Photoshopping has unfortunately been going on for a while now
Regards
Johnathan
it is a great idea! however, having graduated with a master’s in public policy analysis where we combed through data, not only there are many restrictions with statistical methods, but finding data is most often the biggest challenge.
Otherwise, I’ll continue to gorge on my oatmeal muffin while posting photoshopped pics on facebook.
having said that, if you know of anyone who is hiring to do forensic econ on aid money, let me know
We’re trying: http://www.nuruinternational.org/about/research. We’ve been at our first project site for a little over a year, and we’re trying our best to analyze the effectiveness of the work we’re doing. Data gathering happened in December. Boy was it tough, but we have it.
This is (frighteningly) vaguely related to what I work on, albeit from a political scientist’s angle. There’s a growing group of young political scientists trying to figure out how you measure “success” or “effectiveness” in public goods provision when that provision is by non-state actors in weak or failed states.
It turns out that coming up with reliable measures is ridiculously difficult. For example, you can’t say a maternal health project in the eastern DRC failed if you don’t see an immediate reduction in the number of deaths during maternal delivery, for example, because there are so many confounding factors and data collection is such a nightmare that figuring out what exactly killed the mother is often almost impossible. Did she die because the program failed or because she hadn’t eaten in a week or because the lack of birth control access meant that her body was worn out after delivering nine babies in ten years? Or did she die because the program was designed in Washington and doesn’t take cultural factors into account?
So now I’m wondering if forensic methods could be used as a way around some of those problems. Any suggestions on references to learn more about forensic economics methods?
Some technology innovations could make aid data more easily accessible for forensics. See: http://psdblog.worldbank.org/psdblog/2009/10/from-development-20-to-development-squared-what-skills-for-the-aid-sector-of-the-future.html or http://wiki.okfn.org/AidinfoReport for some ideas. Maybe someday there will be enough data ‘fingerprints’ out there for enterprising researchers.
one of my favorite tools in falsified data is benford’s law. I have a spreadsheet tool that is pretty handy to look for spurious errors. bogus data can be caught out this way and least raise suspicion. It would be easy to plot say the raw data from say 20 projects and then say, well we here are the outcomes that look “the cheatiest.” It was used pretty effectively in discussion of vote counting in the last Iranian presidential election.
Oh and I think the world could use AWB (accountants without borders). Forensic accounting does wonders at shining daylight on things and daylight, is the start of a feedback loop.
Forensic accounting, unlike forensic economics usually leads to a very strong feedback loop, when negative findings are made– usually in the form of aid recipients having to pay back money, people losing jobs, etc. Forensic economics usually produces evidence (as in the case of the Suharto regime) that lawyers would consider circumstantial. Michela Wrong’s book on Kenya does a good job of showing how even in the face of strong forensic evidence, donors found reasons to keep giving to the government and lawyers found reasons not to bring cases or try the guilty. I think good old monitoring and evaluation is still the best way to keep tabs on aid agencies.