Every year I give essentially the same concluding short speech at our AQR Insight Award day (when the authors of the five finalist papers present their work, competing to be among the one to three chosen to split the $100,000 award). I tell the big group we’ve gathered that it’s my favorite working day of the year (notice the hedge!). Research, both academic (as highlighted here) and practitioner, is perhaps the deepest part of AQR’s DNA. We engage in it, both consuming and producing, to make our investment processes better, to strengthen our ties with outside researchers, and, as over-the-top as this may sound, because we love it. While I can’t come up with something original each year, I hope I make up for it by delivering this repetitive message with sincerity.
This year we split the prize between two winners (a tie).
One winner is "International Currencies and Capital Allocation" by Maggiori, Neiman, and Schreger (of Harvard, Chicago, and Columbia; talk about non-denominational!). This paper exploits a new and fascinating data set to study global investment positions and what currency they’re denominated in. They find that investors appear to have a strong home currency bias, preferring bonds denominated in their own country’s currency when choosing among bonds issued in different currencies by foreign firms. As a result most firms around the world issue only in their local currency and don’t even bother to access foreign capital. The one exception is the U.S. and the dollar. Even small U.S. companies, borrowing exclusively in U.S. dollars, find it quite easy to source some of that borrowing from abroad. The authors also find this unique role for the U.S. dollar has substantially increased since the Global Financial Crisis of 2008 which is consistent with the world viewing the U.S. dollar as the reserve currency and valuing that perceived status and safety more in the post-crisis world. Novel data and a neat result!
The second winner is "Taming the Factor Zoo" by Feng, Giglio, and Xiu (another three school collaboration, this time between City University of Hong Kong, Yale, and Chicago). This paper deals with one of the most vexing and topical problems in empirical finance — how to judge whether results are "real" or just the output of concerted data mining. Specifically, we mean the notion that if you test enough factor ideas some will ex post look significant due to sheer randomness (i.e., they aren’t real and won’t work going forward). This is a topic anyone serious about factor investing has worried about. AQR is anything but complacent on this topic, and we’ve written about it extensively, including recently here and here. Still, I’d be remiss if I didn’t mention that some of the headlines you might read (500 factors tested!) are somewhat exaggerated. Many, maybe even most, of the "factor zoo" are variations on a much smaller set of themes. No serious factor investor believes in 500 independent factors, not even close. But, just because headlines often overstate the problem, I would certainly not go the other way and be dismissive. Indeed, data mining should be the number one concern of all quantitative researchers. Essentially, the authors propose a new approach that tries to compare the contribution of a new factor to all existing factors (implicitly dealing with the issue I mention above — that factor candidates are often far from independent) while simultaneously trying to avoid data snooping biases. I can’t do it justice here (due to the intended breezy blog format, laziness, and an aging mind). For both papers I certainly strongly encourage interested parties to read them in their original glory.
Hopefully the AQR Insight Award remains my favorite day of the (working) year for a long-time to come. This year it certainly lived up to it.