1. Evidence-Based Policy: So this may seem pretty off-topic as a way to start, but here's a story about the very slow moving revolution in soccer/football analytics, told from the perspective of attending a "bootcamp" put on by one the leading firms in the field. Why is it in the faiV? Because I think there is a lot for those of us who think about evidence-based policy to learn from watching how evidence infiltrates other domains. [Side-note: the RCT apologetics that appeal to "the way it's done in medicine" annoy me to no end, because the use of evidence in medicine is terrible.] And I think in many ways the sports world is a useful mirror to the policy world--if only because there are a lot of people who care a lot, have strong opinions but relatively little expertise. Here's a story about that specifically: what it means to be a fan, psychologically, when there is increasing distance between you and the people who are making decisions (or put another way, how does it feel to live in a technocracy?). Which also allows me to slip in Glen Weyl's recent essay, "Why I Am Not a Technocrat."
I don't worry that much about the pros and cons of a technocracy as we are so far away from living in one--many of the people in positions to make decisions are still a long way away from adopting the evidence that is available, even when their job would seem to depend on listening.
Of course there is another factor delaying evidence-based policy in many domains: the poor quality of the evidence. Here's a newly revised paper from Bradley Shapiro, Gunter Hitsch and Anna Tuchman about, of all things, advertising effectiveness (Twitter thread here). I find it interesting because this is a place where you would expect that there is lots of demand for high quality evidence. And yet, with really painstaking work, the authors are able to show that the published literature is quite biased, and therefore wrong. So wrong that the maxim should possibly be not that "half of my advertising budget is wasted, I just don't know which half", but "Three quarters of my budget is wasted...". Waiting for the revolution indeed.
Finally, since I expressed growing skepticism about nudging last week, here's a paper that finds an effect in a place I would not have expected it at all: reminding seniors with reverse mortgages to pay their property taxes.
2. SMEs: Thanks to David McKenzie, I just learned about a relatively new "book" from the World Bank on High Growth Firms: Facts, Fiction and Policy Options for Emerging Economies. It's a terrific effort to pull together a lot of research from different countries and account for how uneven the data is. Two important evidence-based takeaways: past episodes of high growth are not predictive of future ones, and not even that predictive of survival; and, the link between high growth and productivity is really weak. The only quibble I have with it is that it is framed too much for "emerging economies." Everything I see here is relevant to the US and other developed economies as well, where the thinking on SMEs can be just as wrong.
Policy prescriptions in the book include focusing on managerial skill, which I am increasingly convinced is the crux of the matter. Another is to focus on market linkages, particularly export markets. Here's a J-PAL report on helping small-scale Egyptian rugmakers connect to export markets, which boosts their profits and productivity (2017 QJE paper here). For one more aspect of SME development and policy implications, see item 5 below.
3. The Corrupted Economy: For those of you who were a bit tuned out during the summer, "The Corrupted Economy" is my new header for items that reveal the "great convergence" between the economy faced by the bottom 40% of the US income distribution and that faced by people in middle-income countries. I try not to include it every week just to maintain my own mental health.
Here's a new paper that encapsulates a lot of what I think about under this heading: despite supposedly random assignment, in Chicago and New York, bankruptcy lawyers are able to manipulate case assignment to the benefit of their clients (and the detriment of those who file without legal representation). The poor are different from you and me--even the rules to make sure they are treated fairly aren't fair. Or put another way, even the programs designed specifically to help them--like opportunity zones--are quickly turned into programs that benefit the wealthy.
The premise of the corrupted economy is that there are two different economies now in the United States. Here's a new report from Brookings on the two different economies and how quickly they are diverging. We've long known about the divergence between urban and rural economies (another feature in common with middle-income countries)--this analysis shows that this urban/rural divide is increasingly a Republican/Democrat divide. Over the last ten years, median household income in Democratic districts has risen by about 5%, while marginally falling in Republican districts.
Here's a new paper on another phenomena that has been oft-remarked: the school-to-prison pipeline. Using students that are plausibly exogenously moved from a low-suspension school to a high-suspension school, the authors show that being suspended increases the likelihood of future arrest and incarceration.
What can be done about the corrupted economy. Here's a new paper by Lily Batchelder and David Kamin about the real possibilities (technically, not politically) for taxing the rich.
4. Our Algorithmic Overlords: Where does the use of algorithms increase fairness and where does it mask, perpetuate and amplify unfair discrimination? Here's a Science Friday story about facial recognition in criminal justice.
What about other kinds of recognition? Perhaps we should be concerned about recognition based on medical imaging since apparently a large number of medical images are pretty easily available on the internet.
But back to the main question: Does AI reduce or increase bias? Of course, it depends on a lot of factors, but one of the largest is how the human beings and AI interact. That stretches from how the data sets that AI engines are trained on are generated (often by human beings with a lot of biases) all the way to the discretion that human beings have in following or rejecting the AI conclusions. Here's a story about how human curation is sneaking back in to fix, replace or simply be an alternative to AI recommendations. More to the point however is this paper thatcompares discrimination in face-to-face lending and via fintech platforms. Borrowers of color pay higher prices in both data sets, but the gap is smaller in the fintech data, and while the fintechs do engage in price discrimination they are much less likely to discriminate by denying loans.
5. Migration: This week I finished up the copy-edits on a paper I co-authored with Michael Clemens on rethinking the research agenda on migration and household finance. Michael and I first submitted the paper a little more than 4 years ago, but I was struck by how many of the research questions we posited back then remain quite relevant. Look for a link soon.
In the meantime, Ryan Edwards has a couple of posts reviewing the literature on "brain drain" (or more properly, the positive spillovers of migration) at the DevPolicyBlog of ANU: Part I and Part II. Samuele Giambre and David McKenzie have a new paper looking at the effects of self-employment, and particularly encouraging self-employment (otherwise known as microcredit) has on cross-border migration, finding a small, negative effect. Corina Mommaerts, Melanie Morton, Mushfiq Mobarak and Costas Meghir look at the effects of rural-to-urban migration on informal insurance in Bangladesh. I think most people's priors would be that migration reduces informal insurance arrangements--but here migration improves informal risk sharing, suggesting that benefits from migration subsidies are 40% higher due to spillovers. And bringing all this together in a Great Convergence kind of way, here's Monica Langella and Alan Manning discussing whether or not people "move to opportunity" in the UK and why regional differences persist even when people do so. In summary, there isn't enough moving far enough by the right people: young people are more likely to move and move farther than older people; poorer people are less likely to move and don't move as far. All-in-all it's highly relevant to thinking about migration and household finance in other developed and in developing countries.
On the Great Convergence and Corrupted Economy topics, here's data from the most recent Bureau of Labor Statistics data on project job growth in the US. As pointed out by Heather Long, seven of the 10 jobs projected to grow the most in the US in the next 10 years pay less than $34,000 a year, and 6 of them less than $27,000 per year. Source: BLS .
