Week of January 21, 2019

1. MicroDigitalFinance: Many of you will be familiar with the story of microcredit's rise and sort-of fall, and it's current state of--I don't know, existential angst? But if not, the story is ably told in a new Vox piece by Stephanie Wykstra, with some comments from Jonathan and I included. Not too long after that, the Campbell Collaborative and 3ie issued a "systematic review of reviews" of the impact of financial inclusion, led by Maren Duvendack. I have to say it's kind of weird. The one sentence conclusion is "Financial inclusion interventions have very small and inconsistent impacts." Which apart from appending an "s" to the perfectly plural "impact", I don't disagree with. But this format is a review of reviews which imposes some weird constraints. Ultimately only 11 of 32 identified studies were included, and only one of those was from an economics journal, two are earlier Campbell or 3ie publications, two are specifically only about women's empowerment, and three are about strangely specific topics like HIV prevention. So I'm left really uncertain what to think of it.
Of course, the hot topic isn't generic microfinance but digital finance. The Partnership for Finance in a Digital Africa has an updated "evidence gap map" of research on the impact of digital finance featuring 55 studies (which is more than I have had the time to delve into so I can't compare it to the Campbell/3ie inclusion set). There's a summary of the findings at Next Billion.
Finally, here's an interesting story about Econet, the Zimbabwean mobile money provider--interesting in that it is really about the evolution of mobile money providers from following M-Pesa to following Tencent.
  
2. US Inequality: A big part of the story of understanding US inequality specifically, and inequality in developed countries in general, is understanding what has happened to wages of low-skill workers. The NYTimes has a piece on how cities have shifted from being the "land of opportunity" for such workers to a trap, based on work that David Autor presented in his Ely Lecture at the AEAs (by the way, AEA, it's still a good time to rename the Ely Lecture!).
One policy option for addressing stagnant wages for low-skill workers is to raise the minimum wage. Cengiz, Dube, Lindner and Zipperer continue their long-running work on the effects of 138 minimum wage changes between 1979 and 2016. They find increased earnings and essentially no effect on number of low-wage jobs
That's encouraging. Less encouraging is a new paper from Rodrik and di Tella finding that people are really, really happy to support protectionist policies, regardless of their politics, as a policy response to trade shocks.

3. Our Algorithmic Overlords: Speaking of people's attitudes, there's a big new report on Americans' attitudes on artificial intelligence from something called the Future of Humanity Institute, which as a name is somewhat creepy in my opinion. Maybe I've seen/read too much dystopian fiction. Anyway, they find that Facebook is the least trusted institution when it comes to AI development (no surprise) and the US military is tied for most trusted (big surprise, apparently these people haven't seen/read the same dystopian fiction I have). Also of interest, the median respondent thinks there's a 50% chance that robots will be able to fully replace human beings in less than 10 years. And just because, here's a Night Before Christmas style poem about the future of AI.
Meanwhile, MIT Technology Review "analyzed 16,625" AI papers to predict directions of future research. As someone interested in the future of humanity who doesn't trust either Facebook or the US military to develop AI, I'm encouraged to see cyclical patterns of research consistent with over-confidence.
One of the questions about the development of AI and machine-learning is how it will integrate into existing procedures. Flint, MI is a particularly fascinating case study on those challenges--and they are large. In a community with every incentive possible (they are literally being poisoned by their water and have extreme budget constraints) except politics to adopt the most efficient approach, the machine-learning approach was abandoned. I guess that should make me re-think my attitude toward phrases like "The Future of Humanity."

4. Methods and Evidence-Based Policy: Back in the fall I featured a paper about the effect of political connections on business success because I was so impressed by the method: Abhit Bhandari set up an actual company in Senegal and had his salespeople vary their pitches to signal political connections. Turns out Bhandari is not alone. David McKenzie has a new post at Development Impact on the apparently hot new trend in experimental development economics: setting up your own firm so you can run experiments on it. If you thought the barriers to running and publishing an experiment were high before...
Eva Vivalt has a new paper on specification searching and significance inflation in impact evaluations (see, you don't need to add an "s" to impact!). She finds less bias in economics and health papers than what's been found in political science and sociology. She also finds significance inflation in RCTs is lower than other methods and has fallen over time.
Here's an article from Gelman, Goel and Ho on what statistics can't tell us about affirmative action at Harvard. I'm a sucker for experts writing about the limits of their field.
I mentioned new research earlier punching some gaps into existing evidence bases. Here we go. Money priming, like other forms of priming, doesn't actually have a meaningful effect on behavior. The charts in this one are particularly striking. And an at-scale implementation of CBT for disruptive kids in Chilean schools radically backfired.
And because I have no other place better to put it, but wanted to include it, here's Ray Fisman and Michael Luca on how free pens are killing so many people in the US that average lifespans are falling. And on a related note, ugh, ugh, ugh. There are some things that need to be re-named more urgently than the Ely Lecture.

5. Global Development: When you can write about industrial policy and subsistence agriculture in the same item, you have to take advantage. Thanks VoxDev! Dani Rodrick has an overview on the resurgent economics of industrial policy, which is a very helpful refresher if you've looked at David's post on setting up your own firm to run experiments and are thinking it may be time to change your topics of interest. But VoxDev also has a summary of work reviewing what's been learned about improving agricultural extension services from the Agricultural Technology Adoption Initiative. Which is a very helpful overview if you, like me, have long-standing plans to look at what we can learn from research on subsistence farming to design programs for subsistence retail.
But there's still a long way to go, because even after all this time studying small-scale agriculture we still don't know a lot. Like how much of the difference in productivity from farm-to-farm is real or just mismeasurement. That paper should also be of interest to anyone thinking about studying firms, by setting up their own or otherwise, or in industrial policy.

Apropos of nothing, I found this chart, and  the related blog post  looking at data from lots of different drugs, on the frequency of use of marijuana quite interesting.

Apropos of nothing, I found this chart, and the related blog post looking at data from lots of different drugs, on the frequency of use of marijuana quite interesting.

Return to the Weekly faiV