1. Causality: In this great book I know, Jonathan Morduch describes an obsession over causality as "the marker of the tribe" of economists. Most people outside the field, then, might be surprised to find out how unsettled the science of causality is and how much, after all these years, the practice of academic economics is 80% arguing about causal inference. Well, at least in the circles of applied micro that I run in. Recently Emi Nakamura, an "empirical macroeconomist", won the Clark Medal("American economist under the age of 40 who is judged to have made the most significant contribution to economic thought and knowledge") for her work mapping macro theory to macro reality. One of her more well-known papers is a discussion of the gap between theory and evidence in macro; it has a jaw-dropping section on the best existing "evidence" on the effects of monetary policy. So much for an obsession over causal identification.
Now before getting too holier-than-thou over what is considered evidence in macroeconomics, it's worth pointing out that the experimental micro-crowd is just getting around to measuring general equilibrium effects, the defining feature of macro debates. I've linked multiple times to recent work on GE effects of microcredit (and related programs) on labor markets (See here for links and lots of discussion on that). While I was writing about that the other day, it occurred to me to wonder, given what we know about peer effects in education, whether anyone had looked at whether spillovers/GE effects were responsible for the rapid fade-out of early childhood education interventions. Less than 24 hours later, this new paper from List, Momeni and Zenou showed up in my Twitter feed, finding large spillover effects from an early childhood intervention (1.2 SD! on non-cognitive skills, which are increasingly found to be the more important feature of such programs), which lead to substantial underestimation of program impact. On a related note, here's a short video of Paul Niehaus talking about the value of experiments at scale, including better measurement of GE effects.
Still, there are lots of appealing things about using experiments to establish causality, even if it is somewhat akin to looking for the keys under the streetlights. For instance street lights cause a 36% reduction in nighttime outdoor crime in New York City housing developments. Unfortunately, people really don't like the idea of being experimented on, or even the idea of other people being part of an experiment even when the treatment arms are "unobjectionable." (MR summary here). I'm not really sure how to think about that.
If you want to dig deep into causality discussions, Cyrus Samii's syllabus for hisQuant II class this spring is here. Lots (and lots) of interesting and useful links there. If you're more of the video type, Nick Huntington-Klein has a new series of videos on causal inference, including one on causal diagrams and using Daggity to draw them. If you are among the obsessed and want to be even more so, Macartan Humphreys is looking for a post-doc to work with him on causal inference at WZB Berlin.
2. Academic Publishing: To understand the RCT movement you have to know something about one of the world's least efficient markets: economics journals (Yes, I'm sure someone has a paper/post explaining how the market is actually efficient after all). Seema Jayachandran tweeted this week about stats from her first year as co-editor at AEA: Applied: "4% were R&R, 36% were reject w/ reports, 60% were desk rejects." All of her R&Rs were eventually accepted and average and median time to decision was less than 2 months.
Data on the acceptance rates at all the AEA journals shows that Seema is doing an exceptional job. AEJ: Micro received 415 papers over a 12 month period, made decisions on only 55% of them, which were all rejections. Yes, zero of those 415 papers were accepted. The overall data led to this thread from Jake Vigdor with the provocative question: "If a journal...never accepts a manuscript, does it exist?" Or how about this paper from Clemens, Montenegro and Pritchett that was finally published in REStat after a decade in R&R? For the record, I have a paper with Michael that we got back for R&R after 4 years that I'm supposed to be revising but I'm writing the faiV instead. While I'm grinding an axe, let me also boost this question from Justin Sandefur on why citations still exist and haven't been replaced by hyperlinks. I wonder if an estimation of the dead weight loss from searching for, formatting and copyediting citation details could get published in an economics journal?
One of the reasons for the dismal acceptance rates in journals is the same as the dismal acceptance rates at top ranked universities. Reputation matters a lot. Tatyana Deryugina has a (revised) proposal on a different way of ranking journals that could lead to a more efficient publishing market. It's a start.
And to close out with some positive news: JDE is now prospectively accepting papers based on pre-analysis plans, without requiring the authors to commit to publishing there. It's almost as if the editors aren't maximizing their oligopolistic power. I hope they don't have their economist credentials revoked.
3. Digital Finance/Bangladesh: When the subject turns to mobile money, the country under discussion is still almost always Kenya even 12 years after the founding of m-Pesa. I have a particular axe to grind about counting use of mobile money without including payment cards, but there is now another reason to look beyond Kenya. There are now more people in Bangladesh with mobile money accounts than in Kenya. Of course, that's a function of population--penetration in Kenya is 73% (axe grinding: 70% of Americans have a credit card; this discussion does not include China), while it's just over 20% in Bangladesh. But we should expect adoption to accelerate in Bangladesh, and Kenya to be left well in the dust in terms of accounts.
Also, helpfully, there is an increasing amount of research on digital finance in Bangladesh. I'm a big fan of this particular paper. Here's a review of the state of digital microfinance in Bangladesh. And here's a report on the "opportunities, challenges and way forward" for the digital transformation of MFIs in Bangladesh. The latter builds off an earlier report on "retail micro-merchants" in Bangladesh that I've been meaning to link to but hadn't gotten around to.
But that's also a platform to explain why I hadn't gotten to it: I hate the framing of "retail micro-merchants." That framing allows for comparing the "retail micro-merchant" sector to the garment industry in Bangladesh, which is honestly ridiculous. The better framing--and I am consciously trying to make this a term of art--is subsistence retail. We shouldn't be thinking of this group of people as an industry sector and looking for opportunities for growth for the same reasons that no one talks about boosting the "subsistence agriculture sector." The goal shouldn't be to boost it, but to get rid of it. To get as many people as possible out of subsistence retail as quickly as possible. These aren't frustrated entrepreneurs looking for tools to improve their business, they are frustrated employees. Now I'm not saying I think helping subsistence retailers is a bad idea. We should be looking for ways to boost their productivity, but mainly so they can invest in things that get them out of the sector and into jobs.
4. Jobs: Which is a useful segway into the next item. An open question in my mind is whether we should consider the online-platform sector part of subsistence retail. I'm leaning toward yes. Here's a look back on what has happened to the "servant economy" companies in the 10 years since Uber's founding created a stampede of companies looking to exploit the slack of frustrated employees.
Here's Daron Acemoglu with a short essay on where good jobs come from, and specifically the insane bias in the last 20 years to investing in technology to replace low-wage workers rather than boosting productivity. Here's a new paper from David Kunst on the fall-out of that weird bias, documenting evidence of premature de-industrialization, and which jobs are being lost. Of note, there is an interesting interaction here with formality that may affect how we think about the barriers to formalization.
Getting people into good jobs is certainly a challenge the world over. Here's a new paper on an apprenticeship program in Ghana from Morgan Hardy, Isaac Mbiti, Jamie McCasland and Isabelle Salcher. The results are a bit complicated. In the short-run at least average income falls as youth shift from casual labor into self-employment. But the kids who trained with more experienced and more ex-ante profitable trainers actually do better. The paper's conclusion is that result can be improved by recruiting better trainers. I always have a problem with conclusions like that, because that option is probably not in the choice set. While spillover effects of early childhood education are probably masking some of the gains (see above), it's also likely that the decreased quality of teachers once you move from pilot programs to scale is a factor. I certainly don't know of any programs where the average quality of the staff increased as the program grew.
Take for instance this program from the US which promised to teach coal miners to code and get them good jobs in the tech sector. I'm always amazed that people's initial reactions to programs like this isn't, "I have a better idea--let's teach them to be NBA players, they make even more money than coders!" It's about as plausible a theory of change (a classic study from the 1970s established that the productivity difference between "good" and "bad" programmers was at least 10x). Predictably, the program turns out to have been mostly a scam to enrich the founders.
5. Our Algorithmic Overlords: In case you haven't seen it, Chris Hughes, a co-founder of Facebook, thinks it's time to break up the company. But there's a question about how much Facebook will continue to matter in a lot of the world, asChina perfects the tools of online surveillance and monitoring (I'm beginning to wonder if traveling to China ever again is going to be a good idea for me). Certainly an increasing amount of the hardware and software being used outside of the developed world is being created in China, and with TikTok even in the developed world. I specifically have to give a shout out to the subtle reference to James Scott and Seeing Like a State in that essay to explain the burgeoning black market in old Nokia feature phones.
The view of big tech coming out of China is an interesting lens to look at thediffering perspectives on big tech among right-wing populists in the US and Europe. In the US, where conservative viewpoints are traditionally stronger, big tech is viewed as an enemy of conservatism. In Europe, where the left has long had a stronger grip on more of the institutions of power, big tech is seen as an ally. And after you've thought about that for awhile, think about the new ways that tech in general provides tools of social control--say like deleting references to the ACA from US federal government websites. Why does that ring a bell? Oh yeah, because Intuit does roughly the same thing to con people into paying for their tax prep software.
L-IFT, a diaries research firm, is a proud sponsor of the faiV. See our video on diaries research with 800 women microentrepreneurs in Myanmar.
A proposal from Josh Zumbrun on what the default style for all economic data charts should be, to emphasize measurement error and uncertainty. I kind of like it. Source.
