Week of September 13, 2019

1. Digital Finance: Is a tide turning on digital credit? Old hands in the microfinance world like MicroSave and CGAP have been highlighting concerns about digital credit for the last few years, but the non-specialist community hasn't seemed to notice until recently. In late August Bloomberg had a quick hit piece with an eyebrow-raising headline, "This Nobel-Prize Winning Idea is Instead Piling Debt on Millions," which is likely the way the general public will perceive this despite the protests of insiders that telecoms/fintechs making instant loans at high rates with minimal customer engagement doesn't have much in common with traditional microcredit. A more serious treatment,"Perpetual Debt in the Silicon Savannah" was published in the Boston Review the same week, though it's frustrating in its own ways, notably the lack of engagement with the global/historical context of small dollar lending or with the research from financial diaries.
In both articles there are two additional issues that I wish received more attention. First, the value of liquidity management. The authors of the Boston Review piece, Emma Park and Kevin Donovan (both historian/anthropologists), spend a good deal of time talking about the "zero-balance economy" creating a situation where consumers can be exploited without engaging on the need for services to manage liquidity when you have low and volatile incomes. Second, the kind of default rates being hinted at in these articles raise serious questions about the business models and sustainability of digital lenders. Tala, one of the larger digital credit providers in Kenya (and elsewhere) just raised another $110 million. How much of that money is covering losses? I would love to see some analysis of what sustainable default rates are for digital credit.
Shifting gears a bit, the reason that the Kenya specifically and East Africa more generally remain in the spotlight on digital finance is the ubiquity of access. But ubiquity can't be assumed and in general I would say not enough attention is being paid to what happens when ubiquity fails. Here I don't mean places where everyone knows service is unreliable, but places and times where service is unexpectedly unavailable. Here's a story about the problems that can create in the US with ZipCar customers stranded in the "wilderness" because of a lack of signal leaves them unable to unlock or start the vehicles. More seriously, though, is the concern when access is limited because of political reasons. Here's a story about the rise in government-directed internet shutdowns. Of course there is the big concern of how these shutdowns would affect people who have adopted digital finance and find themselves unable to spend. But I also wonder if Tala investors have priced in the risk to the business model of internet shutdowns.
Internet shutdowns are a blunt tool. We should also be concerned about more fine-grained tools in the hands of governments or private companies. I'm old enough to remember when one of the highlighted "benefits" of digital finance was that it created an audit trail of transactions. Here's a story about how much data about you leaks to unknown parts of the internet when you use the Amazon Prime card and the Apple Card. And finally, here's a new report on cash as a public good from IMTFI, sponsored by the International Currency Association, which I am fascinated to discover exists (though I'm even more fascinated to discover the International Banknote Designers Association, which is one of its members).

2. Our Algorithmic Overlords: There is of course a lot of overlap between concerns about digital finance and privacy and digital everything and privacy. One of the standard mantras of those gathering and selling data is that much of it is anonymized, so we shouldn't be concerned. But, of course, not so much. That's not just a concern in the US, because digital data-gathering is becoming a thing worldwide. Here's a plea to stop "stop surveillance humanitarianism." And here's a story about how a high-tech surveillance approach to improving disaster responseturns out to have not been such a good idea (spoiler: garbage in/garbage out).
One of the major concerns about the use of algorithms in these situations is the garbage in/garbage out problem--combined with the gee-whiz veneer that technology provides obscuring that problem. I'm generally skeptical of that argument as a whole, because my experience is that people are far less likely to trust an algorithm than a human being (In some sense I wrote a whole book about it in a different application: the bogus fears that Toyotas were suddenly accelerating and trying to kill people). But there are other forms that algorithmic discrimination can take. Here's a story about a new US Housing and Urban Development regulation that would exempt landlords from responsibility for the discriminatory results of their screening practices as long as they don't understand the algorithm, which y'know is a given.
Finally, there is a new documentary about the 2016 US election, the Brexit referendum, Facebook/Cambridge Analytica, etc. called The Great Hack. Here's a piece about 7 things the documentary gets wrong which I find pretty convincing.

3. Behavioral Economics: Similar to the expanding skepticism about digital credit, I feel like I'm also seeing some growing skepticism about the efficacy of interventions built on behavioral insights, particularly the broad category of "nudges," (though there is clearly a lot of disagreement about what should be considered a nudge).
For instance, four recent impact evaluations on "behaviorally-informed" interventions in education have come up with not much. few significantly exceeds that of smaller programs. A scale up of a program to encourage completion of financial aid applications, that had been successful at the state-level, found no impact. A program to encourage low- and middle-income students to apply to more colleges found no impact overall (with a very small effect on the quality of schools applied to by African-American and Hispanic students). Nudging college students to study moredoesn't do anything either. But this last one encapsulates the issues for me because it's framed as a success. A program to use behaviorally-informed notices to parents about their kids missing school increased attendance by 40%! Except that amounts to .07 days, or by my rough calculation about 20 minutes.
If you're interested in more specifically finance related nudges, here's an experiment in the Netherlands to encourage more savings with a "social norm nudge." It succeeds in getting people to intend to save more but has a precisely estimated null effect on actual savings.
At a more theoretical level, Dmitry Taubinsky with various co-authors, has two new papers digging a bit deeper on how behavioral insights play out. In the first one they find that people will take up commitment contracts both to exercise more and exercise less! Even worse, educating people about self-control problems reduced demand for commitment contracts. Here's we show that with some uncertainty about the future, demand for commitment contracts is closer to a special case than to a robust implication of models of limited self-control." In the other paper, the authors find that people have "heterogeneous rules-of-thumb" for attention to costs, which complicates a lot of models of limited attention.
And it's not just the empirical research that is getting shakier--some of the key underlying premises of the whole idea that priming and nudging will be effective are also being shaken. Here is a defense of the limited willpower literature from Roy Baumeister, a defense that will shift most readers priors further against that literature being reliable. But more importantly, the neuroscience finding that most undermined classical concepts of free will (brain activity begins before a person makes a conscious decision) has been debunked as an artifact of background mental noise.

4. New Research: I mentioned that there is always a flood of new working papers that appear in late August and early September (and NEUDC is coming fast!). Here are a few that caught my eye. Martin Ravallion has a new paper on measuring global poverty. Closely related is a look at cyclical fluctuations in social indicator measures--some measures of welfare are reflections of global and national business cycles, while others are better measures of long term development. Here's a paper that looks at the longer-term effects of agricultural subsidies in Mozambique, finding thatspillovers account for most of the gains (spillovers are a particular interest of mine lately). An asset transfer program in the Philippines designed to reduce child labor actually increased child labor (as children were pulled out of school to put the asset to use). In Chicago, eviction is a consequence of financial distress not a significant cause of it, i.e. the majority of the negative consequences come before eviction rather than after. And finally, I had been wondering recently about replications of the Drexler, Fischer, Schoar rule-of-thumb experiment, and it turns out there was one recently (hat tip David Mckenzie): Irani Arraiz, Syon Bhanot and Carla Calero tested a similar program in Ecuador and find "significant and meaningful" effects on sales and profits, driven by women. They hypothesize that women have higher cognitive burdens (due to unequal household labor burdens) and therefore are more likely to adopt short-cuts and benefit from them.

5. Other Odds and Ends: Clearly I'm still getting back in the swing of writing the faiV and there are just a whole lot of things out there that I'd love to write about and link to, but it's already 5pm. So some more odds and ends.
The Netflix documentary about a US factory taken over by a Chinese company is getting rave reviews, from the New York Times and Planet Money. There's also a documentary about a factory in India that has been highly recommended but I haven't figure out a way to watch yet. The Economist has a long look at BRAC and it's future. I'm often baffled that BRAC isn't the most famous thing in the modern world. There's a new comprehensive look at "deaths of despair" in the United States that finds they have indeed risen but with a much more complicated story, suggesting that just focusing on opioid epidemic may be more fruitful than attention to "despair." And I'll close out with some confirmation bias--Lauren Willis pushes back on the CFPB's turn to education and calls out financial literacy as a dead end. And so does Caitlin Zaloom specifically on student debt, calling out Steve Mnuchin who is apparently calling for mandatory college financial literacy classes.

Return to the Weekly faiV