Week of June 11, 2021
Editor's Note: When people (OK, a person; Hi Jeff!) starts checking on your safety since it's been so long since you've written the faiV, it's definitely time to bend the dimensions of space and time and finally get another one out. The last faiV turned into an extended rant on financial literacy. Here's a long overdue post on why I'm not convinced by the latest systematic review of finlit interventions, which claims evidence of impact on behavior.
–Tim Ogden
1. Digital Finance: First things first. The next faiVLive is coming up on June 23rd, and the topic will be the FinTech Expectations Gap. What's that you ask? Well, it comes from my obsession with the Great Convergence and domains where perceptions aren't converging. FinTech is one of those places. I've noticed that, in general, FinTech innovation is generally viewed as pro-poor in developing countries, but very skeptically in wealthier countries. So I'll be moderating a discussion of whether that's accurate, if those perceptions are justified, and whether there is a holistic framework for assessing whether digital finance innovations are pro-poor that applies across contexts. Joining me will be Yinka David-West, Tim Flacke, Tavneet Suri, Barbara Magnoni and Lois Bruu. Join us.
You can probably guess where my priors are: that we should be more skeptical of DFS and FinTech in developing countries. That's based on articles like this about predatory practices among Indian lending apps. Regulators are struggling to cope with the proliferation of bad actors in a lot of countries. Here's a review of the regulatory actions that have been taken recently in India, Indonesia and the Philippines to rein in badly behaved lending apps. Regulatory capacity is going to be a big issue for the next few years.
But there are also really positive signs and possibilities. Here's Gates' Seth Garz asking why DFS isn't more popular among policymakers and practitioners given how much evidence there is of positive impact of digital transfers. Next Thursday, June 17th, I'm moderating a discussion on what's stopping small shops in Mexico from adopting digital payments with Shreya Kankanhalli and Pablo Ize, hosted by TEC de Monterrey's new Financial Access, Inclusion and Research Center (what a great name!). The registration page is in Spanish but the main part of the discussion is in English so don't worry if you're not hispanohablante. And here's an assessment from Jakirul Islam of the possibilities for collaboration between DFS providers and microfinance institutions in Bangladesh.
There are a couple of places where DFS is gaining traction in some new ways that we will have to start getting used to. China has introduced a digital yuan, the first major economy to have it's own government-issued digital currency. I will confess trying to figure out how different this is than account-to-account transfers of non-digital currency makes my head hurt. But it is different in one very important way: it enables a new level of tracking of transactions, in real time. El Salvador would probably not be the next country you would expect to come up in a discussion like this--but this week the country announced that it will start treating Bitcoin as legal tender, and apparently not charge taxes on capital gains in Bitcoin. No one seems to know exactly what the thinking is, but there are several main areas of speculation. One is that the measure will attract crypto-millionaires and billionaires and crypto firms to the country. The other is that it will make remittances easier and cheaper. The third, which is very much in line with the second, is that the president has ties to MS13 and is making it easier for the gang's transnational criminal operations. That's becoming an increasingly real fear. It's clearly true that the growing ransomware plague would simply not exist without crypto.
Finally, if you want to stay hip in the world of digital finance, you have to start paying attention to DeFi, or decentralized finance. It's the crypto killer app (well the killer app other than enabling cross-border criminal syndicates) according to Tyler Cowen.
2. Microfinance: A year ago I was hosting a faiVLive dedicated to the impact of the pandemic on the sector, and of course, I'm thoroughly on the record that the microfinance industry has never faced a challenge remotely as difficult as this one. But a strange thing happened on the way to the crisis--it has played out much more slowly than I anticipated. No country that I know of has had a crisis like those documented in Weathering the Storm II (a joint effort from CFI and e-MFP to document lessons learned in prior national crises). Indeed, this snapshot from Symbiotics and CGAP paints a pretty rosy picture of how the industry is coping.
Still, if you have your ears to the ground you hear plenty of reasons for concern: liquidity, staff, meaningless PARs, uncertain business models, disconnection from customers. That, in part, was why I convinced a group of microfinance researchers to collaborate on the Sentinel Project--a more systematic way of having ears to the ground to better understand what is happening with MFIs in various contexts around the world. The Sentinel Project is an effort to document how MFIs are managing through the pandemic, as they are doing so. Rather than a look back a few years from now, we are talking to senior leadership from a variety of (anonymous) MFIs around the world to learn about the challenges and choices they are facing, as they face them. A much better website is coming--and over the next few weeks we'll be posting summaries of what we're learning from the regular interviews. If you have any interest in microfinance I think you'll find it fascinating and helpful.
But this is nominally a research newsletter, so I can't move on without linking to a few interesting recent papers. Well the first isn't a paper per se--it's a VoxDevLit literature review on microfinance co-edited by our very own Jonathan Morduch. It's a great overview of the state of the literature on microfinance.
Bari, Malik, Meki and Quinn have a new paper on asset-based lending in Pakistan. Graduates from standard microcredit loans are given the option to get a much larger (roughly 4x) loan to buy an asset--the asset is collateral for the loan. The larger loans channeled into an asset have significant and persistent effects on firm profits and household income and household consumption. It's another piece of evidence that in general microcredit loans are too small to escape a "poverty trap". Speaking of poverty traps, here's a recent paper by Clare Balboni, et al that finds evidence of poverty traps in Bangladesh using the BRAC Targeting the Ultra Poor program data--a large asset transfer bumps people above the trap with lasting effects.
Part of the logic necessary to believe the poverty trap story is that households have no way to accumulate a lump sum sufficient to acquire an asset that gets them out of the poverty trap on their own. In the Balboni paper the suggestion is that wages from casual labor are just too low to ever reach the lump sum. Bari et al propose that there are a lack of mechanisms for slowly accumulating flows into lumps.
Which leads us to one more recent paper on savings accounts--which would be one way out of the poverty trap--in Mexico. A lot of savings research finds that people don't use formal savings accounts unless there is a specific incentive to do so (and when the incentives stop, the savings flows stop). That wasn't the case here. A prize-linked savings account leads previously unbanked households to open accounts, and even though the prize was only available for two months, there are still increased flows three years later.
3. PAYGO: The VoxDevLit review covers a number of "asset-based" lending experiments, and they seem to be pretty promising. Defaults don't increase, the assets are put to use productively, and there seems to be a functional business model. Conceptually there is another version of "asset-based" lending that leaves me less enthusiastic--PAYGO.
PAYGO or "pay as you go" is generally conceived as a solution to the challenge of acquiring productive assets. Rather than needing a lump sum to acquire the asset, you only pay small sums for the use of the asset. PAYGO is mostly associated with solar lighting systems. The provider installs the solar lighting system at no cost to the household. The household can only use the lighting though after paying for some amount of time. In PAYGO there's no loan, and it technically gives more flexibility to the household. If you hit a rough patch, you can cut back or temporarily stop your payment and you are no worse off than before the installation.
So far, so good. But it's always possible to take a conceptually interesting idea too far. First there are the easy-to-mock ideas. Here's a proposal for PAYGO inductive cooking stoves that just leaves me shaking my head. Not only is it suggesting that people will adopt yet another cook stove technology--somehow ignoring the decades of failed programs to get people to adopt improved cook stoves--but that they will adopt a cook stove that relies on unreliable electricity, but also one that they will not be able to use at all if they don't have the cash on hand to make a payment, or that will not work if there is any break in the payments system.
If you're feeling a little uncomfortable with a system that prevents people from cooking without making a cash payment first, well, I am too. So try this one on: rather than a PAYGO solar lighting system where you pay for lighting, what if the solar lighting system and its use were collateral for an actual loan. So if you miss payments on the loan, you lose the light. It works, according to this new paper from Paul Gertler, Brett Green and Catherine Wolfram. By works, I mean it reduces default rates. And reducing default rates is a piece of making lending business models work. But still. It's uncomfortable.
But it's discomfort we're going feel a lot more, because PAYGO isn't just booming in developing countries, it's booming everywhere (Great Convergence anyone?). Subscription business models were already on the rise but they were pushed into overdrive by the pandemic. Of course, there are differences between a subscription and canonical PAYGO, but they're not that different. If you don't have the monthly fee in any particular month, you are out of luck. No streaming, no razor, no cooking, no lighting for you. Particularly the subscription model pushes low-income households in all the wrong ways: you have to make a fixed payment at a fixed time despite volatile income, you have to keep track of lots of payments that arrive at different times, you have to make constant decisions about value, etc.
4. Our Algorithmic Overlords: The subscription model also seems like it's ripe for scams. In fact, it already is a common scam--getting people to sign up for recurring payments that they won't notice for at least some time. That's just one type of scam app uncovered in the Apple App Store by the Washington Post inspired by the antitrust lawsuit against Apple by Epic Games. Supposedly the Apple App Store is highly curated to protect people from scam apps, but it turns out that the curation is algorithmic. Scammers figure out Apple's Fraud Engineering Algorithms and Risk (yes, that's the real name) process and find ways to get their app approved, and rake in the money. And there's increasing evidence that scam apps are becoming a "software-as-a-service" offering, just like ransomware. If you're interested in scams, how they work and some efforts to fight back, check out this Mark Rober video on trying to prevent a common telephone scam run from call centers in India.
But sometimes it's the scammers that get scammed. Certainly a highlight of this year so far is this story about how the FBI and other law enforcement agencies created a scam app of their own--an app that was "marketed" to criminals as a way of encrypting their conversations and hiding them from law enforcement. What was really happening was the cops were reading all of the criminals messages.
Admittedly those items are somewhat tenuously related to our algorithmic overlords, but this one is not. It's the story of a man in Chicago targeted by a predictive policing algorithm that turned into a self-fulfilling prophecy: he's been shot two times because of the increased law enforcement attention the algorithm drew to him convinced others in his neighborhood he was an informant.
5. Research Methods: To close out, here are a few things for the hardcore faiV readers, the only ones who would read this far in an email newsletter. Here's Dave Evans' review of the volume Randomized Control Trials in the Field of Development. Of course, I'm linking it because he says nice things about my chapter but you should read it because it says interesting things about the many other chapters. Here's Arvind Subramanian and Devesh Kapur on "The Absent Voices of Development Economics"--no extra credit if you know whose voices are absent (see the Graphic of the Week below). I'll have some more reflections on the absent voices and how we're approaching that in the faiV in the coming weeks. It belongs here because they argue that methods play a role in limiting who has a voice.
And this is a very different kind of item on research methods but well worth the read none-the-less. David McKenzie walks through a new paper on why scaling is so hard, and so often yields disappointing results built on an early childhood education intervention that scaled from 70 kids in Jamaica to 700 kids in Colombia to 70,000 kids in Peru.
Via Pam Jakiela, from Randall Munroe aka @xkcd. I'm so in awe of this that I'm considering making it the permanent graphic of the week.
