1. Targeting: I intended for the faiVLive conversation to spend more time on targeting than we did--it's a sort of rushed conversation at the end. Targeting is something that I've been thinking about a lot, but I'm not sure what I think yet. So forgive me for just ruminating on a few things here.
The whole concept of microcredit is based on targeting--every lender has to target not only those interested in taking a loan but those interested in repaying a loan. Hand-in-hand with targeting repayers was targeting borrowers who were "entrepreneurs," people who would start a business, since the belief was a new microenterprise was the only plausible way for these very poor households to repay. But since the rhetoric emphasized that the poor were natural entrepreneurs, targeting repayers substituted 1:1 for targeting entrepreneurs. Given the findings of microcredit impact studies--namely that while average impact is minimal, there are people who see large gains--the focus on targeting has returned. See for instance, asking middle men who the best farmers are, or surveying other microenterprises.
But if your aim is reducing poverty, then you have to care about more than just finding the borrowers who will repay and have the highest returns on capital--you have to care about equity as well and the effect on, or exclusion of, the poorest or least able to generate high returns. Earlier this year I linked to a paper by Hanna and Olken on the equity effects of targeted transfers vs. UBI. Here's an interview with the two that summarizes their findings: for most poor countries, targeted transfers far outperform a UBI in terms of total welfare. And by the way, here's new Banerjee et al paper from Indonesia showing limited distortions from proxy-means tests.
Of course, in targeting microcredit we are doing the opposite essentially: looking for a proxy-means test to exclude the least-able to generate high returns. What effects might that have? If we boost market efficiency, it could be good for most everyone. That's not just theoretical--here's an empirical finding from Jensen and Miller on improving market efficiency in Kerala boat-building finding higher aggregate quality, lower production costs and lower quality-adjusted prices. But maybe not. That paper above on using middle-men to target finds that traditional allocation of loans does better for the poorest. And as we discussed on the faiVLive conversation, there can be systematic differences in market structure that limits who can generate high returns (in this case, among women seamstresses in Ghana). It's why I worry about what exactly is being measured in targeting algorithms like EFL/Lenddo.
The possible gains and losses have to be measured against the cost of targeting. The cost of microcredit as it exists, without targeting, is pretty low. The median subsidy per loan is about $25, not much for spreading access to the liquidity management features of microcredit well beyond those with high returns to capital. And then there is reason to think about the effect of greater targeting on the microfinance business model. Here is one of the few economics papers to make me actually angry, suggesting that microcredit contracts were purposefully designed to limit the growth of borrower's businesses. While I wholly reject that claim, the underlying idea is worth considering: microcredit's low relative costs are based on a mass-lending business model and MFIs have largely failed to find a way to compete higher up the banking value chain. Altering that business model could have unintended consequences. That's not just based on that paper. As I mentioned last week, City of Debtors, a book about small sum lending in New York City during the 20th century confirms the business model problem is real and pervasive.
So I don't really know what I think. I'll keep thinking about it, but as always I appreciate your thoughts if you're willing to share them.
2. US Inequality: I haven't covered US Inequality for several weeks, and so things have been building up. And there's been a whole lot of new stuff in the last few weeks. Let's start with the state of median US income over the last 30 years. The widely held current view is that incomes for all but the top quintile or decile have been stagnant. But that's heavily dependent on all the adjustments that need to be made for taxes, transfers, inflation and innovation. Stephen Rose at the Urban Institute summarizes the past and new work trying to measure changes in median income, and then writes in more detail about the methodological issues. One thing that had particularly slipped by me: Picketty, Saez and Zucman have a newish paper updating the famous results that showed stagnation and find median incomes have increased about 30% over the last 30 years. That shifts the proportion of gains by the top decile from around 90% to around 50% (I'm intentionally rounding these numbers because they are so sensitive to methodological choices, that I think we're all better off not reporting precise numbers because of the illusion of certainty that goes along with them). Perhaps one of the reasons that these new findings didn't seem to get as much attention as the idea of stagnation for the middle class, is that the new paper also finds that stagnation is true for the bottom 50% of the income distribution.
This week the US Census also released it's "Small Area Income and Poverty Estimates" for 2017, with county-level data on incomes and poverty rates. They find that over the last 10 years, median incomes in 80% of US counties were unchanged, with 11% of counties seeing an increase and 8% seeing a decrease. When you look at the maps, it's apparent that a majority of the counties seeing an increase are related to the fracking boom (and thus mostly in places with very few people). On the poverty front, there's a whole lot of stagnation too, with almost 90% of counties seeing no change, but 8% seeing an increase and only 3% seeing a decrease. Not an encouraging picture.
Whenever you talk about incomes and poverty, it's worthwhile to think about the definition of poverty. Here's Noah Smith on updating the definition of poverty to include volatility (though he shockingly fails to mention the US Financial Diaries). And here's Angus Deaton on "How America poverty became fake news"--with some more methodological detail and the horrid engagement of the present administration with international attempts to measure poverty.
There's plenty new on the policy front as well. Here's a new paper estimating the total budget effect of the EITC--finding that the program self-finances 87% of its cost by reducing use of other transfer programs and increasing taxes collected. And here's The Hamilton Project on the work histories of people receiving SNAP and Medicaid benefits, finding that the majority are working, but irregularly and a substantial portion would "fail to consistently meet a 20 hour per week-threshold" because their hours worked vary so much from week-to-week.
3. US Inequality, Part II: I told you things were building up. Here are a few more things that are a bit less connected, to each other at least. People born in the late 1920s have had consistently higher mortality rates beginning at age 55, "rendered vulnerable by being born during and just after the Great Depression."
The Federal government took over the public housing system in Wellston, MO, near St. Louis, 20 years ago because of chronic mismanagement. It didn't get any better and now it's being shut down. That means 20% of the town's population is going to receive vouchers to leave the town and find housing elsewhere. Here's a thread from Jenny Schuetz of Brookings on the issues. She's a lot more concerned about moving people than I am.
Finally, some new data on women's earnings. You probably saw the study that measures the wage gap not based on hourly earnings, but on what people earn cumulatively over 15 years, finding that women earn about 50% of what men do because of lower rates of participation (hey Stephen Rose is a co-author on this one too). It's an interesting way to look at the issue, but I haven't figured out how to think about it yet. But that finding very interestingly dovetails with new work on the effect of attending an elite college. The traditional finding is that on average, the selectivity (I'm purposely avoiding using the world "quality") of the college someone attends doesn't matter. But for women it does matter--it substantially increases wages through the labor participation rate. But it also decreases the chances of marriage.
4. Our Algorithmic Overlords: I haven't been neglecting this category as much as US Inequality but I have been curtailing the entries because of time. Which means that there's also plenty built up here too.
I've frequently covered stories about China's surveillance state, especially when it comes to Uyghurs in Xinjiang province where it's increasingly clear that hundreds of thousands of people are being sent to concentration camps. Here's a first person story about how that surveillance state works.
Most of what I feature here is from academics researching the application of AI or machine learning or skeptics. But I occasionally like to cast my eye over what the business world is saying. Here's how AI can make us more human. Though I have to confess, of late, I'm not sure I can fully endorse anything that makes us more human. For the more traditional, at least for the faiV, perspective here's the new AI Now Institute report. They use the phrase very differently than, say, Prosperity Now: the headline is 10 recommendations for immediate and significant regulation of tech companies in general and AI applications in particular.
The other frequent area of coverage in this heading is mocking blockchain. Was there ever a more perfect item than blockchain projects in development have a 0.0% success rate. Here's a post with more details and less snark, but the same scathing conclusions. In an attempt at a veneer of fairness, here's a thread for Vitalik Buterine making a case that as the transaction costs of blockchain entries fall, there are some compelling use cases. Your mileage may vary.
5. Methods and Evidence-Based Policy: A special edition of the faiV focused on these built-up items is coming later this week.
Very relevant to the inequality conversation, and whether people should move, here's new data from the US Census on the cratering rates of Americans moving geographically. This remains to me one of the great mysteries of the current US economy. Source: Quartz
