1. The Search for Truth, Part II: Last week's opening theme was about how hard social science is. I often find there's an unspoken wistfulness in social science research for the clear questions and clear answers of the "hard sciences."
But cheer up! It's just as bad on the other side of the fence. When you're frustrated that there doesn't seem to be a biological mechanism that explains the long-term positive outcomes of deworming, remember that we have no idea--literally, no idea--what causes "side stitch," that shooting pain we've all had in our abdomen during exercise. And when you're down in the dumps that so many development interventions don't seem to show much effect, remember that the universe shouldn't exist, and we don't know why it didn't explode nanoseconds after coming into being.
On the other hand, Ioannidis, Stanley and Doucouliagos' paper on how vastly underpowered most economics papers are has finally been published (it's been circulating for awhile). If that's not enough to send you back into despair, the fact that economists need to be reminded of basic good practice in presenting their ideas--per this slide deck from Rachael Meager--might do the trick. Don't get me wrong, it's good advice. But I was reminded of the time I attended a conference for PR "professionals" where the advice included such gems as, "Make sure the reporter you're pitching actually covers the topic" and "Read the last few articles the reporter wrote." Last year I was joking with Jessica Goldberg about starting a side-business editing the introductions and slide decks of job market papers. Perhaps I shouldn't have been joking.
2. The Mess that is US Higher Education (or Labor Markets are Broken All Over): Studying labor market inefficiencies is a common topic in development economics (yes, this is clickbait for David McKenzie). But as in so many domains, the problems we study in developing economies also exist in developed ones, just wearing a Halloween mask. Here's a new study on "credentialism" in the US labor market, the demand for college degrees for jobs that have no reason to require a college degree (as demonstrated by the fact that the vast majority of people currently in those jobs don't have one). That's bad for employers who pay some of the cost of the self-imposed mismatch in the labor market, but it's much, much worse for potential employees who are shut out of well-paying, stable jobs for no good reason. Unless, of course, they spend large amounts of money to get a credential. The large, and growing, lifetime earnings gap between those with a credential and those without has justified the incredible growth in student debt to finance these credentials. But if the credential is just an artifact of herd behavior among employers...
And why are those credentials so expensive? One reason is that the universities providing those credentials are spending, and borrowing, huge amounts themselves in order to attract the students who have to get the credential to apply for a job. So the students borrow, and borrow some more. And then they get shut out of programs for loan forgiveness that they are should be eligible for, because the system is a mess. But don't worry, if their debt gets too out of hand, they can discharge those loans in bankruptcy. Oh wait, we changed the bankruptcy law so they can't ever discharge those loans. Don't forget too that large numbers of the people we've pushed into needing a credential are entering universities, taking loans, but never getting the credential (e.g. 70% of single mothers who enroll).
And the advanced degree market may be worse. A few weeks ago I featured some work on English football academies juxtaposed with a paper about the Clark Medal. Perhaps my comparison was too oblique--so here's a piece from Nature making the connection explicit. The chances that a Ph.D. student will land a permanent academic job in the US or UK is well under 10%. The reason it's plausible to offer job market paper editorial consulting is that the premium for a well-written paper is so large. And it's large because there is massive over-supply.
For those newly minted Ph.D.'s taking adjunct teaching jobs just so they can stay marginally attached to academia and perhaps make enough to supplement their food stamps, I have bad news. Current students (bachelor's and master's students that is) teach just as well as adjuncts, suggesting that "student instructors can serve as an effective tool for universities to reduce their costs." Oh right, I was trying to avoid a novella.
Week of October 16, 2017
1. The Search for Truth: The New York Times Magazine has a long piece about Amy Cuddy, the social psychologist of "power posing" fame, and the messy process by which her research has been popularized and then discredited. The piece suggests that Cuddy (though it by no means holds her out as blameless) has been uniquely and personally targeted as the face of unreplicable and bad social science in an era of changing research practices and expectations, perhaps because she is a woman. More broadly it ponders whether the process and social conventions of communication around challenging social science research may do more harm than good. It points specifically to Uri Simonsohn, Joseph Simmons and Andrew Gelman and their roles in both calling out bad social science and in specifically highlighting Cuddy's power posing paper as an example.
It's well worth the long read, careful consideration but also some critical evaluation. The piece comes at a very interesting time, with the Weinstein saga, #MeToo, and more specifically the push back about Econ Job Market Rumors and bad behavior in economics. It's important to read the piece in the context of such things as EJMR and this anecdote from Rohini Pande (in an interview with David McKenzie this week) relating how a "senior male World Bank economist wrote to our senior male colleagues at MIT and Yale asking that they review our work and correct our mistakes" in one of her early papers (with Esther Duflo; see question 4 in the link, but read the whole thing, it's very good on a lot of topics).
But on reflection, I don't think the idea that Cuddy was uniquely targeted or treated more harshly than others holds water. It only appears so to a New York Times reporter because Cuddy's works is the kind that gets broad attention. Remember when Ben Goldacre kicked off "Worm Wars" with an amazingly condescending piece asking people not to point and laugh at Miguel and Kremer for the supposed "errors" in their Worms paper because they shared their data? Or the language and dudgeon around Reinhart and Rogoff's Excel error? Or the intemperate words flowing around the failure to replicate John Bargh's priming work? From another field, here's some pointed language challenging a recent result on gene editing alleging some pretty basic errors.
Of course, the commonality of bad behavior in academic circles doesn't excuse it. But that cuts both ways. Cuddy has also been using this faulty logic in her own defense. As far as I can tell, her main defense has always been "everyone was engaging in bad research practices, so it's not my fault", and that's definitely the implication that the NYT article gives. I don't see much distance between that and people excusing sexual harassment because they were "raised in the '60s and '70s."
Could the practice of social science be better? There's no question, but it's also not clear exactly how, other than the obvious avoidance of misogyny, ad hominem and personal attacks. But that line is difficult to see sometimes because the nature of social science research requires a great deal of personal investment. It's hard not to feel attacked when one's research, quite literally one's life's work, is criticized.
To me, the most thought-provoking part of the NYT piece is when Simmons, reviewing an email he sent to Cuddy about follow-up work on whether the power posing research was reliable, says "that email was too polite" given how serious he thought the problems were. And there is a lot of bad science that needs to be called out. This week, there's yet another update to the Brian Wansink saga--several papers flat out misrepresent who the study participants were (e.g. a paper claiming participants were 8-11 when they were 4-5). Not calling bad science out, I think, is a real contributor to real world problems, like Chief Justice John Roberts being able to call good political science research "sociological gobbledygook."
Here's a Chris Blattman thread on his reactions. Here's Andrew Gelman's response to the NYT piece and for the sake of this topic it is one of the few posts anywhere on the internet where you should read the comments. Someone in one of the Twitter threads wondered about the responsibility of Gelman and other bloggers like Tyler Cowen to police their comments. I'm sympathetic to this idea, but I'm old enough to remember policing comments on my own blog. It's an incredibly time-consuming and soul sucking affair with lots of trade-offs. The "business model" of blogging just doesn't allow it. In fact, in some ways it was the business model required to police commentary, also known as paid journalism, that led to blogging: the gatekeepers of commentary shut out too many voices who should be heard. Science, and the pursuit of truth, is hard.
Week of October 9, 2017
1. Evidence-Based Policy: Yesterday I was at a workshop hosted at Yale SOM and funded by the Hewlett Foundation on how to better connect evidence to policy. The workshop was part of a bigger project and a series of reports are coming that I will share when they are available. There was a lot of good discussion, but I thought I would share two thoughts that I find to be missing appropriate weight in evidence-based policy discussions. First, there is often discussion of a mismatch in the time horizons of researchers, implementers and policy makers. While this is no doubt true, the mismatch between those groups is trivial in comparison to the mismatch all those groups have with the amount of time it takes for change that people can feel to occur. Deworming's important effects--on earnings, not school attendance--are only felt decades after treatment. Moving to Opportunity similarly has a decade-scale effect. Few if any of the researchers, implementers or policymakers are still going to be around when the world really is undeniably different because of them.
Which brings me to the second point. The enterprise of evidence-based policy is grounded in marginal improvements across large groups of people--and that's a good thing! I'm a big believer in the value of marginal improvements (QED). But people have a really, really hard time noticing or caring about marginal improvements. Human beings prefer stories about big changes for a few people with unclear causality a lot more than they do about marginal gains with sound causal inference. I'm more and more convinced (because of evidence!) that hope is a key ingredient for even marginal impact, but hope comes from Queen of Katwe, not from 1/10 a standard deviation improvement in average test scores. So the unanswered question for me in this conversation is, "How do we manage the tension between the policies that are good for people and the policies that people want?"
In other evidence-based policy news, here's a rumination on the difficulty of applying research to practice in democratization (specifically Myanmar). And here's Andrew Gelman on not waiting for peer review, particularly in Economics, to start putting evidence into practice.
2. Evidence-Based Operations: OK, so there's one more thought: the gap between policy and research, and operations. But rather than a long discussion on that topic, here's a very good new piece on the operational choices of front-line social workers and the gap between policy (whether evidence-based or not) and practice. The challenge in the spotlight is not the Marxist-style view of workers dissociated from their work by rules but workers dissociated because of having too many morally-fraught choices. More light-heartedly, here's a piece that illustrates how hard it is to go from evidence to operational choices, as reflected through the failure of the US men's soccer team (I told you it would return). There is growing attention to front-line staff and the "product" as actually experienced by the beneficiary in impact evaluations, but much more is needed as far as I'm concerned.
3. Our Algorithmic Overlords: Speaking of operations, one of the areas where more attention is needed is the way that operations are being instantiated into algorithms that are opaque or entirely invisible. Ruben Mancha and Haslina Ali argue that that the unexamined algorithm is not worth using. Of course, they are arguing from ethics, not from business profits, where it's abundantly clear that unexamined algorithms are worth using.
Here's a piece about technology-related predictions from Gartner, a tech industry research and advisory company. Skip the first three to see some striking predictions about AI-generated false information, such as that people in "mature economies will consume more false information than true information." There's a threat to advancing evidence-based policy that definitely wasn't on the agenda yesterday. I started my career at Gartner way back in 1995 and I remember one of the first things we were given to read was an an article in Scientific American about the coming age of fake photography and video. Apparently that future has finally arrived.
Week of October 2, 2017
1. Abusive Practices: This is the part of the faiV that is different. But, perhaps contrary to the evidence, I have to hang onto the belief that making abusive practices in many domains more visible will in fact play a role in changing those practices. So first up is a piece about abuse of the elderly in Nevada where for years shady operators, aided and abetted by courts, legislators, medical professionals and other nominally civil servants have cooperated to revoke the rights and steal the assets of vulnerable people. That may seem an abstruse topic, but I think it has lots of parallels in many domains. Often, abuse of the vulnerable is tied to weak institutions or institutions that have no duty to those abused. Here we have strong institutions in many cases explicitly designed and supposedly devoted to protecting the vulnerable, which were turned against the people they were supposed to protect and which made challenging what was going on virtually impossible. As an aside, I have to commend Beth Rhyne of CFI who began talking years ago about the challenges that an aging global population would bring to financial inclusion and protection efforts.
At the other end of the age spectrum, here's a piece about the "1% winners/99% losers" labor market of young football/soccer players in England. It's a form of vocational school that consistently lies to 10 year-olds and their families and then dumps the vast majority of them at age 16.
Stretching even further afield, I'm hoping that many folks will find the time to read, or at least scan, the NY Times article on Harvey Weinstein, the movie mogul, and his decades of sexual harrassment and abuse and cover-ups. I'm particularly struck, if not surprised, because Weinstein moved in supposedly progressive circles. His behavior was apparently an open secret but did not dissuade many from working with him and for him and apparently participating in the glossing over of the abusive practices that let him continue. This piece about the lack of criticism coming from Hollywood is particularly pointed.
And now to connect this back to something more faiV-like: I hope the Weinstein saga provides further momentum behind efforts to reform practices and behavior in the social sciences, particularly when it comes to the academic job market. There is a rapidly growing effort particularly in Economics (with offshoots as far as I can tell in Political Science and Sociology) to make job market information more transparent, but more importantly move it away from sites like EconJobRumors which facilitate abusive behavior. Check out the hashtag #EJMinfo for more. This is a rare obvious opportunity to choose between the type of behavior that enabled Weinstein, and the type of behavior that will make such abusive practices and behavior impossible.
2. Economic History, History of Economics, and Evidence: Pseuderasmus, the pseudonym for an economic historian whose real name I don't know, has a long (long, long) post about the productivity gap that opened between India and Japan in the first 30 years of the 1900s. It's filled with fascinating historical details, so even a skim of it will be rewarding. The short version is that the power of unions in Japan was restrained by demographics, culture and the government which allowed manufacturers there to innovate far more quickly and increase productivity. This in turn left Japanese workers eventually far better off than Indian workers where labor unions exerted more power.
Beatrice Cherrier and Andrej Svorencik have a new paper examining the history and evolution of the Clark Medal and it's winners. Again there are plenty of interesting details to reward even a skim. I took particular note of the concentration of winners--eight universities account for all winners in terms of where their degree was earned, and 10 for all winners in terms of employment when they won. So economists are apparently uniquely good at identifying talent early, right? Right?
Finally, this week I stumbled on a newish site, Straight Talk on Evidence, that reviews not the evidence for various programs and policies (for instance as the Cochrane Collaboration, AidGrade or 3ie does), but the claims made by studies that are part of the evidence base. It's a project of the Laura and John Arnold Foundation. Here is their review of an evaluation of CCTs in the United States and of a Heckman paper on the long-run health impact of early childhood education.
Week of September 25, 2017
1. Basic Income: I haven't touched on basic income in what seems like months, but that's because there was little to report. This week Planet Money has an episode (adapted from 99% Invisible) on the details of what basic income is and how it might be delivered. And apparently last week, Y Combinator announced some more details of their US Basic Income study. If details matter to you, you'll be pleased to know that the work in Oakland that received a lot of attention last year was a feasibility study and now they are planning an RCT with 3000 individuals in two different states.
2. Methods and More: My next book of interviews is about big data and machine learning (If you have a better name than "Dated Conversations," let me know). Susan Athey is the first person I interviewed for the new book this past spring (I hope to have some excerpts of that interview available soon) in part because of some things Athey had written on how machine learning will change the field of economics. There's a new version of a (preliminary) paper on the topic. It has details.
More specifically on details and methods, here's a new paper on the use of randomization to study network effects, a quite tricky prospect. But when it comes to methods and details mattering, two items this week really hit the nail on the head. First, Buzzfeed of all places has a lengthy piece examining the myriad problems that have emerged as people examine the details of studies published by Brian Wansink's Food and Brand Lab at Cornell. Missing data, mis-described studies, statistical errors, it's stunning. This week also saw publication of what is many ways the exact opposite of what appears to be have happened at the Food and Brand Lab: David Roodman's incredibly detailed review and replication of the research on the relationship between incarceration (or decarceration) and crime rates for the Open Philanthropy Project. The starkest contrast for me isn't actually the attention to detail but the philosophy. The Wansink saga began with a blog post that indicated that the Lab was torturing data until it said what they wanted; the Roodman review and replication was done because they were concerned that their beliefs were wrong.
3. Microfinance, US and Global: My expertise and knowledge is definitely concentrated in global microfinance rather than microfinance in the US, but because of the work on the US Financial Diaries I'm learning a lot more about the US. This week for instance I got to hang around the outskirts of the Opportunity Finance Network meeting. There are no links here but a couple of things have really struck me and so I wanted to note them, and invite you to tell me what you think/have seen, etc.
First, I was really surprised about how open the US microfinance community is about the presence of and need for subsidy. Globally I see an almost totemic adherence to the idea of self-sustainability, even in the presence of compelling evidence of the prevalence of subsidy. I'm sure that's a consequence of how those industries have evolved but I'm curious about any ideas about the details of the US microfinance history that led to this.
Second, two parallel conversations really struck me. One was about "community investment" in order to create "quality jobs." The second was about how to use technology to cut down costs of making loans, costs that are mostly about staffing--or in other words, how to expand microfinance by lowering the need for quality employees in the lenders. I bring this up not to point fingers about hypocrisy, but to raise the inevitable trade-offs for MFIs everywhere about reach and cost. The tension doesn't seem to exactly be on the surface in the US but it is more apparent than in global conversations, where the value of the jobs created by the global microfinance movement seem to be ignored, especially in the rush to digital finance services.
Week of September 18, 2017
1. Microenterprise and Household Finance: I assume that most of you are familiar with David McKenzie's business plan competition in Nigeria (there's even a Planet Money episode about it!) and his cash drop work (I have to use this self-serving link of course). David and co-authors have a new paper in Science (summary/blog version here) testing the effectiveness of business training for microenterprises in Togo and find that a standard business curricula did not do much (in line with lots of other business training studies, though most are plagued by too little power) but a curriculum based on boosting personal initiative did have large effects.
I see this as lining up with a stream of research finding that boosting aspirations or "hope" can have meaningful impact in many different contexts (see for instance, this recent work on effects of watching Queen of Katwe) and through a variety of interventions (any one know of an overview of recent work in this vein?). It also helps explain why there seem to be only small effects of business training on businesses that objectively should have lots of gains from marginal improvements in operations--if you don't believe that running your microenterprise better will matter...
In other microenterprise/microcredit news, I learned this week about a study (new draft coming soon apparently) that tests allocating microcredit based on peer views of microenterprise owner business skills. Those ranked highly do in fact see large returns to a $100 cash drop (8.8 to 13% monthly returns). I heard about the study from this excellent thread from Dina Pomeranz on a talk by Abhijit Banerjee and Esther Duflo on what new they've learned since that "old" book Poor Economics came out.
Finally, here's a new piece from Bindu Ananth that should go on your "must read" list. I couldn't agree with this statement more: "[T]he field of household finance has failed to examine the financial lives of low-income families in sufficient detail." She examines specifically issues with how to think about insurance vs. savings, high frequency saving and borrowing, and financial complexity. I will continue to beat the drum on two points: 1) low-income households are having to make financial decisions that would challenge a finance MBA, with large consequences for sub-optimal choices, and 2) almost all the advice we have on making wise financial choices is built on an assumption that the life-cycle model holds true, and may not in fact be good advice if the life-cycle model doesn't hold.
2. Premium Mediocre and American Inequality: I'll lead this off with a concept that I'm not quite sure what to make of, but does have me thinking: Premium Mediocre. The post goes on way way too long, but it's worth reading at least through the first couple of scrolls for some new ways to think about the old problems of inequality and mobility, or lack thereof, and what it does to household decision making.
This summer I mentioned but failed to link to a study on how delivering food stamps more frequently lowered the rate of shoplifting in grocery stores in Chicago. Here's a new paper that shows a much larger and long-term effect of food stamp receipt. Children whose families received food stamps for more years (due to staggered roll out of the program in the 60s and 70s) were less likely to be convicted of any crime as an adult, with larger effects on violent crime.
The importance of such safety net programs in the United States is growing as we learn more about how household finances are changing. Not only is year-to-year volatility seemingly increasing, and month-to-month volatility seemingly spreading, but lifetime earnings aren't just stagnant--they're falling. Some new work indicates that since the late 1960's American men's expected lifetime earnings began falling each year (into the present). That can make premium mediocre a stretch for each new cohort. It also perhaps helps explain this new and fairly shocking chart, based on Case and Deaton's work discussed extensively in the faiV this spring, that has been circulating on Twitter this week.
Week of September 11, 2017
1. Digital Finance: There's a regular theme I hit when it comes to digital finance--digital gives much more power to providers, government or private sector, than physical cash does. And that is something we should worry about. So my confirmation bias when into overdrive when this crossed my feed this week: China is detaining ethnic and religious minorities in Xinjiang Province and one of the criteria for detention is people who "did not use their mobile phone after registering it." Brett Scott objects to cashlessness for both its inherent nature as a tool of surveillance and for more pecuniary reasons: unlike cash, every digital transaction generates fees. Which in turn gives power to the organizations that have a seemingly insatiable appetite for categorizing and controlling people. Hey, ever wonder why Facebook is pushing hard into payments, even into fundraising for non-profits?
Scott uses Sweden's progress toward cashlessness as a foil. Want to guess which other country beyond China and Sweden has made the most progress toward digital-only payments? Somaliland. Huh. Elsewhere, the progress of digital finance seems to have slowed to a crawl: 76% of mobile money accounts are dormant, and the average active user only conducts 2.9 transactions a month. Perhaps that's because of a huge gap in usability that will require a similarly large push in education (according to Sanjay Sinha).
Given the near unrelenting negativity above, I feel like I have to say for the record: I don't oppose digitisation. I oppose not recognizing and planning for the negative consequences of digitisation.
2. Global Finance: Digital finance and mobile money is generally about very local transactions. But another important use is long-distance transactions, particularly remittances. But international transfers of funds require banks to have relationships that cross borders. The technical term is "correspondent banks." What correspondent banks do is vastly simplify and accelerate the flow of funds across borders. So it's a problem that correspondent banking relationships are shutting down as a result of "de-risking," which is banking jargon for "avoiding anything that may draw the attention of regulators who have the somewhat arbitrary ability to impose massive fines." The IFC reports that more than a quarter of banks responding to their survey reported losing correspondent bank relationships with compliance costs the most common reason; and 78% expected compliance costs to increase substantially for 2017.
And now for a bit of levity, if you can call it that. Matt Levine has the incredible story of how the Batista brothers, owners of a large Brazilian meat-packing company, made money shorting the Brazilian Real--they knew recordings of their conversations with President Michel Temer about bribes were going to be released. Is that insider trading?
3. US Poverty and Inequality: This week the US Census Bureau released its report on income and poverty in the United States in 2016. The new was good, at least on a relative basis: incomes are growing across the board and poverty is down. But...the majority of gains are still going to upper income groups, and inequality continues to rise as a result. The bottom half of the distribution is only now getting back to where it was in 1999 or earlier. Here's Sheldon Danziger's take on the data and the policy implications. The Economic Progress Institute has a good overview (with good charts) of the poverty data specifically, which focuses on how safety net programs reduce the number of people below poverty by "tens of millions."
The 8+ million who are above the Supplmental Poverty Measure threshold because of refundable tax credits (e.g. the EITC and the Child Tax Credit) particularly caught my eye because of this profile of a US Financial Diaries household that I just finished. Amy Cox, for the year we followed her, is one of those people. For the year, she is above the SPM because of tax credits. But she receives all of that in one lump sum in February. So for 11 months of the year, she's poor. In 9 months of the year, she's around 75% of the SPM threshold. But officially, she's not poor. Makes me think it's time for a Supplemental Supplemental Poverty Measure that takes into account how many weeks a year someone is below the line.
In other US Financial Diaries news, here's Jonathan Morduch speaking about Steady Jobs without Steady Pay at TEDxWilmington this week (skip ahead to 1:30:00).
Week of September 4, 2017
1. Evidence-Based Policy and Methods: One of the reasons I took a few weeks off was in late August I was part of a panel at Stockholm International Water Week sponsored by Water.org on the "evidence base for WASH microfinance." If you've been following the evaluations of microfinance or of WASH you know that evidence base is thin (in more ways than one). Preparing for the panel got me thinking about the strange state of evidence-based interventions. [Warning: I'm going to oversimplify for the next few paragraphs; if you want not oversimplified I recommend the detailed write-ups GiveWell has on both deworming and WASH] Arguably deworming is the sine qua non of evidence-based interventions right now, but the arguably mostly comes not from whether there is some other intervention with a better claim, but that there are large swathes of people who don't believe the evidence for deworming: epidemiologists. Why? Because there isn't a plausible biological mechanism to explain where the gains from deworming come from. There is no consistent detectable effect of deworming on weight or anemia for instance.
In the meantime, there's no question that if you remove bacteria and viruses from water, people won't get sick and will have all sorts of positive short-term health gains. But the most rigorous evaluations of WASH interventions don't find detectable effects on incidence of diarrhea or other health or economic indicators. The most-likely story is that there are so many vectors for infection that people end up consuming contaminated water despite the WASH interventions (and given that doctors in US hospitals still won't wash their hands regularly, that's very plausible). In that way, WASH has a lot in common with microfinance--single point interventions in complex and broken systems are unlikely to produce large long-term effects.
So the state of play is that the intervention with a clear biological mechanism has no effect and the one with no clear biological mechanism has large effects. I hope I'm not the only one who finds that a bit discomfiting.
So what to make of all of this? The point I made at the conference is that building an evidence base isn't just about methodology but about what is being measured. In the WASH + microfinance space, I think the right metrics are about whether well-functioning markets are being created (see my rant about low-quality equilibria, or my "vaccine or antibiotic" theory of change for microfinance piece) where poor households have more actual ability to choose, including the option to not have to think about whether their water is clean.
A second important point is that there is a long way to go figuring out how to measure things we care about. To that point specifically, Rachel Glennerster and Claire Walsh have a post about the difficulty of measuring women's empowerment via surveys and the limitations of how empowerment is currently being measured. They have some useful specific suggestions for improving the current methods. Perhaps there will be some real traction here, as Glennerster was named the new Chief Economist of DfID this week.
Bonus Overflow Links: David McKenzie has a post about re-interviewing participants in unrelated evaluations. Kieran Healy is writing a book about Data Visualization for Social Science and posting most of the content as far as I can tell.
Week of August 7, 2017
This week's faiV is a fun change of pace of just visualizations & graphics - click through to see.
1. The Global Middle Class: By now, Branko Milanovic's elephant chart should be quite familiar. Nancy Birdsall of CGD has a new post about the state of the global middle class that delves into the elephant chart and other data looking at the state of the middle class globally.
2. Global Inequality: Another chart that may be somewhat familiar but certainly should be top of mind these days. Our World in Data looks at inequality, from a lot of perspectives, here before and after taxes and benefits in developed countries.
3. US Inequality (and Debt): Speaking of inequality before and after redistribution, Catherine Rampell at the Washington Post has a couple of interesting recent posts on policy to help (or not) lower-income workers. The first chart here made lots of waves this week in a post by David Leonhardt, and provides the visceral oomph behind the need to reassess policy in the US. Although this data and similar charts have been circulating for quite awhile, it still thankfully grabs attention.
Whether or not the top chart is related to the bottom chart is one of the questions that Aspen's EPIC is taking on this year. Regardless of the direct connection between income inequality and rising debt, the fact that we are back to record levels of credit card debt seems concerning since it's likely not the .001 percent taking on this debt. That being said, rising debt could also be a sign that finally consumer confidence is returning and people feel that their incomes may start rising again.
4. Statistics GIFS: You can't say I don't know my audience--you guys go crazy for things like this, at least that's what the click data says. The two images at the top are from Rafael Irizarry at Simply Stats, in a post about teaching statistics and how to think about data. Helpfully, the post includes the code to recreate each of the images (and he's got a lot more where these came from).
This week there was also a revival of the Autodesk post about how visualizations can mislead that I featured a while back. It's here again because Jeff Mosenskis of IPA made an underappreciated awesome joke about also being wary of violin plots.
5. Low Quality Equilibria: I couldn't pass this one up when I saw it this week, given my recent rants. Who knew that removing frictions from sharing market information would make it impossible to ever tell if any product was good or not?
Week of August 1, 2017
1. More Ranting (Low-Quality Equilibria and Digital Currency): Following up on my rant last week about the prevalence of low-quality or sub-optimal equilibria because people have such a hard time figuring out what matters, here's another paper that caught my attention because it so thoroughly confirms my priors. The basics: a field experiment provided repair technicians with varying amounts and frequency of feedback. Performance suffered when feedback was weekly versus monthly because the technicians overreacted to each report. In other words, they had a hard time figuring out which details mattered to their own performance. The study could inspire another about "isomorphic mimicry" and the technology of management but I'll save that for another time.
Instead, I'll move on to a different rant about digital finance. In my world, there's only a tenuous connection between the digital finance groups and the cryptocurrency (e.g. BitCoin) groups, but the former certainly should be paying attention to the latter. As Matt Levine put it this week (again, he says this a lot): "The job of the cryptocurrency revolutionaries is to re-learn all of the old lessons of modern finance, one at a time, in public, in embarrassing ways." Right now those old lessons being re-learned seem particularly focused on how hard it is to manage and secure a money supply. I really hope that the digital finance advocates are paying attention to how often various "unhackable" and "secure" cryptocurrencies are being hacked. The spirit of Willie Sutton lives on, and as more "money" is stored in digital form, there will inevitably be more theft. And there's very little reason to believe that average users will employ security practices better than the supposed sophisticated users currently adopting cryptocurrencies. I fear though that the fate of much of digital finance is to "re-learn all of the old lessons of financial services, one at a time, in public, in especially embarrassing ways because they ignored the cryptocurrency movement's repeated mistakes."
2. Global Development (rants): On to more traditional faiV-ing. Kevin Starr has a new rant on the many outside groups making hay over government-funded private schools in Liberia (We need a hashtag to go along with #lantrant, I'm proposing #starrant). Someone once told me there were a lot less education experiments in the US than in other countries because more people were paying close attention and fighting any policy experiments where the outcomes were not already known. That may have been true, but it's certainly not true anymore in Liberia at least. Kevin's plea is to let the Education Minister do his job.
And here's a rant (with a link to another) against the "getting better" narrative that points out how much the world has improved, to the point where it is certainly the best time in history to be alive. I find the argument here pretty annoying, but not annoying enough to rant about myself. Pointing out that fewer children are dying of malnutrition and more people can read (for instance) in no way implies "this is fine."
In fact it's far more common for the "getting better" crowd to argue for more and for taking risks to make more progress, rather than settling for the status quo as Kottke says they are. In that vein, philosopher Peter Singer is probably the best known advocate for doing more, particularly associated with the "drowning child" thought experiment. Except it's not always an experiment. Last week, French philosopher Anne Dufourmantelle died while trying to rescue some actual drowning children. She was particularly known for her work on taking risks.
Week of July 24, 2017
1. Low Quality Equilibria: There's an important "new" (e.g. it's been circulating in working paper form for a while, but is now published) paper in QJE about why hobby woodworkers waste so much money...just kidding, it's about why people keep buying cheap Chinese knock-off tech products and IKEA furniture...actually it's about the persistent use of predatory financial products and poor financial decision making...OK, it's really about the bind that the evidence-based policy movement finds itself in. Well, truthfully it's actually about agricultural markets in Uganda and why adoption rates of fertilizer and improved seed are low, but not zero. Really, that's what the paper is about.
But it is also about all of those other things. Here's the basic story:
Fertilizer and improved seeds boost agricultural productivity substantially. But it's hard for farmers to tell whether the fertilizer or seeds they are buying are fake. So there are lots of people willing to sell low quality stuff claiming it's high quality--in Uganda, the fertilizer is regularly diluted (30% of nutrients are missing) and the "improved seed" is fake 50% of the time. Classical economics tells us that markets will drive out the low quality products as people learn who is a reliable seller; or that the market will collapse and no one will be willing to buy the fertilizer or seeds at all. But farming, like almost every other human endeavor depends on lots of factors, not just these inputs. And so it's not only hard for farmers to tell whether they were sold a "lemon" even even after using it. Did their crops underperform because the were sold fake inputs or because the weather was bad, or they used it wrong, or their land was too degraded, or their were too many of a certain kind of pest, or because they were sick during the planting season, etc.? After all some people buying the fertilizer and seeds did get good stuff and have high yields, so it's even harder to tell where the problem lies. So the market doesn't collapse, and low-quality sellers/products don't get driven out of the market but farmers also--for good reason!--don't invest in the inputs as much as would make sense based on the theoretical productivity boost.
Here's where the rant, and the weird introduction to the item, comes in. This situation is incredibly common: in most of life it's hard to tell whether some input--be it technology, or practice, or advice, or an employee--is high quality before you use it, but also after you use it because of the complex nature of most of life. This basic fact seems to be ignored frequently as researchers, policymakers, and advocates try to explain behavior. In almost all our endeavors we are in a Dunning-Krueger low quality equilibrium. We don't know enough to tell high quality from low quality ex ante, or ex post (yes, I'm a Calvinist). Determining causality is hard--even the most highly trained economists and social scientists get it wrong all the time! What hope does the average human have of looking at a complex system and determining which of the hundreds of factors involved was responsible for what portion of the outcomes? Behavioral economics explanations for sub-optimal choices are tempting because they tend to skirt this core issue. True, cognitive biases and limited attention exacerbate these problems and nudges can yield improvement on the margin, but figuring out what matters is hard (an opportunity to link, yet again, to one of my favorite papers, [Not] Learning by Noticing [the wrong things].
This is why Amazon or any crowdsourced product reviews are worthless. And it's why most people, regardless of their financial literacy, can't consistently tell which financial products are good for them and their situation. And it's why evidence-based policy is such a hard sell--when a policy with strong evidence behind it fails to live up to expectations is that because the advice was bad, the implementation was bad or circumstances changed?
Low quality equilibria are everywhere, defeating them is hard, and that's the sobering challenge we face.
Week of July 17, 2017
1. Weaponized Data and American Inequality (Part 3): We learned a lot in reading the faiV’s summary and corresponding links detailing the minimum wage debate consuming economists across the country. While we haven’t reached our own conclusion about whether a $13 minimum wage in Seattle is or isn’t too high, we are following how some state legislatures across the country are actively rolling back minimum wages established by municipal governments. Example? St. Louis was dealt a big blow and the city has received a lot of press this summer.
(ICYMI the debate, here and here are the two papers that offer opposing outcomes of Seattle’s minimum wage increase. If you don’t have time to read the papers, here’s a fun breakdown from Vice.)
2. Living for the City: CityLab profiled recent research on the intersection of urban development and economic inequality, making us think back to Stevie Wonder’s “Living for the City.” Still relevant. And beautiful. A new study out of the University of Idaho looks at 639 urban counties in the US and the factors that determined when they felt the effects of the 2006-2010 recession. Rarely do we see the Gini coefficient being used in the context of domestic inequality – but we should use this metric more often. Consequently, we were really excited to see this interactive map of the Gini coefficients of counties across the US.
For more on cities, another CityLab piece looks at how housing policies worldwide will only exacerbate urban inequality and housing crises. And this story on how inefficient tax codes, high cost of living, and migration, by both companies and residents, are sending the state of Connecticut spiraling, makes us rethink how we view the fiscal policies of traditionally blue, wealthy states.
3. Income Volatility, Short-Term Savings, Retirement (Oh My): Over the last 18+ months, our team has conducted a deep dive on both the impact income volatility – large fluctuations in week-to-week and month-to-month income – has on US households and potential solutions for mitigating the problem. Our latest briefs look at the role wage insurance could play in helping families cope with job loss or reduced wages and how shortfall savings can serve as a buffer during financial emergencies.
Because we care about both short-term financial stability and long-term security, we also spend our days thinking about comprehensive policy solutions to help expand access to retirement savings opportunities. In our process learning about more about income volatility, we’ve realized it’s particularly hard to save for the long-term when short-term savings are lacking. This new paper looks at the effect income shocks have on retirement savings (the stats aren’t pretty: “96 percent of Americans experience four or more income shocks by the time they reach 70”), and *mark your calendars* later this fall, we’ll be publishing two papers on how volatility affects retirement savings.
Week of June 26, 2017
1. Weaponized Data and American Inequality: Last week I linked to a paper finding minimal effects from minimum wage increases, unaware that a huge explosion of debate on this issue was about to occur. If you follow these things at all, you know that last Friday a paper on Seattle's minimum wage increase was released finding no job losses or cuts in hours. Monday, a different paper finding large losses for households with minimum wage jobs was released. There's a whole lot out there now on the two papers so I'm not going to rehash those arguments (if you need to catch up, try this or this or this or just scroll through Twitter). I want to focus on the backstory of why there were two papers released so close to each other because it's important for the future of research and policy-making. As detailed here, what appears to have happened is researchers at UW shared an early draft of their paper (using tax data that is rarely available in minimum wage studies) with the Seattle mayor's office. The mayor's office didn't like the conclusions so asked a different set of researchers to write their own paper--and release it just before the planned date for release of the UW paper. While I have no special insight into the exact details of what happened, the prospect that the report is accurate disturbs me a great deal. It's a blatant step toward what the author of the Seattle Weekly piece calls "weaponized data." Be afraid for evidence-based policy. Very afraid.
In other American inequality news on topics that yield strong confirmation bias reactions, Justin Fox reports on new work suggesting that occupational licensing actually crowded-in historically disadvantaged workers--seemingly the transparent rules of licensing reduced formal and informal discrimination that kept these groups underemployed. That's a very plausible story to me, though I generally also buy the anti-licensure arguments.
There's also new work on school vouchers, from Indiana, finding short-term declines in test scores, but later (over four years) gains. It's worth noting how claims for vouchers have down-shifted to "no harm and some students gain." But keeping on the weaponized data theme, the paper is not publicly available and was only obtained by ChalkBeat through public records requests. Apparently the study authors don't think it should be public until it's peer-reviewed, which illustrates the difference in norms in sociology and economics.
2. Our Algorithmic Overlords: Also a few weeks ago I linked to a story about how to tell if borrowers on online lending platforms were going to default, and to the book, Everybody Lies, from which it came. I said I was going to read the book and I started this week--and was immediately dismayed. The opening of the book discusses what search data--particularly searches on pornography websites--can tell us about Americans' hidden desires. You can see a summary in this deeply disappointing Vox piece (isn't Vox supposed to be better at thinking critically about this stuff?). There is no discussion of how such data might be biased or inaccurate, how a site's interface may interact with what people search for, or why we should believe that search data closely corresponds to "real life." In other words, it's an object lesson in the dangers of using data and algorithms without understanding the data or the people, social structures and institutions that generate it. So of course it's a best seller. Suffice it to say that I have radically revised down my faith in any of the book's conclusions.
In other data-generating processes of uncertain usefulness news, Google will stop showing ads inside Gmail based on scans of email content (illustrating the sucker's game that is attention, I had no idea they were still doing this; I hadn't noticed an ad in years). The nominal reason is combating hesitance from corporates to adopt Gmail and Google's suite of web apps. As someone in my Twitter feed noted, the real reason is that Google already gets better information to drive ads to you than your email.
Week of June 19, 2017
1. Indebtedness: A few weeks ago I mentioned the wave of agricultural loan waivers in a variety of Indian states, a pattern that has been repeated over decades (and not just in India; and perhaps I should say repeated over millennia) with all sorts of moral hazard implications for lenders and borrowers (here's Xavi Gine explaining the impact of the 2008 agricultural debt relief program). Shamika Ravi looks at data from the current round of farmer distress examining how poverty, indebtedness and political power interact since straightforward explanations don't hold up to scrutiny.
2. Our Algorithmic Overlords (and some Data Viz): Sometimes it's helpful to take a step back and see where artificial intelligence is still struggling. Reassuringly while AI can negotiate it still produces aphorisms like: Death when it comes will have no sheep. But maybe that's a negotiating tactic? Meanwhile, apparently machine learning still struggles to tell the difference between labradoodles and fried chicken (I suppose that would be more frightening than funny to chickens and labradoodles).
And while not about algorithms, here's another one of those cool illustrations of how data visualization influences how we interpret data that are so popular.
3. American Inequality: One of the clear themes of recent research on poverty and inequality in the United States is the rise of month-to-month and year-to-year volatility of incomes, while real wages have stagnated. The safety net in the US, such as it is, is especially unable to deal with income volatility. Here's the story of a family in Texas with volatile income who has adopted a number of medically fragile children: because of the way the state administers Medicaid the family has to re-certify eligibility almost every month. While this is somewhat unusual, the language of the Senate Republicans healthcare/Medicaid legislation would enable states to require all recipients to re-certify eligibility monthly.
Meanwhile here's Cengiz, Dube, Lindner and Zipperer with a new look at the perpetual question of what raising minimum wages does to jobs, finding little evidence for job losses or labor substitution. And here's a piece from HBR on the household effects of unstable work.
Week of June 12, 2017
1. St. Monday, American Inequality and Class Struggle: One of my favorite things about writing the faiV is when I get the chance to point readers to something they would likely never come across otherwise. So how about a blog post from a woodworking tool vendor about 19th century labor practices, craft unions and the gig economy? Once you read that, you'll want to remind yourself about this piece from Sendhil Mullainathan about employment as a commitment device (paper here), and this paper from Dupas, Robinson and Saavedra on Kenyan bike taxi drivers' version of St. Monday.
Back to modern America, here's Matt Bruenig on class struggle and wealth inequality through the lens of American Airlines, Thomas Picketty and Suresh Naidu. I feel a particular affinity for this item this week having watched American Airlines employees for a solid 12 hours try to do their jobs while simultaneously giving up the pretense that they have any idea what is going on.
2. Our Algorithmic Overlords: Facebook is investing a lot in machine learning and artificial intelligence. Sometimes that work isn't about getting you to spend more time on Facebook...or is it? With researchers at Georgia Tech, Facebook has been working on teaching machines to negotiate by "watching" human negotiations. One of the first things the machines learned was to "deceive." I use quotes here because while it's the word the researchers use, I'm not sure you can use the word deceive in this context. And that's not the only part of the description that seems overly anthropomorphic.
Meanwhile, Lant Pritchett has a new post at CGD that ties together Silicon Valley, robots, labor unions, migration and development. And probably some other things as well. If I read Lant correctly, he would approve of Facebook's negotiating 'bots since negotiation is a scarce and expensive resource (though outsourcing negotiation is filled with principal-agent problems). I guess that means a world where robots are negotiating labor contracts for low- and mid-skill workers would be a better one than the one we're currently in?
3. Statistics, Research Quality and External Validity: Here's another piece from Lant on external validity and multi-dimensional considerations when trying to systematize education evidence. A simpler way to put it: He's got some intriguing 3-dimensional charts that allow for thinking a bit more carefully about likely outcomes of interventions, given multiple factors influence how much a child learns in school. It closely parallels some early conversations I've had for my next book with Susan Athey and Guido Imbens, so I'm paying close attention. And if you can't get enough Lant, you could always check out my current book. Yes, both of those sentences are shameless plugs.
Week of June 5, 2017
1. Social Enterprise: A few weeks ago I noted that Etsy was under pressure from an activist investor for behaving like a B Corp (which it is (was?)). I missed the notice that the investor won: Etsy layed off 80 employees and fired the CEO/Chairman. Here's a piece reflecting on the Etsy saga that is emblematic of much of what I think is wrong in social enterprise rhetoric. The argument that social enterprises have to be ruthless competitors may sound good (to some) but it ignores the exact issue that is at the heart of social enterprises: how do you manage the trade-offs. It's worthless--less than worthless, I should probably say "actively harmful"--to pretend there are no trade-offs or to imply that there is value in advice like "be ruthlessly competitive except for in these parts of your business model." It's why efforts like B Corporations that don't have any governance teeth are a distraction, and why even efforts life For Benefit Corporations that do have governance teeth are fraught.
In other social enterprise-ish news, I can't resist a story about a star rapper, off-grid solar power in Senegal and Chinese investors. You can't either can you? On a more practical level here's Devanshi Vaid on the lack of information flow on social enterprise in India.
And here's Felix Salmon with some remarkably clear reframing of an important wing of social investment: if a foundation endowment can't get high investment returns in the near term, don't cut back on grantmaking, accelerate it!
2. Our Algorithmic Overlords: The Atlantic has a long piece on how cryptocurrencies like Bitcoin, purportedly designed to limit centralized authority, actually can become tools of authoritarianism. You don't have to go all the way to cryptocurrencies though, as I try to frequently point out. Digital currency of any sort can easily become weaponized by authority, even authority that isn't fully authoritarian.
I wasn't sure whether to include this in "Social Enterprise" or "Our Algorithmic Overlords" because it's a bit of both, through an extraordinary lens: Venezuela's bonds. As Matt Levine relates, Goldman Sachs (sort-of) bought some bonds from Venezuela (sort-of) that (sort-of) prop up an authoritarian government apparently bent on starving people. But no one is really responsible for this decision because of the way governance of the investment funds is set-up and which all point back to an index by which fund manager performance is measured. (I know, this is confusing and complicated, but it's worth it). In this case everyone is pointing to some arbitrary set of decisions as responsible for their behavior and denying any responsibility for moral judgment. If we struggle with these issues already, how much worse are they going to get with the arbitrary set of decisions are made by an algorithm that we don't really understand?
But people are more worried about algorithms driving their cars, than about algorithms ruling their moral decisions.
Week of May 29, 2017
1. Income Instability and the Cost of Living: Those who have studied financial management among low-income people know that instability and unpredictability of income are a main source of difficulty. The U.S. Financial Diaries project and book bring this challenge to life. This week Jonathan Morduch is quoted in the New York Times examining how the norm of a steady paycheck has been replaced by turbulence, affecting pretty much everyone who makes under $105,000 per year.
Speaking of $105,000 being a new marker for financial challenges, according to the U.S. Department of Housing and Urban Development’s new income limits, in parts of the high cost Bay Area, a family of four earning $105,350 is considered low income. On the other extreme, in the recent CGAP national survey of smallholders in Bangladesh, 75% of households reported that an annual income of $1,524 would cover their household needs.
2. A Raise or More Frequent Paychecks?: Here’s an interesting fact: More than one in four millennials prefer real-time pay to a raise. A report by the Aspen Institute’s Expanding Prosperity Impact Collaborative (EPIC), employers and governments are exploring new options for workers to collect their pay more frequently and when they need it. The jury is still out on the effects for managing household finances. We suspect this might be be helpful for short-term volatility management, but may result in difficulty with long-term savings and may diminish the commitment element that some people prefer in being able to keep money at a distance under some circumstances. Let’s hope Uber and Lyft are learning from all the data they have!
3. Global Tech Trends: TechCrunch summarized the 2017 Internet Trends report this week, and there are lots of great insights here. With over 700 million mobile internet users, the volume of mobile pay in China doubled last year to reach $5 trillion. There are 3.4 billion internet users in the world, up 10% since last year. As consumers increasing look online for shopping and retailers offer customer service through “chat”, we are curious what these trends will mean for call centers, a big industry for several developing countries, like the Philippines. There are also increasing apps developed in emerging markets, such as Kampala’s fast-growing Safeboda which allows riders the convenience of the common motorcycle taxi with the promise of a trained and safe driver.
Meanwhile, on-demand car and bike services are exploding in China and throughout Southeast Asia. Bike-for-rent services experienced a 100% month on month growth, reaching 20 million users in 2016 (a scale which could mean meaningful reductions in CO2 emissions. Helpful given recent developments in US politics!).
Week of May 22, 2017
1. The Value of Management: If you pay any attention to the development economics world, you were probably already aware that there was unrest at the World Bank since Paul Romer became Chief Economist. Yesterday that unrest came out into full public view with stories about Romer being relieved of management responsibilities for the Development Economics Group. The news stories make everyone look bad, and don't reflect my experience with the parties involved (which is admittedly quite limited). But rather than adjudicate any of the issues, I'm going to pivot to my ongoing amazement that economists of all people seem to have so little appreciation of the value of management and specifically specialization in management. It's a learned skill! The idea that someone should be managing a department of more than 600 people because they happen to be a leading economist is bonkers.
Just look at what a little bit of management training for school principals can do for schools and test scores. Or what professional management training can do for quality of care in hospitals. That's right, management can save lives! Here's hoping that skilled management will advance the very legitimate goals of clear and useful communication in Bank reports. I can't be the only one glancing through the stories about the gender studies hoax paper and thinking it wouldn't be that hard to do the same thing for a World Bank research report.
In closing, I'm not good enough of a person to avoid noting that "and" is 16% of the World Bank's actual name and linking to Ryan Briggs' Drunk World Bank twitter account.
2. Immigration: If you weren't distracted by counting the number of "and"s in your latest piece of writing, you may have seen another controversy bubbling up in social media: Michael Clemens and Justine Hunt have a new paper suggesting that Borjas' finding of losses for low-wage workers from the Mariel boatlift are actually a result of a change in the composition of wage survey samples. Borjas responded first by accusing Clemens and Hunt of being tools of Silicon Valley open border enthusiasts--and essentially saying that no grant-supported research can be trusted--and only later with an attempt to defend his results with data. That attempt looks plausible until you realize that he ends up charting the outcomes for less than 20 people. David Roodman--whose earlier work on this specific issue Borjas also managed to slander by calling it "fake news"--weighs in with some typically substantive and clear points (maybe he could do some coaching for World Bank writers?). The major one from my perspective being: Borjas already had to pick through data to find a narrow slice of the population that might have been negatively affected by sudden mass immigration, and can only defend that result with a sample better suited to a local news broadcast than serious economic inquiry.
If this kind of thing fascinates you, rather than tires you, Borjas has an additional reply that is more substantive and ultimately arrives at a useful point. But the process to get there remains bizarre.
In other immigration news, here's a look at the effect of differing state approaches to immigration law enforcement, and here's an animation of Mushfiq Mobarrak making the case for the gains from migration.
Week of May 15, 2017
1. Data (and Our Algorithmic Overlords): Many of you probably saw the Economist piece on data becoming the world's most valuable resource. It does a darn good job at producing conflicting reactions for me: Yes, we should be paying more attention to the accumulation and use of data among private companies! But governments aren't to be be trusted with this kind of data any more than private companies! And you're spending way too much time in Silicon Valley--we're a long long way from data being more valuable than physical resources for most of the people in the world!
So I'm going to use it mostly as a foil to introduce two pieces that you should read that you probably don't think are relevant to the faiV. First, here's a piece by Ted Knutson, a protagonist in the development and use of "advanced statistics" in football/soccer, about why he developed and continues to use a terrible visualization of data to evaluate player performance. Second, here's a piece about how adapting behavior based on data in baseball has helped some players but hurt others so that there is zero net gain. The point here being, understanding data is hard enough. Using data is even harder. Figuring out how to help people change based on data--without just turning everything over to our algorithmic overlords--is the toughest of all. And if you don't believe, that let me remind of you of one of my all-time favorite papers about seaweed farmers. Take that, "vast empirical literature"!
2. Theories of Change (and Demonetization): In my book of interviews of development economists on RCTs etc. the throughline is theory of change. How do ideas get translated into policy and into making the world a better place? I argue that a lot of debate about methodology is really debate about theory of change, particularly around the role of experts and the value of small vs. large changes. This Planet Money episode about the Indian demonetization has the most jaw-dropping "theory of change" story I think have ever encountered. The short version is an engineer developed--through divine inspiration--a model of the Indian economy, complete with cheesy illustrations, and just kept talking about it until a powerful politician took notice and decided to introduce one of the biggest economic shocks in modern history. If you know someone graduating from high school or college, perhaps you should make them listen to the episode rather than buying them a copy of Oh, The Places You'll Go. (Oh, and that feeling when you visit the Smithsonian with your kid and get to talk about how even our heroes fail us.)
3. Digital Finance: Over at CGAP, IPA has a post about fees for 21 mobile money services in seven different countries, with an eye to how the highest fees are paid on the smallest transactions, presumably serving as an effective tax on the poorest customers. This of course is the same issue we've been talking about in microfinance for decades: small transactions don't cost less to process than large ones and so small transactions are more expensive. While it's less of an issue in things like digital services than in-person services it doesn't entirely go away and so providers have to make decisions about whether they are going to over-charge their relatively wealthier clients to subsidize their poorer ones, or tax their poorer ones for their inability to transact in larger amounts. The problem with the former is that there is almost certainly going to be a competitor who is willing to take those wealthier clients by not asking them to subsidize costs for smaller transactions.
This also raises one of my long-term fascinations: people tend to react strongly to poor customers being charged more for financial services but not for telephone services--even when it's the same company doing it! The same poor customers who are paying more for mobile money transfers are almost certainly paying more for cellphone minutes by buying them in small increments, but I don't ever see that being charted.
Week of May 8, 2017
Editor's Note: You might think you've seen this announcement before, but it's different. The US Financial Diaries project has an upcoming free webinar on May 16th. If you're in the New York City area, join us on May 23rd at New America NYC.
1. American Inequality: The exceptionalism of the United States in promoting home ownership asthe signifier of middle class status and/or upward mobility, and a generally accepted keystone of building wealth has persisted despite the Great Recession/housing crisis. But that doesn't mean that things haven't changed--the availability of housing that costs less than 30% of a household's income has dramatically decreased. Matt Desmond, author of Evicted, writes in the New York Times magazine that the American emphasis on home ownership has become one of the primary engines of inequality. Non-profits--or at least how we measure and fund them--are another (unintended) engine of inequality. In New York state, non-profits pay wages just above retail and food service (and 80 percent of these workers are women, and 50 percent people of color).
2. Our Algorithmic Overlords: The goal of machine-learning and using algorithms to analyze data is to yield better decisions, at least better than human beings would make given biases and the challenges of causal inference. A(nother) new book looking into the way this works is Everybody Lies. I haven't read it yet, but I'm looking forward to it. In the meantime, there's an excerpt in the Science of Us, taking a look at one of those areas that humans always struggle to make good decisions: who is credit-worthy. The substitution of bias against minorities (or at least people different from the loan officer) and the poor for careful judgment is well documented and wide-spread. Netzer, Lemaire and Herzenstein turn the machine loose on data from Prosper, an online platform for peer-to-peer lending, and find that the words that borrowers use are predictive of repayment behavior. You should read the whole excerpt because it does focus on the unintended consequences of using machine learning and big data. I, of course, immediately wonder how quickly borrowers and lenders will adapt to the findings.
Meanwhile, here's a Quora forum with Jennifer Doleac on the American criminal justice system, which dwells a lot on how machine learning is affecting decisions in another area humans have a lot of trouble with: who's guilty and who is a threat for recidivism. And of course, on the unintended consequences of our efforts to punish people. And here's a speculation that Donald Trump is a dynamic neural network/machine-learning algorithm with narrow goals. Here's an alternate version of the same argument, which in addition to being even more frightening, provides additional insight into the potential unintended consequences of data analysis without theory (of Mind).
3. Digital Finance: The item on Prosper and algorithms determining credit-worthiness based on language used by borrowers is about digital finance of course. But in the domain of more traditional ways of thinking about digital finance, here's a story about M-Pawa in Tanzania, interesting for it's integration of savings, lending and education. The bottom line: more savings, larger loans, better repayment. In other news, M-Pesa is supporting proposed regulations for cross-platform transfers in Kenya. And MicroSave has some ideas on how to enable digital finance among the illiterate, since traditional approaches to inclusion through digital have the unintended consequence of excluding the illiterate.
More specifically on the "unintended consequences" theme, though having relatively little to do with digital finance, here's some new research on how global de-risking in banking has cut the number of correspondent banking relationships (what makes cross-border payments even somewhat efficient) have declined by 25% since 2009, pushing whole regions out of the regulated banking sector.