1. Communications: Marc Bellemare has a new post on how to communicate research titled "The Goal of Scientific Communication Is Not to Impress But to Be Understood." To which I say, the goal of human beings is not to be understood but to impress (hence the faiV). But assuming that you aren't as Calvinist as I am, I've been collecting a few things over the last few weeks that broadly fit the theme of better communicating research and ideas. Here's an experiment on disaster relief communications testing negative and positive imagery for their effect on donations and on donors sense of that change was possible. Unfortunately, there are few conclusions to draw; these are hard experiments to run. Here's a piece from ODI on 9 things you are doing, but shouldn't in research communications. I'm guilty of at least five (with mitigating circumstances, e.g. the funders told me I had to).
But let's get specific. Here's something you should definitely not do: produce a set of guidelines for behavior that have no input from the most important people in the equation. You should also not try to write jargony, provocative headlines without really understanding the context, for instance, saying that "40% of Older Americans Will Experience Downward Mobility." Given that the standard models of retirement planning assume that everyone retiring will have a lower income (hello there Lifecycle theory!), and most people aren't close to saving enough for retirement according to those standard models, I'm willing to bet a lot of money that the figure will be a lot higher than 40%. Don't try to find some way to contextualize a massive ritual sacrifice of children. And finally, definitely don't be one of these Manhattanites caught on video expressing revealed preferences for segregation and inequality, but do be like the principal at the end of the video clip and communicate your disgust in no uncertain terms.
2. Global Convergence: But not in a good way. I often think about the divergence in outcomes (or put another way, growing income and wealth inequality, falling mobility) for Americans as a convergence: for the bottom ~40% of the income distribution, the American economy looks a lot more like the economy in, say South Africa or Brazil, than the economy experience by the upper half of the distribution. That clip above is one example of how far out of reach the tools for mobility can be. Justin Fox has a story about fee-based governance in the United States--government agencies funding themselves through fines and fees. Justin makes the connection to the Gilded Age in the US, but it's a mechanism that will be very familiar to people in developing and middle-income countries. For a ray of hope on that front, you can check out Tishuara Jones, Treasurer of St. Louis, who is fighting back against fines and fees as revenue in her city.
3. Household Finance: This week I guest-taught a class at Haverford on US microfinance. In the post-discussion I learned that students prefer off-campus jobs, because Haverford pays student workers only once-a-month, and those who need the paycheck from a job during the semester, need it more frequently. That makes sense. But people on low-incomes also often prefer infrequent payments, so as to get larger lump-sums. Dairy farmers in Kenya do according to this new work from Casaburi and Macchiavello. To the convergence point earlier, this isn't a difference between the US and developing countries. The demand for income spikes among people in the US can be seen in the low take-up rates for monthly EITC payments, and the high take-up of "overwithholding." It's also evident in the fintech Even's pivot away from consumption smoothing. The bottom line is we still have a long way to go to understand optimal income volatility and we should have weak priors about the interest in and benefits of say, on-demand income or a "rainy day EITC."
Week of April 16, 2018
1. Read, Synthesize, Repeat: Two weeks ago I featured a bunch of links about new and new-ish research about cash transfers, including a synthesis by Berk Ozler which particularly draws attention to the growing evidence of negative spillovers from cash transfers. This week Justin Sandefur wrote up his own synthesis, which disagrees with Berk in important ways, and followed up with a Twitter thread summary, which includes the amazing line: "unless cash recipients literally spent the money on gasoline to set fire to their neighbors farms..." Which of course led to a response from Berk and then lots of further replies--much of which center on how to think about the scope of negative spillovers and what to do with data that doesn't seem to be entirely trustworthy. That's the job of synthesis! But there's a long way to go before there's any consensus on the right synthesis.
The site Straight Talk on Evidence has been working on, if not synthesis, at least part of the work of synthesis, sorting through lots of research on US policy interventions and whether it holds up. A few weeks ago they started a series of blog posts on what the path forward should be "when most rigorous program evaluations find disappointing effects." Here's part two with their proposed steps (I try to avoid using the word "solution" even when it's just quoting others). And here's Chris Blattman's Twitter thread response to their proposed steps.
I may have already linked this but in case I didn't, it's relevance to this conversation in particular compels me to include it: The Political Economy of RCTs. Equally I have to include this short article titled "Evidence-Based Claims About Evidence" which challenges the conventional wisdom on how long it takes for evidence to influence physician behavior.
And yes the connection is tenuous, but here's Ideas42 first ever Impact Report on their first 10 years of work. I think there remains a lot of work to be done on synthesizing behavioral science and other approaches and the real world.
2. Banking: When I first started working with Jonathan at FAI, one of the first things was helping get the book Banking the World out the door--based on work by Jonathan and others estimating that "half the world is unbanked." The World Bank's Findex database has just been updated with 2017 data, with a new report and complete data, and it now seems that the proper statement is "a third of the world is unbanked." Of course, that begs the question of what we mean by unbanked or financial inclusion, and how to think about people who have access to formal accounts but choose not to use them--often because those formal accounts aren't as useful as the alternatives (or in some cases are actively harmful). Obviously, the Findex has a lot to explore and I'm sure I'll be sharing more in the coming weeks as people try to synthesize the findings.
But coming back to that point about how to think about financial inclusion and exclusion, here's the text of a speech from N.S. Vishwanathan, Deputy Governor of the Reserve Bank of India, about evolving regulation of Indian banks and stressed assets, which closes with an all-too-familiar warning: "There appears to be taking hold a herd movement among bankers to grow retail credit and the personal loan segment. This is not a risk-free segment and banks should not see it as the grand panacea for their problem riddled corporate loan book."
Meanwhile, the US Consumer Financial Protection Bureau under Mick Mulvaney has drastically cut back it's enforcement actions, apparently to zero. The latest is dropping charges and sanctions against an abusive payday lender and scaling back regulations of high-cost consumer lending. Perhaps Mick should place a call to India.
3. Philanthropy: Discussions of philanthropy would be improved if there was more synthesis of public choice economics--too often I see writing about philanthropic actors that seems to start with either an assumption of saintly altruism or evil capitalist intent in disguise. A reasonable example of something better is this new report on "what goes wrong in impact-focused projects" and finds roughly half of the "roadblocks" are funder-created obstacles.
Another example is an important set of stories about the Silicon Valley Community Foundation, which has become one of the largest foundations in the world, that illustrate that the world of philanthropy is even messier than most human endeavors where altruism, good intentions, power and self-interest collide. Marc Gunther, writing in the Chronicle of Philanthropy, details many accusations of abusive behavior by SVCF's leading fundraiser, who has resigned in the few days since the article was published. There was a lot of work to get the story published, as Marc details here on his own blog, but like so many other "revelations" in this season, the accusations were well-known and apparently ignored by a great many people, including allegedly by the president of SVCF, Emmett Carson. SVCF is no stranger to controversy. Though I've linked these before, as a refresher here's Marc's earlier reporting on SVCFs' role, or lack thereof, in Silicon Valley itself and an excellent piece by Phil Buchanan of CEP on how to think about community foundations' role in the complicated world of philanthropy. And here's Rob Reich (the Stanford political scientist, not the Berkeley Economist) on interrogating the power of large philanthropy.
Week of April 9, 2018
1. Global Development: Hey, does anybody remember the Millennium Villages Project? It seems an age ago in terms of development fads, now that we're all focused on cash grants and graduation programs, and according to some papers would fall into the "long-run" category. Andrew Gelman has a post about a new retrospective evaluation of the program (that he participated in), including a link to an evaluation of the evaluation. The results are surprisingly good, given what I expect most people's priors were at this point. Though I suppose the TUP evaluations should perhaps have shifted those priors in a positive direction. I guess I'm kind of surprised that the results don't seem to have gotten the attention I would have predicted. Of course, I don't think anyone has argued that the MVP should be a model for other programs since Nina Munk's book, so maybe I shouldn't be so surprised.
Lant Pritchett has a list of six other things in development that people aren't paying (enough) attention to, mostly variations on the continuing large gap between even the lower part of the income distribution in rich countries and the upper part of the distribution in poor countries.
Lant's first point is about the huge gains from moving. Here's a piece from a few weeks ago about the lack of geographic mobility, specifically rural to urban migration, in the United States where the overall tone is exasperation at these benighted people who stay in small towns (and ruin things for everyone else; it's an interview with Robert Wuthnow about his new book). It caught my eye because I can't imagine something like this being written about rural people in developing countries (without touching off a lot of blowback). But perhaps we should see more stuff like this about all forms of poor-to-rich geographic mobility. Speaking of those rural people, here's a new paper from Marc Bellemare about one of the dynamics that may be keeping the poorest people in rural areas (at least in Madagascar)--the intensification of income from agriculture.
2. Jobs: Last week I linked to the recent study of scheduling practices at The Gap that found that encouraging managers to set more stable schedules for retail employees led to higher productivity and sales for the firm. The exact mechanism for increased sales isn't completely clear, but it appears that managers shifted hours to more experienced workers, who unsurprisingly were more productive. While the study is encouraging overall--stable schedules are better for (most) workers and for employers--it also has a dark tinge. To see why, consider this Atlantic article about the future of jobs at Walmart (which, to its great credit, was well ahead of The Gap in experimenting with more stable schedules for its hourly workers, and other efforts to stabilize workers income). The macro trend is toward fewer jobs, at least in terms of how we used to define that term, for less-skilled and less-experienced employees, and declining job quality for those people. That's been happening at many companies (think of outsourcing of janitorial, security and similar jobs) for a long time. It seems an awful lot like what I understand has happened in European labor markets which are more regulated--stable jobs are limited, more workers, particularly the young pushed into contingent labor contracts with limited benefits, stability or security. From a distance this is fascinating: similar outcomes from radically different processes. But from a policy perspective it's frightening. In the economic development world, we've been talking for a long time about how to move more people into formal employment, like in developed economies. Meanwhile the developed economies are making great progress moving people into informal employment, like in developing countries. Maybe I should have called this item Global Undevelopment.
And to play to the academic part of my readership for a moment, here's a piece about how every effort to create better incentives in academic jobs makes things worse. I remain baffled at the general assumption in economics that managers know what they are doing, given the management they experience on a daily basis. While I can't vouch for the management abilities at the Open Philanthropy Project, chances are if you're a reader of the faiV you, or someone you know might be interested in these job openings.
3. MicroDigitalFinance: Is a neologism a step too far? Probably. But check out CFI's fellows program research agenda. There's a whole lot of "microdigital" there. Interestingly, to me at least, is that you could copy and paste these questions into a research agenda for the US financial services marketplace and no one would bat an eye, especially the ones about the changing nature of work.
Week of April 2, 2018
1. Global Development: To start us off, how about some rain on the "rising Kenyan middle class" parade? The core point--that gains from rising incomes that don't translate into durable assets can rapidly be erased, a perspective that should sound familiar to anyone with a passing knowledge of anti-poverty policy in the US.
But the real parade in global development in recent years has been on the value of delivering cash to poor households. This is a train that's been picking up steam for a long while. I would date the current push back to the first studies of Progresa/Opportunidades, the Mexican conditional cash transfer program. Momentum has steadily built around both the positive impact of cash transfers--that recipients don't waste the money, that they use the money productively--and dropping conditions. That momentum was built on many studies, but probably the two most well known in international circles are Blattman, Fiala and Martinez on cash transfers in Uganda, and Haushofer and Shapiro/GiveDirectly in Kenya. Both showed significant gains by recipients of unconditional cash.
Both of those papers were about relatively short-term effects. Both studies included longer-term follow-ups. And you know what's coming: the large positive effects seem to have disappeared in the medium term. Berk Ozler of the World Bank is currently playing the role of Deng (it's the closest I could get geographically) with two lengthy blog posts. The first, keying off comments from Chris Blattman in the recent Conversations with Tyler, but really delving into the recently released update to the Haushofer and Shapiro/GiveDirectly update is the important one for non-specialists. The second is very useful for understanding the specific details of interpretation. The posts also kicked off a number of useful Twitter conversations (here, here, here, here and here, though that's just a sample; just scroll through Chris's and Berk's timelines for more). Berk's first post also takes on the role that academics have played in stoking that momentum and is worth a close read.
I think it's also important to think through what is happening with cash transfers in light of not only of other studies of cash (like this one finding positive effects on the personality of Cherokee Native American kids whose families receive cash that was just officially published) but also other interventions. Deworming is one example--one big source of the controversy over the effects of deworming is that there isn't a medium-term biological effect to explain the the long-term economic effects. The Moving to Opportunity study is another--no short-term or medium-term gains, only long-term ones. And I have to note that the Native American paper is a frustrating example of Berk's critique of the role academics can play in raising expectations too high--the paper's title and abstract simply reference a large positive effect of cash transfers with no indication of when (now? 10 years ago? 30 years ago?), where or who the participants are, or even the size or mechanism of the transfers.
Week of March 19, 2018
1. Household Finance, Debt Specifically: This week I had the chance to talk about the moral dimensions of debt with Fred Wherry, as part of Aspen EPIC's focus on consumer debt in the US (and there are more conversations about debt before and after in that video). One of the things that doesn't get mentioned in the video is that the ancestor of mine who was rescued from debtor's prison later became the official Collector for Jersey City. It's a topic that fascinates me because attitudes toward debt vary so widely across time, culture, context and individual. It often seems like perspectives on debt are pulled from the Wheel of Morality. Just the selective use of the words "credit" and "debt" could be fodder for 100,000 words or more, much less the tension between the lack of access to credit coinciding with troubling debt burdens in many contexts.
To get up to speed on the current situation with consumer debt in the United States, you couldn't ask for a better overview than Aspen EPIC's just published primer. Well, you could ask for one, but given the gaps in the underlying data, you wouldn't get it. And to push some more moral buttons, here's a profile of one of the most influential figures in consumer debt today: Dave Ramsey. If you don't know who that is, you really do need to read the profile.
2. Microfinance and Digital Finance: I suppose I'm sending a message by increasingly conflating these two categories. This piece from NextBillion on the need for Indian MFIs to digitize at least gives me an excuse this week. But while I figure out what message I'm sending (or at least intending to send), here are a couple of recent pieces about digital accounts helping people save more. First, a paper from the job market that I missed about M-Pesa boosting savings among those whose alternatives were most costly. And a new paper about an experiment with female entrepreneurs in Tanzania finding digital savings accounts boosted savings rates. My priors aren't shifted much by these, but they are shifted some.
To maintain some strategic ambiguity, here's a new paper that fights the digital invasion--there's nothing less digital than grain storage. Providing farmers with a way to communally store grain at harvest has high take-up and as a result were able to sell grain later at a higher price. An intervention to allow individual cash savings for inputs was less successful, though possibly because there wasn't much margin to improve on.
3. Methods and Economics: It took a lot of willpower (though apparently not ego-depleting) not to put this item first, but I worry that my excitement over things like this is not normative for the faiV readership. But for those of you in this niche, here's a new comment from Guido Imbens on the Cartwright and Deaton critique of RCTs (and if you prefer a simpler version, here's my interview of Deaton for Experimental Conversations which gives an overview of most of the issues). To give you a flavor of Imbens perspective: "Nothwithstanding the limitations of experimentation in answering some questions, and the difficulties in implementation, these developments have greatly improved the credibility of empirical work in economics compared to the standards prior to the mid-eighties, and I view this as a major achievement by these researchers."
Imbens places RCTs within "the credibility revolution" in empirical economics (which of course is the crux of the debate--how much do RCTs improve credibility?). The credibility revolution, in turn, has played a big role in the growth of empirical economics compared to theory and econometrics. Here's Sylvain Chabe-Ferret with an overview of "the empirical revolution in Economics", some thoughts on the path forward and a treasure trove of links. I have to note here, for those not so enmeshed in the details, that while Deaton is a critic of RCTs, he is a part of the credibility/empirical revolutions through his careful and detailed work with surveys.
Finally, here's something form the Royal Economic Society with the headline "Tweeting Economists Are Less Effective Communicators Than Scientists". I haven't read it yet but how could I not link it when it has such an exquisite combination of direct and implied slights on economists?
Week of March 12, 2018
1. Microfinance and Digital Finance: Apparently the "farmer suicide over indebtedness" hype train is kicking up again in India. That's not to imply that farmer suicides are not a serious issue. But Shamika Ravi delves into the data and points out that indebtedness doesn't seem to be the driver of suicides and so attacking lenders or forgiving debts isn't going to fix the problem. Certainly poverty and indebtedness add huge cognitive burdens to people that affect their perceptions and decisions in negative ways, including despair. Here's a new video about poverty's mental tax--there's nothing new here, but a useful and simple explanation of the concepts.
Last year (or the year before) I noted Google's decision to play a role in safeguarding people in desperate straits from negative financial decisions: the company banned ads from online payday lenders, in effect becoming a de facto financial regulator. This week, Google announced another regulatory action. Beginning in June it will ban ads for initial coin offerings (if you don't know what those are, congratulate yourself). While I'm all for the decision, it's strange for Google to conclude that these ads are so dangerous to the public that they should be banned, but not for three more months. Cryptocurrency fraudsters, get a move on! Meanwhile, the need for Google and Apple (and presumably Facebook, Amazon, Alibaba and every other tech platform) to step up their financial regulation game is becoming clearer. In an obviously self-promotional, but still concerning survey web security firm Avast found that 58% of users thought a real banking app was fraudulent, while 36% thought a fraudulent app was real. I don't really buy the numbers, but my takeaway is: people have no idea how to identify digital financial fraud. I wish that seemed more concerning to people in the digital finance world.
2. Our Algorithmic Overlords: I've had a couple of conversations with folks after my review of Automating Inequality, and had the chance to chat quickly with Virginia Eubanks after seeing her speak at the Aspen Summit on Inequality and Opportunity. My views have shifted a bit: in her talk Eubanks emphasized the importance of keeping the focus on who is making decisions, and that the danger that automation can make it much harder to see who (as opposed to how) has discretion and authority. A big part of my concern about the book was that it put too much emphasis on the technology and not the people behind it. Perhaps I was reading my own concerns into the text. I also had a Twitter chat with Lucy Bernholz who should be on your list of people to follow about it. She made a point that has stuck with me: automation, at least as it's being implemented, prioritizes efficiency over rights and care, and that's particularly wrong when it comes to public services.
I closed the review by saying that "the problem is the people"; elsewhere I've joked that "AI is people!" Well at least I thought I was joking. But then I saw this new paper about computational evolution--an application of AI that seeks to have the machine experiment with different solutions to a problem and evolve. And it turns out that while AI may not be people, it behaves just like people do. The paper is full of anecdotes of machines learning to win by gaming the system (and being lazy): for instance, by overloading opponents' memory and making them crash, or deleting the answer key to a test in order to get a perfect score. I think the latter was the plot of 17 teen movie comedies in the '80s. Reading the paper is rewarding but if you just want some anecdotes to impress your friends at the bar tonight, here's a twitter thread summary. It's funny, but honestly I found it far scarier than that video of the robot opening a door from last month. Apparently our hope against the robots is not the rules that we can write, because they will be really good at gaming them, but that the machines are just as lazy as we are.
To round out today's scare links, here's a news item about a cyberattack against a chemical plant apparently attempting to cause an explosion; and here's a useful essay on our privacy dystopia.
Week of March 4, 2018
1. Crappy Financial Products: The results are no surprise, but it remains troubling to see the numbers. “Color and Credit” is a 2018 revision of a 2017 paper by Taylor Begley and Amitatosh Purnanandam. The subtitle is “Race, Regulation, and the Quality of Financial Services.” Most studies of consumer financial problems look at quantity: the lack of access to financial products. But here the focus is on quality: You can get products, but they’re lousy. Too often, they’re mis-sold, fraudulent, and accompanied by bad customer service. These problems had been hard to see, but they’ve been uncovered via the Consumer Financial Protection Bureau Complaints database, a terrifically valuable, publicly accessible—and freely downloadable—database. (Side note: this makes me very nervous about the CFPB’s current commitment to maintaining the data.)
Thousands of complaints are received each week, and the authors look at 170,000 complaints from 2012-16, restricted to mortgage problems. The complaints come from 16,309 unique zipcodes – and the question is: which zipcodes have the most complaints and why? The first result is that low income and low educational attainment in a zipcode are strongly associated with low quality products. Okay, you already predicted that. On top of those effects, the share of the local population identified as being part of a minority group also predicts low quality. No surprise again, but you might not have predicted the magnitude: The minority-share impact is 2-3 times stronger then the income or education impact (even when controlling for income and education). The authors suspect that active discrimination is at work, citing court cases and mystery shopper exercises which show that black and Hispanic borrowers are pushed toward riskier loans despite having credit scores that should merit better options. So, why? Part of the problem could be that efforts to help the most disadvantaged areas are backfiring. Begley and Purnanandam give evidence that regulation to help disadvantaged communities actually reduces the quality of financial products. The culprit is the Community Reinvestment Act, and the authors argue that by focusing the regs on increasing the quantity of services delivered in certain zipcodes, the quality of those services has been compromised – and much more so in heavily-minority areas. Unintended consequences that ought to be taken seriously.
2. TrumpTown: Another great database. ProPublica is a national resource – a nonprofit newsroom. They’ve been doing a lot of data gathering and number-crunching lately. Four items today are from ProPublica. The first is the geekiest: a just-released, searchable database of 2,475 Trump administration appointees. The team spent a year making requests under the Freedom of Information Act, allowing you to now spend the afternoon getting to know the mid-tier officials who are busily deregulating the US economy. The biggest headline is that, of the 2,475 appointees, 187 had been lobbyists, 125 had worked at (conservative) think tanks, and 254 came out of the Trump campaign. Okay, that’s not too juicy. Still, the database is a resource that could have surprising value, even if it’s not yet clear how. Grad students: have a go at it. (Oh, and I’d like to think that ProPublica would have done something similar if Hilary Clinton was president.)
3. Household Finance (and Inequality): This ProPublica story is much more juicy, and much more troubling. Writing in the Washington Post, ProPublica’s Paul Kiel starts: “A ritual of spring in America is about to begin. Tens of thousands of people will soon get their tax refunds, and when they do, they will finally be able to afford the thing they’ve thought about for months, if not years: bankruptcy.” Kiel continues, “It happens every tax season. With many more people suddenly able to pay a lawyer, the number of bankruptcy filings jumps way up in March, stays high in April, then declines.” Bankruptcy is a last resort, but for many people it’s the only way to get on a better path. Even when straddled with untenable debt, it turns out to be costly to get a fresh start.
The problem will be familiar to anyone who has read financial diaries: the need for big, lumpy outlays can be a huge barrier to necessary action. Bankruptcy lawyers usually insist on being paid upfront (especially for so-called “chapter 7” bankruptcies). The problem is that if the lawyers agreed to be paid later, they fear that their fees would also be wiped away by the bankruptcy decision. So, the lawyers put themselves first. The trouble is that the money involved is sizeable: The lawyers’ costs plus court fees get close to $1500. The irony abounds. Many people tell Kiel that if they could easily come up with that kind of money, then they probably wouldn’t be in the position to go bankrupt. Bankruptcy judges see the problem and are trying to jerry-rig solutions, but nonprofits haven’t yet made this a priority. So, for over-indebted households, waiting to receive tax refunds turns out to be a key strategy.
Read MoreFirst Week of March, 2018
1. Global Development: One of the more encouraging trends in development economics as far as I'm concerned is the growth of long-term studies that report results not just once but on an on-going basis. Obviously long-term tracking like the Young Lives Project or smaller scale work like Robert Townsend's tracking of a Thai village (which continues to yield valuable insights) falls in this category, but it's now also happening with long term follow-up from experimental studies. Sometimes that takes the form of tracking down people affected by earlier studies, as Owen Ozier did with deworming in Kenya. But more often it seems, studies are maintaining contact over longer time frames. A few weeks ago I mentioned a new paper following up on Bloom et. al.'s experiment with Indian textile firms. The first paper found significant effects of management consulting in improving operations and boosting profits. The new paper sees many, but not all, of those gains persist eight years later. Another important example is the on-going follow up of the original Give Directly experiment on unconditional cash transfers. Haushofer and Shapiro have new results from a three year follow-up, finding that as above, many gains persist but not all and the comparisons unsurprisingly get a bit messier.
Although it's not quite the same, I do feel like I should include some new work following up on the Targeting the Ultra Poor studies--in this case not of long-term effects but on varying the packages and comparing different approaches as directly as possible. Here's Sedlmayr, Shah and Sulaiman on a variety of cash-plus interventions in Uganda--the full package of transfers and training, only the transfers, transfers with only a light-touch training and just attempting to boost savings. They find that cash isn't always king: the full package outperforms the alternatives.
2. Our Algorithmic Overlords: If you missed it, yesterday's special edition faiV was a review of Virginia Eubanks Automating Inequality. But there's always a slew of interesting reads on these issues, contra recent editorials that no one is paying attention. Here's NYU's AINow Institute on Algorithmic Impact Assessments as a tool for providing more accountability around the use of algorithms in public agencies. While I tend to focus this section on unintended negative consequences of AI, there is another important consideration: intended negative consequences of AI. I'm not talking about SkyNet but the use of AI to conduct cyberattacks, create fraudulent voice/video, or other criminal activities. Here's a report from a group of AI think tanks including EFF and Open AI on the malicious use of artificial intelligence.
3. Interesting Tales from Economic History: I may make this a regular item as I tend to find these things quite interesting, and based on the link clicks a number of you do too. Here's some history to revise your beliefs about the Dutch Tulip craze, a story it turns out that has been too good to fact check, at least until Anne Goldgar of King's College did so. And here's work from Judy Stephenson of Oxford doing detailed work on working hours and pay for London construction workers during the 1700s. Why is this interesting? Because it's important to understand the interaction of productivity gains, the industrial revolution, wages and welfare--something that we don't know enough about but has implications as we think about the future of work, how it pays and the economic implications for different levels of skills. And in a different vein, but interesting none-the-less, here is an epic thread from Pseudoerasmus on Steven Pinker's new book nominally about the Enlightenment.
Book Review Special Edition: Automating Inequality
1. Algorithmic Overlords (+ Banking + Digital Finance + Global Development) book review: I'd like to call myself prescient for bringing Amar Bhide into last week's faiV headlined by questions about the value of banks. Little did I know that he would have a piece in National Affairs on the value of banks, Why We Need Traditional Banking. The reason to read the (long) piece is his perspective on the important role that efforts to reduce discrimination through standardization and anonymity played in the move to securitization. Bhide names securitization as the culprit for a number of deleterious effects on the banking system and economy overall (with specific negative consequences for small business lending).
The other reason to read the piece is it is a surprisingly great complement to reading Automating Inequality, the new book from Virginia Eubanks. To cut to the chase, it's an important book that you should read if you care at all about the delivery of social services, domestically or internationally. But I think the book plays up the technology angle well beyond it's relevance, to the detriment of very important points.
The subtitle of the book is "how high-tech tools profile, police and punish the poor" but the root of almost all of the examples Eubanks gives are a) simply a continuation of policies in place for the delivery of social services dating back to, well, the advent of civilization(?), and b) driven by the behaviors of the humans in the systems, not the machines. In a chapter about Indiana's attempt to automate much of its human services system, there is a particularly striking moment where a woman who has been denied services because of a technical problems with an automated document system receives a phone call from a staffer who tries very hard to convince her to drop her appeal. She doesn't, and wins her appeal in part because technology allowed her to have irrefutable proof that she had provided the documents she needed to. It's apparent throughout the story that the real problem isn't the (broken) automation, but the attitudes and political goals of human beings.
The reason why I know point a) above, though, is Eubanks does such an excellent job of placing the current state in historical context. The crucial issue is how our service delivery systems "profile, police and punish" the poor. It's not clear at all how much the "high tech tools" are really making things worse. This is where Bhide's discussion is useful: a major driver toward such "automated" behaviors as using credit scores in lending was to do an end-run around the discrimination that was rampant among loan officers (and continues to this day, and not just in the US). While Eubanks does raise the question of the source of discrimination, in a chapter about Allegheny County, PA, she doesn't make a compelling case that algorithms will be worse than humans. In the discussion on this point she even subtly undermines her argument by judging the algorithm by extrapolating false report rates from a study conducted in Toronto. This is the beauty and disaster of human brains: we extrapolate all the time, and are by nature very poor judges of whether those extrapolations are valid. In Allegheny County, according to Eubanks telling, concern that case workers were biased in the removal of African-American kids from their homes was part of the motivation for adopting automation. They are not, it turns out. But there is discrimination. The source is again human beings, in this case the ones reporting incidents to social services. The high-tech is again largely irrelevant.
I am particularly sensitive to these issues because I wrote a book in part about the Toyota "sudden acceleration" scare a few years ago. The basics are that the events described by people who claim "sudden acceleration" are mechanically impossible. But because there was a computer chip involved, many many people were simply unwilling to consider that the problem was the human being, not the computer. There's more than a whiff of this unjustified preference for human decision-making over computers in both Bhide's piece and Eubanks book. For instance, one of the reasons Eubanks gives for concern about automation algorithms is that they are "hard to understand." But algorithms are nothing new in the delivery of social services. Eubanks uses a paper-based algorithm in Allegheny County to try to judge risk herself--it's a very complicated and imprecise algorithm that relies on a completely unknowable human process, that necessarily varies between caseworkers and even day-to-day or hour-to-hour, to weight various factors. Every year I have to deal with social services agencies in Pennsylvania to qualify for benefits for my visually impaired son. I suspect that everyone who has done so here or any where else will attest to the fact that there clearly is some arcane process happening in the background. When that process is not documented, for instance in software code, it will necessarily be harder to understand.
To draw in other examples from recent faiV coverage, consider two papers I've linked about microfinance loan officer behavior. Here, Marup Hossain finds loan officers incorporating information into their lending decisions that they are not supposed to. Here, Roy Mersland and colleagues find loan officers adjusting their internal algorithm over time. In both cases, the loan officers are, according to some criteria, making better decisions. But they are also excluding the poorest, even profiling, policing and punishing them, in ways that are very difficult to see. While I have expressed concern recently about LenddoEFL's "automated" approach to determining creditworthiness, at least if you crack open their data and code you can see how they are making decisions.
None of which is to say that I don't have deep concerns about automation and our algorithmic overlords. And those concerns are in many ways reinforced and amplified by Eubanks book. While she is focused on the potential costs to the poor of automation, I see two areas that are not getting enough scrutiny.
First, last week I had the chance to see one of Lant Pritchett's famous rants about the RCT movement. During the talk he characterized RCTs as "weapons against the weak." The weak aren't the ultimate recipients of services but the service delivery agencies who are not politically powerful enough to avoid scrutiny of an impact evaluation. There's a lot I don't agree with Lant on, but one area where I do heartily agree is his emphasis on building the capability of service delivery. The use of algorithms, whether paper-based or automated, can also be weapons against the weak. Here, I look to a book by Barry Schwarz, a psychologist at Swarthmore perhaps most well-known for The Paradox of Choice. But he has another excellent book, Practical Wisdom, about the erosion of opportunities for human beings to exercise judgment and develop wisdom. His book makes it clear that it is not only the poor who are increasingly policed and punished. Mandatory sentencing guidelines and mandated reporter statutes are efforts to police and punish judges and social service personnel. The big question we have to keep in view is whether automation is making outcomes better or worse. The reasoning behind much of the removal of judgment that Schwartz notes is benign: people make bad judgments; people wrongfully discriminate. When that happens there is real harm and it is not obviously bad to try to put systems in place to reduce unwitting errors and active malice. It is possible to use automation to build capability (see the history of civilization), but it is far from automatic. As I read through Eubanks book, it was clear that the automated systems were being deployed in ways that seemed likely to diminish, not build, the capability of social service agencies. Rather than pushing back against automation, the focus has to stay on how to use automation to improve outcomes and building capability.
Second, Eubanks makes the excellent point that while poor families and wealthier families often need to access similar services, say addiction treatment, the poor access them through public systems that gather and increasingly use data about them in myriad ways. One's addiction treatment records can become part of criminal justice, social service eligibility, and child custody proceedings. Middle class families who access services through private providers don't have to hand over their data to the government. This is all true. But it neglects that people of all income levels are handing over huge amounts of data to private providers who increasingly stitch all of that data together with far less scrutiny than public agencies are potentially subject to. Is that really better? Would the poor be better off if their data was in the hands of private companies? It's an open question whether the average poor person or the average wealthy person in America has surrendered more personal data--I lean toward the latter simply because the wealthier you are the more likely you are to be using digital tools and services that gather (and aggregate and sell) a data trail. The key determinant of what happens next isn't, in my mind, whether the data is held by government or a private company, but who has the power to fight nefarious uses of that data. Yes, the poor are often going to have worse outcomes in these situations but it's not because of the digital poorhouse, it's because of the lack of power to fight back. But they are not powerless--Eubanks stories tend to have examples of political power reigning in the systems. As private digital surveillance expands though, the percentage of the population who can't fight back is going to grow.
So back to the bottom line. You should read Automating Inequality. You will almost certainly learn a lot about the history of poverty policy in the US and what is currently happening in service delivery in the US. You will also see lots to be concerned about in the integration of technology and social services. But hopefully you'll also see that the problem is the people.
Week of February 12, 2018
1. Banking: In case you missed it, here's that link from last week finding that banks would be better off if they did a lot less. Well, a lot less of the complicated financial stuff that most (large) banks spend a lot of time doing. Matt Levine sees a generalized trend in a positive direction--that is that the financial engineering that financial services companies are engaged in is focused much less on engineering complex financial instruments and a lot more on software and technology engineering. Even the cool project names are being reserved for technology projects rather than hard-to-understand derivatives-of-futures-of-insurance-of-bonds-of-weather-derivatives.
That does raise some questions about the evolution of fintech--if the banks themselves are more focused on the technology of service delivery, what does that mean for the technology firms? I do feel a bit of unease that these are the same banks that don't seem to be able to add value to themselves in their core area of expertise (and it's not just the banks, remember that Morningstar's ratings are negative information). How much should we expect from their wading into technology and advice? More on that below, in item 2.
There's another concern with banks moving in this direction. While it's not always the case, the kind of engineering that banks are doing now tend to increase consolidation: returns to scale tend to be bigger and matter more in software, data and high-volume/low-margin activities. And when consolidation happens it tends to be bad for lower-income customers. Here's a recent paper examining the impact of bank consolidation in the US (particularly large banks acquiring small banks): higher minimum account balances and higher fees, particularly in low-income neighborhoods. Those neighborhoods see deposits flow out of bank accounts (justifying closing branches) and later see increases in check-cashing outlets and decreased resilience to financial shocks. But wait there's more: the current version of the Community Reinvestment Act regulations tend to focus on places where banks have a physical presence. So closing branches and delivering more services through technology means, well, that those banks have less worries about complying with CRA. Hey did you know that the Treasury Department is considering making changes to the CRA regulations? I'm guessing the first priority isn't going to be expanding the CRA mandates.
And just to throw in a little non-US spice, here's a story about massive bank fraud at the Punjab National Bank in India.
2. Our Algorithmic Overlords: I've made jabs in the faiV pretty regularly about fintech algorithms ability to make good recommendations, particularly for lower income households. It turns out I'm not alone in distrusting machine-generated recommendations. Human beings tend to believe pretty strongly that humans make better recommendations than machines particularly when it comes to matter of taste. But we're all wrong. Here's a new paper from Kleinberg, Mullainaithan, Shah and Yeomans testing human versus machine recommendations of jokes(!). The machines do much better. Perhaps I should shift my concern away from machine-learning-driven recommendations and spend more time on a different preoccupation: why humans are so bad at making recommendations. There is perhaps another way: making humans and machines both part of the decision-making loop. A great deal of work in machine learning right now is organized around humans "teaching" a machine to make decisions, and then turning the machine loose. An alternative approach is having the "machine-in-the-loop" without ever turning it loose. That is the approach generally being used in such things as bail decisions. The big outstanding question is where we should allow humans (and which humans) to overrule machine recommendations and when we should allow the machines (and which machines) to overrule the humans.
Key to making such decisions is whether the human is able to understand what the machine is doing, and whether humans should trust the machine. Both are dependent on replicability of the AI. You might think sharing data and code in AI research would be standard. But you'd be as wrong as I was about recommendations. There's a budding replication crisis in AI studies because it is so rare for papers to be accompanied by the training data (about 30%) used in machine-learning efforts, much less the source code for their algorithms (only 6%!). Of note if you click on the paper above about recommendations, on page two there is note that all of the authors' data and code are available for download.
