Editor's Note: Week 2 of the Monday experiment and things keep happening on the weekends, which makes it a bit more complicated. There won't be a faiV next week, as one of my other "side gigs" is running the Bardet Biedl Syndrome Foundation, and we have our first in-person conference since 2018 this coming weekend. I raise this here because there's an item related to that, though it's kind of outside the usual faiV envelope, below.
- Tim Ogden
1. Economic Development
It's telling that I'm introducing an item on the death of a political scientist under the heading of Economic Development. James Scott died this weekend, and while he was certainly a political scientist he also had a big influence on thinking about economic development and particularly about the downsides of large, planned development schemes of the sort that defined the 60's-80's (at least).
He's perhaps most famous in economic circles for Seeing Like a State, but there's also The Moral Economy of the Peasant, Weapons of the Weak, The Art of Not Being Governed and Against the Grain, and more. I can't begin to do him justice, so here are some things to bring you up to speed. First, a blend of the professional and personal from a former student. Here's a paper from Marcel Fafchamps that puts the challenge that Scott created to prevailing economic models of proto-insurance in context, and gives you a sense of how much he shifted thinking here (given how much this is now "of course, that's how that works.". Here's a review of Seeing Like a State by Brad DeLong, and a critical review from Paul Seabright. The Annual Reviews has published (for free) Scott's reflection on his academic journey.
For me personally, the two ideas that I constantly carry around with me that I attribute back to Scott are: 1) Operating anything (until possibly the near future, see below) at scale requires a flattening of difference that necessarily involves costs, and possibly very large costs; this isn't a flaw in design or knowledge or anything else, it is a fundamental feature of scale. 2) People will be remarkably creative and often incur very large costs in order to not be flattened by the scale machines.
And on a tangential note, here's a link to a recent book that is related in that it illustrates part of the reason that people will so creatively and expensively revolt against being flattened—they often fundamentally misunderstand the workings of anything at scale (this is specifically about how Americans don't understand their own government). I just really really wish this had been titled Seeing Like a Citizen.
2. Our Algorithmic Overlords
Given that last week I was noting some of the successes of AI, it seems that I was especially tuned in to things this week that undermine the narrative. Specifically, I linked a paper about predictive policing. Immediately after, there was a NYT story on the failures of a Spanish algorithmic system for predicting risk of domestic violence. This connects to the flattening point above: algorithms necessarily flatten down a huge amount of data to produce an outcome that may be close to reality, even if it is not always correct. We see in many places evidence that the flattening that an algorithm/AI does is better than the flattening a human does. But there is still a flattening. I gave a nod to one of the things I think about a lot these days last week that also applies here. The most interesting thing to me about AI's progress is not that it can flatten better than humans (possibly, in some circumstances, with the right training data, etc.) but the possibility that it does less flattening. It potentially provides the means to truly change the scale challenges that underlie Scott's Seeing Like a State. And again raises the question of how much personalization in things like prices and credit and risk scores is "fair." Did I mention the possibly collusive apartment rental pricing algorithm/platform?
You may have heard that there was bit of a snafu with some PCs on Friday—no more than a couple billion—that had some effect on those computers' ability to function. This is a trade-off that Scott would recognize: the safety of end nodes in a network at scale requires flattening differences and centrally controlling the environment, which yields huge benefits, until it doesn't and yields huge costs.
I won't link to any of the many many stories on that which I'm sure you've already seen, but here is a story about the impact on small businesses, which are not insignificant (though probably, because of compounding at scale, smaller in total while more devastating to specific businesses). Also of note is a new paper about how badly large and medium enterprises have been doing, well before the advent of AI-enabled bad actors, at keeping their computer systems up-to-date with the latest patches and fixes (can you imagine the range of emotions these authors felt when this outage hit?). Now let's think about how we do cost/benefit analysis of putting several billion people's household finances, payment mechanisms, business records, etc. into the cloud. Is it better to centralize or decentralize?
3. Basic Income
What would you expect to be the effect of providing low-income people (in the US) with $1000/month in unconditional cash over 3 years? Would "not much" be high on your list? There are two new papers on one of the largest randomized experiments on unconditional cash transfers, and probably the largest in a wealthy country. The first looks at employment effects—small but negative (e.g. people work a few hours less, and total income falls). The second looks at health—and finds no change in health, including measures of both subjective well-being and biomarkers from blood draws. You might already be thinking of some confounders that would muddy the interpretation of the results. But the researchers already thought of every one I could come up with and measured that. Still, nothing to see. Here's Eva Vivalt's thread on the employment paper; here's David Brookmans' thread on the details of the experiment and how to successfully run an RCT given all the challenges. There are also several more papers coming including what people predicted the effects would be vs. what was found, and specific effects on other measures of well-being and household finance. I guess the only question now is will some of the authors become consultants on the basic income comedy cartoon linked in the last edition.
I'll also note that Jonathan is involved in another cash transfer study with quantitative and qualitative methods and the findings are broadly consistent with this study—papers coming in the next few weeks.
4. Demographics
I've mentioned a couple of times in recent faiVs that I'm finding myself thinking a lot about how the mass personalization that AI will enable will change our perceptions of how financial services should operate, and what is fair. There's another thing that I also find my mind wandering to a lot these days: how rapidly and radically demographics are changing and what that means for our basic intuitions about how the world will work.
These aren't anywhere close to fully formed thoughts—and I'm definitely interested in your thoughts, or pointers to things you think are worth reading on this. But here's some sampling of what has my brain working when I'm not working. First, the shift is more radical than you may realize, in part because the UN's Population Division has a long record of ignoring trends and issuing forecasts based on fertility rates stabilizing wherever they happen to be when the forecast was made. Check out this chart about India's Total Fertility Rate—and read the thread for more.
What happens when populations begin falling? For instance, what happens in communities when the arguments turn from under-resourced schools to a physical plant that's too big and expensive for the needs? How does culture change when a very large portion of the population doesn't have children at all? That would certainly have a big effect on household financial behavior (and on the theoretical optimal behavior). It certainly will have a big effect on the politics of child care. What happens in labor markets when the population isn't growing? Interestingly, it may have short term effects on reducing the gender wage gap (because the gap is bigger for older generations and as they retire the average gap can fall without any current progress on equal pay).
There's a lot more on this topic—I'm hoping that I'll make it a bit more salient to you and you'll start noticing (and sending me) things where the implications of this are just starting to play out.
5. Market Design
Okay, I'm going to ask for your indulgence here a bit. This is the item I referred to above that's outside the usual faiV envelope. As long time readers know, I have a son with an ultra-rare genetic disease. If you take a few moments, it will become obvious to you that one of the challenges in addressing ultra-rare diseases is the vanishingly small incentives for investing in treatments appropriately specific to those diseases. For instance, there's good news that significant scientific progress is being made on gene therapies that would address two of the most common genes that cause this disease. On the other hand, gene therapies are very expensive to develop and bring to market and with small possible target populations...well, I don't have to spell it out for you, faithful faiV subscriber. How expensive might these gene therapies be? The few gene therapies that have been approved so far, for instance for sickle cell disease, cost at least $2 million per patient.
So an effort to bring a group of experts to focus on how to bring down the cost of gene therapy is very welcome. But to be honest I am decidedly unimpressed with the recommendations. These include, hidden behind lots of jargon, price controls. But also new business models, which would find investors willing to offer capital at lower rates of return based on selection-effect biased estimates of risk. And they would also depend on "well-defined governance structures [...] to maintain public benefit goals and remain in compliance with relevant laws." If James Scott were still with us, I think he might have something to say about the way this group is seeing things.
Gene therapies are going to break health insurance even more than it already is. While there are some ways to bring down the cost (including regulatory reform that is included in that report), I think the more promising and impactful development is significant investment in gene-agnostic therapies. Take retinal degeneration for instance. There are hundreds of genes that cause retinal degeneration. We can make custom gene therapies for each of those genes, each with a (relative) handful of possible patients, or we could find a way to ramp up investment in approaches that work regardless of the genetic cause. That, would in fact, be a pretty wise investment for payers (whether government or private). But given the large risks in developing such approaches, even the funds available through that channel may not be up to the task.
If only we knew of a way to crowd in additional investment by reducing risk. Say, like some sort of contract to purchase such a gene agnostic therapy, were it to be developed. We could call it something catchy, like maybe an "Advanced Market Commitment." I'm probably just dreaming, something like that would never work.
P.S. It would be really great if someone out there reading the faiV wanted to collaborate on making the case for payers to invest in alternatives to gene therapy for rare genetic diseases. Call me.
Video of the Day
I'm honestly a bit conflicted by this—there being certain historic tragedies/monstrous evil that it's OK to joke about and others where it's not (cf. this moment in tragi-evil-comedy. But the AI-generated mash-up of Mao-era propaganda with uber-American jingoism embodied by Toby Keith just leaves me amazed every time I click on it.
The faiV is written by Timothy Ogden and produced by the Financial Access Initiative at NYU's Wagner Graduate School of Public Service
Email: fai-wagner@nyu.edu
To read this in your inbox, subscribe to the faiV.
