What we’re reading, April 3, 2026
Factory-built housing, immigration success, and forecasting AI's economic impact
Happy Friday! Here’s what caught our attention this week:
Industrial Policy for Housing Policy is a new blueprint for tackling the housing crisis from Arpit Gupta and Steve Teles. One bit I particularly liked was the five-part diagnosis of why factory-built housing has failed to really take off in a sustainable way: (1) housing’s boom-and-bust cycle makes it hard for large investments into factory production to survive lean times, especially because (2) the government has not traditionally moved in to smooth fluctuations; (3) a patchwork of different regulations makes it hard for learning by doing to get as much traction, (4) it’s expensive to transport housing components away from factories, and (5) it’s hard to assemble lots of little (separate) parcels into a contiguous area that can support dense new buildings. The diagnosis suggests several possible fixes and Gupta and Teles have a bunch of other ideas about how to overcome these barriers. — Matt Clancy
Why America is so much better at immigration than Europe is a fascinating new piece by Kelsey Piper and Alexander Kustov (who also writes the great Popular by Design substack on immigration). Piper and Kustov point to a number of factors that have led to better outcomes and more public support for immigration in the United States (I know, right?), not least of which is America’s flexible labor market. When immigrants can get to work, they are more likely to integrate, impose less of a fiscal burden, and less likely to engage in crime. That, in turn, leads to more public support for immigration. Indeed, Piper and Kustov argue that the turn in public opinion against immigration in the USA coincides with shifts that made it harder for immigrants in the US to work! The labor market is only one part of the argument Piper and Kustov make, but it’s highly relevant to our thinking about policy work we could support in Europe. — Matt Clancy
We at AGF don’t work as directly on AI issues as many of our peers at Coefficient Giving, but you can’t really think about the medium to long-run state of the US and world economies without thinking seriously about AI. I found a big new paper from the Forecasting Research Institute (summary thread here) asking AI experts, economists, “superforecasters,” and the general public to predict AI’s effects on the economy by 2030 and 2050 very helpful in calibrating my own views. Perhaps the most surprising thing to me is that the four groups weren’t terribly far apart from each other. In a world of rapid AI progress, economists expect 3.5 percent GDP growth in 2050; AI experts expect 5.3 percent. This pales next to forecasts of double-digit annual economic growth caused by transformative AI that more optimistic commentators have made. For more, see Matt, Alex, and my follow-up posts on the survey. — Dylan Matthews
This week, there were two very different pieces on reforming how we fund science. In City Journal, Michael Gibson (former VP of Grants at the Thiel Foundation) argues that Jim O’Neill’s nomination as NSF director is a chance to shake up the agency’s approach to talent identification and grant design. His specific proposals will be familiar to anyone who follows metascience, and include partial lotteries, scout programs, open access, and indirect cost caps. Meanwhile, in PNAS, Harvey Fineberg (former president of the Gordon and Betty Moore Foundation) echoes the need to reinvigorate the U.S. scientific ecosystem, but presents a more macro vision for getting there: sustained increases in federal science funding, networked innovation clusters, state-level funding programs, and a focus on domestic and global talent. Notably, both emphasize the need for increased support of younger scientists, but otherwise their priorities diverge. — Jordan Dworkin
However you want to reform science funding, structural shake-ups require that there are functioning agencies to reform; to keep tabs on how things are going on that front, Grant Witness is now tracking NIH and NSF funding curves in real time. — Jordan Dworkin
It’s a big week for electricity price data - the Lawrence Berkeley National Laboratory updated its Retail Electricity Price Trends and Drivers with 2025 data and new analysis, and a Heatmap/MIT/CleanEcon collaboration launched an Electricity Price Hub with electricity bill and price data down to the utility level. Both illustrate how fragmented price patterns continue to resist a coherent story: the most dramatic price increases continue to be geography-specific, like outsize generation costs where PJM load growth has hit supply constraints. LBNL’s update does flag a few trends that will likely continue to shape prices going forward: investor-owned utilities requested their largest revenue increases since the 1980s in 2025 (and regulatory approval rates for such increases have gone up significantly since 2020, particularly in New England and the Southeast); and equipment prices increases for transmission and distribution inputs continue to far outpace inflation. — Willow Latham-Proenca
Research in action: Why permitting reform matters. Evan Soltas translated his recent LA permitted reform paper, that we featured a few weeks ago, into a localized estimate of the value of permitting reform in NYC. Each year of permitting time saved in NY is equivalent to an 8% drop in construction costs. With state SEQRA reform and local NYC streamlining, 2.5 years of promised permitting time savings could save quite a bit. Chris Elmendorf further notes the Governor & State Senate would need to prioritize SEQRA in budget negotiations with a still-recalcitrant Assembly to get clean reform over the finish line — Alex Armlovich
We might finally get a vaccine against Lyme disease! Maybe. Hopefully? Unfortunately, the confidence intervals in the phase three trial were really wide: after four doses, its efficacy was estimated at 73.2%, with a 95% confidence interval of 15.8% to 93.2%. (The trial’s primary endpoint was to surpass a lower bound of 20% efficacy, which it failed to do.) That huge uncertainty is probably because Lyme disease is so rare that you’d need a huge sample size to reach a precise estimate with a standard trial design. This trial had over 12,000 participants, but it seems even that wasn’t enough for more precision. But also… we actually had a Lyme vaccine in the ‘90s! And then it was withdrawn after a lack of public demand for it. (Not because people don’t care about Lyme disease, but because of fears that the vaccine caused arthritis, even though no connection was found by FDA analysis and later research…) If that sounds incredibly frustrating, that’s probably because it is. Maybe we’ll have better luck this time. — Saloni Dattani
We also wanted to share a few updates from our team and grantees:
Our grantee Witold Więcek co-authored a perspective in JAMA on the FDA’s new Bayesian statistics guidance for clinical trials. The piece argues that Bayesian methods can improve both trial design and regulatory decisions by formally incorporating prior information from related studies.
DC’s City Council passed single stair reform on first reading, which our grantee Greater Greater Washington has long supported. If it clears second reading and the mayor’s signature, the reform will unlock small multifamily-zoned sites where lot assembly for double-loaded stairs is difficult or impossible - particularly valuable given DC’s history of allowing creative exterior stair workarounds.
The Institute for Progress released a new report on Department of Labor wage proposals that affect H-1B visa eligibility, analyzing how different wage ranking methodologies would impact skilled immigration flows.
Alex Armlovich published Zoned Capacity Is Like an Artificial Organ Donor Registry, exploring how cities create the illusion of housing supply through theoretical zoned capacity that often fails to translates into actual construction.
And if you’re looking for abundance-related job opportunities, the Abundance Network maintains a job board with openings across housing, energy, infrastructure, and innovation policy organizations.