Peter Foreshaw Brookes

At @MarketUrbanism , @salimfurth comments on the disappointing fade of the English New Homes Bonus program. I had never heard of this program until I was working on something else, asked Claude to recommend me some literature, and it recommended me something on the NHB program... and then I was flabbergasted I'd never heard of it. The NHB at its peak provided English localities with SIX YEARS of property-tax-revenue-equivalents for every new home they permitted. This was thousands of pounds of direct revenue for every home. They spent billions of pounds on this program from 2011-2026 (it has just this year been eliminated). So, did paying English localities to permit housing work? Well, I had Claude figure it out. The answer is... maybe slightly? I instructed Claude to study 4 different ways of analyzing if the program boosted homebuilding. First, leveraging a rule change in 2017 that excluded a bunch of localities, we can see if losing the NHB reduced permitting. Answer: no, in fact they maybe increased their permitting after losing the NHB. However, this isn't a clean test: they may have increased their permitting in order to get back into the NHB program by getting over the 0.4% stock additional threshold. So this one is a "Maybe the program kinda worked?" answer. Second, I was curious about a learning mechanism. You can imagine that pre-NHB, localities that had a "big year" of permitting may have then had "big years" of paperwork, disruptions, and costs, and so "big years" would cause subsequent reductions in permitting. But post-NHB, you can imagine that "big years" of permitting would cause big windfalls of revenues, so might nudge localities to "learn" to permit more. Did Claude find evidence for this? No, not at all. In fact "big years" for permitting seem associated with more NEGATIVE post-year permitting in the post-NHB than the pre-NHB period. Yikes. Talking with @salimfurth he noted that in the post-2017 era, localities might have incentivizes to cluster their permitting all in one year to get over the 0.4% threshold, which could bias these results. Fair. Third, the NHB was a WAY bigger budgetary impact for some localities (Shire districts) than for others, just because the ratio of new housing construction to municipal budgets varies widely. You'd expect the "more exposed" districts to increase permitting more. The effect here at first seems promising: Shire districts DID have big increases in permits granted! But this is because they had big increases in permits applied for; the actual *rate of granting a permit* DECLINES for Shire districts, despite the NHB literally *paying them to approve*. Moreover, the rise in approvals clearly has a pre-trend before NHB rolled out. NHB also paid localities to clear out long-term empty homes, but we see no Shire-specific reduction in long-term empty homes. So this all suggests little to no effect. Fourth and finally, the NHB applied *in England only*. We can look at England vs. Wales and Scotland, and then specifically at *municipalities near the borders of England* to see if they changed in a unique way. The answer here is.... maybe? It does look like both the whole regions AND the border-regions only did have higher permitting in England, but it is VERY modest and noisy. On the whole, this seems like a very big, expensive experiment telling us that "pay for permitting" may not work very well at all. cc @aarmlovi @SCP_Hughes @PeterRFB @johnarnold As an aside, this is, I think, a fun case of what Claude allows for researchers. Claude caused my awareness of this program which gets shockingly little attention. I didn't have time to do tons of elaborate analyses myself, so I literally just came up with four possible tests of how the program could be evaluated, and told Claude "here are four ways you could test the program to see if it worked; go do these tests and make sure you use best research practices and very good methods along the way" Claude didn't do perfectly, and I hope a human housing scholar will go investigate in more detail. But being able to run basically four different quasi-causal models on an intervention *the same afternoon I learned of its existence* and doing so at <1 hour time cost to me, the researcher... y'all, AI is the real deal.

Young Koreans want love but they don’t want babies