Saturday, January 23, 2016

You weren't wrong, Adam Ozimek

Adam Ozimek has a nice post today at Modeled Behavior, discussing the discipline of objectivity.  His recommendations of imagining yourself changing your mind and talking about things you have been wrong about are both difficult and useful practices to maintain.

His example of something he has been wrong about is that he was somewhat sanguine about the loss of manufacturing jobs, and that the housing bust convinced him that the shift from manufacturing has been more painful than he had acknowledged.  He cites a study with the following excerpt:
….We also find that housing booms significantly reduce the likelihood that displaced manufacturing workers remain non-employed, suggesting that housing booms masked non-employment growth that would have otherwise occurred earlier in the absence of the booms… Collectively, our results suggest that much of the non-employment growth during the 2000s can be attributed to manufacturing decline and these effects would have appeared in aggregate statistics earlier had it not been for the large, temporary increases in housing demand.
So, ironically, as IW readers know, Adam wasn't wrong at all.  The boom wasn't unsustainable.  It didn't mask anything.  The bust was avoidable and unnecessary.  So, without the presupposition about the unsustainability of the housing boom, this study actually reinforces Adam's prior beliefs.  The lost manufacturing employment wasn't masked by housing.  It was shifted to other sectors, just as Adam would have thought.

This is what is so important about getting the housing issue right.  So much new research on so many issues is built on presuppositions which are false, that this one mistake is affecting many areas of economics and finance.  Since we had two housing booms - the boom with little construction and high prices, and the boom with abundant construction and moderate prices - any presumption of how prices and supply related to one another on a national level is tainted.

The authors make the following comment (page 19):
given that our housing demand change measure is constructed with the assumption that there are no housing supply shocks, it is likely an error-ridden version of true housing demand changes.

They purport to try to fix that.  The statistics, admittedly, are a bit above my understanding without some assistance, but they appear to try to adjust for existing supply and demand elasticities in various cities.  I think the problem this may create is that supply elasticity is not stable.  In any given city, supply tends to be very elastic, up to some maximum level of bureaucratic or political ceiling, at which point it shifts to very inelastic.  In Dallas and Houston, that ceiling is very high.  In San Francisco, it is very low.  They interpret sharp price changes in places like Phoenix as shifts in demand, so that the price changes can be attributed to demand changes, but Phoenix has had high demand for housing for decades.  I think what we see in 2005 in Phoenix is probably more of a shift in demand that triggered a shift in supply elasticity.

The idea that construction employment was positively correlated to home price increases just isn't plausible, given how much of the price boom was due to cities that were extreme outliers in limiting housing supply.  Outside the Closed Access cities, which had prices well beyond what we saw in other cities, the places that did see price shifts tended to be sharp, late, and temporary (2004-2005).  For instance, even if we attribute the sharp rise in home prices in Phoenix in 2005 to housing demand, clearly this had little to do with construction employment and its relationship to changes in manufacturing employment, which would have been steadily shifting throughout the period.  And housing starts among MSA's just didn't shift that much in relation to local price shifts.

The housing boom wasn't unsustainable.  The high prices were coming from the places that wouldn't accommodate a booming construction sector.  The idea that the housing boom masked manufacturing losses because of construction employment in the Closed Access cities (the red lines in the housing permits graph above) is implausible.

Adam, you weren't wrong.  Maybe you should be more confident about your intuition, after all!


  1. Phoenix clearly appears to be an open access city from your graph of housing permits/population. But I happened upon this figure (or 'interactive graph' I guess) from the NY Times from 2014 ( It shows housing price changes for some 20 cities between 2000 and 2014, and Phoenix appears to follow a pattern (during the housing crisis time period) more similar to cities like LA and SF than Dallas or Atlanta, in terms of the extent to which prices dropped after 2007.

    Is Phoenix an outlier then among open access cities in its relatively sharp rise and decline in housing prices? Something peculiar about Phonix that explains this, I wonder?

    Also, I'm curious, do you know where one would find (optimally, downloadable) tables with housing prices from various cities over time, and housing starts per year from various cities over time? I'd kind of like to do a linear regression of the decline in housing prices from 2007-2008 (or 2009) as a function of housing starts/population prior to 2007.

    1. You are right. Phoenix - and also Florida, Nevada, and inland California - are special cases. It does appear that there were price movements late in the boom in those areas that were not justifiable by reasonable long term rent expectations, given their longstanding openness to new building. So, I think these few areas are the one place where we might ascribe some sort of "demand-side" causality to home price increases.

      On the other hand, the extreme price movements in these areas tended to be in late 2004 and 2005, after Fed rate hikes had begun. That doesn't really fit some of the demand-side explanations. Can we blame loose money when the sharpest price increases were happening when the yield curve had flattened and the Fed Funds rate was hitting 4-5%?

      These are areas that are the main recipients of migration out of New England (Florida) and coastal California (inland Cal., AZ, and NV). And, there were clearly sharp supply constraints at the time. I hope to eventually get some qualitative answers. In Phoenix, at the time, builders simply couldn't get enough permits to meet demand. I believe that migration from California was the trigger here. According to the BEA, population flows into the Open Access cities was moving up during this period. I think this is the product of the migration pattern that develops when we have extended economic expansion. High income migration into the Closed Access cities keeps ratcheting up, creating more stress on local rents, and as local home prices went through the roof, lower income households moved out, to the secondary areas, creating a temporary spike in demand that local bureaucracies couldn't handle in the areas that were the main recipients.

    2. I wonder if housing starts/population *growth* in addition to just population might be informative. Off the top of my head, I don't remember Phoenix having markedly higher growth rate than Texas's major cities, but maybe it did near the peak? For Las Vegas I think it definitely makes sense to consider growth rate; LV definitely would require a lot more building than probably any other city in the country of comparable size to keep pace with demand.

      There's also the question of how much builders try to preempt anticipated growth in demand: if it's 2004, and builders expect demand to grow as much in 2005 as it is in 2004, and they buy up land and build to account for expected growth, then the rates go up, quashing demand, effectively rendering that city retroactively 'overdeveloped', at least for the new post-rate-hike demand level.

      But I have no clue to what extent or how far in the future developers do indeed try to anticipate growth. But this might explain why some high growth cities might have experienced such a boom and bust in prices.

    3. I think you're right about population growth. I posted a graph on that a few days ago that I had received from a reader. At shorter time scales, though, the data is probably a little messy on that.

      The issue with homebuilders is why I think the sharp drop in housing starts is an early sign of a negative demand shock. Why wouldn't builders drop prices in order to keep volumes up? Prices did drop in the cities, like Phoenix, that had a late run-up. But, in the other open cities, prices dipped a little bit, while starts collapsed as strongly as anywhere.

      If interest rates were artificially high (and I think issues that keep the yield curve from inverting fully kept even long term rates high), then the value of the lot with residential improvements on it falls below the value of the lot with less intensive alternative usage. So, regardless of the economics of the inventory they were sitting on, homebuilders in Texas would have been faced with costs on replacement lots that would have reduced their incentives for price concessions. Why unload a bunch of inventory that you have to replace with more expensive lots? That's a theory. I'm hoping to get more information on that issue.