Edward Glaeser, Joseph Gyourko, and Albert Saiz construct a model of housing bubbles that is consistent with movements in housing prices and quantities during the two most recent housing bubbles (the current episode and the prior episode in the 1980s). Looking at the data, they note that areas with inelastic housing supply had large price run-ups and subsequent long, drawn out crashes in both episodes. However, "The fact that highly elastic places had price booms is one of the strange facts about the recent price explosion." Because it is unprecedented, there is considerable uncertainty about how much prices might fall in the areas where supply is elastic. However, using the model as a guide, they find that "If these markets return to their historical norm..., then they will experience further sharp price declines," though there is a lot of uncertainty surrounding this prediction.
Maybe another way to think about this is that in some areas, those areas where supply is termed inelastic, the quantity response is essentially symmetric -- housing supply moves sluggishly whether prices are rising or falling. However, other areas could have housing supply that responds elastically when prices are rising (though sometimes there can be bubbles in these markets anyway - see below), but inelastically when prices are falling. In these markets, housing comes online relatively easily when prices are rising, but quantity responds much more sluggishly when prices fall, and the response could be similar to the symmetrically inelastic cases.
Why might the two sets of markets have similar responses on the down-side? Think about the inelastic markets where supply cannot increase due to geographic limitations (they use a geographic measure to sort the data). Geography limits the expansion of housing, but when there is an oversupply of housing, geography does not prevent the supply from falling, so it must be something else that prevents quantity adjustment and whatever it is could certainly be present in markets where geography is not an issue (i.e. the markets that are elastic when prices rise). If this is right, then there's reason to believe that the two sets of markets will generate similar responses for price and quantity on the down-side, and this would explain the finding in the paper that "elasticity was uncorrelated with either price or quantity changes during the bust" (so long as other factors such as regulation are similar, e.g., it's equally easy to replace a house with a restaurant after remodeling in the two markets). This would mean that - as predicted (with qualifications) in the paper - the elastic markets may mimic the inelastic markets and be in for a sharp price decline.
One more note from the paper about elastic markets, "Even though elastic
housing supply mutes the price impacts of housing bubbles, the social welfare
losses of housing bubbles may be higher in more elastic areas, since there will
be more overbuilding during the bubble." Thus, it's possible for markets with sharp price adjustments to fare better in a welfare sense than markets where price changes are more muted. Here's some of the introduction from
the paper [
can anyone find an open link?] [Update: Richard Green: Mark Thoma thinks housing supply elasticities may be asymmetric ... I have reason to think
Mark is right. My 2005 paper with Mayo and
Malpezzi found evidence of this; cities that appeared inelastic included
Pittsburgh, Toledo, Albany, Buffalo and Providence. None of these cities had
upward pressure on housing production; rather, they were losing population and
the housing stock took a long time to adjust to the loss.]:
Housing Supply and Housing Bubbles, by Edward L. Glaeser, Joseph Gyourko, and Albert Saiz, NBER WP 14193, July 2008: Introduction In the 25 years since Shiller (1981) documented that swings in stock prices were extremely high relative to changes in dividends, a growing body of papers has suggested that asset price movements reflect irrational exuberance as well as fundamentals (DeLong et al., 1990; Barberis et al., 2001). A running theme of these papers is that high transactions costs and limits on short-selling make it more likely that prices will diverge from fundamentals. In housing markets, transactions costs are higher and short-selling is more difficult than in almost any other asset market (e.g., Linneman, 1986; Wallace and Meese, 1994; Rosenthal, 1989). Thus, we should not be surprised that the predictability of housing price changes (Case and Shiller, 1989) and seemingly large deviations between housing prices and fundamentals create few opportunities for arbitrage.
The extraordinary nature of the recent boom in housing markets has piqued interest in this issue, with some claiming there was a bubble (e.g., Shiller, 2005). While nonlinearities in the discounting of rents could lead prices to respond sharply to changes in interest rates in particular in certain markets (Himmelberg et al., 2005), it remains difficult to explain the large changes in housing prices over time with changes in incomes, amenities or interest rates (Glaeser and Gyourko, 2006). It certainly is hard to know whether house prices in 1996 were too low or whether values in 2005 were too high, but it is harder still to explain the rapid rise and fall of housing prices with a purely rational model.
However, the asset pricing literature long ago showed how difficult it is to confirm the presence of a bubble (e.g., Flood and Hodrick, 1990). Our focus here is not on developing such a test, but on examining the nature of bubbles, should they exist, in housing markets.
House price volatility is worthy of careful study in its own right because it does more than just transfer large amounts of wealth between homeowners and buyers. Price volatility also impacts the construction of new homes (Topel and Rosen, 1988), which involves the use of real resources that could involve substantial welfare consequences. When housing prices reflect fundamentals, those prices help migrants make appropriate decisions about where to live. If prices, instead, reflect the frothiness of irrational exuberance, then those prices may misdirect the migration decisions that collectively drive urban change.
Most asset bubbles also elicit a supply response... Models of housing price volatility that ignore supply miss a fundamental part of the housing market. Not only are changes in housing supply among the more important real consequences of housing price changes, but housing supply seems likely to help shape the course of any housing bubble. We show this is, indeed, the case in Section II of this paper, where we develop a simple model to investigate the interaction between housing bubbles and housing supply. ...
We model irrational, exogenous bubbles as a temporary increase in optimism about future prices. Like any demand shock, these bubbles have more of an effect on price and less of an effect on new construction where housing supply is more inelastic. Even though elastic housing supply mutes the price impacts of housing bubbles, the social welfare losses of housing bubbles may be higher in more elastic areas, since there will be more overbuilding during the bubble.
We also endogenize asset bubbles by assuming that home buyers believe that future price growth will resemble past price growth. Supply inelasticity then becomes a crucial determinant of the duration of a bubble. When housing supply is elastic, new construction quickly comes on line as prices rise, which causes the bubble to quickly unravel. The model predicts that building during a bubble causes post-bubble prices to drop below their prebubble levels. ...
We then examine data on housing prices, new construction and supply elasticity during periods of price booms and busts. While this empirical analysis is not a test for bubbles, much of the evidence is consistent with the conclusions of our models. For readers who resolutely do not believe in bubbles, the empirical results in the paper still provide information on the nature of housing price volatility across markets with different supply conditions.
In performing the empirical analysis, we distinguish between areas with more or less housing supply elasticity using a new geographical constraint measure developed by Saiz (2008). We also investigate differences in price and quantity behavior between the most recent boom and that which occurred in the 1980s.
During both the 1980s boom and the post-1996 boom, more inelastic places had much larger increases in prices and much smaller increases in new construction. Indeed, during the 1980s, there basically was no housing price boom in the elastic areas of the country. Prices stayed close to housing production costs. If anything, the gap in both price and quantity growth between elastic and inelastic areas was even larger during the post-1996 boom than it was during the 1980s. However, in the years since 1996, there were a number of highly elastic places (e.g., Orlando and Phoenix) that had temporary price explosions despite no visible decline in construction intensity.
The fact that highly elastic places had price booms is one of the strange facts about the recent price explosion. Our model does not suggest that bubbles are impossible in more elastic areas, but it does imply that they will be quite short, and that is what the data indicate.
While the average boom in inelastic places lasted for more than four years, the average duration of the boom in more elastic areas was 1.7 years.
Does the housing bust of 1989-1996 offer some guidance for the post-boom years that are ahead of us? During that period, mean reversion was enormous. For every percentage point of growth in a city’s housing prices between 1982 and 1989, prices declined by 0.33 percentage points between 1989 and 2006. The level of mean reversion was more severe in more inelastic places, but on average, elasticity was uncorrelated with either price or quantity changes during the bust.
Relatively elastic markets such as Orlando and Phoenix have not experienced sharp increases in prices relative to construction costs in the past, so they have no history of substantial mean reversion. Yet some insight into their future price paths might be provided by the fact that for two decades leading up to 2002-2003, prices in their markets (and other places with elastic supply sides) never varied more than ten to fifteen percent from what we estimate to be minimum profitable production costs (MPPC), which is the sum of physical production costs, land and land assembly costs, and a normal profit for the homebuilder.
Those three factors sum to well under $200,000 in these markets. If these markets return to their historical norm of prices matching these costs, then they will experience further sharp price declines. ...