Jump to:Page Content
As of December 2011, the S&P Case- Shiller housing price index had declined in 18 of the past 19 months. In constant dollars, its 10-city index is about the same as it was in 1989, which suggests two lost decades.
Yet the national housing blahs obscure the fact that some markets, such as New York and Washington, have seen prices rise substantially since 2000, while others, such as Detroit and Las Vegas, have experienced dramatic price declines.
The real value of housing in Detroit is 30 percent less than it was when Case-Shiller (ETSLTOTL) started tracking the city in 1991, and 52 percent less than it was a decade ago. Any real estate agents who told homebuyers that housing prices only go up deserve a place in the pantheon of great liars along with Pinocchio, Joe Isuzu and the Grand Duchess Anastasia.
The great housing boom and bust of the last decade wasn’t a unified national phenomenon. Figure 1(attached) relates growth in real housing prices between 2001 and 2006 with declines in prices between 2006 and 2011. I use the Federal Housing Finance Agency price indexes, because they are publicly available for far more metropolitan areas than the Case-Shiller data, and I include only those areas with more than 200,000 people in the 2000 census. To keep the units similar, I divided price growth in both periods by the price index for 2001.
Joseph told Pharaoh to expect seven lean years after seven fat years; that’s also pretty normal in housing markets. Typically, markets give back 32 cents of any extra price growth over a five-year period during the next five years.
But the past 10 years have been far more undulating than usual. Figure 1 shows that, on average, between 2006 and 2011, markets gave back 96 percent of their growth between 2001 and 2006. The boom of the early 2000s was a tsunami that came and went, leaving the average market more or less where it started.
The figure also illustrates the remarkable range of metropolitan experiences. Some markets, such as Dallas and Houston, had barely any movement during either period. Others, such as Las Vegas and Riverside, California, experienced extraordinary swings.
Those areas that lie significantly above the line experienced less decline between 2006 and 2011 than their earlier growth would have predicted, which means that they did relatively well over the entire decade. Washington and Midland, Texas, are particularly notable in this group (being one of President George W. Bush’s hometowns seems to have been good for housing prices). The areas below the line, such as Detroit and Las Vegas, experienced post-2006 price declines that were lower than their previous price growth would warrant. In these areas, real prices were down over the decade.
The best way I have found to explain why certain areas experienced the biggest price swings over the past decade is by looking at temperatures in January. Figure 2 shows the relationship between that variable and percentage growth in prices (again relative to the 2001 baseline) and January temperature.
Obviously, there were exceptions -- Dallas and Houston -- but overall, for every additional 10 degrees of January temperature, prices rose by an additional 12 percent. But after 2006, the connection between warmer temperatures and price decline was just as strong as the warmth-growth connection before that year. Over the entire decade, there is no statistical link between warm Januaries and price growth.
The fact that the great booms and busts occurred disproportionately in warmer places is a puzzle. Why should irrational exuberance dislike snow? One explanation is that because the Sun Belt has been growing rapidly for decades, buyers may have overestimated the impact that this trend would have on price growth. Certainly, buyers in Las Vegas and Phoenix don’t seem to have understood that the essentially unlimited housing supply in those areas will ultimately keep prices low even if population growth continued.
Although January temperature doesn’t explain price-growth variation over the entire decade, there are some variables that correlate with housing price growth between 2001 and 2011.
I’ve put together a little statistical model that can, with five variables for 2000, explain the majority of the heterogeneity in price growth between 2001 and 2011 for metropolitan areas with more than 400,000 people (as of 2000); the same five variables can explain over four-tenths of the variation in price growth for areas with more than 200,000 people. Figure 3 shows the relationship between actual price growth and the growth predicted by the model.
There were some big misses. Midland and Washington did much better than expected (the Bush effect) and Detroit did worse. Yet overall, the model seems to do a reasonable job “explaining” the data.
Before I discuss the model, a few disclaimers. Above all, this is a data-mining exercise, which means that it should be taken with plenty of grains of salt. Some of the correlations (even though they are “statistically significant”) may be spurious. Moreover, there is no reason to have any confidence that past price relationships will hold in the future. My model just tries to make sense of what happened over the last decade.
The five factors that best predicted price growth between 2001 and 2011 were the share of adults with college degrees, housing permits per capita, median family income, median housing value and the closest distance to the Atlantic Ocean or the Gulf of Mexico. To reduce the power of outliers to shape results, I took logarithms of all of the variables, except the percent of adults with college degrees. Apart from distance to the ocean, the variables all come from the U.S. Census Bureau.
The only variable that had a positive correlation with price growth was the share of adults with college degrees as of 2000. On average, if the share of adults with college degrees was 10 percentage points higher in 2000, price growth was 15 percent higher over the past decade. Given the strong correlation between education and urban success over the past four decades, we shouldn’t be surprised when skills predict price growth.
But college education only has such a positive link with price growth once we control for income, and income (controlling for education) is strongly negatively associated with price growth over this time period. Holding education constant, 10 percent higher family incomes in 2000 are associated with about 4 percent lower price growth between 2001 and 2011.
If places were earning more than their skills would suggest, as of 2000, then they experienced less price growth. This result is in line with decades of economic research that finds that -- holding skills constant -- poorer places experience faster economic growth.
Technology catches up in backward areas, and companies reduce labor costs by coming to places with low wages relative to skills. As previously poorer places catch up economically, their housing prices also grow more quickly.
Housing prices also typically converge, and higher prices in 2000 were strongly associated with less price growth between 2001 and 2011. Places with prices that were 10 percent higher in 2000 experienced, on average, 2 percent less price growth.
There are two plausible explanations for why price growth typically declined by 3 percent as distance to the coasts doubled. One explanation is that coastal America contains the most economically productive parts of the country, and demand for housing in these areas has been more robust. A second explanation is that there is just so much empty space between the oceans that abundant supply limits the growth of housing prices.
The fact that abundant supply limits price growth is certainly the best explanation why permits per capita are so strongly associated with lower price growth between 2001 and 2011. Figure 4 shows the relationship between average price growth and permits per 100 people in 2000, where I have grouped those areas with more than 10 permits per 1,000 together in the right-most bar. The areas with the least permitting activity in 2000 had average real price growth of about 15 percent; the areas with the most permitting activity had real price declines of 13 percent.
Oversupply and Demand
No matter which way the macroeconomic winds may blow, housing price will ultimately be determined by the intersection of supply and demand. Places that permit a lot of new homes, such as Las Vegas and Phoenix, will ultimately see limited housing price growth even if demand is robust. Places that permit very little will tend to be expensive, as long as they have the kind of dynamic skill-driven economy that keeps housing demand high.
Indeed, it’s precisely because I like affordable housing just as much as I like inexpensive computers and cars, that I typically favor fewer restrictions on housing construction.