Monetary Policy and House Prices: Evidence from the States

Did the Fed cause the housing bubble?  I don’t know the answer but the issue has received renewed interest since Governor Bernanke gave a talk at the AEA meetings categorically denying any culpability on the part of the Fed.  I too question the ability of monetary policy to create a bubble in housing (or stocks) without, over a long period of time, also creating a high level of inflation.  However, if we examine evidence across U.S. states instead of across countries, where the data may or may not be comparable, there is a statistically significant relationship between the easiness of monetary policy and the rate of house price appreciation.

Bernanke’s Comparison Across Countries

In addition to a weak model-based defense, Bernanke showed the following picture.  The scatter plot shows the relationship between the average residual in the Taylor rule between 2001Q4 and 2006Q4 against the cumulative change in real house prices.  The relationship shows a weak but positive relationship between the change in house prices and the looseness of monetary policy.  The lack of statistical significance I cannot debate; however, Bernanke also referred to a lack of economic meaning, which seems an absurd statement given the Fed’s long history of speaking on the relationship between house prices and monetary policy.  Indeed, in the early 2000s, the Fed’s favorite transmission mechanism for monetary policy was through the housing market. 

Although the lack of relationship may be real, cross-country comparisons of this nature are plagued by data and comparability problems.  The measurement of house prices (as well as the measurement of the deflator) differs tremendously across countries.  Consider the data for the United States.  If the slide had used the Case-Shiller index instead of the FHFA index, the data point for the United States would be much higher.  As well, if the house price data had been deflated by the PCE deflator instead of CPI the data point would move. 

Comparison Across States

To eliminate a host of comparability problems, I replicated Bernanke’s picture using U.S. states in lieu of countries.  The BEA produces comparable annual estimates of state GDP and the FHFA produces comparable house price indices for every state.  Using these two sources, I calculate Taylor-rule residuals and average house price growth across the states and plot the results below.  This exercise is very much akin to Bernanke’s picture considering euro area countries alone (an exercise which yields a statistically significant relationship). 

States with the “loosest” monetary policy are exactly those with the highest house price appreciation.  Unlike the picture across countries, here the relationship is statistically significant.  Indeed, those States that are “looser” than the U.S. average (-2.5) experienced more than 30 percent more house price appreciation on average than those states “tighter” than the U.S. average. 

By the way, the intercept of the line, 23.2, is only a touch below the average U.S. house price growth over the last 30 years.  In other words, as the simple regression above predicts, when monetary policy is neither too loose nor too tight, house prices should grow about 4 percent per year (or 23.2 percent over five years).  If we believe this result, then the Fed’s loose policy stance over this period added 32 percentage points to the growth rate of house prices between 2001 and 2006.  Since total growth was 48 percent, the Fed’s policy accounted for two-thirds of the total increase in house prices, according to this calculation. 

House Prices vs. GDP and Inflation

In his speech, Bernanke pointed out that the relationship might be spurious:  those economies with the highest growth might be precisely those with the loosest policy.  I am not sure the logic follows for certain; the statement seems to imply some additional unstated constraints on monetary policy. 

In any case, the picture below shows the scatter plot for the states where average annual GDP growth is substituted for the Taylor residuals.  Note that the R² is actually zero.  The slope is not only statistically insignificant, the point estimate is zero. 

Likewise, no substantial relationship can be found plotting state average annual inflation.  Indeed, the statistically insignificant slope goes in the wrong direction:  states with the highest inflation had, on average, the lowest nominal appreciation of house prices. 

Conclusion

These pictures are far from proof that monetary policy caused the house price boom.  But, they can also not be easily dismissed.  Although I personally do not believe monetary policy is capable of producing these types of long-term effects, the case is far from closed. 

I find the statistical fit across states disturbing.  The simple Taylor model yields not only a believable distribution, out of sample, it nails the long run average growth of house prices.  I know most economists reject this exercise within the United States but I would like to hear Bernanke explain why it is okay for the euro area but not for the U.S. 

In my next post, I will explore the theoretical link between house prices and monetary policy.  Conditional on the Fed having some control over the long-term rate, the link is direct and large.  The post will show that Bernanke's interest rate chart in his speech was likely misleading.