Tuesday, January 01, 2008
Paul Carrillo, Dirk Early and Ed Olsen have created a great new data set on Rent
It gives indexes of quality adjusted rents across American metropolitan areas. I am discussing the paper at the ASSA meetings this week; the authors finds a method for constructing hedonic price indexes for rent for MSAs and rural areas. They then compare costs for a typical, constant quality unit across MSAs. This is a very useful exercise--I will speak on issues involved with index construction, but my criticisms will be more quibbles than anything else.
The data they produce are fun to play with. The graph to the left is a scatter plot of rent against per capita income for 90 cities. The relationship is pretty tight, the correlation coefficient being about .55. Not bad for one variable.
If we look at another scatter plot between population and rent, we find a rather non linear relationship: there seems to be a critical population above which rents rise quite dramatically. These findings are rather good news for those of us who think the standard urban model gives us insight into systems of cities.
A regression using both income and population to explain rent produces an equation where an increase of 1 million population adds 2 percent to rent; while an income change of $1000 produces also a 2 percent change in cross section. Highest household income in the data was 22000 more than the lowest, meaning that income could explain a 44% difference in rent.
Still, one of the striking things about the CEO data is how little variation in rental costs there are in the US relative to house prices. This is the puzzle that is worth substantial examination.