Most new borrowing during America's housing boom was for spending

The Economist:

HOW big an influence on spending is housing wealth? Hopes for a consumer revival in countries where house prices have slumped rest, in part, on the answer. A purist view is that the value of homes has no "wealth effect" on consumption. An increase in house prices only raises the future cost of shelter. Those about to trade down or sell out receive a windfall, but first-time buyers and those hoping to buy a bigger home are worse off. The overall effect on wealth is a wash. But even if that is correct, house-price increases may still have an impact as they create housing collateral for consumers who could not otherwise borrow. A study* by Atif Mian and Amir Sufi of the University of Chicago's Booth School of Business pins down the size of this effect, using the credit records of almost 70,000 American borrowers.

It could be that an unseen influence, such as greater optimism about future earnings, pushed up both house prices and debt. So the authors use their granular data to first establish a link between the two, which is apparent in the aggregate figures up to 2006 (see chart). They found that house prices and household debt increased most where the supply of new housing was limited--in places that are hemmed in by hills, rivers or the ocean. But in cities where housing supply is very "elastic"--where homes can easily be built to meet demand and prices did not rise--debt barely rose either. This suggests that house-price rises led to more borrowing.

How much of this was simply down to new buyers needing bigger home loans? By limiting their sample to those who were already homeowners in 1997, before the boom in housing and credit, the authors were able to measure how much of the rise in debt was the result of cashing in on higher home values. They reckon almost 60% of increased debt between 2002 and 2006 came from this source. Put another way, every $1 increase in home values led to a rise of 25-30 cents in borrowing. That is far bigger than some long-standing estimates of the wealth effect from rising asset values, which are in the 3-5 cent range (though these include the response of renters, too).