³¹The specification is motivated by Calomiris, Longhofer, and Miles (2009), p. 17, who argue that our measure of housing wealth is problematic because “more than half of ‘housing wealth’ consists of non-housing wealth” and claim that this fact biases upward our estimates of the housing wealth effect; they propose an alternative measure which they claim is a better measure of housing wealth, and show that using their measure the housing wealth effect is smaller. But their measure, oddly, subtracts mortgage debt from financial assets to yield a supposed measure of ’financial wealth,’ while counting housing assets as though they were unencumbered by debt. We considered several measures of housing versus nonhousing wealth in earlier drafts of this paper, but confess that this surprising configuration did not occur to us. Most choices that we did try yielded results similar to or stronger than those we present as our baseline. For example, our estimate of the immediate MPC out of housing wealth of 0.022 for model 7 suggests that the results are quite robust to measuring wealth. (The eventual housing MPC is larger than for the baseline specification because of the high χ for model 7. We see no reason to change our original view that, on balance, the evidence is mildly supportive of a larger effect for housing wealth than for nonhousing wealth, but that the hypothesis that the two wealth effects are equal cannot be rejected. And, as noted at the outset of the paper, we are sympathetic to the possibility that what this literature calls a ’wealth effect’ may instead be a reflection of a correlation between wealth and other variables that are harder to measure, like growth expectations or credit conditions. Those are questions unlikely to be answerable using aggregate data, though microeconomic and regional data offer hope of eventual resolution of the problem.)