3Dynan [2000] is the best known micro study; many others are reported in Havranek et al. [2017]. It is well-known that household-level data on consumption are subject to substantial measurement error. Consequently, attenuation of the micro estimates of χ  might be expected (even though many studies use estimation techniques robust to measurement error). Micro studies have been sensitive to this challenge from Dynan [2000] onward, and have not found any method of controlling for measurement error that makes the estimates appreciably higher.