46Essentially similar results are obtained if we assume that households have heterogeneous discount factors, in the style of Carroll et al. [2017]. Using a calibration of the distribution of β  that approximately matches the distribution of net worth in the U.S., the results presented in Table 6 are effectively unchanged (table available upon request). The main results hold whether β  is chosen to match aggregate asset holdings, the wealth of the median household, or the entire distribution of wealth; it is not sensitive to the particular calibration of the model.