⁶We should note here that personal saving rates tend to be one of the most heavily revised data series in the NIPA accounts, and that substantial revisions can occur even many years after the BEA’s “first final” estimates are made (Nakamura and Stark (2007), Deutsche Bank Securities (2012)); the revisions are systematically upward-biased, and often as much as 1–2 percent, so it would not be surprising if some years from now the saving decline in the 2000s appears to be smaller than in the data used here. It seems very unlikely, however, that either the broad trends or the business-cycle frequency movements will be revised greatly; past revisions have tended to be at medium frequencies, and not at either the very low or very high frequencies that tend to provide most of our identification.