Numerous cross sectional studies have identified significant saving differences between the rich and poor suggesting that saving is affected by distributional influences. However, these influences cannot be measured with current versions of aggregate national income data. While the Census Bureau provides annual income distribution estimates by quintile, expenditure estimates by quintile, the other half of the saving calculation, have not been undertaken.
Annual Bureau of Labor Statistics (BLS) Consumer Expenditure surveys are the only source of nationally representative expenditure data. After considering various objections and showing similarity with Census income distributions, survey shares are used to disaggregate National Income and Product Accounts (NIPA) data into annual income and expenditure quintiles for years between 1984 and 2014. While expenditure shares cannot be validated with U.S. data, they are not statistically different from United Kingdom and Canadian shares for the same time periods.
Basically, saving by the fifth quintile is sufficient to offset dissaving by the lower four. The aggregate saving decline is explained by a collapse of saving by the middle three quintiles, a result also found by Nelson using Survey of Consumer Finances (SCF) wealth data. In terms of disposable income shares, saving by the top quintile with a 48 percent share has been insufficient to offset the saving decline by the middle three quintiles, also with a 48 percent share. The collapse is consistent with changes in household wealth as a consequence of housing speculation beginning in the late 1990s and ending around 2010 as found by Nelson, or the rise in middle quintile negative net worth as shown in the survey data.
Deconstruction of NIPA data is a topic of current interest. Efforts are underway at the Bureau of Economic Analysis, by Piketty, Saez and Zucman and by Nelson. Rather than revising values, this paper focuses on shares. The resulting estimates are unique. Annual quintile expenditure and saving shares consistent with NIPA figures have not been published. Disposable income shares have not been published since 2009. The paper shows that consumer expenditure survey data can be successfully used for aggregate analysis despite widespread reservations and implies some serious limitations on aggregate analysis of household spending behavior.