Measurement Error and the Flow of Funds Accounts: Estimates of HouseholdAsset Demand Equations

In the household sector of the Flow of Funds Accounts, the difference between net acquisition of financial assets and net financial savings is equal to a statistical discrepancy which is often quite large relative to the reported changes in asset holdings. This means that the budget restrictions emphasized in the Brainard-Tobin approach to specifying asset demand equations are not satisfied by the data commonly used to estimate such equations. The view adopted in this paper is that the statistical discrepancy should be thought of as resulting from measurement error in the Flow of Funds data. By imposing a structure on the measurement error, a consistent estimator is developed and used to estimate asset demand equations for the household sector. The demand equations are similar in specification to those used by others so that the results allow a direct assessment of the effects of alternative treatments of the statistical discrepancy. The empirical results suggest that qualitative conclusions about the effects of financial flows and interest rates on asset demands are not affected by the way the statistical discrepancy is treated. Quantitative conclusions are, however, affected.