|
Abstract:
|
Federal law requires statistical agencies to mask the sensitive and identifiable information in order to protect household confidentiality when the agency releases survey data to the public. The Consumer Expenditure (CE) Survey employs a statistical disclosure limitation (SDL) process called "top-coding" for this purpose. For example, in the microdata released to the public, CE substitutes high values of household property tax with the average of all high household property tax values. Top-coding can have numerical impacts on the utility and quality of the publicly released data, especially for analyses that are sensitive to values in the tails of the distribution. For instance, the bias in the propensity of consumption model of a top-coded expenditure can induce inaccurate inferences from economic models that involve the estimated coefficient of propensity to purchase. In this study, we examine the consequences of top-coding on three economic models often used for analyzing CE data in the literature and where the coefficient estimate of the propensity of consumption is a component.
|
