A Quick and Dirty Estimate of Measurement Error in Household Survey Consumption Expenditures: Application to Indonesian Data.

Measurement error is an enormous problem in empirical work. In some types of analysis, it is often ignored for various reasons. In some others, however, it cannot be ignored because it affects the results of analysis significantly. We use a simple procedure to estimate the extent of measurement error in household consumption expenditures data by comparing the coefficients obtained from OLS and instrumental variable regressions. We applied the procedure into school participation, durable goods expenditures, and clothing expenditures models using 10 different household survey data sets from Indonesia. The results indicate that measurement error accounts for less than 20 percent of total variance found in cross section household expenditures. In general, surveys which use detailed consumption questionnaire have smaller measurement error than those which use aggregated consumption questionnaire.