Application of distributed lag and autocorrelated error models to short-run demand analysis
The objective of the research reported here was to investigate the usefulness of distributed lag economic models and autocorrelated error statistical models for analysis of monthly and quarterly food demand. Distributed lags are a way of incorporating dynamic considerations into econometric models of consumer demand. In the distributed lag model used here, current consumption is the dependent variable, and lagged consumption is one explanatory variable. Testing the significance of the coefficient of lagged consumption tests the hypothesis of a lag in consumer adjustment to conditions affecting demand.
The presence of autocorrelated errors can have serious effects on least squares (L.S.) estimates of coefficients. Autocorrelated errors may frequently occur in equations fitted to monthly and quarterly data. Therefore, equations were estimated by autoregressive least squares (A.L.S.) as well as by least squares. A.L.S.-1 assumes the errors ut to follow a first order autoregressive scheme, ut = β1ut-1 + et. It provides simultaneous estimates of β1 and of the coefficients in the demand equation. A.L.S.-2 assumes the errors to be generated by a second order autoregressive process, ut = β1ut-1 + β2ut-2 + et. It provides simultaneous estimates of β1, β2 and the coefficients in the demand equation.