Time Series Autocorrelation for Panel Data with Stata (Panel)

Stata time-series operator can be applied to panel data when both panel and time identifier are set with the xtset command.

 The model we estimates;

We use the data Paneldata01.

To regress Eq(1), first, sort id and t and then regress;

xtset id t

sort id t

regress lwage exp exp2 wks ed,vce(cluster id)

 

To compute first-order autocorrelations for panel data for certain variable;

correlate lwage L.lwage

 

To calculate autocorrelation at all lags (six periods) for the residual Eq(1);

quietly regress lwage exp exp2 wks ed,vce(cluster id)

predict uhat,residuals

forvalues  j = 1/6 {

     quietly corr uhat L `j’.uhat

     display “Autocorrelation at lag `j’ = “%6.3f r(rho)

     }

 

Wooldridge (2002, 282–283) derives a simple test for autocorrelation in panel-data models. 

Drukker (2003) provides simulation results showing that the test has good size and power properties in reasonably sized samples. 

There is a user-written program, called xtserial, written by David Drukker to perform this test in Stata;

xtserial lwage exp exp2 wks ed, output

 

The null hypothesis for the test is there is no serial correlation in the model.

The results show that the null hypothesis of no serial correlation is strongly rejected, means that there is serial correlation in the model.