PANEL Procedure

In general, there must be at least one cross section that has more than one time series observation. Some estimation methods might have more stringent requirements; for example, the Amemiya-MaCurdy estimator requires data that are balanced. Some estimators require that there be more cross sections than time series values. When the data are insufficient for an estimator, check the log for error messages that provide further details.

If you are using the Parks method (by specifying the PARKS option in the MODEL statement) and the number of cross sections is greater than the number of time series observations per cross section, then PROC PANEL produces an error message that states that the matrix is singular. This is analogous to a seemingly unrelated regression that has fewer observations than equations in the model. To avoid this problem, reduce the number of cross sections.

It is vitally important that you sort your data by cross sections and by time periods within cross sections. As PROC PANEL steps through the observations in the data, it treats any change in the value of the cross section ID variable as a new cross section, regardless of whether it has encountered that value previously. If you do not sort your data, the results might not be what you expect.

PROC PANEL is not supported for data sets that have duplicated time values within cross sections. If data with such duplication are encountered, PROC PANEL issues an error message such as the following:

"The data set is not sorted in ascending sequence with respect to time series ID. The current time period has year=1955 and the previous time period has year=1955 in cross section firm=1."