Pooling Time-Series Cross-Section Data Data sets may combine time series and cross section data. Two types of data sets are: A data set with cross-sections such as states, provinces or countries. A micro-panel or longitudinal data set constructed from a survey of the same micro-units over time. This type of data may typically contain a large number of cross-sectional units (for example, families or individuals) and relatively few time periods. A pooled regression model assumes common coefficients across the cross-section units. The POOL command in SHAZAM provides features for estimating models that combine time series and cross section data. The general command format is: POOL depvar indeps / NCROSS=n options where depvar is the dependent variable, indeps is a list of the explanatory variables and options is a list of desired options. The NCROSS= option specifies the number of cross-section units. The data must be arranged so that all observations of a cross-section are together. That is, the complete time series for the first group must be followed by the complete time series for the second group etc. Examples The examples in this section use the investment demand data set that is analyzed in Greene [2000, Chapter 15]. Pooling by OLS with Panel-Corrected Standard Errors and Dummy Variables Estimation with AR(1) Errors Cross-Section Heteroskedasticity and Time-Wise Autoregression Notes Including lagged variables as explanatory variables Estimation with a sub-set of time series observations References Nathaniel Beck and Jonathan N. Katz, "What to do (and not to do) with Time-Series Cross-Section Data", American Political Science Review, Vol. 89, 1995, pp. 634-47. Nathaniel Beck, Jonathan N. Katz, R. Michael Alvarez, Geoffrey Garrett and Peter Lange, "Government Partisanship, Labor Organization and Macroeconomic Performance: A Corrigendum", American Political Science Review, Vol. 87, 1993, pp. 945-948. A. Bhargava, L. Franzini and W. Narendranathan, "Serial Correlation and the Fixed Effects Model", Review of Economic Studies, Vol. 49, 1982, pp. 533-549. A. Buse, "Goodness of Fit in Generalized Least Squares Estimation", American Statistician, Vol. 27, 1973, pp. 106-108. William H. Greene, Econometric Analysis, Fourth Edition, 2000, Prentice-Hall. J. Kmenta, Elements of Econometrics, 1986, Macmillan. Richard W. Parks, "Efficient Estimation of a System of Regression Equations when Disturbances are both Serially and Contemporaneously Correlated", Journal of the American Statistical Association, Vol. 62, 1967, pp. 500-509. [SHAZAM Guide home] Including lagged variables as explanatory variables When the POOL command is used, if lagged variables are included as explanatory variables then they should be specified using the special form: var(first.last) where var is a variable name and the numbers in parentheses specify the first and last periods to use for lags. For example, the next command implements pooled estimation with a lagged dependent variable. POOL Y Y(1.1) X / NCROSS=4 [Back to Top] [SHAZAM Guide home] Estimation with a sub-set of time series observations When the POOL command is used, a sub-set of time series observations can be selected. This is demonstrated in the next list of SHAZAM commands. SAMPLE 1 100 * Set the number of time periods GEN1 NT=20 * Generate an index for each cross-section GENR CSINDEX=SUM(SEAS(NT)) * Generate a time index for each cross-section GENR TINDEX=TIME(0)-NT*(CSINDEX-1) * Estimate over the time period 1 to 15 SET NOWARNSKIP SKIPIF (TINDEX.GT.15) POOL I F C / NCROSS=5 DELETE SKIP$ STOP [Back to Top] [SHAZAM Guide home]