You can use the CPANEL procedure in SAS^® Econometrics to perform instrumental variable regression with panel data. The CPANEL procedure is available only in SAS Viya and requires a license for both SAS Viya and SAS Econometrics.

The CPANEL procedure currently supports the following instrumental variable panel regression models using the specified option in the MODEL statement:

• Instrumental variables pooled regression model (IVPOOLED option)
• Instrumental variables between-groups regression model (IVBTWNG option)
• Instrumental variables one-way fixed effects model (IVFIXONE option)
• Instrumental variables one-way time fixed effects model (IVFIXONETIME option)
• Instrumental variables two-way fixed effects model (IVFIXTWO option)
• Instrumental variables one-way random effects model (IVRANONE option)
• Instrumental variables two-way random effects model (IVRANTWO option)
• Instrumental variables generalized two-stage least squares model (IVG2SLS option)

The CPANEL procedure performs instrumental variables regression by the method of either two-stage least squares (2SLS) or the efficient generalized method of moments (GMM). 2SLS is a special case of GMM and is equivalent to the one-step GMM method. The efficient GMM method, which is also called two-step GMM, is a GMM that uses the optimal weighting matrix. You request the efficient GMM method for instrumental variable regression by using the GMM=TWOSTEP option in the MODEL statement together with an instrumental variable regression model option in the list above. For the two-stage least squares instrumental variable method, you do not need to specify the GMM= option because GMM=ONESTEP is the default when an instrumental variable regression model is specified.

You can obtain robust standard errors for instrumental variable models using the ROBUST option in the MODEL statement. The robust covariance matrix obtained with the ROBUST option is robust against both heteroscedasticity and serial correlation. Similar to two-step GMM estimation for a dynamic panel model, you can also obtain a bias-corrected variance of two-step GMM instrumental variable models by using the BIASCORRECTED option in the MODEL statement. The bias-corrected standard errors only correct for the bias from the two-step GMM. They are not robust to serial correlation.

For more details about panel data instrumental variable regression models that are supported in the CPANEL procedure, refer to the CPANEL procedure chapter in the SAS Econometrics documentation. For an example of fitting an instrumental variable regression model to panel data and obtaining robust covariance matrix estimates using the CPANEL procedure, see the example titled "Crime Rates in North Carolina: Instrumental Variables Regression" in the CPANEL procedure chapter in the SAS Econometrics documentation.