CPANEL Procedure

You perform instrumental variables (IV) regression by specifying one or more of the IVBTWNG, IVFIXONE, IVFIXONETIME, IVFIXTWO, IVG2SLS, IVPOOLED, IVRANONE, and IVRANTWO options in the MODEL statement.

When you use the ENDOGENOUS statement for IV regression, it selects the subset of effects from the MODEL statement that are endogenous—that is, the effects that are correlated with the observation-level error term.

If you specify an effect in the ENDOGENOUS statement but not in the MODEL statement, then the effect is added to the list of model effects and treated as endogenous. If you have multiple MODEL statements, then each ENDOGENOUS statement is associated with the MODEL statement that immediately precedes it.

The ENDOGENOUS statement is used in conjunction with the INSTRUMENTS statement to perform IV regression. The following statements perform IV fixed-effects regression with the model effects X1, X2, Z1, and Z2. The effects X2 and Z2 are endogenous, and G1 and G2 are external instruments.

proc cpanel data =  mylib.a;
   id firm year;
   model Y = X1 X2 Z1 Z2 / ivfixone;
   endogenous X2 Z2;
   instruments G1 G2;
run;

You perform dynamic panel estimation by specifying the DYNDIFF or DYNSYS option in the MODEL statement.

When you use the ENDOGENOUS statement in dynamic panel estimation, it specifies a set of variables whose instruments follow a prescribed form that is appropriate for endogenous regressors.

If you have multiple MODEL statements, then each ENDOGENOUS statement is associated with the MODEL statement that immediately precedes it.

You can specify the following option after a slash (/):

The following statements perform dynamic panel estimation by using GMM-style instruments for the endogenous variable Sales and standard instruments for the exogenous variable Price. Both sets of instruments are included only in the difference equations.

proc cpanel data =  mylib.a;
   id firm year;
   model Sales = Price / dyndiff;
   endogenous Sales / eq = diff;
   instruments Price / eq = diff;
run;

For both IV regression and dynamic panel estimation, you can also specify classification variables, interactions, and nested effects in the ENDOGENOUS statement. For more information about constructing model effects, see the section Specification and Parameterization of Model Effects in Chapter 4, Shared Concepts.