The PDLREG Procedure

This example estimates the demand for money by using the following dynamic specification:

where

The following DATA step reads the data and transforms the real money and real GNP variables using the natural logarithm. Refer to Balke and Gordon (1986) for a description of the data.

data a;
   input m1 gnp gdf r @@;
   m    = log( 100 * m1 / gdf );
   lagm = lag( m );
   y    = log( gnp );
   p    = log( gdf / lag( gdf ) );
   date = intnx( 'qtr', '1jan1968'd, _n_-1 );
   format date yyqc6.;
   label m    = 'Real Money Stock (M1)'
         lagm = 'Lagged Real Money Stock'
         y    = 'Real GNP'
         r    = 'Commercial Paper Rate'
         p    = 'Inflation Rate';
datalines;
187.15  1036.22    81.18     5.58

   ... more lines ...   

Output 21.2.1 shows a partial list of the data set.

The regression model is written for the PDLREG procedure with a MODEL statement. The LAGDEP= option is specified to test for the serial correlation in disturbances since regressors contain the lagged dependent variable LAGM.

title 'Money Demand Estimation using Distributed Lag Model';
title2 'Quarterly Data - 1968Q2 to 1983Q4';

proc pdlreg data=a;
   model m = lagm y(5,3) r(2, , ,first) p(3,2) / lagdep=lagm;
run;

The estimated model is shown in Output 21.2.2 and Output 21.2.3.