** Chapter 22 ** for W.H. Greene, Econometric Analysis 6th ed. *****************
* (c) Noel Roy 2003, 2008
*
*                        NONSTATIONARY DATA
*
*===============================================================================
* 22.3 COINTEGRATION
* 
* Example 22.5 (p. 757) Cointegration in Consumption and Output
*
READ (TableF5-1.prn)Year qtr realgdp realcons realinvs realgovt realdpi cpi M1 i unemp pop pi realint / SKIPLINES=1
GENR loggdp=LOG(realgdp)
GENR LOGCONS=LOG(REALCONS)
* Replicate Figure 22.10.
GRAPH LOGGDP LOGCONS /LINEONLY TIME
*
* The COINT command with option TYPE=RESD performs Dickey-Fuller tests on the 
* residuals of cointegrating regressions. The first variable in the list of
* variables is regressed on the remaining ones to obtain the cointegrating
* regression. SHAZAM tests cointegrating equations both with (labelled COINTEGRATING 
* REGRESSION - CONSTANT, TREND) and without (labelled COINTEGRATING REGRESSION - 
* CONSTANT, NO TREND) a time trend. The Example in the text is without a time trend.
* The number of lagged first differences is set by the NLAG= option; if not specified,
* this value is set as the highest significant (at 95%) lag order from either the
* autocorrelation function or partial autocorrelation function of the first
* differenced series (up to the dquare root of the sample size).
* By default, SHAZAM does not report the estimate of the cointegrating
* vector, but will do so if the MAX option is specified.
*
COINT LOGCONS LOGGDP /TYPE=RESD SIGLEVEL=5 MAX NLAG=1
*
*===============================================================================
* 22.3.3 Testing for cointegration
*
* The Johansen Test can be generated by a SHAZAM Procedure that can be 
* downloaded from http://shazam.econ.ubc.ca/examples/johansen. See also pp.174-44
* of the SHAZAM Manual.
* The test would be implemented as follows:
* FILE PROC JOHANSEN
* NLAG: 2
* VARS: realcons realgdp
* EXEC JOHANSEN
* Thw test generates the following output, where r is the number of 
* cointegrating vectors, LTRACE_ is a test of r>=R_, and LMAX_ is a test of r=R_.
*
*       R_             LTRACE_        LMAX_
*        1.             0.041        0.041
*        0.            20.784       20.743
* The hypothesis r=1 is clearly accepted (5% critical value is 8.1 for both tests),
* and r=0 rejected (5% critical value is 17.8 for LTRACE and 14.6 for LMAX).
*
*
stop
*
*===============================================================================
* Updated November 21, 2008