Test for cointegration                                            (STB-7: sts2)
----------------------


	^coint^ varlist [, ^l^ags(#) ^t^rend ^v^ector regress-options]



Description
-----------

^coint^ tests the null hypothesis that the variables in varlist are NOT coin- 
tegrated.  The Engle-Granger test is used.  In this test, a cointegrating 
vector is estimated by ordinary least squares with the normalization that the 
first variable in varlist is the left-hand-side variable.  Next, ^unitroot^ is 
called to test the null hypothesis that the residuals from this regression have
a unit root.  The test statistics have a non-standard distribution.  Critical 
values can be found in Engle and Yoo (1987).



Options
-------

^lags(^#^)^ specifies the number of lags to be considered.   The default is 4.

^trend^ includes a time trend term in the regression.

^vector^ displays the cointegrating regression.

regress_options refers to any of the options of ^regress^; see ^help regress^.




(Examples follow)








Examples
--------


 . ^coint m2 cpi^

tests the hypothesis that M2 and the price level (as measured by the consumer 
price index) are not cointegrated.  The Dickey-Fuller test is calculated using 
0 to 4 lags of the first difference of the residual from the first-stage 
regression (the cointegrating regression).


 . ^coint m2 cpi, lags(6)^

would do the same except it would consider using 0 to 6 lags.



References
----------

Engle and Yoo.  1987.  Forecasting and testing in co-integrated systems. 
    ^Journal of Econometrics^ 35(1), May, Table 2, p. 157.



Author
------
 
Sean Becketti, Federal Reserve Bank of Kansas City.



Also see
--------

    STB:  sts2 (STB-7)
On-line:  ^help^ for ^findlag^, ^lag^, ^unitroot^