Skip to content

Extension of Johansen Procedure ca.jo

2 messages · Johannes Lips

#
Dear list,

I am currently working on an extension of the basic ca.jo() function to 
also make it possible to incorporate structural breaks, like in Johansen 
et al. (2000). What I basically do is to add a matrix, which 
incorporates possible structural breaks in the cointegration vector. 
Therefore I added the function paramter break.matrix, which so far takes 
the break matrix according to the H_l(r) case of Johansen et al. (2000).
I construct the dummy matrix according to Joyeux (2007) and add them to 
the dumvar and the break.matrix parameter of the ca.jo function.
So far it "works", but I am unsure if I made all the adjustments of the 
ca.jo function correctly and I would be really glad if someone could 
also take a look at the function and see if it does, what it's supposed 
to do.
I uploaded the function to my github repo at [1], where I also provide 
the code to create the dummy matrix incorporating the possible 
structural breaks. Please note, that the creation of the dummy matrix is 
also not yet finished and most probably also needs some changes and is 
only for the purpose of testing the ca.jomoni() function.

Thanks in advance,
Johannes

[1] https://github.com/hannes101/CointegrationAnalysis



References:
Johansen, S?ren, Rocco Mosconi, and Bent Nielsen (2000). ?Cointegration 
analysis in the presence of structural breaks in the deterministic 
trend?. en. In: Econometrics Journal 3.2, pp. 216?249. doi: 
10.1111/1368- 423X.00047 (cit. on p. 6).

Joyeux, Roselyne (2007). ?How to Deal with Structural Breaks in 
Practical Cointegration
Analysis?. English. In: Cointegration for the Applied Economist. Ed. by
B. Bhaskara Rao. 2nd edition. Palgrave Macmillan, p. 256 (cit. on p. 6).
10 days later
#
Dear all,

I extended the basics of the ca.jo function to incorporate five 
different cases:
nc: no constant
rc: restricted constant, i.e. constant in cointegration vector
uc: unrestricted constant, i.e. constant in the deterministic part of 
the model
crt: restricted constant + trend, i.e. constant and trend in the 
cointegration vector
ct: constant + unrestricted trend, i.e. constant and trend in the 
deterministic part of the model
I checked the results against the results of JMulti with the same data 
set and the same case.
Please mind, that the results can't be used for further analysis in the 
"vars" package yet, since this will need some more work on the 
underlying ca.jo class of results.
Also I didn't pay a lot of attention to the "transitory" specification, 
so if there are any errors, please let me know.

Best,
Johannes
On 16.10.2015 14:27, Johannes Lips wrote: