Skip to content

Normalization and Cointegrating Vectors from VECM analysis

2 messages · Hammed, Guy Yollin

Original
Hammed
# 
On page 24 of Bernhard Pfaff's  "VAR, SVAR and SVEC Models: Implementation
Within R Package vars" the analysis noted the existence of one cointegration
relationship. Subsequently a VECM with the restriction that r=1 and a
normalization of the long-run relationship with respect to total wages was
re-estimated. This was done via

R>vecm <-cajo(Canada[, c("rw", "prod", "e", "U"], type="trace",
ecdet="trend", K=3, spec="transitory")
R> vecmR1 <-cajorls(vecm, r=1). 

Here are the questions I have:

1. Does placing "rw" first signal that it is the variable on which the VECM
estimation will be normalized? If  not, how is normalization conducted?

2. How does one go about obtaining the t-statistics (and standard errors) of
the alpha and beta parameters as provided in Table 5? I have searched the
net and noted that other users have had similar issues in obtaining
something similar to Table 5 when one gets results such attached below
LY.d        LPVI.d      LGC.d       LPI.d     

ect1      -1.281e-01   2.857e-01  -7.791e-01   1.412e+00
Guy Yollin
# 
Hammed,

Pfaff presents the details of the SVEC model in section 8.3 of his text "Analysis of Integrated and Cointegrated Time Series with R".

All of the scripts used in the text are available from his website:

http://www.pfaffikus.de/ex2.html

Rcode-8-8.R through Rcode-8-17.R perform the analysis of the Canada dataset

Specifically, Rcode-8-13.R performs the calculation of t-stats and SEs for alpha and beta.

Best,

-- G