Accounting for regime shifts in a panel cointegration model

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

maragloria
Posts: 101
Joined: Tue Jun 29, 2010 7:09 am

Accounting for regime shifts in a panel cointegration model

Postby maragloria » Sat May 13, 2017 2:00 pm

Hello there,

I would like to get comments on whether or not the panel estimation approach I describe below make econometric sense :

- I have a macroeconomic data for 30 countries (G10 + 20 Emerging), from 2002Q1 to 2016Q3 (59 observations)
- Since the 30 countries are quite heterogeneous, instead of running the estimations on a single panel of 30 countries, I instead split it into smaller panels, the smallest containing 4 countries (thus, 236 observations)
- I’m estimating different specifications with up to 6 explanatory variables
- I’ve run unit root tests (both panel and single-equation) on all variables and concluded that most are I(1), with some evidence of I(0) for a few variables for certain countries
- I started by running individual cointegration tests (EG/PO) on each cross-section. The null of no cointegration was not rejected for most countries, across different specifications. I suspected that this is due to structural breaks (regime shifts in the cointegrating vector). So, I’ve proceeded with Gregory-Hansen cointegration test and indeed the evidence of cointegration increases

My question is: I would like to incorporate the breakpoint dates obtained from the Gregory-Hansen cointegration test when running panel estimations. My intention is to create dummies for each individual cross-section and include them on my panel regressions (both in the constant and the slopes). This implies that in specifications with 6 explanatory variables I would have 14 regressors (constant + dummy + 6 regressors + 6 regressors-dummies interactions). Does anyone see any econometric issue with this approach?

Also, since Pedroni panel cointegration test in Eviews only accepts a maximum of 7 regressors, I was thinking of using Kao and 2-step EG/unit root test on residuals when testing for cointegration in the panel specifications with dummies.

Many thanks in advance.

Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 20 guests