EViews.com

Hello,

I've noticed that there are a lot of unanswered threads in this forum with people asking for various help. I'm sorry to add to this list but quite frankly I'm quite desperate. I'm in my final year and currently working on my dissertation project which is due at the end of this month. I'd just like a confirmation from someone on the methodology I'm taking and whether or not this is correct.

I'm studying the effect of the exchange rate on the trade balance between Brazil and Argentina with the United States. I aim to estimate the long-run relationship between 4 variables [Trade Balance, Home GDP, Foreign GDP (USA), and the Real Exchange Rate]. All data is quarterly and expressed in log-form. The trade balance value is the ratio of exports to imports and real exchange rate is the bilateral rate between each country and the US, adjusted for prices using the relative CPI index.

Here's the process (which I think is correct):

1. Unit root test for stationarity - Augmented Dickey Fuller
- Look at graph plots of the series and determine whether to use ADF equation with trend and constant or just constant
- Carry out test on series in levels and then first differences
- If all are stationary in first differences (i.e. integrated of order 1) I can proceed to the Johnasen Cointegration test

2. Determine number of lags for VECM model:
- Estimate a VAR model with a large number of lags (Is this arbitrary? - I've been choosing
- Use the lag order selection criteria in EViews and compare optimal number of lags selected by each criterion

3. Carry out Johansen test for cointegration
- I've read this from another forum post:

As noted in the Eviews6 User Guide II page 368, "the lags are specified as lags of the first differenced terms used in the auxiliary regression, not in terms of the levels." Therefore, when you determine the lag length for Johansen cointegration test, you should run the VAR with first differenced variables. Alternatively, you may run the VAR with the variables at levels, then subtract the suggested lag length by one (1), for example, if the level VAR suggested that 3 lags is the best, then you should use lags 2 for your Johansen's cointegration test. Caution: You must use the variables at levels when running the Johansen cointegration test.

So does this mean that I need to remove a lag from the VAR lag order taken from the previous stage? (Currently it suggests I use only 1 lag, using no lags for a cointegration test seems nonsensical).

- The results of the test should reveal the number of cointegrating vectors amongst the variables.

4. VECM
- Calculate a VECM with the number of lags as suggest in step 2, normalising my results on the TB and noting the number of cointegrating vectors as found in step 3.
- Interpret results

5. Impulse response
- 1 st. dev. shock on variables and observe the graphs


As you may be able to tell my understanding tends to diminish as I progress through the process. They haven't actually covered VECM as part of my course, nor have we really looked time-series analysis in great detail. I chose it because I genuinely found the topic interesting but i'm now feeling as if I've bitten off more than I can chew.

I would really REALLY appreciate if anyone could offer their opinion or share their insight on this as I'm becoming really stressed thinking of the prospect of going screwing up my dissertation.

Anyway, I really appreciate if you've read to the end of this post and will be even mroe grateful if you can offer any sort of help. After all, I'm sure there are plenty of other more rewarding things you could be spending your time on rather than helping an ignorant student. Nonetheless, thank you.