2
$\begingroup$

So the central question of my project is to what extent does a country's level of export contibution towards GDP (i.e. exports as a % of total GDP) affect its GDP growth.

I'm comparing this hypothesis for two different countries, China and India.

I have annual time-series data for exports as a % of total GDP for both countries and their GDP growth for the last 30 years, but I have no idea how to choose a correct model to begin to anaylse it with.

I was thinking of pehaps using Autoregressive model one, any thoughts? If this is correct then I how would I begin to model it using my data?

| improve this question | |
$\endgroup$
  • 1
    $\begingroup$ I wonder if this is really more of a substantive / theoretical question for developmental economists than a statistical question. Are you wondering if growth in the portion of GDP due to exports correlates with growth in the portion not due to exports, or if it correlates at some lag, or if it Granger causes, etc? After all, exports won't be constant & unless the rest of the economy correspondingly shrinks, GDP will grow. $\endgroup$ – gung Apr 30 '15 at 16:26
  • $\begingroup$ @gung is right: Although we would like to help, there is not sufficient information given to do so. If you could articulate, in a quantitative way, an economic theory concerning the relationships among your variables, then we could help you formulate it in a statistical fashion, but the stats site is not the place to be asking what economic models reasonably apply. $\endgroup$ – whuber Apr 30 '15 at 16:44
  • $\begingroup$ Thank you all for your quick responses, I agree with the fact that my title itself is flawed. As I don't want to deviate too much from the topic and must perform some kind of econometric analysis, I was wondering as to the viability of changing the title to simplify it slightly. I could do this by simply looking at the total value of exports, and how they correlate with the total GDP of a country. I already have data for the last 30 years for both variables, and a preliminary graphical analysis does suggest a correlation. so surely I could model this new permutation? $\endgroup$ – Babar Sattar May 1 '15 at 17:41
2
$\begingroup$

Plot things. Do the raw data look like they have cycles? (I assume so, given that you are talking about economic data and want to use an autoregressive model.) Next, try to get the order of the autoregressive model by taking differences until the cycles disappear. This can also be tested formally using (a version of) the Durbin-Watson test. There will likely be further patterns. What do you see? Does the variability stay constant over time (likely when you have percentages) or is there a tendency for the variability to grow? What smooth (and simple) function follows the general distribution of the data? Fit it, assess the fit, repeat with different functions, until you have something reasonable.

Model selection is far more of an art than a science, and there often are several reasonable models that you can use.

| improve this answer | |
$\endgroup$
1
$\begingroup$

I think you can use VAR models in these kind of applied policy issues. I think you are interested by some questions like "how about the effect of exports on GDP and GDP on export ?" So, if it is the case, I think the most appropriate way is to use VAR models in which you can analyze the causality (like Granger Causality) between your key varibles. But you must look at your data if your time series are cointegrated, if yes, you can use a VECM model (Vector Error Correction Model) which can be considered as a different version of VAR models.

| improve this answer | |
$\endgroup$
  • $\begingroup$ I'm debating now to change my title simply to the effect of exports on GDP. My data is still a time series but one country displays cointegration between the two variables whereas the other does not. In your opinion is VAR still the best choice? $\endgroup$ – Babar Sattar May 1 '15 at 17:52
  • $\begingroup$ How many variables do you have in your model ? If there are at least some cointegrated variables in your models, you can use VAR models. $\endgroup$ – optimal control May 1 '15 at 21:16
  • $\begingroup$ Just two. Dependant is GDP, Independent is Exports, and an error term. If could add the other determinants of GDP such as consumption, Investment and Gov Spending. $\endgroup$ – Babar Sattar May 2 '15 at 14:40
  • $\begingroup$ If you have GDP and Exports which are cointegrated, you can use VAR approach with VECM model. Because you can not use another approach if you have at least 2 cointegrated series in your data. Your estimations will be biaised. So, the best way is to use a VAR modelling. $\endgroup$ – optimal control May 2 '15 at 22:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.