Dear all,
I want to know if it is econometric-ally correct to use a predicted value as an EXPLANATORY variable in another regression in order to test a hypothesis within a time series and panel data analyses. For the time series, the estimation technique is ARDL while I am using both pooled OLS and dynamic fixed effects estimators for the panel data.
My research hypothesis is that trade (an outcome of favourable exchange rate) is a predictor of income inequality and in order to test this, I need to first analyse the determinants of trade openness after which I test its impact on income inequality in order to establish the nexus.
Using the Gini index as the measure of income inequality, and trade openness as the value of imports and exports over GDP, I formulate 2 equations:
First equation: trade = intercept + exchange rate + per capita income + u
Second equation: gini = tradehat + per capita income + education + u
So, is my 2nd equation appropriate, with:
1) tradehat as an explanatory variable?
2) the inclusion of "per capita income" as an explanatory variable in the second equation having used it in the first equation?
What questions/criticisms are likely to emanate from reviewers as regards using this approach? Suggestions and comments will be immensely appreciated.
Thank you.
Dear Ngozi,
The answer to your first question of using trade hat (the predicted values) as predictor is that: it is better to use the observed actual data instead of using the predicted values due to the fact that the predicted values are not the actual data, there would be some error (difference between actual and predicted). My question regarding this is that: what is the reason to use the predicted values of trade instead of its actual values?
Well, as far as your second question is concerned, I strongly suggest to consult the theory or existing literature for the selection of variables. If to examine the effect of per capita gdp on trade is your objective, it is for sure in the model. However, if you are taking it as a control variable, (obviously it affects trade volume in a country), but still control variables must come from the theory or from the literature.
Feel free to ask again if I am not able to answer your questions.
Best Wishes,
It seems to be ok. However, you should explicitly write down time indices of your variables. I assume you have panel data. Am I right?
Dear Ngozi,
The answer to your first question of using trade hat (the predicted values) as predictor is that: it is better to use the observed actual data instead of using the predicted values due to the fact that the predicted values are not the actual data, there would be some error (difference between actual and predicted). My question regarding this is that: what is the reason to use the predicted values of trade instead of its actual values?
Well, as far as your second question is concerned, I strongly suggest to consult the theory or existing literature for the selection of variables. If to examine the effect of per capita gdp on trade is your objective, it is for sure in the model. However, if you are taking it as a control variable, (obviously it affects trade volume in a country), but still control variables must come from the theory or from the literature.
Feel free to ask again if I am not able to answer your questions.
Best Wishes,
Dear
I have found some papers may be similar to your topic please have a look.
https://www.imf.org/external/np/seminars/eng/2014/lic/pdf/Lim.pdf
https://link.springer.com/chapter/10.1007/978-3-319-27573-4_32?no-access=true
http://ohlininstitutet.se/wp-content/uploads/2008/09/amandajakobsson.pdf
http://scientificpapers.org/download/286/
Yes Peter I have a panel dataset of 20 countries from 2000 to 2015...and what do you imply by time indices?
Thanks Muhammad Atif Nawaz for your contribution.
My reason for using the predicted value of trade is to prove the trade-inequality nexus within the confines of the model. I am of the opinion that if I use the level of trade as an explanatory variable in the inequality equation, there will be a disconnection in this nexus. This is because having explained what determines trade in the 1st equation, it is imperative that the outcome variable (now tradehat) should explain income inequality.
This is my hypothesis....I feel the way I specified the two equations seems plausible but I am open to criticisms and comments on how to improve this study. Thanks again, I sincerely appreciate your taking the time to respond. I have learnt something of value from your response.
Thank you Muhammad Saqib Irshad for these suggested papers. I will go through them and revert. Thanks a lot!!!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression