Good morning. And good luck...

This field of research is full of traps. Measurement issues, theoretical mis-specification and statistical traps.

For example, the widely used Solow-type models give very good econometric fit but are statistically and theoretically flawed. Indeed, aggregate production function (e.g, Cobb and Douglas) fail the Cambridge Capital Theory Controversies Test. More importantly, the estimated parameters of the production function are merely capturing an underlying accounting identity. Hence, no reliance can be placed on the good fit you usually get, and even less on the residual (the total factor productivity).

This said, most people still use it, so why not?

Endogenous growth provides a more multivariate approach, with human capital added to physical one, with the quality of soft and hard public infrastructure (governance, macroeconomic stability, education, transport and communication, etc...). Here, the issues are also of measurement (we cannot observe the depth of trust between economic agents nor their animal spirits) plus econometric issues (endogeneity, basically).

You must also include trade, not only for its short term impact on GDP, but also for its long term influence of total factor productivity (as long as you can measure it).

Speaking about trade, if you work from an international perspective (several countries), you may look into GDP/head convergence and productivity catching-up. Absolute, conditional, through convergence clubs... Huge literature, again...

I do not know of any good solution to the econometrics. The method of moments was “the” solution a couple of decades ago, but has its limits. Instrumental variables are usually not convincing...

This said, it provides for a good thesis subject: lots of different approaches, and lots of issues to be described and explained.

So, you are on the right path.

Dear Kamran,

Most model use growth in GDP/GNP as the measure. By national accounting, GDP is the sum of consumption (C), Investment (I), Government spending (G) and net export (NX). Serious regression models explain each of those components and their subcomponents. The end result is a system of equation regression model. There are many of these systems around. For instance, you have the Wharton Economic Model, the Federal Reserve Bank model etc. I suggest that you look into those and similar ones to see what variables they use. If you want to use net FDI flows, which you are familiar with, then that could be one of your investment equation.

More modern growth models look at human development factors--education, health, etc., as driver of growth. A good book on growth theory usually covers this aspect.

Good luck in your research

Thank you Mr. Ramrattan for your helpful answer. I will certainly look into the models you have included and reflect over it on my work. Much obliged.

Dear Kamran Io

In addition to expenditure approach as explained by Lall B. Ramrattan , we can also use production function, i.e. economic growth (measured by GDP growth) is determined by capital (K), labor (L), and other factors, such as quality of institutions (A). Your FDI inflow data can measure capital accumulation.

If you want to add neighboring countries data (eg. Afghanistan and India) it will be a good strategy to increase number of observations, since I guess that Pakistan data is relatively limited to use time series regression. Accordingly you can use panel data regression.

If you were using many countries (cross-section data set), then the baseline would be the Solow growth model:

y = a0 + a1 * s + a2 * (n + d) + v, where y is per capita GDP, s is per capita investment, n is population growth and d the depreciation rate.

However, it seems your using only one country therefore you have time series data. Therefore, as people again just expressed, a production function/Solow residual approach is good, that is:

DY = b0 + b1 * DK + b2 * DL * b3 * t + u, where Y is GDP, L is Labour, K is Capital Stock, t is a linear trend and u the regression's residual, which also measures technological level.

Since all variables are expressed on logs, then DY is GDP growth and you may add foreign investment (your factor of interest) in the equation.

other than Labour and its quality,Capital stock and other factors suggested by Pandelis Mitsis you could also add the role of bright /unique ideas and role of governance and it's quality in promotion of growth.

Thank you Chandras Pant. I will look into incorporating the governance variable into my regression. Are there any websites you would advise that I can obtain the data from?

Hi, Since you are preparing a undergraduate thesis you could use Multiple linear regression model (MLR) or maybe to consider ARMA models, but these models (ARMA) are on more advanced level.

Good morning. And good luck...

This field of research is full of traps. Measurement issues, theoretical mis-specification and statistical traps.

For example, the widely used Solow-type models give very good econometric fit but are statistically and theoretically flawed. Indeed, aggregate production function (e.g, Cobb and Douglas) fail the Cambridge Capital Theory Controversies Test. More importantly, the estimated parameters of the production function are merely capturing an underlying accounting identity. Hence, no reliance can be placed on the good fit you usually get, and even less on the residual (the total factor productivity).

This said, most people still use it, so why not?

Endogenous growth provides a more multivariate approach, with human capital added to physical one, with the quality of soft and hard public infrastructure (governance, macroeconomic stability, education, transport and communication, etc...). Here, the issues are also of measurement (we cannot observe the depth of trust between economic agents nor their animal spirits) plus econometric issues (endogeneity, basically).

You must also include trade, not only for its short term impact on GDP, but also for its long term influence of total factor productivity (as long as you can measure it).

Speaking about trade, if you work from an international perspective (several countries), you may look into GDP/head convergence and productivity catching-up. Absolute, conditional, through convergence clubs... Huge literature, again...

I do not know of any good solution to the econometrics. The method of moments was “the” solution a couple of decades ago, but has its limits. Instrumental variables are usually not convincing...

This said, it provides for a good thesis subject: lots of different approaches, and lots of issues to be described and explained.

So, you are on the right path.