I want to create one model for this relationship. Is this model is acceptable and FIT? or any suggestions?

RealGDPt = α0 + α1POPt + α2LABFt + α3LIFt + α2INFMORt + εt

real gdp= gdp / cpi * 100
pop = population
labf = labor force participation rate
lif = life expectancy
infmor = infant mortality rate

I give you better one
RealGDPt = α0 + aN RealGDP(t-n)

Nothing predicts the future than the past. Just brush up on time series diagnostics.

Forgot to add, you can try ARIMA models also.

This post has been edited by berzerk: Mar 24 2011, 05:42 PM

i need to put at least 3 independent variables.

use more lag variables GDPt-1, GDPt-2, GDPt-3 etc etc etc.

I'm curious why you chose your original variables of pop, life exp, infant mortality, etc.

i need 3 independent variables. coz i wanna test the relationship between population and economic growth. not GDP.

In the first post you have Real GDP as the dependent variable. Now you say you want to test the relationship between economic growth & population.

Which is which?

Just add the usual growth model factors like capital, labour and savings if you need more variables.

hi. now im doing my thesis title population and economic growth. i wanna asking what is the best recommended model specification for my research title.

my propose model is : Real GDP = a + LaborParticipationRate(b1) + LifeExpectancy(b2) + InfantMortalityRate(b3) + e

when I regress the data, it seems that my data have bad outcome such as autocorrelation, heteroskedasticity.

Is there any propose model for my research title?

To see how well it fits, you have to refer to the R ^2 value after you run your regression.

You can correct for heteroskedasticity using White's correction for an example.

Look at your data values. Have you made them into logs? Decide whether you need to make them into logs or not.

Your model is too general and insufficient. You would need more independent variables (X's). Refer to more quality journals and see what they have used. I'm sure they've used more. You can select more variables from there and craft your own model.

How much data do you have? Usually for time series, you need data from at least 1900 - 2010 and for every month as you need thousands of data for your time series model to be accurate. You can use yearly data, but the model would not be accurate. Try another alternative.

However, if you do not have that much data, you can use panel data.

This post has been edited by copenhagen1: Jan 16 2012, 08:00 PM

I agree I think you need to run the model first and move on from there to see if it fits. Are you doing your thesis on Malaysia? Because if you are it's not going to be very easy to get the sufficient amount of data for those variables. Unless if you wanna do quarterly or monthly as it's hard to find (doubt it exist) old data on Malaysia what with us being relatively newly independent and so on..