Log initial gdp per capita

[Tugba Celik]

Hi everyone, i try to make a growth analysis with five-year panel data; i uploaded the gdp per capita data like: 1960; 1965; ...;2015.
i took five year average growth;; for example, 1960 in my data= average of 1960-1964 years growth and so on..
i want to ask what is the meaning of "log initial gdp per capita" or gdp per capita in base year?
i try to find convergence coefficient , but i couldnt understand whether if i should take lag(1) of my gdp per capita data? or Lag(5)?? (due to time interval is 5 like 1960,1965..)
or as initial value, should i take value of 1960 gdp per capita for all years (namely 1960 gdp per capita= 1965=1970=...=2015?) What would you do if you were? please help..

[Phil Bromiley]

You'll generally get more helpful answers by following the FAQ on asking questions including providing Stata code in code delimiters, readable Stata output, and sample data using dataex.

You question is very hard to understand. With the info requested above, we'd have a better idea what you are asking. Also, assume we are not in your area so instead of "growth analysis" tell us specifically what model you're trying to estimate.

If this is what they do in your area, then you probably should do it, but I"m not sure about averaging growth over multiple years. This creates statistical problems. Your annual observations are clearly not independent by construction. Actually, you've created by construction extremely high associations between observations. You also create the ambiguous issue you note at the end - what is the meaningful lag when you have this kind of dv? My preference is generally to do annual models with annual data.

[Tugba Celik]

Dear Bromiley, thank you very much for your reply.
I try to analyze economic growth and demography with five-year panel data. It is like a cross section convergence analysis.
I use 70 countrys' data and 1960-2015 time period for anaylsis. My dependent variable is gdp per capita growth; my independents are initial gdp level, initial working age population level, working age pop increase, population growth and demographic other variables; also trade opennes, education, investment etc and some regional and time dummies.. In literature, this analysis is made by five or ten year panel data analysis to investigate the impact of demographics on economy (it is impossible to see this effect in short time). As i understand, this is partial adjustment process and i should use lag (1) of gdp per capita for "initial". What's your opinion?

[Carlo Lazzaro]

Tugba:
I'm not sure whether I got your query right.
Anyway, if you paln to include a lagged dependent variable among your predictors, you will be better off with a dynamic panel data model, such as the ones performed via -xtabond-.

[Tugba Celik]

Dear Lazzaro, thank you for yoru reply. I want to ask you whether if i take lag(1) of gdp "level" when dependent is growth, is it dynamic model?namely i dont put lag of growth but i put lag of level, then is it dynamic?

[Carlo Lazzaro]

Tugba:
thanks for providing further clarifications.
I have to correct myself, though: if you lag an independent variable the panel is not dynamic.
However, I still find difficult to follow your regression model.
I do share Phil's concern about woorking with 5-year aggregated data rather than with annual data.
Maybe the literature in your research field supports your approach, though.