Hi Kostas,

Generally in databases growth refers to the change from one period to another which is also what you do when taking the first difference. Logarithmising a variable does not change it into growth.

Based on your description of coefficients, the GDPpcgr is in absolute values, so if you take the ln of this, it should match what you get from d1_lnGDPpc. If this is not the case, have a look at the data source and see what the unit(s) of the variables are.

Whether or not you should go with the change depends on the problem you are studying and somewhat on the econometric method you are using. If you have the change of income inequality as an independent, it generally makes sense to study the change in economic growth as well.

Hopefully this helps.

The GDP is growing exponentially. So the difference between the log GDP is a clearer proxy for growth. Moreover, it's a better approximation that GDP changes multiplicatively than that it changes additively, so analysis on logarithmic scale is helpful. Hence, I think you can continue your regression analysis using d1_lnGDPpc.

As stated by the answers above, the log of per-capita GDP (LGDP) is not the growth rate of GDP. This period's LGDP minus last period's LGDP (that is, LGDP(t) - LGDP(t-1)) is an approximation to the growth rate of GDP that is valid for small growth rates. To see the degree of approximation for different growth rates you can compare the difference of LGDP approximation with the more usual formula for the growth rate, being [GDP(t) - GDP(t-1)] / GDP(t-1)

You can find this video useful for understanding the issue:

https://www.youtube.com/watch?v=MdALpwQAV_o

You need to understand the level-level, log-level, level-log and log-log regression result interpretation regarding this an important fact regarding your question is how is your income inequality is expressed, as percentage or absolute number?

The natural log just formats your data in a different way it does not transform absolute values into averages as you implied.

When you take the ln of number 5, the resulting ln(5) is the exponent of the base number, if natural log (almost always is) then the base number is the transcendental number (look it up), that makes the base trans. number turn into five. Ln(5), (trans number=2.718281828459)1.609438 = 5

Ln(5) = 1.609438

If a variable is already a percentage you must not log transform it, because the base number can not grow to a percentage which is a relative number, you just leave it as a percantage and use the log transform on the absolute number then interpret both numbers as a log-log model.

Dear Kostas,

GDP per capital growth is not the same as log of GDP per capita.

They are different.

I recommend @Chris Stewart.

Regards