Before regression (like fixed-effects, random-effects), for these two kinds of variables:

1) Price levels (e.g. price level of capital formation) They are usually ratios like 0.987, 1.022..., based on USA 2010=1

2) % of GDP (e.g. broad money as % of GDP) They are also ratios (%) like 80.5, 99.7 (or in decimal form 0.805, 0.997)

Do I need to log transform them, and how can I log transform them?


You don't need to log transform these variables. Consider the following model

$\log y = \beta x.$

Then, you can interpret $\beta$ as a semi-elasticity, i.e., the percentage change in $y$ when $x$ increases by 1 unit.

To see this you can transform the previous equation as: $y = \exp(\beta x),$ in which case $dy/dx = \beta \exp(\beta x)$. Thus, the percentage change in $y$ when $x$ changes is $\frac{dy/dx}{y} = \frac{\beta\exp(\beta x)}{\exp(\beta x)} = \beta$.

Note also that

$\Delta \log y = \beta \Delta x$ so $100.\Delta \log y = (100.\beta) \Delta x$

It follows that

$\% \Delta y \approx (100.\beta) \Delta x$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.