Interpretation of coefficients in a log-level model

[Emil Bruhn]

Dear Statalist

I have a problem when interpreting the coefficients of my model. My dependent variable is LOGGDP and my independent variable is the share of the population who sees themself as a religious person in % (0-100). My coefficient is 0.0175. I have the two following interpretations in mind:

1. When the share of the population who sees themself as a religious person increases with 1 pct. point, LOGGDP is increasing by 0.0175 pct.

2. When the share of the population who sees themself as a religious person increases with 1 pct. point, LOGGDP is increasing by 1.75 pct.

Or maybe a third option?

Thank you in advance!

[Phil Bromiley]

The easiest way to work with this is to use margins and the exponent of the predicted value.

But the question you ask is not really about the log issue. It depends on whether you've coded the percent as a decimal (0 to 1) or as a percent (1 to 100). If it is 1 to 100, then a coefficient of .0175 means a one percent increase (i.e., a one unit in the rhs variable) results in a .0175 increase in predicted LogGDP. I have trouble interpreting a change in a log, so I'd normally translate it back to GDP.