# Wage regressions: Nominal versus real wage

I'm wondering do I need to convert nominal wages to real wages when running wage equations using panel data sets? My dependent variable is log of wage and I'm looking at returns to education across countries over time. Does the decision to use nominal versus real wage depend on what model I'm using- namely random, fixed effects or pooled OLS?

Because you want to measure the real impact of education.

If there is inflation, the nominal wages most likely will go up over time, but you don't want to think that your return to education does that, it's just inflation.

If, on the other hand, you're comparing wage differentials, say $\log w_s - \log w_u$, where $s$ indicates skill and $u$ indicates unskilled, you do not need to correct for the price level.

To see, note that

$$\log w^s_t - \log w^u_t = \log \frac{w^s_t}{w^u_t} \\ \frac{w^s_t}{w^u_t} = \frac{\frac{w^s_t}{p_t}}{\frac{w^s_t}{p_t}}$$

• If you are using wage differentials you will still have inflation. Or am I missing something? – Rud Faden Mar 24 '15 at 22:22
• @RudFaden I was referring to log difference, I emphasized that now. – FooBar Mar 24 '15 at 22:40