Hello,

I am conducting a fixed effects regression on panel data of 17 countries within the Eurozone, denoted c, using quarterly time periods between 1996 - 2016, so roughly 80 time periods. I plan to use a fixed effects regression and control for country fixed effects as well as time fixed effects. One of my explanatory variables is the exchange rate which is time fixed due to all nations sharing the Euro. When specifying the model to input into Stata, I cannot include this and the quarterly time dummies at the same time as I believe this would lead to collinearity. If this assumption is correct, then do I just omit the exchange rate variable and just include the quarterly time dummies, and perhaps mention that the coefficient on the quarterly time dummies encompasses the effect of the exchange rate? Although exchange rate it a key determinant of my explanatory variable, its effect is not crucial for my study given that it is not my main explanatory variable (my main explanatory variable is country-varying and time-varying). Please see the screenshot file below where explanatory variable 4 is the exchange rate.

On a separate matter, as mentioned before, I have a panel data of 17 countries using quarterly data between 1996 - 2016 and so roughly 80 time periods. I have been told that because the large majority of my variation is time variation, then I should not be using a fixed effects regression and should look more so towards time series techniques. However, I have no familiarity with time series techniques whatsoever, and as a result I was told that if I am to use a fixed effects regression, then I should look towards using yearly data rather than quarterly data so that the large majority of the variation does not pertain to mainly time variation, but rather both cross-sectional and time variation. Given that this is the time period 1996 - 2016 is the only time period that I have found available data for all of my variables, this would mean that I would only have 20 time periods, accompanied by the 17 countries, which drastically reduces my sample size to n = 340. The person has told me to double-check their suggestions as their specialty is not econometrics, and so I am wondering if anyone could kindly shed some light as to whether their advice is correct or not, or if anyone can offer any additional guidance on this at all?

Any answers to the two questions above would be massively appreciated.

Many thanks,
Ley