# Panel data with country fixed effects

I am wondering about the estimation of a fixed effects model. It is just given in the paper that estimation is done via OLS with robust standard errors. Which method is meant by such explanation? Did they use pooled OLS or fixed effects estimator (using also OLS to estimate after building the difference) or first-difference estimator (also using OLS after having built the difference)? Are they using robust standard errors to take care of serial correlation?

Thank you for your answer. I should maybe clarify my question: I have got an fixed effects model dealing with it from an economics perspective. My main question is regarding the estimation used to find the results. It is said that they are obtained using OLS. However, I think that this does not exactly clarify if they used Pooled OLS, Fixed Effect estimator or first difference estimator. How can I identify which of them was used? It wuold be great if you stated how you identified the estimation method used.

The given model is:

$y_{i,t}=\alpha_0+\alpha_i+\beta_1X_{i,t}+\gamma_1S_{j,t}+(\delta_0+\delta_i+\beta_2X_{i,t}+\gamma_2S_{j,t})D_t+\epsilon_{i,t}$

One observes the change in the parameters after a special event.. therefore the dummy. The whole sample is for 10 years. The paper I am referring to is the following:

• Can you give a complete citation for the paper in question? – gung - Reinstate Monica Apr 11 '16 at 0:43
• It's not clear what you are asking here. I see several different questions. It's usually better to break these into separate questions. One question I see is, "What is meant by robust standard errors? Another (for a separate post) might be, "What is a fixed effects model? Yet another might be, "please help me understand how the authors of this paper carried out their analysis." Note too that there are at least two different interpretations of the "fixed effects" models. The interpretation from an economist is usually different than that of a statistician. – StatsStudent Apr 11 '16 at 0:44
• That citation would be great. There is little we can do without it, I'm afraid. – StatsStudent Apr 11 '16 at 2:19
• scholar.google.de/… This is the paper I am refering to – Meier Apr 11 '16 at 2:25

## 1 Answer

Given that a dummy $\alpha_i$ for each country is included (or rather the deviation of each country from a common mean, which would be the way that Stata includes a constant term $\alpha_0$), this is a fixed effects model. You can estimate the fixed effects model either by subtracting the country specific mean of each variable from itself (this is called the within transformation) and use the demeaned variables in an OLS regression - or, as they do here, you can run OLS with a dummy for each country. The latter is referred to as the least squares dummy variables model.

They also write:

"Based on this conceptual framework, we examine the determinants of the pricing of sovereign risk both in non-crisis and crisis states for a range of advanced and emerging economies, using a standard panel model with country fixed effects" (p. 64)

• Thank you very much for your answer! Can you tell be how you obtained that they used LSDV? For me it is not applicable where they stated to use this method. Is this meant by "standard panel model with country fixed effects"? – Meier Apr 12 '16 at 9:34
• Yes, country fixed effects means that there is a dummy for each country (except for one). So the country specific fixed effect is modeled as a country specific intercept which does not vary over time. – Andy Apr 12 '16 at 19:58