Hello,
Can I regress company data (as firm performance - the dependent variable) on macroeconomic data (as macroeconomic uncertainty - the independent variable)? Is it econometrically right? Because all firms are subject to the same macroeconomic uncertainty at a given point in time (cross-sectionally invariant), which effect should I adopt for panel data? I'm not sure if fixed effects fits.
Thanks.
You will able to estimate it without problems, unless you have time fixed effects. Remember you are estimating the effect of regressors that do not have variation across individuals but do vary across time. In that case, the time demeaning is possible and you will be able to estimate the model. In other words, it is econometrically ok. However, the best procedure would be to find some regressors that vary not only across time but also across individuals.
ECONOMY >>>> INDUSTRY >>>> SECTOR >>>> Firm
A B C D
There is a missing links of industry and sector in your proposed model. Although you can regress economy and firm, the measure of the influence or effect that you will see might be small because there are two other intervening factors: industry and sector.
FIRM >>>>> SECTOR >>>>> INDUSTRY >>>>> ECONOMY
D C B A
You may want to do multi stages by running regression economy and industry. Then industry to sector where the firm is located, Finally sector to firm level. For the 3 elements (economy, industry, and sector), secondary data may be available. For firm, you may want to define what indicator to use. If using publicly traded companies, then all financial statements may be available through the exchanges website or company's website (annual report). In fact, firm sector index, and industry index may also be available through stock exchanges.
The choice of methodology depends on several criteria. Please, refer the attachment that clearly shows the which model you have to adopt for panel data based on those criteria. Good Luck.
You will able to estimate it without problems, unless you have time fixed effects. Remember you are estimating the effect of regressors that do not have variation across individuals but do vary across time. In that case, the time demeaning is possible and you will be able to estimate the model. In other words, it is econometrically ok. However, the best procedure would be to find some regressors that vary not only across time but also across individuals.
From the perspective of random effects modelling you have a three level model 1: occasion, 2: firm 3: country and in this formulation you could have time varying and time constant variables
Article Explaining Fixed Effects: Random Effects Modeling of Time-Se...
Why not. It all boils down to what the researcher is looking at
Sure... why not, if there is something has economic meaning just go for it. But first of all, I suggest, you had better derive a strong conceptual framework then go to model such panel ..etc. I think that would be better. Good luck.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression