Dear all,
I am running panel data analysis and facing some difficulties. My goal is to find out drivers of firms' decision on leverage. I am using Eviews software for my analysis.
My data has 9 years timespan with yearly frequency and consists of 75 companies, represented by 23 industries.
My dependent variables are: long-term debt, short-term debt .
Independent variables: profitability(x1), size(x2), tangibility(x3), liquidity(x4), non-debt tax shield(x5) median industry leverage(x6). All my variables are appropriately transformed.
I estimated Y C X1 X2 X3 X4 X5 X6 random effect equation (as suggested by Hausman test) and got satisfying result. Next, I decided to include dummy variables (k-1) for each industry to control individual intercept and find out effect of each industry.
The problem is that I can no more
estimate an equation as Eviews gives me 'Near singular matrix'. I dropped variable X6 and managed to estimate pooled regression. However, random effect regression can no longer be estimated as 'there are not enough cross-sections'. From now on Eviews allows just for fixed effect regression.
After introduction of dummy variables, Eviews does not let me to conduct Heteroscedasticity and Hausman tests.
Thank you for dedicating your time to read this question. Any help is appreciated.
You cannot include fixed effects and firm level variables as you have used up all the information at this level (that is the time invariant information). use random effects and drop the fixed effects; see these three papers:
Article Understanding and misunderstanding Group mean centering: a c...
Article Fixed and Random effects models: making an informed choice
Article Explaining Fixed Effects: Random Effects Modeling of Time-Se...
You cannot include fixed effects and firm level variables as you have used up all the information at this level (that is the time invariant information). use random effects and drop the fixed effects; see these three papers:
Article Understanding and misunderstanding Group mean centering: a c...
Article Fixed and Random effects models: making an informed choice
Article Explaining Fixed Effects: Random Effects Modeling of Time-Se...
Dear Kelvyn Jones ,
Thank you for the reply. My firm level information is from 2010 to 2018, does this still invariant? As I mentioned, after introduction of dummy variables for each industry k-1 (NOT COMPANY), the software no longer lets me estimate random effect equation.
You may use Stata as it is better in panel analysis than E-Views. You may also read this attachment it could be helpful.
Dr Ajit kumar Roy ,
Does this mean I should drop industry dummies, and proceed with random effect equation?
Best regards
Hausman test is not so important. Whether you will take a ramdom effects or fixed effect it completely depands on your data. If you have some omitted variables you should you fixed effect. If you have the data on all the variables you should use random effect model.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression