Hi All,
I am dealing with company level unbalanced panel data (attached). There are broadly two types of control variables in the study - a) firm level control variables, FIV1 (market cap, return on assets, cash holdings, carbon risk etc.) which varies for a firm cross-sectionally as well as in time and (b) macro or market variables, MIV1 (risk free rate, excess market return etc.) which do not vary across firm level but vary only across the time.
When I run the regression, I get an estimated coefficient of ‘0’ for macro-economic variables if I take the time and industry fixed effect. Stata provides the warning message that such variables are perfectly correlated with the fixed effects, hence cannot be estimated and provides ‘0’ coefficients for such variables. This serial correlation is understood because the values of these control variables will remain same for the time effect dummy across all the firms. Hence, when I remove the time fixed effects, then as expected this issue does not occur and I get the estimated coefficients.
Similar macro-economic variables are used in the other studies where the results always mention that the time fixed effects and cross-sectional fixed effects are being considered. I am not able to understand why these studies did not face the same issue that I am facing? I need to have time fixed effect in my analysis but the collinearity issue is giving '0' value for the market level coefficients. Please guide me.
Code:
reghdfe DV FIV1 MIV1, absorb(IndustryId Date) vce(cluster CompanyId Date)
Warning from Stata - MIV1 is probably collinear with the fixed effects
MIV1 omitted because of collinearity
Related Posts with Coefficient omitted due to time fixed effect and multi-collinearity
test Code: . xtnbreg collab_weight lie_dummy lmie_dummy umie_dummy hie_dummy crudeoil distance exp_good…
Measuring women's empowerment: creating an index for psychological well being, social support and decision-making powerDear Statalisters, For my thesis I am working on a dataset in which empowerment is measured from th…
Problem in duplicate case using HIES 2018-19My data has duplicate cases, I want to remove duplicate case if they report '2' as response in varia…
longitudinal analysis with measurements taken at different timesdear stata users, I'm dealing with a longitudinal dataset with measurements taken in different time…
Bootstrap Maximum Likelihood Estimation_ time-seriesDear Statalist users, I would like to conduct a bootstrap maximum likelihood to estimate a linear t…
Subscribe to:
Post Comments (Atom)
0 Response to Coefficient omitted due to time fixed effect and multi-collinearity
Post a Comment