Hi all,
I am pretty new to Stata and as part of a research project, I want to analyse the impact of ESG ratings of M&A targets on the organisational performance of the post-merger entity (var:ROE), by controlling with variables such as total assets at the moment of the merger and the deal value.
Since my dataset considers a temporality I followed what I thought was right by doing a Hausman test which led me to the following fixed-control effects model:
. xtset deal_id YEAR
. xtreg ROE TargetESG DealValueUSDMillions TotAssetsy0 i.YEAR, fe vce(robust)
Nevertheless, when I enter the command the system omits variables because of collinearity (namely: TargetESG, DealValue & TotAssets). I tried to solve the issue online, but it seems conceptually problematic for me to drop something other than the control variables (Deal value or/and total assets). Even when I do, my dependent variable still gets cancelled by Stata.
To see which variable is creating a collinearity issue I ran a pwcorr command and a VIF and none seems excessively correlated though:
. pwcorr ROE TargetESG DealValueUSDMillions TotAssetsy0
| ROE Target~G DealVa~s TotAss~0
-------------+------------------------------------
ROE | 1.0000
TargetESG | -0.0078 1.0000
DealValueU | 0.0793 -0.0432 1.0000
TotAssetsy0 | -0.0030 0.0092 -0.0007 1.0000
. reg ROE TargetESG i.YEAR, vce(robust)
. estat vif
Variable | VIF 1/VIF
-------------+----------------------
TargetESG | 1.00 0.999967
YEAR |
2 | 1.31 0.760534
3 | 1.31 0.760553
-------------+----------------------
Mean VIF | 1.21
Does it mean I should change the type of statistical analysis (online research has pointed towards ridge or lasso regression)? Or did I miss something when coding my variables/ writing the xtreg command (since I didn't expect return on equity (ROE) and ESG grades to be that correlated)?
Please do not hesitate if you need more information!
And thank you in advance to those who will help on that matter.
Best,
I am pretty new to Stata and as part of a research project, I want to analyse the impact of ESG ratings of M&A targets on the organisational performance of the post-merger entity (var:ROE), by controlling with variables such as total assets at the moment of the merger and the deal value.
Since my dataset considers a temporality I followed what I thought was right by doing a Hausman test which led me to the following fixed-control effects model:
. xtset deal_id YEAR
. xtreg ROE TargetESG DealValueUSDMillions TotAssetsy0 i.YEAR, fe vce(robust)
Nevertheless, when I enter the command the system omits variables because of collinearity (namely: TargetESG, DealValue & TotAssets). I tried to solve the issue online, but it seems conceptually problematic for me to drop something other than the control variables (Deal value or/and total assets). Even when I do, my dependent variable still gets cancelled by Stata.
To see which variable is creating a collinearity issue I ran a pwcorr command and a VIF and none seems excessively correlated though:
. pwcorr ROE TargetESG DealValueUSDMillions TotAssetsy0
| ROE Target~G DealVa~s TotAss~0
-------------+------------------------------------
ROE | 1.0000
TargetESG | -0.0078 1.0000
DealValueU | 0.0793 -0.0432 1.0000
TotAssetsy0 | -0.0030 0.0092 -0.0007 1.0000
. reg ROE TargetESG i.YEAR, vce(robust)
. estat vif
Variable | VIF 1/VIF
-------------+----------------------
TargetESG | 1.00 0.999967
YEAR |
2 | 1.31 0.760534
3 | 1.31 0.760553
-------------+----------------------
Mean VIF | 1.21
Does it mean I should change the type of statistical analysis (online research has pointed towards ridge or lasso regression)? Or did I miss something when coding my variables/ writing the xtreg command (since I didn't expect return on equity (ROE) and ESG grades to be that correlated)?
Please do not hesitate if you need more information!
And thank you in advance to those who will help on that matter.
Best,
Comment