Hi everyone,
I'm doing a research on the relationship between equity contribution in a joint venture and parent firm performance (measured by CAR, continuous variable). Also testing for the moderation effects of the level of industry relatedness between 2 partners of a JV, and also previous JV experience (measured by the number of previous formed joint ventures). You can find the data in the file attached
For the equity contribution, I have 3 categories: majority (>50%), equal (50%) and minority (<50%), coded as dummy variables. Since i want to test if the choice of each type could have a negative or positive relationship with the CAR, i choose the regression model. OLS regression model is often used to test these kind of relationships but it can only compare two out of three choices to the reference group given my independent variables.
Running the correlation matrix, there's no significant relationship between these above choices with CAR.
Initially I run the OLS regression model for each category: reg CAR majority_owned c.ROA c.Leverage c.Culture_Dist c.International c.market_size, vce(robust)
however, the p-value of the model is super high -> showing that the model is statistically insignificant.
Afterwards, I decided to treat the data as panel data, as of the observations were made for specific firms, and the control variables are firm-specific effects. but the time effects are omitted, since there are multiple observations happens at different dates in a year, the only identifier is the key-id of each company
xtreg CAR majority_owned c.ROA c.Leverage c.Culture_Dist c.International c.market_size, re vce(robust)
xtreg CAR equally_owned c.ROA c.Leverage c.Culture_Dist c.International c.market_size, re vce(robust)
The statistical power of the model is better. However, some of the coefficients are statistically insignificant.
Therefore, i want to ask whether my choice of model is reasonable given the relationship i want to test. Also, what could be the interpretation for the insignificance or tests i could do further? Thank you very much for your help
I'm doing a research on the relationship between equity contribution in a joint venture and parent firm performance (measured by CAR, continuous variable). Also testing for the moderation effects of the level of industry relatedness between 2 partners of a JV, and also previous JV experience (measured by the number of previous formed joint ventures). You can find the data in the file attached
For the equity contribution, I have 3 categories: majority (>50%), equal (50%) and minority (<50%), coded as dummy variables. Since i want to test if the choice of each type could have a negative or positive relationship with the CAR, i choose the regression model. OLS regression model is often used to test these kind of relationships but it can only compare two out of three choices to the reference group given my independent variables.
Running the correlation matrix, there's no significant relationship between these above choices with CAR.
Initially I run the OLS regression model for each category: reg CAR majority_owned c.ROA c.Leverage c.Culture_Dist c.International c.market_size, vce(robust)
however, the p-value of the model is super high -> showing that the model is statistically insignificant.
Afterwards, I decided to treat the data as panel data, as of the observations were made for specific firms, and the control variables are firm-specific effects. but the time effects are omitted, since there are multiple observations happens at different dates in a year, the only identifier is the key-id of each company
xtreg CAR majority_owned c.ROA c.Leverage c.Culture_Dist c.International c.market_size, re vce(robust)
xtreg CAR equally_owned c.ROA c.Leverage c.Culture_Dist c.International c.market_size, re vce(robust)
The statistical power of the model is better. However, some of the coefficients are statistically insignificant.
Therefore, i want to ask whether my choice of model is reasonable given the relationship i want to test. Also, what could be the interpretation for the insignificance or tests i could do further? Thank you very much for your help
