Dear Community,
I'm currently working on my master's thesis where we aim to study the effect that certain rules have on the number of leveraged buyouts that take place in a given country.
The panel is structured as follows (numbers are made up in this example):
I originally intended to use the absolute `no. of leveraged transactions` as the dependent variable (count data). 'Rule dummy' is the main independent variable of interest and I've also included some time-varying controls like GDP and population.
Running the regression with country and year fixed effects I get:
DV: No. of Leveraged Transactions
Estimate Std. Error t-value Pr(>|t|)
rule -4.17157 0.96766 -4.311 1.807e-05 *** (No controls)
When adding e.g. GDP and the 10Y yield as controls I instead get:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 8.3480 3.6333 2.298 0.02186 *
rule 1.2833 1.1805 1.087 0.27736
gdp_trn 2.9497 0.2662 11.080 < 2e-16 ***
US10y 0.2163 1.4732 0.147 0.88331
The rule becomes insignificant, which is not surprising as more transactions take place as GDP increases, i.e, GDP is highly correlated with the number of leveraged transactions.
I was wondering if there's a way to improve the model to avoid the aforementioned issue.
Here are some ideas I've come up with:
1) Use 5th column as the dependent variable (leveraged / total transactions).
2) Use gdp growth instead of the absolute GDP figures.
Here's what I'd like to know:
-The DV would now be proportions (from 0 to 1) instead of count data. Would this be a problem in my fixed effects set up? I assume that beyond the change in the interpretation of the coefficient, the regression output would also change as the proportions incorporate the total number of transactions vs. the original DV?
- Alternatively, is it also sensible to keep the original DV (count data) but replace GDP with GDP_growth?
- If I use GDP_growth, should I also use population_growth (otherwise absolute population would explain the number of leveraged buyouts)?
I'd really appreciate any comments or feedback! Please let me know if something is not clear.
Best regards,
Diego
I'm currently working on my master's thesis where we aim to study the effect that certain rules have on the number of leveraged buyouts that take place in a given country.
The panel is structured as follows (numbers are made up in this example):
| Country | Year | No. of leveraged Transactions | Total Transactions (leveraged+non-leveraged) | Leveraged / Total | Rule (dummy) | GDP ($trn) | Population | US 10Y Yield | ... | |
| A | 2000 | 3 | 10 | .3 | 0 | 1 | 200 | 7 | ||
| A | 2001 | 2 | 8 | .25 | 1 | 1.2 | 344 | 6 | ||
| A | 2002 | 0 | 3 | 0 | 1 | 1.7 | 355 | 5 | ||
| B | 2000 | 6 | 6 | 1 | 0 | .4 | 73 | 7 | ||
| B | 2001 | 5 | 20 | .25 | 0 | .7 | 75 | 6 | ||
| B | 2002 | 3 | 9 | .33 | 1 | 1 | 76 | 5 |
Running the regression with country and year fixed effects I get:
DV: No. of Leveraged Transactions
Estimate Std. Error t-value Pr(>|t|)
rule -4.17157 0.96766 -4.311 1.807e-05 *** (No controls)
When adding e.g. GDP and the 10Y yield as controls I instead get:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 8.3480 3.6333 2.298 0.02186 *
rule 1.2833 1.1805 1.087 0.27736
gdp_trn 2.9497 0.2662 11.080 < 2e-16 ***
US10y 0.2163 1.4732 0.147 0.88331
The rule becomes insignificant, which is not surprising as more transactions take place as GDP increases, i.e, GDP is highly correlated with the number of leveraged transactions.
I was wondering if there's a way to improve the model to avoid the aforementioned issue.
Here are some ideas I've come up with:
1) Use 5th column as the dependent variable (leveraged / total transactions).
2) Use gdp growth instead of the absolute GDP figures.
Here's what I'd like to know:
-The DV would now be proportions (from 0 to 1) instead of count data. Would this be a problem in my fixed effects set up? I assume that beyond the change in the interpretation of the coefficient, the regression output would also change as the proportions incorporate the total number of transactions vs. the original DV?
- Alternatively, is it also sensible to keep the original DV (count data) but replace GDP with GDP_growth?
- If I use GDP_growth, should I also use population_growth (otherwise absolute population would explain the number of leveraged buyouts)?
I'd really appreciate any comments or feedback! Please let me know if something is not clear.
Best regards,
Diego
Comment