Assignment Task
1. Summary Statistics:
Obtain a final sample by excluding any missing observations in all the variables you will use in the regression analysis. Compute any appropriate ratios and perform the functional form transformations of the variables that you will use in the subsequent statistical analysis. Winsorize the variables at the 1st and 99th percentiles for variables taking on both positive and negative values, and at the 99th percentile for variables that only take on positive values. Report two tables showing the summary statistics before and after winsorization for all the variables – raw and transformed, including the number of observations, mean, standard deviation, median, minimum and maximum.
2. Pairwise Correlations
Report a correlation matrix for all the variables you will use in the regression analysis (after winsorization)
3. You want to assess whether firms with better ESG practices have an effect on performance. Based on a strong literature review and data availability, choose a variable that proxies for ESG practices and two relevant control variables. One of the control variables should be used as a dummy variable, so make a sensible choice. Control for firm size and add time dummies. Using all these variables run a firm-fixed effects regression and include this table in your project. Label this table: Table 3: Fixed Effects. This table does not count towards the word count. Make sure all the variables are fully defined at the end or beginning of the table in the form of table notes
4. Run the same regression as in task 3 above but report clustered standard errors. Label this table: Table 4: Fixed Effects with Clustered Std Errors. This table does not count towards the word count. Clearly explain your rationale for choosing this clustering and write down the STATA command used to make this calculation.
5. Add an interaction term involving your ESG proxy and your chosen dummy variable to the regression reported in Table 4. Label this table: Table 5: Interaction Effects. This table does not count towards the word count. Carry out an analysis using the margins command showing the marginal effect of your dummy variable when the ESG variable is at its 25th and 75th percentiles. Are these marginal effect statistically significant (answer Yes or No for each effect)? Illustrate the economic magnitude of the variables involving your interaction.
6. Compare the results of the models in Table 3, Table 4 and Table 5: the magnitudes and significance of the coefficients. Come up with at least two plausible explanations for the differences between the estimates involving ESG in tables 4 and 5.
7. Give an example of an unobserved omitted variable that could lead to your ESG estimated coefficient being biased. Spell out in detail the mechanism through which the bias works.
In your regression analysis you are using a cleaned reduced dataset and the reader is therefore not aware of the steps undertaken to reach this final sample. In this part of the project you will assess whether the sample you used in your regressions is representative of the population and how the undertaken data management decisions (e.g. dropping missing observations) that were taken could potentially affect the inference you carried out in the first part of the project.
8. Based on the table above, explain whether the sample you used in Part A is likely to lead to unbiased estimations.
9. List your references (at least 5 references). These have to be different to the ones listed in the section Scenario.
10. Add a table of variable definitions. Label this table: Table 6: Variable Definitions. Make sure that every variable used in every table and regression is clearly defined. Label the variables consistently throughout the project.
