FIN41330: Examine If a Causal Relation Exists between Firm- Level Governance: Empirical Corporate Finance Assignment, UCD

Topic 1

Project Objective: Examine if a causal relation exists between firm-level governance and firm value?

Topic 2

Project Objective: Examine if a causal relation exists between firm-level governance, CEO pay, the value of firm investments (Capex and R&D), and cash-holdings?

Topic 1 – background

The literature suggests that a firm’s governance quality predicts future firm value and returns (e.g., Gompers, Ishii and Metrick, 2003; Bebchuk and Cohen, 2005). Since governance quality is publicly known (and can be measured and presumably diversified), this suggests that markets may be inefficient, or that typical asset pricing models are potentially misspecified, with governance as an omitted risk factor. The empirical analysis should examine if indeed there is a causal relation between governance quality and firm value.

Related empirical papers

• Bebchuk, L. and Cohen, A., ‘The costs of entrenched boards’, Journal of Financial Economics 78, 409-433, 2005.
• Core, J.E., W.R. Guay, and T.O. Rusticus. ‘Does weak governance cause weak stock returns? An examination of firm operating performance and investors’ expectations’, Journal of
  Finance 61, 2006
• Gompers, Paul A., Joy L. Ishii, and Andrew Metrick. ‘Corporate governance and equity prices’, Quarterly Journal of Economics 118, 2003,107–155.

The key papers are Bebchuk and Cohen (2005), Core et al (2006) and Gomper et al (2003). You may want to search for other more recent papers (e.g., Cremers et al., 2015; Larcker et al., 2015 – also on Brightspace) on the relationship between governance, firm value and performance. More recent work can be found on SSRN.com (click on Finance link and do a search on keywords).

Your first objective should be to measure firm governance. Prior studies (including those above) use anti-takeover provisions (ATPs) to define governance quality. Firms with a greater number of ATPs are argued to have more entrenched (dictatorship) boards, giving rise to potential agency problems and value destruction. Your dataset contains information on 3 ATPs that are argued to be the most potent (classified board, dual-class, poison pill). You can use the total number (sum) of these 3 ATPs to create dictator and democracy dummy variables (e.g.,=1 > median or 75th percentile, 0 otherwise). However, you may also want to examine combinations (classified board and pill) or examine the measures separately since some prior recent work suggests that only 2 provisions that do not require shareholder approval (e.g., poison pills) are correlated with lower firm value.

Your empirical analysis should begin creating (generate command) the set of variables required in Stata to operationalize your model. All the components of the variables are within the dataset. Be careful about how you deal with outliers, and it is expected that all continuous variables would be winsorized at the 0.5% level. You should first look at summary statistics and bivariate analysis to determine if indeed governance is related to firm value, which is typically measured as industry adjusted Tobin’s q. You can also examine the relationship with other performance measures (e.g., return on assets, long-run returns, etc). Your regression analysis should follow closely the approach used by Bebchuk and Cohen (2005) and Core et al., 2006, if you also examine firm
performance). Note that it’s important that you try and establish if a causal relation exists, and I expect tests to deal with potential endogeneity.

Topic 2 – background

The literature shows that badly governed or entrenched firms are more likely to destroy shareholder value, and this largely arises due to the agency problem in listed firms when owners no longer manage their firms (Jensen and Meckling, 1976). This value destruction can occur through several channels, including poor investment decisions (i.e., investors undervalue CAPEX, R&D, M&A, or stockpiling cash), or over-generous CEO (pecuniary benefits) pay relative to firm performance.

Related empirical papers

• Masulis, R., Wang, C. and F. Xie, ‘Agency problems at dual-class companies’, Journal of Finance 62, 2008, 1851-1889.
• Faulkender M., and R. Wang, ’Corporate financial policy and the value of cash, Journal of Finance 61, 2006, 1957-1990.

Similar to Topic 1, you should first calculate a governance score or index for each firm based on the 3 ATPs (classified board, dual-class, poison pill) in the dataset. Alternatively, you could focus on combinations (classified board and pill) or individual ATPs separately since prior recent work suggests that only provisions that do not require shareholder approval (e.g., poison pills) are
correlated with lower firm value.

Your empirical analysis should begin creating (generate command) the set of variables required in Stata to operationalize your model(s). All the components of the variables are within the dataset. Be careful in how you deal with outliers, and it is expected that all continuous variables would be winsorized at the 0.5% level. You should first look at summary statistics and bivariate analysis to determine if indeed governance is related to CEO pay and investment outcomes. Note that it’s 3 important that you try and establish if a causal relation exists between governance and the different outcomes, and I expect tests to deal with potential endogeneity. Additional marks will be given for innovation in modelling and empirical testing.

Dataset Information for both topics

The dataset is a large “merged” dataset comprising financial statement data (Compustat and CRSP,) governance data (Risk Metrics), and compensation data (Execucomp) for all US-listed firms for the years 1992 to 2015. The dataset is a Stata data file so can be directly opened in Stata. The definitions of the variables are provided in a separate spreadsheet. When defining variables, you should carefully consult the related paper(s) on definitions of ratios/variables used. You will also have to compute industry averages (e.g., means), and possibly lag and lead variables if you need to examine the relation between governance and prior and future value. Note that to help deal with
endogeneity it’s fairly common for independent variables to be measured as lagged values (t-1), rather than a contemporaneous model when all variables are measured at time t. The ‘by’ command in Stata is useful for this purpose. The ‘by’ and ‘egen’ commands can also be used to calculate industry-year averages.

If you require additional financial data (e.g., if you want to examine long-run monthly returns) this can be extracted from datasets (CRSP) available from the ‘Data Centre’.