Corporate Finance

Managerial Attributes and Executive Compensation Graham, Li, and Qiu Review of Financial Studies, 2012

Manager Compensation | Motivation Motivation

Dataset

Methodology

Results

Parallel Trends Robustness

Contribution

Motivation

Literature

■ Revisit findings of studies on managerial compensation

■ Large body of literature on managerial compensation ■ Claim: research fails to account correctly for fixed effects

■ Assess the (relative) importance of unobserved effects on managerial compensation

■ Introduce labor-economics methodology to finance ■  Identify advantages and shortcomings

■ Bertrand and Schoar (2003) ■ Use sample of firm-moving managers ■ Relate manager fixed effects to corporate activities ■ Find that managers have heterogeneous „styles“

■ Abowd, Kramarz, and Margolis (1999) ■ Sample of ~1m French worker‘s compensation in ~500k firms ■ Provide econometric framework for compensation decomposition 1

08.05.2015, Mannheim

Manager Compensation | Dataset Motivation

Dataset

Methodology

Dataset ■ 1992-2006 ExecuComp-Compustat panel dataset

■ ~25.5k managers, ~2.3k firms ■ Crucial: Manager movement across firms

■ Subsamples constructed for estimations similar in nearly all dimensions but size  Size controlled for in estimation

■ Mean (median) compensation 1.9m (0.9m)  right skewed distribution, ■ 17% are CEOs, 5% women 2

08.05.2015, Mannheim

Results

Parallel Trends Robustness

Contribution

Manager Compensation | Methodology 1/2 Motivation

Dataset

Methodology

Results

Parallel Trends Robustness

Empirical Strategy ■ Rationale: expected wage = manager‘s human capital * rental rate

■ Baseline model: exponential production function; manager- & firmobservables and unobservables determine stock of human capital ■ Empirical model: ln(yit) = Xitβ + Witγ + firm FE + manager FE + time FE + errorit

■ Further testing (1): manager FE regressed on personal characteristics

■ Further testing (2): manager FE regressed on „management styles“ (i.e., policy FE) ■ Empirical model: FE(comp)i = a + β*FE(z)i + errori 3

08.05.2015, Mannheim

Contribution

Manager Compensation | Methodology 2/2 Motivation

Dataset

Methodology

Results

Parallel Trends Robustness

Contribution

Why Can Fixed Effects be Identified? ■ Spell fixed effects method: One dummy variable for each unique manager-firm combination

Pros: Full sample Cons: FEs indistinguishable

■ Mover dummy variable method: Use mobility of managers; Sample reduced to those managers who switch companies

Pros: Disentangles FEs Cons: Sample very small, movers could be systematically different from non-movers

■ „Connectedness“ (AKM) method: Form group connections to also determine effects for non-movers if their firm has at some time hired a mover ■ Starting individual  all firms  all individuals of those firms  all firms of these individuals  until exhausted ■ Connectedness within groups, no mobility across

4

08.05.2015, Mannheim

Pros: Disentangles FEs, estimates effect for non-movers too, sample shrinks moderately Cons: limited mobility leads to imprecision (also for MDV!)

Manager Compensation | Results 1/2 Motivation

Dataset

Methodology

Results

Parallel Trends Robustness

Base Results ■ FEs increase R2 substantially

■ Using AKM: Manager FEs contribute 54% to model fit > Firm FEs at 5% ■ Manager FEs most important, firm characteristics second ■ Economic terms: 1 std.-dev. in manager FE  $2.5m rise in total compensation

■ Theoretical model (Gabaix & Landier, 2008): Larger firm size  Much higher wage ■ Here: OLS coefficient ~0.37 vs. FEs-coefficient ~0.22 ■  Impact of size overstated by factor ~2

■ CEO variable coefficient now represents promotion effect ■ 150% pay increase for CEO, but 35% (promotion) for the same manager 5

08.05.2015, Mannheim

Contribution

Manager Compensation | Results 2/2 Motivation

Dataset

Methodology

Further Results ■ Manager FEs vary substantially

■ Education significant on manager FEs ■ But: model fit only 1% ■ Several personal characteristics unobservable, possibly important determinants

■ Higher manager FEs associated with higher R&D, Investment, Leverage, & dividend yield; associated with lower cash holdings

6

08.05.2015, Mannheim

Results

Parallel Trends Robustness

Contribution

Manager Compensation | Limitations Motivation

Dataset

Methodology

Results

Limitations

Is The 3-Way FE Methodology Unequivocally Advantageous? ■ Research dependent: If variable‘s variation is cross-sectionally  No benefit

■ Methodology cannot cope with time-variant unobservables

■ Estimation precision dependent on proportion of movers

■ Firm-FEs susceptible to matching bias  Problematic if based on unobservables

7

08.05.2015, Mannheim

Contribution

Manager Compensation | Contribution Motivation

8

Dataset

Methodology

Results

Parallel Trends Robustness

Contribution

Contribution

Criticism

■ Enhances the „toolbox“ to deal with unobserved heterogeneity issues

■ What exactly drives fixed effects?  How to get data though?

■ Makes a case for revising/extending research on key questions

■ How do sample companies compare to average U.S. company?  ExecuComp tracks large firms…

■ Shows that manager and firm FEs are substantial determinants of manager pay

■ Closer look at time-variation of variables could help understanding what is absorbed by FEs and what is not

08.05.2015, Mannheim