Replicating Bebchuk, Cremers & Peyer (2011): The CEO Pay Slice¶

U.S.E. Finance Data Hub / WRDS Database: ExecuComp¶

Paper replicated: Bebchuk, L. A., Cremers, K. J. M. & Peyer, U. C. (2011). The CEO Pay Slice. Journal of Financial Economics, 102(1), 199–221.

Database used: ExecuComp (Data Hub guide)

---

What this notebook does¶

Bebchuk, Cremers & Peyer (2011) introduce the "CEO Pay Slice" (CPS), defined as the fraction of the aggregate compensation of a firm's top five executives that is captured by the CEO. They show that a higher CPS is associated with lower firm value (industry-adjusted Tobin's Q), worse accounting performance, and weaker corporate governance. The paper has been cited over 1,400 times and CPS has become a standard measure of CEO dominance in the governance literature.

The beauty of this paper for a database tutorial is that the entire analysis can be done within a single ExecuComp query. All the compensation variables needed live in the anncomp table.

1. Pull executive-level compensation data from ExecuComp (comp_execucomp.anncomp) via the WRDS API
2. Identify the CEO and the top-5 paid executives for each firm-year
3. Construct the CEO Pay Slice (CPS) as the CEO's share of top-5 total compensation
4. Visualize the distribution and time trend of CPS across S&P 1500 firms
5. Compare our findings to the published 2011 results

Learning objectives¶

• Practice connecting to WRDS and querying executive compensation data from ExecuComp
• Understand the structure of the anncomp table and key variables like tdc1, ceoann, and execid
• Learn how to aggregate executive-level data to the firm-year level
• Build intuition for how CEO compensation is structured and measured in empirical governance research

Requirements to run this notebook¶

• A valid WRDS account with ExecuComp access (ask the Finance Data Hub if you don't have one yet)
• pip install wrds pandas numpy matplotlib seaborn
• You will be prompted for your WRDS username/password the first time you connect (or set up a .pgpass file, see the WRDS Python guide)

1. Setup and WRDS connection¶

In [ ]:

import wrds
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

sns.set_theme(style="whitegrid", palette="deep")
pd.set_option("display.max_columns", 50)

db = wrds.Connection()

import wrds import pandas as pd import numpy as np import matplotlib.pyplot as plt import seaborn as sns sns.set_theme(style="whitegrid", palette="deep") pd.set_option("display.max_columns", 50) db = wrds.Connection()

2. Identifying the ExecuComp variables we need¶

All variables come from the comp_execucomp.anncomp table (Annual Compensation), which is the main table in ExecuComp. Each row represents one executive in one firm in one fiscal year.

Paper concept	Description	ExecuComp field
Total Compensation	Total compensation (salary + bonus + stock awards + option awards + non-equity incentive + pension change + other)	tdc1
CEO Flag	Whether the executive served as CEO during the fiscal year	ceoann
Executive ID	Unique executive identifier	execid
Company ID	Compustat firm identifier	gvkey
Fiscal Year	Fiscal year	year
Company Name	Company name	coname
Executive Name	Full name of executive	exec_fullname
Title	Executive title	title

The variable tdc1 is the standard total compensation measure used in the literature. It is calculated by S&P and includes salary, bonus, the grant-date fair value of stock and option awards, non-equity incentive plan compensation, the change in pension value, and all other compensation. The ceoann field takes the value 'CEO' if the executive served as CEO during the fiscal year.

In [ ]:

# Pull all executive compensation records for the analysis period.
# We need all top executives (not just CEOs) to compute CPS.
query = """
    SELECT gvkey, year, execid, exec_fullname, coname, title,
           ceoann, tdc1
    FROM comp_execucomp.anncomp
    WHERE year BETWEEN 2006 AND 2023
      AND tdc1 IS NOT NULL
"""

df = db.raw_sql(query)
print(f"Rows pulled: {len(df):,}")
print(f"Unique firms: {df['gvkey'].nunique():,}")
df.head()

# Pull all executive compensation records for the analysis period. # We need all top executives (not just CEOs) to compute CPS. query = """ SELECT gvkey, year, execid, exec_fullname, coname, title, ceoann, tdc1 FROM comp_execucomp.anncomp WHERE year BETWEEN 2006 AND 2023 AND tdc1 IS NOT NULL """ df = db.raw_sql(query) print(f"Rows pulled: {len(df):,}") print(f"Unique firms: {df['gvkey'].nunique():,}") df.head()

3. Cleaning the sample¶

In [ ]:

# Remove records with zero or negative total compensation
df = df[df["tdc1"] > 0]

# Identify the CEO for each firm-year
df["is_ceo"] = (df["ceoann"] == "CEO").fillna(False).astype(int)

# For each firm-year, rank executives by tdc1 and keep only the top 5
df["rank"] = df.groupby(["gvkey", "year"])["tdc1"].rank(ascending=False, method="first")
top5 = df[df["rank"] <= 5].copy()

# Require that we have exactly one CEO and at least 4 other executives
# (so there are 5 named executives total, as in the paper)
firm_year_stats = top5.groupby(["gvkey", "year"]).agg(
    n_execs=("execid", "nunique"),
    n_ceos=("is_ceo", "sum")
).reset_index()

valid = firm_year_stats[(firm_year_stats["n_execs"] == 5) & (firm_year_stats["n_ceos"] == 1)]
top5 = top5.merge(valid[["gvkey", "year"]], on=["gvkey", "year"], how="inner")

print(f"Firm-years with valid top-5 teams: {valid.shape[0]:,}")
print(f"Executive-year observations: {len(top5):,}")

# Remove records with zero or negative total compensation df = df[df["tdc1"] > 0] # Identify the CEO for each firm-year df["is_ceo"] = (df["ceoann"] == "CEO").fillna(False).astype(int) # For each firm-year, rank executives by tdc1 and keep only the top 5 df["rank"] = df.groupby(["gvkey", "year"])["tdc1"].rank(ascending=False, method="first") top5 = df[df["rank"] <= 5].copy() # Require that we have exactly one CEO and at least 4 other executives # (so there are 5 named executives total, as in the paper) firm_year_stats = top5.groupby(["gvkey", "year"]).agg( n_execs=("execid", "nunique"), n_ceos=("is_ceo", "sum") ).reset_index() valid = firm_year_stats[(firm_year_stats["n_execs"] == 5) & (firm_year_stats["n_ceos"] == 1)] top5 = top5.merge(valid[["gvkey", "year"]], on=["gvkey", "year"], how="inner") print(f"Firm-years with valid top-5 teams: {valid.shape[0]:,}") print(f"Executive-year observations: {len(top5):,}")

4. Constructing the CEO Pay Slice¶

The CEO Pay Slice is simply the CEO's total compensation as a fraction of the sum of the top five executives' total compensation.

$$\text{CPS} = \frac{\text{CEO Total Compensation (TDC1)}}{\sum_{i=1}^{5} \text{Top-5 Executive Total Compensation (TDC1)}_i}$$

By construction, CPS falls between 0 and 1. A value of 0.20 means the CEO captures exactly one-fifth of the top-5 pay (equal sharing). Values above 0.20 indicate that the CEO captures a disproportionate share. Bebchuk et al. find a mean CPS of about 0.35 in their sample (1993 to 2008), meaning CEOs on average earn roughly 35% of the top-5 total.

In [ ]:

# Compute sum of top-5 compensation per firm-year
top5_sum = top5.groupby(["gvkey", "year"])["tdc1"].sum().reset_index()
top5_sum.columns = ["gvkey", "year", "top5_total_comp"]

# Get CEO compensation per firm-year
ceo_comp = top5[top5["is_ceo"] == 1][["gvkey", "year", "tdc1", "exec_fullname", "coname"]].copy()
ceo_comp.columns = ["gvkey", "year", "ceo_comp", "ceo_name", "company"]

# Merge and compute CPS
cps = ceo_comp.merge(top5_sum, on=["gvkey", "year"], how="inner")
cps["cps"] = cps["ceo_comp"] / cps["top5_total_comp"]

print(f"Firm-years with CPS: {len(cps):,}")
print(f"\nCPS summary statistics:")
print(cps["cps"].describe())

# Compute sum of top-5 compensation per firm-year top5_sum = top5.groupby(["gvkey", "year"])["tdc1"].sum().reset_index() top5_sum.columns = ["gvkey", "year", "top5_total_comp"] # Get CEO compensation per firm-year ceo_comp = top5[top5["is_ceo"] == 1][["gvkey", "year", "tdc1", "exec_fullname", "coname"]].copy() ceo_comp.columns = ["gvkey", "year", "ceo_comp", "ceo_name", "company"] # Merge and compute CPS cps = ceo_comp.merge(top5_sum, on=["gvkey", "year"], how="inner") cps["cps"] = cps["ceo_comp"] / cps["top5_total_comp"] print(f"Firm-years with CPS: {len(cps):,}") print(f"\nCPS summary statistics:") print(cps["cps"].describe())

5. Visualizing the results¶

In [ ]:

fig, axes = plt.subplots(1, 2, figsize=(14, 5))

# Distribution of CPS
axes[0].hist(cps["cps"], bins=50, color="steelblue", edgecolor="white")
axes[0].axvline(0.20, color="red", linestyle="--", label="Equal share (0.20)")
axes[0].axvline(cps["cps"].mean(), color="green", linestyle="--",
                label=f"Mean ({cps['cps'].mean():.2f})")
axes[0].set_title("Distribution of CEO Pay Slice\n(ExecuComp, 2006-2023)")
axes[0].set_xlabel("CEO Pay Slice (CPS)")
axes[0].set_ylabel("Number of firm-years")
axes[0].legend()

# Mean and median CPS over time
yearly = cps.groupby("year")["cps"].agg(["mean", "median"])
axes[1].plot(yearly.index, yearly["mean"], "o-", color="steelblue", label="Mean CPS")
axes[1].plot(yearly.index, yearly["median"], "s--", color="darkorange", label="Median CPS")
axes[1].axhline(0.20, color="red", linestyle=":", alpha=0.5, label="Equal share")
axes[1].set_title("CEO Pay Slice Over Time")
axes[1].set_xlabel("Fiscal Year")
axes[1].set_ylabel("CPS")
axes[1].legend()

plt.tight_layout()
plt.show()

fig, axes = plt.subplots(1, 2, figsize=(14, 5)) # Distribution of CPS axes[0].hist(cps["cps"], bins=50, color="steelblue", edgecolor="white") axes[0].axvline(0.20, color="red", linestyle="--", label="Equal share (0.20)") axes[0].axvline(cps["cps"].mean(), color="green", linestyle="--", label=f"Mean ({cps['cps'].mean():.2f})") axes[0].set_title("Distribution of CEO Pay Slice\n(ExecuComp, 2006-2023)") axes[0].set_xlabel("CEO Pay Slice (CPS)") axes[0].set_ylabel("Number of firm-years") axes[0].legend() # Mean and median CPS over time yearly = cps.groupby("year")["cps"].agg(["mean", "median"]) axes[1].plot(yearly.index, yearly["mean"], "o-", color="steelblue", label="Mean CPS") axes[1].plot(yearly.index, yearly["median"], "s--", color="darkorange", label="Median CPS") axes[1].axhline(0.20, color="red", linestyle=":", alpha=0.5, label="Equal share") axes[1].set_title("CEO Pay Slice Over Time") axes[1].set_xlabel("Fiscal Year") axes[1].set_ylabel("CPS") axes[1].legend() plt.tight_layout() plt.show()

In [ ]:

# Top 10 highest-CPS firm-years in the sample
top_cps = cps.nlargest(10, "cps")[["company", "ceo_name", "year", "ceo_comp", "top5_total_comp", "cps"]]
top_cps["ceo_comp"] = top_cps["ceo_comp"].apply(lambda x: f"${x:,.0f}")
top_cps["top5_total_comp"] = top_cps["top5_total_comp"].apply(lambda x: f"${x:,.0f}")
top_cps["cps"] = top_cps["cps"].apply(lambda x: f"{x:.2%}")
print("Top 10 Highest CEO Pay Slices in the Sample:\n")
print(top_cps.to_string(index=False))

# Top 10 highest-CPS firm-years in the sample top_cps = cps.nlargest(10, "cps")[["company", "ceo_name", "year", "ceo_comp", "top5_total_comp", "cps"]] top_cps["ceo_comp"] = top_cps["ceo_comp"].apply(lambda x: f"${x:,.0f}") top_cps["top5_total_comp"] = top_cps["top5_total_comp"].apply(lambda x: f"${x:,.0f}") top_cps["cps"] = top_cps["cps"].apply(lambda x: f"{x:.2%}") print("Top 10 Highest CEO Pay Slices in the Sample:\n") print(top_cps.to_string(index=False))

6. Comparing to Bebchuk, Cremers & Peyer (2011): What's the same, what's different¶

What replicates cleanly

The CPS construction maps perfectly onto ExecuComp. The tdc1 field and the ceoann flag are exactly the variables the authors use, and the anncomp table provides all the information needed in a single query. The mean CPS in our sample should be in the neighborhood of 0.35 to 0.40, consistent with the paper's findings. The time trend should show a gradual increase in CPS over the past two decades, reflecting the well-documented growth in CEO pay relative to other executives.

Where a modern WRDS-based replication necessarily differs from the original

1. Sample period. Bebchuk et al. use 1993 to 2008. Our sample starts from 2006, which means we cover different economic regimes (including post-financial-crisis changes to executive compensation like say-on-pay requirements under Dodd-Frank). The level of CPS may differ as compensation practices have evolved.

2. Firm value regressions. The paper's main contribution is showing that CPS is negatively associated with industry-adjusted Tobin's Q. This requires merging ExecuComp with Compustat to get firm-level financial data (at, mkvalt, ceq). We focus on constructing and visualizing CPS itself, but the valuation regression is a natural extension students can build by joining on gvkey and year.

3. Compensation definition. The paper uses tdc1 throughout, which is the standard measure. However, S&P has revised the calculation of tdc1 over time, particularly after the 2006 SEC disclosure rule changes (which altered how option awards are valued). Our post-2006 sample uses the revised calculation consistently, while the original paper spans both the old and new regimes.

4. ExecuComp coverage. ExecuComp covers S&P 1500 firms, so results may not generalize to smaller or private companies. The original paper also restricts to S&P 1500 firms, so this is actually a feature rather than a limitation.

---

References¶

• Bebchuk, L. A., Cremers, K. J. M. & Peyer, U. C. (2011). The CEO Pay Slice. Journal of Financial Economics, 102(1), 199–221.
• WRDS ExecuComp guide: https://uufinance.github.io/data/wrds/databases/execucomp/
• WRDS ExecuComp access page: https://wrds-www.wharton.upenn.edu/pages/get-data/compustat-capital-iq-standard-poors/compustat/execucomp/
• WRDS Python/API setup guide: https://uufinance.github.io/data/wrds/notebook/

Prepared for the U.S.E. Finance Data Hub as a database-tutorial template.