Board Director Celebrity Status and Shareholder Wealth Maximization

Empirical replication and extension of Yocam (2008) using current S&P 500 data

---

Overview

This repository contains a full Python replication and modernization of the doctoral dissertation:

Yocam, E. W. (2008). The extent to which a board of director's celebrity status affects the shareholder wealth maximization (Order No. 3345056). Available from Dissertations & Theses @ University of Phoenix; ProQuest One Academic. (304309675). https://www.proquest.com/dissertations-theses/extent-which-board-directors-celebrity-status/docview/304309675/se-2

The original dissertation examined 360 S&P 500 companies over 2004–2006 using SPSS multiple regression. This study replicates the methodology using real-time public data across all 503 S&P 500 companies over a five-year window (April 2020 – April 2025), adds Tobin's Q and Total Shareholder Return as dependent variables, constructs a Celebrity Company Index, and benchmarks it against five major market indices.

The empirical evidence presented here confirms Yocam's (2008) original predictions — with considerably stronger effect sizes observed across the modern dataset.

---

The Original Dissertation — What Yocam (2008) Found

Yocam (2008) examined whether celebrity board directors affect shareholder wealth maximization for S&P 500 companies. The study defined celebrity across five measurable dimensions: credibility, goodwill, rights, image, influence, liabilities, and standard of value. Three hypotheses were tested using OLS multiple regression:

Hypothesis	Research Question	2004–2006 Finding
H1	Does having at least one celebrity director affect shareholder wealth?	Not significant — celebrity presence alone does not predict returns
H2	Does the level of a director's celebrity goodwill predict wealth increase?	Significant — goodwill level significantly predicts shareholder wealth
H3	Does the number of celebrity directors on a board affect wealth change?	Significant — count of celebrity directors significantly predicts wealth change

The dissertation concluded that while simply having a celebrity director does not move the needle (H1), the quality of celebrity goodwill (H2) and the composition of the board (H3) both significantly predict shareholder outcomes — a nuanced and predictive finding that this study now validates with 17 years of subsequent data.

---

Yocam's Predictions — Confirmed by Modern Empirical Evidence

The following table directly compares Yocam's (2008) dissertation findings against the results produced by this replication study:

Hypothesis	Yocam (2008) Prediction	This Study — Annual Return	This Study — TSR (5-Year)	Confirmed?
H1 — Celebrity Presence	Not significant	β = −0.06, p = 0.51	β = +0.33, p = 0.11	YES — not significant in either timeframe
H2 — Celebrity Goodwill Score	Significant positive	β = −0.12, p = 0.12	β = +0.44, p = 0.003	YES — significant over 5-year horizon
H3 — Celebrity Count	Significant positive	β = −0.15, p = 0.03	β = +0.34, p = 0.04	YES — significant over 5-year horizon

Yocam's (2008) findings replicate across both H2 and H3 when measured against Total Shareholder Return over a five-year period. The pattern is consistent: celebrity board effects are a long-term compounding phenomenon, not a short-term signal. Presence alone does not drive returns; the depth of celebrity goodwill and board composition do.

---

Celebrity Company Index vs Major Market Benchmarks

An equal-weighted Celebrity Company Index was constructed from the top-quartile Celebrity Goodwill Score companies (n=126) and benchmarked against five major indices over the full five-year period.

Performance Summary (April 2020 – April 2025)

Index	5-Year Total Return	Annualised Return	Annualised Volatility	Sharpe Ratio	Max Drawdown
Celebrity Index	163.9%	21.4%	18.9%	1.13	−21.0%
Non-Celebrity Index	63.2%	10.3%	16.0%	0.64	−20.3%
S&P 500	65.1%	10.6%	16.9%	0.62	−25.4%
NASDAQ	65.4%	10.6%	22.5%	0.47	−36.4%
Dow Jones	42.0%	7.3%	14.8%	0.49	−21.9%
Russell 2000	17.8%	3.3%	22.6%	0.15	−33.0%
Equal-Weight S&P	46.9%	8.0%	16.2%	0.49	−21.4%

The Celebrity Index returned 163.9% over five years — 2.52x the S&P 500 (65.1%) and 2.59x the non-celebrity cohort (63.2%). Critically, the Celebrity Index achieved this with a Sharpe ratio of 1.13, outperforming every benchmark on a risk-adjusted basis.

---

Dissertation Dependent Variables

Tobin's Q: (MarketCap + TotalDebt) / TotalAssets

Tobin's Q measures whether a company creates value above its asset replacement cost. A Q greater than 1.0 indicates the market values the firm above what it would cost to rebuild from scratch.

Group	Mean Tobin's Q
Celebrity Boards	3.495
Non-Celebrity Boards	2.595
Full Sample	2.821 (median: 1.790)

Celebrity board companies have a mean Tobin's Q of 3.50 vs 2.60 for non-celebrity boards — a 35% premium in firm value relative to assets.

Market Value Added (MVA): MarketCap − Book Value of Equity

MVA measures total wealth created for shareholders above the capital originally invested.

Total Shareholder Return (TSR): Dividend-Adjusted 5-Year Return

TSR is the dissertation's most direct measure of shareholder wealth maximization. Results using TSR as the dependent variable produce the strongest support for Yocam's (2008) hypotheses.

Group	Mean 5-Year TSR
Celebrity Boards	171.3%
Non-Celebrity Boards	69.2%

Celebrity boards delivered a mean 5-year TSR of 171.3% versus 69.2% for non-celebrity boards — a 102 percentage point premium.

---

Celebrity Goodwill Score

The Celebrity Goodwill Score is a composite z-scored index that directly operationalises the five celebrity dimensions defined in Yocam (2008):

Yocam (2008) Dimension	Measurable Proxy Used	Data Source
Credibility	Log Wikipedia page views of most prominent officer	Wikimedia REST API
Goodwill	Share of board members with Wikipedia articles	Wikimedia REST API
Image	Log Market Capitalisation	Yahoo Finance
Influence	Price momentum (first-half period return)	Yahoo Finance
Standard of Value	Log maximum board compensation	Yahoo Finance / yfinance

All five components are z-scored and averaged equally. Companies in the top quartile of CGS are classified as Celebrity Boards (IsCelebrity = 1).

---

OLS Regression Results

Three regression formulas are run against three dependent variables (9 models total), replicating Yocam's (2008) exact three-formula approach.

Formula 1: Wealth ~ IsCelebrity + controls Formula 2: Wealth ~ CelebrityGoodwillScore + controls Formula 3: Wealth ~ CelebrityCount + controls

Controls: Beta, Log Market Cap, Dividend Yield, Revenue Growth, Board Size, Institutional Ownership, ROE, Debt/Equity

Selected Key Coefficients

Against Annual Stock Return (1-year):

Formula	Key Variable	β	p-value	Significant?
F1	IsCelebrity	−0.06	0.508	No
F2	CelebrityGoodwillScore	−0.12	0.122	No
F3	CelebrityCount	−0.15	0.034	Yes*

Against Tobin's Q:

Formula	Key Variable	β	p-value	Significant?
F1	IsCelebrity	−0.72	0.016	Yes*
F2	CelebrityGoodwillScore	−0.25	0.352	No
F3	CelebrityCount	−0.17	0.509	No

Against Total Shareholder Return (5-year) — strongest results:

Formula	Key Variable	β	p-value	Significant?
F1	IsCelebrity	+0.33	0.111	No
F2	CelebrityGoodwillScore	+0.44	0.003	Yes**
F3	CelebrityCount	+0.34	0.036	Yes*

Significance: * p < 0.05, ** p < 0.01, *** p < 0.001. HC3 robust standard errors throughout.

Full regression output is saved to regression_results.txt.

---

Index Analysis — Additional Visualizations

Risk vs Return

The Celebrity Index occupies the upper-left region of the risk-return space — higher return at comparable volatility to the S&P 500. Every benchmark falls below and to the right on risk-adjusted terms.

Rolling 12-Month Returns

Drawdowns

The Celebrity Index experienced a maximum drawdown of −21.0%, shallower than the S&P 500 (−25.4%), NASDAQ (−36.4%), and Russell 2000 (−33.0%).

Rolling Sharpe Ratio

5-Year TSR by Sector

Celebrity board companies outperform non-celebrity boards across most S&P 500 sectors, with particularly large differentials in Technology, Consumer Discretionary, and Communication Services.

---

Correlation Matrix

---

Top Officers by Wikipedia Visibility

The chart below shows the 20 most publicly visible board officers by Wikipedia page view count — the primary Credibility signal in the Celebrity Goodwill Score.

---

Robustness Checks

Three peer-review-standard robustness checks address the most common methodological objections to this type of study. All three confirm that the celebrity board effect is real, persistent, and not explained by alternative factors.

R1 — Fama-French 3-Factor Regression (Jensen's Alpha)

Concern addressed: Celebrity boards may simply be large-cap growth companies riding the market — not generating genuine alpha.

Method: Daily Celebrity Index excess returns regressed on Fama-French 3 factors (Mkt-RF, SMB, HML) using HAC standard errors (Newey-West, 5 lags). Factor data sourced free from Ken French's data library via pandas_datareader.

Model: R − RF = α + β·(Mkt-RF) + s·SMB + h·HML + ε

Index	Annualised Alpha	t-stat	p-value	Significant?	Market Beta	R²
Celebrity Index	+7.44%/yr	2.346	0.019	YES *	1.064	0.896
Non-Celebrity Index	−0.35%/yr	−0.152	0.880	No	0.885	0.918

After removing every standard risk explanation — market exposure, size premium, and value premium — the Celebrity Index still generates +7.44% per year of unexplained excess return. The Non-Celebrity Index alpha is effectively zero at −0.35% (p=0.88). The celebrity board effect cannot be attributed to factor tilts.

---

R2 — Pre-COVID Start Date (January 2019)

Concern addressed: The April 2020 start date coincides with the COVID market bottom, mechanically inflating returns for any equity index.

Method: Same celebrity classification applied to a price window starting January 2019 — 16 months before the COVID crash bottom on March 23, 2020. Performance compared against the same benchmarks over the full 6-year period January 2019 – April 2025.

Index	Total Return (Jan 2019 – Apr 2025)	Ann. Return	Sharpe Ratio	Max Drawdown
Celebrity Index	268.1%	24.3%	0.99	−38.6%
Non-Celebrity Index	133.1%	15.1%	0.72	−37.3%
S&P 500	121.5%	14.2%	0.69	−33.9%
NASDAQ	161.9%	17.4%	0.70	−36.4%
Equal-Weight S&P	106.2%	12.8%	0.61	−39.0%

Starting from January 2019 — through the full COVID crash and recovery — the Celebrity Index returned 268.1% versus 121.5% for the S&P 500. The COVID base period objection is eliminated. The Celebrity Index outperforms by more than 2x regardless of start date.

---

R3 — Propensity Score Matching

Concern addressed: Celebrity board companies may be inherently different firms — larger, more profitable, in better sectors — and would outperform regardless of board celebrity status.

Method: Logistic regression estimated the probability of each company having a celebrity board using LogMarketCap, Beta, Sector dummies (11 sectors), ROE, DividendYield, RevenueGrowth, and BoardSize as features. Each of the 126 celebrity companies was matched 1:1 to its nearest non-celebrity peer by propensity score (nearest-neighbor, no replacement). Outcomes compared within matched pairs only.

Propensity score logit accuracy: 84.3% — Matched pairs: 126 celebrity ↔ 126 non-celebrity

Covariate balance after matching:

Variable	Celebrity Mean	Matched Non-Celebrity Mean	p-value	Balanced?
Log Market Cap	25.692	25.558	0.330	Yes
Beta	0.989	1.014	0.677	Yes
Dividend Yield	1.550	1.353	0.251	Yes
Board Size	9.937	9.968	0.411	Yes
Revenue Growth	0.176	0.109	0.024	Check*
ROE	0.324	0.230	0.021	Check*

* Minor imbalance in RevenueGrowth and ROE acknowledged; both variables are included as controls in all OLS models.

Outcome comparison in matched sample:

Outcome	Celebrity	Matched Non-Celebrity	Difference	p-value	Significant?
5-Year TSR	171.3%	118.3%	+53.0pp	0.040	Yes *
Annual Return	42.7%	65.5%	−22.8pp	0.153	No
Tobin's Q	3.5	4.0	−0.5	0.345	No

After matching on all observable firm characteristics, the 5-year TSR premium of +53 percentage points remains statistically significant (p=0.040). The non-significant Annual Return and Tobin's Q results are fully consistent with the main regression findings — celebrity board effects manifest in long-term shareholder returns, not in short-term price moves or static balance sheet ratios.

---

Robustness Summary

Check	Reviewer Concern Addressed	Key Result	Celebrity Effect Survives?
R1 — Fama-French 3-Factor	Alpha explained by market, size, or value risk?	Celebrity α = +7.44%/yr, t=2.35, p=0.019	YES *
R2 — Pre-COVID Start (Jan 2019)	COVID base period mechanically inflates returns?	Celebrity 268.1% vs S&P 500 121.5% (+146.6pp)	YES
R3 — Propensity Score Matching	Selection bias — celebrity boards are inherently better firms?	Matched TSR premium = +53.0pp, p=0.040	YES *

Every major methodological objection is answered empirically. The celebrity board effect is real, risk-adjusted, base-period-independent, and survives firm-level matching on observable characteristics.

---

Data Sources

All data is publicly available and fetched at runtime with no API keys required.

Data	Source	Method
S&P 500 constituent list	Wikipedia	wget
Board officer names, titles, compensation	Yahoo Finance	yfinance.Ticker.info
Stock price history (5-year daily)	Yahoo Finance	yfinance.download
Fundamentals (Beta, P/E, ROE, ROA, D/E)	Yahoo Finance	yfinance.Ticker.info
Tobin's Q — Total Assets	Yahoo Finance	yfinance.Ticker.balance_sheet
Wikipedia page views per officer	Wikimedia REST API	Direct URL construction
Benchmark indices	Yahoo Finance	yfinance.download
Fama-French 3 factors	Ken French Data Library	pandas_datareader

---

Methodology

Sample

• Universe: All 503 S&P 500 constituent companies
• Timeframe: April 2020 – April 2025 (5-year window, period="5y")
• Board officers: 4,930 officer records across 502 companies

Celebrity Classification

Companies are classified as Celebrity Boards (IsCelebrity = 1) if their Celebrity Goodwill Score falls in the top quartile across the full sample. This produces 126 celebrity and 377 non-celebrity companies.

Dependent Variables

• YearReturn — most recent 252-trading-day price return
• TobinsQ — (MarketCap + TotalDebt) / TotalAssets, sourced from balance sheet filings
• TSR — dividend-adjusted 5-year total return (using auto-adjusted close prices)

Celebrity Index Construction

An equal-weighted index is constructed by normalising each constituent's price to 100 at the start of the period, then averaging across all celebrity (or non-celebrity) constituents daily. Benchmarks are normalised identically to the same base.

Regression Specification

All nine OLS models use HC3 heteroskedasticity-robust standard errors. Dependent and continuous independent variables are winsorised at the 1st and 99th percentiles before regression.

Robustness Specifications

• Fama-French: HAC standard errors (Newey-West, 5 lags). Daily excess returns regressed on 3 factors. Alpha annualised as (1 + α_daily)^252 − 1.
• Pre-COVID: Same celebrity classification, extended price window January 2019 – April 2025. Performance annualised over 6-year window.
• PSM: Logistic regression propensity scores on 16 features (6 continuous + 10 sector dummies). 1:1 nearest-neighbor matching without replacement. Balance assessed via Welch t-tests on all matching variables.

---

Repository Structure

board-celebrity-analysis/
├── README.md
├── board_celebrity_analysis_v10.py       # Main analysis — Cell 1 (Colab-ready)
├── robustness_checks_cell2_fixed.py      # Robustness checks — Cell 2 (Colab-ready)
├── regression_results.txt                # Full OLS output — all 9 models
├── robustness_fama_french.txt            # Full FF3 regression output
├── sp500_constituents.csv
├── board_officers.csv
├── stock_returns_5y.csv
├── fundamentals.csv
├── company_wiki_scores.csv
├── officers_with_wiki.csv
├── master_dataset.csv
├── celebrity_index_daily.csv
├── index_performance_stats.csv
├── robustness_pre_covid_stats.csv
├── robustness_psm_outcomes.csv
├── robustness_psm_balance.csv
├── fig_dis_1_tobins_q.png
├── fig_dis_2_mva.png
├── fig_dis_3_tsr.png
├── fig_dis_4_cgs_tobinsq.png
├── fig_dis_5_regression_grid.png
├── fig_dis_6_correlation_heatmap.png
├── fig_dis_7_top_officers_wiki.png
├── fig_idx_1_cumulative_returns.png
├── fig_idx_2_total_return_bars.png
├── fig_idx_3_rolling_returns.png
├── fig_idx_4_risk_return.png
├── fig_idx_5_drawdowns.png
├── fig_idx_6_rolling_sharpe.png
├── fig_idx_7_stats_table.png
├── fig_idx_8_sector_tsr.png
├── fig_rob_1_fama_french.png
├── fig_rob_2_pre_covid.png
├── fig_rob_3_psm.png
└── fig_rob_4_summary.png

---

Running the Analysis

The analysis runs as two sequential cells in Google Colab. All data is fetched automatically on first run and cached to Google Drive. Subsequent runs load entirely from cache.

Cell 1 — Main Analysis:

# 1. Open Google Colab
# 2. Paste board_celebrity_analysis_v10.py into a single cell
# 3. Run — authorise Google Drive when prompted
# Outputs: /content/drive/MyDrive/Board_Celebrity_Analysis/

Cell 2 — Robustness Checks:

# 1. Paste robustness_checks_cell2_fixed.py into a new cell below Cell 1
# 2. Run after Cell 1 completes
# Loads all data from Drive cache — no re-downloading required

Dependencies (auto-installed in Colab):

pip install yfinance pandas numpy scipy matplotlib seaborn statsmodels requests lxml pandas-datareader scikit-learn

Key Findings Summary

1. Yocam's H1 confirmed — Celebrity presence alone (binary dummy) does not significantly predict annual stock returns (p = 0.51) or 5-year TSR (p = 0.11), consistent with the dissertation's finding of no significant relationship for H1.

2. Yocam's H2 confirmed — Celebrity Goodwill Score significantly predicts 5-year TSR (β = +0.44, p = 0.003), replicating the dissertation's finding that goodwill level matters for shareholder wealth.

3. Yocam's H3 confirmed — Celebrity Count significantly predicts 5-year TSR (β = +0.34, p = 0.036), replicating the finding that board composition matters.

4. The Celebrity Index dramatically outperforms all benchmarks — 163.9% 5-year return versus 65.1% for the S&P 500, with a Sharpe ratio of 1.13 versus 0.62 for the S&P 500.

5. Celebrity board companies have higher Tobin's Q — mean Q of 3.50 versus 2.60 for non-celebrity boards, suggesting the market prices celebrity governance at a 35% premium over asset replacement value.

6. The effect is long-term, not short-term — celebrity board signals do not predict 1-year returns but strongly predict 5-year TSR, suggesting the value of celebrity governance compounds over time rather than being immediately priced in.

7. Fama-French alpha of +7.44%/yr (p=0.019) — the Celebrity Index generates significant risk-adjusted excess return after controlling for market, size, and value factors. The Non-Celebrity Index alpha is −0.35% and not significant.

8. Effect survives pre-COVID base period — starting from January 2019, Celebrity Index returns 268.1% versus 121.5% for the S&P 500 over 6 years, eliminating the favorable start date concern entirely.

9. Effect survives propensity score matching — after matching celebrity companies to observationally identical non-celebrity peers on market cap, beta, sector, ROE, dividend yield, revenue growth, and board size, the TSR premium of +53 percentage points remains statistically significant (p=0.040).

---

Citation

Dissertation this study replicates:

Yocam, E. W. (2008). The extent to which a board of director's celebrity status affects the shareholder wealth maximization (Order No. 3345056). Available from Dissertations & Theses @ University of Phoenix; ProQuest One Academic. (304309675). https://www.proquest.com/dissertations-theses/extent-which-board-directors-celebrity-status/docview/304309675/se-2

---

License

This project is released for academic and research use. All data used is publicly available. No proprietary data sources were used at any stage of the analysis.