Human capital equity returns: Evidence from the

Jordan Moore 1 June 5, 2019

Abstract Human capital comprises a substantial portion of global wealth, but equity claims to human capital are not publicly traded. Between April 2014 and April 2016, ten active players in the National Football League (NFL) sold equity claims to their future labor income. The proceeds from the IPOs of these equity contracts total more than $38 million and value the human capital of these athletes at more than $369 million. These transactions represent the largest, most transparent dealings in human capital equity contracts by private parties. I estimate the expected returns of these human capital IPOs using NFL contract data and player performance statistics. The average expected return of these IPOs is 3.9% before expenses. Expected returns are larger for human capital equity investments in players from teams in smaller geographic markets.

1Rowan University, Rohrer College of Business. Email: [email protected]. I am grate- ful to James Byrnes, Ron Kaniel, Arthur Kim, Einar Kjenstad, Hunter Land, Anisha Nyatee, and Jerry Warner for helpful comments and suggestions.

1 1. Introduction

Between April 2014 and April 2016, six active players in the National Football League

(NFL) sold equity claims on a percentage of their future labor income to Fantex, a private

firm. The proceeds from the IPOs of these acquired brand income (ABI) contracts total more than $38 million and value the human capital of these six athletes at more than $369 million. These transactions represent the largest, most transparent dealings in human cap- ital equity contracts between private parties. Fantex publicly reported these transactions to the SEC in extensive detail because they were considering going public. Because there are publicly available data on both the performance and contract terms of NFL players, I am able to estimate the expected returns and realized returns from these investments in human capital equity contracts.

Human capital, the present value of future labor income, comprises a substantial portion of total wealth. Jorgenson and Fraumeni (1989) decompose the US National Income and

Product Accounts and estimate that human capital makes up around 93% of wealth. Di

Giovanni and Matsumoto (2011) estimate the total human capital for US citizens in 2007 at $145 trillion (in current dollars), or $482,000 per capita. To compare, the World Bank estimates total market capitalization of listed US equities at around $20 trillion in 2007.

2 Workers and their families typically finance their human capital investments through precautionary savings, inter-family transfers, and federal or private loans. Friedman (1955) suggests the use of equity contracts to finance human capital investment, reflecting the riskiness of individual labor market outcomes.

Although human capital is a gigantic asset class, there is no significant human capital equity market. The 13th Amendment of the United States Constitution abolishes slavery.

2The data are from: http://data.worldbank.org/

2 In the absence of clear federal guidelines, individual states could interpret the sale of claims to future labor income as unconstitutional. Schultz (1961) argues that ambiguity over the interpretation of anti-slavery laws is a substantial factor in the decline of formal apprenticeships. Adverse selection is a concern if athletes who expect to have below average career outcomes are more likely to sell shares of their future income. Moral hazard is a problem if athletes who sell contracts are more likely to retire early or deliberately default.

High administrative costs of monitoring and collecting labor income would substantially reduce returns to investors. Little is known about human capital equity returns and its relation to the expected returns of other assets. Roll (1977) argues that the absence of publicly traded human capital is a substantial hindrance to testing asset price models.

The setting of the Fantex ABI contracts minimizes many concerns about the sale of equity human capital. The SEC did not issue any objections to the validity of the Fantex ABI contracts during the IPO registration process. In 2018, Fantex won a lawsuit against a player who did not make all of the contractually stipulated payments.3 There are minimal concerns about adverse selection because college performance, professional performance, and injury history of athletes negotiating ABI contracts are publicly available. Moral hazard is a small problem because professional athletes are public figures who would incur severe reputational costs from reneging on their contractual obligations. Finally, the very high earnings of individual athletes reduces the proportional cost of administration.

The expected return on equity human capital is an interesting and open question. An athlete selling equity claims to future labor income is like a laborer purchasing life insur- ance. Campbell (1980) argues that the best hedge for a household concerned about future wealth is purchasing life insurance for the primary earner. Households should be willing to

3https://www.nashvillepost.com/sports/sports-business/blog/21018687/kendall-wright-ordered-to-pay- in-fantex-suit

3 overpay for life insurance contracts, implying high returns for the sellers. Matvos (2009) acknowledges that NFL player contracts are not guaranteed. NFL players have highly volatile labor incomes and some players may be willing to sell equity claims to future in- come at a substantial discount. However, the athlete’s fans may derive some benefit from purchasing ABI contracts that is more like consumption than investment. For instance,

NFL fans wager on games and willingly participate in fantasy football leagues.4 These activities provide risky cash flows and negative expected returns.

The NFL is a giant and highly visible business. earns around $7 billion dollars in annual revenues and more than 100 million American viewers watch the every year. There are publicly available player-level data on both performance and compensation.

Using these data, I estimate that investors in these human capital equity IPOs earn average expected returns of 3.9% before transactions costs. I estimate that the expected return on individual contracts ranges from +15.3% to -8.2%. Notably, the expected returns line up with the market size of the player’s team on the IPO date. The two players whose contracts have the highest expected returns play in Buffalo and Cincinnati while the two players whose contracts have the lowest expected returns play in Chicago and San Francisco.

These results are consistent with a “size effect” in human capital equities, just like the size effect in equities first identified by Banz (1981).

2. Human Capital Equity in Theory and Practice

Economists have advocated for equity financing of human capital for decades. Friedman

(1955) advocates for equity financing of education and other investments in human capital.

He argues that government should subsidize “education for citizenship” because primary

4The International Business Times estimates that Americans wagered $95 billion on NFL and col- lege football in the 2015 season. See http://www.ibtimes.com/legalized-sports-gambling-americans-bet-95- billion-nfl-college-football-season-mostly-2089606

4 and secondary education produce large positive externalities. This argument applies less to professional and vocational education, as benefits primarily raise the economic productivity of the individual student. In frictionless markets, investments in physical capital and human capital should earn similar rates of return. Abnormally high returns to human capital investment imply some students are unable to fund their education. Schultz (1961) argues that investments in human capital explain large increases in output that aren’t attributable to changes in capital and labor hours. However, Schultz (1961) stipulates that some education expense is consumption rather than investment.

Friedman (1955) argues that equity financing of human capital investment is appropriate for the high variance in individual labor outcomes. Palacios (2002) notes that an individual’s future income after education is highly uncertain because he or she could fail to complete a degree program, could learn skills that are not highly valued, or could voluntarily choose a career that does not maximize wages. Subsidized federal student loans transfer adverse selection risk to the taxpayer, while equity contracts align the interests of the student and investor. Friedman (1955) argues that governments should administer equity human capital contracts because of the ability to enforce payment through income taxes. Moral hazard remains a risk at the individual student level, which is only mitigated through diversification.

There are significant economic benefits to increasing investment in human capital. Becker

(1962) and Mincer (1974) show that more educated workers earn higher wages at any age or level of experience. The wages of more educated workers grow for longer before leveling off. More educated workers have less frequent and shorter periods of unemployment. Card

(2001) reviews the large literature that measures the causal relation of education on wages using both ordinary least squares (OLS) and instrumental variables (IV) methods. The omitted variable of ability produces an upward bias in the OLS estimate and error in

5 measuring educational attainment produces a downward bias, as acknowledged by Griliches

(1977). To address these biases, popular IVs include quarter of birth, proximity to the nearest college, changes to mandatory schooling laws, and differences in cost of education.

Card (2001) finds that IV estimates are typically similar or higher than OLS estimates, although differences aren’t statistically significant because the IV estimates are imprecise.

An additional year of education causes 7-8% higher wages using the OLS estimate, or

9-10% higher wages using the IV estimate.

Equity contracts that finance higher education could appeal to both investors and students.

Investors benefit from higher yields than fixed income securities and low return correlations with other asset classes. Students would face lower stress from the burden of defaulting on fixed payments. Palacios (2002) argues for a regulatory environment to support these contracts: enforceability in the courts, designation as securities suitable for investment funds, tax deductibility, and bankruptcy protection. A transparent marketplace for these securities would make the relative economic value of certain degrees more transparent and encourage students to make productive educational investments. Educational institutions could initiate truly needs-blind admissions policies, since all students could fund their education at market rates. Financial institutions could structure and market portfolios of individual contracts according to university, course of study, or student demographics to appeal to investors.

In 1989, the Australian government was the first to start an income-based repayment pro- gram for university and graduate education. The Higher Education Contribution Scheme

(HECS) addresses the needs of students whose parents are unable or unwilling to help

finance their higher education. Under the HECS, students can pay the nominal tuition amount at enrollment, with a discount that fell from 25% in 1989 to 9% in 2012 . Alter- natively, students can defer payment until their income exceeds a threshold. Students pay

6 a percentage of their income in excess of the threshold until the debt is repaid. This debt accrues interest at the rate of inflation. Chapman (1997) finds that HECS enrollment in- creases by more than 4% annually. Also, the program is efficient to administer because the government receives the borrower’s details upon enrollment and automatically implements collections via the tax system. Chapman (1997) estimates that administrative costs of the

HECS are less than 1% of revenue.

Other countries, including the United States, United Kingdom, Hungary, South Africa,

South Korea, and New Zealand, have since adopted income-based repayment of education loans. Yale University initiated a Tuition Postponement Program including a partial equity component in 1971, but ended the program after subsidized federal loans became widely available. As of 2014, all borrowers of US Federal Student Loans can choose an income- based repayment plan. Relative to the HECS, the US income-based repayment options are less appealing to students because of lower income thresholds, a higher required percentage repayment, and higher interest rates. Also, it is substantially more difficult to sign up for income-based repayment in the US, and less than 20% of eligible borrowers use the program.

5

Although government issued student loans with income-based repayment are the largest human capital equity investments, several private entities have offered human capital equity contracts. In the 1990s, Roy Chapman organized a firm called Human Capital Resources.

The company never issued contracts because of concerns that residents of states that pro- hibit the sale of future income might legally avoid repayment. My Rich Uncle, a firm founded in 2001, sold human capital contracts using a forecasting model to determine ap- propriate income repayment rates. The firm grew quickly and went public in 2004, but

5See a discussion of income-based repayment at: http://www.theatlantic.com/education/archive/2016/03/australia- college-payment-model-exposes-shortcomings-of-new-american-version/473919/

7 it filed for bankruptcy in 2008 because it was obtain financing during the financial crisis.

Another private firm, Lumni, offered “income share agreements” (ISAs) for US students to access higher education, but no longer offers these contracts in the US due to weak demand and regulatory uncertainty. Pave and Upstart are two other US firms which sold

ISA contracts, but stopped in 2014. According to Upstart CEO Dave Girouard, “while many regulatory and policy efforts are underway to facilitate the development of the (ISA) market, these efforts will likely take many years - a time frame ill-suited for a startup like ours.” 6

Lumni continues to sell ISAs to students in Chile, Colombia, Mexico, and Peru. According to a 2012 US News and World Report article, Lumni earns 7-11% on its ISA investments in

Latin America.7 Education Equity, Inc. offers ISAs to students attending specific educa- tional programs with promising risk-adjusted returns, such as MIT Sloan School’s MS de- gree in finance and Columbia University Teachers College Summer Principals Academy. A nonprofit organization called 13th Avenue Funding organizes community leaders to provide

ISAs for educational expenses to local, socioeconomically disadvantaged students. Purdue

University is exploring whether to offer ISAs to their students based on the 13th Avenue business model.8 App Academy provides vocational training for computer programmers.

They accept approximately 5% of applicants for three-month intensive non-degree pro- grams in New York or San Francisco. Accepted students pay no tuition, but they must pay 18% of their first-year gross salary within six months of graduation. App Academy students earn average starting salaries of $105,000 in San Francisco and $89,000 in New

York. 6The full blog post is from May 7, 2014 and can be found at http://blog.upstart.com 7http://www.usnews.com/education/best-colleges/paying-for-college/articles/2012/09/24/college- students-get-a-hand-from-investors 8http://www.purdue.edu/newsroom/releases/2015/Q4/purdue-research-foundation-signs-letter-of- intent-to-explore-college-funding-alternative.html

8 Cumulus Funding sells ISAs with repayment obligations of up to 10% of gross income and term lengths from two to six years. Borrowers have unrestricted use of the funds and Cumulus Funding provides assistance with employment searches. The ISA contract includes a buyout provision, so there is a hard cap on the borrower’s future financial obligations. Thrust Fund is a platform to bring together individual borrowers and ISA investors. The Thrust Fund ISA contracts also include buyout provisions allowing the borrower to terminate the contract for a payment of between three and 10 times the initial investment. Kenneth Michael Merrill is an Oregon-based entrepreneur who sold shares of himself for $1 in a 2008 IPO. The shares, which trade in a secondary market, fetch up to

$18 and trade for around$5 in 2016. Mr. Merrill does not pay dividends, but shareholders own proportional voting rights.

In April 2014, Florida Senator Marco Rubio and Wisconsin Representative Tom Petri intro- duced legislation to clarify regulations for ISA transactions. The Congressional Research

Service (CRS) summarizes the main provisions of the Investing in Student Success Act

(ISSA).9 If two parties agree to an ISA that meets ISSA standards, then the agreement is valid, binding, enforceable, and not subject to state usury laws or state laws regulating sales of future income. The maximum ISA repayment period is 30 years and the maxi- mum payment is 15% of annual income above $10,000. All ISAs must specify terms and conditions for early termination. All ISAs must list the specific sources of revenue that are included in official income and the sources which are excluded. Side payments are explicitly prohibited.

The ISSA includes many provisions that address anti-slavery concerns. The ISA firm must provide the seller with a document specifying the terms of the agreement. These terms must explicitly state that the ISA is an equity agreement and not a debt agreement. The

9The text can be found at https://www.congress.gov/bill/114th-congress/senate-bill/2186/text

9 terms must also clearly express that, like student loans, obligations cannot be discharged in bankruptcy. Because the ISA borrower is essentially agreeing to higher future tax rates, future repayments are tax deductible. The ISA investors cannot coerce or harass borrowers about future employment choices. Investors are not allowed to coerce or harass investees.

As of March 2016, neither the ISSA nor any similar legislation is in effect and there is no registered organization lobbying for its passage. The public is largely unfamiliar with ISAs and related arrangements. If ISA legislation passes, financial institutions that originate

ISAs could market other financial products, such as annuities or life insurance, to the

ISA borrower. The ISA lender could utilize information from standardized test scores, admissions committee decisions, previous employers, and credit ratings agencies. It is hard to evade income-based repayment obligations if the IRS assists in enforcing the contract.

Until ISA legislation passes, the Fantex ABI transactions provide the best environment to evaluate the market for human equity capital.

3. Background on Fantex and Acquired Brand Income Tracking Stocks

Fantex Holdings was incorporated in Delaware in April 2012 with headquarters in San

Francisco. Fantex Holdings is the parent company of Fantex Brands, a subsidiary incor- porated in Delaware in September 2012. Fantex Brands negotiates and executes contracts with professional and amateur athletes. These contracts pay the athletes a lump sum today in exchange for a fixed percentage of all future ABI, contingent on financing. Once Fantex secures financing, there is an IPO for the athlete’s individual “tracking stock.” After the

IPO, the tracking stocks trade on an exchange operated by a wholly-owned subsidiary of an affiliated broker-dealer, Fantex Brokerage Services, LLC.10

10Throughout the paper, I refer to the entire organization as “Fantex”

10 Table 1 summarizes the terms of the Fantex ABI contracts described in the Fantex 2015

10-K report, filed on March 10, 2016. Fantex signed its first brand contract with Arian

Foster, a running back for the Houston Texans, on February 28, 2013. The contract stated that Fantex would make a gross lump sum payment of $10.53 million to Mr. Foster in exchange for 20% of his future ABI. The actual payment would be only $10 million because Fantex withholds 5% of the gross amount as an underwriting fee. Shortly after the contract was signed, Mr. Foster seriously injured his back. Fantex was unable to secure

financing under the contract terms and Mr. Foster was unwilling to renegotiate terms to appeal to investors. After more than two and a half years of failed negotiations, the two parties agreed to terminate the brand contract.

Between October 2013 and July 2015, Fantex successfully negotiated and financed six IPOs for ABI contract tracking stocks. All six of the contract parties were current NFL players on their IPO dates. The first successful IPO closed on April 28, 2014. The agreement paid

Vernon Davis a lump sum of $4 million dollars net of underwriting fees in exchange for 10% of all ABI after the effective date of the contract. The percentage of ABI that Mr. Davis earned between the effective date and IPO date, including earnings from approximately half of an NFL season, were deducted from the lump sum payment he received on the IPO date.

Unlike Mr. Davis, who was playing in his ninth NFL season when he signed his ABI contract, the next five NFL players completed their IPOs during their second, third, or fourth NFL seasons. Players at this stage of their NFL careers may be very motivated sellers of ABI contracts because their future income streams are incredibly volatile. Players accrue a season of experience for each year they are on an active roster, inactive roster, or injured reserve list for at least six NFL games. A player who accrues three seasons of experience becomes a restricted free agent (RFA). An RFA who is not under contract can

11 receive new contract offers from other teams, but the RFA’s current team can sign the player by matching the other team’s offer. A player who completes four or more seasons of experience becomes an unrestricted free agent (UFA). A UFA who is not under contract can sign with any other team even if the UFA’s current team matches or exceeds the other team’s offer. A player with two to four years of NFL experience could soon earn an eight

figure guaranteed contract in free agency or never earn another dollar in the NFL because of injuries or poor performance.

The Fantex ABI contracts do not require the athlete to maximize brand income. The athlete does not have fiduciary responsibility to Fantex or any individual ABI investors.

Such language is necessary to mitigate concerns about slavery and coercion. The ABI contracts stipulate that the player cannot pay more than 15% of income to agents and other advisors, ensuring that athletes retain a large majority of claims to their future earnings. Some of the athletes who signed brand contracts took subsequent actions that failed to maximize the contract value. For instance, Vernon Davis did not participate in an optional 2014 offseason training program with the , which cost him a $200,000 workout bonus. Later in 2014, Mr. Davis failed to attend a mandatory training program and was fined $70,000. These two decisions cost his shareholders $27,000.

However, since Mr. Davis still keeps the vast majority of his future football earnings, he is still highly motivated to maximize those earnings. Mr. Davis may have chosen to forego bonuses or pay fines to rest an injury or to gain bargaining power in future contract negotiations.

Table 2 lists the sources of revenue that the player is obligated to pay to his investors

The most obvious source of included revenue is compensation as a professional athlete.

Depending on the terms of the player’s contract, these payments include a salary and a variety of bonuses. Player contracts include a signing bonus and other bonuses associated

12 with player or team performance. Since the total salaries that each NFL team can pay of all their players each season is subject to a fixed salary cap, it is common for teams and players to restructure contracts. Players who agree to restructure their contracts typically earn a special restructuring bonus.

Other sources of income are included or excluded depending on whether the earnings are directly related the seller’s occupation as a professional athlete. Because professional athletes are public figures, endorsement income is included. Endorsement income can be cash, merchandise, equity, or other investment opportunities. The full value of any cash payments are included. Merchandise payments are included and valued at the prevailing retail price, however players can deduct up to $40,000 in merchandise income. Players who receive equity stakes as part of an endorsement opportunity must provide Fantex the opportunity to invest in a pro-rata share. For instance, Vernon Davis was allowed to invest in a Jamba Juice franchise as part of an endorsement contract, and Fantex purchased a 10% stake in the investment.11. Sometimes players make appearances in exchange for charitable donations rather than cash. These donations are included as brand income.

After the player’s athletic career ends, income from endorsements or acting is included, since this income is attributable to the player’s persona as a former professional athlete.

Income from employment directly related to the player’s athletic expertise is also included.

Examples of included post-career employment include coaching at the professional or college level, broadcasting, and operating of a sports camp or athletic training business. ABI contract payments are proportional to gross earnings. Taxes for payroll, Medicare, FICA, retirement contributions, and payments to agents or other advisors are not deductible.

Under the NFL’s Collective Bargaining Agreement (CBA), players can defer some portion

11The press release can be downloaded from https://fantexbrands.com/about-fantex-inc/investor- relations/news-details/2015/Fantex-Inc-Acquires-Ownership-Interest-In-Additional-Jamba-Juice- Franchise-On-Behalf-Of-Fantex-Series-Vernon-Davis-Stockholders/default.aspx

13 of their income, but they cannot delay the associated ABI contract payment. The regular pension benefits that players earn after three credited seasons are excluded from ABI income.

Since the ABI contract functions as an income tax and not as a wealth tax, certain revenue is specifically excluded. Income from employment unrelated to athletics or to the player’s notoriety as a former athlete is excluded. This includes coaching at any level below college.

Although former athletes may be more likely to win political elections or appointments, the salaries for political positions are excluded. Other excluded income includes earnings for di- rectorships, capital gains, or inheritance. Players who are married must have their spouses sign an exhibit to the contract, and players cannot use the status of marital community property to evade payment. Finally, ABI contracts allow players to exclude some income sources that the ABI contracts would normally include. For instance, some players exclude income from endorsements that predate the effective date of the contract. If an athlete spends money to fulfill the ABI contract requirements, such as fees for legal or accounting services or mandatory self-employment taxes, these expenses are deductible.

Fantex initially sest the fair value of the ABI contracts to the gross IPO price inclusive of the underwriting fee. In each subsequent 10-Q or 10-K filing, Fantex updates the fair value of each investment as of the quarter end, but provides limited methodological details.

However, Fantex filed a prospectus with the SEC on November 24, 2015 that discloses more details of its valuation model. Table 3 provides a detailed example of how Fantex estimates the fair value of its ABI contract with Vernon Davis as of 9/30/2015. The ABI contract with Mr. Davis entitles Fantex to 10% of the ABI that Mr. Davis earns from included contracts. Mr. Davis’ ABI contract includes five categories of included revenue: current NFL contract, current included endorsements, projected NFL contracts, projected endorsements, and projected post-career income. Panel A of Table 3 shows the Fantex

14 estimates of the gross and net cash flows for these line items. Fantex does not provide any details about how they estimate endorsements or post-career income. However, Fantex discloses their method for estimating projected NFL contracts in extensive detail.

Fantex estimates a player’s future NFL contract earnings in a three-stage procedure. First,

Fantex constructs a sample of comparable players. For Mr. Davis, the sample of compara- ble players includes 40 retired tight ends who had 30 or more catches in both their eighth and ninth NFL seasons. Fantex uses statistics including receptions, yards, ,

All-Pro selections, and draft round, to construct a proprietary similarity score between

Mr. Davis and each player in the sample. Panel B of Table 3 lists the six players most similar to Mr. Davis, the length of their careers, and each player’s weighting in the simi- larity score calculation. In the second stage, Fantex estimates the expected length of Mr.

Davis’ career as the weighted average career length of these six comparable players (12.06), rounded up to the next integer (13). Since Mr. Davis’ contract at the time of filing ran through his tenth season, Fantex estimates that he will play for three additional seasons.

I have two concerns with this stage of the estimation. First, rounding the estimate of Mr.

Davis remaining career up to the next integer biases the estimate upwards. Second, only two data points contribute about 50% of the estimate of Mr. Davis’ career length. Even if these two players have very similar statistics to Mr. Davis, other factors such as injuries, the willingness of a veteran player to switch teams, and the willingness of a veteran player to play for a rebuilding team influence individual career outcomes.

In the third stage, Fantex estimates Mr. Davis’ expected earnings in the final three years of his career. Fantex constructs a sample of 14 tight ends who signed NFL contracts between season 10 and season 12 of their careers and selects the four most similar players.

For each player, they calculate the average annual compensation in their new contracts.

Fantex compounds this average compensation to the present at an annual rate of 5.58%,

15 the average increase in the NFL salary cap from 2000-2014. Fantex uses the similarity score to calculate the weighted average inflation-adjusted salary. Fantex allocates the inflation- adjusted salary to a signing bonus and three annual salaries according to the average proportions for three-year NFL contracts. Panel C of Table 3 presents the comparable players, average inflation-adjusted salaries, and similarity weights. The contracts of two players determine 68% of the estimate for Mr. Davis’ future contract compensation, which could magnify errors driven by unobservable characteristics that affect the compensation of individual veteran players. Also, Matvos (2009) shows that most NFL contracts pay higher salaries in the later years. Since NFL contracts are not guaranteed, teams are especially likely to terminate the contract of a veteran whose skills decline faster than expected.

Fantex completed six ABI IPOs with individual NFL players at the time of their prospec- tus filing. Among these contracts, approximately 91% of the total fair value come from current and future NFL contract salaries. Fantex discounts the guaranteed portion of the player’s existing contract at 4.5% and the unguaranteed portion at 7.5%. Fantex discounts estimated income from future contracts at rates between 11.4% and 16.2%. Fantex uses higher rates to discount future contracts for players who have less NFL experience. Fantex discounts current endorsements at 10%, future expected endorsements at rates between

10% and 16.5%, and expected post-career earnings at rates between 15% and 20%.

The estimated fair value of the ABI contracts changes because of the passage of time, changes in expected cash flows, or changes in the discount rates. Players could earn lower than expected future income due to poor performance, injuries, criminal or scandalous behavior off the field, death, strikes, or lower future salary cap growth. An athlete could acquire debt that is senior to his ABI contract payment obligations. ABI contracts are un- secured and investors could incur significant costs in trying to collect any missed payments.

Fantex can use income from one player’s contract to settle another player’s liabilities. In-

16 vesting in any individual player’s IPO ultimately represents taking an equity stake in Fantex itself. Fantex has approximately 40 full-time employees and incurs significant operating costs.

Fantex filed for an IPO of Sports Portfolio I units on November 24, 2015. These financial assets would consist of proportional ownership in the six funded NFL player’s ABI tracking stocks as well as the ABI contracts of Kendall Wright, Andrew Heaney, Terrance Williams, and Ryan Shazier that would be funded by the IPO with any remaining proceeds allocated to funding Fantex working capital. The contract with Mr. Heaney was notable because he was the first player in a sport outside the NFL. Mr. Heaney, a Major League Baseball

(MLB) pitcher for the Los Angeles Angels of Anaheim, agreed to receive a lump sum of

$3.06 million, net of underwriting fees, in exchange for 10% of his future ABI. Mr. Heaney’s deal was funded in February 2016. When a player earns brand income and pays the agreed percentage to Fantex, 95% of the payment is allocated to the associated tracking stock and 5% is allocated to the common platform stock. Since Fantex Holdings is the parent of Fantex Brands, Fantex is required to employ a qualified independent underwriter for the IPO. UBS served as the qualified independent underwriter in the IPO. After the IPO closes, Fantex Holdings will retain 98.6% of voting rights.

Fantex announced that they entered into agreements with ten new athletes on April 27,

2016. , a with the Jacksonville Jaguars, agreed to sell 12% of his future brand income for $4.6 million. Fantex agreed to deals with five new MLB players: Tyler Duffey, Maikel Franco, Collin McHugh, Jonathan Schoop, and Yangervis

Solarte. The MLB players agreed to sell 10-11% of future brand income for lump sums of $2.23-$4.91 million. Finally, Fantex signed four professional golfers: Kelly Kraft, Scott

Langley, Jack Maguire, and Kyle Reifers. The golfers agreed to sell 11-15% of future brand income for lump sums of $1.74-$3.06 million. Fantex also announced that they would

17 withdraw their IPO registration statement for the Sports Portfolio I units.12

Fantex successfully completed a private placement of the Sports Portfolio I units consisting of earnings from all 20 athletes with signed brand agreements. UBS acted as the intermedi- ary in the private placement. Fantex announced the deal on July 25, 2016. The remainder of this paper will focus on the ten completed IPOs with NFL athletes.

4. Estimating Realized and Expected Human Capital Equity Returns

I focus on estimating gross returns before fees because Fantex is a start-up company which spends around $4 million annually in operating expenses. These expenses are large relative to the expected income from its portfolio of brand contracts. In addition, Fantex ultimately abandoned their strategy of operating a secondary market in the individual athlete tracking stocks. Therefore, my analysis focuses on estimating the realized and expected returns based on the IPO price in these transactions.

4.1 Realized Returns

Table 4 provides estimates of realized gross returns from investing in the portfolio of ABI contracts. These estimates use data from Fantex 10-Q and 10-K filings to calculate the return of an investor who fully participates in every completed IPO. I calculate returns as of the end of each calendar quarter. For each completed IPO, I assume the investor makes the IPO payment at the start of the calendar quarter. This payment is net of the 5% withholding for underwriting and is the actual amount Fantex pays the player within five business days of the IPO date. At the end of each calendar quarter, I estimate the value of the investor’s portfolio, which consists of three components. The first component is the brand income payments from all completed IPOs received during the calendar quarter,

12https://www.businesswire.com/news/home/20160427006579/en/Fantex-Announces-Brand- Agreements-Ten-Athletes-Professional

18 which is the sum of cash receipts from brand contracts and changes in accounts receivable.

Cash receipts from brand contracts come from the cash flow statement and accounts receiv- able comes from the balance sheet. The second component is the quarterly interest earned on the investor’s cash holdings at the start of the quarter. I assume the investor earns interest at the weighted average discount rate for the Fantex portfolio of ABI contracts.

This assumption is consistent with dividend reinvestment. The discount rate ranges from

12.4% to 14.5% in the time series. The third component of the investor’s portfolio is the estimated fair value of the brand contracts. Fantex provides updated estimates of fair value in each quarterly balance sheet. See Table 3 for an example of how Fantex estimates the fair value of an ABI contract.

I calculate the investor’s quarterly return using the portfolio values at the end of each quarter. I also compound these realized returns over the time series, and annualize each quarter’s cumulative return. An investor who participates proportionally in the six com- pleted IPOs earns an annualized return of -13.26% over the seven quarters from April 1,

2014 to December 31, 2015. However, these results are very sensitive to the performance of any individual offering. For instance, the -24.05% quarterly return in the fourth quar- ter of 2014 comes primarily from the poor performance of a single player, E.J. Manuel.

Mr. Manuel was replaced as starting of the Buffalo Bills. As a result, the discounted value of both his future expected NFL contracts and endorsement income is substantially lower going forward.

Table 5 presents estimated realized returns for each individual IPO as of December 31,

2015. The returns for the first two completed IPOs are abysmal. Investing in the Vernon

Davis IPO yields a total return of -34.4% and investing in the EJ Manuel IPO yields a truly horrible -76.7% return. EJ Manuel has high expected earnings when he completes his IPO because he plays quarterback, the position with the highest average salary, and

19 his selection in the first round of the draft. In the four most recent completed IPOs, the average return is +12.5% for an average holding period of about nine months. It’s possible that Fantex is better at marketing, valuation, or due diligence and are now better at pricing these securities. It’s also possible that Fantex deliberately negotiated high prices for the

first IPOs to gain publicity. The Fantex IPO of Arian Foster’s ABI contract never closed.

However, investor’s realized returns would have been very poor. The terms of this deal provide Mr. Foster with $10 million in exchange for 20% of all brand income on or after

February 28, 2013. In the 2013, 2014, and 2015 season, Mr. Foster earned a total of just under $18 million. In March 2016, Mr. Foster’s team, the Houston Texans, chose to release him from the final year of his contract. Because of Mr. Foster’s history of serious injuries, including a torn Achilles tendon in October 2015, it is unlikely he will earn the additional compensation that it would’ve required for his IPO to be a profitable investment.

Table 6 summarizes NFL contract terms for NFL players with completed IPOs. The NFL contract earnings data suggest that investors in the Vernon Davis and EJ Manuel IPOs are not likely to recover their investments. Vernon Davis needs to earn more than $40 million for his investors to turn a profit. For the included portion of the 2013 season as well as the complete 2014 and 2015 seasons, he earned a total of about $14 million. Mr. Davis signed a 1-year contract with the Washington Redskins for the 2016 season, but the compensation terms were not disclosed. However, his statistics declined substantially in 2014 and 2015 and it is unlikely he will earn a lucrative multi-year guaranteed contract at this stage in his career. EJ Manuel must earn about $50 million for his investors to turn a profit. His guaranteed included ABI for the 2014, 2015, and 2016 seasons total less than $4 million.

The Buffalo Bills hold a team option for the 2017 season. Mr. Manuel lost his starting job in the middle of 2014 and lost his only two starts in 2015. For investors to eventually turn a profit on the IPO, Mr. Manuel will need to remain in the NFL as a backup, earn

20 another opportunity to start, and then play out a typical starting quarterback’s contract for two or three seasons.

The other four IPO investments are more promising. Mohammed Sanu signed a five-year contract with ATL in March 2016. Sanu’s compensation for 2016 and 2017 are guaranteed, and his compensation for 2018, 2019, and 2020 are not guaranteed. CHI offered Alshon

Jeffery a non-exclusive franchise tag for 2016. If Jeffery signs another team’s offer sheet,

CHI can match the contract. If CHI doesn’t match the contract, they receive two extra

first-round picks. If Jeffery doesn’t sign another team’s offer sheet, Jeffery and CHI can negotiate a multi-year deal, or Jeffery can play one year for CHI for 14.6 guaranteed and will be an unrestricted free agent in 2017. STL exercised a team option and Michael Brockers

2016 compensation is guaranteed. Brockers will be an unrestricted free agent in 2017. Jack

Mewhort’s compensation in 2016 and 2017 is not guaranteed. After 2017, Mewhort will be an unrestricted free agent.

In the next section, I estimate the expected return of the brand contract IPOs. Because there are extensive data available for both player performance and player contracts, it is feasible to forecast expected future earnings for active NFL players.

4.2. Expected Returns

I estimate expected returns by finding an internal rate of return (IRR) for each ABI contract that equates the lump sum payment with the sum of discounted future cash flows. The IPO completion date and lump sum payment amount are observable for the completed IPOs.

The expected future cash flows of the contract depend on the probabilities of employment in the NFL in subsequent years and expected compensation conditional on employment:

Pt→∞ E[CFi,t] Pt→∞ E[P r(1i,t)]E[CFi,t|1i,t] CFi,t=IPO = t = t t=IPO+1 (1+IRRi) t=IPO+1 (1+IRRi)

21 The above decomposition assumes that the probability of player i surviving for t more seasons [P r(1i,t)] and the player’s cash compensation in year t conditional on survival

(CFi,t|1i,t) are uncorrelated.

There is an NFL draft during each offseason where teams select players for the upcoming season. NFL teams have access to statistics and game film for all the games. NFL teams employ scouting departments to collect and evaluate statistics, watch game film, and otherwise evaluate potential draft picks. There is an annual six-day NFL combine where eligible players perform position-specific skills exercises and standardized strength and speed tests. In addition, major college football programs have special Pro-

Days, when NFL teams send scouts to campus to watch workouts and interview potential draftees. For these reasons, a player’s draft position should provide a good measure of his

NFL potential relative to other players in the same draft.

Table 7 charts average career outcomes for all players drafted in the first seven rounds of any NFL draft between 1990 and 2015. The NFL draft has twelve rounds until 1992, eight rounds in 1993, and seven rounds from 1994 to 2015. The outcomes measured in Table

7 are useful because they are comparable for players with different positions. Football teams consist of players with a variety of specialized skills. Many popular statistics, such as touchdowns and tackles, are only relevant for players at certain positions. The columns

P1 . . . P10 show the probabilities of eligible players surviving at least 1 . . . 10 NFL seasons.

If there are only three completed NFL seasons since a player’s draft, the player’s career is only considered in the calculations of P1, P2, and P3. AP1 is the average number of

First-Team NFL All-Pro Selections. This honor goes to one NFL player per position per season. PB is the number of Selections. This honor goes to about 3.5 players per position per season. The number of selections depends on whether replacement selections are required to fill the roster. St is the number of seasons that a player is the primary

22 starter at his position for his team. G is the number of games the player appears in during his career. All statistics are from pro-football-reference.com. All fourteen of these position- independent outcomes are almost perfectly monotonically decreasing with increasing draft round.

I use a probit model to estimate the probability that a player survives one additional year in the National Football League. I estimate the model separately for nine position groups: quarterback (QB), offensive linemen (OL), running backs or fullbacks (RB/FB), wide receivers (WR), tight ends (TE), defensive linemen (DL), linebackers (LB), defensive backs (DB), and kickers or punters (K/P). It is common for players to switch positions within a position group, but very rare for players to switch positions across position groups.

I divide players into position groups because it is conceivable that position groups influence a player’s conditional survival rate. Position groups require various profiles of size, speed, strength, durability, and mental toughness. Table 8 summarizes the NFL career length distribution by position group. The sample consists of all players selected in the first seven rounds of all NFL drafts between 1990 and 2014 who are not in the NFL as of the start of the 2015 season. It is clear that players in different position groups have different survival rates over their careers. For example, kickers and punters who are drafted are especially likely to have long careers. Kickers and punters have minimal strength and speed requirements. They do not frequently tackle players and are not frequently tackled by other players. Kickers and punters who prove their mental toughness are likely to have long careers.

Table 9 presents results of NFL survival rate probit regressions for all players drafted in the

first seven rounds of any NFL draft between 1990 and 2015. I estimate the probit model separately for each of the nine position groups. For each player in the sample, I construct data on their survival in the NFL. The dependent variable is a binary survival indicator.

23 A player-season observation qualifies for the sample if the player is selected in the first seven rounds of the previous draft or is credited with a year of experience in the previous season. The independent variables include the player’s draft round and years of experience.

Draft order should have a negative relation to survival rates because of the extensive and publicly available body of work available for each player at the time he is drafted. Players drafted in earlier (lower) rounds should have a greater marginal one-year survival rate at any point in their careers. Experience is measured as the number of NFL seasons since the player is drafted. Experience should have a positive relation to the one-year survival rate for at least two reasons. First, with each year of football experience, a player’s physical condition deteriorates relative to that of an incoming rookie. Second, the minimum NFL salary under the CBA increases with tenure, which makes replacing the veteran with a rookie more economically attractive.

Consider a player who was drafted in 2010 and played the 2010, 2011, and 2012 seasons before retiring. The player’s career would be represented by four observations: three successes in years 1, 2, and 3, and one failure in year 4. Using these data, I estimate separate probit models for each position group, using Rogers (1994) standard errors clustered by draft year. A player’s conditional one-year survival probability is Φ(βx) where Φ is the normal CDF operator, beta is a vector of coefficient estimates from the probit, and x includes an intercept, the player’s draft round, and experience. For instance, an offensive lineman who is drafted in the second round and has four years of NFL experience has an estimated probability of Φ(2.14 − 0.14 ∗ 2 − 0.10 ∗ 5) = Φ(1.36) = 0.913 of surviving for a fifth season. All statistics on draft round and NFL experience are from pro-football- reference.com.

For each of the nine position groups, the intercept is positive and statistically significant.

This means that a theoretical player drafted in the first round has a probability much

24 higher than 50% ($Phi(0)) of surviving his rookie season. Table 7 shows that the average player drafted in the first seven rounds has an 89.3% survival rate for the first year, rising to 99.5% for the average player selected in round 1. As expected, the coefficient on draft round and experience are both negative and highly significant for all position groups. The coefficient estimate on draft round is smallest in magnitude for kickers and punters, who experience minimal wear and tear relative to other position players. Kickers and punters with the mental and psychological toughness to play in the NFL can play well past the age of 40. Thus, there is relatively low demand for rookie kickers and punters. Teams rarely draft kickers or punters early, and can often fill either roster spot with an undrafted free agent. The coefficient estimate on draft round is largest in magnitude for wide receivers and defensive backs. These two positions are the most dependent on running speed. Because speed is a congenital trait, teams may use early draft rounds to select the fastest players at these positions. The coefficient estimate on experience is negative and significant for all position groups. The probability of surviving one additional year in the league decreases with each additional year of experience. The coefficient estimate is particularly small in magnitude for kickers and punters because of the limited speed and strength requirements and the lack of physical punishment. The coefficient estimate is also particularly small for , likely due to the incredibly high mental requirements of playing the position.

Next, I model the expected annual compensation for a player on an NFL roster using NFL contract data. Table 10 describes the sample of NFL contract data from 2000-2015. The contract data are from Spotrac 13. If a player has contracts with two or more different teams in the same NFL season, I consolidate the contracts into a single entry. For each player- year observation, I estimate the contract compensation. If the player’s cash compensation

13www.spotrac.com

25 for that season is available, I use that number. If the cash compensation is not available, but the player has a contract in the data which includes the particular year, I estimate the player-year compensation as the yearly average of the total salary and guaranteed bonus in the contract. I only include players who were drafted in the first seven rounds of an NFL draft between 1990 and 2015 and who are playing in their first 10 NFL seasons. There must be an exact match between the player’s first and last names in the Spotrac data and in the historical NFL draft data, which are from www.pro-football-reference.com. The average, median, minimum, and maximum salaries are quoted as percentages of the NFL salary cap. The salary cap is the amount each team is allowed to pay its active roster of

53 players, so the average compensation of all players is approximately 1.89. The NFL salary cap data are from www.nfl.com and www.nfllabor.com and are quoted in millions of dollars.

Table 11 shows results of the OLS compensation regressions for the same model for each of the nine position groups. The dependent variable is a player’s cash compensation for a particular NFL season, as a percentage of that season’s NFL team salary cap.14 Scaling by team salary cap accounts for a time trend in player salaries due to the salary cap increase from around $62 million in 2000 to around $143 million in 2015. Each NFL team is allowed 53 players on its active roster each season, so the expected average value for the dependent variable is constant over time at 1.89. The independent variables are an intercept, Draft Round, measured as the round when the player is initially drafted, and

Experience, measured as the number of NFL seasons since the player is drafted.15 Standard errors are clustered by draft year as in Rogers (1994).

Unsurprisingly, a player’s draft round is not only related to lower probability of future

14Salary cap data are from either www.nfllabor.com or www.nfl.com. 15All statistics on draft round and NFL experience are from pro-football-reference.com

26 employment, but also lower expected compensation conditional on employment. This is true for all players, although the magnitude of the coefficient estimate is especially large for quarterbacks and especially small for kickers and punters. The quarterback has more responsibility for the game’s outcome than any other player and a true franchise quarter- back is the scarcest resource in the NFL. On the other hand, kickers and punters are the easiest players to replace and teams often find adequate replacements among undrafted free agents. The coefficient estimate on draft round is very similar for the other position groups. The coefficient estimate on experience is positive and significant for every posi- tion group. The combined effects of mandatory increases to the minimum salary based on tenure and the loss of marginal players from the league each season outweighs the effect of physical deterioration among those players who remain employed. The coefficient estimate is particularly small in magnitude for kickers and punters. Their salaries do not increase as much as the salaries of other players over time. They are simply too easy to replace.

I use the model estimates of one-year survival probabilities and expected compensation conditional on survival to construct a stream of expected future NFL earnings for each of the six players with completed IPOs. Because the initial cash flow is of the opposite sign as all future cash flows, there is a single IRR which sets the net present value (NPV) of all current and future cash flows equal to zero. Table 12 shows the IRRs for each of the six completed IPOs. A few other modeling assumptions are necessary to calculate IRR. First of all, future NFL earnings are not the only revenue included in the ABI contract. The contract specifies that the player owe a fixed percentage of all future “brand income,” which includes other sources of income besides future NFL earnings. The IPO price and required repayment percentage of future ABI imply a total present value of total ABI payments.

To estimate the implied present value of future NFL earnings, I reduce the implied present value of ABI payments by the estimated present value of non-NFL earnings from the

27 earliest published Fantex 10-K or 10-Q SEC filing following the completed IPO. Based on these data, the present value of future NFL earnings makes up 91% of the present value of all sources of included revenue. Also, I assume players will survive for a maximum of 20 additional NFL seasons. Out of more than 1500 players drafted between 1990 and 1996, only 4 had NFL careers lasting 20 seasons or longer. Since my estimate of expected NFL earnings is scaled by salary cap, I have to make an assumption about future salary cap growth. I estimate the cap will grow at a rate of 5% annually from a base of $133 million in 2014. Finally, I assume there are annual cash payments. The estimated t=0 cash flow is the lump sum payment net of payments the player owes for ABI earned between the effective date and the IPO date. I assume the NFL season spans from September 1 through

December 31. If the effective date falls in the middle of the season, I prorate the initial cash flow accordingly.

For the six completed IPOs, the IRR estimates range from -8.2% to 15.3% with an average of 3.9%. These are gross returns; an actual investor’s returns would be lower because of administrative costs and default risk. Given the substantial uncertainty in career outcomes for an individual NFL player, the equilibrium price for human capital equity contracts is high if buyers view the contracts purely as investments. If investors require a higher rate of return for an investment with similar risk, then some portion of the ABI contract purchases are really consumption rather than investment. With only six data points, it is unrealistic to explain any statistically significant variation in expected returns. However, there is a strong negative correlation between expected returns and the market size of the player’s team on the effective date. The two ABI contracts with negative expected returns belong to players for teams in two of the largest US markets: Chicago and the San Francisco Bay

Area. The contracts with positive expected returns belong to players for teams in much smaller markets: St. Louis, Indianapolis, Cincinnati, and Buffalo. Players on big-market

28 teams may have more fans willing to purchase equity in their future income regardless of the price of the contract. This fan buying pushes up the equilibrium price and lowers the expected return.

I also calculate the IRR with the current cash flow equal to the implied present value of total earnings based on the IPO price. The IRR from this analysis is a lower bound. It is the expected rate of return investors should expect to earn if the only player only earns included income from NFL contracts. If the player earns any income from endorsements or post-career employment, this would increase the IRR. Using this calculation, the minimum

IRR from the ABI contracts ranges from -10.6% to 11.3% with an average of 2.3%. There is still a strong negative relation between minimum ABI investment returns and the size of the player’s home market. I also calculate the value-weighted average of expected returns. I weight each expected return by the implied present value of total earnings. In the expected

IRR estimate, the total earnings includes all included revenue sources under the terms of the ABI agreement. In the minimum IRR estimate, the total earnings includes only the present value of current and future NFL contracts. In both cases, the value-weighted average expected return is roughly two percent lower than the equal-weighted average expected return. This reflects another size effect. On average, smaller deals offer more attractive expected returns to investors.

Finally, I estimate expected future compensation using an alternative specification. The previous estimation techinque uses two distinct models for expected survival probabilities and expected compensation conditional on survival. Modeling expected compensation as the product of expected survival probability and expected compensation conditional on survival requires assuming that the residual terms in these estimates are uncorrelated. To address this, I estimate expected compensation using a one-stage tobit regression with data specific to each player. For each player, the specific sample includes all players in the same

29 position group with at least the same amount of experience as the contract player as of the effective date. For instance, Vernon Davis is in his eighth NFL season on the effective date of his contract. The player-specific sample for Mr. Davis consists of all contract data for tight ends in their eighth season or later. I construct data for the total compensation each player earns through their 20th season. If a plays 11 seasons and retires before

2015, he contributes 13 data points to the sample. Four of these data points represent the compensation he earns in his eighth, ninth, tenth, and eleventh seasons, and nine data points represent the zero compensation he earns in seasons 12 through 20. If another tight end has nine seasons of experience and is still active in 2015, the player only contributes two data points for earnings in his eighth and ninth seasons.

Table 13 shows the statistics from the player-specific compensation Tobit regressions. For all players, the coefficient estimate on draft round is negative and signficant. This is expected since draft round is negatively related to both survival probability and expected compensation conditional on survival. The coefficient estimate on experience is negative and significant for all players. When estimated separately, experience relates negatively to survival probability and positively to expected compensation conditional on survival.

The negative coefficients on experience reflect the fact that players exit the league at a faster rate than wages increase for those players who remain. The player-specific Tobit regressions describe a fair amount variation in future compensation, with pseudo-R2 values ranging from 0.1 to 0.27. When the same model is estimated using OLS, the R2 values are generally around 20% to 30% smaller in magnitude. The pseudo-R2 value is much smaller for quarterbacks than for the other samples. This reflects the fact that quarterbacks are far more likely than other players to earn especially high compensation.

Next, I will use the coefficient estimates from the Tobit model to estimate expected returns for the contract parties.

30 To be continued . . .

5. Conclusion

This paper evaluates large transactions in human capital equity contracts between private parties. Because the details of the transactions are publicly available and there are histori- cal data on both performance and compensation, it is feasible to estimate expected returns.

I find that the average expected return on the completed human capital equity IPOs is

3.9%. This expected return does not account for administrative costs, default risk, and taxes; an investor’s actual return would be lower as a result. Cross-sectional variation in expected returns suggests a return premium to investing in players on small-market teams, similar to the size effect in traditional equities.

It is also feasible to use NFL compensation data to estimate returns for asset-backed se- curities representing claims to future cash flows for portfolios of athletes. For instance,

firms could structure and price an asset-backed security representing claims to 1% of NFL compensation for all active players drafted in the third round with three years of experi- ence over the next five seasons. If the ISSA or similar legislation passes, human capital equity transactions may soon become commonplace among students and workers in all occupations, not just professional sports.

31 6. References Banz, Rolf W. “The relationship between return and market value of common stocks.” Journal of Financial Economics 9.1 (1981): 3-18. Becker, Gary S. Investment in Human Capital: A Theoretical Analysis. Journal of Political Economy 70.5 (1962): 949. Caetano, Gregorio Silva, Miguel Palacios, and Harry A. Patrinos. “Measuring aversion to debt: An experiment among student loan candidates.” World Bank Policy Research Working Paper Series, Vol (2011). Campbell, Ritchie A. “The demand for life insurance: An application of the economics of uncertainty.” The Journal of Finance 35.5 (1980): 1155-1172. Cameron, Stephen V., and James J. Heckman. ”The dynamics of educational attainment for black, hispanic, and white males.” Journal of political Economy 109.3 (2001): 455- 499. Card, David. “Estimating the return to schooling: Progress on some persistent econometric problems.” Econometrica 69.5 (2001): 1127-1160. Chapman, Bruce. “Conceptual issues and the Australian experience with income contin- gent charges for higher education.” The Economic Journal 107.442 (1997): 738-751. Di Giovanni, Julian, and Akito Matsumoto. ”The value of human capital wealth.” Global COE Hi-Stat Discussion Paper Series 174 (2011). Friedman, Milton. The role of government in education. Rutgers University Press, 1955. Gould, Eric D., Omer Moav, and Bruce A. Weinberg. ”Precautionary demand for edu- cation, inequality, and technological progress.” Journal of Economic Growth 6.4 (2001): 285-315. Griliches, Zvi. ”Estimating the returns to schooling: Some econometric problems.” Econo- metrica: Journal of the Econometric Society (1977): 1-22. Jacobs, Bas, and Sweder JG van Wijnbergen. ”Capital-market failure, adverse selection, and equity financing of higher education.” FinanzArchiv/Public Finance Analysis (2007): 1-32. Jorgenson, Dale, and Barbara M. Fraumeni. ”The accumulation of human and nonhu- man capital, 1948-84.” The measurement of saving, investment, and wealth. University of Chicago Press, 1989, 1989. 227-286. Matvos, Gregor. ”Renegotiation Design: Evidence from NFL roster bonuses.” Unpublished working paper (2009).

32 Mincer, Jacob A. ”Age and Experience Profiles of earnings.” Schooling, experience, and earnings. NBER, 1974. 64-82. Palacios, Miguel. ”Human Capital Contracts: Equity like Instruments for Financing Higher Education.” Policy Analysis 462 (2002): 1-12. Rogers, William. ”Regression standard errors in clustered samples.” Stata technical bul- letin 3.13 (1994). Roll, Richard. ”A critique of the asset pricing theory’s tests Part I: On past and potential testability of the theory.” Journal of financial economics 4.2 (1977): 129-176. Schultz, Theodore W. ”Investment in human capital.” The American Economic Review (1961): 1-17.

33 Table 1: Acquired Brand Income (ABI) Contract Transactions The Acquired Brand Income (ABI) contract stipulates that the contract party pay Fantex Holdings a fixed percentage of all pre-tax “brand income” earned on or after the Effective Date. Brand income includes salary and bonus payments as a professional athlete, endorsement earnings, and earnings from employment, such as coaching and broadcasting, related to the athlete’s sport. See Table 2 for a list of revenue sources which are included in or excluded from acquired brand income. The IPO Date is the date when Fantex secures funding and the transaction closes. Proceeds is the amount of the lump sum payment, as stipulated in the ABI contract. The Proceeds are paid to the athlete within five business days of the IPO date. The Proceeds are 95% of the nominal lump sum in the ABI contract. The remaining 5% are withheld for underwriting services, payable to Fantex Brokerage Services, LLC, an affiliated broker-dealer. Under the terms of the ABI contract, any payments the player owes Fantex for brand income he earns between the effective date and the IPO date are deducted from the lump sum payment. PctInsider is the percentage of shares held by Fantex Brands or its directors on 12/31/2015. PctInsider is not reported for the Andrew Heaney IPO as of 3/10/2016. The IPO for Arian Foster’s tracking stock never closed due to the inability to obtain financing. The two parties mutually agreed to terminate the brand contract on 11/10/2015. The IPOs for Kendall Wright, Terrance Williams, Ryan Shazier, and Scott Langley are active pending financing as of 3/10/2016. Player League Effective Date IPO Date Proceeds %ABI PctInsider Arian Foster NFL 2/28/2013 $10.00M 20% Vernon Davis NFL 10/30/2013 4/28/2014 $4.00M 10% 24.3% EJ Manuel NFL 2/14/2014 7/21/2014 $4.98M 10% 47.7% Mohamed Sanu NFL 5/14/2014 11/3/2014 $1.56M 10% 47.5% Alshon Jeffery NFL 9/7/2014 3/19/2015 $7.94M 13% 47.9% Michael Brockers NFL 10/15/2014 5/29/2015 $3.44M 10% 45.0% Jack Mewhort NFL 2/15/2015 7/14/2015 $2.52M 10% 46.3% Kendall Wright NFL 12/1/2014 $3.13M 10% Andrew Heaney MLB 1/1/2015 2/12/2016 $3.06M 10% N/A Terrance Williams NFL 2/1/2015 $3.11M 10% Ryan Shazier NFL 9/1/2015 $3.11M 10% Scott Langley PGA 10/25/2015 $3.06M 15%

34 Table 2: Included and Excluded Revenue under the ABI Contract Terms The Acquired Brand Income (ABI) contract stipulates that the player pay Fantex Holdings a fixed percentage of all pre-tax “brand income” earned on or after the effective date. Brand income includes salary and bonus payments as a professional athlete, endorsement earnings, and earnings from employment related to the athlete’s sport, such as coaching and broadcasting. The contract is contingent on Fantex Holdings securing sufficient funding. The Proceeds are 95% of the nominal lump sum in the ABI contract. The remaining 5% are withheld for underwriting services, payable to Fantex Brokerage Services, LLC, an affiliated broker-dealer. Under the terms of the ABI contract, any payments the player owes Fantex for brand income he earns between the effective date and the IPO date are deducted from the lump sum payment. This table lists potential revenue sources that are specifically included or excluded from the ABI contract. Included revenue sources which are marked with (*) refer to athlete expenses which cannot be deducted from gross income. Disputes are resolved by good faith negotiations, then by arbitration. Included Revenue Excluded Revenue Professional sports salary Legal fees associated with fulfilling Contract Professional sports bonuses Self-employment taxes Endorsement cash income Per-diem or travel expenses Endorsement equity income $40,000 merchandise income deduction Endorsement merchandise income Income from interest, dividends, and capital gains Appearance fees (cash/donation) Coaching at any level below college Co-investment opportunities Employment unrelated to the Field Acting in TV or film Owned businesses unrelated to the Field Professional coaching Directorships College coaching Political positions Broadcasting Marital community property Owned businesses related to the Field Inheritance Deferred income Pensions Payments to agents or advisors(*) Designated excluded income Retirement contributions(*) Taxes for payroll, Medicare, or FICA (*)

35 Table 3: Estimated Fair Value of the Vernon Davis ABI Contract as of 9/30/2015 At the end of each quarter, Fantex updates the fair values of investments in Acquired Brand Income (ABI) contracts. These contracts entitle Fantex to a fixed percentage of the contract party’s revenue from included contracts. Vernon Davis is an NFL player who signed an ABI contract on 10/30/2013, and the funding for this contract was secured on 4/28/2014. The included contracts in the Vernon Davis ABI contract fall into five broad categories: current NFL contracts, current endorsements, projected NFL contracts, projected endorsements, and projected post-career income from related employment. See Table 2 for a detailed list of included and excluded revenue sources. Panel A shows Fantex estimates of gross and net cash flows associated with these line items. Fantex does not provide information about how they estimate endorsements or post-career income. However, Fantex discloses details of its NFL contract valuation model in a prospectus filed with the SEC on November 24, 2015. They project Vernon Davis’ future NFL contract income in a three-stage procedure. First, Fantex identifies a sample of 40 tight ends with 30 catches per season in their eighth and ninth NFL seasons who have since retired. Fantex uses statistics including receptions, yards, touchdowns, All-Pro selections, and draft round, to identify the six most comparable players. Panel B lists the six players, their total career length, and the weight assigned to each player, based on a proprietary measure of statistical similarity to Vernon Davis. The career length estimate for Vernon Davis is the weighted average of these six career lengths (12.06), rounded up to the next integer (13). Since his current contract runs through his tenth season, Fantex estimates Mr. Davis to sign a three-year contract. They identify 14 tight ends who sign NFL contracts between their tenth and twelfth seasons. Among this sample, they identify the four most comparable players, using their proprietary similarity measure. Panel C lists the four players, and the average annual compensation, adjusted for the 5.58% average increase in NFL salary cap from 2000 to 2014. Fantex allocate the estimated inflation-adjusted salary to a signing bonus and three annual salaries according to the average proportions for three-year NFL contracts in their sample. Panel A: Source Gross Amount Discount Rate Net Amount Current NFL Contract $3,948,053 4.5% $3,903,409 Current Endorsements $40,700 10.0% $39,167 Projected NFL Contracts $23,617,653 12.7% $19,192,764 Projected Endorsements $2,359,300 10.0% $1,952,695 Projected Post-Career $5,500,000 15.0% $2,107,913 Panel B: Comp Player Career Length Player Weight 11 29.36% Ben Coates 10 20.68% Rodney Holman 14 13.34% Wesley Walls 15 12.70% Steve Jordan 13 12.09% Jay Novacek 12 11.82% Panel C: Comp Player Average Adj Comp Player Weight $9,063,802 37.12% Heath Miller $5,298,870 31.30% Jason Witten $9,270,509 19.05% $8,646,925 12.53%

36 Table 4: Estimated Realized ABI Contract Portfolio Returns, March 2014-December 2015 This table provides estimates of gross realized returns for investments in acquired brand income (ABI) contracts. All data used to calculate estimated returns are from Fantex 10-Q and 10-K SEC filings. Gross returns are calculated from the standpoint of an investor who takes the opposite position to the ABI contract party for all cash flows. Quarterly returns are calculated based on the total value of cash and investments at the start and at the end of the quarter. Return calculations assume that the investor participates in 100% of each ABI IPO. See Table 2 for a summary of ABI IPO transactions. Start Q Cash is the total value of cash in the investment portfolio at the end of the previous quarter. Start Q Investments is the total fair value of all investments in ABI tracking stocks. These investments are initially marked to the lump sum payment that the athlete receives on the IPO date. This lump sum is the gross IPO price, less a 5% underwriting fee. In subsequent quarters, these investments are marked to fair value based on Fantex’s proprietary model. See Table 3 for an example of how Fantex estimates the fair value of investments in ABI tracking stocks. At the end of each quarter, the portfolio of investments in ABI tracking stocks generates cash. I estimate this cash by summing the cash receipts from brand contracts (from the statement of cash flows) and the changes in accounts receivable (from the balance sheet). During all subsequent quarters, I assume the cash balance grows at the weighted average discount rate used in Fantex’s proprietary model. This discount rate ranges from 12.4%-14.5% in the time series. Return Q is the growth in the investor’s portfolio of cash and investments from the start to the end of the quarter. Return Cumulative is the growth in the investor’s portfolio of cash and investments from April 1, 2014 until the end of the quarter. Return Annualized is the investor’s cumulative return, converted to an annual rate based on quarterly compounding. 37 Start Q End Q Return Quarter Cash Investments Total Cash Investments Total Q Cumulative Annualized 2014Q2 $0 $4,000,000 $4,000,000 $406,869 $3,768,312 $4,175,181 4.38% 4.38% 18.70% 2014Q3 $406,869 $8,743,312 $9,150,181 $548,947 $8,902,342 $9,451,289 3.29% 7.81% 16.24% 2014Q4 $548,947 $10,462,342 $11,011,289 $1,141,777 $7,221,182 $8,362,959 -24.05% -18.12% -23.39% 2015Q1 $1,141,777 $15,271,982 $16,413,759 $1,418,644 $16,571,702 $17,990,346 9.61% -10.25% -10.25% 2015Q2 $1,418,644 $20,011,702 $21,430,346 $1,607,953 $19,199,133 $20,807,086 -2.91% -12.86% -10.43% 2015Q3 $1,607,953 $21,719,133 $23,327,086 $1,877,211 $21,940,727 $23,817,938 2.10% -11.03% -7.49% 2015Q4 $1,877,211 $21,940,727 $23,817,938 $2,791,348 $17,869,172 $20,660,520 -13.26% -22.82% -13.76% Table 5: Estimated Realized ABI Contract Returns by Player, March 2014-December 2015 This table provides estimates of gross realized returns for the six completed IPOs for acquired brand income (ABI) contracts of NFL players. In each completed transaction, the player receives the Lump Sum on the IPO Date. See Table 1 for details of the completed IPO transactions. Fair Value is the Fantex estimate of the discounted value of future expected revenues for each ABI contract. The Fair Value estimates, taken from the Fantex 2015 10-K filing, are as of 12/31/2015. See Table 2 for a list of included and excluded revenue streams and see Table 3 for an example of how Fantex estimates fair value of an ABI contract. Payments are the total amount the player pays to Fantex between the IPO date and 12/31/2015. WADR is the weighted average discount rate Fantex uses to estimate the fair value of each player’s ABI contract on 12/31/2015. Adj Payments are the value of the payments on 12/31/2015, assuming payments are uniformly distributed between the IPO date and 12/31/2015 and accumulate interest at the WADR. Total Return is the total realized return from the IPO date to 12/31/2015. Player Lump Sum IPO Date Fair Value Payments WADR Adj Payments Total Return Vernon Davis $4,000,000 4/28/2014 $1,036,757 $1,450,810 11.3 $1,589,028 -34.4% EJ Manuel $4,975,000 7/21/2014 $950,303 $188,228 13.7 $206,591 -76.7% Mohamed Sanu $1,560,000 11/3/2014 $1,750,146 $210,345 15.3 $227,964 26.8% Alshon Jeffery $7,940,000 3/19/2015 $7,789,801 $175,679 16.2 $186,133 0.5% Michael Brockers $3,440,000 5/29/2015 $3,138,539 $225,268 14.6 $234,509 -1.9% Jack Mewhort $2,520,000 7/14/2015 $3,092,826 $47,966 15.6 $49,306 24.7% 38 Table 6: Player NFL Contract Compensation since Effective Date This table summarizes NFL contract terms for NFL players with completed IPOs. See Table 1 for a summary of ABI IPO transaction terms and see Table 2 for a list of included and excluded revenue sources. NFL teams are abbreviated as follows: SF (San Francisco 49ers), DEN (), WAS (Washington Redskins), BUF (Buffalo Bills), CIN (), ATL (Atlanta Falcons), CHI (Chicago Bears), STL (St. Louis Rams), and IND (). Cash is the total compensation a player earns in a particular season, in $millions. Asterisks (*) denote seasons where the IPO effective date is midseason and the associated cash value is the total NFL compensation for that season. Vernon Davis signed a 1-year contract with WAS for the 2016 season, but the compensation terms were not disclosed. EJ Manuel has a guaranteed contract with BUF through 2016. BUF holds a team option for 2017. Mohammed Sanu signed a five-year contract with ATL in March 2016. Sanu’s compensation for 2016 and 2017 are guaranteed, and his compensation for 2018, 2019, and 2020 are not guaranteed. CHI offered Alshon Jeffery a non-exclusive franchise tag for 2016. If Jeffery signs another team’s offer sheet, CHI can match the contract. If CHI doesn’t match the contract, they receive two extra first-round picks. If Jeffery doesn’t sign another team’s offer sheet, Jeffery and CHI can negotiate a multi-year deal, or Jeffery can play one year for CHI for 14.6 guaranteed and will be an unrestricted free agent in 2017. STL exercised a team option and Michael Brockers 2016 compensation is guaranteed. Brockers will be an unrestricted free agent in 2017. Jack Mewhort’s compensation in 2016 and 2017 is not guaranteed. After 2017, Mewhort will be an unrestricted free agent. Davis Manuel Sanu Jeffery Brockers Mewhort Year Team Cash Team Cash Team Cash Team Cash Team Cash Team Cash

39 2013 SF 6.67* 2014 SF 5.30 BUF 0.81 CIN 0.59 CHI 0.90* STL 1.26* 2015 SF/DEN 4.91 BUF 1.21 CIN 1.54 CHI 1.05 STL 1.73 IND 0.58 2016 WAS 2.38 BUF 1.62 ATL 8.00 CHI 11.16 LAR 9.15 IND 0.75 2017 WAS 6.00 OAK 0.80 ATL 6.00 PHI 16.00 LAR 10.25 IND 0.91 2018 WAS 3.97 OAK 0.10 ATL 6.00 PHI 7.90 LAR 10.25 IND 0.30 2019 WAS 4.97 KC 0.81 ATL 6.00 PHI 11.75 LAR 10.00 2020 ATL 6.50 PHI 13.00 2021 PHI 13.00 Williams Shazier Robinson Wright Year Team Cash Team Cash Team Cash Team Cash 2013 DAL 1.02 TEN 0.76 2014 DAL 0.53 PIT 5.65 JAX 1.29 TEN 1.14 2015 DAL 0.79 PIT 0.91 JAX 0.75 TEN 1.53 2016 DAL 1.67 PIT 1.28 JAX 0.74 TEN 7.32 2017 DAL 6.00 PIT 1.72 JAX 0.90 CHI 2.00 2018 DAL 2.88 PIT 8.70 CHI 15.81 MIN/ARI 0.73 2019 CHI 13.00 2020 CHI 13.00 2021 Table 7: NFL Career Production by Draft Round, 1990-2015 This table charts average career outcomes for all NFL players selected in rounds 1-7 of the NFL draft between 1990-2015. The NFL draft has 12 rounds in 1990-1992, 8 rounds in 1993, and 7 rounds from 1994-2015. The columns P1-P10 are the probabilities of a drafted player completing greater than or equal to 1-10 NFL seasons. These probabilities are conditional on the completion of that number of seasons since the draft. For instance, if there are only 3 completed NFL seasons since a particular player is drafted, that player is only included in the sample to estimate P1, P2, and P3. AP1 is the average number of First-Team NFL All-Pro Selections. The Associate Press selects the NFL All-Pro teams, and the First Team consists of one player per position per season. PB is the number of Pro Bowl Selections. Fans, players, and coaches all vote for Pro Bowl players. There is an average of 3.5 Pro Bowlers in the NFL per position per season. The number of selections depends on the position and whether replacement selections are required. If the player initially voted to the team is unable to play in the Pro Bowl game, the player with the next-highest votes is chosen for the team, and both players are Pro Bowl selections. St is the number of seasons a player has more starts than any other player at his position for his team. G is the number of games the player appears in during his career. All statistics are from pro-football-reference.com. Round P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 AP1 PB St G All 0.893 0.827 0.740 0.649 0.556 0.482 0.409 0.338 0.263 0.202 0.088 0.272 1.95 55.5 1 0.995 0.991 0.973 0.933 0.879 0.831 0.747 0.667 0.552 0.468 0.340 1.113 5.05 96.3 2 0.988 0.959 0.917 0.861 0.788 0.702 0.616 0.525 0.414 0.300 0.100 0.375 3.42 80.2 40 3 0.964 0.941 0.868 0.765 0.653 0.550 0.458 0.372 0.284 0.224 0.065 0.227 2.10 63.9 4 0.941 0.876 0.766 0.660 0.555 0.462 0.390 0.331 0.250 0.185 0.027 0.108 1.50 54.4 5 0.873 0.772 0.679 0.559 0.456 0.382 0.297 0.229 0.163 0.105 0.033 0.117 1.07 43.1 6 0.803 0.689 0.576 0.489 0.384 0.307 0.253 0.184 0.137 0.105 0.016 0.088 0.76 35.8 7 0.753 0.649 0.511 0.400 0.304 0.260 0.211 0.159 0.124 0.090 0.010 0.040 0.61 29.5 Table 8: NFL Career Length by Position Group, 1990-2014 For each player drafted in rounds 1-7 of the NFL draft between 1990-2014, I construct data on their career length in the league. To qualify for the sample, a player must be out of the league before the start of the 2015 NFL season. All statistics on draft round and years in the NFL are from pro-football- reference.com. Players are segregated into nine position group: quarterbacks (QB), offensive linemen (OL), running backs or fullbacks (RB/FB), wide receivers (WR), tight ends (TE), defensive linemen (DL), linebackers (LB), defensive backs (DB), and kickers or punters (K/P). For each position group, the table shows the percentage of players in that group with careers of 0 years, 1-3 years, 4-6 years, 7-9 years, 10-12 years, and more than 13 years. 0 1-3 4-6 7-9 10-12 13+ QB 18.6 26.8 25.5 13.4 9.5 6.1 OL 18.2 23.8 21.1 20.3 12.4 4.1 RB/FB 11.2 30.1 28.9 18.7 9.0 2.2 WR 16.0 33.8 25.7 15.9 5.5 3.1 TE 10.3 28.3 27.9 18.3 11.7 3.4 DL 11.8 29.3 22.8 19.1 12.9 4.1 LB 10.9 31.5 24.4 21.4 8.6 3.2 DB 11.6 29.3 28.2 18.6 9.9 2.4 K/P 17.5 33.3 15.9 7.9 11.1 14.3 41 Table 9: NFL Survival Rate Probit Regressions, 1990-2015 For each player drafted in rounds 1-7 of the NFL draft between 1990-2015, I construct data on their survival in the league. The dependent variable is a binary player-season survival indicator. The indicator is 1 if the player is on an active roster, inactive roster, or injured reserve list for at least six NFL games during a particular season. The indicator is only calculated for players who are in the league the previous season or selected in rounds 1-7 of the most recent draft. The independent variables include the player’s draft round and experience. Experience is measured as the number of NFL seasons that have passed since the player was drafted. Consider a player who was drafted in 2010 and played the 2010, 2011, and 2012 seasons before retiring. The player’s career is included in four observations: three successes in years 1, 2, and 3, and one failure in year 4. Using these data, I estimate probit models separately for each of nine position groups: quarterbacks (QB), offensive linemen (OL), running backs or fullbacks (RB/FB), wide receivers (WR), tight ends (TE), defensive linemen (DL), linebackers (LB), defensive backs (DB), and kickers or punters (K/P). Errors are clustered by draft year using Rogers (1994) standard errors. A player’s one-year conditional survival probability is Φ(βx) where Φ is the normal CDF operator, beta are the coefficient estimates from the probit, and x includes an intercept, the player’s draft round, and experience. For instance, an offensive lineman drafted in the second round with four years of NFL experience has an estimated probability of Φ(2.14 − 0.14 ∗ 2 − 0.10 ∗ 5) = Φ(1.36) = 0.913 of playing the next season. All statistics on draft round and years in the NFL are from pro-football-reference.com. QB OL RB/FB WR TE DL LB DB K/P Intercept 1.85 2.14 2.16 2.11 2.28 2.11 2.19 2.38 1.95

42 [16.98] [33.92] [24.51] [22.62] [16.98] [28.03] [27.56] [30.44] [6.86] Draft Round -0.13 -0.14 -0.14 -0.17 -0.13 -0.12 -0.15 -0.18 -0.10 [-4.91] [-16.10] [-9.72] [-10.42] [-6.77] [-10.06] [-9.96] [-15.38] [-2.15] Experience -0.04 -0.10 -0.12 -0.10 -0.11 -0.10 -0.11 -0.13 -0.02 [-3.80] [-14.43] [-10.77] [-10.03] [-9.29] [-15.27] [-13.59] [-13.68] [-1.85] Table 10: Summary Statistics for NFL Contract Data, 2000-2015 This table describes the NFL contract data from Spotrac (www.spotrac.com). Spotrac has data on NFL contracts from 2000-2015. If a player has contracts with two or more different teams in the same NFL season, I consolidate the contracts into a single entry. For each player-year observation, I estimate the contract compensation. If the player’s cash compensation for that season is available, I use that number. If the cash compensation is unavailable, but the current contract information is available, I estimate the player-year compensation as the yearly average of the total salary and guaranteed bonus in the contract. I only include players who were drafted in the first seven rounds of an NFL draft between 1990 and 2015 and who are playing in their first 10 NFL seasons. There must be an exact match between the player’s first and last names in the Spotrac data and in the historical NFL draft data, which are from www.pro-football-reference.com. The average, median, minimum, and maximum salaries are quoted as percentages of the NFL salary cap. The salary cap is the amount each team is allowed to pay its active roster of 53 players, so the average compensation of all players is approximately 1.89. The NFL salary cap data are from www.nfl.com and www.nfllabor.com and are quoted in millions of dollars. Year N Avg Median Min Max Cap 2000 165 3.52 1.40 0.07 19.22 62 2001 195 2.64 1.01 0.36 19.14 67 2002 253 2.92 1.12 0.36 21.64 71 2003 332 3.00 1.33 0.03 20.13 75 2004 459 2.52 0.87 0.03 43.48 81 2005 650 2.05 0.71 0.06 32.16 86 2006 829 1.95 0.61 0.04 24.20 102 2007 974 1.76 0.60 0.02 28.21 109 2008 1078 1.71 0.61 0.03 23.88 116 2009 1191 1.64 0.59 0.03 21.14 123 2010 1250 1.83 0.80 0.03 27.08 120 2011 1299 2.1 0.72 0.02 28.33 120 2012 1365 1.84 0.58 0.04 21.77 121 2013 1276 1.78 0.67 0.02 31.10 123 2014 1298 1.73 0.66 0.02 27.44 133 2015 1285 1.78 0.66 0.02 22.12 143

43 Table 11: NFL Cash Compensation OLS Regressions, 2000-2015 Each column of this table reports results of a panel OLS regression. The dependent variable is the player’s cash compensation for a particular NFL season, scaled as a percentage of that season’s NFL team salary cap. Salary cap data are from either www.nfllabor.com or www.nfl.com. NFL teams have 53 players on their active rosters each season, so the average value for the dependent variable is 1.89 (100/53). To qualify for the sample, the player must be drafted in rounds 1-7 of an NFL draft between 1990 and 2015, and his contract details must be in the www.spotrac.com data. These data cover around 20% of NFL players in 2000 and increase to nearly 100% of players from 2010 to 2015. Independent variables include Draft Round, the round in which the player was initially drafted, and Experience, the number of NFL seasons since the player was drafted. Using these data, I estimate OLS models for each of nine position groups: quarterbacks (QB), offensive linemen (OL), running backs or fullbacks (RB/FB), wide receivers (WR), tight ends (TE), defensive linemen (DL), linebackers (LB), defensive backs (DB), and kickers or punters (K/P). Errors are clustered by draft year using Rogers (1994) standard errors. All statistics on draft round and years in the NFL are from pro-football-reference.com. QB OL RB/FB WR TE DL LB DB K/P Intercept 4.14 2.39 2.12 1.90 1.95 2.26 2.25 1.82 0.83 [6.25] [10.67] [7.94] [7.65] [8.73] [8.21] [9.87] [9.33] [1.91] Draft Round -0.74 -0.43 -0.39 -0.41 -0.33 -0.40 -0.46 -0.37 -0.10 [-5.28] [-10.70] [-8.75] [-9.32] [-11.05] [-8.22] [-10.49] [-10.63] [-1.40] Experience 0.35 0.26 0.27 0.41 0.21 0.28 0.33 0.32 0.15 44 [3.72] [13.33] [9.71] [9.53] [6.34] [7.51] [15.43] [14.34] [6.38] N 836 2511 1237 1534 854 2150 1943 2496 338 R2 0.15 0.21 0.21 0.24 0.24 0.17 0.23 0.24 0.25 Table 12: Acquired Brand Income (ABI) Contract IRR Estimates This table estimates the internal rate of return (IRR) for the completed IPOs of ABI contracts. Table 1 summarizes the completed IPO transactions, and Table 2 provides details about sources of income included in the ABI contract terms. The Acquired Brand Income (ABI) contract stipulates that the contract party pay Fantex Holdings a fixed percentage of all pre-tax “brand income” earned on or after the Effective Date. Brand income includes salary and bonus payments as a professional athlete, endorsement earnings, and earnings from employment, such as coaching and broadcasting, related to the athlete’s sport. See Table 2 for a list of revenue sources which are included in or excluded from acquired brand income. The IPO Date is the date when Fantex secures funding and the transaction closes. Proceeds is the amount of the lump sum payment, as stipulated in the ABI contract. The Proceeds are paid to the athlete within five business days of the IPO date. The Proceeds are 95% of the nominal lump sum in the ABI contract. The remaining 5% are withheld for underwriting services, payable to Fantex Brokerage Services, LLC, an affiliated broker-dealer. Under the terms of the ABI contract, any payments the player owes Fantex for brand income he earns between the effective date and the IPO date are deducted from the lump sum payment. Implied PV Total Earnings is Proceeds/%ABI. Implied PV NFL Earnings is Implied PV Total Earnings less the Fantex estimate of the present value of brand income from current endorsements, future endorsements, and post-career earnings. I estimate the present value of non-NFL earnings from the first Fantex 10-K or 10-Q SEC filing published after the completed IPO. Implied IRR is a single rate of return that equates Implied PV NFL Earnings with the sum of all future discounted expected cash flows. I estimate expected NFL earnings from the effective date of the ABI contract up to a maximum career of 20 additional seasons. The expected compensation in a season is the product of

45 the expected survival rate and expected NFL earnings conditional on survival. The survival rate each season is estimated using a probit model for each position group, where the explanatory variables are the player’s draft round and experience. Table 8 provides more details about the probit model. The expected NFL earnings conditional on survival is estimated using an OLS model for each position group, where the explanatory variables are the player’s draft round and experience. Details of the OLS model are in Table 9. I estimate the NFL salary cap will grow at 5% annually from $133 million in 2014. EW Average is the equal-weighted average IRR of the six completed transactions. VW Average is the value-weighted average of the six completed transactions, using the implied present value of NFL earnings. MSA Rank is the ranking of the metropolitan statistical area (MSA) population, using 2010 US census data.

Player Effective IPO Implied PV Implied PV Expected Minimum MSA Rank Date Date Proceeds %ABI Total Earnings NFL Earnings IRR IRR Vernon Davis 10/30/2013 4/28/2014 $4.00M 10% $40.00M $32.17M -6.4% -10.6% 11 EJ Manuel 2/14/2014 7/21/2014 $4.98M 10% $49.75M $39.82M 15.3% 11.3% 47 Mohamed Sanu 5/14/2014 11/3/2014 $1.56M 10% $15.61M $15.37M 10.4% 9.9% 28 Alshon Jeffery 9/18/2014 3/19/2015 $7.94M 13% $61.08M $59.04M -8.2% -8.6% 3 Michael Brockers 1/9/2015 5/29/2015 $3.44M 10% $34.41M $34.15M 5.1% 4.9% 18 Jack Mewhort 3/26/2015 7/14/2015 $2.52M 10% $25.16M $25.44M 7.2% 7.1% 33 EW Average 3.9% 2.3% VW Average 2.1% 0.5% Table 13: Player-Specific Compensation Tobit Regressions, 2000-2015 Each column of this table reports results of a player-specific tobit regression. The dependent variable is the player’s cash compensation for a particular NFL season, scaled as a percentage of that season’s NFL team salary cap. Salary cap data are from either www.nfllabor.com or www.nfl.com. NFL teams have 53 players on their active rosters each season, so the average value for the dependent variable among active players is 1.89 (100/53). If the player is no longer in the NFL prior to the 2015 season, I include observations in which the player earns compensation of year 0 for all years from his first year out of the NFL to his 20th season. I construct a sample which is specific to each player based on position and experience as of the Effective Date of the player’s ABI contract IPO. See Table 1 for details of the completed IPO transactions. Players are divided into nine position groups: quarterbacks (QB), offensive linemen (OL), running backs or fullbacks (RB/FB), wide receivers (WR), tight ends (TE), defensive linemen (DL), linebackers (LB), defensive backs (DB), and kickers or punters (K/P). For example, the player-specific sample for Vernon Davis consists of all tight ends with eight or more years of experience. Furthermore, to qualify for the sample, the player must be drafted in rounds 1-7 of an NFL draft between 1990 and 2015, and his contract details must be in the www.spotrac.com data. These data cover around 20% of NFL players in 2000 and increase to nearly 100% of players from 2010 to 2015. See Table 10 for summary statistics on the contract data. Independent variables include Draft Round, the round in which the player was initially drafted, and Experience, the number of NFL seasons since the player was drafted. Errors are clustered by draft year using Rogers (1994) standard errors. All statistics on draft round and years in the NFL are from pro-football-reference.com. Davis Manuel Sanu Jeffery Brockers Mewhort

46 Intercept 12.55 8.58 5.68 6.98 6.18 4.90 [9.79] [7.71] [9.84] [10.40] [9.48] [15.75] Draft Round -0.62 -1.31 -0.75 -0.87 -0.47 -0.59 [-2.75] [-4.51] [-6.37] [-5.78] [-3.48] [-12.17] Experience -1.00 -0.71 -0.64 -0.78 -0.71 -0.47 [-6.90] [-12.84] [-13.46] [-14.53] -[18.80] [-17.35] N 371 2226 4189 3873 3991 6622 Position TE QB WR WR DL OL Years 8+ 1+ 2+ 3+ 3+ 1+ Pseudo R2 0.27 0.10 0.16 0.16 0.16 0.19