Sample Proposal to University to Franchise by hbz73105

VIEWS: 32 PAGES: 241

More Info
									LA Lakers’ salaries, 2003-04
Shaquille O’Neal               28,515,000
Kobe Bryant                    13,500,000
Gary Payton                    4,917,000
Rick Fox                       4,549,000
Devean George                  4,545,000
Derek Fisher                   3,000,000
Stanislav Medvedenko           1,500,000
Karl Malone                    1,483,000
Kareem Rush                    1,096,000
Bryon Russell                  1,070,000
Samaki Walker                  1,400,000
Brian Cook                     752,000
Jamal Sampson                  583,000
Luke Walton                    366,000
Ime Oduka                      366,000
LA Clippers’ salaries, 2003-04

Elton Brand                      10,960,000
Cory Maggette                    5,475,000
Pedrag Drobnjak                  2,500,000
Chris Kaman                      2,394,000
Keyon Dooling                    2,256,000
Chris Wilcox                     2,066,000
Marko Jaric                      1,900,000
Quentin Richardson               1,805,000
Melvin Ely                       1,628,000
Olden Polynice                   1,070,000
Doug Overton                     974,000
Eddy House                       775,000
Matt Barnes                      366,000
Josh Moore                       366,000
Core principles of the economic
way of thinking
• People optimize – they try to maximize some
  objective, subject to constraints, i.e. to the
  fundamental fact that they can’t have everything
  they want. The object of maximization often is,
  but need not be, money.
• This leads to the idea of efficiency, or how to
  create the most value.
• Value is often best thought of as the sum of
  producer surplus (space between price line and
  supply curve) and consumer surplus (space
  between price line and demand curve).
Markets and the creation of value
• Unrestricted operation of markets generally create the
  most value. There are three exceptions:
• (1) Restricted competition. In markets where there is
  imperfect competition and restricted entry, the price will
  not be the competitive one.
• (2) Externalities. When an action or transaction affects
  others, those costs or benefits are not properly
  internalized by the price system.
• (3) Public goods. Some goods are nonrivalrous and
  nonexclusive, and are therefore underprovided. (More to
  come in unit on stadiums and mega-events.)
Regression – the basics
• Regression is a statistical technique to measure the extent of the
  relationship between variables that theories say are supposed to be
  related. Wages, for example, might be related to years of
  experience and to education. Mathematically, we might write w =
• There are two criteria that are often used in regressions:
• (1) Statistical significance for particular variables. Is the relation
  between a single right-hand variable (schooling, say) and the left-
  hand variable (wages) strong enough in the data that it is probably
  not due to random chance?
• (2) Overall explanatory power of the model. Usually measured by
  R2, a figure which varies between 0 and 1. A high R2 means that
  movement in the right-hand variables coincides with most of the
  movement in the left-hand one. An R2 of close to zero means that
  collectively little of the movement in the right-hand variables
  coincides with movement in the left-hand ones.
NFL, 1996

Gate Revenue
   Mean: $23 million
   Variance: $14.97 million
Broadcast Revenue
   Mean: $43 million
   Variance: $40.41 million
Licensing Revenue
   Mean: $6 million
   Variance: $58.61 million
MLB, 1996

Gate Revenue
     Mean: $25.73 million
     Variance: $140.144 million
Broadcast Revenue
     Mean: $25.23 million
     Variance: $93.69 million
Licensing Revenue
     Mean: $12.65 million
     Variance: $41.74 million
NBA, 1996

Gate Revenue
   Mean: $23 million
   Variance: $64.07 million
Broadcast Revenue
   Mean: $21 million
   Variance: $27.06 million
Licensing Revenue
   Mean: $8 million
   Variance: $26.97 million
Economic idea: vertical and horizontal
• Vertical Integration occurs when a
  company in one stage of a production
  process purchases a firm in another stage.
• Horizontal Integration occurs when one
  company at a particular level of the
  production process (especially the final
  product) purchases another company at
  the same stage.
• Accounting profits are money in (total
  revenue) minus money out.
• Economic profits are total revenue minus
  opportunity cost of resources used.
Economic ideas: Standardization and network externalities

• A network externality is when the benefits
  a consumer derives from owning a product
  depend on the number of other users.
  Standardization enables gains from
  network externalities. Examples: fax
  machines operating on a common
  standard, the Windows operating system.
• In this case, a “product” is the number of
  teams playing by the same set of rules.
  The more players playing by a given set of
  rules, the more appealing the product is.
Economic idea: Path dependence

• Path dependence occurs when the product that
  arrives first is able to maintain dominance because it
  becomes the industry standard, even if it is inferior.
  The costs of adopting a technology that others don’t
  use are too great. Superior technologies thus can’t
  get off the ground.
• Alleged examples: The QWERTY typewriter,
  DOS/Windows, the VHS videocassette.
• A sports league, from the point of view of setting
  rules, thus represents a tradeoff – gains from single
  set of rules, possible losses from rules being inferior.
        Economic Idea: Productive Efficiency – For a
given amount of inputs, how close does a firm come to
achieving the maximum level of output? Equivalently,
for a given amount of input, how close does a firm come
to producing it at minimum cost?
Sample Production Function for the NBA
        To measure productive efficiency, first
estimate a production function relating wins to
statistical productivity. Then for each team, use its
statistical productivity and the production-function
estimates to calculate an expected number of wins.
The gap between these two is the total inefficiency in
the use of inputs. A higher number indicates better
Ruggiero et al.

Winning percentage = 2.116 + 1.175 * Slugging
percentage + 0.501 * Batting average + 0.054 *
Stolen bases + 7.465 * Fielding percentage –
0.826 * ERA
Ruggiero et al. Table 14.4, Efficiency in Win

AL                          NL
Oakland           .863      Philadelphia        .832
Boston            .839      Houston             .830
California        .836      St. Louis           .829
New York          .836      Colorado            .822
Chicago           .833      San Diego           .819
Baltimore         .832      San Francisco       .813
Detroit           .830      New York            .809
Milwaukee         .825      Atlanta             .808
Texas             .808      Los Angeles         .806
Kansas City       .805      Cincinnati          .805
Cleveland         .804      Chicago             .800
Minnesota         .804      Florida             .795
Toronto           .791      Montreal            .793
Seattle           .784      Pittsburgh          .793
Source: Lawrence Hadley, Marc
Poitras, John Ruggiero and Scott
Knowles, “Performance
Evaluation of National Football
League Teams,” Managerial and
Decision Economics 21, 2000, 63-
Source: Lawrence
Hadley et al.
Source: Hadley et al.
Above-average coaches   Below-average coaches
1. John Madden          John McKay
2. Art Shell            Dick Nolan
3. Don Shula            Bart Starr
4. Bud Grant            Sam Wyche
5. Chuck Knox
6. Joe Gibbs
7. Dan Reeves
8. Tom Landry
9. Mike Ditka

Source: Hadley et al.
Arbitration and Free Agency: Hadley and
Gustafson (1991)
• Question: How do arbitration and free
  agency affect player salaries?
• Test: for a given level of statistical
  productivity, are pre-arbitration, arbitration
  and free-agent players paid differently?
Source: Hadley and Gustafson
Hadley and Gustafson, Figure 1.
Hadley and Gustafson, Figure 2.
Measures of competitive balance

• Dispersion of team winning percentage in
  a season
• Concentration of championship titles
• Dispersion of lifetime franchise winning
Quirk and Fort, Table 7.1.
Quirk and Fort, Table 7.1 (continued).
• Economic idea: A Lorenz Curve relates
  the percentage of the population to the
  percentage of some other variable it
  possesses. A 45-degree line represents
  perfectly equal distribution of the other
  variable. Most commonly used example:
  distribution of income.
Why does competitive imbalance exist at

• Dynasties and dominant teams have value
  for fans, too.
• Resistance of major-franchise owners.
• Owners substantially maximize profits, and
  Coase theorem applies.
Economic Idea: The Coase Theorem explores
whether the initial distribution of property
rights matters with respect to its final use. He
argues that as long as transaction costs are
low, the initial allocation of a right to a
particular piece of property does not matter.
Applications include legal disputes, especially
nuisance suits, allocation of broadcast
frequencies, and many others.
Ways to improve competitive balance

• Universally used:
  -   Reverse-order draft
  -   League-wide TV contracts
• Less widely used
  -   Salary caps (NBA, NHL, NFL)
      - Soft salary cap: exemption for own free agents.
      - Hard salary cap: no exemption.
  -   Reverse-order scheduling (NFL only)
  -   Luxury taxes (MLB only)
• Seldom used
  -   More franchises
Problems of salary caps
• Owners cheat.
• Benefits mostly low-end and highest-end
• Given franchise-entry restrictions, it is
  more efficient for bigger cities, with bigger
  audiences, to win more.
• Disincentive to improve.
Economic Idea: Marginal Revenue Product
is the addition to a firm’s revenue by a
particular worker, equivalent to his marginal
product (i.e., his addition to output) times
the unit price of the output. Economic
theory holds that in perfect competition
workers’ wages will always be bid up to
marginal revenue product. If workers are
paid less than MRP, it is a sign of market
Scully (1974) – Estimating MRP

A three-stage process:
- Step 1: Find the relation between a team’s
  statistical production and its winning
- Step 2: Find the relation between wins and
  team revenue.
- Step 3: Estimate a player’s contribution to
  team statistics, therefore to wins, and
  therefore to revenue.
Scully (1974) (continued)

• Scully, estimated relation between winning percentage
  and player statistical performance
  PCTWIN = 37.24 + 0.92 * SLUGGING + 0.90 * K/W –
  38.57 * NL + 43.78 * CONTEND – 75.64 * OUT
• Estimated relation between wins, team revenue
  REVENUE = -1,735,890 + 10,330 PCTWIN + 494,585 *
  POPULATION + 512 * FAN INTENSITY + 580,913 * NL
  – 762,248 * OLD PARK - 58,523 * BLACK PLAYERS
• One player’s MRP assumed to be:
  1/12 * SLUGGING * $10,330 (position players)
  1/8 * K/W * $10,330 (starting pitchers)
Scully (1974) (continued)

• There is now a predicted relation between
  statistical production and salaries if
  players are paid MRP.
• Next step: Regress actual individual
  salaries on statistical productivity and
  these other considerations. Compare
  what typical players are actually paid to
  what they should be paid according to the
  MRP they generate.
End result (Table 2): The least talented
players were actually overpaid after
deducting their maintenance costs
(transportation, training equipment, etc.)
Their “net MRP” (gross additions to
revenue minus these costs) was
negative. But average and star players
got only about 15-25 percent of their net
“The most radical proposal is a completely free labor
market with all contracts for a full season being
negotiated off-season. The proposal would eliminate
player economic rents. Organized baseball argues that
such a scheme would destroy the game. They point to
the rich owner, who could not be prevented from buying
all of the good players. They argue that investments in
teams would be unattractive. Teams would fold and no
buyers would be found. They also forecast the end of
player development and minor league subsidies and
hence long-term damage to the sport.”
        - Scully (1974), p. 930.
“The Wages of Sin…” (Jones et al.)

• Question: What are the features of a hockey
  player that teams are willing to pay for? Is
  violent play one of them?
• Test: Regress hockey salaries on measure of
  experience, playing skill, physical size and
  market structure. “Playing skill” includes
  penalty minutes, which for most players
  should have a negative effect on salary.
  Defensemen and forwards are assumed to
  have different salary equations.
• An econometric technique called “switching
  regression” allows you to look at a group of
  observations and see if they can be allocated
  into two distinct groups (“goons” versus non-
• After conducting the regressions, such
  differences exist. In other words, there are two
  different salary processes determining the
  compensation of goons and non-goons. Goons
  are clearly a distinctive input into win production.
Statistical differences in the two groups:


      Non-goons: 0.74 points/game
      Goons: 0.31 points/game
      Non-goons: 0.51 points/game
      Goons: 0.25 points/game

Penalty Minutes

      Non-goons: 0.88 minutes/game
      Goons: 3.14 minutes/game
      Non-goons: 0.98 minutes/game
      Goons: 2.31 minutes/game
Determinants of pay for non-goons

• Forwards
   – Points per game and penalty minutes are positively
     significant for goons and non-goons. But experience
     is only significant for non-goons, suggesting that there
     is a greater search cost problem in identifying quality
     non-goon players that doesn’t exist with goons.
• Defensemen
  - Experience, penalty minutes and weight are positive
  and significant determinants of salary. But scoring is
  only significant for non-goons.
What determines your probability of being in
the goon category?

•       A “probit” estimation finds that weight
    and penalty minutes are positively and
    significant determinants, while scoring is,
    for defensemen at least, negatively and
    significantly correlated.
The marginal product of figure skaters

• Economic idea: Revealed preference
  refers to the use of people’s choices to
  infer their preferences, rather than vice
Economic idea – the elasticity of demand
measures the responsiveness of quantity
demanded to changes in price. When a small
change in price (in either direction) causes a big
change in quantity (always in the other
direction), we say that demand is very elastic.
When a big change in price causes a small
change in quantity, demand is very inelastic.
Elasticity of demand is a function of the number
and quality of the available substitutes.
• “We’re not really going to worry about
  what the hell [the fans] think about us.
  They really don’t matter to us. They can
  boo us every day, but they’re still going to
  ask for our autographs if they see us on
  the street. That’s why they’re fans and
  we’re NBA players.”
  - Former Portland Trailblazer Bonzi Wells.
The economics of rank-order tournaments

• Assumption: Workers/competitors can’t be paid their
  marginal revenue products because MRPs and effort are
  unobservable. The only thing the firm can observe is the
  relative order of output (i.e., who produces the most, who
  produces the second-most, etc.)
• The marginal cost of effort by the worker, while
  unobservable, is known by the worker and is forever
  increasing. In other words, the marginal cost of effort is
  always getting bigger.
• Therefore, to obtain significant effort the prizes to
  finishing high in the ranking must be large, and then
  must decline quickly as the final ranking increases. In
  other words, the top finisher must be paid much more
  than the second-place finisher, who must get
  significantly more than the third-place finisher, etc.
PGA Prize Structure (Ehrenberg and
• Two tests of whether the tournament model
  describes PGA rewards:
- (1) Does more prize money induce better scores
  (i.e., more effort)?
- (2) Do individual players try harder when the
  marginal reward to effort is greater?
Ehrenberg and Bognanno
• Testing (1):
- Regress scores on measures of course difficulty, golfer skill and
  rewards. All variables are significant in the expected direction. It
  takes more work to do well on a harder course, and prize money
  reflects that.
- If the tournament is a major, then for a given starting position,
  golfer skill and monetary reward scores are better.
• Testing (2):
- Look at a player’s ranking entering the last day of the
  tournament. Use available prize money and number of golfers
  near him in the standings to calculate the marginal reward of
  effort, where effort is assumed to mean a given number of
  improvement in strokes.
“It may cost you $5 million to get to the track, but it
might cost you an additional $3 million for a few tenths
better lap times…It’s pretty cost-effective to a certain
point, but that extra little bit is where it is starting to get
Bill Elliott, driver and owner.
Von Allmen

• Thus, old NASCAR point system is rising, but
  not as much as PGA. Why?
• (1) Improving performance in NASCAR is so
  expensive that success early in the season
  could be largely determinative of later success,
• (2) There is a negative externality to one type of
  increased effort – riskier driving. The
  compensation system must keep this under
Economic idea: two models of strikes

• The resistance model: Each side in a labor dispute has
  some threat point, i.e. an ability to maintain a standoff. This
  is a function of alternatives available without production.
  Example: rival leagues such as the USFL always bring
  dramatic salary escalation.
• The length of a strike and the resulting wage increase are a
  function of the strength of each side’s threat point. The
  asymmetric information model: Only management knows
  the true profit level, and they have an incentive to overstate
  it. A strike is a way to solicit this information from
  management. A long strike is a sign profits are low, and a
  quick settlement indicates that management is understating
  them. If labor usually wins that is an implication that
  management does in fact have a lot of profits not being
  distributed as compensation.
A history of North American sports work
• 1972: MLB strike nominally over pensions, really over an attempt to
  crush MLBPA, which was increasingly assertive. Almost complete
  victory for players.
• 1987: NFL strike primarily over free agency. NFL resorted to
  replacement players for several games. Eventually threat of striker
  defections led to owners winning.
• 1994: NHL lockout over free agency and salary cap. Rookie salary
  cap and limits on free agency resulted.
• 1994: MLB lockout over salary cap, which was fought successfully in
  favor of a luxury tax.
• 1998: NBA lockout over player desires to eliminate salary cap and
  draft. In the end salary cap was tightened with restrictions on pay
  related to years in league. Owners helped by contract requiring
  NBC and TNT to continue to make payments during lockout.
• 2004-5: NHL lockout over salary cap, which owners got.
Average salaries, 2001-2 (NHL 2002-3)

•   NFL 1,200,000*
•   NBA 4,500,000
•   MLB 2,550,000
•   NHL 1,640,000

*NFL contracts not guaranteed.
Free-agency rules

• MLB: Unlimited after six years, arbitration
  years 3-6, reserve clause years 0-2.
• NFL: Eligibility after four years, subject to
  franchise-player exception.
• NBA: Teams have right of first refusal after
  four years, players unrestricted after five.
• NHL: After seven years players become
  free agents.
Salary caps

• NFL: 64% of revenues, plus prorated signing
• NBA: 55% of revenues Plus Larry Bird
  exception. Separate cap on rookie salaries.
• NHL: 54-57% of revenues, depending on how
  big revenues are. Team salaries must be
  between $21-$39 million.
• MLB: No cap, but luxury tax starting at $117
  million payroll.
Economic idea: economic theories of
• Taste for discrimination: Firms pay less for
  workers certain groups either because firm
  owners don’t like them (employer discrimination)
  or because customers don’t (customer
• Statistical discrimination: Firms are in principle
  willing to pay workers according to productivity
  (assuming customer discrimination doesn’t
  exist), but group membership is used as a
  statistical shortcut when individual productivity
  data are too costly.
Older tests of customer
• Baseball cards – do people pay more for
  white players with equal career statistics?
• All-Star voting – do fans vote more for
  white players with equal career statistics?
• In both cases the answer is yes, with
  diminishing margins (perhaps to zero now)
  over time.
Testing for employer discrimination: To
what extent to workers from some
groups get paid less because they are
less qualified, versus because they are
being discriminated against?
The “Oaxaca decomposition”: To try to estimate
how much of a wage difference between two
groups is due to discrimination and how much to
differences in qualifications, multiply the higher-
paid group’s regression coefficients times the
lower-paid group’s average qualifications. This
tells us what the lower-paid group would make if it
had the qualifications of the higher-paid group.
Any wage difference that remains after doing this
is a rough measure of discrimination.
• Fundamental question: do black football players get paid
  less than white ones, after standardizing for other
  relevant differences?
• Test: Regress players’ salaries on player-skill measures
  (Pro Bowl appearances, draft position, injuries), market
  size, injuries, experience measures and race.
• Years in league, games started, draft position, Pro Bowl
  appearances and being white are positively and
  significantly related. Number of injuries, and being white
  multiplied by the fraction of the city that is nonwhite, are
  negatively and significantly related.
Kahn (continued)

• However: most salary difference is due to
  positional segregation. Once player
  position is controlled for, race effects are
  not statistically significant.
• Other things equal, whites make more in
  cities with bigger white population
  percentages, nonwhites make more in
  cities with bigger nonwhite pop.
Hoang and Rascher
• Question: Do white players stay in league
  longer, after standardizing for other relevant
• Test: Regress probability of exiting league in a
  given year of one’s career on player statistics,
  injuries, team record, draft position, race,
  experience measures, and player position.
• Result: Points/minute, number of injuries, games
  played, and being white are negatively and
  significantly correlated with the probability of
  exiting the league. Lower position in draft is
  positively and significantly correlated.
Why? Customer or employer
• Test: Regress number of black (white)
  players on team as function of percentage
  of city’s population that is black (white).
  This turns out to be a significantly positive
  predictor of team’s racial makeup.
• Regress attendance on winning
  percentage, arena size, and extent to
  which city has more whites and team has
  more whites.
Kanazawa and Funk
• Another test for customer discrimination: relation
  between the number of white players and TV ratings.
• Test: Regress Nielsen ratings of locally televised games
  on quality of both teams and their players, game time,
  market size, degree of competition from other sports,
  number of whites in the city, and number of white players
  on each team.
• Result: Quality, prime-time games, market size, number
  of white players positively and significantly correlated.
  Degree of competition negatively and significantly
From: Kanazawa and Funk (2001)
Aldrich et al. – Discrimination in
favor of black quarterbacks
• Overall, Monday Night Football games with at least one
  black starting quarterback have Nielsen ratings at least
  as high as those with 2 white QBs, despite generally
  featuring smaller-market teams.
• Standardizing for week in season, QB ratings and
  rushing yards, average team scoring and team wins, the
  presence of a black QB still has a statistically significant,
  positive effect on ratings – approximately two million
  viewers out of roughly 20,000,000.
• Effect is most pronounced for males age 18-34.
• Over time, having a black QB has gone from making a
  team less likely to more likely to be scheduled on MNF.
• Own-race effect (preference of black viewers for
  black QB)? Unlikely. Black viewership of MNF
  would have to increase 67% to account for this.
• It is thus probably significantly due to increased
  white viewership, which may therefore be a taste
  for diversity.
• Because whites are a minority in the NBA, the
  Kanazawa/Funk results – greater viewership
  when there are more white NBA players – can
  be interpreted in the same way.
Shmanske – Male and female golfers

• Question: Male golfers are better-paid
  than female golfers. Is it because men put
  out a more attractive product, or is it
  because of intrinsic customer
• Test: Conduct Oaxaca decomposition to
  see what female players would be paid if
  they played like males.
Shmankse – Comparison of PGA and LPGA
tours, 1999
• Events – 45 PGA events, with none less
  than 3 rounds and 2 with 5 rounds. 36
  LPGA events, 12 of which have only three
• Average yardage: 6998 (PGA), 6282
• Average purse: 2,144,444 (PGA),
  $788,500 (LPGA)
Shmankse – Comparison of PGA and LPGA
tours, 1999 (continued)
• Average putts/round: 29.148 (PGA),
  30.300 (LPGA)
• Average score: 70.902 vs. 72.918
• Average driving distance: 271.25 vs.
• Greens in regulation: 65.624% vs.
• Drive accuracy: 69.774% vs. 69.065%.
• Sand saves: 52.455% vs. 39.208%.
Shmanske (continued)

• Oaxaca decomposition indicates that the percentage
  of LPGA compensation due to differing payment for
  the same amount greens in regulation, sand saves
  and the number of putts is actually negative – i.e.,
  LPGA golfers are paid more for a given amount of
  these skills than PGA golfers. (PGA golfers are paid
  more for a given amount of driving distance and
• Overall, the amount of the compensation gap
  explained by differential PGA golfer productivity is
  129%. In other words, given their productivity LPGA
  golfers are actually compensated 29% more than PGA
Economic idea: Moral hazard occurs when
one party to a contractual relationship has
incentives to behave in a way that harm the
second party, and when the first party’s
behavior cannot be observed. When he
fails to perform the way the second party
wants him to, he is said to shirk. Examples:
insurance, agricultural labor, committee
members or participants in group projects.
The trick is to come up with a contract that
provides the first party with the proper
Economic idea: Market failure

• Market failure occurs when voluntary
  exchange does not lead resources to be
  used in the way that creates the most
  value. In theory (a critical qualifier),
  government can restrict or redirect
  resource use to enhance overall social
Two of the three types of market failure

• Externalities occur when the effects of a transaction spill
  over and affect nonparticipants. (Example: network
  externality, typically positive.)
• Public goods are goods with two features:
• (1) They are nonrivalrous, meaning consumption by one
  person doesn’t leave less available for others.
• (2) They are nonexcludable, meaning nonpayers can’t
  be excluded from consumption. They are thus subject to
  free-riding, an unwillingness to pay the full value of the
• Public goods, for these reasons, are under-provided by
  the market.
The Cincinnati Stadium Deals –
Paul Brown Stadium
• Paul Brown Stadium
• The process
• 1993 – Mike Brown is paying $2,500,000 annually in rent to be the second
  tenant at Riverfront Stadium. He gets no income from parking or
  advertising, while many teams got 100 percent of it. He threatens for the
  first time to move the Bengals.
• 1994 – The Bengals are revealed to have the worst stadium revenue deal of
  all NFL teams, at $53,000,000 annually. (Dallas, a luxury-box pioneer, led
  with $101,000,000.) Brown agrees to stay if the Bengals have a new
  stadium in place by 2000.
• 1995 and 1996, throughout – Mike Brown threatens to move the Bengals,
  especially to Baltimore.
• June 30, 1995: At seven minutes until midnight, Brown’s deadline for
  Cincinnati to reach a decision, the City Council passes a deal for a new
  stadium to be financed by a one-cent increase in the sales tax, later changed
  to a half-cent in 1996.
Paul Brown Stadium (continued)
•   The process (continued)
•   March, 1996: After public protest, a referendum is held on a half-cent
    increase, combined with a property-tax rollback, to fund two stadiums for
    $544,000,000, an estimate that would prove to be ridiculously low. With Art
    Modell’s help, referendum passes with 60 percent approval.
•   1997: A new dispute arises over city-owned land in the middle of the
    proposed stadium site. City council refuses to transfer the land until Brown
    negotiates a better deal regarding surrounding construction rights. Brown
    says cough the land up by 1/31/98 or he is gone.
•   Feb. 1, 1998, 1:00 AM: The land is transferred.
•   Mid-2000: Lease changed to eliminate county responsibility for ticket
    shortfalls, but requiring county to pick up Bengals federal tax obligations.
•   August, 2000: After huge cost overruns, stadium opens.
•   Summer, 2004: Hamilton County sues Bengals and NFL for unfair lease.
Paul Brown Stadium (continued)
• The contract:
• - County pays all maintenance other than on
  game day.
• - County pays for new technology as soon as 14
  other stadiums get it, or 7 get it with public
• - Bengals get all concession and advertising
• - Bengals can veto events and get 50% of
  revenue from any other events (concerts, high-
  school games, etc.)
Paul Brown Stadium (continued)
• The costs
• - Bengals contributed $25 million (through seat licenses) of
  $455 million construction expense.
• - From 2009-2026 (with ten-year Bengals option), Bengals will
  pay no rent.
• - By overseeing construction, Hamilton County brought about
  $50 million on cost overruns.
• - Sales-tax revenue less than expected, perhaps forcing a
  choice between funding stadium from general revenues or
  canceling the property-tax cut that was given in exchange for
  the sales-tax increase. Projected $8 million (or more) shortfall
  by 2007.
• Sales tax has already been extended 16 years longer than
  originally planned.
Great American Ballpark

• Naming rights: $75,000,000 over thirty years.
• Estimated cost: $361 million, after eliminating
  upper-deck sunscreen, front-entrance canopy,
  cheaper counter tops in the luxury boxes and
  other cost-cutting measures, as well as re-
  tendering some construction contracts. It was
  $11 million under budget, as opposed to $51
  million over budget for Paul Brown Stadium.
  Public subsidy was $300 million.
• But Reds paid for any cost overruns, which were
  then unsurprisingly few.
Why can cities be expected to
• Entry limitation gives franchises market
• Winner’s curse problem – without a
  strategy to correct for the problem, the
  party that ones an auction for an asset of
  unknown value is likely to be the one that
  most overestimated its value.
The economic impact of mega-events

• Direct benefit: People come to town and
  spend on hotels, restaurants, etc.
• Direct benefit: Multiplier effects.
• Indirect benefit: For Super Bowl in
  particular, corporate movers and shakers
  get a chance to scout out the city as a
  business location.
Reasons for doubt:

• Cost: From 1995-2003, cities spent $6.4 billion on stadium
  construction and refurbishment. Reliant Stadium in Houston,
  for example, cost over $400 million. Mega-events also require
  extra public services, which are often not included in the
• Crowding out: The opportunity cost of a mega-event guest is
  that some other guest might not be using the hotel room and
  attending the restaurants. People who might otherwise visit a
  city for some other reason will refuse to do so when the Super
  Bowl is in town.
• Substitutions: Much of the income is from local residents,
  who are not spending it elsewhere in the local economy.
• Leakage: Many of the employment and spending effects to
  some extent benefit people outside the area.
Source: Robert A. Baade and Victor A. Matheson, “Super Bowl or Super
(Hyper)Bole: Assessing the Economic Impact of America’s Premier Sports Event”
The evidence

• Phil Porter (1999)
   - Look at cities that host Super Bowl. Compare retail
   sales and sales taxes to what those cities had a year
   earlier. Finding: no detectable effect.
   - Hotel rental rates are the same, although room prices
   are higher.
• Baade and Matheson (2003)
   - Use a regression to measure the relation between
   income growth in a group of host cities as a function of
   other variables (new business hiring, e.g.) and whether
   or not the city hosted a mega-event. Finding: At most,
   benefits are 25 percent of what the NFL claims.
• “Thanks to Super Bowl XXXIII, there was a $670
  million increase in taxable sales in South Florida
  compared to the equivalent January-February
  period in 1998.” – NFL, 1999.
• But nominal sales taxes grow anyway because
  of population, inflation and expected economic
  growth. According to one study (Baade and
  Matheson 2005), accounting for these lowers
  impact of 1998 Super Bowl to $37 million.
“Wright State University will be a catalyst for educational
excellence in the Miami Valley, meeting the need for an
educated citizenry dedicated to lifelong learning and
service. To those ends, as a metropolitan university,
Wright State will provide: access to scholarship and
learning; economic and technological development;
leadership in health, education, and human services;
cultural enhancement, and international understanding
while fostering collegial involvement and responsibility
for continuous improvement of education and research.”
      - The WSU mission statement
Multi-Sport Male Athletes
Multi-Sport Female Athletes
I AM a DIE-HARD Clemson Alum and fan. I love everything about the
University. In the issue of football, I've always been against firing any of our
coaches because changes bring a lot of instability; case in point, Vince
Dooley at Georgia, Barry Switzer at Oklahoma, and of course, Danny Ford at
CU. But enough is enough! I've supported Coach West since he first became
our coach, but I haven't seen any signs of improvement or that he is the man
to take CU football BACK TO THE TOP. I'm sorry to say it. I didn't even realize
that we are ONLY VERY few wins above .500. I guess I've been fooled by the
fact that we have been going to bowl games (AND LOSING). I think this year
is JUDGMENT YEAR for Coach West. Mediocrity is unacceptable at Clemson
University. Speaking of mediocrity, our chicken "friends," who at one time
made fun of us for scheduling Ball State, seemed to have shown once again
just how pathetic they are. After glancing at my brand new issue of Sporting
News, it caught my eye that those u SCum chickens have scheduled TWO (2, I
say) Mid-American teams this year. Well, I guess even when Kentucky and
Vandy have gained a leg up on you, chickens have no choice but to RETREAT
to a more comforting area by TRYING (and I do mean TRYING) to beat up on
Mid-American teams; break a chicken leg. CLEMSON U owns U SCum
chickens! Go Tigers!!
Tiger (
USA - Monday, June 22, 1998 at 17:57:30 (EDT)
Unidentified poster named “Tiger,” on a Clemson University sports web site.
• “Education.”
  - University of Chicago president Robert Hutchins, when
  asked what the university could provide to excite
  students after it dropped the football team.
• “A college racing stable makes as much sense as
  college football. The jockey could carry the college
  colors; the students could cheer; the alumni could bet;
  and the horse wouldn’t have to pass a history test.”
  - Hutchins again, when asked to assess the consistency
  of intercollegiate football with the university’s mission.
“The NCAA television plan on its face constitutes a restraint
upon the operation of a free market, and the District Court's
findings establish that the plan has operated to raise price
and reduce output, both of which are unresponsive to
consumer preference. Under the Rule of Reason, these
hallmarks of anticompetitive behavior place upon the NCAA
a heavy burden of establishing an affirmative defense that
competitively justifies this apparent deviation from the
operations of a free market. The NCAA's argument that its
television plan can have no significant anticompetitive effect
since it has no market power must be rejected.”
- NCAA v. Board of Regents of U. of Oklahoma, 1984, 468
U.S. 85; 104 S. Ct. 2948 .
"It is clear from the evidence that were it not for the NCAA controls,
many more college football games would be televised. This is
particularly true at the local level. Because of NCAA controls, local
stations are often unable to televise games which they would like to,
even when the games are not being televised at the network level. The
circumstances which would allow so-called exception telecasts arise
infrequently for many schools, and the evidence is clear that local
broadcasts of college football would occur far more frequently were it
not for the NCAA controls. This is not a surprising result. Indeed, this
horizontal agreement to limit the availability of games to potential
broadcasters is the very essence of NCAA's agreements with the
networks. The evidence establishes the fact that the networks are
actually paying the large fees because the NCAA agrees to limit
production. If the NCAA would not agree to limit production, the
networks would not pay so large a fee. Because NCAA limits
production, the networks need not fear that their broadcasts will have to
compete head-to-head with other college football telecasts, either on
the other networks or on various local stations. Therefore, the Court
concludes that the membership of NCAA has agreed to limit production
to a level far below that which would occur in a free market situation."
- NCAA v. Regents of OU.
  We find that the problems of big-time college sports have
grown rather than diminished. The most glaring elements of
the problems outlined in this report - academic
transgressions, a financial arms race, and commercialization
- are all evidence of the widening chasm between higher
education's ideals and big-time college sports.
  Clearly, more NCAA rules are not the means to restoring
the balance between athletics and academics on our nation's
campuses. Instead, the Commission proposes a new "one-
plus-three" model for these new times - a Coalition of
Presidents, directed toward an agenda of academic reform,
de-escalation of the athletics arms race, and de-emphasis of
the commercialization of intercollegiate athletics.
  - Knight Foundation Commission on Intercollegiate
Athletics, June 2001
Problems associated with big-time college

•   Admissions standards
•   Displacement of other students
•   Low graduation rates
•   Academic dishonesty
    – Cheating on assignments
    – Easy courses
• Point shaving.
Adjusted Admission Advantages
GPA Bottom Third of Class
Athlete SAT Divergence
Fig. 2.6b – Division III Athlete
2.6a Ivy League Athlete
Fig. 2.6c Division IA Private Athlete
Fig. 2.6d Division IA Public Athlete
Recent efforts to reform admission
• 1986: Proposition 48 required that admitted athletes
  have 2.0 high-school GPA and at least 700 (15) on SAT
• 1992: In response to protests, Proposition 16 allows
  some tradeoffs between standardized-test scores and
• 1999: In Cureton et al. v. NCAA, a federal district judge
  holds that Prop. 16 violates federal civil-rights laws on
  “disparate impact” grounds.
• Although decision is later overturned on procedural
  grounds, a federal appeals court without overturning the
  reasoning that led to the decision. In 2002, the NCAA
  drops any minimum SAT score requirement.
• At selective co-ed liberal-arts schools in one
  study, athletes are one-third of male and one-
  fifth of female students.
• At same schools walk-ons are almost
• According to one study in Social Science
  Quarterly, to be an athlete is worth about 200
  SAT points at five elite private colleges. (For
  comparison, being a legacy is worth 160 points,
  being black is worth 230 points, being Hispanic
  is worth 185 points, and being Asian is worth -50
Displacement – Graduation rates of athletes and
other students, 1996-97 freshman cohort

• Overall: 59%
• Athletes: 60% overall, 70% female, 55%
• Basketball: 44% overall, 52% female, 41%
• Football: 54%
Six-Year Graduation Rate
Academic dishonesty - cheating
• “In the two years I was there, I never did
  anything. The coaches knew. Everybody
  knew. We used to make jokes about it. ... I
  would go over there some nights and get,
  like, four papers done. The coaches would
  be laughing about it.”
  - Russ Archambault, Minnesota basketball
  player, in The Cincinnati Enquirer, 3/11/99
Academic dishonesty – joke classes.

Excerpts from final exam, Jim Harrick Jr.'s “Coaching Principles and
Strategies of Basketball” class, Fall 2001, University of Georgia.
•   1. How many goals are on a basketball court?
•   a. 1
•   b. 2
•   c. 3
•   d. 4
•   2. How many players are allowed to play at one time
    on any one team in a regulation game?
•   a. 2
•   b. 3
•   c. 4
•   d. 5
The Harrick final (continued)

• 3. In what league to (sic) the Georgia Bulldogs
• a. ACC
• b. Big Ten
• c. SEC
• d. Pac 10
• 5. How many halves are in a college basketball
• a. 1
• b. 2
• c. 3
• d. 4
The Harrick final (continued)
• 8. How many points does a 3-point field goal account
  for in a Basketball Game?
• a. 1
• b. 2
• c. 3
• d. 4
• 11. What is the name of the exam which all high
  school seniors in the State of Georgia must pass?
• a. Eye Exam
• b. How Do The Grits Taste Exam
• c. Bug Control Exam
• d. Georgia Exit Exam
Given substantial reputational
costs, why have athletics?
• Alumni donations (McCormick and Tinsley,
  “Athletics and Academics…”)
• To distract students while research is
  emphasized (Sperber)
• To promote undergraduate enrollment
  (Osborne; McCormick and Tinsley,
  “Athletics vs. Academics…”)
Economic idea - signaling
• Signaling occurs when you have a desirable
  characteristic that is unobservable to someone who
  values it. You can signal when your characteristic can
  be proven to the observer by engaging in an activity that
  is too costly for someone without the characteristic You
  might be a potential high-quality employee, in contrast to
  other low-quality applicants, but the employer can’t tell
  which you are just from the qualifications he can
  observe. But if only high-quality applicants can graduate
  from college, you may incur the expense of a degree,
  even if what you learn has no relevance for the job for
  which you are applying.
McCormick and Tinsley, “Athletics and
• Colleges provide two types of value to students:
• Production (human) capital: Improvement in students’ skills
  enable them to earn more money after they graduate.
• Consumption capital – College is enjoyable, and the memories of
  these experiences provide utility over the alum’s entire life.
• Once a student graduates, his college can still make his degree
  valuable by signaling. The need to signal occurs when some
  valuable attribute of a seller is unobservable, but there is some
  costly procedure that he can pay for that can differentiate
  between high-value and low-value sellers. In this case, athletics
  are said to be a way of signaling that the university intends to
  continue maintaining Clemson academic quality, and hence a
  way to encourage alums to continue to donate.
McCormick and Tinsley, “Athletics and
Academics…” (continued)
• Test of above hypotheses:
• Regress per-alum contributions to Clemson U.
  athletic programs in each S. Carolina county.
• Findings: Contributions are:
- positively and significantly correlated with
  income (because richer people give more);
- positively and significantly correlated with
  number of farms in county (because Clemson is
  first and foremost an agricultural university,
  meaning that farmers have more incentive to
  preserve the value of the Clemson name);
McCormick and Tinsley, “Athletics and
Academics…” (findings, continued)
• Contributions are:
- negatively and significantly correlated with
  population; small towns don’t need the signal
  value of a college degree as much;
- positively and significantly correlated with
  athletic contributions.
- The last finding suggests that academics and
  athletics are complements, not substitutes.
  People are willing to contribute to both of them
Sperber: “Beer and Circus”
• Thesis: major research universities value
  research. They are hostile to undergraduate
  teaching, which they view as a distraction, but
  need the tuition money, especially as state
  funding is declining. Big-time athletics is a
  distraction that keeps the tuition money rolling in
  while allowing the faculty to put out second-rate
  teaching without student complaint, thus
  enabling faculty to concentrate on research.
Sperber’s evidence that undergraduate education is de-
emphasized at universities with athletics, especially, Big
State Universities

• Star professors get lower teaching “loads.”
• Despite a huge surplus of Ph.D.s in most fields, more and
  more universities create and promote doctoral programs.
• Many courses, especially ones with many undergraduates,
  are mostly taught by adjuncts, “gypsy faculty” and graduate
• Principles courses, the most basic and important
  knowledge in any field, are the most likely to be taught in
  huge sections with little faculty contact.
• At many Big State Universities (BSUs), high grades are
  traded for low expectations.
• Small honors programs are emphasized in university
  publicity, when in principle this is what should be offered to
  all undergraduates.
“The Honors Program exists to serve the needs of capable,
hardworking, ambitious students who want to make the most of
their undergraduate education. In addition to offering Honors
classes, the Honors Program provides several other services for
Wright State's outstanding students. You may be surprised to
learn of some of the things the Honors Program can do for you.
- Small classes
- Selective enrollment
- Priority registration
- Student lounge and study area
- Special advising
- Strong peer group
- Honors housing
- Opportunities for travel, leadership development, and community
        From the WSU Honors Program Web site.
How to distract undergraduates
• College sports provides utility via game attendance,
  watching games on television, celebrating after victories.
  Many students in Sperber’s surveys report that sports
  was the best part of college at BSUs.
• Alcohol is an independent distraction in its own right, and
  interacts with sports and the Greek system.
  - 80% of fraternity members “binge drink” at least
  occasionally and they average 20.3 drinks/week in one
  - There are 94 liquor stores within one mile of the
  University of Iowa campus.
  - In regression analysis by others, best predictors of
  alcohol use and abuse in campus are big dorms, the size
  of the Greek community and Division I status.
“Party, Party, Party [at LSU]: Nearly every [student]
organization on campus hosts parties throughout the
year…[For football weekends] all of the campus streets
are closed to accommodate the massive number of
people tailgating, drinking, and partying…Such frenetic
activity and enthusiasm extend to all aspects of student
life at LSU, and often preclude more serious activity like
        What is a typical weekend schedule? Friday – drink,
fall asleep in someone else’s bathtub; Saturday – leave
bathtub, watch the game, drink; Sunday – drink lightly.”
        - The Insider’s Guide to the Colleges, 2000 edition.
“Every semester here I have encountered a professor
who uses an overhead projector and writes
continuously on it for the whole class, every class. No
questions allowed, no eye contact made. I always
feel compelled to ask the profs why they do not simply
hand out all the notes they’re going to write on the
overhead at the beginning of the semester, and just
let the students show up for the tests? Not one of
these instructors has ever answered the question.
They just walk away from me.”

- One Indiana University student’s response to Prof.
Sperber’s survey.
Top Party Schools, 2003, Princeton Review

•   Colorado
•   Wisconsin
•   Indiana
•   Illinois
•   Washington and Lee University
•   Texas-Austin
•   The University of the South
•   DePauw University
•   Saint Bonaventure
•   Florida
McCormick and Tinsley (Athletics vs. Academics)

•         Question: Is big-time college athletics negatively related to
    academic quality?
    -     Test: Regress SAT scores of incoming freshmen (measure of
    student quality) on membership in big-time athletic conferences and
    on other measures of school quality.
•   Results:
•   Tuition is positively and significantly related to SAT scores.
•   Professors’ salary is positively and significantly related.
•   Age of the school is positively and significantly related.
•   University endowment per enrolled student is positively and
    significantly related.
•   Student/faculty ratio is negatively related, but only marginally
•   Membership in a big-time athletic conference is positively and
    significantly related.
Conclusion: While the market for
universities appears to respond to
consumer desires, athletics also seems to
perform an advertising function. Athletic
effort and academic quality are friends, not
Osborne, “Motivating College
• Question: What is the relation between
  athletic spending and school’s
Top 20 D-I schools, athletic spending per undergraduate student, 2002
Wake Forest
Holy Cross
Oregon St.
Miami (FL)
Bottom twenty D-I schools, athletic spending per undergraduate
S. Alabama
Long Beach St.
Florida International
Florida Atlantic
Southwest Texas St.
Central Michigan
Chicago St.
Cal St.-Fullerton
UC-Santa Barbara
UT-San Antonio
SE Louisiana
St. Bonaventure
• Test: Regress tuition per student (a measure of ability to pass
  price along) on several university features: athletic, teaching
  and school’s own research spending.
• Result: It is positively and significantly related to teaching and
  athletic spending, not significantly related to research
• Test: Regress SAT scores of applicants on spending
  measures, joint effect of teaching and research spending and
  public-university status.
• Result: Teaching, research and athletic spending are positive
  and significant. But the joint effect of teaching and research is
  negative, suggesting that more spending on research, while it
  raises SAT scores by itself, lowers the effectiveness of
  teaching. Implication: research effort comes at the expense
  of teaching.
Economic idea: cartel
 A cartel consists of independent producers
 who cooperate to restrain output or
 increase price rather than compete.
 Examples: OPEC, mafia garbage
 contracting in New York City.
Cartels are subject to defections because of
the prisoners’ dilemma problem. This
problem occurs when breaking an
agreement is profitable for each party even
if they would both be better off if they both
stuck to the agreement.
Does the NCAA restrict output or
• Brown (1996, on reading list in section
  IVC) estimates that a premium college
  football player generated between
  $400,000-$600,000 in 1988-89.
• The same author estimates elsewhere that
  in 2003-4 Jameer Nelson generated
  perhaps $1,000,000 in revenue for St.
  Joseph’s University.
How does the NCAA police its cartel
(DeBrock and Hendricks)?

• Individual schools make more money from
  better teams.
• But the total revenue pool increases when
  there is more competitive balance.
DeBrock and Hendricks (continued)
• Implications of previous assumptions:
• (1) The NCAA should admit more members up
  to a point, but then place entry barriers. The
  more members there are, the more unbalanced
  competition becomes beyond a certain point.
• (2) In addition, once the number of members
  has been decided on the NCAA will enact
  minimum quality standards and maximum
  quality standards.
DeBrock and Hendricks (continued)
• To achieve (1), the NCAA creates
  separate divisions and requires more
  investment in college athletics to move up
  into a higher division.
• To achieve (2), the NCAA limits
  scholarships and coaching staff.
Are college athletes exploited? In
economics, the only way to interpret
this question is to ask whether their
pay is less than their marginal
product. Outside of economics, one
might ask whether student-athletes
are led to make bad choices because
they are deceived by college athletic
Brown (athletes’ MRPs)
• Question 1: What is the MRP of a premium college football
• To answer, regress a team’s revenues from ticket sales, TV
  and radio, donations and miscellaneous sources on number
  of premium players (which the author defines as players
  drafted by the NFL), on the size of the city in which it is
  located, on its past AP poll ranking and on the ranking that
  season of its opponents.
• Opposition quality, NFL picks and the team’s own ranking in
  previous years are statistically significant. A premium player
  generates between $400,000-$600,000 in annual revenue.
• His compensation consists of tuition, books and the increase
  in his value acquired from the human and physical capital he
  acquires while he is there.
Brown (continued)
• Another finding is that the number of premium players
  obtained is positively related to the percentage of its
  team admissions that are “special authority” -- i.e. lower-
  standards -- admissions, and negatively related to team
  high-school GPA.
• Admitting one player (without yet knowing whether he
  will be premium) on special authority increases revenue
  by $90,000-$126,000.
• Decreasing the team’s high-school GPA requirements by
  0.21 increases revenues between $800,000-$1,120,000.
• Inference: colleges have a clear incentive to admit
  players with less chance of doing well in school in order
  to improve team revenues.
But what about athletics overall and
student prospects? (Long and Caudill)
• Athletics and future earnings potential:
• On the one hand, college athletics may divert
  time in college from classroom-based human-
  capital acquisition, due to demands of practice
  and games.
• On the other hand, college athletics provides
  some human capital on its own, e.g. enhancing
  self-discipline, teamwork.
• In addition, even if college athletics does not
  provide this human capital, it may serve as a
  costly signal of greater existing possession of
  the above traits.
Long and Caudill (continued)
• Test: Regress annual income on various demographic
  characteristics, extent of education, college GPA, self-
  reported measures of ambition and goals, and whether or not
  student received a varsity letter in athletics.
• Result: For males, earnings positively and significantly
  associated with athletics, being married, having children,
  working for a big company, college grades, having a graduate
  degree, and personal characteristics. They are negatively
  related to working part-time. Black males also earn less than
  others. Athletics increases income by about 4%. For
  females, athletics is statistically insignificant. The other
  variables have the same effect, except that black females
  earn higher wages than non-black females after standardizing
  for other characteristics.
Long and Caudill (continued)
• With respect to graduation probability, athletics,
  high-school grades, ACT scores, parental
  income and education and personal
  characteristics are positively and significantly
  related. For females, athletics is again
  significant in addition to the other variables that
  are statistically significant for men.
• Conclusion: for athletics overall, it is hard to
  argue that exploitation exists in terms of the
  university’s athletic goals distracting students
  from improving their earnings prospects.
Theories of international trade
• Comparative-advantage theory says that nations
  have different levels of technology or different
  resource bases. Each nation maximizes its
  prosperity by specializing in what they do
  relatively well.
• Product-lifecycle theory says that all advanced
  nations have production structures that are
  roughly the same, and all pass through the
  same stages on the way to becoming modern.
Comparative advantage in the
Olympics (Tcha and Pershin)
• Over three Olympics, define six sport
  groups: swimming, track and field,
  weightlifting, ball games, gymnastics,
• Specialization is defined as the
  percentage of medals a country wins in
  one group divided by the total medals
  available in that group.
Top performers overall
Swimming         Track      Weights   Ball Games
Costa Rica       Bahamas    Iran      Ghana
Hong Kong        Ethiopia   Turkey    India
Iceland          Jamaica    Israel    Indonesia
Ireland          Namibia    Greece    Argentina
New Zealand      Zambia     Algeria   Lithuania

Tcha and Pershin (continued)
• Result (1) Comparative advantage in
  Olympic success. Left-hand variable is
  relative specialization in medals. Right-
  hand variables are land mass, coast
  length, altitude, GDP, GDP per capita, and
  dummy variables for former communist
  countries, Asia and Africa.
• Swimming: only Asia dummy is significant.
• Track: Land mass, altitude, temperature, per capita
  income and Africa dummy are positive and significant.
  Coastline is negative and significant.
• Weights: Temperature, GDP, Asia and Africa are
  positive and significant. Altitude, GDP per capita,
  communist, Asian and African dummies are negative
  and significant.
• Ball games: Population is positive and significant.
• Gymnastics: Communist dummy is positive and
Tcha and Pershin (continued)
• Result (2): Richer nations have less
  variance in their specialization than poor
  ones do. Specifically, the variance in
  comparative advantage across sports in a
  given country is a negative and significant
  function of its per capita GDP. Poorer
  nations specialize in only a small number
  of sports, rich countries spread out their
  success more.
Osborne, “Baseball’s
• Question: Can comparative-advantage or
  product-lifecycle theory explain the
  statistical productivity patterns of different
  countries that contribute to the major
• Statistical problem: unit of analysis.
  Should all players be equally weighted, or
  should total statistical productivity for a
  country be simply added together?
Foreign-born players, 1950, 1970, 2002
                          1950           1970   2002
Aruba                     0              0      2
Australia                 0              0      3
Canada                    6              7      10
Colombia                  0              0      3
Cuba                      9              24     11
Dom. Republic             0              16     74
Japan                     0              0      11
Korea                     0              0      2
Mexico                    2              6      18
Neth. Antilles            0              0      2
Nicaragua                 0              0      2
Panama                    0              8      7
Puerto Rico               1              23     38
Venezuela                 1              11     38
U.S. Virgin Islands       0              3      1
San Pedro de Macoris, 1962-2002

Positions               Games
Right-handed pitchers    2053
Left-handed pitchers       73
First Base               1071
Second base              4926
Shortstop                5644
Third base               1791
Outfield                 6075
Catcher                    79
Designated hitter        1546
Measuring Specialization – A Country’s
Player’s At-Bats/Innings Pitched as
Test 1: Add all player productivity
together. A player with 3000 AB will
therefore contribute much more to a
country’s total productivity than a player
with 100 career AB. What is produced,
in international-trade terms, is therefore
major-league statistical output.
Test 2: Assume that what is produced is major-
league players, regardless of career length. Take
each player’s productivity as a separate
unweighted observation.
Relative at-bats/innings pitched

              1940-59      1960-79    1980-2002

Overall       3.826486     3.77804    3.911677
Dom. Rep.                  1.986451   1.358929
Mexico                     1.352569   0.263868
Venezuela                  13.31242   2.002334
Puerto Rico   1.112291     3.210672   3.004303
Cuba          0.991405     1.654358   1.283594
Canada        0.530192     0.201441   0.387893
Osborne - hitting for average and
• Because nations do not generally stay on
  the same side of 1 in their relative
  production of batting average and HR/AB,
  the conclusion is that these are not skills
  governed by comparative advantage.
• Only possible exception is Puerto Rico
  batting average, but overall comparative-
  advantage model performs poorly by these
  measures, in this specification.
Osborne (continued)
• But by method 2, in which each player is analyzed
  separately, results are different.
• Using a statistical technique called analysis of variance,
  it is shown that there are sustained country differences in
  average hitting, and changes over time in power hitting.
  The latter is a skill developed later, as product-lifecycle
  theory would predict.
• Specifically, Puerto Rico and Venezuela consistently
  produce more hits than expected, and Canada and
  Mexico produce fewer. Every nation except Canada
  produces more HR/AB from one interval to the next.
Osborne (continued) –
specialization within pitching
• Test 1 : specialization in handedness. No pattern is
  shown. The only strange result is that the Dom. Rep.
  produces surprisingly few left-handers between 1980-
• Specialization in strikeouts and walks: again, analysis of
  variance shows no detectable pattern.
• Same holds for Games started/Total appearances, a
  measure of specialization in starting.
• Conclusion: The differences in human capital required to
  produce different types of pitching do not appear to be
Table 8
Lefties and righties
          Dom. Rep.          Mexico    Venezuela          P.R.      Cuba     Canada
1940-59                                                             7/31     4/19
                                                                    (.226)   (.211)
1960-79 3/20                1/14*                          6/16     3/22     6/25
        (.150)              (.071)                         (.375)   (.136)   (.240)
1980-02 22/136**            26/41* 11/41         9/36      4/16     10/30
        (.162)              (.366)      (.268)   (.250)    (.250)   (.333)
Note:   * denotes statistical significance at ten-percent level.
        ** denotes statistical significance at one-percent level.
Specialization by position
• Test: Use multinomial distribution to see if
  distribution of players is statistically
  different from random chance.
Table 10
Multinomial Х2 components, fielding
            Pitchers                1B/3B     2B/SS       OF         C           DH      Total
1940-59     1.4                     0.233     .305        .001       .343        N/A     2.282
(n = 56)
1960-79     .694                    .601      7.079(+)    .086       .006        .624    9.090
(n = 52)
1980-02     .75                     .004      1.277       .297       .355        .243    2.926
(n = 30)
Puerto Rico
            Pitchers                1B/3B     2B/SS       OF         C           DH      Total

1940-59     .167                   .25        .694        .375       .014        N/A      1.500
(n = 15)
1960-79     5.143(-)               1.052      9.163(+)    3.665(+)   0.917       .030     29.970***
(n = 70)
1980-02     9.345(-)               1.792(-)   18.608(+)   .004       12.330(+)   0.747    31.729***
(n = 149)
         Pitchers 1B/3B     2B/SS      OF         C       DH     Total
1940-59 3.003(+) 1.633(-)   4.800(-)   0.450      0.067   N/A    9.953*
(n = 32)
1960-79 10.548(+) .706      1.676      3.841(-)   .852    .396   18.017***
(n = 33)
1980-02 5.518(+) 4.281(-)   2.251(-)   .114       .066    .019   12.249**
(n = 49)
         Pitchers 1B/3B     2B/SS      OF         C       DH     Total
1960-79 .236      .063      .595       2.442(-)   .602    .001   3.939
(n = 26)
1980-02 7.014(+) .092       1.049      6.776(-)   1.400   .585   16.966***
(n = 64)
Dominican Republic
          Pitchers   1B/3B      2B/SS      OF         C          DH         Total
1960-79 1.329        .323       6.190(+)   .708       1.282      .708       10.540*
(n = 59)
1980-2002 .918       14.909(-)34.197(+)    3.472(-)   .137       4.989(+)   58.622***
(n = 306)
          Pitchers   1B/3B      2B/SS      OF         C          DH         Total
1960-79 1.376        .002       4.960(+)   .089       .198       .252       6.877
(n = 21)
1980-02 .229         2.284(-)   17.329(+) 3.569(-)    4.097(+)   4.158(-)   31.666***
(n = 126)
Positional specialization patterns
• The Dominican Republic and Venezuela
  produce middle infielders.
• Puerto Rico and Venezuela produce
• Canada produces pitchers.
• Puerto Rico produces outfielders.
Osborne (conclusions)
• The pattern of specialization in pitching
  versus hitting clearly conforms to a
  comparative-advantage interpretation, as
  does specialization in fielding positions.
• Specialization within pitching and hitting is
  not detectable, although power hitting is a
  late-stage industry in the product-lifecycle
Economic idea: An efficient market
occurs when profits (or expected
profits) have been eliminated by
competition. Efficient financial
markets – e.g., stock markets,
currency markets – occur when no
trading strategy offers expected
profits over any period of time.
This is a question of information. Do trading
markets collectively reveal all available
information? We must distinguish between
perfect information, which eliminates all
uncertainty beforehand, and symmetric
information, which allows uncertainty, but
where all traders, observing the market price,
have the same expected probability of making
money. Symmetric information does not
eliminate uncertainty, but it means that all
available information is revealed via the
asset’s price. Thus, no strategy can expect to
earn profits.
Economic idea: Rationality. Rational
economic agents have consistent preferences,
and act in accordance with those preferences,
given those constraints. Most propositions in
economics about choices people make and
the effects of various kinds of economic policy,
assume that decision-makers are rational.
When they are irrational in a systematic way,
markets need not be efficient.
Example of irrationality: the magical-thinking
prisoners’ dilemma

• In a prisoners’ dilemma game, when told that
  partner has defected in advance, 97% of players
• When told that partner has cooperated in
  advance, 16% cooperate.
• When told that partner has already chosen
  strategy, but not told what strategy is, more
  people (37%) choose cooperate.
• Often interpreted as a belief in “magical thinking”
  – that a decision to cooperate could cause the
  other player to choose cooperate, even though
  he has already made his choice.
Test 1 of streakiness: Does the probability of a
made shot vary with the fate of the previous shot?
• Mathematically, is Pr(Hitt|Hitt-1) > Pr(Hitt|Misst-1),
  given player’s general probability of making a
• Test: Use last three shots. Compare probability
  of a hit given 0,1,2 or 3 hits in the last three
  shots. If there is streakiness then for any player
  the probability should be increasing in the
  number of previous hits.
• Only for one player out of eight is there any such
Test 2 of streakiness: how many runs of consecutive
misses or makes are there for each player?

• A run is a streak of identical results. For
  example, HHHMMMMH contains three runs;
  HMHMHHMM contains six. Streakiness would
  imply relatively few runs.
• There is no significant difference between actual
  and expected number of runs for eight players
  over 48 home games. One has more than
  expected, i.e. the opposite of streakiness.
But, perhaps the lack of correlation is due to hot
players taking harder shots, or being more
aggressively defensed.
• Test 1: Look for streakiness in free-throw shooting.
• Result: For the Boston Celtics over two seasons no player has a
  statistically significant difference between his chance of making a
  second free throw given that he made or missed the prior one.
• Test 2: Let college basketball players take shots (and bet on the
  results) from a single spot on the floor. There is again no evidence
  of the results of prior shots affecting chances of making the current
• But players mistakenly believe they can predict results – i.e., they
  think they’re hot when they’re not. Players can bet a larger amount
  when they’re more confident they’ll make shot and less otherwise.
  Only 5/26 had a statistically sig. relation between bets and
  outcomes, and one player’s relationship was negative. Overall,
  there was no relation. But the way both players and observers on
  the sidelines bet is highly related to the outcome of the previous
Conclusion: there is no hot hand, but
people often mistakenly think they are
observing one. A clear sign of a sustained
consistent mistake, which is a sign of
irrational behavior.
• Irrationality – college names
Sports betting offers a test of the efficient
markets hypothesis. The price of a stock or
currency is equivalent to the odd on or the
point spread of a game. If betting markets
are efficient, no betting strategy should yield
expected profits. This is a possible result
even with irrationality, as long as some
people are less irrational (even perfectly
rational) and can take advantage of the
mistakes of those who aren’t.
Preliminary question: Do independent
ratings of teams or individuals reflect
their true strength? Answer: only
NCAA Men’s Tournament Records By Seed
       RECORD       WIN %

1      328-94       .777
2      238-103      .698
3      172-105      .621
4      146-106      .579
5      128-108      .542
6      145-106      .578
7      92-108       .460
8      82-107       .434
9      63-107       .371
10     70-107       .395
11     46-106       .303
12     44-104       .297
13     21-83        .202
14     16-84        16
15     4-85         .045
16     0-84         .000

Boulier and Stickler – testing the ability of tennis
seedings to predict match outcomes

• First: Use regression to create expected
  probability of the higher seed winning. If
  information were perfect higher seeds would
  always win. If information is symmetric then the
  probability that a higher seed wins is always
  greater than 0.5, but declines as the higher seed
  is less and less strong. For example, the
  estimated probability that a 1 should beat a 16 is
  higher than the estimated chance that a 5
  should beat a 12.
Boulier and Steckler - Overall winning percentage, 1985-1995, seeds in Grand Slam
tennis tournaments (rankings among seeds in parentheses)
                           Men               Women
1                          .756 (#2)         .891 (#1)
2                          .810 (#1)         .810 (#2)
3                          .705 (#3)         .615 (#4)
4                          .674 (#4)         .565 (#5)
5                          .614 (#5)         .632 (#3)
6                          .600 (#6)         .532 (#7)
7                          .500 (#10)        .544 (#6)
8                          .487 (#11)        .491 (#8)
9                          .577 (#8)         .459 (#9)
10                         .452 (#12)        .411 (#10)
11                         .541 (#9)         .333 (#12)
12                         .594 (#7)         .353 (#11)
13                         .258 (#15)        .276 (#13)
14                         .375 (#14)        .231 (#15)
15                         .421 (#13)        .217 (#16)
16                         .150 (#17)        .273 (#14)
Unseeded                   .163 (#16)        .111 (#17)
Appearances (out of 44) in round of 16 by seeded tennis players, 1985-1995
                       Men’s            Women’s
1                      38 (#1)          43 (#1)
2                      36 (#2)          32 (#2)
3                      33 (#4)          36 (#4)
4                      34 (#3)          38 (#3)
5                      19 (#10)         29 (#6)
6                      22 (#8)          29 (#6)
7                      23 (#7)          32 (#5)
8                      20 (#9)          29 (#6)
9                      24 (#5)          20 (#12)
10                     19 (#10)         22 (#10)
11                     18 (#13)         26 (#9)
12                     16 (#15)         22 (#10)
13                     24 (#5)          21 (#14)
14                     19 (#10)         20 (#12)
15                     12 (#16)         18 (#15)
16                     17 (#14)         16 (#16)
Boulier and Stickler – A Brier score measures how
close to perfection the estimate probabilities are.

• Defined as:

   n = number of observations
   pi = estimated probability of higher seed winning
   di = 1 if higher seed wins, 0 if higher seed loses.
Boulier and Stickler

• 0 is thus a perfect Brier score. Actual Brier
• Women’s tennis 0.140
• Men’s tennis 0.160
• Women’s NCAA basketball: 0.170
• Men’s NCAA basketball: 0.180.
• Conclusion: because of consistent errors in
  seedings as predictors of victory, they do not
  completely use available information.
The basics of gambling markets
• Football and basketball bets are placed against
  point spreads. In other words, you bet as to
  whether the favored team will win by at least a
  certain number of points.
• Bookmakers generally adjust the line – the point
  spread against which you bet – to equalize the
  money bet on each side.
• Bookmakers take a commission for every bet, so
  that bettors have to win roughly 52.5 percent of
  the time to break even.
Zuber, Gandar and Bowers

• Question: do profit opportunities exist in NFL betting
• Solution: regress actual winning margins on statistics
  available before the game is played.
• Problem: For the model to be profitable it must not
  simply look backward, i.e. it must not simply conduct
  bets within the sample. Instead, out of sample tests
  must be conducted, where the model’s predictions are
  used to test games not included in the regression used
  to create the model.
• Solution: come up with a predicted point spread for first
  half of 1983 season, then test it using games in the
  second half of season.
Zuber, Gandar and Bowers
• Home team’s predicted victory margin =
  1.547 + .047*Net yards rushing + .044*Net yards
  passing + .697*Net previous wins – 2.299*Net fumbles –
  2.619*Net interceptions -.424*Net penalties - .217*Net
  pass play percentage -.319*Net number of rookies
• All variables are statistically significant in the expected
• The winning percentage when the model has a
  prediction at least 0.5 points different from the betting
  line is 59 percent, which is economically profitable.
  Conclusion: the market for NFL betting is not efficient.
Gandar, Dare, Brown and Zuber.

• Question: Do markets improve in efficiency as trading
  proceeds? In other words, are profitable opportunities
  eliminated over the course of trading?
• The strong version of the efficient-markets hypothesis
  (EMH) requires that prevailing prices always reflect all
  available information about asset prices, so that opening
  betting lines should never change.
• But a weaker version of EMH contends that prices
  eventually incorporate all available information.
• Implication: There should be no way to make money on
  closing betting lines, although there may be using
  opening lines.
Gandar, Dare, Brown and Zuber
• Finding 1: 80% of opening lines change.
  Of these, 40% move 0.5 points, 31% move
  1 point, 29% move more than one point.
  Implication: strong EMH does not hold,
  unless these movements are random.
• Are they? To test this, see whether
  closing lines better predict results than
  opening lines.
Difference in forecast error, opening and closing lines, NBA, 1986-
Season        Avg. Opening Error          Avg. Closing Error
1985-86              8.85                                8.76
1986-87              9.16                                9.06
1987-88              8.72                                8.68
1988-89              8.99                                8.92
1989-90              8.88                                8.76
1990-91              8.93                                8.87
1991-92              9.06                                8.92
1992-93              9.22                                9.21
1993-94              9.14                                9.01
All seasons          9.00                                8.92
Home team winning percentages versus line for different changes in opening and
closing lines, NBA, 1985-1994
Line change                               Home team beats opening Home team
                                          Line                        beats closing
<-4                                       0.25                        0.46
-3.5                                      0.37                        0.43
-3                                        0.39                        0.49
-2.5                                      0.43                        0.54
-2                                        0.37                        0.43
-1.5                                      0.43                        0.49
-1                                        0.47                        0.51
-0.5                                      0.48                        0.5
0                                         0.51                        0.51
+0.5                                      0.51                        0.49
+1                                        0.53                        0.49
+1.5                                      0.59                        0.52
+2                                        0.58                        0.48
+2.5                                      0.69                        0.56
+3                                        0.52                        0.36
+3.5                                      0.57                        0.47
>+4                                       0.74                        0.69
Gandar, Dare, Brown and Zuber
• Closing lines are more accurate than opening ones. Thus,
  traders capitalize on available information to achieve more
  accurate price.
• When opening line never moves, favored team covers point
  spread almost exactly 50% of time.
• Collectively, when opening line changes in direction of home
  team by any amount, you win by betting on the home team
  against the opening line 54% of time. When the opening line
  changes against the home team, then if you bet against the
  home team on the opening line you win 55% of time. These
  are both statistically significant.
• Conclusion: while this paper does not identify a strategy to
  win against the opening line, it indicates that such
  opportunities exist. Since closing lines do not present
  profitable opportunities, the market moves to efficiency. The
  weak version of the EMH is supported.
NFL betting - Osborne
• Basic approach: instead of regressing
  victory margins on team statistics, simply
  regress them on the bottom line – points
  scored and allowed in previous games by
  the two teams.
Osborne (betting, continued)
• Regress the margin of victory on average points
  scored and allowed by home and road team in
  previous games in that season, starting with
  week six. Use 1980-1990 to estimate results.
• All four variables statistically significant, but the
  R2 (percentage of variance in margins explained
  by variance in points scored and allowed in
  previous games) is small, less than 10 percent.
Osborne (betting, continued)
• The same regression of the betting line on
  these measures of previous performance
  during the season explains over 70
  percent of the variance in that variable.
• Implication: bookmakers use this
  information extensively to set the line, but
  game results are still highly unpredictable.
Osborne (betting, continued)
• To test for profits (as usual) use the model’s predicted margins to
  try to predict the margin in games outside the sample period.
• Betting with the predicted margin overall wins 51.2 percent of the
  time, which is not profitable. But betting with at least a 3-point
  difference between predicted margin and betting line wins 57.5
  percent of the time, and betting with at least a 5-point difference
  wins 54.3 percent of the time.
• In another version of the model (in which regression is updated
  every week), results are roughly the same. However, error in
  model’s predictions declines until bottoming out in roughly week
  15. It then rises in weeks 16-17.
• Implication: over the course of the season, bettors get better at
  predicting game results. However, in last two weeks of season
  many teams have been eliminated or clinched playoff spots by
  then. They engage in many experiments (e.g., trying out new
  players) and this makes games harder to predict.
Biases in sports betting that have been
found in the economic literature
• The favorite/long shot bias: cognitive scientists have discovered that
  people systematically overestimate the chances of some low-
  probability events (e.g., plane crashes, winning the lottery). One
  sports example is the chance that a big underdog will win. Bettors in
  horse racing and football tend to over-bet on long shots and under-
  bet on favorites.
• The referent bias: People tend to evaluate comparisons on the basis
  of the reference for comparison. They might give different answers,
  for example, to the question “By how much do you think the
  Dolphins will beat or lose to the Cowboys” then “By how much do
  you think the Cowboys will lose to or beat the Dolphins?” In each
  case the first team mentioned will be the focus, and people will
  either overestimate its strengths (if it is favored) or its weaknesses (if
  it is the underdog). If betting by one team’s fans dominate, then if it
  is favored the spread will be too high and if it is an underdog the
  predicted loss margin will also be too high.

To top