See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/24074967

Measuring Discrimination in Major League Baseball: Evidence from the

Baseball Hall of Fame.

Article  in  Applied Economics · February 2002

DOI: 10.1080/00036840010034244 · Source: RePEc

CITATIONS READS

21 251

3 authors, including:

Todd Jewell Robert Brown

Texas State University California State University, San Marcos
73 PUBLICATIONS   1,013 CITATIONS    19 PUBLICATIONS   369 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Robert Brown on 23 April 2016.

The user has requested enhancement of the downloaded file.

Applied Economics, 2002, 34, 167 ±177

Measuring discrimination in major league

baseball: evidence from the baseball hall of
fame
R . T O D D JE W E LL * , R O B E R T W . B R O W N { and S C O T T E . M I L E S}
Department of Economics, University of North Texas, Denton, TX 76203, USA,
{ Economics Programme, California State University, San Marcos, CA 92096, USA
and } Verizon, Irving, TX 75028, USA
E-mail: tjewell@unt.edu, rbrown@coyote.csusm.edu , miles.scott@verizon.com

This paper examines the eVects of race on player induction into the National Base-
ball Hall of Fame, concentrating on a player’s ®rst ballot. Past research has found
some evidence of discrimination in voting against retired Major League Baseball
players who were born in a Latin American country and against American-born
black players. This study ®nds that estimates of discrimination in Hall of Fame
voting are sensitive to speci®cation. However, we do ®nd limited evidence that
retired players who are both black and Latin face some voting discrimination.
Furthermore, the results show that race does not seem to aVect whether a player
actually receives enough votes to get into the Hall of Fame on his ®rst ballot.
Instead, it appears that any discrimination in voting is concentrated among those
players who would not have received enough votes to enter the Hall of Fame based
solely on their career statistics.

I. INTRODUCTION researchers to investigate the eVect of race on labor

Each year, members of the Baseball Writer’s Association of
America (BBWAA) vote on the most signi®cant perform-
ance-related award in Major League Baseball (MLB),
induction into the National Baseball Hall of Fame. In
this paper, we examine whether a retired player’s race
aVects his probability of entering the Hall of Fame. If
voting members of the BBWAA have biases against min-
ority-race players, then these players may be less likely to
enter the Hall of Fame. In general, the sports industry has
proven to be a fruitful area in which to test for racial biases
and discrimination since detailed data exist on player and
team performance. In the context of Hall of Fame voting,
the availability of detailed measures of player performance
allow us to isolate the eVect of a player’s race on the
BBWAA’s voting outcome.
Unlike many other industries, available data on
player productivity in the sports industry have allowed
market outcomes such as salaries and hiring. Kahn and
Sherer (1988) and Koch and Vander Hill (1988) ®nd
that white players in the National Basketball Association
(NBA) are paid signi®cantly more than equally-skilled
black players. Using more recent data, Hamilton (1997)
®nds that the premium paid to white NBA players is
most notable at the upper end of the salary distribution.
Using National Football League (NFL) data, Kahn
(1992) ®nds no evidence of statistically signi®cant salary
discrimination. Several studies have also investigated
race-based salary diVerentials in MLB. Kahn (1991)
reviews these studies and ®nds little evidence of
signi®cant salary discrimination. Singell (1991) shows
that black ex-players are less likely to be hired as
coaches in MLB than their white counterparts. Brown
et al. (1991) and Burdekin and Idson (1991) show that
players are sorted to NBA teams partially based on
their race.

* Corresponding author.
Applied Economics ISSN 0003±6846 print/ISSN 1466±4283 online # 2002 Taylor & Francis Ltd 167
http://www.tandf.co.uk/journals
DOI: 10.1080 /0003684001003424 4
168
Following Becker (1971), economists generally think of
discrimination as emanating from one of three sources:
employers, employees or customers. Most evidence points
to customers as the primary source of any salary and hiring
discrimination in the sports industry. For example, if white
customers are willing to pay more to see white athletes and
the majority of fans are white, then teams with race-neutral
preferences may respond by paying white players more
than non-white players or by preferentially hiring white
players. Results using NFL data imply that white players
earn higher salaries in largely white metropolitan areas,
while non-white players earn higher salaries in largely
non-white areas (Kahn, 1992). Evidence from MLB sug-
gests that teams with more black players have lower atten-
dance and revenue (Scully, 1973; Gwartney and Haworth,
1974; Sommers and Quinton, 1982). Anderson and La
Croix (1991), Gabriel et al. (1995, 1999) and Nardinelli
and Simon (1990) examine the market for baseball cards
and ®nd evidence that baseball card purchasers discrimi-
nate against both Latin and black players in favour of
Anglo players. Kahn and Sherer (1988) show that NBA
teams with more black players have lower attendance.
Furthermore, Brown and Jewell (1994, 1995) show that
fans of college basketball teams pay a premium to see
white players.
Race-based discrimination in voting on performance-
related awards has received relatively little attention from
researchers. Hanssen and Anderson (1999) study the be-
haviour of fan voting for the MLB All-Star Game. If
fans are willing to pay more to see players of their own
race, then they may also be more likely to vote for players
of their own race as All-Stars. The authors ®nd that white
players tend to receive more votes than equally-quali®ed
black players, although the vote diVerential has narrowed
over time. This result is important to player compensation
since many baseball players have incentive clauses in their
contracts tied to performance-relate d awards. In addition,
performance-relate d awards can give a player national
recognition that may lead to greater compensation through
endorsements.
In the same way that an appearance in the MLB All-Star
Game provides an active player greater recognition, induc-
tion into the National Baseball Hall of Fame provides a
retired player greater recognition and possibly greater
earning power. Each year the BBWAA chooses among
eligible retired MLB players to induct into the Hall of
Fame, any player receiving a 75% vote being elevated to
Hall of Fame status. According to Hall of Fame rules,
`voting should be based upon the player’s record, playing
ability, integrity, sportsmanship, character and contribu-
tions to the team(s) on which the player played.’ In evalu-
ating a player, each writer±voter attaches some subjective
importance to these factors and then decides whether a
player is worthy of being inducted into the Hall of Fame.
Thus, racial bias among writers will impact a player’s
R. T. Jewell et al.
chances of induction. In addition, if writers have race-
based preferences, they may indirectly aVect the prefer-
ences of fans (customers) and, therefore, be a source of
any observed discrimination in MLB.
Findlay and Reid (1997) are the ®rst to examine whether
the voting members of the BBWAA exhibit racial biases in
their voting behaviour. The authors categorize players as
`Latin’ if their birthplace is a Latin American country,
`black’ if they are of African descent and American-born,
or `white’ if they do not fall into the ®rst two categories.
The authors ®nd limited evidence that African American
and Latin American players are treated diVerently than
white players in Hall of Fame voting. In particular, there
is some evidence of discrimination against Latin American
players, though it seems to have diminished over time.
Alternatively, African American players receive favourabl e
treatment in voting, but this result is inconsistent across
speci®cations. Desser et al. (1999) also study the eVect of
race on voting for the Hall of Fame. Using the same racial
categories as Findlay and Reid, the authors ®nd that both
Latin American and African American players receive
fewer votes than equally-quali®ed white players.
However, these results are only marginally signi®cant and
of relatively small magnitude. According to the authors,
the estimated eVect of discrimination is too small to sig-
ni®cantly alter the racial composition of the Hall.
This paper extends existing research on voting for the
Hall of Fame in several ways. First, more extensive meas-
ures of race are utilized. Speci®cally, a distinction is made
between American-born black players and black players
born in Latin American countries. Second, the sensitivity
of the estimates to speci®cation changes are examined by
using an alternative estimation technique, by using a dif-
ferent measure of time, and by including additional meas-
ures of player performance. Third, the present authors’
model is used to predict which players are most aVected
by race-based discrimination. The results presented here
indicate that minority-race players are not less likely than
other players to be in the Hall of Fame. However, the
measured eVect of race on Hall of Fame voting seems to
be extremely sensitive to model speci®cation.
II. DATA AND METHODOLOGY
The data for this study consists of 290 non-pitchers
(hitters) who were eligible for election into the Hall of
Fame over the years 1962±1998. The sample includes 200
hitters who received votes and 90 hitters who appeared on
a ballot but did not receive votes from 1962±1998.
Information on voting rules and the total number of
votes in annual elections are from the Hall of Fame’s
web page (www.baseballhalloVame.org). Each eligible
player’s name and number of votes received are from the
Total Baseball web page, an online version of the o cial
Measuring discrimination in major league baseball
encyclopedia of MLB (www.totalbaseball.com). Players
that were placed on ballots but did not receive votes (no-
vote players) were found in various issues of The Dallas
Morning News, The New York Times, USA Today and
The Boston Globe. Player statistics are collected from the
Total Baseball web page. 1
A player must have played in MLB for at least 10 sea-
sons, must have been retired for at least 5 years, and must
have been an active player within the last 20 years to be
considered for election into the Hall of Fame by the
BBWAA.2 Eligible players must then be nominated by
two of the six members of the BBWAA Screening
Committee to be placed on the ballot in a given year.
Players who are placed on the ballot must be named on
75% of the ballots returned by the writer-voters to enter
the Hall of Fame. Voters must have been active as baseball
writers and members of the BBWAA for at least 10 years.
Although the actual number is a closely guarded secret,
there are over 500 voting members in the BBWAA, and
voting is anonymous. Each year, the number and composi-
tion of voters varies. In addition, each writer is able to vote
for up to 10 diVerent players.
Since subjective measures such as integrity, sportsman-
ship and character are di cult to measure, we assume that
each writer casts a vote for each player based on that
player’s career performance statistics and other noteworthy
accomplishments. In addition, a writer’s vote may be par-
tially in¯uenced by the player’s race. Let i represent voters
and j represent players. A voter’s evaluation of any player,
EVALij , is a continuous function of the career statistics of
player j …STATSj † and the race of player j …RACEj †.
Assuming that EVALij is a linear function of STATSj
and RACEj results in Equation 1:
EVALij ˆ STATSj ¬i ‡ RACEj ­ i …1†
where coe cients ¬i and ­ i are in¯uenced by voter prefer-
ences. EVALij is unobserved; instead we observe a variable,
VOTEij , equal to 1 or 0 (yes or no) based on the following
voting rule:
) play in both leagues may be perceived negatively by writers
VOTEij ˆ 1 if EVALij >ˆ HOFi
…2†
VOTEij ˆ 0 if EVALij < HOFi
where HOF i is voter i’s evaluation of the minimum level
necessary for Hall of Fame status.
VOTEij is a binary response (yes or no) for each voter.
However, individual votes are unobserved in the voting
169
data; instead, the total number of `yes’ votes each player
receives are observed. Therefore, the observed dependent
variable for each player is the percentage of votes received
out of total votes …Pj †.3 This type of data is referred to as
grouped or proportions data. Grouped-data models are
normally estimated with maximum-likelihood techniques
or with weighted least squares, the two methods being
asymptotically equivalent. The minimum logit chi-square
method is employed (Maddala, 1983) which consists of
transforming the dependent variable to a logistic and
applying weighted least squares to the following equation:
ln …Pj =1 ¡ Pj † ˆ STATSj ¬i ‡ RACEj ­ i ‡ "i …3†
with weights equal to …NPj …1 ¡ Pj ††1=2 where N equals
total votes cast. The random error term is "i . The logistic
transformation is also discussed in Fox (1997) as a way to
`linearize’ the relationship between Pj and the dependent
variables. Both Findlay and Reid (1997) and Desser et al.
(1999) utilize a similar logistic transformatio n in the voting
variable. A potential problem, however, is that for obser-
vations in which Pj is zero (i.e. no-vote players) the depen-
dent variable cannot be transformed since the natural log
of zero cannot be computed. This transformation problem
is similar to sample selection and will be treated as such in
the estimation discussed below.
The vector STATSj contains player j’s career perform-
ance measures: seasons played (SEASONS), home runs
(HOMERS), runs batted in (RBIS), batting average
(BATTING), stolen bases (STEALS), and World Series
championship appearances (WORLD). The performanc e
measures are expected to positively aVect a voter’s evalua-
tion of each player, an increase in any of these measures
raising the probability of any writer voting for a player. A
dummy variable for players who spent at least 10% of their
careers in both leagues (BOTH) is included. Spending time
in both leagues may increase a player’s visibility and, thus,
his number of votes received. Alternatively, players who
who value loyalty to a single team. Dummy variables
are also included for primary ®elding position during a
player’s career (FIRST, SECOND, THIRD, SHORT,
CATCHER). Fielding position measures control for visi-
bility eVects, with out®eld being the excluded category.
STATSj also includes squared terms for all relevant
performance measures (e.g. SEASONS2 , RBIS2 , etc.) to

1
The sample excludes the no-vote players for the years 1962, 1964 and 1966 since these names were unavailable. Pete Rose is excluded
from the sample; Rose was on the ballot ®rst in 1992, when he received 42 votes (9.74%). He never received votes commensurate with his
on-the-®eld accomplishments, possibly due to his gambling and legal problems. Roberto Clemente is also excluded from the sample since
he was elected to the Hall in a 1973 special election, shortly after his death in 1972 while still an active player. The initial year 1962 is
chosen since this is the ®rst year a player of African ancestry, Jackie Robinson, was eligible for election.
2
Players who are retired for at least 23 years are eligible for election by the Committee on Baseball Veterans.
3
Data on the votes of individual voters are not made available by the BBWAA. Therefore, we can make no statements about the race-
based preferences of individual writer±voters.
170
control for any non-linearities in the eVects of these
variables.
The vector RACEj contains measures of a player’s
race. The variable BLACK equals one if a player’s skin
tone is perceived as `black’ regardless of country of birth.
In addition, the variable LATIN is equal to one for a
player who was born in a Latin American country.
However, BLACK and LATIN are not mutually exclusive,
i.e. a player can be both black and born in a Latin
American country. Therefore, an interaction variable,
BLACK*LATIN, captures the eVects of players who are
both black and Latin. If raced-based discrimination is
present, then these measures should be negatively cor-
related with votes.4
Two measures of the eVect of time on Hall of Fame
voting are included. First, the year a player was placed
on the ballot (TREND) was used as a linear time trend.
TREND equals zero if the player’s ®rst vote was in 1962,
equals one if the player’s ®rst vote was in 1963, and so
forth. Second, dummy variables measuring the ®rst year
in which a player was on the ballot (YEAR) are
included. Inclusion of these yearly time categories
allows us to estimate a ®xed eVect for each year.5
Findlay and Reid (1997) and Desser et al. (1999) include
performance-relate d awards (most valuable player awards
won, number of appearances in All-Star games, etc.) as
explanatory variables. A performance-relate d award is nor-
mally given as a result of a voting process, which may be
aVected by discriminatory preferences of voters, and
including performance-relate d awards as explanatory vari-
ables may bias the estimation results. Thus, no measures of
performance-relate d awards are included in this study.
Summary statistics for the entire sample are provided in
Table 1.6
R. T. Jewell et al.
Both Findlay and Reid (1997) and Desser et al. (1999)
estimate models of Hall of Fame voting in the ®rst year a
player was eligible. This study also concentrate s on a
player’s ®rst Hall of Fame ballot. Findlay and Reid use a
two-stage estimation with Heckman’s (1976) sample selec-
tion correction. In the ®rst stage, the authors address a
potential sample selection problem due to the fact that
some players on the ballot do not receive votes. This is
not traditional sample selection, however, in the sense
that the dependent variable on players who do not receive
votes is observed, i.e. these players receive a `no’ vote on
every ballot. Therefore, a player receiving zero votes is a
player whom all writers evaluate as undeserving of Hall of
Fame status. If these players are fundamentall y diVerent
from players who receive votes, and if the diVerence results
from an unobservabl e component, then it is logical to
treat no-vote players as having been selected out of the
sample.
Desser et al. (1999) estimate a logit model with trunca-
tion from below to include the no-vote players. The
authors use Heckman’s method to correct for another
potential sample selection problem, which arises from the
procedure by which players are nominated and placed on
the ballot each year. The authors ®nd limited evidence that
both African Americans and Latin Americans are less
likely than other players to appear on a ballot. This result
suggests racial bias in the nomination process, although the
results are sensitive to speci®cation. Neither the present
paper nor Findlay and Reid’s deals with this issue. Thus,
the results presented here, and those presented by Findlay
and Reid, are eVectively conditional on being nominated
and placed on the ballot.
Like Findlay and Reid, the present authors employ
Heckman’s correction for sample selection to account for

4
We use the baseball cards pictured in Slocum and Foley (1990) to categorize the black players. For Latin American players, we use the
place of birth listed by Total Baseball. A complete list of players by race is given in the Appendix. We make the following comments with
respect to the categorization of players. First, LATIN includes two players born in the Virgin Islands (Jose Morales and Horace Clark);
we also estimate a model categorizing these two players as non-LATIN with no signi®cant diVerences from the results presented in this
paper. Second, Rod Carew is categorized as non-LATIN, since he was born in the Panama Canal Zone, which was an American
Protectorate at the time of his birth. The model is also estimated categorizing Carew as LATIN with no signi®cant diVerences from the
results presented in this paper. Third, we estimate a model in which LATIN equaled one if a player had an Hispanic surname regardless
of birthplace. The results from this estimation are not signi®cantly diVerent from those presented in this paper. Fourth, there is some
di culty in determining whether or not to categorize some players as BLACK. We re-estimate the models reported in Table Three
changing the racial category of some players; a complete list of these players is available from the authors. In three cases (changing Bert
Campaneris from BLACK to non-BLACK, changing Orlando Cepeda from BLACK to non-BLACK, changing Dave Concepcion from
non-BLACK to BLACK), the re-categorization leads to insigni®cant coe cients on all race variables. The remainder of the changes
leads to either no change in or an increase in the signi®cance of the racial coe cients; however, it is clear that the results presented in this
paper are sensitive to racial categorization. All of the estimates commented on in this footnote are available from the authors.
5
Hall of Fame voting took place in each year from 1962 to 1998 except 1963 and 1965. The coe cients on the year dummies (YEAR) are
not reported for the sake of brevity. Also, we do not list the results on the position categories, which are included in all estimations
reported in this paper.
6
Although not reported in this paper, we estimate our models including the performance-related award variables used by Findlay and
Reid (1997) and Desser et al. (1999); the inclusion of these variables lead to insigni®cant coe cients on the race measures in Table Three.
Findlay and Reid (1997) estimate models interacting race and a linear time trend. Their results indicate that discrimination against Latin
American players has decreased over time. Also not reported here, we estimate our models using these race/time interactions but ®nd no
evidence that discrimination has increased or decreased over time. These results are available from the authors.
Measuring discrimination in major league baseball 171
Table 1. Summary statistics …N ˆ 290†

Variable Mean Std. dev. Min. Max.

Percentage vote received 0.0919 0.2314 0 0.9783

BLACK 0.3345 0.4726 0 1
LATIN 0.1103 0.3139 0 1
BLACK*LATIN 0.0862 0.2812 0 1
TREND 20.700 10.009 0 36
HOMERS 174.10 135.63 6 755
HOMERS2 /1000 48.643 74.400 0.036 570.025
RBIS 784.14 399.21 66 2297
RBIS2 /1000 773.70 764.05 4.356 5276.209
BATTING 0.2691 0.0214 0.193 0.344
BATTING2 0.0729 0.0115 0.0372 0.1183
STEALS 95.231 120.88 0 938
STEALS2 /1000 23.630 75.548 0 879.844
WORLD 1.9586 2.0373 0 12
WORLD2 7.9724 17.475 0 144
SEASONS 15.383 3.3205 10 24
SEASONS2 247.62 105.22 100 576
BOTH 0.3793 0.4861 0 1
FIRST 0.1414 0.3490 0 1
SECOND 0.1103 0.3139 0 1
THIRD 0.1241 0.3303 0 1
SHORT 0.0931 0.2911 0 1
CATCH 0.1276 0.3342 0 1

the no-vote players.7 The ®rst stage is a probit estimation
of the determinants of a player receiving at least one vote.
The independent variables in the ®rst stage include all
independent variables used in the second stage plus the
exclusion restrictions discussed below. The second stage
is a minimum logit chi-square estimation; only players
who received votes are included and weighted least squares
is used to correct for heteroscedasticity. The inverse Mills
ratio constructed from the ®rst-stage estimates is included
as a dependent variable in the second stage.
In the Heckman two-stage model, the ®rst-stage equa-
tion is theoretically identi®ed owing to its non-linear nat-
ure. The model is on ®rmer ground, however, if exclusion
restrictions exist. Two variables are included in the ®rst-
stage that are excluded in the second stage. The ®rst is a
dummy variable equal to one if a player was a MLB man-
ager prior to his Hall of Fame ballot. It is expected that this
variable will be positively correlated with receiving at least
one vote since MLB managers have greater visibility. The
second is a dummy variable equal to one if a player spent
over half of his playing career in the National League. This
is included to indicate any exposure eVect of National
versus American League Players. It is found that players
who were managers were more likely to receive at least one
vote in all ®rst-stage speci®cations. Only the second-stage
estimates are reported; the ®rst-stage estimates our
available upon request.
III. RESULTS
The speci®cation in Column A of Table 2 includes a linear
time trend (TREND) and is similar to Findlay and Reid’s
(1997) and Desser et al. (1999) speci®cations. The present
paper concentrates on the results with respect to the race
measures. The positive coe cient on BLACK means that
black players receive more votes than non-black players,
everything else held constant, contradicting race-based dis-
crimination against black players; Findlay and Reid (1997)
®nd limited evidence of a signi®cantly positive eVect of
BLACK, while Desser et al. (1999) ®nd a negative eVect.
Alternatively, LATIN is signi®cant and negative, suggest-
ing that players from Latin America are discriminated
against. The sign on this coe cient is consistent with
Findlay and Reid (1997) and Desser et al. (1999) who
®nd evidence of discrimination against Latin players.
The negative coe cient on TREND indicates that
players on the ballot in later years receive fewer votes;
Findlay and Reid (1997) and Desser et al. (1999) ®nd a
similar result, suggesting that voting standards have

7
The Heckman correction is performed using LIMDEP. The estimation is based on method of moments and is discussed in detail in
Greene (1981). This method produces a likelihood function value, reported in Tables 2 and 3, that can be used to test goodness of ®t.
172 R. T. Jewell et al.
Table 2. Weighted least squares estimates with Heckman correction …N ˆ 200† …standard errors in parentheses†

A B C

Constant 712.357*** (2.5326) 713.048** * (2.4027) 63.303*** (13.056)

BLACK 0.3999* (0.2397) 70.3095 (0.2693) 70.1373 (0.2377)
LATIN 70.5580* (0.3096) 0.3030 (0.3756) 70.1057 (0.3271)
TREND 70.0317*** (0.0113)
YEAR (Not reported) (Not reported)
HOMERS 0.0028 (0.0019) 0.0018 (0.0020) 70.0169** * (0.0047)
HOMERS2 /1000 0.0247** * (0.0061)
RBIS 0.0027*** (0.0009) 0.0040** * (0.0009) 0.0171** * (0.0027)
RBIS2 /1000 70.0052** * (0.0011)
BATTING 19.110** (7.9971) 20.194** * (7.2585) 7432.30*** (88.787)
BATTING2 793.26** * (155.02)
STEALS 0.0032*** (0.0006) 0.0045** * (0.0010) 0.0036* (0.0022)
STEALS2 /1000 0.0011 (0.0029)
WORLD 0.0573 (0.0360) 0.1009** (0.0419) 0.1891** (0.0889)
WORLD2 70.0063 (0.0084)
SEASONS 0.0918* (0.0522) 0.0939* (0.0581) 72.1735** * (0.3061)
SEASONS2 0.0640** * (0.0089)
BOTH 70.7618*** (0.2385) 70.3669 (0.2324) 70.6725** * (0.2103)
LAMBDA 70.6354 (0.7012) 70.0043 (0.6296) 70.9653 (0.6072)
Wald test À2 …2† ˆ 4:82* À2 …2† ˆ 1:54 À2 …2† ˆ 0:61
Log-likelihood 7314.809 7253.696 7213.319

*** Signi®cant at the 1% level.

** Signi®cant at the 5% level.
* Signi®cant at the 10% level.

increased over time.8 The coe cient signs on the perform-
ance statistics are all positive as expected. The inverse Mills
ratio (LAMBDA) is included in the second stage to
account for potential sample selection. The coe cient is
negative and insigni®cant. This result indicates that there
is no sample selection associated with getting at least one
vote. Findlay and Reid (1997) generally report positive
coe cients on LAMBDA. Desser et al. (1999) report nega-
tive coe cients on LAMBDA, but in their paper, it meas-
ures sample selection due to the nomination procedure.
Column B of Table 2 replaces the linear time trend with
yearly time dummies (YEAR). Including a dummy for each
time period accounts for ®xed time eVects. Since the com-
position and number of voters change each year, the time
dummies may also control for unobserved diVerences in
voters across years. The most important changes from
Column A are that the coe cients on BLACK and
LATIN become insigni®cant. Column C of Table 2 adds
squared terms for each performance measure. Including the
squared terms does not change the signi®cance of the race
variables. Table 2 reports the chi-square statistic from a
Wald test restricting the race measures in the relevant spe-
ci®cation to be zero; we also report Wald tests for the
speci®cations in Table 3. The results indicate that the
race variables are jointly insigni®cant in Columns B and
C. Thus, utilizing YEAR instead of TREND eliminates
any signi®cant eVect of race on Hall of Fame voting.
We might expect the coe cients on the squared terms to
be negative, indicating decreasing returns to performanc e
statistics. There are some implausible signs on HOMERS,
BATTING and SEASONS, which may indicate that the
model in Column C is miss-speci®ed. However, this
seems to be less of a problem when looking at the marginal
eVects. An increase in home runs increases the number of
votes for any player with more than 342 homers, which is
the case for 35 players in our sample. Any player with a
batting average of more than 0.272 (135 players in our
sample) has an increase in votes when his average increases.
Any player with more than 17 years of experience sees an
increase in votes with another year of service; 110 players in
our sample played more than 17 years in MLB. In addi-
tion, it is noted that including the squared terms improves
the goodness of ®t; the chi-square statistic is 80.75 with a
critical value of 16.81 at ¬ ˆ 0:01.
If discrimination exists in voting, it is expected that the
eVect will be compounded for players who are both black
and born in Latin American. To test for such an eVect,
Column A of Table 3 reports results from an estimation
that includes YEAR and the dummy variable
BLACK*LATIN. The coe cients on BLACK*LATIN

8
The coe cient on TIME is reported as positive in Table 2.2 in Findlay and Reid’s paper; however, Dave Findlay assures us that this is a
typographical error and that the sign is negative.
Measuring discrimination in major league baseball 173
Table 3. Weighted least squares estimates with Heckman correction …N ˆ 200† …standard errors in parenthesis†

A B C

Constant 713.961** * (2.4749) 62.267** * (12.98) 63.486** * (13.147)

BLACK 70.1686 (0.2845) 0.0665 (0.2524)
LATIN 1.2339* (0.7280) 1.3049** (0.6373)
BLACK*LATIN 71.2414 (0.8329) 71.8729** (0.7464)
YEAR (Not reported) (Not reported) (Not reported)
HOMERS 0.0027 (0.0021) 70.0149** * (0.0047) 70.0176** * (0.0047)
HOMERS2 /1000 0.0238** * (0.0061) 0.0254** * (0.0062)
RBIS 0.0035** * (0.0010) 0.0159** * (0.0027) 0.0173** * (0.0027)
RBIS2 /1000 70.0050** * (0.0011) 70.0052** * (0.0011)
BATTING 23.160** * (7.5192) 7431.96** * (88.318) 7433.39** * (89.352)
BATTING2 802.19** * (154.29) 793.97*** (156.20)
STEALS 0.0040** * (0.0010) 0.0022 (0.0022) 0.0031 (0.0021)
STEALS2 /1000 0.0020 (0.0029) 0.0013 (0.0029)
WORLD 0.1093** * (0.0420) 0.2307** * (0.0900) 0.1852** (0.0897)
WORLD2 70.0093 (0.0085) 70.0058 (0.0085)
SEASONS 0.1142* (0.0594) 72.1160** * (0.3057) 72.1700** * (0.3081)
SEASONS2 0.0634** * (0.0088) 0.0638** * (0.0089)
BOTH 70.3055 (0.2349) 70.6006** * (0.2118) 70.6705** * (0.2050)
LAMBDA 70.1099 (0.6190) 70.9833 (0.6040) 71.0294* (0.6161)
Wald test À2 …3† ˆ 3:77 À2 …3† ˆ 6:94*
Log-likelihood 7251.922 7209.827 7215.136

*** Signi®cant at the 1% level.

** Signi®cant at the 5% level.
* Signi®cant at the 10% level.

and BLACK are negative and insigni®cant, while the coef-
®cient on LATIN is positive and signi®cant. However, the
coe cients on the race measures are shown to be jointly
insigni®cant. The results reported in Column B are from a
model that adds in the squared terms (similar to Table 2,
Column C). It now appears that players who are both
black and Latin face voting discrimination, black players
face no discrimination, and Latin players actually get more
votes than non-Latin players. Additionally, the race vari-
ables are jointly signi®cant at the 10% level. Again, it is
important to note that the measured eVect or race is extre-
mely sensitive to changing the independent variables in
Tables 2 and 3.9
Most of the performance variables show the expected
results, a positive ®rst derivative and a negative second
derivative. However, HOMERS, BATTING and
SEASONS have negative ®rst derivatives and positive sec-
ond derivatives. As in Table 2, this may indicate a mis-
speci®cation. In any case, the results indicate that an
increase in home runs increases the number of votes for
any player with more than 313 homers; 45 players in our
sample have enough home runs to be positively aVected.
Any player with a batting average of more than 0.269 (148
players in our sample) has an increase in votes when his
average increases. Any player with more than 17 years of
experience sees a increase in votes with another year of
service; this result is the same as reported in Table 2,
Column C. Finally, we report estimates in Column C of
a model that drops all race variables from Column B. A
likelihood ratio test comparing the models in Columns B
and C shows that the race variables improve the goodness
of ®t; the chi-square statistic is 10.62 with a critical value of
9.35 at ¬ ˆ 0:025.
IV. MODEL PREDICTIONS
Tables 2 and 3 contain interesting results, but they provide
no clue about the quantitative eVects of any race-based
voting on Hall of Fame voting outcomes. These results
give limited evidence that players who are black and were
born in a Latin American country tend to receive fewer
votes than other players. To examine the quantitative
eVects of discrimination, it is necessary to ascertain
whether the actual vote tally for these players was aVected.
Using the results from Column C of Table 3, the predicted
vote tally is computed, as a percentage of the total vote, for
each BLACK*LATIN player under the assumption that

9
To further test for discrimination in Hall of Fame voting, we estimate a model that interacts player statistics and player race. This
estimation reveals extremely limited evidence that racial minorities receive fewer additional votes for increased performance than other
players. Speci®cally, BLACK players are punished more than other players for playing in both leagues and BLACK*LATIN players
receive less credit than other players for more stolen bases. These results are available from the authors.
174 R. T. Jewell et al.
Table 4. Predicted outcomes for BLACK*LATIN players

Name Actual vote (%) Predicted vote (%) DiVerence (%)

F. Alou 0.78 0.88 0.10

J. Alou 0.25 0.03 70.22
M. Alou 1.30 0.50 70.79
B. Campaneris 3.13 45.72 42.58
J. Cardinal 0.24 2.13 1.90
L. Cardenas 0.25 0.37 0.12
R. Carty 0.25 1.44 1.18
P. Casanova 0.00 1.33 1.33
C. Cedeno 0.47 11.67 11.20
O. Cepeda 12.47 23.55 11.08
H. Clarke 0.00 2.66 2.66
C. Geronimo 0.00 0.14 0.14
P. Guerrero 1.27 3.09 1.82
M. Minoso 1.77 13.31 11.55
W. Montanez 0.00 4.76 4.76
J. Morales 0.00 0.00 0.00
M. Mota 4.22 3.45 70.77
I. Murrell 0.00 0.07 0.07
B. Oglivie 0.00 3.15 3.15
T. Oliva 15.18 11.00 74.18
T. Perez 50.00 45.21 74.79
V. Power 0.56 0.72 0.16
M. Sanguillen 0.47 9.20 8.73
C. Tovar 0.00 2.39 2.39
M. Trillo 0.00 1.85 1.85
Mean 3.70 7.54 3.84

race plays no part in BBWAA voting. The results are then
compared to the actual vote percentage received. Table 4
presents the predicted outcomes in an alphabetical listing
of BLACK*LATIN players.
The BLACK*LATIN players are largely clustered at the
lower end of the vote distribution (22 of these 25 players
received less than 5% of the votes). These players seem to
have neither the actual votes nor the statistics to enter the
Hall of Fame. The mean actual vote for this group is
3.70%, while the mean predicted vote is 7.54%. The player
in this group most aVected by potential racial bias is Bert
Campaneris, whose actual vote was 3.13% but is predicted
to receive 45.72% of the vote based solely on his statistics.
A few of these BLACK*LATIN players actually received
more votes than their statistics would predict. Tony Perez
provides the most interesting example. Before being elected
to the Hall of Fame in 2000 on his ninth ballot, Perez
received attention for his claim that the BBWAA discrimi-
nated in voting for the Hall of Fame (Desser et al. 1999,
p. 592). According to our vote predictions, however, Perez
(who received 50% of the vote on his ®rst ballot) would
have received fewer votes (45.21%) if voting were race-
neutral. Although most of the players in this group should
have received more votes based on their statistics alone,
racial bias in voting clearly did not keep these players
from entering the Hall of Fame on their ®rst ballot.
However, racially-biased voting may not have allowed
some players to the receive the 5% vote necessary to be
on the next ballot.10
Next, the predicted outcomes for all players (regardless
of race) is investigated, looking for racial bias by compar-
ing actual vote percentages to predicted vote percentages as
discussed above. The present paper does not report the
predictions for all players here; the full set of predicted
vote outcomes is available from the authors. No players
were left out of the Hall of Fame due to racial bias; speci®-
cally, there are no players for whom the predicted vote is
greater than 75% and the actual vote is less than 75%.
Racially-biased voting seems to have no negative eVect
on the ®rst vote by the BBWAA in terms of which players
were inducted into the Hall of Fame.
However, the authors did ®nd several players who were
elected to the Hall of Fame on the ®rst ballot and who had
a predicted vote percentage of less than 75%: Johnny
Bench (actual vote ˆ 96%, predicted vote ˆ 67%); Al
Kaline (actual vote ˆ 88%, predicted vote ˆ 71%); Willie

10
The 5% minimum vote rule was put into place in 1981. Prior to 1981, players could remain on the ballot for up to 15 years no matter
how many votes they received in prior elections.
Measuring discrimination in major league baseball
McCovey (actual vote ˆ 81%, predicted vote ˆ 70%);
Jackie Robinson (actual vote ˆ 78%, predicted
vote ˆ 64%), Mike Schmidt (actual vote ˆ 97%, predicted
vote ˆ 58%), and Willie Stargell (actual vote ˆ 82%; pre-
dicted vote ˆ 74%). It is noted that half of these players are
categorized as black (McCovey, Robinson and Stargell).
These players may have received votes based on accom-
plishments other than those in our data set. For instance,
Jackie Robinson was, obviously, an important ®gure due
to his breaking of the colour line in baseball. The other
players may have entered the Hall of Fame simply based
on `integrity, sportsmanship, character, and contributions
to the team(s) on which the player played.’ None the less, it
is clear that the race of these players did not preclude them
from entry into the Hall of Fame on their ®rst ballot.11
V. C ONCLUSION
This paper extends recent research on the eVects of race
on player induction into the National Baseball Hall of
Fame. Speci®cally, the authors tested for discrimination
in voting against African American players, Latin
American players, and players who are both black and
born in Latin America on their ®rst Hall of Fame ballot.
A major ®nding is that model speci®cation has a large
in¯uence on the measured eVect of race on voting out-
comes. Limited evidence was found that players born in
Latin American countries receive fewer votes on their
®rst ballot; however, the results are extremely sensitive to
the estimation speci®cation. For instance, including ®xed
time eVects signi®cantly reduces the negative eVect of
Latin American birth. In addition, it is found that any
discrimination against Latin players is limited to those
who are also black. No evidence was found of bias
among voters against black players, with the exception of
black players born in a Latin American country. Finally
the paper shows that race did not aVect whether the players
in our sample received enough votes to get into the Hall of
Fame on the ®rst ballot. Instead, any discrimination in
voting is concentrated among those players who receive
few votes and may not have the statistics to warrant
entry into the Hall of Fame.
A C K N O W LE D G E M E N TS
The authors wish to thank David Finlay for help in ®nding
the names of the no-vote players. We thank David Finlay,
11
The speci®cations in Tables 2 and 3 are re-estimated with a dependent variable equal to one if the player was ever elected into the Hall
of Fame. The model will not run with the full set of independent variables, and the results are highly sensitive to the independent
variables included. The only consistent result is that Black players are more likely to be inducted than non-black players. This result is
not surprising given the predicted voting results discussed above. These results are available from the authors. Findlay and Reid (1997)
attempt a similar estimation, ®nding no signi®cant eVect of race on Hall of Fame induction.
175
Ed Lopez, JeV Rous and an anonymous referee for helpful
comments and suggestions. In addition, we thank Christal
McNutt for providing excellent research assistance.
REFERENCES
Anderson, T. and La Croix, S. J. (1991) Customer racial discrimi-
nation in major league baseball, Economic Inquiry, 24, 665±
77.
Becker, G. (1971) The Economics of Discrimination, second edi-
tion. University of Chicago: Chicago, IL.
Brown, E., Spiro, R. and Keenan, D. (1991) Wage and non-
wage discrimination in professional basketball: do fans
aVect it? American Journal of Economics and Sociology, 50,
333±45.
Brown, R. W. and Jewell, R. T. (1994) Is there customer discrimi-
nation in college basketball? The Premium fans pay for white
players. Social Science Quarterly, 75, 401±13.
Brown, R. W. and Jewell, R. T. (1995) Race, revenues, and college
basketball, The Review of Black Political Economy, 23, 75±
90.
Burdekin, R. C. K. and Idson, T. L. (1991) Customer preferences,
attendance, and the structure of professional basketball
teams, Applied Economics, 23, 179±86.
Desser, A., Monks, J. and Robinson, M. (1999) Baseball hall of
fame voting: a test of the customer discrimination hypothesis,
Social Science Quarterly, 80, 591±603.
Findlay, D. W. and Reid, C. E. (1997) Voting behaviour,
discrimination, and the national baseball hall of fame,
Economic Inquiry, 35, 562±78.
Fox, J. (1997) Applied Regression Analysis, Linear Models, and
Related Methods. Sage: Thousand Oaks, CA.
Gabriel, P. E., Johnson, C. and Stanton, T. J. (1995) An
examination of customer racial discrimination in the
market for baseball memorabilia, Journal of Business, 68,
215±30.
Gabriel, P. E., Johnson, C. and Stanton, T. J. (1999) Customer
racial discrimination for baseball memorabilia, Applied
Economics, 31, 1331±35.
Greene, W. H. (1981) Sample selection bias as a speci®cation
error: comment. Econometrica, 49, 795±98.
Gwartney, J. and Haworth, C. (1974) Employer costs and dis-
crimination: the case of baseball, Journal of Political
Economy, 82, 873±81.
Hamilton, B. H. (1997) Racial discrimination and professional
basketball salaries in the 1990’s, Applied Economics, 29,
287±96.
Hanssen, F. A. and Andersen, T. (1999) Has discrimination
lessened over time? A test using baseball’s all-star vote,
Economic Inquiry, 37, 326±52.
Heckman, J. J. (1976) The common structure of statistical
models of truncation, sample Selection, and limited
dependent variables and a simple estimator for such
models, Annals of Economic and Social Measurement, 5,
475±92.
176
Kahn, L. M. (1991) Discrimination in professional sports: a sur-
vey of the literature, Industrial and Labour Relations Review,
44, 396±418.
Kahn, L. M. (1992) The eVects of race on professional football
players’ compensation, Industrial and Labour Relations
Review, 45, 295±310.
Kahn, L. M. and Sherer, P. D. (1988) Racial diVerences in profes-
sional basketball players’ compensation, Journal of Labour
Economics, 6, 40±61.
Koch, J. V. and Warren Vander Hill, C. (1988) Is there discrimi-
nation in the `black man’s game? Social Science Quarterly,
69, 83±94.
Maddala, G. S. (1983) Limited-Dependent and Qualitative
Variables in Econometrics. Cambridge University,
Cambridge, UK.
R. T. Jewell et al.
Nardinelli, C. and Simon, C. (1990) Customer racial discrimi-
nation in the market for memorabilia: the case of baseball,
Quarterly Journal of Economics, 105, 575±96.
Scully, G. W. (1973) Economic discrimination in pro-
fessional sports, Law and Contemporary Problems, 38,
67±84.
Singell, L. D. (1991) Baseball-speci®c human capital: why good
but not great players are more likely to coach in the major
leagues, Southern Economic Journal, 58, 77±86.
Slocum, F. and Foley, R. (1990) Topps Baseball Cards, The
Complete Picture Collection: A 40 Year History, 1951±1990.
Warner Books, New York.
Sommers, P. M. and Quinton, N. (1982) Pay and performance in
major league baseball: the case of the ®rst family of free
agents, Journal of Human Resources, 17, 426±36.

APPENDIX

Race categories for all players

Name BLACK LATIN Name BLACK LATIN Name BLACK LATIN

Hank Aaron 1 0 Bert Campaneris 1 1 Joe Ferguson 0 0

Bobby Adams 0 0 Chris Cannizzaro 0 0 Curt Flood 1 0
Dick Allen 1 0 Jose Cardenal 1 1 George Foster 1 0
Felipe Alou 1 1 Leo Cardenas 1 1 Nellie Fox 0 0
Jesus Alou 1 1 Rod Carew 1 0 Bill Freehan 0 0
Matty Alou 1 1 Chico Carrasquel 0 1 Jim Fregosi 0 0
Luis Aparicio 0 1 Gary Carter 0 0 Carl Furillo 0 0
Richie Ashburn 0 0 Rico Carty 1 1 Phil Gagliano 0 0
Bob Bailey 0 0 Paul Casanova 1 1 Augie Galan 0 0
Bob Bailor 0 0 Dave Cash 1 0 Oscar Gamble 1 0
Dusty Baker 1 0 Norm Cash 0 0 Phil Garner 0 0
Sal Bando 0 0 Phil Cavaretta 0 0 Steve Garvey 0 0
Ernie Banks 1 0 Cesar Cedeno 1 1 Cesar Geronimo 1 1
Bob Barton 0 0 Orlando Cepeda 1 1 Jim Gosger 0 0
Don Baylor 1 0 Ron Cey 0 0 Dick Green 0 0
Hank Bauer 0 0 Chris Chambliss 1 0 Bobby Grich 0 0
Glenn Beckert 0 0 Jack Clark 0 0 Ken GriVey 1 0
Mark Belanger 0 0 Horace Clarke 1 1 Dick Groat 0 0
Buddy Bell 0 0 Rocky Colavito 0 0 Greg Gross 0 0
Johnny Bench 0 0 Dave Concepcion 0 1 Jerry Grote 0 0
Yogi Berra 0 0 Cecil Cooper 1 0 Pedro Guerrero 1 1
Ken Berry 0 0 Walker Cooper 0 0 Mike Hargrove 0 0
Paul Blair 1 0 Clint Courtney 0 0 Tommy Harper 1 0
John Boccabella 0 0 Del Crandall 0 0 Toby Harrah 0 0
Bobby Bonds 1 0 Terry Crowley 0 0 Bud Harrelson 0 0
Bob Boone 0 0 Jose Cruz 0 1 Jim Ray Hart 1 0
Larry Bowa 0 0 Alvin Dark 0 0 Grady Hatton 0 0
Clete Boyer 0 0 Rich Dauer 0 0 Richie Hebner 0 0
Ken Boyer 0 0 Vic Davalillo 0 1 Jim Hegan 0 0
Rocky Bridges 0 0 Tommy Davis 1 0 Tommy Helms 0 0
John Briggs 1 0 Doug DeCinces 0 0 Solly Hemus 0 0
Lou Brock 1 0 Joe DeMaestri 0 0 George Hendrick 1 0
Gates Brown 1 0 Rick Dempsey 0 0 Keith Hernandez 0 0
Larry Brown 0 0 Bucky Dent 0 0 Jim Hickman 0 0
Bill Bruton 1 0 Larry Doby 1 0 Larry Hisle 1 0
Bill Buckner 0 0 Brian Downing 0 0 Gil Hodges 0 0
Al Bumbry 1 0 Walt Dropo 0 0 Bob Horner 0 0
Smoky Burgess 0 0 Johnny Edwards 0 0 Willie Horton 1 0
Rick Burleson 0 0 Del Ennis 0 0 Frank House 0 0
JeV Burroughs 0 0 Darrell Evans 0 0 Elston Howard 1 0
Johnny Callison 0 0 Dwight Evans 0 0 Frank Howard 0 0
Roy Campanella 1 0 Ron Fairly 0 0 Art Howe 0 0
Measuring discrimination in major league baseball 177

Name BLACK LATIN Name BLACK LATIN Name BLACK LATIN

Roy Howell 0 0 Bill Melton 0 0 Manny Sanguillen 1 1

Randy Hundley 0 0 Denis Menke 0 0 Ron Santo 0 0
Ron Hunt 0 0 Eddie Miksis 0 0 Hank Sauer 0 0
Reggie Jackson 1 0 Felix Millan 0 1 Paul Schaal 0 0
Sonny Jackson 1 0 Norm Miller 0 0 Mike Schmidt 0 0
Jackie Jensen 0 0 John Milner 1 0 Red Schoendienst 0 0
Alex Johnson 1 0 Minnie Minoso 1 1 Tony Scott 1 0
Dave Johnson 0 0 Rick Monday 0 0 Roy Sievers 0 0
Deron Johnson 0 0 Don Money 0 0 Ted Simmons 0 0
Cleon Jones 1 0 Willie Montanez 1 1 Duke Sims 0 0
Willie Jones 0 0 Wally Moon 0 0 Ken Singleton 1 0
Mike Jorgensen 0 0 Jose Morales 1 1 Enos Slaughter 0 0
Al Kaline 0 0 Joe Morgan 1 0 Roy Smalley 0 0
George Kell 0 0 Manny Mota 1 1 Reggie Smith 1 0
John Kennedy 0 0 Thurman Munson 0 0 Duke Snider 0 0
Terry Kennedy 0 0 Bobby Murcer 0 0 Chris Speier 0 0
Don Kessinger 0 0 Ivan Murrell 1 1 Jim Spencer 0 0
Harmon Killebrew 0 0 Stan Musial 0 0 Mickey Stanley 0 0
Ralph Kiner 0 0 Graig Nettles 0 0 Willie Stargell 1 0
Dave Kingman 0 0 Ron Northey 0 0 Rusty Staub 0 0
Ted Kluszewski 0 0 Jim Northrup 0 0 John Stearns 0 0
Ray Knight 0 0 Ben Oglivie 1 1 Champ Summers 0 0
Andy Kosco 0 0 Tony Oliva 1 1 Jim Sundberg 0 0
Ed Kranepool 0 0 Al Oliver 1 0 Garry Templeton 1 0
Harvey Kuenn 0 0 Amos Otis 1 0 Gene Tenace 0 0
Hal Lanier 0 0 Andy Pafko 0 0 Gorman Thomas 0 0
Carney Lansford 0 0 Dave Parker 1 0 Bobby Thomson 0 0
Chet Lemon 1 0 Larry Parrish 0 0 Andre Thornton 1 0
JeVrey Leonard 1 0 Tony Perez 1 1 Bobby Tolan 1 0
Whitey Lockman 0 0 Rico Petrocelli 0 0 Earl Torgeson 0 0
Davey Lopes 0 0 Lou Piniella 0 0 Joe Torre 0 0
John Lowenstein 0 0 Vada Pinson 1 0 Cesar Tovar 1 1
Greg Luzinski 0 0 Darrell Porter 0 0 Manny Trillo 1 1
Fred Lynn 0 0 Boog Powell 0 0 Del Unser 0 0
Garry Maddox 1 0 Vic Power 1 1 Ellis Valentine 1 0
Bill Madlock 1 0 Terry Puhl 0 0 Elmer Valo 0 0
Mickey Mantle 0 0 Doug Rader 0 0 Mickey Vernon 0 0
Roger Maris 0 0 Willie Randolph 1 0 Bill Virdon 0 0
Billy Martin 0 0 Johnny Ray 1 0 Dixie Walker 0 0
Eddie Mathews 0 0 Rick Reichardt 0 0 Claudell Washington 1 0
Gary Matthews 1 0 Pee Wee Reese 0 0 John Wathan 0 0
Dal Maxvill 0 0 Ken Reitz 0 0 Bob Watson 1 0
Carlos May 1 0 Jerry Remy 0 0 Vic Wertz 0 0
Lee May 1 0 Del Rice 0 0 Wes Westrum 0 0
John Mayberry 1 0 Jim Rice 1 0 Frank White 1 0
Willie Mays 1 0 Bobby Richardson 0 0 Roy White 1 0
Bill Mazeroski 0 0 Jim Rivera 0 0 Billy Williams 1 0
Dick McAuliVe 0 0 Bill Robinson 1 0 Ted Williams 0 0
Bake McBride 1 0 Brooks Robinson 0 0 Maury Wills 1 0
Tim McCarver 0 0 Frank Robinson 1 0 Jim Wynn 1 0
Willie McCovey 1 0 Jackie Robinson 1 0 Carl Yastrzemski 0 0
Gil McDougald 0 0 Cookie Rojas 0 1 Steve Yeager 0 0
Roy McMillan 0 0 Joe Rudi 0 0 Rudy York 0 0
Ken McMullen 0 0 Bill Russell 0 0 Richie Zisk 0 0
Hal McRae 1 0 Mike Ryan 0 0

(continued )
Copyright of Applied Economics is the property of Routledge and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission. However, users
may print, download, or email articles for individual use.

View publication stats