G Model

SMR-254; No. of Pages 9

Sport Management Review xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Sport Management Review

journal homepage: [Link]/locate/smr

When the crowd evaluates soccer players’ market values:

Accuracy and evaluation attributes of an online community
Steffen Herm a,*, Hans-Markus Callsen-Bracker b,1, Henning Kreis c,2
a
Hochschule für Technik und Wirtschaft Berlin, Fachbereich Wirtschaftswissenschaften I (VG 440), Treskowallee 8, 10318 Berlin, Germany
b
Technische Universität Berlin, Lehrstuhl für Finanzierung und Investition, Str. des 17. Juni 135, 10623 Berlin, Germany
c
Freie Universität Berlin, Fachbereich Wirtschaftswissenschaft, Marketing Department, Otto-von-Simson-Str. 19, 14195 Berlin, Germany

A R T I C L E I N F O A B S T R A C T

Article history: Evaluating the value of athletes to determine whether they should be transferred from one
Received 16 June 2013 club to another has become a major challenge for managers of sports teams. In the context
Received in revised form 21 December 2013 of German soccer, aficionados have built a large online community that evaluates
Accepted 26 December 2013
professional soccer players’ market values. The community has become the main source
for reporting market values in the media and has a strong impact on sports economy: it is
Keywords:
used in real market transactions and wage negotiations, indicating the power of crowd
Crowd valuation
wisdom in the sports management context.
Market value
Human capital This research describes the evaluation process of the community, investigates the
Professional athletes accuracy of its estimated market values, and shows which attributes are most important for
Soccer market-value evaluations. After demonstrating that the community’s market-value
estimates can predict actual transfer fees, we show that community evaluations can
largely be explained by an econometric model that contains two blocks of determinants:
variables that are directly related to players’ talent and variables that result from judgments
by external sources (e.g., journalists). Reorganizing variables that were used in previous
studies into two blocks (talent vs. external determinants) provides a more differentiated look
at the popularity of players than recent literature on the ‘‘superstar phenomenon.’’
ß 2014 Sport Management Association of Australia and New Zealand. Published by
Elsevier Ltd. All rights reserved.

1. Introduction

We define the market value of a professional athlete – such as a soccer player – as an estimate of the amount of money a
club would be willing to pay in order to make this athlete sign a contract, independent of an actual transaction. However,
evaluating an individual’s value within any kind of team – such as a soccer team – is a challenging task. This research
investigates how and how well an online community with special interest in professional soccer performs on such a complex
task of human capital valuation.
Drawing on research on information aggregation mechanisms for teams (e.g., Brunswik’s lens model) we explain and
discuss the special approach that a community uses to evaluate professional athletes. Furthermore, we discuss variables that
are important for evaluating soccer players, drawing on prior research on sports management in the context of soccer and

* Corresponding author. Tel.: +49 30 5019 2420.

E-mail addresses: herm@[Link] (S. Herm), markuscallsen@[Link] (H.-M. Callsen-Bracker), [Link]@[Link] (H. Kreis).
1
Tel.: +49 30 314 28904.
2
Tel.: +49 30 838 54460.

1441-3523/$ – see front matter ß 2014 Sport Management Association of Australia and New Zealand. Published by Elsevier Ltd. All rights reserved.
[Link]

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

2 S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx

focusing on the ‘‘superstar phenomenon.’’ This phenomenon is not only important for professional sports but also in many
other areas of application such as top management salaries or compensations for artists and managers (Franck & Nüesch,
2012). Following the superstar literature, a team member’s value is driven by two distinct factors: ‘‘talent’’ indicated by
objective performance (Rosen, 1981) and ‘‘network externalities of popularity’’ (Adler, 1985). Although talent – in the
context of soccer – is often reflected by simple statistics such as the number of assists, precise passes, and won ‘‘duels’’
(taking or defending the ball from an opponent), popularity has been operationalized by the number of citations in
newspapers and weekly magazines (Franck & Nüesch, 2012). We suggest that the number of press citations is only one of a
number of factors that influence popularity and have an impact on athletes’ market values. Thus, we enhance this view and
discuss various external variables that can reflect players’ popularity, e.g., decisions of team coaches or club managers and
evaluations by sports journalists. For example, the number of games an athlete plays for a team does not necessarily depend
on his or her talent, but might be up to the coach’s discretion. The evaluation of a player’s performance by a coach is highly
subjective, because the player might not fit into the team coach’s strategy or because of personal tensions between player
and coach. Also a community (crowd) that evaluates athletes based on a variable ‘‘number of games played’’ can make
subjective interpretations and does not necessarily interpret a low number of games played to be an indicator of low market
value. This example stresses the importance of enhancing our knowledge on the mechanisms of information aggregation and
evaluation process in the context of crowdsourcing. More specifically, this research investigates (1) how – with respect to
their organizational principle – and (2) how well – with respect to the accuracy of the evaluation output – an online
community estimates professional soccer players’ market values, and (3) which attributes are most important for the
evaluations made by the crowd. We also contribute to the literature on the superstar phenomenon by reconsidering various
well-known attributes as being external variables and not talent variables. Hence, the superstar phenomenon can be
measured by more variables, not just by the number of press citations.
This manuscript is organized in the following manner: in the next section we describe the relevance of crowdsourcing as
well as an organizational principle that we call the judge principle, which theorizes how crowds can conduct evaluation
tasks. Afterward, we evaluate how accurately an online community performs the complex task of human capital evaluation.
In particular, we test if market-value estimates made by the online community predict actual transfer fees. Finally, we
present an econometric model that helps to explain which talent attributes and which external attributes are most indicative
for market-value estimations. After presenting and discussing our findings we offer ideas for further research.

2. The wisdom of crowds

Sir Francis Galton (1907) discussed the idea of ‘‘collective wisdom’’ when he asked ‘‘the crowd’’ to gauge the weight of an ox.
He successfully employed a democratic principle (an equal say) to arrive at a precise estimation. Since Galton (1907), the
concept of ‘‘collective wisdom’’ has nourished a rich and interdisciplinary stream of research (Saavedra, Duch, & Uzzi, 2011) and
popular science literature (Surowiecki, 2005). Although the concept of ‘‘crowdsourcing’’ has been around for a long time, the
advent of the Internet has opened up many new possibilities. Self-selected members of ‘‘the crowd’’ work together on problems
such as improving algorithms, translating websites, and evaluating soccer players – quickly and worldwide (Afuah & Tucci,
2012). On the one hand, the concept of crowdsourcing has been described as an open call to identify innovative input from non-
obvious sources (Jeppesen & Lakhani, 2010). On the other hand, Adams and Ferreira (2009) compare guesses from individual
bettors with guesses from groups of bettors and find that group decisions are more accurate and more moderate, ‘‘either
because groups have to reach a compromise when their members disagree or because individuals with extreme opinions are
less likely to be part of a group’’ (Adams & Ferreira, 2009, p. 882). In a similar vein, other studies that have compared group and
individual decision making provide evidence that groups produce more rational output than individuals (Charness & Sutter,
2012). In addition, previous research on prediction markets has also indicated that crowds perform very well in information
aggregation tasks (e.g., Wolfers & Zitzewitz, 2004).1 Estimating the market value of a professional athlete is such an information
aggregation task, because all available information about the athlete could be integrated into a single variable. Hence, a crowd
should perform well in predicting market values. This research investigates how and how well an online community with
special interest in professional soccer performs the complex task of human capital valuation.

3. The judge principle – a pragmatic conception for information aggregation

Conceptually, communities can arrive at a judgment using different approaches (Schenk & Guittard, 2011). In an
integrative approach, a crowd pools input (compare to prediction markets). In contrast, Schenk and Guittard (2011) call
crowdsourcing selective when a client chooses an input from a set of options that the crowd has provided.
The market value of a professional athlete can be defined by applying integrative crowdsourcing in a democratic but
inflexible way. That is, a community can calculate the mean (or median) of all suggested market values. The market value

1
Large companies such as Google, General Electric, Siemens, and Chrysler – to name just a few – have used prediction markets. These companies aimed to
harvest information provided by ‘‘the crowd’’ (their employees), e.g., estimates of the number of Gmail users in the future (Google). However, Cowgill,
Wolfers, and Zitzewitz (2009) have found that predictions of market results are prone to numerous biases. For example, new employees are overly
optimistic or employees working closely together exhibit highly correlated views and estimates.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx 3

can also be defined in a selective approach to information aggregation, e.g., by weighting and selecting from the set of options
that the crowd has provided. This would imply a more hierarchic, but also more flexible approach to determine market
values that we call the judge principle. That is, an empowered community member – we call him or her ‘‘judge’’ – could
evaluate an athlete’s market value differently than the average suggestion in the community and choose to aggregate
information provided by the community on a case-by-case basis. The judge has the opportunity to reduce the influence of
outliers in the final valuation or can attribute higher weights to the opinions expressed by more qualified members of the
community. For example, a judge could infer that a member with a very high number of posts in an online forum could have a
better ability to estimate market values correctly than a member with a very low number of posts. Aside from judgments in
respect to the ability to estimate a market value, a judge can correct for some community members who appear to have some
personal interest in a player’s market value and also in competing players’ market values, e.g., sports agents.
A conceptual model that has captured such a subjective policy is Brunswik’s lens model (Brunswik, 1952; Cooksey, 1996;
Mosier & Fischer, 2010). Since the 1960s the model’s influence has shown a steady increase in a variety of research areas,
most notably in research on human judgment. In line with the selective approach of information aggregation from crowds
that Schenk and Guittard (2011) have described, Brunswik conceptualized that observers do not rely on all possible cues
when making judgments about other individuals or objects, but they rely instead on selected, probabilistic cues or attributes.
Based on Brunswik’s lens model, researchers such as Brehmer and Hagafors (1986) and Hollenbeck et al. (1995) have
presented multilevel models of decision making in teams in which members possess distinctive roles, have different
expertise, or have different access to decision-relevant information. These adaptations of the lens model (for further
adaptations, see Cooksey, 1996) help to explain how team members make subjective recommendations (YeA and YeB in
Fig. 1) in respect to a player’s market value (Yt). They might build these recommendations on evaluations of different
attributes (Xi) and pass them onto a team leader (judge), who then renders the team’s decision (Yf).
Research on different information aggregation approaches for crowds has shown that a selective approach outperforms
an integrative approach when the crowd is heterogeneous in respect to knowledge and experience (Bachrach, Graepel,
Kasneci, Kosinski, & Van Gael, 2012). Consequently, in case of the evaluation of soccer players within an online community
that is not selecting members based on expertise but gives access to any interested individual, a selective approach to
information aggregation (e.g., the judge principle) should outperform an integrative approach (e.g., a prediction market).
[Link], Germany’s leading online soccer community and one of the largest sport websites on the German-
language Internet, applies the judge principle. Every interested person can register for free on the [Link]
community and discuss market values of thousands of soccer players playing in various countries and divisions. Commonly,
a member proposes a market value for a player and provides a rationale for this number, for example, by arguing about one or
more of the player’s attributes to justify his or her market-value estimation. Every community member can follow the
discussion threads of all the players’ market values and users can contribute based on criteria that are most important to
them personally. However, only a few merited community members are discussion leaders – ‘‘judges’’ in the language of our
research. Judges have earned superior rights and make the final decisions about market-value suggestions that are posted by
community members. Thus, the community creates market values based on criteria that are different from case to case and
that are based on their judges’ decisions, rather than creating values based on a formal approach, for example, if all
community members evaluated a fixed set of variables and decided democratically. A democratic decision process would
also imply an equal say for each community member that participates in the discussion about a player’s market value
(Morrell, 1999; Sherry, 2007).

[(Fig._1)TD$IG]
Market value
True market Attributes (talent Judge’s final
estimations by
value of and external evaluation of
community
an athlete variables) market value
members A and B

X1

X2 YeA

Yt X3 Yf

X4 YeB

X5

Fig. 1. Brunswik’s lens model of decision making for team decisions.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

4 S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx

According to the language of the literature on prediction markets (e.g., Wolfers & Zitzewitz, 2004), the judges are able to
estimate the dispersion of market value estimates by analyzing the discussion threads. Judges can evaluate if ‘‘the market’’ is
rather sure or if participants indicate a high degree of uncertainty. Additionally, a mere statistical approach might be
inappropriate, especially for non-prominent players with only a handful of estimates by community members (compared to
a prediction market with low liquidity). For these players the judge’s own valuation expertise is even more valuable than it is
for superstar players with hundreds of value suggestions.
To sum up, the judge performs the complex task of filtering, weighting, and aggregating information by taking into
account the source of information and the reasons given as justifications for particular estimates. Thus, decision making by
team leaders (judges) might be a more efficient organizational principle, compared to a democratic and thereby inflexible
procedure, e.g., computing the mean or median of all discussed market-value suggestions.
Hollenbeck et al. (1995) have shown that decision-making team leaders are sensitive to the quality of advice they receive
from team members and that they adjust their judgments in subsequent decisions according to the feedback they get. In the
[Link] community, leaders (judges) receive external feedback when a player is transferred and the actual market
value becomes publicly available. Thus, the information aggregation ability of the [Link] community can be
evaluated by comparing the a priori estimated market value Yf with the corresponding actual transfer fee.

4. Accuracy of the community’s market-value evaluations

Market values provided by [Link] have a good reputation in the sports industry and have a high economic
relevance; they are used in actual transfer and salary negotiations. Also they have been used as a proxy for income or market
value in scientific research (Bryson, Frick, & Simmons, 2013; Franck & Nüesch, 2012). In order to evaluate the quality of
market-value estimates that the users of the online community [Link] (Yf in Fig. 1) provide, we examined the
relation of these market values to actual transfer fees the clubs paid.
It is important to note the conceptual difference between actual transfer fees and estimated market values. A transfer fee
– the actual value of a player for a particular club during a particular purchase/selling decision – can differ from that value,
e.g., due to the length of the remaining contract between the player and his or her club, due to strategic reasons (weakening a
competitor by buying key players, even for a higher price), or due to the bargaining power of the buying and selling club
(Bryson et al., 2013; Frick, 2007). Despite some conceptual differences, actual transfer fees serve as a reasonable proxy to
validate the online community’s market-value estimates.
For the analysis, we considered every single actual transfer that took place in the top German soccer league ‘‘Bundesliga’’
during the winter transfer window 2011–2012 (n = 67). We did not find any systematic differences between transferred and
non-transferred players in our data. Due to lack of evidence of a bias, a correction was not necessary. A regression of market
values that were estimated by the community before the transfers became publicly known on actual transfer fees, using
robust standard errors (White, 1980), indicates that these market values explain most of the variance of actual transfer fees
(R2 = 0.90, b = 0.95, p < 0.01). Thus, we find that the community’s market-value estimates are related to actual transfer fees
and may serve as predictors, justifying their high relevance for sports managers.

5. Indicators of market value

Brunswik’s lens model suggests that not all available attributes are equally used for final judgments on the target variable.
Both regular community members and judges ‘‘look through’’ their individual lenses in order to evaluate market values. The
next section analyzes attributes (Xi in Fig. 1) that are most frequently used in justifying market-value estimations made by
the online community [Link] (Yf in Fig. 1). Beforehand, we provide some explanation for our categorization of
these attributes.

5.1. Sources and conceptual background for the selection of attributes

Frick (2007) has identified two streams of research on human capital in professional sports, especially on professional
soccer players: On the one hand, researchers have analyzed the determinants of transfer fees. On the other hand, researchers
have explained professional soccer players’ remuneration. Frick has found that ‘‘the observable variation in transfer fees can
largely be explained by the same variables that also affect player salaries. Player age, career games played, and international
caps all have a positive yet decreasing influence on the amount of money paid for the services of a player’’ (Frick, 2007,
p. 431).
A number of empirical studies on the values or salaries of sports players have relied on differences in individual
characteristics, such as age or footedness (Brandes & Franck, 2012; Bryson et al., 2013; Carmichael, Forrest, & Simmons,
1999; Dobson & Gerrard, 1999; Frick, 2007; Speight & Thomas, 1997). However, most of these studies mix individual
characteristics with rather external variables that provide additional quality signals, e.g., games played. Adler (1985) has
stressed the importance of external variables in the superstar formation in addition to superior talent. Superstars might
emerge among equally gifted performers because of positive network externalities of popularity. Also subsequent literature
on superstar effects (e.g., Franck & Nüesch, 2012) indicates a distinction between more objective, talent-associated measures
(e.g., age, percentage of duels won, goals scored, percentage of successful passes) and popularity as an external, more

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx 5

subjective variable. Franck and Nüesch (2012) have operationalized popularity by examining citations in newspapers and
weekly magazines. We enhance this view and discuss various external variables that can reflect players’ popularity (in
addition to press citations). In previous studies on transfer fees, talent and external variables were used undifferentiated
(e.g., Carmichael et al., 1999). We enhance this view by noting that some of these variables reflect players’ popularity rather
than their talent (e.g., number of appearances as decisions by team coaches). Following this approach, we analyze the
predictive power of talent as well as of external attributes that are most frequently used to justify market-value evaluations
by the [Link] community.

5.2. Method and measures

This study tested the influence of different player attributes (Xi in Fig. 1) on market values that are proposed by the
[Link] online community. We focused on 338 active athletes (all players who participated at least in one match)
who played in the top German soccer league (Bundesliga) during the first half of the season 2011/2012. We used data from
the [Link] community, combined with data from other Internet sources. Because the community renews or
confirms market values for each player every two to six months, and we assume that market values respond to talent and
external factors, we included a time gap in the data collection. In particular, data concerning the predictor variables was
collected in the winter break of the season 2011/2012, while data for the dependent variable (market value) was collected in
July 2012 (for data sources see Table 1).
In order to include the most potential indicators (attributes) in our research design we selected variables that were most
frequently mentioned in the online community’s discussion forums about market values (for frequencies also see Table 1).
These variables have also been used in previous research on market values and salaries in sports (for sources in literature also
see Table 1). In particular, we focused on five talent variables: age, precision, success, assertion, and flexibility, and we
employed five external variables that reflect players’ popularity with experts such as club managers, coaches, agents, and
sports media as well as public attention.
While preparing the analysis, we checked if all predictor variables were related to the dependent variable ‘‘market value’’
in a linear way. All variables except age fulfilled this criterion. For age, we found a non-linear (quadratic) relationship with
market value, which is in line with previous research. In order to include the variable in the regression model, we used the
empirical quadratic function for calculating the positive deviation (absolute value) from the age with maximum market
value (vertex = 24.16 years, comparable to 25.4 years, in Lehmann & Schulze, 2008). When we compared this corrected linear

Table 1
Player attributes and descriptives.

Attributes (indicators for Description and operationalization Talent indicator/ M (SD) Number
market value or income, external variable of posts
e.g., in (source))

X1: Age (Franck & Age (years)/absolute deviation from Talent 25.30 (3.96)/3.32 (2.45) >500
Nüesch, 2012) age with maximum value (24,16)
X2: Precision % Successful passes/corrected by mean Talent 72.75 (23.03)/0.00 (5.63) >500
(Franck & Nüesch, 2012) at positiona
X3: Scoring Number of goals and assists/corrected Talent 1.97 (3.03)/0.00 (2.78) >500
(Franck & Nüesch, 2012) by mean at positiona
X4: Assertion Won duels or save to shots ratio of Talent 51.48 (11.58)/0.00 (7.03) ca. 200
(Franck & Nüesch, 2012) goalkeeper/corrected by mean at posi-
tiona
X5: Flexibility Ability to play predominantly with one Talent 0.24 (0.43) ca. 50
(Bryson et al., 2013) foot (0), or equally good with both feet
(1)
X6: Club management Financial success/recruitment effec- External 88 960 661.76 (61 474 927.19) >500
(Treadway et al., 2012) tiveness operationalized by the sum
of market values of all teammates (s)
X7: Team coach Number of league games played (Bun- External 11.16 (4.87) ca. 250
(Franck & Nüesch, 2012) desliga)a
X8: Sports agent Average market value of all other External 1 780 392.82 (1 500 542.02) ca. 150
(Mason & Slack, 2001) players in the portfolio of his sport
agent (s)
X9: Experts in media Average grades (1–6) from experts of External 3.70 (0.52) ca. 350
(Bryson et al., 2013) three different sports portalsa
X10: Public attention Total number of links reported by the External 536 265.59 (953 119.50) ca. 400
(Garcia-del-Barrio & Internet search engine Google
Pujol, 2007)
Yf: Market value (Franck Market value (s), defined by the online Depend. 4 447 205.88/(5 990 666.25) All
& Nüesch, 2012) communityb variable
Data sources: Yf; X1; X5; X6; X7; X9: [Link]; X2; X3; X4: [Link]; X8: [Link]; [Link]; [Link]; X10:
[Link].
a
In winter break of season 2011/12.
b
At the end of season in July 2012.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

6 S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx

Table 2
Impact of talent and external attributes on market values.

Dependent variable Market value according to [Link]

Model and estimation Model 1 Model 2 Model 2

(OLS, White robust s.e.) (OLS, White robust s.e.) (95% quantile reg.)

N = 338 Coefficient t Coefficient t Coefficient t

Talent
Age 539 439.66*** 5.96 518 397.09*** 5.79 523 858.46*** 3.94
Precision 215 718.34*** 4.61 42 744.08 1.40 14 745.82 0.41
Scoring 1 242 983.01*** 6.88 356 796.37** 2.25 397 328.82** 2.40
Assertion 71 288.01** 2.10 14 068.05 0.51 39 595.77 1.13
Flexibility 1 425 951.13* 1.95 637 542.95 1.41 553 709.52 0.79

External attributes
Club management 0.03*** 5.19 0.04*** 3.65
Team coach 169 514.16*** 4.03 42 528.77 0.54
Sports agent 0.17 0.79 0.08 0.20
Experts in media 1 166 920.06*** 2.41 2 298 657.88*** 3.16
Public attention 2.27*** 3.69 3.93*** 6.04
Intercept 5 897 393.31*** 14.24 3 975 657.77** 1.81 11 397 816.18*** 2.84
R2/pseudo-R2 0.42 0.70 0.65

* p < 0.1
** p < 0.05.
*** p < 0.01.

function of age with the original quadratic function we found both functions explained the same amount of market value
variance (R2 = 0.024). To avoid multicollinearity, we used the corrected linear term of age instead of both, age and age2.
Requirements in respect to some talent variables (e.g., strikes, assists, precision, and assertion) vary depending on the
players’ positions on the pitch (Tavana, Azizi, Azizi, & Behzadian, 2013). The main task of a forward is to score (goal or assist),
whereas defenders are expected to win duels and pass the ball precisely to other players on their teams. Thus, we corrected
these four variables with mean values for the particular positions (for further possibilities see Bryson et al., 2013; Franck &
Nüesch, 2012). For the fifth talent variable, ‘‘footedness,’’ Bryson et al. (2013) have argued that being adept with both feet
(‘‘two-footedness’’) is a generally advantageous skill that directly affects player performance and that might be rewarded in
the labor market. ‘‘Two-footedness’’ gives team management the opportunity to use a player in various positions on the pitch
and this utility may raise a player’s market value. Consequently, we did not correct this talent variable with mean values for
position.
Aside from talent variables, our model includes five external attributes, that is, player evaluations by experts as team
coaches or the public. For example, sports experts from the media evaluate player performances using discrete grades from
1 = excellent to 6 = very bad and publish them in various online and offline media. We found a high correlation of grades
among three different sources (three widely read German special-interest sports media sources: [Link], [Link], and
[Link]: rkicker-sportal = 0.83, rkicker-bild = 0.76, rbild-sportal = 0.75; all p < 0.01). Thus, we used average grades in our model. Before
the final analysis, we checked for differences between positions with respect to external variables. In contrast to prior
research (Bryson et al., 2013), we did not find any systematic differences; thus we did not correct the external variables for
position.
Furthermore, soccer players add to the economic profitability of their clubs – through broadcast rights, merchandizing
contracts, etc. – in addition to direct contributions in sporting performances. Due to their revenue-generating capability,
clubs are willing to pay more for popular players (‘‘superstars’’). In line with previous work (e.g., Garcia-del-Barrio & Pujol,
2007), we also measured this popularity by the total number of hits reported by Google.
In addition, financially successful clubs and those with excellent recruitment skills hire valuable players. Thus, being
under contract with such a club is also a quality signal for the player (Garcia-del-Barrio & Pujol, 2007). We propose that the
same interdependency exists for a player’s relationship with his or her sports agent. The literature remains almost silent on
the impact of such agents on the sport economy (for an exception see Mason & Slack, 2001), despite the fact that almost every
professional soccer player has signed a contract with a sports agent.
Table 2 shows the results of a regression-based analysis on the market values of 338 active Bundesliga soccer players in
the season 2011/2012.2 Table 1 provides descriptive statistics and Table A.1 (appendix) offers correlations for all variables
used in this study. To account for heteroskedasticity in the OLS-estimation we used heteroskedasticity-consistent standard
errors as proposed by White (1980). The results are provided in Table 2. Although Model 1 (R2 = 0.42) includes talent
indicators only, Model 2 (R2 = 0.70) includes both talent indicators and external variables and is able to explain market value

2
To integrate the possible role of the judges on [Link], we did set up a two-level regression model including judge experience (number of posts
of the respective judge) as a judge-level explanatory variable. However, we did not get any additional insights, as the judges had neither a significant impact
on explaining market value nor could we find any significant cross-level interactions.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx 7

to a high degree. Model 2 demonstrates a unique contribution of external variables to explain the variance of market values
(significant difference in R2; p < 0.01). In both models, the variance inflation factors as well as the tolerance measures
indicated no problems of multicollinearity.
Age, precision, and scoring – the most easily identifiable (see number of posts in Table 1) and probably the most
‘‘objective’’ talent variables – have a sizable influence on market value in Model 1. That is, the more a player’s age diverges
from the average age the lower his or her market value is. The more precise a player passes and the more goals or assists he
scores the higher his value is. Furthermore, the data replicates the findings of Bryson et al. (2013) who find that the talent
variable ‘‘flexibility’’ (the ability to play equally well with both feet) positively influences the market value of a talent. In
addition, we found a significant effect of ‘‘assertion.’’
When entering the second block of external variables into the model we found a vanishing influence of talent variables
(precision, assertion, and flexibility) and a strong positive influence of external quality signals from club managers (hiring a
talent for a successful team), team coaches (selecting a talent to play in most matches), and most importantly, from attention
from the general public (Google hits). Good grades from experts in the media positively influenced a talent’s market value
(low numbers indicated high performance). However, we did not find significant effects for quality signals from personal
sports agents (e.g., agents that contract with valuable clients only).
For further investigation on the superstar phenomenon we applied a 95% quantile regression, as suggested by Lehmann
and Schulze (2008). The results in the third column of Table 2 show the equation coefficients for the 95 percentile of the same
sample and indicate that the impact of the explanatory variables on market value (as determined by the crowd on
[Link]) differs for the top-market value players, mostly in accordance with the rationale of Adler (1985). Whereas
the talent variables roughly exert the same influence on market value in comparison to the mean-regression, the impact of
the external attributes ‘‘grades from experts’’ and ‘‘public attention’’ is stronger.

6. Discussion

6.1. Summary and managerial implications

A major challenge for sports managers is to evaluate the value of an individual athlete, e.g., in the context of a transfer
between two sport clubs. This research investigates how and how well an online community performs the complex task of
human capital valuation, and it shows which attributes are most indicative of a player’s value. In the context of German
soccer, aficionados – not professionals – have built an online community that evaluates professional soccer players’ market
values. The community has become the main source for reporting market values in the media and has a strong impact on the
real sports economy: In fact, it is used in real market transactions and wage negotiations, indicating the power of crowd
wisdom in the sports management context.
This research introduces an adaptation of Brunswik’s lens model for studying organizational principles in online
communities. After applying this adaptation, we discovered that an online community – which does not follow a democratic
principle (‘‘an equal say’’) – performs very well on the complex task of human capital evaluation. In particular, we found that
the community’s market-value estimates are highly predictive for actual transfer fees. Furthermore, our research describes
the judge principle (selective approach) as a powerful alternative to prediction markets (integrative approach). This principle
takes into account which individual provides what kind of information and leaves room for interpretation and weighting by
appointed experts. Consequently, the selective design of collecting crowd wisdom in online communities combined with
interpretation by expert judges could possibly outperform a design where market participants trade a virtual stock based on
their private information.
Moreover, this research shows which attributes are most important for the evaluations of the online community. When
we looked for attributes that were indicative of most of the market-value judgments, we found that the community’s
estimates can largely be predicted using an econometric model that contains two blocks of determinants: talent-related
measures and variables that result from judgments of external experts (e.g., coaches or journalists). In Model 2, the selected
10 variables explain 70% of variance, and we derive our main finding from a comparison of Model 1 and 2. That is, various
external variables (e.g., sum of market values of teammates) are indicative for market values and contribute independently
from talent variables to explain market-value judgments that are made by an online community. This puts forward the idea
of a more balanced perspective when evaluating a soccer players’ market values. Aside from talent, external factors
associated with the players need to be given more attention. It is important to note that these external variables – and not
talent variables only – have a large impact on market values. In particular, the variables club management (sum of market
values of all teammates), experts in media (grades), and public attention (Google hits) have a significant impact, which is
even stronger for the superstars of the league (top 5% of market value). Therefore, organizing variables into two blocks of
talent and external determinants provides a more differentiated look at the ‘‘popularity’’ of players than recent literature on
the superstar phenomenon.
Given the limitations of our dataset, prediction is not the main focus of our research. However, we believe that our model
provides sports managers with indicators for potentially under- or overvalued players (e.g., in order to prepare an actual
transfer), provides a possibility of assuring quality management in the community that predicts market values, and provides
practitioners in the field of sports management and the scientific community with a starting point to develop better tools and
methods for market-value prediction.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

8 S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx

6.2. Limitations and further research possibilities

Recent literature has demonstrated that a number of additional variables (mental abilities, speed, etc.) are likely to affect
player selection (Tavana et al., 2013) and salaries (Frick, 2011). However, even systematic research on metrics on human
capital evaluation lacks agreement on variables that should be included (Lim, Chan, & Dallimore, 2010). By organizing the
potential drivers of market value into two main categories, the contribution of our approach goes beyond the coefficients of
our regression models and lies in a call for a more balanced evaluation of soccer players by using a number of talent and
external variables. Including additional variables in the model that belong to one of these two blocks (talent variables vs.
external variables) is therefore left for future research. Additional variables may increase model fit but will not enhance our
knowledge in respect to this main finding.
This research is the first to evaluate the influence of sports agents on market values. Surprisingly, our results indicate that the
average value of a player in an agent’s portfolio does not influence the market value of each individual client. This is also of high
practical interest because almost every professional soccer player – as well as many athletes in other disciplines – has signed a
contract with a sports agent. Because the literature remains almost silent on the impact of sport agents on the sport economy, a
further evaluation of the relationship between sports agents and market values is an interesting area for further research.
Furthermore, all kinds of alternative econometric models can only be approximations of actual communication in online
communities. Further research should use methods (e.g., content analysis) that allow researchers to analyze actual discussions
(Saavedra et al., 2011). A deeper analysis of the content that is discussed in communities would also allow for further comparisons
of the effectiveness of more structured ‘‘democratic’’ evaluation processes – where attributes and discussants are treated equally
– and more flexible less ‘‘democratic’’ decision processes as demonstrated in the judge principle described in this manuscript.

Appendix A

Tables A.1 and A.2.

Table A.1
Correlation matrix.

Yf X1 X2 X3 X4 X5 X6 X7 X8 X9

X1 0.181*
X2 0.191* 0.097*
X3 0.567* 0.026 0.017
X4 0.083 0.060 0.283* 0.061
X5 0.111* 0.001 0.025 0.038 0.089*
X6 0.599* 0.019 0.268* 0.210* 0.145* 0.072
X7 0.372* 0.049 0.080 0.449* 0.118* 0.094* 0.056
X8 0.357* 0.007 0.111* 0.246* 0.058 0.039 0.264* 0.124
X9 0.472* 0.020 0.132* 0.489* 0.120* 0.026 0.291* 0.420* 0.207*
X10 0.682* 0.090* 0.143* 0.592* 0.006 0.143* 0.433* 0.230* 0.392* 0.342*

* p < 0.05 (two-tailed).

Table A.2
Standardized coefficients.

Dependent variable Market value according to [Link]

Model and estimation Model 1 (OLS, White robust Model 1 standardized Model 2 (OLS, White robust Model 2 standardized
s.e.) s.e.)

N = 338 Coefficient t Coefficient Coefficient t Coefficient

Talent
Age 539 439.66*** 5.96 0.22 518 397.09*** 5.79 0.21
Precision 215 718.34*** 4.61 0.20 42 744.08 1.40 0.04
Scoring 1 242 983.01*** 6.88 0.58 356 796.37** 2.25 0.17
Assertion 71 288.01** 2.10 0.08 14 068.05 0.51 0.02
Flexibility 1 425 951.13* 1.95 0.10 637 542.95 1.41 0.05

External attributes
Club management 0.03*** 5.19 0.34
Team coach 169 514.16*** 4.03 0.14
Sports agent 0.17 0.79 0.04
Experts in media 1 166 920.06*** 2.41 0.10
Public attention 2.27*** 3.69 0.36
Intercept 5 897 393.31*** 14.24 3 975 657.77** 1.81
R2/pseudo-R2 0.42 0.70

* p < 0.1
** p < 0.05.
*** p < 0.01.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006
G Model
SMR-254; No. of Pages 9

S. Herm et al. / Sport Management Review xxx (2014) xxx–xxx 9

References

Adams, R. B., & Ferreira, D. (2009). Moderation in groups: Evidence from betting on ice break-ups in Alaska. Review of Economic Studies, 77, 882–913.
Adler, M. (1985). Stardom and talent. American Economic Review, 75, 208–212.
Afuah, A., & Tucci, C. L. (2012). Crowdsourcing as a solution to distant search. Academy of Management Review, 37, 355–375.
Bachrach, Y., Graepel, T., Kasneci, G., Kosinski, M., & Van Gael, J. (2012). Crowd IQ – aggregating opinions to boost performance. In Proceedings of the 11th
international conference on autonomous agents and multiagent systems (AAMAS) (pp. 535–542).
Brandes, L., & Franck, E. (2012). Social preferences or personal career concerns? Field evidence on positive and negative reciprocity in the workplace. Journal of
Economic Psychology, 33, 925–939.
Brehmer, B., & Hagafors, R. (1986). Use of experts in complex decision making: A paradigm for the study of staff work. Organizational Behavior and Human Decision
Processes, 38, 181–195.
Brunswik, E. (1952). The conceptual framework of psychology. Chicago: The University of Chicago Press.
Bryson, A., Frick, B., & Simmons, R. (2013). The returns to scarce talent: Footedness and player remuneration in European soccer. Journal of Sports Economics, 14(6),
606–628.
Carmichael, F., Forrest, D., & Simmons, R. (1999). The labour market in association football: Who gets transferred and for how much? Bulletin of Economic Research,
51, 125–150.
Charness, G., & Sutter, M. (2012). Groups make better self-interested decisions. Journal of Economic Perspectives, 26, 157–176.
Cooksey, R. W. (1996). Judgment analysis: Theory, methods, and applications. San Diego, CA: Academic Press.
Cowgill, B., Wolfers, J., & Zitzewitz, E. (2009). Using prediction markets to track information flows: Evidence from Google. Working Paper.
Dobson, S., & Gerrard, B. (1999). The determination of player transfer fees in English professional soccer. Journal of Sport Management, 13, 259–279.
Franck, E., & Nüesch, S. (2012). Talent and/or popularity – what does it take to be a superstar? Economic Inquiry, 50, 202–216.
Frick, B. (2007). The football players’ labor market: Empirical evidence from the major European leagues. Scottish Journal of Political Economy, 54, 422–446.
Frick, B. (2011). Performance, salaries, and contract length: Empirical evidence from German soccer. International Journal of Sport Finance, 6, 87–122.
Galton, F. (1907). Vox populi. Nature, 450.
Garcia-del-Barrio, P., & Pujol, F. (2007). Hidden monopsony rents in winner-take-all markets – Sport and economic contribution of Spanish soccer players.
Managerial and Decision Economics, 28, 57–70.
Hollenbeck, J. R., Ilgen, D. R., Sego, D. J., Hedlund, J., Major, D. A., & Phillips, J. (1995). Mutlilevel theory of team decision making: Decision performance in teams
incorporating distributed expertise. Journal of Applied Psychology, 80, 292–316.
Jeppesen, L. B., & Lakhani, K. R. (2010). Marginality and problem solving effectiveness in broadcast search. Organization Science, 21, 1016–1033.
Lehmann, E. E., & Schulze, G. G. (2008). What does it take to be a star? – The role of performance and the media for German soccer players. Applied Economics
Quarterly, 54, 59–70.
Lim, L., Chan, C., & Dallimore, P. (2010). Perceptions of human capital measures: From corporate executives and investors. Journal of Business & Psychology, 25, 673–
688.
Mason, D. S., & Slack, T. (2001). Industry factors and the changing dynamics of the player–agent relationship in professional ice hockey. Sport Management Review,
4, 165–191.
Morrell, M. E. (1999). Citizen’s evaluations of participatory democratic procedures: Normative theory meets empirical science. Political Research Quarterly, 52,
293–322.
Mosier, K. L., & Fischer, U. M. (2010). Judgment and decision making by individuals and teams: Issues, models, and applications. Reviews of Human Factors and
Ergonomics, 6, 198–256.
Rosen, S. (1981). The economics of superstars. American Economic Review, 71, 845–858.
Saavedra, S., Duch, J., & Uzzi, B. (2011). Tracking traders’ understanding of the market using e-communication data. PLoS ONE, 10, e26705.
Schenk, E., & Guittard, C. (2011). Towards a characterization of crowdsourcing practices. Journal of Innovation Economics, 7, 93–107.
Sherry, S. (2007). Democracy and the death of knowledge. University of Cincinnati Law Review, 75, 1053.
Speight, A., & Thomas, D. (1997). Football league transfers: A comparison of negotiated fees with arbitration settelments. Applied Economics Letters, 4, 41–44.
Surowiecki, J. (2005). The wisdom of crowds: Why the many are smarter than the few and how collective wisdom shapes business, economies, societies and nations. New
York.
Tavana, M., Azizi, F., Azizi, F., & Behzadian, M. (2013). A fuzzy inference system with application to player selection and team formation in multi-player sports.
Sport Management Review, 16, 97–110.
Treadway, D. C., Adams, G., Hanes, T. J., Perrewé, P. L., Magnusen, M. J., & Ferris, G. R. (2012). The roles of recruiter political skill and performance resource
leveraging in NCAA football recruitment effectiveness. Journal of Management [Link]
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48, 817–838.
Wolfers, J., & Zitzewitz, E. (2004). Prediction markets. Journal of Economic Perspectives, American Economic Association, 18, 107–126.

Steffen Herm is a professor of marketing at the Hochschule für Technik und Wirtschaft Berlin. He holds a Ph.D. and a postdoctoral degree in Marketing from the
Technische Universtät (TU) Berlin. His research, related to innovation, product design, entrepreneurship, branding, and consumer behavior is published in
journals as the Journal of Retailing and the Journal of Product Innovation Management.

Hans-Markus Callsen-Bracker is an project manager Projektträger Jülich and was an assistant professor of finance at the TU Berlin. His research is related to
pricing and risk management.

Henning Kreis is a assistant professor for market communications at Freie Universität Berlin. He holds a Ph.D. in Marketing from Humboldt-Universität zu Berlin.
His research, related to word-of-mouth communication, customer loyalty and quantitative methods for data analysis is published in journals as the Journal of
Business Market Management and Marketing ZFP.

Please cite this article in press as: Herm, S., et al., When the crowd evaluates soccer players’ market values: Accuracy and
evaluation attributes of an online community. Sport Management Review (2014), [Link]
[Link].2013.12.006