Opportunistic behaviour of players’ agents in football and its monitoring by the players—an empirical analysis from the perspective of the players

This article explores the following questions: (1) To what extent and in relation to which activities do football players perceive actions of their agents that are directed against their own interests depending on their performance levels? (2) Are players’ agents monitored by their players, and if so, in relation to which activities? (3) Which factors influence the monitoring behaviour of players? The relationship between the actors is modelled as a principal–agent relationship. The explorative study includes a dataset of 336 professional German football players. A logistic regression model was used to consider relevant factors influencing players’ monitoring behaviour. The results indicate that players’ agents across all leagues take actions potentially harmful to the players and show that approximately 50–60% of the players had their current agent monitored. Monitoring behaviour is influenced by ‘selection criteria’ and ‘framework conditions of the working relationship’, but not to the extent that could be expected. The research helps to understand the increasingly important relationship between players and agents. Players need to be aware that agents pursue their own objectives, which could be against the players interests. Therefore, monitoring is necessary. But data also show a lack of transparency, so that players do not know what agents are allowed to do and what is forbidden. If the associations and/or the players’ unions want to prevent possible damage to the players by their agents, it seems advisable to intensify and support information and awareness-raising campaigns or to adopt new statutes that regulate the players’ agents market more closely.


Introduction
Players' agents are important contacts and business partners in football. Players rely on their agents to execute tasks in their favour. However, if an agent puts his or her own interests above the player's interests, this may result in disadvantages for the player. For example, the French national footballer N'Golo Kantè describes that his former agent tried to demand money from him for negotiations with betting providers regarding the exploitation of image rights, although the agent already knew that Kantè would reject such an idea on principle (Baumeister, 2019). Not only absolute top players are affected by possible misconduct by their agents. The TV channel BR24 reported that several Brazilian soccer talents were lured to Germany with promises by agents. The players were required to pay a fee and were then unable to contact the agents after payment was made (Hennl, 2017). The examples show that players cannot unconditionally rely on their agents to always act in their interests. It therefore appears advisable for players to consistently monitor the work performance of their agents to avoid monetary, career-specific, or image-specific damage. It has not yet been clarified if and to what extent players monitor their agents and which factors influence monitoring. This article examines the following central research questions:

1.
To what extent and in relation to which activities do football players perceive actions of their agents that are directed against their own interests depending on their performance levels?
2. Are players' agents monitored by their players, and if so, in relation to which activities?

Literature review
Findings on the actual service provision of players' agents and on the quality assessment of these services from the perspective of footballers essentially relate to one publication. Kelly and Chatziefstathiou (2018) asked 25 footballers, five agents, and 20 club managers for their opinions on players' agents. Most managers report feeling that the agent market is largely corrupt and that players do not need agents. The players and the agents themselves have a rather ambivalent view of the market. In their opinion, there is a balance between "good" and "bad" agents. Regardless of this, however, both player and agent statements indicate agent actions that are not in the interests of the players.
Studies of the National Hockey League (NHL) present a more general examination of the state of research regarding players' agents in sport. Mason (1999) considers the principal-agent theory to be well suited to identify and understand problems within the player-agent relationship. Building on this, Mason and Slack (2001a) show that industry factors affecting league and team revenue, information asymmetry, agent monitoring, and the ability of principals and agents to enter into contracts significantly affect principal-agent relationships. Subsequently, players are more likely to pay their agents a salary or a flat fee than to use the conventional (result-based) commission method. In a further study, Mason and Slack (2001b) state that the available surveillance solutions for agent opportunism were found to be flawed, even though National Hockey League Players' Association (NHLPA) regulations initially appear to be the best alternative. Combined with the results of the Mason and Slack (2003) study, it appears that concerns about agent reputation, agent competition, agent certification, and salary disclosure cumulatively reduce information asymmetry, favour the agent, and reduce the likelihood of agent opportunism.
In summary, the academic work to date clearly shows that the principalagent theory is a good basis for working out problems between players and their agents, and that players-agent markets have been the focus of previous studies. So far, however, there is little knowledge regarding possible problems in the performance relationship between player and agent, especially in football. The available results of the qualitative (Kelly & Chatziefstathiou, 2018) study indicate that possible problems are not merely isolated cases, but instead they affect the industry.
Therefore, it seems advisable to further investigate actions that are not in the players' interest, and to analyse subsequent consequences for the players-in the sense of monitoring agents.

Theoretical framework
The relationship between the actors is modelled as a principal-agent relationship (Ross, 1973). Building on this, factors that can influence the watching (monitoring) behaviour of the players are reflected. The factors derived analytically with reference to the principal-agent theory are assigned to the superordinate categories 'selection criteria' (before conclusion of the contract) and 'framework conditions of the working relationship' (after finalising the contract).

Opportunistic behaviour and monitoring in the relationship between player and agent
The principal (footballer) assigns certain tasks and subsequent decision-making powers to the agent (players' agent), so the agent can support the footballer in achieving his or her goals. For example, footballers want to play in clubs that are as successful as possible, sign high-paying contracts, make successful investments, or open up post-football careers. In return, the agent receives remuneration for fulfilling these tasks (Breuer, 2015) in the form of a negotiated percentage of the player's basic gross income for the entire duration of the relevant employment contract (FIFA, 2015, Art. 7). For the player, this relationship harbours the risk that the agent will not act in the sense of the agreed assignment (Göbel, 2002).
In general, the problems that can arise between the principal and the agent are referred to as agency problems (Göbel, 2002). The basic problem is that individual maximisation of benefits can lead to conflicting goals. For example, the player as a principal would like to optimise their own benefits in terms of remuneration, sporting success and prestige as quickly as possible, which is why the agent is contacted (Parensen, 2013). By contrast, the agents' engagement aims at earning the highest possible remuneration and developing a reputation to differentiate themselves in the competition for lucrative contracts from other footballers.
In doing so, the agent offsets individual disadvantages (workload, costs, time) against individual advantages (remuneration, reputation; Eisenhardt, 1989;Heath, 2009). This deliberation process can lead to opportunistic agent behaviour and disadvantages for the player (Williamson, 1975). Based on these basic problems and the tasks that an agent usually takes on (Jungels, Förch, & Riedl, 2017) it is conceivable that, for example an agent might pressure the player to sign a contract that the player is not convinced about. In addition, an agent's misconduct could result in a contract between the player and the club or sponsors not being finalised or in a player having to pay a penalty to a contractual partner (Buschmann & Wulzinger, 2019). Furthermore, if an agent urges the player to sign an exclusive contract with them, then the footballer's options for action and choices are severely restricted, as he or she can only fall back on the services of one agent (Parensen, 2012). If a player believes they must comply with this clause, it is also irrelevant that exclusive contracts-at least with regard to the brokerage activity-are not permitted by law ( § 297 No. 4 SGB III).
These potential post-contract problems are exacerbated by an asymmetrical distribution of information. Ultimately, the player can often only insufficiently observe and assess the actions/efforts of their agent (hidden action or hidden information; Keser & Willinger, 2007), which increases the incentive for opportunistic behaviour (moral hazard; Sappington, 1991). Another incentive for agents to behave opportunistically may arise from the fact that clubs (sometimes) pay the agents, instead of the footballers who originally hired the agent, although Fédération Internationale de Football Association (FIFA) regulations say that players (principals) must bear the costs (Art. 7). From the players' point of view, this could be seen as good negotiation skills by the agent, since they do not have to pay the agent themselves. Or it could be an incentive for the agents to act in the club's interest, because the agents see the club as their principal. 1 It can also be assumed that players report more problems with their previous players' agent than with their current agent, as possible problems are only noticed with a delay, or the problems were so serious that they contributed to the termination of the contractual relationship (Göbel, 2002). According to the findings of Mason and Slack (2003), the performance level of the players could also influence agents' opportunistic behaviour, through the effect that the incentive for opportunistic behaviour decreases as the player's performance level increases, since the opportunity costs (loss of reputation, loss of income, compensation payments) for the agent increase significantly when misconduct is discovered.
Monitoring is a tool to fight against such problems. It means that the principal (the player) is watching the agent (Göbel, 2002). According to agency theory in football, monitoring is the attempt to create tools to control agent behaviour that can be seen as an effort to increase information to the player, by reducing information asymmetries that favour the agent (Mason & Slack, 2001b). Since the player usually lacks both the specific expertise and the time to monitor the agent, the player will delegate control of the agent to a third party. The control mostly relates to specific outputs of the agent, such as club offers, contract negotiations, sponsorship agreements, or investment offers, and can be carried out by other agents or experts (lawyer, finan-1 A third option is that an agent can also be commissioned directly by a club (principal). In this case, an agent receives a fixed amount that is negotiated before the contract is signed. This article does not focus on this constellation. cial, tax consultant) as well as confidants (family, friends) without specific expertise.
Against the background of possible problems, it appears expedient from the players' point of view to use monitoring measures, even if this involves additional work (monitoring costs) for them and reduces their own benefit (Mason & Slack, 2001b). Monitoring enables the footballer to uncover agent misconduct or to increase the likelihood of detection, so that the agent refrains from acting in a manner that is potentially harmful to the player.
As another solution to avoid future opportunistic behaviour, a reporting office could help players to reduce information asymmetries about an agent's qualities. Thus, players can report incidents and the reporting office can inform other players. This would make agents' opportunistic behaviour less attractive in the long run, due to the risk of losing current and future clients (Göbel, 2002). However, in football, there is no official reporting office. Players can report incidents with agents to the FIFA, the DFB/DFL, or their club, but without being able to fall back on a structured and established process.
Monitoring influencing factors 1. Selection criteria: Monitoring is only possible after signing a contract. Nevertheless, the player has the option of reducing the likelihood of opportunistic behaviour and thus the need for monitoring even before the contract is signed-by selecting the 'right' agent. The player can use screening instruments (Auronen, 2003) to counteract the problem of hidden characteristics. 2 For example, the recommendation of an agent by someone trusted (players, clubs, or parents) could represent a quality characteristic for the player (Schölermann, 2003) and makes further monitoring seem unnecessary. In this context, an of the working relationship' , but not to the extent that could be expected. The research helps to understand the increasingly important relationship between players and agents. Players need to be aware that agents pursue their own objectives, which could be against the players interests. Therefore, monitoring is necessary. But data also show a lack of transparency, so that players do not know what agents are allowed to do and what is forbidden. If the associations and/or the players' unions want to prevent possible damage to the players by their agents, it seems advisable to intensify and support information and awareness-raising campaigns or to adopt new statutes that regulate the players' agents market more closely.

Keywords
Players' agents · Principal-agent relationship · Monitoring · Intermediaries · Football agents agent's good reputation is a quality characteristic and builds trust with the player, which is why the player may tend to forgo a specific control (Iossa & Rey, 2014;McLean, 1997). Furthermore, the formal qualification of the agent may act as a quality attribute, so that the player forgoes additional monitoring of the agent (Bathi, 2015). 2. Monitoring depending on the framework conditions of the service relationship: The respective framework conditions of the service relationship between the player and the agent are likely to have an impact on monitoring. It can be assumed that a personal relationship with the agent (relative, spouse or close friend) will reduce the need for targeted monitoring of the agent (Chrisman, Chua, & Kellermanns, 2007). If opportunistic behaviour becomes known, additional social costs arise for the agent in a personal relationship (e.g. conflicts in the family or among friends, social ostracism; Kallmuenzer, 2015). Therefore, the need for additional active control should be less than if there is no personal relationship. It should be noted that people from an athlete's immediate environment who act as agents have an increased risk of inadequate qualifications (Heidtke, 2013), which in turn increases the need for monitoring. If the agent serves other players in the player's club, the direct contact between the players makes it easier to compare and evaluate the services provided by the agent (Ang, Cole, & Lin, 2000). In the case that (all) players are satisfied with the agent's services, there is little incentive for the individual to invest in further monitoring. However, if individual players provide indications of an agent's opportunistic behaviour, then the probability increases that other players will also have reviewed the agent's performance (Göbel, 2002). Another factor that could affect monitoring is the number of players' agents  a player has had in their career so far. If a player has already worked with several agents, it can be assumed that there is a high probability of having had negative experiences with at least one previous agent (Gohritz, Ehnold, & Hovemann, 2018;Parensen, 2013). This leads to a high probability of the player monitoring the current agent's performance (Lietke, 2009). The intensity of the cooperation may also influence the need for a separate control. Increased cooperation should have greater transparency, which enables the player to evaluate the agent's results on several levels. The more tasks a player assigns to the agent and the more often the agent carries them out, the closer and more transparent the cooperation should become (Petersen, 1989). As the intensity of the cooperation increases, additional monitoring of the agent becomes less likely. A similar influence can also be assumed for the duration of the working relationship.
As the length of the working relationship increases, trust in the agent's actions grows and the likelihood of monitoring subsequently decreases (Nachbar, 2005).

Sample
The underlying data for the empirical study come from an online survey, which was carried out from April 2018 to January 2019. To create the questionnaire, a pre-test (n = 12) was carried out with active and former footballers and with clubs. The questionnaire was sent to players in the Bundesliga, the 2. Bundesliga, the 3. Liga and the Regionalliga via the Vereinigung der Vertragsfußballspieler (VDV; German Union of Contract Footballers), the transfermarkt.de platform, and private channels. Only those players who currently had an agent, or at least had an agent previously in their career, were considered in the evaluation. In total, 336 questionnaires (German: n = 333; English: n = 3) were included in the analysis, which corresponds to approx. 9.6% of the addressed population.
Of the 336 questionnaires included in the analysis, 152 were from players who currently had a player's agent and already had (at least) one other agent, 150 from footballers who were currently working with an agent for the first time, as well as 34 players who did not currently employ an agent but who had worked with one in the past. In terms of player numbers, the Bundesliga was adequately represented (Δ = -0.9%), the 2. Bundesliga (Δ = 16%) and the 3. Liga (Δ = 6.6%) were over-represented, and the Regionalliga was under-represented (Δ = -21.7%; for the distribution of the individual leagues, see . Table 1). The average age of the survey participants was 25.5 years 3 and thus corresponds to the average age (25.6) of the addressed population.

Incidents
Possible negative incidents were determined using a question battery with a total of 13 items on a three-level scale (happened: never, once, several times). The items were determined according to the principal-agenttheory (Göbel, 2002), Table 2 Incidents (0 = never, 1 = once, 2 = several times) for the current and the previous agent, differentiated according to the performance level of the player Incident Bundesliga (in %) 2. Bundesliga (in %)

Regionalliga (in %)
Overall ( (Kelly & Chatziefstathiou, 2018;Heidtke, 2013, Mason & Slack, 2001b and media contributions (e.g. Buschmann & Wulzinger, 2019). As an additional incident, players were asked if they had signed an exclusive contract with the agent concerning the player's placement. The incidents with the agent were queried separately for the current and-if available-for the players' previous agent (see . Tables 2 and 3). The players were also asked if they had ever reported incidents with one of their agents to, for example, the FIFA, Deutscher Fußball-Bund (DFB), Deutsche Fußball Liga (DFL), or to their club (yes/no). If players answered yes, a follow-up open question asked to whom they had reported the misconduct.

Monitoring
Players were first asked whether they had their agents' decisions and recommendations reviewed by another person, as a way to monitor their agents' actions (no = 0/yes = 1). In this way, the players were easily able to understand the question and had no problems understanding the meaning of 'monitoring' in this context. In addition, players who used monitoring measures were asked who reviews the decisions and recommendations of the agent. The players had different answer options (family/ spouse, friends/colleagues, other player's agents, jurists, club/association officials), as well as an open response option, which was not used by the players (multiple responses were possible). In addition, the players should indicate in relation to which overarching fields of decisions and recommendations they had their agents monitored (Mason & Slack, 2001b). The superordinate subject areas 'Contract-related activities' (extension/new contracts with clubs and/or sponsors) and 'Recommendations from the agent' (for [new] clubs and/or sponsors; career; legal advice) were available for selection, as well as an open response option, which was, however, not used by the players.

Independent variables
The description of all independent variables is presented in .

Regionalliga (in %)
Overall ( 1 A contract with a club was not concluded due to the behaviour of the agent 2 A contract with a sponsor was not concluded due to the behaviour of the agent 3 Agent recommended loss-making investments 4 Player received a penalty from the association or club for the behaviour of the agent 5 Agent has received a percentage of a transfer payment for the player 6 Agent wanted money from the player without giving anything in return 7 Agent withheld requests from clubs or sponsors from player 8 Agent withheld payments from player 9 Agent knowingly gave player incorrect information 10 Agent urged player to change club 11 Agent urged player to sign a contract that he was not fully convinced of 12 Agent advised player to put pressure on the club (e.g. not to go to training) 13 Agent recommended player a contract, although the likelihood of playing frequently was very low 'selection criteria' included in the analysis (recommendation, reputation, formal qualifications) were collected using a fivepoint Likert scale (1 = not at all important, 5 = very important; Göbel, 2002). The variables of the 'framework conditions of the working relationship' were determined differently. The dichotomous assignment (yes/no) as to whether the player and agent were in a personal relationship was made using the question, 'Who is your players' agent?' The extent to which the agent serves other players in the club of the footballer was recorded using a five-point Likert scale (0 = does not apply at all, 5 = applies fully). The number of agents that a player had so far in their career was asked directly (Jungels et al., 2017). To map the scope of the agent's tasks (intensity of cooperation), a weighted total score was formed, which considers the frequency of the various tasks. This ranged from 0 = minimum (agent does nothing for a player) to 30= maximum (always does everything for the player). The duration of the working relationship (in years) was asked directly. In addition, the age and performance level of the player (operationalised by affiliation to the league) were included in the analysis as control variables.

Data analysis
In addition to the descriptions, the first step was to use the Kruskal-Wallis test (with Dunn-Bonferroni post hoc test) and Pearson chi-square test to determine the extent to which incidents against players' interests and their frequency varied between leagues. Subsequently, the Pearson chi-square test was used to analyse how the monitoring behaviour differed depending on the league (performance level) of the players. All descriptions and analyses were carried out for both the current and previous agents. Finally, the bivariate analysis was expanded by adding a logistic regression model (blockwise), which was estimated for the current players' agent 4 toconsiderrelevantfactors that influence the player's control behaviour (monitoring). These estimates were used to investigate the extent to which different variables influence monitoring in terms of direction and strength. The blockwise approach also allows the stability of possible effects of the various factors included in the model to be tested. Nevertheless, the respective effects within and between the models should not be interpreted or compared using the coefficients or odds ratios (OR; Allison, 1999). 5 Instead, the average marginal effects (AME) should be calculated, which can be expressed as 'the average influence of the independent variable on the probability of occurrence' P (y = 1 | x; Best & Wolf, 2012, p. 387). Multicollinearity was tested for all regression models. The variance inflation factor (VIF) did not have values higher than 2.1, which means that there was no multicollinearity between the individual explanatory variables. The number of cases per predictor was regarded as acceptable (Vittinghoff & McCulloch, 2007). There were no outliers in the dataset (all standardised residuals were -2 ≤ SResid ≤ 2; Pardoe, 2012, p. 166). All continuous predictors were found to follow a linear relationship to the logit of the dependent variable (Box & Tidwell, 1962). To improve comparability between the models, cases with missing values were excluded from the analysis so that all blockwise models were estimated using 233 cases. Nagelkerke's pseudo-R 2 and the Hosmer-Lemeshow adaptation test were reported for all models.

Descriptions and bivariate analyses Incidents
Overall, 40.3% of the players said they had experienced at least one incident 6 with their current agent, compared with 44% of the players with their previous agent. Looking at the incidents in detail (. Table 2), the players reported more incidents with their previous agent than with their current agent. One exception was the problem that agents participated in the 'transfer fees' (incident 5). This behaviour, which is forbidden according to FIFA statutes (Art. 7), occurred more frequently with the current agent than with the previous one (24.1-17.2%). Players in the Bundesliga (37.5%) and the 2. Bundesliga (30.8%) reported this incident for their current agents, while this behaviour was reported less in the 3. Liga (20.4%) and the Regionalliga (14.6%).
When looking at the incidents between players and their previous agents across all leagues, it was particularly common for a contract with a club not to be finalised due to the agent's behaviour (incident 1) and for players to feel that their agent was pushing them to change clubs more often (incident 10; 18.5% each). Players in the 2. Bundesliga in particular reported this type of incident (1 = 29.3% and 11 = 31.7%), while in the Bundesliga these incidents were the least common (11.8% each). According to the data, the next most common incident was players feeling pressured by their agent to 'sign a contract' (incident 11) they were not sure of (17.5%), closely followed by receiving the recommendation to sign a contract with a club where the player was getting little playing time (incident 13; 17.3%). These incidents also occurred most frequently in the 2. Bundesliga (11 = 30% and 13 = 22.5%) and least often in the Bundesliga (11 = 5.9% and 13 = 11.8%).
In terms of specific differences between leagues, there was only one significant difference in the incident 'transfer fees' of the current agent. The post hoc test (Dunn-Bonferroni) showed that the Regionalliga differed significantly from the Bundesliga with a mean effect size according to Cohen (1992) of (r = 0.229; z = 2.669, p = 0.046).
Despite the legal prohibition of exclusive contractual relationships across all leagues, 35.4% of all players stated that such a contractual relationship exists or existed for the current agent and 34.4% for their previous agent. It is also interesting that 19.2% and 15.6% of the players reported not even knowing whether such a clause exists/existed for their current or for their previous agents, respectively. This problem was observed in all leagues (see . Table 3). With regard to players reporting incidents, for example to FIFA, DFB, DFL, or their club, only one player said that he reported at least one incident to his club. All other players said they had not yet reported the incidents.

Monitoring
Regarding monitoring behaviour (yes/ no), across all leagues it appears that the current agent (52.6%) was monitored more strongly than the previous one (47.5%), with an increase for the 2. Bundesliga and the Regionalliga, while the values for the Bundesliga and 3. Liga remained constant. Overall, monitoring of the current agent was strongest in the Regionalliga (62.2%) and the Bundesliga (50%). However, the differences were not significant (. Table 4).
Looking more closely at monitoring behaviour, it becomes clear that across all leagues both the current and the previous agents were monitored by third parties, particularly the 'suggestions for new clubs' (current agent = 77.6%, previous agent = 73.6%), the 'contract extension' issue (current agent = 75.7%, previous agent = 70.8%) and the 'contract clubs' task (current agent = 72.4%, previous agent = 86.1%). Significant differences between the leagues can only be seen in relation to the task 'suggestion Table 5 Variables predicting monitoring (no = 0/yes = 1) for the current agent (logistic regression, blockwise; odds ratios [OR]) and average marginal effects (AME) and the standard error in brackets are reported Model 1 Model 2 Model 3 OR AME OR AME OR AME Control variables (Intercept) 5.784 ( sponsors' for the current agent. These suggestions were monitored most in the Bundesliga (55%) followed by the 3. Liga (42.9%). In all other leagues, the control behaviour was under 20%.
In general, players used a family member/spouse (current agent = 90.8%, previous agent = 90.3%) or friends/ colleagues (current agent = 35.5%, previous agent = 33.3%) to monitor the agents' recommendations and decisions. Other player's agents, jurists, or club/association employees were only rarely consulted to monitor the agent (1.4-9.9%). There were differences between the leagues, but there was only a significant difference between leagues regarding using friends/colleagues to monitor the current agent. In total, 57.1% of players from the 3. Liga said that they asked their friends/colleagues to monitor their agent, while only 19% of players in the first division did the same.

Model 1
It became apparent that the only significant influence was playing in the 2. Bundesliga. Thus, it was about 19% (AME = -0.194) less probable for a player in the 2. Bundesliga to monitor their agent in comparison with a player in the Regionalliga (model 1). All other control variables show no significant influence on monitoring behaviour.

Model 2
When including the factors related to the 'selection criteria' , it became apparent that the 'recommendation of the club' had a significant influence on monitoring behaviour (model 2). According to this, the probability of monitoring was reduced by about 10% (AME = -0.096) if the club recommended the agent. The formal 'qualification' of the agent also had a significant influence. An agent was about 7% more likely (AME = 0.068) to be monitored, the more important the agent's formal qualification was to the player as a 'selection criterion' . Looking at the control variables, playing in the 2. Bundesliga still had a significant influence (AME = -0.164). All other factors had no significant influence.

Model 3 (overall model)
In the overall model, the factors 'recommendation of the club' (AME = -0.109) and the formal 'qualification' (AME = 0.084) remained significant. In contrast to model 2, the influence of the factor 'playing in the 2. Bundesliga' decreased and was no longer significant. Regarding the newly included factors for the 'framework conditions of the working relationship' , the factor 'number of players' agents' had a significant influence. As the number of agents that a player had so far in their career increased, the probability of monitoring the current agent also increased by approx. 8% (AME = 0.083) for every additional agent before the current agent. All other factors had no significant influence on the players' monitoring behaviour.

Discussion
The results of the present study indicate that players report that agents across all leagues take actions potentially harmful to the players. Looking more closely at individual incidents with the current agents, the corresponding actions initially do not directly indicate a systematic form of moral hazard, in contrast to the data of the previous agent, which indicate a systematic problem very well, as Kelly and Chatziefstathiou (2018) suggest in their study. It is interesting that the agents behave increasingly opportunistically with regard to task areas (e.g. exclusive contract, change of club, contract signing, participation in transfer fees) that generate direct monetary added value for the agent, without immediately recognising a personal disadvantage for the player. This speaks in favour of a rational weighing process (Gerasimou, 2018), even if this can be associated with disadvantages for their clients (Eisenhardt, 1989;Heath, 2009). In this context, it is not surprising that the overall intensity and frequency of incidents were higher with the previous agent compared with the current one. After all, incidents due to information asymmetries can increasingly only be perceived in retrospect or, if perceived, lead to the termination of the cooperation, which would also have been expected in advance from a theoretical perspective (Petersen, 1989). The results for the previous agent therefore already indicate a systematic problem from the players' point of view and should be regarded as decisive for the whole players' agent market due to the nature of the problems, which only become apparent over time.
Although players reported having noticed several incidents with their agent(s), it is interesting that the players did not report those incidents. This could be because the players did not know where to report the incidents, but it would have been expected that more than one player would at least have tried. The regulatory bodies are FIFA and the DFB. Therefore, a corresponding body for (anonymous) reporting of incidents should be set up here. This would help the players and reduce the information asymmetries.
In terms of monitoring, around 50-60% of the players had monitored their current agent and around 40-53% had monitored their previous agent. The values show that almost every second player simply does not trust the work of their agent, but actively monitors it with the help of a third person. Against the background of the shown incidents, this behaviour appears rational, but from an economic point of view it represents a loss of welfare, since in an optimal equilibrium, monitoring should theoretically be superfluous (Balmaceda, Balseiro, Correa, & Stier-Moses, 2010).
In general, players also used family members/spouses or friends/colleagues to monitor the agents' recommendations and decisions for their previous agent(s). By contrast, experts such as other player's agents, jurists and club/association employees were rarely used for monitoring. Monitoring by family members/ spouses or friends/colleagues is associated with lower costs than if it is carried out by professionals. On the other hand, it is questionable whether family members/spouses or friends/colleagues have the knowledge to value the agent's out-put. The players seem to prefer informal and more cost-effective monitoring processes, while a significantly higher proportion of formal experts was expected from a quality point of view (Romano, 1999).
Among the factors influencing the monitoring behaviour, the factors 'recommendation of the club' , the formal 'qualification' of the agent and the 'number of players' agents' prove to be significant. It is interesting that the players put trust in the 'recommendation of the club' . For the players, expertise seems to be an important quality feature (Schölermann, 2003), although clubs and players could have different interests during negotiations. The 'qualification' of the agent as a search criterion of a player significantly influences the monitoring behaviour. A player who pays particular attention to the formal 'qualifications' of the agent in the run-up to a collaboration is more likely to allow the agent to be monitored while the service is being provided. Since the formal 'qualification' should theoretically function as a quality feature, it could have been expected that the monitoring probability would decrease (Bathi, 2015). Apparently, the 'qualification' does not function as an instrument that makes subsequent monitoring less necessary. Rather, like monitoring itself, it is an expression of an overall critical basic attitude that does not want to 'blindly' trust the agent. As the 'number of agents' increases, so does the likelihood of third-party control. This seems particularly plausible against the background that a change of agent usually takes place because of dissatisfaction. Therefore, based on their experience, the players should be more cautious in dealing with players' agents and therefore let the current agent be monitored (Lietke, 2009). It is also conceivable that, based on their experience, the players themselves increasingly monitor the agent and do not delegate control to third parties (Nachbar, 2005). The fact that the good 'reputation' of the agent has no influence is surprising insofar as reputation is viewed in the principal-agent theory as a central solution mechanism regarding the problem of choosing the 'right' agent (Eisenhardt, 1989;Heath, 2009). Whether the players do not trust that reputation or whether the reputation does not have the necessary selectivity must remain open at this point. It has also been shown that, compared with the other leagues, the highest probability of agents being monitored is in the Regionalliga. However, the differences to the higher leagues are comparatively small and insignificant (model 3).

Conclusion
1. The results clearly show that players report that agents show opportunistic behaviour and also take actions that are not in the interest of the players. The problems become particularly apparent when looking at the previous agent. More in-depth regulatory interventions, as required in some cases (MacInnes, 2020), could represent the first approaches to a solution variant. 2. Around half of the players seem to be aware of possible opportunistic behaviour by their agents and do not 'blindly' trust them. Instead, the monitoring is mostly carried out by people without formal professional expertise. To what extent the other players have decided against a control by third parties or are not aware of possible problems cannot be answered from the data. 3. The existence of exclusive contracts and the uncertainty of whether there is an exclusive contract at all indicates that at least some players lack knowledge about which agent actions are legally permissible and which are against their own interests. If this problem occurs not 'only' with exclusive contracts, but also in other areas, then this could also lead to an underestimation of the problem of opportunistic behaviour, since the players do not classify such actions as such. This clearly shows a lack of transparency that needs to be solved. If the associations and/or the players' unions want to prevent possible damage to the players by their agents, it seems advisable to intensify and support information and awareness-raising campaigns. This approach would enable the players to better recognise the opportunistic behaviour of agents and to make decisions about whether and to what extent an agent's actions require monitoring. Educating the players on how to work with agents and to make them aware of possible problems with agents appears to be a target-oriented measure. More transparency would also be achieved through regulatory measures, although the legal hurdles would be higher and agents would again search for ways to bypass the regulations. Regardless of these problems, FIFA is already considering implementing suitable measures to be adopted in mid-2021 (MacInnes, 2020). This should include capping agent commissions and setting up so-called clearing offices to increase transparency (Lades, 2020).

Limitations and outlook
It is important to note here that: 1. The data in this study refer to professional leagues in Germany. Since only three questionnaires were completed in English, the focus of this study seems to be for the German-speaking players. A stronger international orientation appears to be advisable for further investigations. In this way, for example, comparisons could be made between different leagues in different countries and regarding different nationalities of the players. 2. In the end, the opportunistic behaviour of the agent is only made possible by an asymmetrical distribution of information. In addition, incidences can only be reported if they are also recognised by the players as such. Therefore, the actual extent of potentially damaging agent actions also cannot be revealed in the present study. Rather, the knowledge gained relates 'only' to the incidents that are recognised and reported by the players. Furthermore, the results show that (some) players' agents bypassed FIFA statutes and it is unclear how they did this. One way of dealing with these problems is to ask the agents directly. However, strongly distorted answers are to be expected in surveys on socially deviant and/or prohibited behaviour. The 'randomized response technique' has proven effective in countering this problem (Pitsch, Emrich, & Klein, 2007). 3. Central variables of the present study, such as incidents or monitoring factors, are available in a relatively highly aggregated form. For example, it therefore remains unclear how the agents' specific opportunistic behaviour manifests and what (negative) consequences this entails for the player. In addition, the data allow no conclusions to be drawn regarding how agents are monitored. This limitation could be countered in further studies by examining the working relationship between player and agent in more detail using individual case analyses. A more qualitative research approach can help to shed more light on the topic and to make individual logic and constructions of both the players and the agents visible. 4. In addition, individual incidents and monitoring processes can be analysed more precisely using a case study design. In this way, both the costs of agents' opportunistic behaviour and the costs of monitoring processes can be assessed in monetary terms for the players. As a result, these costs can be compared to determine the welfare loss and to optimise the form and intensity of monitoring processes.

Declarations
Conflict of interest. A. Gohritz, G. Hovemann and P. Ehnold declare that they have no competing interests.
For this article no studies with human participants or animals were performed by any of the authors. All studies mentioned were in accordance with the ethical standards indicated in each case.
Open Access. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.