Combining Egalitarian and Proportional Sharing Rules in Team Tournaments to Incentivize Energy-Efficient Behavior in a Principal-Agent Context

Abstract

Triggering the energy-efficient behavior of agents in firms simultaneously decreases costs and mitigates CO₂ emissions. If firms use team tournaments to increase energy-efficient behavior and thus employee performance, they may face unintended consequences, like a bifurcation effect: Individuals drop out if they believe that they cannot win the contest. By contrast, high-performing employees may overexert themselves. Additionally, some individuals might be tempted to free-ride. In a field experiment with truck drivers, we analyze whether proportional sharing of the bonus within teams based on individual effort instead of egalitarian sharing reduces both bifurcation and free-riding during team tournaments. Our results reveal that (1) the team contest improves performance; (2) this increase in performance is overall slightly stronger under the proportional than under the egalitarian sharing rule, using ceteris paribus comparisons; and (3) the performance increase is mainly driven by the team member performing worse.

Keywords

environmental performance business strategy and the environment environmental and natural resource economics greening business functions field experiments environmental attitudes and behavior

Introduction

Energy efficiency in firms creates a win–win situation wherein energy costs decrease for the firm, and CO₂ emissions are mitigated. Thus, firms need to know how to incentivize energy-efficient behavior of their employees. This applies in particular to energy-intensive industries such as the transportation sector. In 2016, the transportation sector was responsible for approximately 25% to 28% of greenhouse gas emissions in Western industrialized countries (cf. European Environment Agency, 2018; U.S. Environmental Protection Agency, 2018). Moreover, CO₂ emissions of mobility are estimated to increase by 60% until 2050, even in a scenario with significant technological progress (Organisation for Economic Co-operation and Development, 2017). At the same time, fuel comprises approximately 40% of the variable transportation costs. Naturally, numerous attempts are being made by the main players of this sector to reduce fuel consumption, ranging from optimizing routes to technical innovations in trucks. However, the extent of the potential of truck drivers’ good driving behavior is often underestimated. Driving in an energy-efficient manner can reduce fuel consumption by up to 25% (Schall et al., 2016). Thus, attempts to increase driver performance are worthwhile to reduce fuel costs for firms and mitigate CO₂ emissions simultaneously.

When attempting to increase energy-efficient behavior, firms may apply tournaments. In organizations, team tournaments are frequently used to increase effort. Teamwork is increasingly viewed as a precondition for achieving high performance in organizations (Cappelli & Tavis, 2018). Pure individual tournaments may also not fit to a company culture that encourages cooperation (Blader et al., 2020). Contrary to individual tournaments, team tournaments may foster potential learning opportunities in teams (Merchant & Van der Stede, 2007) and fairness-based considerations about effort choices: Individuals may put more effort into their work as they feel obliged toward their team mates (Livio & De Chiara, 2019). However, also team tournaments have some side effects and can be distinguished with respect to whether there is also an individual pay component or not. They may be designed as (1) egalitarian sharing of tournament prizes among the team members (egalitarian tournaments) or (2) as proportional sharing of prizes among the team members according to their individual contribution. Proportional prize-sharing tournaments (proportional tournaments) have yet to be sufficiently researched (Majerczyk et al., 2019), although they are intensively used in practice (Hwang et al., 2009; Majerczyk et al., 2019; Sisk, 2005) and offer the possibility to eliminate at least three of the following severe adverse side effects of egalitarian team tournaments:

Egalitarian team tournaments, particularly, with possible communication between teams, offer the potential for collusion that can result in harmful drawbacks for the efficiency of these incentive systems (Cason et al., 2017; Harbring, 2006).

Egalitarian team tournaments may cause a bifurcation effect on effort (Müller & Schotter, 2010). Although some team members overexert themselves, some also may drop out and refuse to exert any extra effort at all.

Egalitarian team tournaments may result in free-riding. Contrary to dropping out because of demotivation, free-riding in pure between-team tournaments is a strategic choice: Individuals reduce their effort if they expect their teammates to execute high levels of effort (Gunnthorsdottir & Rapoport, 2006).

Consequently, many empirical studies on team tournaments report high levels of variance in efforts (e.g., Nalbantian & Schotter, 1997; Orrison et al., 2004). Majerczyk et al. (2019) showed both theoretically and experimentally that proportional sharing rules in team tournaments are effective in increasing sustainable effort by reducing free-riding, increasing cooperation, and decreasing collusion. However, this only empirical evidence so far stems from a laboratory experiment with student participants and abstract decisions on effort levels without conducting real-effort tasks. Also, whereas performance increases in a field setting require changes in working habits, the laboratory investigation was not able to draw a conclusion regarding the impact of between- and within-team tournaments on actual changes in performance levels. The present study contributes to and complements the preliminary evidence on proportional sharing rules in team tournaments by analyzing a tournament for energy efficiency in a field experiment in a real and ongoing firm.

Our results reveal that tournaments are suited to improve energy efficiency in the firm. We compare egalitarian and proportional sharing rules, and under both tournament types, individuals overall improved in their performance. The improvement in energy efficiency was ceteris paribus significantly better under the proportional than under the egalitarian sharing rule. The worse-performing team member mainly drove this effect, while already well-performing individuals do not improve much under a team tournament. The magnitude of improvement is larger in proportional sharing rule tournaments. Additionally, we show that free-riding is a problem under egalitarian sharing. However, better drivers, who simultaneously increase their performance slightly but significantly under egalitarian sharing, compensate this negative effect.

Egalitarian and Proportional Sharing Rules in Team Tournaments

Related Literature

Since the seminal theoretical work of Lazear and Rosen (1981) on rank-order tournaments, considerable theoretical and empirical research has been conducted to understand the behavior of agents competing for a prize. Bull et al. (1987) found in the laboratory that effort levels in tournaments are high relative to piece-rate incentives. However, they also found a comparable high variance of effort in tournaments. Their results were sufficiently replicated in the laboratory by Nalbantian and Schotter (1997), Orrison et al. (2004), Harbring and Irlenbusch (2003, 2005, 2008), and Agranov and Tergiman (2013).

Müller and Schotter (2010) found in the laboratory a bifurcation effect in individual tournaments over time: In their experiment, low-ability workers were dropping out by exerting little or no effort and high-ability workers overexerting themselves. This led to a bifurcation: Low-ability workers exerted little or no effort, and high-ability workers tried too hard. Similar evidence was found for team tournaments: Irlenbusch and Ruchala (2008) observed in the laboratory a significant decline in contributions under egalitarian sharing rules, which could only be suspended by introducing a high bonus for individual performance but also came along with the downside of bifurcation.

Next to the payment scheme, the relative performance differences of team members may also affect bifurcation. Sheremeta (2011) found that heterogeneity in team members’ ability leads to higher effort provision for all team members than theoretically predicted. On the other hand, team tournaments may stimulate strategic effort reduction. Individuals may intentionally reduce their contribution and free-ride. Comparing individual and team tournaments, Van Dijk et al. (2001) found this effect, but other simultaneously overexerting individuals compensated free-riding in groups. Additionally, tournaments induced higher but more variable effort, which was also found by Nalbantian and Schotter (1997) and Orrison et al. (2004).

While these adverse side effects of egalitarian tournaments were frequently reported, only a few systematically analyze how these adverse side effects affect overall performance and how they can be reduced. Gunnthorsdottir and Rapoport (2006) and Kugler et al. (2010) studied egalitarian versus proportional sharing in a public good context and found efforts to be higher under the proportional sharing rule because free-riding was prevented. The intention to free-ride is particularly high if it is difficult to observe and verify the contribution of individuals in the group. Nalbantian and Schotter (1997) analyzed a situation in the lab in which two groups with six members competed for a single prize to be shared equally between the group members. They showed that individuals in teams exerted more effort, and the free-riding problem was reduced when monitoring of team members was possible (Nalbantian & Schotter, 1997). Next to monitoring, communication among team members can reduce free-riding. Sutter and Strassmair (2009) allowed for communication in an experimental tournament with teams and showed that communication within a team increased efforts. In contrast, communication between teams led to collusion and, thus, a reduction of effort.

Majerczyk et al. (2019) analyzed how communication interplays with egalitarian and proportional sharing rules theoretically and in the lab. Their theoretical model predicted a free-riding problem in the case of an egalitarian sharing rule. However, the proportional sharing rule is effective in reducing free-riding, but only if individuals cannot collude. Their experimental results showed that proportional sharing rules were effective in increasing effort, in sustaining effort over time, and in reducing the free-riding problem by increasing cooperation and decreasing collusion within teams.

Rosenbaum et al. (1980) conducted an extralaboratory experiment with students in a natural environment (see, Charness et al., 2013, for this typology of experiments) and found that teams performed relatively worse under proportional rewards than under egalitarian prize-sharing team tournaments. The task was a real-effort task; however, the individuals’ tasks were highly independent. The first and, to the best of our knowledge, only experiment to test proportional rewards in the field was conducted by Wageman (1995). She found adverse effects on the effort level if individual rewards were introduced on top of egalitarian prize-sharing tournaments.

Rosenbaum et al. (1980), Wageman (1995), Irlenbusch and Ruchala (2008), and Majerczyk et al. (2019) have research questions, and experimental designs derived thereof that are similar to ours. However, these authors conducted (extra)laboratory experiments in which agents made decisions in abstract situations. Wageman (1995) and Irlenbusch and Ruchala (2008) introduce individual incentives on top of an egalitarian sharing rule. We combine both streams but directly compare egalitarian and proportional incentive setting mechanism. Our contribution to the literature is to analyze the different effects of egalitarian and proportional prize-sharing tournaments in the field. We conduct our experiment in a real and ongoing firm, with truck drivers in their daily work routine over a long duration of 11 months. We offer essential insights into the impact of the combination of egalitarian and proportional prize-sharing team tournaments. We first observe a control phase in which no incentives were given to anyone. In the treatment phase, half of the truck drivers were allocated to a team tournament with egalitarian sharing rules within their team. The other half of the drivers were allocated to a team tournament with proportional sharing rules in their team. In the posttreatment phase, teams are dissolved, and no incentives paid—drivers work under the same circumstances as in the control period. We observe both the development of every treatment group’s performance over the three experimental phases (control, treatment, and posttreatment) and compare the performance of both treatment groups in each of the three phases with each other. Thus, we can have clear evidence which sharing rule led to higher efforts in our sample.

Our study also allows us to observe individual effort levels, and we can thus analyze whether certain types of drivers react differently to the egalitarian and proportional sharing rules. Additionally, we have a 3-month observational period posttreatment, which allows us to identify any sustainable performance effects triggered by either egalitarian or proportional prize-sharing rules in teams.

Theoretical Considerations

Our theoretical model is informed by the models of Moldovanu and Sela (2001) and Flamand and Troumpounis (2015). Consider all agents in the firm to be risk-neutral and divided into one of the treatment groups, $t = [Q, P]$ (egalitarian sharing rule and proportional sharing rule) that take part in a team tournament to compete for a prize with value V. Each treatment group consists of j teams, with $j = [1, 2, .., N]$ in treatment group Q and $j = [1, 2, .., N]$ in treatment $P .$ Each team j has i members, with $i = [1, 2]$ . Both treatments, as well as membership in a specific team, were exogenously given to agents who were uniquely identified by $[t, i, j]$ . All agents in each team simultaneously chose an effort level $e_{i j}^{t}$ from the closed interval $[1, 100]$ . The teams’ effort is defined by $E_{j}^{t} = \sum_{i} e_{i j}^{t}$ ; thus, individual efforts are perfect substitutes. In the field, we observe the agents’ performance, which is a function of effort, ability, and luck (Ericsson & Charness, 1994). Let team j’s probability of success in this team contest be given as $π_{j} (E_{j}) = \frac{E_{j}}{\sum_{j} E_{j}}$ .

Exerting effort is costly. Assuming linear cost functions, agent $[t = Q, i, j]$ maximizes as follows:

\frac{E_{j}}{\sum_{j} E_{j}} \frac{V}{2} - e_{i j}^{Q}

(1)

Agents $[t = P, i, j]$ also maximize as follows:

\frac{E_{j}}{\sum_{j} E_{j}} V \frac{e_{i j}^{P}}{e_{i j}^{P} + e_{- i j}^{P}} - e_{i j}^{P}

(2)

with the first-order condition for the agent $[t = Q, i, j]$ being

\frac{V}{2} (\frac{\sum_{j} E_{j} - E_{j}}{{(\sum_{j} E_{j})}^{2}}) = 1

(3)

and the first-order condition for the agent $[t = P, i, j]$ being

V (\frac{\sum_{j} E_{j} - e_{i j}^{P}}{{(\sum_{j} E_{j})}^{2}}) = 1

(4)

In a symmetric Nash equilibrium, identical members of team j receive the same payoff. Assuming the proportional sharing rule, Equation (4) reduces to $V (\sum_{j} E_{j} - e_{i j}^{P}) = {(\sum_{j} E_{j})}^{2}$ , and the equilibrium effort is the result of a standard Tullock contest, which reads as follows:

e_{i j}^{P *} = \frac{(N - 1) V}{N^{2}}

(5)

where N is the total number of individuals in the contest. Under the egalitarian sharing rule, the equilibrium is the following:

E_{j}^{*} = \frac{n - 1 [N - (n - 1) 2] V}{N^{2}}

(6)

where n is the number of teams in the contest. Therefore, under the egalitarian sharing rule, free-riding becomes more attractive. From this result, we derive our first hypotheses:

Hypothesis 1: Free-riding is higher under the egalitarian than under the proportional sharing rule.

At the end of each team tournament phase, every member of a team j is informed about the individual effort of both team members, the team’s effort $E_{j},$ and the ranking in comparison with all other teams. Thus, team members will form beliefs about $π_{j} (E_{j}) = \frac{E_{j}}{\sum_{j} E_{j}}$ , the probability to win the team tournament. Thus, informing individuals about their team’s effort level and ranking may result in bifurcation. However, we assume overexertion to be stronger under the proportional sharing rule, as added to the motivation to win the team tournament, individuals are motivated by their relative high individual payoff. By contrast, dropping out and refusing to make any effort at all is more likely under the egalitarian sharing rule, as individuals are not additionally motivated on an individual level. Consequential, we derive our second hypotheses:

Hypothesis 2: Overexertion is more prevalent under the proportional than under the egalitarian sharing rule.

Hence, we assume that overall proportional sharing rules will improve performance better than egalitarian sharing rules as free-riding is less prevalent and more individuals overexert themselves.

Experimental Design

Setting

In our experiment, we observe the performance of 104 workers on 100 trucks over 11 months. These 11 months consist of a control phase (3 months), a treatment phase (5 months), and a posttreatment phase (3 months). After the control phase, workers were combined into two-person teams, and tournaments were run. In half of the teams, we introduced an egalitarian sharing rule; in the other half, we introduced a proportional sharing rule of winning prizes. Under egalitarian sharing, every team member received an equal share of the winning prize. Under proportional sharing, every team member received the share of the winning prize that was proportional to his contribution. After the 5 months of the treatment phase, teams were dissolved, and tournament incentives were removed. The individual effort levels were observed during the 3 months’ posttreatment phase.

The truck drivers in our experiment work for a medium-sized German company. All trucks were produced from one manufacturer and were maximal 3 years old. Therefore, we assumed that trucks were as homogeneous as possible in terms of technology. Moreover, the 104 drivers were assigned to different trucks throughout the experiment. Approximately 2 years ahead of our field experiment, the firm installed in its trucks a telematics system that was developed by the truck manufacturer. All drivers at the firm were familiar with the system before the start of the experiment.

Among other features, the telematics system measures the energy efficiency of the driving style. The telematics system uses data from the truck to compute a so-called “fuel efficiency score.” The fuel efficiency score ranges from 0 to 100 points, with 100 points representing the best driving performance. The score is a weighted average of the four categories: acceleration, gear usage, brake usage, and consumption at idle. Fuel consumption is an absolute measure and depends on the driving style as well as factors exogenous to the driver (weight, route topography, etc.), whereas the fuel efficiency score is endogenous to the driver’s effort and is solely influenced by driving behavior. Therefore, this variable is the most important one for our analysis. Also, we received access to the driving behavior component of the telematics system through a back-end solution. From this, we extracted data on the driving performance and technical characteristics of each truck and route on a daily, weekly, and monthly basis.

The fleet manager did not monitor driving behavior in advance of our investigation. The telematics system was predominantly used because of its scheduling deliveries and routing features. According to the firm, fuel makes up approximately 40% of its total variable costs. Therefore, fuel consumption is one of the main productivity measures, and fuel-efficient driving, next to punctuality and avoiding accidents, is the main component of drivers’ performance. Given the increased pressure on costs in the sector, the firm sought to reduce the variable costs of transportation and to increase fuel efficiency.

Phases and Treatments

The experiment was split in a control phase (May 1, 2018, to July 31, 2018), a treatment phase (August 1, 2018, to December 31, 2018) and a posttreatment phase (January 1, 2019, to March 31, 2019). For an overview, see Figure 1. On July 19, 2018, we announced the team competition to both treatment groups. We organized our communication with drivers through postal mail. Because drivers were either German or Polish, we sent personalized letters in both languages.¹ We always communicated simultaneously with all drivers and used the same content for all drivers.

The telematics system recorded the performance of all drivers during the whole experiment, thus, also in the pre- and posttreatment phases. This information was potentially visible to the fleet manager, the telematics supplier, and us. Neither the fleet manager nor the telematics supplier analyzed the data.

Figure 1.

Overview of experimental timing.

We used a stratified randomization approach to assign drivers to both treatment groups accounting for two influences on the performance we were already aware of.

Some drivers installed an app on their firm’s mobile phones before the control phase and thus could inform themselves about their fuel efficiency scores. Obtaining information on the actual use of this app was not possible without influencing drivers’ attention; however, we knew who had installed the app. Because self-monitoring could impact performance, we used a stratified randomized approach. We controlled for app installation by (1) ensuring that the number of app users was equal in both treatment groups, (2) ensuring that teams comprised either two potential app users or two nonapp users, and (3) controlling for the potential impact of the app in our multivariate analysis. Additionally, we reached an agreement with the provider of this app that the status quo of app installation that was confirmed before the start of the control phase did not change throughout the experiment. Thus, drivers did not uninstall or install the app.

Second, because we assumed that routines strongly influence driving behavior, we used preperformance as an indicator for routines and randomly assigned drivers to treatments to ensure that average control phase performance was equal. However, because we intended to create homogenous groups and teams concerning the use of the app, both experimental groups showed differences in their mean performance during the control phase. Users of the app were randomly chosen; therefore, differences in the average control phase performance were not due to self-selection. During the subsequent analysis, we considered this issue by controlling for control phase performance.

In both treatment groups, we built teams by matching the best- and the worst-performing drivers (according to drivers’ average fuel efficiency scores during the last 4 weeks driven). This procedure resulted in a similar means of the team fuel efficiency scores for both groups. In half of all the two-person teams, we introduced a team tournament with an egalitarian bonus-sharing rule. Among the remaining two-person teams, we introduced a tournament with a proportional sharing rule (see Table 1 and Figure 1).

Table 1.

Treatment Groups.

Treatment	N drivers	N drivers with app	N teams
1. Team proportional (between- and within-team tournament)	50	24	25
2. Team egalitarian (only between-team tournament)	50	24	25

Control Phase: May 1, 2018, to July 31, 2018

During the control phase, truck drivers were working as usual in their trucks. Individual performance was recorded, and we used the data obtained as control measures of performance. We also used the data here for the stratified randomization approach described above. On July 19, 2018, all drivers received one postal mail over the tournaments starting on August 1, 2018, and their teammate was announced. Drivers here received a fixed wage. The average wage of a driver was €1,812 per month (gross salary). Depending on a driver’s situation, this amount equaled a net income of approximately €1,200.

Treatment Phase: August 1, 2018, to December 31, 2018

In each tournament, the best team received a monthly bonus of €50, the second-best team received a monthly bonus of €30, and the third-best team received a monthly bonus of €20. Drivers also received their fixed wage as in the control phase. Thus, the tournament prize per month equaled approximately 5% of net income in the most extreme case.

Posttreatment Phase: January 1, 2019, to March 31, 2019

After the treatment phase, drivers were no longer part of a team contest and received no bonus payments or any other incentive, just their fixed wage as in the control phase.

Data and Estimation Strategy

The experiment started on May 1, 2018, with 104 initial drivers. In total, our data set contained 2,365 person-week observations.² The number of drivers changed weekly because of turnover, sick leave, national holidays, and holiday weeks. All the drivers in our experiment were male, and the average age was 43.94 years³ (SD = 9.94). Approximately 80% of the drivers were Polish; the rest were German. On average, drivers had work experience as truck drivers of 8.13 years (SD = 7.22). The average tenure at this firm was 3.6 years⁴ (SD = 4.55).

Dependent Variable

To measure drivers’ performance, we used the fuel efficiency score provided by the telematics system as our dependent variable. The fuel efficiency score was measured as between 0 and 100 in one-unit steps. Although the firm’s variable of interest was fuel consumption, we intentionally did not use it as the dependent variable because it was influenced by both driving behavior and exogenous factors, such as the truck’s weight and route (e.g., topography). We used stratified randomization that controls for the app and the preperformance of drivers. In consequence, performance during the control phase was not equal for both treatment groups in this phase: The mean fuel efficiency score in the proportional sharing treatment groups was 80.58 (SD = 13.29), and it was 77.78 (SD = 15.10; p = .00, Mann–Whitney U test; p = .004 for the variance comparison test, also robust) in egalitarian sharing treatment groups. Therefore, we control for preperformance in the following analysis.

Independent Variable

We differentiated treatment groups concerning the sharing rules for winner prizes among team members. In compliance with theoretical predictions, we assumed the egalitarian sharing rule to motivate free-riding. The individual payoff of a team member in the proportional sharing treatment groups depends on the sum of both team members’ fuel efficiency scores compared with those of other teams and the share of the individual fuel efficiency score in the team’s performance. However, the individual payoff of a team member in egalitarian sharing treatment groups depends only on the sum of both, the team member’s fuel efficiency score compared with those of other teams. To estimate the impact empirically, we defined a treatment variable that takes the value of 0 if the driver’s performance is observed during the control phase (control), 1 if the driver is in the proportional sharing treatment group (team proportional), and 2 if the driver is in the egalitarian sharing treatment group (team egalitarian).

Control Variables

Time-Invariant Variables

Because half of our drivers could inform themselves about their performance via an app, we controlled for this impact with the dummy variable app, which takes the value of 1 if the driver could use the app and 0 if the driver has no access to the app but is informed monthly about his performance. This variable was fixed over time.

We assumed that very good drivers, relative to rather poor drivers, have fewer means to improve their performance. On one hand, bad drivers may be more likely to consider free-riding, and the better driver of a team may anticipate this. On the other hand, the better driver may also help and motivate the worse driver to increase his performance. To control for within-team heterogeneity, we included the dummy variable better one, which takes the value of 1 if a driver is the better one of both in a team according to his preperformance and 0 if not. This variable did not vary over time. One may also argue that the lead performance in one team changes over time. Although drivers have had varying performances, the relative ordering within all teams remained constant during the experiment. Thus, although the variable is theoretically time-varying, it was fixed over time in practice.

Time-Varying Variables

We assumed that reaching a prize in the contest motivates drivers, at least not to worsen their performance in the subsequent period. Therefore, we introduced the variable prize in the previous period (t − 1), a dummy variable that takes the value of 1 if the driver’s team won a prize in the contest in the last period (month).

Although the fuel efficiency score accounts as much as possible for a route’s difficulty, reaching better fuel efficiency scores on routes with more highway driving and with a lighter truck is still easier. Also, the longer the distance traveled, the higher the probability that the fuel efficiency score measures actual driving behavior. To account for the aforementioned, we controlled for distance traveled during the week in 1,000-km increments using the time-dependent variable weekly distance. The telematics system also provides information on the share of the distance that was covered with light, medium, or heavy weight. We used the time-varying variables weight medium and weight heavy with the reference of weight empty to report the share of the distance driven at each weight category as additional controls.

Results

Descriptive Statistics

Using descriptive statistics without controlling for important variables, we found that the teams and individuals under the egalitarian sharing rule improve more than under the proportional sharing rule. On an aggregated level, team fuel efficiency scores increased significantly from a mean of 162.10 to 167.73 under the proportional sharing rule and from a mean of 155.23 to 163.12 under the egalitarian sharing rule (both p = .00, Mann–Whitney U test). On average, the team fuel efficiency score increased significantly stronger under the egalitarian than under the proportional sharing rule (by 7.46 points vs. 5.31 points, p = .00, Mann–Whitney U test). On an individual level, we found that fuel efficiency scores increased significantly from an individual mean of 81.03 to 83.91 under the proportional sharing rule and increased significantly from an individual mean of 77.80 to 81.69 under the egalitarian sharing rule (both p = .00, Mann–Whitney U test). Also, on the individual level, the fuel efficiency score increased on average significantly stronger under the egalitarian than under the proportional sharing rule (by 3.69 points vs. 2.89 points, p = .00, Mann–Whitney U test).

The descriptive results previously presented seem to indicate that the egalitarian sharing rule outperforms the proportional sharing rule on average. However, due to initial performance levels, especially of the worse driver, this result may be driven by simply more room for improvement for worse drivers in the egalitarian sharing rule treatment. Deriving multivariate results will be essential in order to derive ceteris paribus interpretations. To understand the underlying mechanism of the results observed, we aim to analyze the free-riding and bifurcation behavior of drivers by comparing the performance improvement of the better and the worse drivers of a team separately.

Figure 2 presents the monthly average of the fuel efficiency score for the best and the worse drivers. While there seems to be not much room for improvement for the better driver, worse drivers improve clearly under both treatments.

Figure 2.

Mean fuel efficiency score by month, treatment, and team member.

On an aggregated level, the worse driver of a team improved his fuel efficiency score significantly by 5.20 points under the proportional sharing treatment (p = .000, Mann–Whitney U test compared with performance in the control phase). However, the fuel efficiency score of the better driver stayed, on average, nearly constant (M = +0.57 points; p = .295, Mann–Whitney U test). Under the egalitarian sharing rule, the worse driver of a team improved his fuel efficiency score significantly by an average of 6.72 points (p = .000, Mann–Whitney U test). The better team member’s fuel efficiency score again remained nearly constant (M = +0.87 points; p = .194, Mann–Whitney U test). While the difference in the improvement is not statistically significant between both treatments for the better driver (p = .208, Mann–Whitney U test), it is for the worse driver (p = .000, Mann–Whitney U test).

Difference-in-Difference Between Proportional and Egalitarian Sharing Rules

Figure 2 also reveals that the performance in the control phase was different for different types of drivers. We, therefore, reanalyze the data and take differences in the control phase into account for ceteris paribus comparisons. First, we run a difference-in-differences (DiD) treatment effect estimation of the team and individual fuel efficiency score from a pooled set (see Table 2). On the team level, the negative difference between fuel efficiency scores in the control and treatment phase, as well as the positive DiD estimate, are significant. On the individual level, the negative difference between fuel efficiency scores in the control and treatment phase remains significant. However, the smaller DiD estimate is clearly not significant. This result on the individual level is clearly driven by performance changes of the worse driver, while the performance of the better one remains nearly constant.

Table 2.

Difference-in-Difference (DiD) Estimates.

	Difference (egalitarian − proportional)
Estimate	Control phase	Treatment phase	DiD	p Value
Team fuel efficiency score	−6.874*** (0.887)	−4.607*** (0.825)	2.267 (1.212)	.062
Fuel efficiency score	−2.233*** (1.042)	−2.221*** (0.739)	1.012 (1.277)	.429
Fuel efficiency score: better one	−0.022 (0.679)	0.278 (0.456)	0.300 (0.818)	.714
Fuel efficiency score: worse one	−6.338*** (1.274)	−4.817*** (1.087)	1.521 (1.675)	.364

Note. DiD results, weekly data, robust standard errors in parentheses, control variables: distance, weight, and ranking in the previous period.

***

p < .01. ^**p < .05. ^*p < .1.

The matching rule used to build teams within each treatment group resulted in teams that differed considerably concerning the difference between the better and the worse drivers. There was some variation across teams. However, because worse drivers caught up and better drivers nearly performed at a constant level (see Figure 2), the within-team difference in performance was reduced in both treatment groups (see Figures 3 and 4).

Figure 3.

Within-team difference (only positive values) in fuel efficiency score by treatment group and phase.

Figure 4.

Standard deviation of fuel efficiency score by phase and treatment group at an individual level (including Levene’s test for equal variances).

Analyzing Bifurcation

To better understand the bifurcation effect of our treatment and to test for Hypotheses 1 and 2, we report on drivers who drop out and refuse to make any effort at all as well as on those who overexert themselves. Dropouts are defined as drivers whose performance is worse than a specified threshold from any week during the treatment until the end of the treatment phase. In accordance with Müller and Schotter (2010), drivers who overexert themselves are denoted as workaholics. In the context of the present investigation, workaholics are drivers whose performance is better than a specified threshold from any week during the treatment until the end of the treatment phase. The thresholds used are the fleet’s lowest and highest quintiles of the fuel efficiency score during the control phase, which is in accordance with the literature on high performers (see, e.g., Kwon & Rupp, 2013). The lowest quintile of the fleet’s fuel efficiency score ended at 67 points, and the highest quintile started at a fuel efficiency score of 91. Figure 5 reports the dropouts and workaholics by treatment group.

Figure 5.

Cumulated dropout/workaholic rates by the sharing rule; dashed lines mark months.

While the proportional sharing rule results in more and earlier workaholics, the egalitarian sharing rule results in more and earlier dropouts. These results support our second hypothesis.

Multivariate Analysis

The previous results suggest that there was already a significant difference during the control phase. Additionally, we know that weather, traffic, and loading conditions affect the driver’s fuel efficiency score. Therefore, applying a multivariate approach in order to understand the effects of proportional and egalitarian sharing rules in team contests is crucial to analyze ceteris paribus conditions. We applied generalized linear panel data models to control for pretreatment differences between the proportional and the egalitarian sharing rule treatment, as well as for the impact of introducing team incentives at all. The dependent variable fuel efficiency score was not normally distributed (Shapiro–Wilk p = .00). The treatment variable and the essential explanatory variables (app, better one) were not time-varying at all, which prevented the use of fixed-effects models.

First, we observed that participating in a team contest significantly increased performance by 3.7 points under the proportional sharing rule and by 2.3 points under the egalitarian sharing rule (Model I, Table 3). This positive and significant effect remained nearly constant after controlling for the impact of the app, the distance, and the truck weight. Moreover, we tested for an interaction effect between the app and our treatment and found no evidence for such an effect.⁵ In Model III, we additionally controlled for the impact of the better team member, who reaches on average a 16.5 points higher fuel efficiency score than his worse peer. This significant effect is of similar size to the results shown in Figure 2. Winning a prize in the team contest in the last period significantly increased performance by approximately 6.1 points. Adding these two control variables reduced the impact of the proportional sharing rule to a significant increase by 2.6 points and the impact of the egalitarian sharing rule to a significant increase by 0.9 points.

Table 3.

Generalized Linear Models I to III.

Fuel efficiency score	Model I	Model II	Model III
Reference: Control
Team: proportional	3.721*** (0.619)	3.753*** (0.607)	2.594*** (0.479)
Team: egalitarian	2.300*** (0.687)	2.518*** (0.676)	0.914* (0.509)
Better one			16.45*** (0.410)
Prize in the previous period (t − 1)			6.283*** (0.648)
App		3.378*** (0.528)	2.773*** (0.399)
Weekly distance		3.737*** (0.763)	2.610*** (0.678)
Reference: Weight empty
Weight medium		−3.920 (16.90)	−24.72** (12.64)
Weight heavy		−0.0916 (16.91)	−22.91* (12.64)
Constant	79.56*** (0.455)	69.86*** (16.85)	86.66*** (12.58)
Observations	2,365	2,365	2,365

Note. Generalized linear model with time-constant variables (treatment, better one, and app) and time-varying variables—prize in previous period (t − 1), distance in 1,000 km, and share of distance driven per weight. Robust standard errors are in parentheses (clustered at individual levels).

***

p < .01. ^**p < .05. ^*p < .1.

To understand whether team performance is affected differently, we reran Models I and II using team values for the fuel efficiency score and team means for distance and weight (Table 4). Model IIIa differs from Model III because we controlled for within-team heterogeneity by using the variable standard deviation of team fuel efficiency score instead of the variable better one in analyzing the team level.

Table 4.

Generalized Linear Models Ia to IIIa.

Team fuel efficiency score	Model Ia	Model IIa	Model IIIa
Reference: Control
Team: proportional	8.443*** (0.723)	8.566*** (0.681)	5.510*** (0.592)
Team: egalitarian	3.655*** (0.777)	4.340*** (0.743)	3.093*** (0.663)
The standard deviation of team fuel efficiency score			−1.340*** (0.0536)
Prize in the previous period (t − 1)			9.791*** (0.872)
App		5.977*** (0.596)	4.358*** (0.498)
Team’s weekly distance		7.971*** (1.142)	7.346*** (1.009)
Reference: Team mean—weight empty
Team mean: weight medium		7.949 (33.824)	−8.995 (36.14)
Team mean: weight heavy		15.333 (33.787)	−8.832 (36.11)
Constant	159.083*** (0.462)	123.228*** (33.858)	163.6*** (36.17)
Observations	1,846	1,846	1,846

Note. Generalized linear model with time-constant variables (treatment and app) and time-varying variables (prize in previous period [t − 1], distance in 1,000 km, and share of distance driven per weight). Robust standard errors are in parentheses (clustered at team levels).

***

p < .01. ^**p < .05. ^*p < .1.

Both the direction and significance of the coefficients remain on the aggregated team level. The coefficients increased, as can be expected, although the coefficient for the proportional sharing rule was clearly higher, as was expected from the individual-level regression. An increase in within-team heterogeneity resulted in a decrease in the mean fuel efficiency score by approximately 1.3 points. As already predicted by theory, Figure 2 and Model III revealed that the egalitarian and proportional sharing treatments had different effects on the better and worse drivers of a team due to free-riding or bifurcation effects. To analyze this impact in a multivariate setting, we reran a variation of Model III separately for both groups.

Table 5 shows the results for the lower performing team member in Model IVa and for the higher performing team member in Model IVb. The fuel efficiency score of the worse driver increased significantly by about 5 points under the proportional sharing rule, but not under the egalitarian sharing rule. In contrast, the fuel efficiency score of the better driver increased significantly by about 1 point under egalitarian sharing rule, but not under the proportional sharing rule. Thus, the better driver seems to anticipate the free-riding intention of his peer and improved his performance slightly more than his peer under the egalitarian sharing rule. Being successful in the contest in the last period increased the performance of the worse driver more than the performance of the better driver. To understand the effect of the sharing rule on the worse and better drivers of a team, we applied this information to the sharing rule treatment—in the form of Model V. There is a strong and highly significant interaction effect between the treatment variable (sharing rule) and the impact of being the better peer of a team.

Table 5.

Generalized Linear Models IVa, IVb, and V.

Fuel efficiency score	Model IVa (worse driver)	Model IVb (better driver)	Model V
Reference: Control
Team: proportional	5.028*** (0.872)	0.316 (0.399)	6.040*** (0.860)
Team; egalitarian	0.986 (0.935)	0.990** (0.414)	2.003** (0.910)
Better one			19.24*** (0.657)
Team: proportional#			−6.672*** (0.941)
Better one
Team: egalitarian#			−2.127** (1.007)
Better one
Prize in previous period (t − 1)	12.19*** (0.933)	0.150 (0.470)	6.226*** (0.604)
App	4.938*** (0.723)	0.643** (0.324)	2.772*** (0.395)
Weekly distance	2.803*** (1.042)	2.241*** (0.831)	2.599*** (0.674)
Reference: weight empty
Weight medium	−29.25* (15.93)	−7.505 (12.90)	−26.99** (12.43)
Weight heavy	−29.91* (15.94)	−4.051 (12.86)	−25.20** (12.43)
Constant	89.77*** (15.88)	88.25*** (12.84)	87.55*** (12.38)
Observations	1,141	1,224	2,365

***

p < .01. ^**p < .05. ^*p < .1.

Figure 6 reports the marginal effects of the interaction assumed in Model V. We plotted the predicted mean performance depending on the team member and the sharing rule. This plot illustrates that free-riding among bad performers strongly declined with the proportional sharing rule, thus, with a combination of between- and within-team tournaments. Also, we again found that better performers anticipated free-riding from their peers under the egalitarian sharing rule and increased their performance slightly more in this treatment condition.

Figure 6.

Predictive margins of mean fuel efficiency score by treatment and team member.

Results of the Posttreatment Phase

After our intervention, we observed the drivers’ performances for another 3 months. During this phase, drivers were treated similarly to the control phase. The posttreatment phase still included a group of drivers who were given the choice of using an app to inform themselves about their driving behavior. However, the drivers were not paired in teams anymore. This phase also removed the contest and thus the opportunity to win a prize. If the increase in performance and the overall reduction in performance variation were observed only because of the effect of an intervention and not because of our treatment, we would have expected performance to decrease at least to the control phase level.

Figure 7 shows that the worse drivers caught up, an impact that remained even after our intervention (posttreatment). We show that the impact on already good and very good performing drivers is also positive and significant but not as strong as that on the worse drivers. In the posttreatment phase, the better drivers’ performances were higher than the performance during the control phase but somewhat lower compared with the treatment phase. In both groups, the worse driver caught up in the posttreatment period. We also found a reduction in variation to persist at least on the level of the treatment or to decrease even further (see Figure 8).

Figure 7.

Mean fuel efficiency score by treatment, phase, and team member.

Figure 8.

Standard deviations of fuel efficiency score by treatment group, phase, and team member.

Discussion and Conclusion

We ran a natural field experiment in the transportation sector to examine whether a combination of between- and within-team tournaments improves performance. In particular, we analyzed how adding a within-team tournament component (proportional sharing rule) to a pure between-team tournament (egalitarian sharing rule) affects potential shortcomings such as free-riding and bifurcation. We found that introducing team tournaments, in general, improves individual performance. This finding is in line with previous results on tournaments (e.g., Nalbantian & Schotter, 1997). We also found a positive and sustainable effect of tournaments on performance: In egalitarian and in proportional tournaments, drivers remained at a higher level or even improved further even after the tournament was stopped.

Our results, obtained in a field setting, support the overall results of Majerczyk et al. (2019) from a laboratory experiment to large extents, and provide thus external validity for their results. Due to our setting, we add to the previous results that the effects are sustainable. Moreover, we are able to analyze differences in performance more nuanced: Using descriptive statistics, the increase in performance seems to be slightly stronger under the egalitarian sharing rule than under the proportional sharing rule. However, using DiD estimates, we already show that this difference is only small and also only significant on the team, not the individual level. Applying multivariate analysis that allows us to control for essential impact factors like weight and distance drove, we find: (1) the team contest improves individual as well as team performance; (2) this increase in performance is overall stronger under the proportional than under the egalitarian sharing rule when we apply multivariate analysis and thus, under ceteris paribus interpretation; and (3) the performance increase is mainly driven by the team member performing worse. Especially the last finding is new and essential in our view: When firms introduce team tournaments, it is essential to know that these tournaments will not affect or demotivate good performers, but that especially worse performers catch up. When the firm’s goal is to improve overall performance (the mean), team tournaments will be a good means.

For whole firms, our results are also crucial in order to foster performance equality: The results obtained support our first hypotheses, that free-riding is higher under the egalitarian than under the proportional sharing rule. In the mean, this effect is mitigated as good-performing drivers increase their performance slightly but significantly under the egalitarian sharing rule than under the proportional sharing rule.

Müller and Schotter (2010) stressed that bifurcation is a possible downside of tournaments: Low performers further decrease their effort and drop out of the tournament. Simultaneously, previous high performers increase their effort. Low performers’ dropping out is a problem in tournaments because increasing performance variance may void overall increases in performance. Dropouts are problematic not only in individual tournaments but also in team tournaments under the egalitarian sharing rule. In the latter, some individuals may drop out, and others may be tempted to free-ride on their teammates’ efforts. One possibility to prevent dropouts and free-riding has a proportional sharing rule. Under this rule, low-performing drivers increased their performance more than under the egalitarian sharing rule. Given that the number of workaholics was higher in proportional sharing than in egalitarian sharing, we still found a positive effect for proportional sharing.

In our setting, the term workaholics is somehow odd: Driving well and not wasting fuel is associated with effort. However, this is in contrast to “workaholics” in the sense of Müller and Schotter (2010) because it is possible to drive well over a very long period without running into the danger of cracking under pressure. In other work situations, exerting substantial effort is probably more associated with physical or psychological exhaustion compared with our setting of eco-friendly and fuel-efficient driving. Thus, we concluded that proportional sharing rules are qualitatively better suited for our setting than egalitarian sharing rules.

Compared with the results obtained in the laboratory, our field setting has some crucial advantages. First, we catch temporal dynamics: Because laboratory experiments only last about an hour, incentive schemes may play out very differently than long-term observations. In our first month, our results are close to the laboratory findings. Significant changes and departures in performance set in only after the third month. Thus, we assume that the short- and long-term consequences of proportional and egalitarian sharing rules are different from each other. However, if we want to understand better the incentive effects and, in particular, temporal dynamics, our results suggest that replicating laboratory experiments in the field with real workers in a long-term observational setting is crucial.

From the firm’s perspective, introducing tournaments not only improved performance but also paid out in fuel consumption (Table 6). The average fuel use is reduced from 30.34 L/100 km during the control phase to 30.27 L/100 km under the egalitarian and to 30.26 L/100 km under the proportional sharing treatment. As a result, the firm saved 9.5 cents per 100 L of fuel in the proportional sharing tournament and 8.5 cents per 100 L in the egalitarian tournament. Given the weekly mean of 2,453 km per week and truck, this reduction in average fuel use translates into weekly savings for the firm of about €233.035 under the proportional sharing treatment and €208.505 under the egalitarian sharing tournament.

Table 6.

Mean Fuel Usage and Money Saved by Phase and Treatment Relative to the Control Phase.

Mean fuel price May 12, 2018: €1.2831		Mean L/100 km	Euros saved per 100 km	Euros saved per week for 100 trucks
Control phase (May 7, 2018)		30.34
Treatment phase (August 12, 2018)	Egalitarian	30.27	.095	233.035
Treatment phase (August 12, 2018)	Proportional	30.26	.085	208.505

Although fuel consumption is also subject to exogenous factors, our experiment helped save fuel for the company (Table 6 and Appendix D) without generating side effects on speed or punctuality (Appendix E). Thus, we increased the overall efficiency of the fleet.

Behaving energy-efficient becomes increasingly important, not only in the transportation sector but also in other jobs. However, in most jobs, energy efficiency is not the main goal of employees. But the closer tasks are related to productivity or added value the more central energy efficient performance becomes. Yet also energy efficiency as a goal is important and many organizations offer a wider variety of measures (Maki et al., 2019). These measures are important, but seldom accompanied by organizational incentives to make use of them. Recently, gamification elements are brought forward to encourage energy efficiency at work (Lounis et al., 2017). These gamification elements frequently entail some type of tournament or ranking, mostly in individual tournaments. Individual tournaments, however, are dangerous as they may undermine cooperation in organizations (Blader et al., 2020). Our study points up a different avenue that is team tournaments. We show that team tournaments improve energy efficiency, regardless of how they are designed. In both design options, those who were already energy-efficient did not decline or improve. Proportional sharing rules may be more suited than egalitarian sharing rules, as they are slightly better to improve bad-performing individuals. Noteworthy, from our point of view, performance improvements in both settings were sustainable, implying that energy efficiency as a goal needs to be made salient once, but individuals seem to understand the value and adhere to a more energy-efficient behavior even after the incentives were removed.

In our field experiment, the reduced fuel usage goes hand in hand with mitigation of the fleet’s greenhouse gas emissions. This mitigation is especially important because the transportation sector is, as stated in the “Introduction” section, one of the most energy-intensive sectors. Understanding how to motivate workers to behave in an energy-efficient manner could be critical to mitigating CO₂ emissions in this industry because saving fuel is linearly related to their reduction. Given the decrease in mean fuel consumption per 100 km, the mean CO₂ emissions declined by 4.58% between the control phase and the treatment.

Footnotes

Appendix A

Appendix B

Appendix C

Interaction App × Treatment

Fuel efficiency score	Model II
Reference: Control
Team: proportional	3.487^*** (0.917)
Team: egalitarian	2.479^** (0.899)
App	3.187^*** (0.899)
Team: proportional#App	0.530 (1.219)
Team: egalitarian#App	0.054 (1.350)
Weekly distance	3.737^*** (0.763)
Reference: Weight empty
Weight medium	−4.002 (16.874)
Weight heavy	−0.205 (16.884)
Constant	70.065*** (16.845)
Observations	2,365

Appendix D

Impact of Fuel Efficiency Score on Fuel Consumption

	Fuel consumption in L/100 km
Fuel efficiency score	−0.0694*** (0.00391)
Weight (Reference = empty)
Medium	15.85*** (2.622)
Heavy	7.694*** (2.624)
Constant	23.08*** (2.638)
Observations	3,411
Number of ID	92
R² within	.3492
R² between	.1658
R² overall	.2877

Note. Fixed-effects estimation, robust standard errors in parentheses.

***

p < .01. ^**p < .05. ^*p < .1.

Appendix E

Acknowledgements

We are deeply grateful to the owner, management, and all employees of the Euba Logistics GmbH. In particular, we thank Ronald Garkisch, Kerstin Nitz, and Elke Westenberger for their strong encouragement throughout the project and generous allowance to analyze the firm’s data.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Christin Hoffmann

Notes

Author Biographies

Christin Hoffmann is a Post-doctoral Researcher at the Brandenburg University of Technology Cottbus-Senftenberg. She holds a PhD in economics from the University of Bern. Her research interests lie at the intersections of Behavioral, Environmental and Resource Economics as well as Transportation Economics. Recurring themes in her research are the (behavioral) adaptation to combat climate change impacts. Her work has been published in journals such as Ecological Economics, Economic Letters, and Climate Change Economics.

Kirsten Thommes is a Professor for Organizational Behavior at the University of Paderborn. Before this, she studied business administration at the Phillips University of Marburg, received her doctorate at the Friedrich Schiller University in Jena and worked as a post-doc at the Radboud University of Nijmegen as well as the RWTH Aachen University. Her research interests include Organizational Behavior and Organizational Identity, Human-Machine Interaction, Teamwork and Time Preference. Her work has been published in journals such as Organization Studies, Journal of Economic Behavior & Organization, Economic Letters, and Industrial and Labor Relations Review.

References

Agranov

Tergiman

(2013). Incentives and compensation schemes: An experimental study. International Journal of Industrial Organization, 31(3), 238-247. https://doi.org/10.1016/j.ijindorg.2012.06.001

Blader

Gartenberg

Prat

(2020). The contingent effect of management practices. Review of Economic Studies, 87(2), 721-749. https://doi.org/10.1093/restud/rdz034

Bull

Schotter

Weigelt

(1987). Tournaments and piece rates: An experimental study. Journal of Political Economy, 95(1), 1-33. https://doi.org/10.1086/261439

Cappelli

Tavis

(2018). HR goes agile. Harvard Business Review, 96(2), 46-52. https://hbr.org/2018/03/the-new-rules-of-talent-management

Cason

T. N.

Sheremeta

R. M.

Zhang

(2017). Asymmetric and endogenous within-group communication in competitive coordination games. Experimental Economics, 20(4), 946-972. https://doi.org/10.1007/s10683-017-9519-2

Charness

Gneezy

Kuhn

M. A.

(2013). Experimental methods: Extra-laboratory experiments-extending the reach of experimental economics. Journal of Economic Behavior & Organization, 91, 93-100. https://doi.org/10.1016/j.jebo.2013.04.002

Ericsson

K. A.

Charness

(1994). Expert performance: Its structure and acquisition. American Psychologist, 49(8), 725-747. https://doi.org/10.1037/0003-066X.49.8.725

European Environment Agency. (2018). Greenhouse gas emissions from transport. https://www.eea.europa.eu/data-and-maps/indicators/transport-emissions-of-greenhouse-gases/transport-emissions-of-greenhouse-gases-10

Flamand

Troumpounis

(2015). Prize-sharing rules in collective rent seeking. In Congleton

R. D.

Hillman

A. L.

(Eds.), Companion to political economy of rent seeking (pp. 92-112). Edward Elgar.

10.

Gunnthorsdottir

Rapoport

(2006). Embedding social dilemmas in intergroup competition reduces free-riding. Organizational Behavior and Human Decision Processes, 101(2), 184-199. https://doi.org/10.1016/j.obhdp.2005.08.005

11.

Harbring

(2006). The effect of communication in incentive systems: An experimental study. Managerial and Decision Economics, 27(5), 333-353. https://doi.org/10.1002/mde.1266

12.

Harbring

Irlenbusch

(2003). An experimental study on tournament design. Labour Economics, 10(4), 443-464. https://doi.org/10.1016/S0927-5371(03)00034-4

13.

Harbring

Irlenbusch

(2005). Incentives in tournaments with endogenous prize selection. Journal of Institutional and Theoretical Economics, 161(4), 636-663. https://www-jstor-org-443.web.bisu.edu.cn/stable/40752544

14.

Harbring

Irlenbusch

(2008). How many winners are good to have? On tournaments with sabotage. Journal of Economic Behavior and Organization, 65(3-4), 682-702. https://doi.org/10.1016/j.jebo.2006.03.004

15.

Hwang

Erkens

D. H.

Evans

J. H.

III . (2009). Knowledge sharing and incentive design in production environments: Theory and evidence. Accounting Review, 84(4), 1145-1170. https://doi.org/10.2308/accr.2009.84.4.1145

16.

Irlenbusch

Ruchala

G. K.

(2008). Relative rewards within team-based compensation. Labour Economics, 15(2), 141-167. https://doi.org/10.1016/j.labeco.2007.02.003

17.

Kugler

Rapoport

Pazy

(2010). Public good provision in inter-team conflicts: Effects of asymmetry and profit-sharing rule. Journal of Behavioral Decision Making, 23(4), 421-438. https://doi.org/10.1002/bdm.667

18.

Kwon

Rupp

D. E.

(2013). High-performer turnover and firm performance: The moderating role of human capital investment and firm reputation. Journal of Organizational Behavior, 34(1), 129-150. https://doi.org/10.1002/job.1804

19.

Lazear

E. P.

Rosen

(1981). Rank-order tournaments as optimum labor contracts. Journal of Political Economy, 89(5), 841-864. https://doi.org/10.1086/261010

20.

Livio

De Chiara

(2019). Friends or foes? Optimal incentives for reciprocal agents. Journal of Economic Behavior & Organization, 167, 245-278. https://doi.org/10.1016/j.jebo.2018.07.012

21.

Lounis

Kotsopoulos

Bardaki

Papaioannou

T. G.

Pramatari

(2017, May). Waste no more: Gamification for energy efficient behaviour at the workplace. In Tuomi

Perttula

(Eds.), Proceedings of the first International GamiFIN Conference (pp. 129-134). Creative Commons.

22.

Majerczyk

Sheremeta

Tian

(2019). Adding tournament to tournament: Combining between-team and within-team incentives. Journal of Economic Behavior & Organization, 166, 1-11. https://doi.org/10.1016/j.jebo.2019.09.002

23.

Maki

McKinney

Vandenbergh

M. P.

Cohen

M. A.

Gilligan

J. M.

(2019). Employee energy benefits: What are they and what effects do they have on employees? Energy Efficiency, 12(5), 1065-1083. https://doi.org/10.1007/s12053-018-9721-x

24.

Merchant

K. A.

Van der Stede

W. A.

(2007). Management control systems: Performance measurement, evaluation and incentives. Pearson Education. https://books.google.de/bookshl=de&lr=&id=ymY2cQN3iEMC&oi=fnd&pg=PR13&dq=Merchant+van+der+Stede&ots=Q6RSfFQsyu&sig=eagmGq2htha_MJ55HEBwXtvk9uo#v=onepage&q=Merchant%20van%20der%20Stede&f=false

25.

Moldovanu

Sela

(2001). The optimal allocation of prizes in contests. American Economic Review, 91(3), 542-558. https://doi.org/10.1257/aer.91.3.542

26.

Müller

Schotter

(2010). Workaholics and dropouts in organizations. Journal of the European Economic Association, 8(4), 717-743. https://doi.org/10.1111/j.1542-4774.2010.tb00538.x

27.

Nalbantian

H. R.

Schotter

(1997). Productivity under group incentives: An experimental study. American Economic Review, 87(3), 314-341. https://www-jstor-org-443.web.bisu.edu.cn/stable/2951348

28.

Organisation for Economic Co-operation and Development. (2017). ITF Transport Outlook 2017. https://www.itf-oecd.org/transport-outlook-2017

29.

Orrison

Schotter

Weigelt

(2004). Multiperson tournaments: An experimental examination. Management Science, 50(2), 268-279. https://doi.org/10.1287/mnsc.1030.0128

30.

Rosenbaum

M. E.

Moore

D. L.

Cotton

J. L.

Cook

M. S.

Hieser

R. A.

Shovar

M. N.

Gray

M. J.

(1980). Group productivity and process: Pure and mixed reward structures and task interdependence. Journal of Personality and Social Psychology, 39(4), 626-642. https://doi.org/10.1037/0022-3514.39.4.626

31.

Schall

D. L.

Wolf

Mohnen

(2016). Do effects of theoretical training and rewards for energy-efficient behavior persist over time and interact? A natural field experiment on eco-driving in a company fleet. Energy Policy, 97, 291-300. https://doi.org/10.1016/j.enpol.2016.07.008

32.

Sheremeta

R. M.

(2011). Perfect-substitutes, best-shot, and weakest-link contests between groups. Korean Economic Review, 27, 5-32.

33.

Sisk

(2005). Reward systems that really work. Harvard Management Update, 10(9), 1-4.

34.

Sutter

Strassmair

(2009). Communication, cooperation and collusion in team tournaments—An experimental study. Games and Economic Behavior, 66(1), 506-525. https://doi.org/10.1016/j.geb.2008.02.014

35.

U.S. Environmental Protection Agency. (2018). Fast facts on transportation greenhouse gas emissions. https://www.epa.gov/greenvehicles/fast-facts-transportation-greenhouse-gas-emissions

36.

Van Dijk

Sonnemans

Van Winden

(2001). Incentive systems in a real effort experiment. European Economic Review, 45(2), 187-214. https://doi.org/10.1016/S0014-2921(00)00056-8

37.

Wageman

(1995). Interdependence and group effectiveness. Administrative Science Quarterly, 40(1), 145-180. https://doi.org/10.2307/2393703