Elsevier

Games and Economic Behavior

Volume 91, May 2015, Pages 97-113
Games and Economic Behavior

Firing threats: Incentive effects and impression management

https://doi.org/10.1016/j.geb.2015.02.015Get rights and content

Highlights

  • We study the effect of firing threats in a real-effort/real-leisure experiment.

  • Firing threats sharply increase worker production and decrease on-the-job leisure.

  • Workers are inclined to impress their boss in the presence of firing threats.

  • Firms endowed with firing threats underperform those using individual incentives.

  • After the removal of firing threats, production drops and on-the-job leisure surges.

Abstract

We study the effect of firing threats in a virtual workplace that reproduces features of existing organizations. We show that organizations in which bosses can fire up to one third of their workforce produce twice as much as organizations for which firing is not possible. Firing threats sharply decrease on-the-job leisure. Nevertheless, organizations endowed with firing threats underperformed those using individual incentives. In the presence of firing threats, employees engage in impression management activities to be seen as hard-working individuals in line with our model. Finally, production levels dropped substantially when the threat of being fired was removed, whereas on-the-job leisure surged.

Introduction

In settings in which employers are unable to provide individual incentives to workers, the threat of being fired becomes an essential feature of an employment contract (Becker and Stigler, 1974, Klein and Leffler, 1981, Shapiro and Stiglitz, 1984, MacLeod and Malcomson, 1989). At the empirical level, researchers have attempted to assess the effectiveness of firing threats by studying employment protection legislation. For example, Ichino and Riphahn (2005) study absenteeism levels of the workers of an Italian bank before and after a probationary period. The authors report an increase in the level of absenteeism after the probationary period suggesting a negative incentive effect of employment protection. Similarly, Riphahn (2004) shows a negative relationship between employment protection and the level of absenteeism of German workers. Recently, Jacob (2013) reports a 10% decrease in absenteeism levels after principals were allowed more flexibility in firing teachers in the public schools system in Chicago. Interestingly, Jacob (2013) distinguishes incentive and selection effects of the newly implemented policy, and shows the prevalence of selection over incentive effects. These studies provide field evidence that restricting dismissal policies is likely to foster absenteeism which could in turn reduce workers' performance.

In this paper, we propose a different approach by assessing the impact of firing threats in a laboratory environment in which we directly assess the incentive effects of firing threats. In addition to obtaining direct measures of productivity and shirking behavior, the laboratory setting allows us to control for possible confounding factors such as firm size, industry structure, job characteristics, demand shocks, and organizational design.

Our aim is to complement previous research by measuring directly workers' effort, productivity and shirking behavior and assess the magnitude of incentive effects which were found to be modest in the field (Jacob, 2013). To that end, we study firing threats within a virtual workplace that reproduces features of existing organizations such as real effort tasks, real-time monitoring, on-the-job leisure (Internet browsing) and chatting (Corgnet et al., forthcoming).1 We study organizations in which bosses are endowed with a real-time monitoring technology so as to assess the work of their nine employees in each of the five periods of the experiment. In addition, we gave organizational members access to an electronic chat room to exchange messages during the experiment.

We consider three types of incentive structures. In all cases subjects received the same fixed wage at the beginning of each period regardless of pending work productivity. Under the fixed wage treatment employees received no further incentives, while under firing threats they could be fired. Under individual incentives employees could not be fired and each employee was rewarded the entire income generated by his or her individual production on the work task. In the fixed wage and firing threats treatments the boss kept all the income generated by all members of the organization. Under firing threats, the boss was also given the option to dismiss one employee at the end of each of periods 2, 3, and 4.2 Bosses saved on labor costs after firing employees as they would not have to continue to pay their fixed wages. Subjects who made it to the start of the last period without being fired found themselves with de facto tenure for that final production period.

Our analysis relates to the seminal work of Brown et al. (2004) which studies a repeated principal-agent model à la Fehr et al. (1993) in which there is an excess supply of agents. In this setting, principals and agents can sign one-period contracts which specify a fixed payment from the principal to the agent and a desired but non-enforceable level of transfer from the agent to the principal. A crucial difference with the original setting of Fehr et al. (1993) is that the authors allow for reputational concerns and long-term contracts by keeping subjects' identification numbers constant across periods. The authors find that principals and agents were willing to develop long-term relationships which in turn resulted in high levels of transfers. The findings in Brown et al. (2004) are in line with the disciplining version of the efficiency wage hypothesis according to which a combination of high wages and threat of dismissal leads to high levels of effort.

In a recent experimental study, Falk et al. (2011) extend the work of Brown et al. (2004) by introducing barriers to dismissals. The authors show that dismissal barriers tend to deter principals from building long-term relationships with agents. This is the case because agents' transfers are significantly reduced when the threat of being dismissed by the principal is eliminated. Note that in Brown et al. (2004) and in Falk et al. (2011) dismissals occur either because the principal signs a one-period contract with another agent or because the agent quits. Even though this contractual design constitutes a privileged setting for studying relational contracts and dismissal barriers, it cannot isolate the effect of firing threats from the effect of quitting. In the present study, we abstract away from career concerns and labor markets and focus on the impact of firing threats on organizational behavior.

Outside the laboratory, a number of studies using archival data have documented pervasive incentive effects by comparing work performance under hourly wages and piece rates (e.g. Lazear, 2000, Shearer, 2004). Also, recent research using data from a large US firm during the 2008 recession suggests that the intensification of firing threats during the economic downturn may explain the increase in productivity and work effort during the period (Lazear et al., 2013). In contrast to previous research, we study firing threats in a laboratory environment in which we can directly control for the intensity of firing threats as well as observe managers' firing decisions. In our setting, we assess the effect of different incentive schemes on workers' behavior inside firms. This includes the analysis of work effort, on-the-job leisure and workplace communication in addition to standard measures of work performance such as productivity.

We found that organizations in which bosses were allowed to fire their employees produced twice as much as organizations which only relied on the payment of fixed wages. This was the case even though by the end of the experiment organizations which could fire employees were about 30% smaller than those that could not. Firing threats also decreased Internet usage and chatting activities by 77.7% in those periods in which the threat of being fired was present. Remarkably, firing threats reduced leisure activities and increased production levels for both low- and high-ability workers. Nevertheless, the incentive effect of firing threats was not as compelling as those of individual incentives, as organizations endowed with individual incentives outperformed those endowed with firing threats by 43.3%. Also, the implementation of firing threats required significantly more monitoring effort (12.5% of managers' time) than the implementation of individual incentives (only 2.5% of managers' time) making it a less appealing option for managers. Interestingly, leisure activities were as low in the presence of firing threats as they were under individual incentives. As a result, the difference in workers' production levels between the two treatments was due to a discrepancy in productivity levels rather than to a difference in working time. These findings suggest that in the presence of firing threats, employees were willing to signal themselves as hard-working individuals who spend long hours at their workstation without browsing the Internet. Social psychologists refer to this process by which people attempt to influence others' perceptions of themselves as impression management.3

When firing threats disappear during the last production period of the firing threats treatment, workers' production collapsed, and Internet usage surged, to levels which were similar to those of organizations which solely relied on the payment of fixed wages.

Section snippets

Virtual organizations

We develop a framework in which subjects can undertake a real-effort organizational task while having access to Internet browsing and chatting activities at any point in time during the experiment. Each activity was undertaken in a separate window so that the experimenter had a precise measurement of the time spent on each activity. We consider organizations with ten subjects, nine of which were referred to as B subjects (employees) while the remaining subject was referred to as the C subject

Hypotheses: firing threats and impression management

In order to establish predictions regarding production levels and Internet usage across treatments, we rely on standard incentive theory (see Laffont and Martimort, 2002 for a review). We build on a two-period model of an organization composed of multiple workers and a supervisor.15 In the firing treatment, the first

Results

In Sections 4.1 Production and Internet usage, 4.2 Firing decisions, 4.3 Selection effects, 4.4 Monitoring, 4.5 The effect of firing threats across ability levels we analyze the first four periods of our experiment which correspond to probationary periods in Treatment F. In Section 4.6, we present the results of the last period when the threat of being fired is removed. In Appendix A we also report our findings when pooling the data for the five periods of the experiment. However, the

Conclusions

In this paper, we investigated the impact of firing threats in a virtual organization characterized by real-effort tasks, access to leisure activities and real-time supervision. We showed that the introduction of firing threats significantly affected organizational behavior. In particular, production was more than twice as high in the presence of firing threats as in their absence while on-the-job leisure activities (Internet browsing and chatting) were almost eradicated. These results show

References (37)

  • T. Eriksson et al.

    Feedback and incentives: experimental evidence

    Lab. Econ.

    (2009)
  • R. Riphahn

    Employment protection and effort among German employees

    Econ. Letters

    (2004)
  • F. Van Dijk et al.

    Incentive systems in a real effort experiment

    Europ. Econ. Rev.

    (2001)
  • J. Abeler et al.

    Reference points and effort provision

    Amer. Econ. Rev.

    (2011)
  • B. Bartling et al.

    Egalitarianism and competitiveness

    Amer. Econ. Rev.

    (2009)
  • G.S. Becker et al.

    Law enforcement, malfeasance, and compensation of enforcers

    J. Legal Stud.

    (1974)
  • M. Brown et al.

    Relational contracts and the nature of market interactions

    Econometrica

    (2004)
  • G. Charness et al.

    Competitive preferences and status as an incentive: experimental evidence

    (2010)
  • B. Corgnet et al.

    Peer pressure and moral hazard in teams: experimental evidence

    Rev. Behav. Econ

    (2013)
  • S. Datta et al.

    Rank-sum tests for clustered data

    J. Amer. Statistical Assoc.

    (2005)
  • E.L. Deci

    Effects of externally mediated rewards on intrinsic motivation

    J. Pers. Soc. Psychol.

    (1971)
  • E.L. Deci

    Intrinsic Motivation

    (1975)
  • E.L. Deci et al.

    A meta-analytic review of experiments examining the effects of extrinsic rewards on intrinsic motivation

    Psychol. Bull.

    (1999)
  • D. Dickinson

    An experimental examination of labor supply and work intensities

    J. Lab. Econ.

    (1999)
  • T. Dohmen et al.

    You get what you pay for: incentives and selection in the education system

    Econ. J.

    (2010)
  • T. Dohmen et al.

    Performance pay and multi-dimensional sorting: productivity, preferences and gender

    Amer. Econ. Rev.

    (2011)
  • A. Donner et al.

    Randomization by cluster: sample size requirements and analysis

    Amer. J. Epidemiol.

    (1981)
  • A. Falk et al.

    Studying labor market institutions in the lab: minimum wages, employment protection, and workfare

    J. Inst. Theoretical Econ.

    (2007)
  • Cited by (0)

    The authors acknowledge financial support from the International Foundation for Research in Experimental Economics, the Argyros School of Business and Economics at Chapman University, the Spanish Ministry of Education [Grant 2012/00103/001], Spanish Plan Nacional I+D MCI [ECO2013-44879-R], 2014-17, and Proyectos de Excelencia de la Junta Andalucía [P12.SEJ.1436], 2014-18. This research was developed when the second author was visiting the Economic Science Institute at Chapman University and working at the University of Granada.

    View full text