Using Experiments to Study Shared Capitalism

1. Using Experiments to Study Shared Capitalism 2011 Mid-year Fellows Workshop February 25th School of Management and Labor Relations Rutgers University

2. Introduction • 35 year old theoretical literature developed that investigates how aspects of shared capitalism affects employee motivation (e.g. Vanek, 1977; Ellerman, 1982, Ben-Ner, 1995). • Mounting evidence that employee ownership is associated with a weak-positive effect on firm performance • Employee ownership (e.g. Kruse, (1993), Freeman and Dube, (2000), Blasi, Kruse, and Bernstein, (2003)). • Participation and productivity (e.g. Levin and Tyson (1990))

3. A “dauntingly difficult” question • The identification of causal relationships linking structural firm characteristics to employee motivation

4. A “dauntingly difficult” question • The identification of causal relationships linking structural firm characteristics to employee motivation • Measuring performance • Task specific • Multiple tasks • Ability

5. A “dauntingly difficult” question • The identification of causal relationships linking structural firm characteristics to employee motivation • Measuring performance • Task specific • Multiple tasks • Ability • Unobservable firm characteristics • Social capital • Production technique • Monitoring practices

6. Experimental Advantages: Control • Take a specific theoretical model • Compare predictions of the model to what happens in the experiment • If theory is rejected, it is relatively easy to test competing explanations (e.g. inequality aversion, reciprocity, framing)

7. Experimental Advantages: Test-bedding • Institutions that are predicted to yield identical results may not do so • Experiments can be used to show that institutions/mechanisms that work in theory don’t always work in practice (e.g. Plott, 1987)

8. (Behavioral) Personnel Economics • (1) Wide variance in effort under rank-order compensations schemes (e.g. Bull et al. (1987); van Dijk et al. (2001); Shearer (2004)) • (2) Strong tendency for workers to reciprocate ‘gift-wages’ with higher effort (e.g. Fehr et al. (1998); Fehr et al. (1993); Gachter and Falk (2001); Charness and Haruvy (2002); Maximiano et al. (2007)) • (3) ‘Pay-for-performance’ not at straightforward (e.g. Gneezy and Rustichini (2000); Titmuss, (1971); Fehr, et al. (1998))

9. What 1/n problem? (1) Sufficient levels of other-regarding preferences • (e.g. Fehr and Schmidt (1999); Pech (2009); Bolton and Ockenfels (2000); Charness and Rabin (2002))., (2) Subjects can communicate • (e.g. Ledyard (1995); Sally (1995); Cooper et al. (1992); Croson et al. (2003); Suetens (2005); Duffy and Feltovich (2002); Frohlich and Oppenheimer (1998)) (3) A fraction of subjects display a willingness to enforce cooperation at a monetary and/or social cost • (Anderson and Stafford (2003); Dickinson (2001); Andreoni et al. (2003); Carpenter and Seki (2005)) (4) Subject have developed some type (even very superficial) group identity • (Eckel and Grossman (2005); Cox et al. (1991); Charness and Jackson (2007); McLeish and Oxoby (2007))

10. Ownership/Control and Motivation

11. Reduced form Mellizo, Carpenter, Matthews (2011) Groups of workers able to participate in determining their own compensation scheme increased effort significantly relative to groups that had this decision implemented by an exogenous process. These effects persist after controlling for gender, compensation scheme, and ability.

12. Experimental Controls • By design, we control for all issues that pertain to monitoring, punishment, threats, or other forms of coercion that might also accompany many types of systems of control in real-world firms. • No team production (minimize confounds that could arise in social dilemmas • Restrict the menu of potential compensation schemes, • All claims on residual profits are held by labor.

13. Percentage of Company Stock Owned by Employee Ownership Plans Reproduced from NCEO “A Statistical Profile of Employee Ownership” http://www.nceo.org/main/article.php/id/2/

14. Basic Hypothesis of the Incentive Effect • Ho: Equity holdings do not affect performance

15. Protocol Summary • 3 basic treatments (group size = 3) • Control: each subject is paid a flat-wage of $1.50 • Treatment 1: each subject is paid a $1.50 flat-wage and is allocated 1 out of 9 total equity shares (33% employee ownership; 11% private stake) • Treatment 2: each subject is paid a $1.50 flat-wage and is allocated 3 out of 9 total equity shares (100% employee ownership; 33% private stake)

16. Experiment Design • Between-subject design • Real-effort task: Subjects asked to perform a simple production task of adding up sets of five, 2-digit, numbers • Each period lasted for 5 minutes • 5 work periods

17. Experiment Design • Asset value: The value of the company stock is assumed to be a function of both the combined effort its workforce and a vector of uncertainty • The value of the firm’s stock is the sum of the output provided by all three subjects measured by the number of correct answers provided in their work task multiplied by a random factor r that between 0.10 and 2.00 (akin to MPCR in VCMs) • The product is then divided by the total number of shares (9) to determine the value of each share.

18. Expectations • “Straw-man” model assuming rational and selfish individuals would not predict any performance differences • We would not be surprised if a worker with a small ownership share were to “irrationally” confuse their ability to affect the value of the share by offering a high level of effort when their contribution cannot be teased out from the firm value.

19. Expectations 2 • Illusory correlation fort) in an attempt to give structure to unpredictable situations (e.g. Glass and Singer (1972); Whitson and Galinsky, (2008)). • Quasi-magical thinking • Over-valuation of low-probability events

20. Pilots • Total n = 18 • Treatment 1 n = 9 • Treatment 2 n = 9 • Abbreviated pilot study (2 round) conducted at The College of Wooster in later in the Fall 2010 term. • While the number of observations is far too limited to draw any (weak) conclusions, there are no differences between effort offered in Treatment 1 and Treatment 2 (Mann-Whitney Prob>|z|= 0.421).

21. What is next