Preferences and organization structure: Toward behavioral economics micro-foundations of organizational analysis

The paper proposes micro-foundations for organizational analysis grounded in behavioral economics. As Simon (1985) pointed out it, “nothing is more fundamental in setting our research agenda and informing our research methods than our view of the nature of the human beings whose behavior we are studying.” The paper examines optimal workplace-level organization structure (decision-making delegation, incentives and monitoring) relative to four common types of individuals, just selfish, civil, decent and dedicated employees (characterized in terms of their social preferences, selfversus other-regarding, reciprocity, trusting and trustworthiness). Four principal propositions arise from this analysis. (1) Mismatch between organization structure and employee preferences reduces productivity and profits. (2) The less prosocial employees in an organization, the more complex and sophisticated and therefore expensive the organization structure must be. (3) The less complex and less interdependent are employees’ tasks, the less dependent is organization structure on employee social preferences. (4) Heterogeneity of preferences poses a design a dynamic challenge as practices generally have to be tailored to one type of employee, and will be associated with exit of other types or adverse-selection by types that will seek to exploit it. © 2013 Elsevier Inc. All rights reserved. 1. Preferences and organization structure: toward behavioral economics micro-foundations of organizational analysis Cooperation, reciprocity, trust, truth-telling and virtuous behavior are ubiquitous in organizations. Yet also common are competition when cooperation is called for, distrust, cheating and other bad behaviors. Is it that “nice” people engage in laudable behaviors and “not-nice” people engage in detrimental behaviors (Kendrick and Funder, 1988)? But nice people do not always do nice things; for example, they may withdraw cooperation in response to others’ negative behaviors. And not-nice people may do nice things in response to incentives. In organizations, much effort goes toward encouragement of desirable behaviors and prevention of unwanted behaviors through allocation of decision-making, incentives, monitoring, hiring, promoting, firing, and more (Williamson, 1975; Ouchi, 1979; Brickley et al., 2009). Those who believe in the ubiquity of self-interest regard good organizational behavior as the consequence of good organization design; bad behavior can be explained as residual problems that are too expensive to eradicate. Those who believe that individuals are generally benevolent and driven by good values see good behavior as the natural state; bad behaviors reflect ∗ Tel.: +1 612 624 0867. E-mail address: [email protected] responses to dysfunctional design that is grounded in bad assumptions about individual preferences (Ferraro et al., 2005; Ghoshal, 2005). The literature in organization theory has generally acknowledged that there is more to individual behavior than concern for self-interest, but has not put forth a comprehensive view of human nature or described how it influences organization structure or architecture (see, for example, leading textbooks by Baron and Kreps, 1999; Hodge et al., 2003; Daft, 2007). The organization behavior literature acknowledges a range of human motivations but does not link them in a systematic way to organization structure (see, for example, Greenberg and Baron, 2000). The recognition that individuals are more than self-interest machines is, of course, not new in economics. Oliver Williamson’s and other new institutional economists’ contributions to organizational analysis (e.g., Williamson, 1985) have been partly grounded in a richer characterization of individuals in organizations. However, their objective has not been to ground comprehensive organization design in a broad view of employee preferences. Researchers from different scholarly traditions have been seeking to establish micro-foundations for organizational analysis. Bromiley (2005), in the tradition of Simon (1985) and Cyert and March (1963), grounds organizational analysis in individuals’ information processing abilities; Abell et al. (2008) ground their micro-foundations in routines practiced by individuals, extending Nelson and Winter’s (1982) macro-oriented analysis; and Van de Ven and Lifshits (2013) ground organizational analysis and 1053-5357/$ – see front matter © 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.socec.2013.08.003 Author's personal copy 88 A. Ben-Ner / The Journal of Socio-Economics 46 (2013) 87– 96 management theory in what constitutes fair, just, and proper behavior of a person in a given role and situation. A special issue of Academy of Management Perspectives (2013) was dedicated to the issue of micro-foundations to management thought and practice. However, an organizational analysis closely grounded in evidenceinformed assumptions about human nature is yet to be advanced. Aspects of the link between some dimensions of human nature – certain social preferences – and organization structure have been investigated. Ferraro et al. (2005), Ghoshal (2005), Frey and Osterloh (2005) and Nahapiet et al. (2005) call for grounding theories of management in a better understanding of human nature and values. Nahapiet et al. (2005) emphasize individuals’ motivation to excel, which leads them to pursue cooperation. Osterloh and Frey (2000) focus on participation in decision-making and contingent rewards in the presence of possible crowding-out of intrinsic motivation by extrinsic incentives; their argument regards the dynamic relationship between organization design and preferences. Other scholars have studied the connection between aspects of human nature and management practices as personjob or employee desires and job supplies fit (Kristof-Brown et al., 2005). Shaw et al. (2000) examine the links between task environment and preferences for group work, and the impact of the fit between them on performance and satisfaction. Bandiera et al. (2005), Fehr and Falk (2002), Tabellini (2008), Baron (2010) and others have demonstrated that the effectiveness of various incentives and other elements of organization structure depend on the precise nature of preferences. Benabou and Tirole (2006) examine the relationship between intrinsic and extrinsic preferences and the design of incentive schemes; Cunyat and Sloot (2011) compare organization structure based on different social preferences of managers; Englmaier and Leider (2012) test the tradeoff between reciprocal incentives and financial incentives. These contributions pave the way for – as well as invite – a more comprehensive examination of the relationship between preferences and organization structure. The paper builds on these and other contributions to establish behavioral-economics micro-foundations for organizational analysis. The paper describes four broad types of individuals, constructed on the basis of combinations of social preferences detected in experimental economics and that correspond to people that readers may easily recognize. One type is the familiar Homo economics, the purely-selfish individual who does not trust, is not trustworthy and does not reciprocate – the type that is central to standard economic analysis and therefore its findings. In experiments, and most likely in most workplaces, this type represents only a minority of people. Other types are more prosocial: they are largely selfish but may also care to varying degrees about others – co-workers and the organization – and, to some degree, trust, trustworthy and reciprocate in their interactions with them. These types are termed here civil, decent and dedicated. Other types, such as nasty ones, are not discussed in this paper. The central argument of the paper is that organization structure should be matched to the types of employees in the organization; an imperfect fit results in suboptimal performance. The paper further argues that properly structured organizations composed of just (purely) selfish employees perform less well than properly structured organizations composed of employees who are more prosocial, so that organizations that can attract, select and retain such employees will be more successful than organizations that do not do so. The importance of prosociality for performance increases with the complexity of employees’ tasks, which compounds the difficulty of controlling their actions. Finally, in organizations with employees who have heterogeneous preferences the structure may fluctuate and gravitate toward a match with the most numerous type if safeguards can be implemented to prevent exploitation by just-selfish employees. The rest of the paper is organized as follows. The next section introduces the key concepts, preferences and organization structure. The following section presents the core arguments and propositions. The final section offers conclusions and ideas for further research. 2. Preferences and organization structure This section elaborates on the concept of preferences as developed in the behavioral and experimental economics literature, with reference to behavior in the workplace. Using experimental findings, I propose four profiles of preferences or types of employees that represent important theoretical prototypes and have empirical relevance. Organization structure is also defined, emphasizing three components: decision-making, incentives and monitoringperformance evaluation. 2.1. Employee preferences Human nature in general and the motivation of employees in particular are characterized in various ways in different disciplines. In the traditional economics literature, including much of organizational economics, employees are viewed parsimoniously as self-interested individuals who care about their income, effort level, and (rarely) various features of their jobs. In organizational behavior and industrial/organizational psychology, employees are viewed as complex and diverse humans who possess a multitude of motives and exhibit a broad range of behaviors. In organization theory, there is a diversity of characterizations of employees, some closer to the model in economics and others closer to models in organization behavior; often the emphasis is laid on cognitive features of decision-making, which are outside the scope of this paper. Behavioral scholars, particularly in economics, use the term ‘preferences’ to describe essential aspects of human nature and values. Preferences are commonly classified as self-regarding, other-regarding and process-regarding or social preferences (BenNer and Putterman, 1998; Charness and Rabin, 2002; Fehr and Fischbacher, 2002). Self-regarding preferences concern a person’s own self-interest in all matters monetary and non-monetary, including income, job characteristics and other factors that directly affect the individual’s well-being. Other-regarding preferences concern the well-being of others (sometimes referred to as altruism); these complement self-regarding. Social, prosocial or processregarding preferences include trusting and trustworthiness, fairness and equity, reciprocity, honesty and more. An individual lacking process-regarding preferences is one who never trusts and is never trustworthy (unless there is an expected gain from reputation), does not consider fairness when making decisions, lies when it is expedient to do so, etc. The large experimental literature demonstrates that although self-regarding preferences are very strong, otherand processregarding preferences are not negligible, and in contexts such as the workplace their influence on behavior is powerful (Frohlich and Oppenheimer, 2006). Two recent meta-analyses of the common dictator and trust games provide important evidence in support of the summary above. In the dictator game, each subject is given an endowment, typically $10, which she can keep to herself or share with another person alternatively described as another subject in the experiment, a certain charitable organization, etc. In the great majority of implementations of this experiment, like in most economic experiments, subjects make decisions in complete anonymity. Engel (2011), summarizing the results of 616 experiments, finds that on average subjects give away 28% of their endowment. Author's personal copy A. Ben-Ner / The Journal of Socio-Economics 46 (2013) 87– 96 89 In the trust game, a subject in the role of trustor can keep to herself the initial endowment (frequently $10) or share it with another person in the role of trustee who receives a multiple of what the trustor sent him; for example, if the trustor sends $5 and the multiplier is three then the trustee receives $15. Now the receivertrustee decides how much to send back, if anything, to the initial sender-trustor. Johnson and Mislin (2011), in their analysis of 143 trust experiments, find that trustors sent on average 49% of their endowment and trustees returned on average 37% of what they received. The strength of individuals’ otherand process-regarding depends on the relationship between the identities of the partners to an interaction. In the workplace context, how an employee will treat others depends on whether they are co-workers, supervisors, management, shareholders, the organization as a whole or other stakeholders. Generally, individuals act more favorably toward those who are similar to them with respect to an important identity than toward those who are dissimilar.1 Virtually all experimental research, including the dictator game and trust game meta-analyses cited above, finds that there is considerable heterogeneity in the behavior of subjects.2 Generally, the common explanation for heterogeneity is that for diverse reasons (from evolutionary pressures, random mutations, upbringing and social influences to life experiences) the strength of a given preference may vary across individuals, so for example some individuals are more selfish or trusting or honest than others (Andreoni and Miller, 2002; Fehr, 2009; Ben-Ner and Halldorsson, 2010; Ben-Ner and Kramer, 2011).3 Behavioral economists have begun to explore classification of individuals into “types” in order to capture regularities in the distribution of various preferences and to summarize continuous multi-dimensional preferences in a limited number of recognizable types, which is the purpose and benefit of most systems of classification (Fryer and Jackson, 2003). Casari and Plott (2003) use versions of the public goods game to classify individuals as “spiteful” or “altruistic;” Kurzban and Houser (2005) use multiple rounds of a public-goods game to classify individuals into “reciprocators,” “free riders” and “cooperators;” Burlando and Guala (2005), on the basis of multiple sessions of public goods and other instruments, classify 92 student subjects as “free riders” (25%), “reciprocators” (35%), “cooperators” (17%) and “noisy” (18%); Rustagi et al. (2010), on the basis of behavior in versions of the public good game, classify 679 Ethiopian forest workers into “conditional cooperators” (34%), “weak conditional cooperators” (12%), “free riders” (11%), “hump-shaped” (3%), “altruists” (2%) and “other” (38%). To investigate the relationship between preferences and organization structure I construct four types that bear clear relationship to the types discussed above but are more directly relevant to the context of the workplace. This classification does not describe all 1 For the effect of identity on other regarding, see Buchan et al. (2006) and Ben-Ner et al. (2009) in diverse contexts, including the workplace; for the effect on trusting and trustworthiness, see Güth et al. (2008) and Hargreaves and Zizzo (2009). 2 As Casari and Plott (2003) put it, “The classical model of identical selfish agents does not capture the data as well as a model with heterogeneous and linear otherregarding preferences. Altruism and especially potentially dysfunctional behavior, such as spite and mistakes, play important positive roles.” Henrich et al. (2010) document similar tendencies in non-Western subject pools. 3 Most economic experiments are carried out with student subjects. Experiments with non-student samples obtain similar findings, often exhibiting greater otherregarding and stronger process-regarding preferences. For instance, Hoffman and Morgan (2010) find that business people in domain-trading and online pornography industries “were significantly more altruistic, more trusting, reciprocate more, lie less, and respond differently to shame then students” at Berkley. Field studies generally produce findings that are within the range of results obtained in the lab (Falk et al., 2013), although some argue that experimental findings overstate the degree of otherand process-regarding (Levitt and List, 2007). individuals; for example, it does not include “nasty” types or those who wish harm onto others, although these clearly exist. The classification is grounded in behaviors expressed in the dictator and trust games described earlier, as well as in reciprocity games.4 The behaviors expressed in these two games reflect a mixture of attitudes toward conditional and unconditional cooperation, equity and distributive fairness, reciprocity, trusting, and trustworthiness. They are well correlated with behaviors captured in public goods and ultimatum games, among other methods of elicitation of prosocial preferences (Charness and Rabin, 2002; Kamas and Preston, 2011). This classification is intended as a concrete illustration of a parsimonious description of important tendencies that manifest themselves in behaviors that determine the productivity and profitability of organizations (Williamson, 1985; Zaheer et al., 1998; Baron and Kreps, 1999). The four types are “just selfish,” “civil,” “decent” and “dedicated.” The types are described in general terms rather than specific numerical values with reference to the games described above. The P1 type employee is just selfish, without any otheror process-regarding preferences. P2, is as selfish as P1 but reciprocates, trusts and is trustworthy to a limited extent – basically, a civil employee. The third type, P3, possesses approximately average preferences: she cares mostly about herself and a little about her coworkers but not about management or the organization at large, and has an average inclination to reciprocate, to trust and to being trustworthy. P3 is what one may term a decent average employee, one who cares about others but not to the point of self-effacement, one who reciprocates but imperfectly so, one who trusts most of his or her endowment in an anonymous interaction when the promise of gain – but also the risk of loss – is high, and one who is trustworthy in that she or he does not take advantage of trust placed implicitly in her or him. The fourth profile, P4, the dedicated employee, is a version of the employee whom Whyte (1956) famously termed “the organization man:” he or she is loyal and dedicated to the organization and pursues its interests as specified by management, and is reliably reciprocal, trusting and trustworthy.5 The four types mat be naturally ordered in terms of the “prosociality” of their preferences, so that P1 is least prosocial, P2 more prosocial (with the addition of process-regarding preferences), P3 even more prosocial (with stronger process-regarding preferences), and finally P4 being the most prosocial. This ordering depends on how the change in the orientation of other-regarding preferences from co-workers to organization is regarded, but it probably makes common sense to call P4 more prosocial than P3. How are these types distributed among employees? There is no information from experiments in which subjects participated in all these games and roles to be able to characterize these types distribution in the population or even particular samples. But we can bound the distribution of type P1 individuals: approximately 36% of subjects are purely selfish, but only about 6% do not trust at all, 20% are not at all trustworthy and about 15% do not reciprocate 4 One version of the reciprocity game is a two-stage dictator game, whereby a receiver in a stage one dictator game plays a stage two one-shot dictator game with his or her sender-dictator. The dictator in the first stage was not aware that there will be a second-stage game where the receiver will have a chance to reciprocate (Diekmann, 2004). The receiver’s behavior is analyzed for the degree of reciprocity in terms of how much he or she send from a new endowment, so the more he or she sends, the greater the tendency to reciprocate is. 5 The four types do not map entirely on the classifications discussed earlier, although “just selfish” is close to “free rider,” “civil” is closely related to “weak cooperator,” “decent” is related to “conditional cooperator” and “dedicated” is a combination or overlap between “altruist” and “conditional cooperator.” As noted in the text, these four types these do not describe all possible individuals. In particular, I do not discuss many possible variations that arise from other-regarding toward groups other than co-workers and the entire organization and preferences that reflect a desire to harm others. Author's personal copy 90 A. Ben-Ner / The Journal of Socio-Economics 46 (2013) 87– 96 at all.6 So, by these estimates, at most 6% of subjects can be justselfish without any process-regarding preferences, P1-types.7 The other three types are likely to represent larger minorities than that, so an obvious case exists for studying organization structure that fits such types of employees. 2.2. Organization structure Organizations employ various measures to engage and direct employees’ efforts toward the promotion of organizational goals. Collectively these measures constitute the organization structure, also referred to in various literatures as organization design or architecture, and human resources management system. In the interest of concreteness and focus I will examine organization structure at the unit level in a profit-maximizing firm that employs a business strategy and production technology that imply considerable task complexity as well as reciprocal interdependence among the tasks of employees. Three organizational/management practices at the unit or workplace level – allocation of decision-making, incentives and monitoring – are considered central by organization theorists (e.g., Williamson, 1975; Ouchi, 1979; Rivkin and Siggelkow, 2003) and economists (e.g., Prendergast, 2002; Brickley et al., 2009; Lazear and Oyer, 2008). Allocation of decision-making concerns the degree of delegation of decision-making from managers to individual employees or teams; it determines the degree of autonomy and discretion afforded by management to employees. Incentives represent rewards conditioned on performance; they may target individuals or groups and may be based on an employee’s absolute performance or performance relative to other employees (Lazear and Rosen, 1981). Monitoring is observation of performance (results) or effort correlated with it; it is performed by supervisors or peers and supplies the basis for performance evaluation that is used for feedback and determination of incentives. These three practices are related to each other through complementarity and substitution: incentives and monitoring complement the proper discharge of decision-making, some forms of monitoring substitute for some incentives while others complement them, etc. (Ichniowski et al., 1997; Prendergast, 2002; DeVaro and Kurtulus, 2010; Ennen and Richter, 2010; Ben-Ner et al., 2012). Organization structure can be said to be optimally designed if it minimizes the cost of engaging the organization’s workforce to produce a certain level of output. An alternative but equivalent formulation is that given a certain cost of operating an organization structure, the output that can be produced with the organization’s workforce is maximized. The cost of the optimal organization structure is a function of the ease or difficulty of engaging employees in the pursuit of organizational goals. This, in turn, depends on the preferences of employees, which is the principal concern of this paper and to which we turn shortly, as well as on the nature of employees’ tasks, which we discuss in the following section. 3. Matching organization structure to employee preferences This section examines design of organization structure best suited for a firm with a homogeneous workforce consisting, 6 These figures are derived from Engel (2011), Ben-Ner and Putterman (2009), Johnson and Mislin (2011) and Diekmann (2004). 7 Even if we were to ignore process-regarding preferences altogether and describe P1 only in terms of giving in the dictator game, the average figure from the metaanalysis cited above, of 36%, can hardly be characterized as anything other than a minority, in contrast with the 100% assumed in standard economic organizational analysis. alternately, of P1, P2, P3 or P4 employees. It is instructive to analyze the optimal structure of firms that have just one type of employees because through the processes of selection and turnover, both voluntary and forced, a certain degree of homogeneity is achieved. This is especially true of units within organizations, such as production or marketing, which are the level to which the analysis of this paper is directed. Most organizations rely on instruments such as personality traits (Guion and Gottier, 1965), honesty (Sackett and Wanek, 1996), cooperation and cognitive ability (Salgado et al., 2003). Furthermore, hiring is done by existing employees who tend to hire employees like themselves, perpetuating the existing workforce’s main characteristics. Such an analysis also provides a benchmark for investigating organization design in the more complicated situation of employee heterogeneity, which concludes this section. 3.1. Organization structure for a workforce of just selfish (P1)