The following is the fourth in a series of guest posts from job market candidates working in Institutional and Organizational Economics (check out the first, second, and third). Watch for the rest of the series over the next couple weeks, and think about interviewing one of these fine students if you have an opening. (-PLW)
guest post by Teck Yong Tang
When imagining the work culture at high-performing startups or "tech firms", one usually has the image of little monitoring with the flexibility to choose one's working hours and locations. A study by LRN found that companies with "high freedom index" are indeed associated with more innovation and better long-term success.
Why do people perform better under more freedom? There is an extensive work in the psychology and management literature studying this phenomenon, and they generally attribute the positive effects from giving workers freedom at work to their intrinsic motivation. But if freedom provision always improves performances, why do some firms still choose to monitor their workers closely?
My job market paper argues that workers also have extrinsic motivation (i.e. pecuniary incentives) for freedom at work, and it offers an explanation to why different firms have different monitoring cultures. The issues stem from the difficulty in incentivizing workers to innovate when the firm cannot credibly promise rewards for workers who successfully improve upon the existing capability in the firm. This is a common feature as innovations are often impossible to describe before they are realized, so it is difficult to contract on them.
The basic intuition is the following. A worker faces a ratchet effect of innovating when being watched. If the worker uncovers a more efficient work method, his firm will know about it. His firm will then rationally respond by raising the worker's performance requirement in future without increasing his pay. Anticipating this, the worker never tries to innovate. If the worker works in a culture with little monitoring instead, then any discovery becomes the worker's private information. If the worker uncovers a more efficient work method, his firm is compelled to give him information rent if it wants to tap on his privately known discovery. This potential information rent in turn channels back as the worker's incentive to innovate. Perhaps counter-intuitively, the firm can benefit from encouraging asymmetric information against itself.
Why do different monitoring cultures exist? Although freedom provision encourages innovation, the asymmetric information about the innovation outcome is costly to the firm. The screening of workers with varying (post-innovation) capabilities leads to inefficiency in the use of the innovations, and the firm also shares part of the innovation surplus with the innovating worker as information rent. Whether freedom provision is worthwhile thus depends on whether the gains and likelihood of innovation are sufficiently high. For example, the founder of a new startup is unlikely to fully understand the operational needs of her startup. Hence, the expected gains from giving freedom to encourage her employees to be innovative and explore for more efficient operating procedures are high. On the other hand, a neighborhood restaurant owner can readily mimic best practices of other similar restaurants; it is also unlikely that her employees can significantly improve on these existing practices even if they try. Hence the restaurant owner would rather monitor her employees closely and give up on any possibility of improving upon existing practices.
What about the moral hazard problem? While a tight monitoring culture allows the firm to observe its worker's innovations, it remains difficult to enforce production effort, which is often non-verifiable. Thus a firm always requires a pay-for-performance scheme to mitigate the moral hazard problem. That a principal can gain from having less information about the agent's actions has been noted in various contexts (e.g. Crémer, 1995 ; Aghion-Tirole, 1997), but the impact of the principal's level of oversight on the optimal incentive structure has been largely unexplored. I show that the optimal pay-for-performance structures under different monitoring cultures are different. A culture of tight monitoring induces the worker to constantly exploit the existing work method in the firm. Since the worker works under the same "technology" every period, the pay-for-performance structure under a tight monitoring culture would be constant across time. On the other hand, a "freedom-at-work" culture is accompanied by low incentives in the early period and then dispersions in the power of incentives in the later period. This dispersion post-innovation is part of the "menu" used by the firm to screen the workers with private information about their innovation outcomes. Pre-innovation, the incentive for good performance is low because, asHolmström (1989) points out, innovation attempts are risky and tend to produce short-term low-quality performances; hence if the early incentive is too powerful, the worker will be deterred from taking on innovative ventures.
Implications. These features of the pay-for-performance structures resonate well with the observation that working under less oversight is often associated with varying degrees of changes in wage structures and job scopes over time (e.g. promotion opportunities), while working in tightly monitored environments is associated with constant wage structures and limited career progressions. Holmström's classic career-concern model provided an explanation to the stylized observation that incentives tend to be low in the early part of one's career. My paper suggests that early low incentives also serves to encourage innovations. (I should note that Manso also showed in a very nice paper that firms must exhibit tolerance to early failure in order motivate their workers to innovate, but the mechanism there is different from here.) Moreover, the early low incentives also implies that the firm-worker relationship starts off with low levels of production. That economic relationships often "start small" is typically attributed to reputation-building (e.g. Gosh and Ray, 1996; Watson 1999, 2002; Halac, 2012) where there is ex-ante asymmetric information about the other person's type, so the parties’ stakes at the start of the relationship are small to gradually resolve the asymmetric information. In contrast, the rationale for starting small here is to encourage the worker to accrue private information (about innovations) rather than resolving it.
Sketch of model. I conclude with a description of some key features of my model. A risk-neutral agent produces a binary output (high or low) each period for a risk-neutral principal for two periods. The probability of getting the high output is the agent's effort chosen from the continuum [0,1]. The agent chooses between using an old technology (o) or a new technology (n). The effort cost for o is commonly known, whereas the effort cost for n is ex-ante unknown to both parties, but they share a common prior about it. I assume that n, which represents the innovative venture, is in expectation more cost-inefficient than o (thus reflecting its riskiness), but it has some probability to be more cost-efficient than o. Hence using n in the first period creates an option value for the second period: if n turns out to be more cost-efficient, the worker can continue to use n in the second period; if not, the worker can revert to using o. The firm can only write spot contracts on the output. Two cultures of monitoring are considered. Under the tight-monitoring culture, the firm also learns the effort cost of n if it is realized; under the freedom-at-work culture, it does not. A technical difficulty arises in the second period under the freedom-at-work culture when the principal solves an adverse selection problem (on the agent's privately known effort cost on n) with moral hazard (on the agent's effort and technology choices). The optimal menu of contracts exhibits some pooling due to a mass point of agents who failed in their innovation attempt and revert to using o in the second period. The pooling feature would imply that low-success innovations are not recognized by the principal.
Teck Yong Tan is a Ph.D candidate in Economics at Columbia University.