Review status
This summary is based on the full article text.
The DOI and source URL have been verified through the official SAGE journal page.
The article is a practitioner-oriented conceptual article published in California Management Review. It synthesizes a probability-based model of innovation incentives, experimental evidence, corporate governance research, bankruptcy law evidence, and scientific funding research. It does not present a single new empirical dataset in the article itself; instead, it translates Manso’s broader research program into managerial implications.
Research question
How should organizations design incentives when they want employees and managers to innovate rather than only improve routine productivity?
More specifically, the article asks why standard pay-for-performance schemes may be suitable for exploitation but poorly suited for exploration, and what kinds of contracts, organizational cultures, governance structures, and policy environments better support innovation.
Hypotheses
Not specified as a formal hypothesis-testing study.
The article develops a clear theoretical argument: incentives that motivate exploitation are fundamentally different from incentives that motivate exploration. For innovation, incentive systems should tolerate early failure, reward long-term success, provide job security, and give regular feedback.
Method
The article is a conceptual and practitioner-oriented management article.
Manso develops the argument using a model based on the bandit problem from probability theory. In this model, an agent chooses between a conventional method with a known probability of success and a new method with an unknown probability of success. The new method is more likely to fail early, but it may reveal information that creates larger long-term payoffs.
The article uses this exploration-exploitation logic to compare two kinds of incentive systems:
- incentives for exploitation, where the goal is reliable execution of known methods;
- incentives for exploration, where the goal is experimentation, learning, and discovery of better methods.
The article also synthesizes several streams of evidence.
First, it reviews productivity evidence showing that pay-for-performance can improve routine work. Examples include agricultural workers in the Philippines, Safelite Glass windshield installers, Canadian tree planters, professional golf players, and laboratory typing tasks.
Second, it summarizes Manso’s theoretical model of innovation incentives, originally developed in related academic work. The model shows why innovation-oriented incentives should tolerate early failure and reward long-term success.
Third, it summarizes laboratory experimental evidence from a computerized lemonade stand task with 379 participants. Participants worked under fixed-wage, standard pay-for-performance, or exploration-oriented compensation schemes.
Fourth, it summarizes evidence from corporate governance research, including studies on independent boards, shareholder litigation, and innovation outcomes.
Fifth, it summarizes evidence from scientific research funding by comparing the Howard Hughes Medical Institute Investigator Program with National Institutes of Health funding.
The article does not report a new original regression analysis inside the article. Its contribution is to integrate theory and prior empirical evidence into a practical framework for designing incentives for innovation.
Results / key findings
The article’s central finding is that innovation requires incentive systems that differ from those used for routine productivity.
Pay-for-performance can be effective for routine, repetitive work. The article reviews evidence showing that workers often exert more effort when compensation is more closely tied to output. Examples include piecework in agriculture, Safelite Glass installers switching from fixed wages to piece-rate pay, tree planters, professional golfers, and laboratory typing tasks. These examples support the idea that standard incentives can increase effort when tasks are familiar and outputs are predictable.
However, Manso argues that innovation is different. Innovation requires exploration of new, untested approaches. Exploration often fails early, takes time, and produces value only after learning has occurred. Standard pay-for-performance can therefore discourage innovation because it penalizes the early failures that are often necessary for discovery.
The article’s probability-based model shows why time horizon matters. If the agent has only a short horizon, exploration is unattractive because early failure dominates the payoff. As the time horizon expands, experimentation becomes more valuable because the agent has more opportunities to use what is learned from trying new methods.
The article identifies four core elements of innovation-oriented incentives.
First, organizations should tolerate early failure. If failure is punished immediately, employees and managers have strong incentives to choose safe, conventional methods.
Second, organizations should reward long-term success. Innovation may look bad at first but valuable later. Incentives should therefore be tied to longer-term performance rather than only short-term results.
Third, organizations should provide job security. The threat of termination can prevent shirking, but it also discourages exploration. If employees believe failed experiments will cost them their jobs, they are more likely to exploit known methods.
Fourth, organizations should provide regular informational feedback. For exploitation, feedback may be less important because the agent is expected to repeat known actions. For exploration, feedback helps the agent learn, adjust, and improve new methods. Manso emphasizes that this kind of feedback should be informational, not simply a disguised performance review tied to punishment or reward.
The article translates these ideas into executive compensation. In the context of senior management, innovation-oriented incentives may include stock options with long vesting periods, option repricing, golden parachutes, and managerial entrenchment. These mechanisms are often criticized in corporate governance debates, but Manso argues that they can support innovation by protecting managers from premature punishment while they pursue risky long-term projects.
The laboratory experiment with the computerized lemonade stand provides concrete evidence. Participants under exploration contracts found the best location for the lemonade stand 80% of the time. This was higher than the fixed-wage group at 60% and the standard pay-for-performance group at 40%. The exploration contract worked because participants could fail at no cost in the first half of the experiment and were rewarded based on profits in the second half.
The experiment also showed that exploration contracts encouraged more systematic learning. Under fixed wages, 55% of participants regularly used the provided table to track business decisions and profits. Under exploration contracts, 82% used the table. This suggests that exploration incentives did not merely produce random experimentation; they encouraged more organized search.
The pay-for-performance group behaved more conservatively. During the first ten periods of the experiment, 80% of participants under the exploration contract left the original location to search for a better one. Under the pay-for-performance contract, only 50% relocated. This supports the article’s argument that standard performance incentives can push people toward fine-tuning known methods rather than exploring uncertain alternatives.
The termination experiment provides further support. Two groups received exploration contracts but faced early termination if first-half profits fell below a threshold. One group also received a golden parachute payment if the experiment ended early. In the golden parachute group, 65% discovered the best location, compared with 45% in the termination group without golden parachutes. This suggests that protection against downside risk can restore some willingness to take chances.
The article also discusses corporate governance evidence. In research on independent boards, companies forced by stock exchange and Sarbanes-Oxley requirements to adopt more independent boards increased patenting by 20% to 30%. Patent citations rose by 40% to 60%. However, the increase was concentrated in familiar technology areas rather than more speculative exploratory innovation. This suggests that stronger board oversight may increase measurable productivity while pushing managers toward safer incremental innovation.
The shareholder litigation evidence points in a similar direction. Manso, Chen Lin, and Sibo Liu studied the adoption of universal demand laws in 23 U.S. states from 1989 to 2005 using a sample of 4,506 U.S. public companies from 1976 to 2006. These laws reduced the threat of derivative lawsuits. After adoption, companies increased R&D spending by about 11% of the sample mean relative to firms in states without such laws. Average citations per patent rose, more patents became highly cited, and more patents appeared in experimental and unfamiliar technology areas. The article interprets this as evidence that litigation risk can discourage exploration.
The article also connects incentive design to bankruptcy law. Innovation-friendly bankruptcy rules can encourage entrepreneurs to try again after failure. The European Council’s 2000 European Charter for Small Enterprises explicitly framed failure as part of responsible initiative and risk-taking. By contrast, the article notes that the United States adopted more creditor-friendly bankruptcy rules in 2005, making debt discharge harder for failed entrepreneurs.
The scientific research section extends the logic beyond business. Manso, Pierre Azoulay, and Joshua Graff Zivin compared HHMI investigators with similar NIH-funded scientists. HHMI funding is more tolerant of early failure, offers five-year cycles, provides feedback from renowned scientists, and funds people rather than specific projects. NIH grants are typically shorter, tied to specific projects, and more vulnerable to nonrenewal after failure.
The comparison shows that HHMI investigators produced more high-impact research. Participation in the HHMI program increased overall publication output by 39%. For publications in the top percentile of citations, the increase was 96%. At the same time, HHMI researchers also produced more low-impact work: compared with the control group, 35% more of their articles were cited less frequently than their own least-cited pre-appointment publication. Manso interprets this as evidence of more exploration: more breakthroughs, but also more flops.
The table on page 12 summarizes the NIH-HHMI contrast. NIH R01 grants involve three- to five-year funding, project-specific funding, more similar first and later reviews, funds ending upon nonrenewal, and some feedback. HHMI offers five-year funding, a more forgiving first review, a two-year phase-down after nonrenewal, feedback from renowned scientists, and a “people, not projects” funding logic. This table visually reinforces the article’s broader argument that long horizons, tolerance for failure, and feedback support exploration.
Overall, the article argues that organizations cannot demand innovation while punishing the process that produces it. If leaders want exploration, they must build incentive systems and cultures that allow failure, learning, patience, and long-term value creation.
Practical implications
For managers, the article’s main implication is that incentive systems should match the type of work.
If the goal is routine productivity, standard pay-for-performance may work well. For tasks where employees already know the right method and the main issue is effort, short-term performance incentives can increase output.
If the goal is innovation, standard pay-for-performance can backfire. When employees are rewarded for immediate results and punished for early failure, they may avoid experimentation and rely on familiar methods. This can produce short-term performance but weaken long-term renewal.
Managers should therefore distinguish between exploitation roles and exploration roles. Exploitation roles need discipline, efficiency, and reliable execution. Exploration roles need slack, patience, feedback, and protection from premature punishment.
For executive compensation, the article suggests that some controversial governance mechanisms may have innovation value. Long-vesting stock options, option repricing, golden parachutes, and managerial entrenchment can be abused, but they can also protect managers while they pursue risky long-term strategies. The key is not to use these mechanisms blindly, but to align them with genuine innovation goals.
For R&D teams and lower-level employees, culture may matter more than formal contracts. Employees need to believe that experimentation is truly valued. If leaders say they want innovation but punish failed attempts, employees will learn to avoid risk. Credibility comes from repeated organizational behavior, not slogans.
Feedback should be designed carefully. For innovation, feedback should help employees adjust and learn. It should not simply become a short-term evaluation tool that punishes experiments before they have time to develop.
The article is also useful for boards. Directors need to balance accountability and patience. Strong oversight can increase productivity, but if it becomes too focused on short-term results, it may discourage exploratory innovation. Boards should ask whether they are pushing management toward safe incremental patents or supporting more uncertain long-term opportunities.
For practitioners, useful diagnostic questions include:
- Is the task mainly exploitation or exploration?
- Are employees punished for early failure even when they are pursuing reasonable experiments?
- Are incentives tied too strongly to quarterly or annual performance?
- Do managers have enough time to learn from new initiatives before being evaluated?
- Does the organization provide feedback that helps people adjust, or feedback that mainly threatens them?
- Are failed experiments treated as learning opportunities or career damage?
- Do governance systems reward only predictable productivity, or also long-term innovation?
- Are compensation systems encouraging safe incremental work while discouraging breakthrough attempts?
Theoretical implications
The article contributes to innovation management by clarifying why innovation incentives differ from productivity incentives.
Standard agency theory often focuses on the problem of effort: the principal wants the agent to work rather than shirk. Manso adds a different problem: the principal may want the agent to explore rather than exploit. This changes the incentive logic because exploration requires risk-taking, experimentation, and learning from failure.
The article also connects agency theory with the exploration-exploitation problem. Exploitation involves repeating known methods and improving predictable performance. Exploration involves trying unknown methods whose value becomes clear only over time. Incentive systems that work well for exploitation may be poorly suited to exploration.
The article contributes to corporate entrepreneurship by showing why internal innovation depends on both formal incentives and informal culture. Employees may not innovate if they distrust organizational promises about tolerating failure. This makes credibility and consistency central to innovation management.
The corporate governance discussion adds theoretical tension. Stronger governance and shareholder accountability may reduce agency problems, but they may also increase short-termism and risk aversion. This suggests that governance systems involve tradeoffs: mechanisms that discipline managers can also discourage exploration.
The article also extends the incentive logic beyond firms. The comparison between HHMI and NIH shows that similar principles apply in scientific research. Funding systems that support long-term exploration, tolerate failure, and give strong feedback can produce more breakthrough output, but also more failed attempts.
Limitations
The article is a conceptual synthesis rather than a single new empirical study.
Some evidence is drawn from prior studies with different designs, samples, and contexts. This makes the article broad and useful, but it also means the findings should not be interpreted as one unified empirical test.
The bandit-problem model simplifies real organizational innovation. Actual innovation involves politics, teams, interdependence, market uncertainty, resource constraints, organizational routines, and external competition.
The article emphasizes incentives, but innovation also depends on capabilities, knowledge, leadership, psychological safety, customer insight, technology access, and organizational structure.
Some recommended mechanisms, such as managerial entrenchment, option repricing, and golden parachutes, can create agency problems if poorly governed. The article acknowledges that these tools can be abused, but it focuses more on their innovation benefits than on detailed safeguards against misuse.
The corporate governance evidence is nuanced. Independent boards increased patenting and citations, but not necessarily exploratory innovation. This suggests that more innovation output does not always mean more radical or exploratory innovation.
The laboratory experiment provides clean evidence but uses a simplified lemonade-stand task. The findings may not fully generalize to complex corporate innovation projects that involve teams, long development cycles, and strategic uncertainty.
The article does not provide a detailed implementation framework for deciding exactly how much failure tolerance, job security, or long-term compensation is optimal in different firms.
Future research
Future research could examine how organizations can distinguish productive failure from careless or low-effort failure.
Researchers could test how much job security is needed to encourage exploration without creating complacency or weak accountability.
Future studies could examine how innovation incentives should differ across levels of the organization, from frontline employees to middle managers to executives.
Another useful direction would be to study how boards can support long-term innovation while still preventing managerial self-dealing and poor governance.
Researchers could examine whether different types of innovation need different incentive systems. Incremental process improvements, new product development, business model innovation, scientific research, and radical technology bets may require different degrees of failure tolerance and long-term reward.
Future research could also study how feedback should be delivered in exploratory work. The article argues that feedback should be informational, not punitive, but more evidence is needed on what feedback frequency, content, and source work best.
Another research direction would be to compare incentive systems across industries with different innovation cycles, such as software, pharmaceuticals, automotive, consumer goods, energy, and professional services.
Finally, future research could examine how artificial intelligence and data analytics affect innovation incentives. More measurement may improve feedback, but it may also increase short-term monitoring and discourage exploration if used poorly.