The Value of Knowledge: How Nth Best Solutions Affect Organizational Search

João Duarte & Thorsten Wahle Università della Svizzera italiana, Department of Economics, Institute of Management Year of Enrollment in PhD: 2014, Expected Completion: 2019 joao.paulo.duarte@usi.ch, thorsten.wahle@usi.ch State of the Art While neoclassical economic theories are theories about the optimal (or 1st best) solution, management and organization theories are, more or less explicitly, theories about the nth best solution. This idea is perhaps most explicit in Levinthal’s (1997) theory of adaptation in rugged landscapes: organizations will tend to converge only to a local peak (or an nth best solution) and fail to find the global peak (i.e., the 1st best solution). The notion that the current best solution is most likely not the optimal solution is also also implicit in other theories. Take for example the literature on practice transfer: Even when we call a solution a “best practice” (e.g., Spender & Grant, 1996; Szulanski, 1996) it is often not literally the 1st best solution but, most of the times, only the best currently known alternative, i.e., at best, a 2nd best solution. Research Gap Even though most organizational decisions are done on the basis of knowledge of nth best solutions, our models of search and learning often focus on the case where agents are endowed with knowledge of the 1st best solution (e.g., Ghemawat & Levinthal, 2008; Rivkin, 2000). If agents had access to the 1st best solution, then the best strategy for future action would be to always adopt that solution, i.e., to not search for a better solution. However, when dealing with nth best solutions, the optimal search strategy is far from obvious. For example, if the agent does not know whether the current solution is the 1st, 2nd, or 3rd best solution, but only that its payoff is already above a set performance target, or aspirational level, should the agent look for a better solution, or should it exploit its current knowledge? Theoretical Arguments Organizational decisions tend to be made on the basis of some knowledge about existing performance targets and the value of potential solutions. Therefore, agents must search and learn, while incurring the costs associated to this process. Without being able to determine how close they are to the optimal solution, we argue that agents will not make better choices, even when they know that the current solution is already above a performance target. Method I tap into these questions by conducting two separate experiments based on the n-armed bandit model. Using this setting, I reproduce search in organizational contexts by requiring subjects to handle the tradeoff between the exploitation of solutions with a known payoff, or the exploration of unknown solutions. In particular, I contrast the performance of human subjects endowed with positive knowledge about nth best solutions, with subjects who are ignorant to this knowledge. Results I find that positive knowledge about the payoff of nth best solutions (knowledge like, “Solution 1 has a payoff of 2.8”) does not improve performance, as it leads to strategies that involve high levels of exploration. However, when subjects have reason to believe that their current knowledge is already close to the optimal solution (e.g., knowledge like, “Solution 1 has a payoff of 2.8 and is the 2nd best solution”) they will tend to adopt this solution more often and, in turn, reach a better performance. Finally, I also find that, in both cases, these subjects will tend to find the optimal solution less often than subjects who are ignorant to knowledge about above-average, nth best solutions, even though they have better chances of finding the first-best solution. Jelcodes: M10,M10 THE VALUE OF KNOWLEDGE: HOW POSITIVE KNOWLEDGE ABOUT N BEST SOLUTIONS AFFECTS ORGANIZATIONAL SEARCH JOÃO DUARTE Università della Svizzera italiana Via Giuseppe Buffi, 13 CH-6904 Lugano Tel: (+41) 58 666 4000 e-mail: joao.paulo.duarte@usi.ch THORSTEN WAHLE Università della Svizzera italiana Via Giuseppe Buffi, 13 CH-6904 Lugano Tel: (+41) 58 666 4000 e-mail: thorsten.wahle@usi.ch