Learning from Peers: Knowledge Transfer and Sales Force Productivity Growth

We study how peers impact worker productivity growth among salespeople in the cosmetics department of a department store. We first exploit a shift assignment policy that creates exogenous variation in salespersons’ peers each week to identify and quantify sources of worker learning. We find that peer-based learning is more important than learning-by-doing for individuals, and find no evidence of forgetting. Working with high-ability peers substantially increases the long-term productivity growth of new salespeople. We then examine possible mechanisms behind peer-based learning by exploiting the multiple collocated firms in our setting that sell products with different task difficulties, and compensate their sales force using either team-based or individual-based compensation systems. The variation in incentives to compete and cooperate within and across firm boundaries, combined with variation in sales difficulty for different product classes, allows us to suggest two mechanisms behind peer-based learning: observing successful sales techniques of peers and direct teaching. Our paper advocates the importance of learning from one another in the workplace, and suggests that individual peer-based learning is a foundation of both organizational learning curves and knowledge spillovers across firms. 1 Tat Y. Chan is Associate Professor of Marketing and Lamar Pierce is Associate Professor of Strategy at Olin Business School, Washington University in St. Louis. Jia Li is Assistant Professor of Marketing at Krannert School of Management, Purdue University. Authors can be reached at [email protected], [email protected], or [email protected]. A well-developed literature in economics and management studies learning curves at the group and organizational level. The primary focus of the research has been on identifying learning-by-doing through cost decreases with cumulative production (Arrow 1962; Argote and Epple 1990; Thompson 2001), as well as organizational forgetting in periods of low production (Darr et al. 1995; Benkard 2000; Thompson 2007). We have little evidence, however, on underlying individual mechanisms that generate this phenomenon (Adler and Clark 1991). While individual learning-by-doing through experimentation and observation of own outcomes is a common explanation for productivity growth (Blume and Franco 2007), social learning theory suggests peers may also play a critical role (Ellison and Fudenberg 1993; Bala and Goyal 1998; Young 2009). This peer-based learning can involve both observing peers’ practices and active teaching by peers (Arrow 1994). Despite evidence of peer-based learning in education (Sacerdote 2001), crime (Bayer et al. 2009), health care (KC et al. 2013), and economic development (Bandiera and Rasul 2006), direct evidence in commercial and industrial settings is rare. In this paper we dissect the separate roles of peer-based learning and individual learning-by-doing (and forgetting) in the productivity change of salespeople, with learning from peers as our primary focus. We extend this focus to study peer-based learning across firm boundaries, motivated by more macro-level evidence of knowledge transfer across firms (Argote et al. 1990), store locations (Darr et al. 1995), and work shifts (Epple et al. 1996). Exploiting firm boundaries also helps disentangle the mechanisms that might drive peer-based learning. Our empirical setting is four years of individual sales data for 92 salespeople working at 11 co-located cosmetics counters in a department store. This research setting matches our research agenda because multiple cosmetics manufacturers employ salespeople at counters on the same retail floor, allowing us to observe peer-based learning both within and across firms. More importantly, the difficult task of selling cosmetics products with often unobservable quality makes learning from peers critical for productivity growth. We conceptualize the sales task in the cosmetics counters as a stochastic production function with parameters linking output (sales) with input (selling activities) variables. Salespeople have heterogeneous