Measuring Tacit Knowledge

| Peter Klein |

The concept of tacit knowledge — knowledge that is difficult or impossible to parameterize, or to express in words or numbers — is central to organization theory, as well as philosophy (Polanyi) and social theory more generally (Hayek). Most of the research literature on tacit knowledge is conceptual and theoretical, such as Hayek’s famous “Use of Knowledge in Society” (1945) or more recent pieces like Jensen and Meckling’s “Specific and General Knowledge, and Organizational Structure” (1992). Empirical studies of tacit knowledge are rare, which is not surprising given the idiosyncratic, personal, subjective, and often ephemeral nature of such knowledge.

An interesting new NBER paper by David Chan estimates the effects of tacit knowledge using matched pairs of physician trainees with similar levels of explicit knowledge but different levels of experience and hence accumulated know-how. The hospital setting allows for some clever tricks, e.g., exogenous sorting into occupational roles by experience, rather than ability. Measuring outcomes via spending is problematic to me, though standard in the medical economics and management literatures. Check it out:

Uncertainty, Tacit Knowledge, and Practice Variation: Evidence from Physicians in Training
David C. Chan, Jr
NBER Working Paper No. 21855, January 2016

Studying physicians in training, I investigate how uncertainty and tacit knowledge may give rise to significant practice variation. Consistent with tacit knowledge accruing only with experience, and empirically exploiting a discontinuity in the formation of teams, experience relative to a peer substantially increases the size of variation attributable to the physician trainees. Among the same physician trainees, convergence occurs for patients on services driven by specialists, where there is arguably more explicit knowledge, but not on the general medicine service. This difference is unexplained by formally coded patient information. In contrast, rich physician characteristics correlated with preferences and ability, and quasi-random assignments to high- or low-spending supervising physicians explain little if any variation.