An exploratory analysis of learning from peers: Radial vs. nonradial efficiency measures and convex vs. nonconvex technologies

This work investigates to which extent the known substantial differences between technical efficiencies on convex and nonconvex technologies translate into different learning possibilities. We also study whether radial and nonradial efficiency measures lead to a different learning experience. To our knowledge, these questions have never been investigated. Our empirical research is guided by three working hypotheses regarding how the analysis of peers facilitates learning by comparing on the one hand convex versus nonconvex technologies, and on the other hand radial versus nonradial efficiency measures. These working hypotheses are investigated using three distinct metrics: peer count, peer similarity, and peer dominance. We employ five existing secondary data sets and one large sample of more than 10,000 observations on Belgian traffic control centres in an effort to refute our three working hypotheses using these three metrics. Anticipating our conclusion, the combination of the logical, the statistical, and the managerial arguments against convexity is rather overwhelming in our data and we think that convexity is an axiom that should be scrutinized in all these three respects in all future methodological innovations as well as in empirical applications.