Decision Bias in Project Selection : Experimental Evidence from the Knapsack Problem

The selection of projects from a set of possible alternatives is a key decision in any organization. However, all entities that allocate resource budgets across several projects and initiatives face the same fundamental problem: to select the most valuable projects given a finite budget. The literature has long recognized the mathematical complexity of this knapsack problem, and has proposed efficient algorithms to solve it. Yet, the non-intuitive behavior of these algorithms, compounded with the complexity of the real-world selection problems, has considerably limited their adoption. Instead, decision makers resort to heuristics in their portfolio decisions: mental searches across the available options, which are influenced by behavioral factors. We perform computer lab experiments for the standard knapsack problem to study the resource productivity vs. budget utilization trade-off that lies at the core of project selection. We show that decision makers' mental search is biased towards selecting too many small projects, which suggests that real-world equivalent decisions may lead to overly fragmented and suboptimal allocation of available resources. We provide evidence that this bias stems from the structure of the search process: decision makers naturally build up a feasible portfolio of projects (i.e., forward search), as opposed to eliminating projects from an infeasible set of projects (i.e., backward search). Furthermore, we show that advising participants to search equally in forward and backward direction suffices to reduce the small-project bias, and leads to better allocation decisions