Measuring Skills in Develping Countries

Measures of cognitive, noncognitive, and technical skillsare increasingly used in development economics to analyzethe determinants of skill formation, the role of skills ineconomic decisions, or simply because they are potentialconfounders. Yet in most cases, these measures have onlybeen validated in high-income countries. This paper teststhe reliability and validity of some of the most commonlyused skills measures in a rural developing context. A surveywith a series of skills measurements was administered tomore than 900 farmers in western Kenya, and the samequestions were asked again after three weeks to test thereliability of the measures. To test predictive power, thestudy also collected information on agricultural practicesand production during the four following seasons. Theresults show the cognitive skills measures are reliable andinternally consistent, while technical skills are difficult tocapture and very noisy. The evidence further suggests thatmeasurement error in noncognitive skills is non-classical,as correlations between questions are driven in partby the answering patterns of the respondents and thephrasing of the questions. Addressing both random andsystematic measurement error using common psychometricpractices and repeated measures leads to improvementsand clearer predictions, but does not address all concerns.The paper provides a cautionary tale for naïve interpretationsof skill measures. It also points to the importanceof addressing measurement challenges to establish therelationship of different skills with economic outcomes.Based on these findings, the paper derives guidelines forskill measurement and interpretation in similar contexts.