AN ECONOMETRIC ASSESSMENT OF GUIDE TRAINING, SERVICE QUALITY, AND TOURIST SATISFACTION: THREE STANDARD TOOLS AND WHAT EACH OF THEM REVEALS

Abstract

The proposition that guide training improves service quality and tourist satisfaction is intuitively appealing but empirically challenging to estimate because guides may self-select into different levels of training, with more capable guides potentially undertaking more training. This article applies three standard econometric tools to a dataset comprising 60 guides, 300 tours, and 1,500 tourist ratings and evaluates the insights and limitations associated with each method. A nested regression specification for the service-quality score produces the expected pattern: the coefficient associated with 100 hours of training declines from 13.904 points in the baseline specification to 6.363 points after controlling for mentored trips, language proficiency, work experience, refresher training, and group size. This result indicates that more than half of the initial association between training hours and service quality is related to other characteristics that accompany greater training intensity. An instrumental-variable strategy uses distance to the nearest training centre as the instrument and yields a first-stage F-statistic of 21.6. The estimated coefficient for tourist satisfaction is 0.489 under the instrumentalvariable specification compared with 0.343 under ordinary least squares. The control-function test does not reject exogeneity (t = -1.05), and with 60 guide-level clusters the two estimates are not statistically distinguishable. Accordingly, the instrumental-variable results do not provide clear evidence that the ordinary least-squares estimate is materially biased in the expected direction. The mediation analysis, however, does not provide a stable basis for interpreting the mediated effect. Once the service-quality score is included, the direct training coefficient changes from 0.343 to -0.213, while the implied mediated share rises to 162%. This pattern is consistent with a situation in which the mediator and the treatment may share an unobserved determinant. Consequently, a reliable mediated share should not be reported without stronger identification assumptions. An Oaxaca-Blinder decomposition of the 0.579-star satisfaction gap between the better-trained half of guides and the remaining guides attributes 75.9% of the difference to observed characteristics. The largest contributions are associated with mentored trips (0.198 stars) and language proficiency (0.167 stars). Finally, a bounded scenario in which every guide is raised to the 75th percentile of training hours-an average increase of 39 hours-is associated with an estimated gain of 2.45 points in service quality and 0.188 stars in tourist satisfaction.