Efficiency in Education - A Multilevel Analysis

Sammanfattning: Sweden spends altogether about 4.5% of GDP on primary, secondary and upper secondary school. Considering the magnitude of this investment it is important that these resources are used efficiently. In this thesis efficiency in public education is analysed on three different levels: The municipal level, the school level, and the individual level. Efficiency is estimated using a linear programming technique, Data Envelopment Analysis (DEA). Each producer is compared with the best performing producers in the sample, and for each producer an efficiency score is estimated. The scores are here used for further analysis of the estimated efficiency. In the first essay, efficiency is analysed for primary and secondary school in 287 Swedish municipalities in the school year 1999/2000. Average efficiency is between 0.90 and 0.96, implying that the municipalities on average could decrease their input use with 4-10 percent without decreasing their results. Focus in the paper is on competition and political factors influencing efficiency in the production of public education. Competition from private schools has no significant influence on efficiency. Among the political variables we find efficiency to be related to political majorities but we find no effects from taxes or voter characteristics. In the second essay, efficiency is estimated for 840 Swedish secondary schools in 1994/95. Taking differences in the production environment into account in the efficiency modelling, mean efficiency is between 0.84 and 0.87. Focus in the analysis is on how private school competition and teacher characteristics influence efficiency. Private school competition has a positive effect, implying that schools facing more private school competition are more efficient producers. Among the teacher characteristics, positive effects are found for variables related to teacher experience. In the third essay the individual student is thought of as being responsible for her own education and inefficiency is estimated for 59 354 students graduating from upper secondary school in 1999. In order to get estimates for larger groups of students, the individual inefficiency is aggregated. Problems with aggregation when estimating inefficiency in multiple dimensions are discussed. In the analysis special attention is given educational programs where one of the sexes is in a clear minority. We find no systematic evidence that the sex in minority has either higher or lower mean inefficiency, but the sexes seem to allocate outputs differently in these programs.