COMPARISON OF EFFICIENCY ASSESSMENTS OBTAINED BY DATA ENVELOPMENT ANALYSIS AND STOCHASTIC FRONTIER ANALYSIS

Purpose of the article The purpose of the article is to assess efficiency of pension fund management Baltic countries by using data envelopment analysis and stochastic frontier analysis based models under various return on equity scenarios and compare results. Methodology/methods Methods used are data envelopment analysis (DEA) including its constant returns to scale (CRS) and variable returns to scale models (VRS), stochastic frontier analysis (SFA) as well as regression analysis. The research covers pension fund management companies in Latvia, Lithuania and Estonia in post financial crisis years from 2009 to 2014. Efficiency is studied from the operational perspective in accordance with the financial models and underlying assumptions. Scientific aim The aim is to apply the stochastic frontier analysis and data envelopment analysis as efficiency assessment methodology in the pension funds industry in the Baltics and compare efficiency results obtained by various models. Findings All regression equations demonstrated strong correlation and significance, what means that in general results obtained by various efficiency assessment models are consistent. Pearson correlation coefficient for regression equations comprising SFA and DEA CRS was higher than for equations with SFA and DEA VRS. There are rather constant returns to scale than variable returns to scale. High correlation between DEA CRS and VRS regression equations also provides evidence that returns to scale are actually close to be constant. Conclusions In case of DEA CRS model any deviation from the top performers is considered to be inefficiency disregarding the fact whether it is caused by a smaller size of operations or inefficient processes. DEA VRS model is comprehensive from a production economic point of view. However, it is quite tricky from a managerial point of view because it assumes not only increasing returns to scale, but also decreasing returns within the same model. SFA model results proved to be closer to DEA CRS than VRS.