| Abstract | Glavni cilj ovog rada je bio analizirati efikasnost javne potrošnje na visoko obrazovanje koristeći se DEA metodom za izračun tehničke efikasnosti, te na temelju toga rangirati zemlje Europske Unije od najefikasnije prema najneefikasnijoj i pokušati odrediti razloge (ne)efikasnosti. S obzirom na to da je problem istraživanja efikasnost državnih izdvajanja na visoko školstvo, postavljene su dvije hipoteze. Prva od tih hipoteza sugerira da sve zemlje članice EU postižu visoku tehničku efikasnost javnih izdataka na visoko obrazovanje; u drugoj hipotezi se pretpostavilo da zemlje sa višim javnim izdacima postižu i viši koeficijent efikasnosti u korištenju resursa. DEA je ne-parametrijska metoda koja koristi linearno programiranje i sa lakoćom kombinira višestruke inpute i višestruke outpute, što je čini pogodnom za ovo istraživanje. DEA dodjeljuje pondere inputima i outputima i na temelju dobivenih rezultata, rangira zemlje po efikasnosti. Analizirani modeli u ovom radu bili su modeli varijabilnih prinosa na opseg, sa tri inputa (broj diplomiranih studenata, akademsko osoblje i javni izdaci na visoko obrazovanje po studentu) i dva outputa (broj diplomiranih studenata i zapošljivost) u dvije godine: 2013. (Model 1) i 2016. (Model 2). Modeli su bili izlazno usmjereni. Rezultati istraživanja pokazali su da zemlje Europske Unije, uistinu, postižu visoki koeficijent tehničke efikasnosti, međutim druga hipoteza je pokazala da je gotovo polovica zemalja u oba promatrana modela bila u donjoj polovici što se varijable izdvajanja za visoko školstvo po studentu (u EUR) tiče. U Modelu 1, 16 od 28 zemalja je postiglo maksimalnu tehničku efikasnost, dok je u Modelu 2 to uspjelo 15 zemalja. Zemlja sa najslabijim rezultatom efikasnosti iz Modela 1 se pokazala maksimalno efikasnom u drugom testiranom modelu, dok su određene zemlje koje su se pokazale efikasnima u prvom modelu, bile i dalje na zavidnoj razini efikasnosti, ali ipak nisu dosegle maksimum. |
| Abstract (english) | The main goal of this research was to explore and analyse the efficiency of public expenditures on higher education in the EU – 28 countries using Data Envelopment Analysis (DEA), and rank countries from the most efficient ones to the inefficient ones as well as to try to determine the reasons for (in)efficiency. Considering that the main problem of this paper was researching the efficiency of public expenditure on higher education, there were two hypotheses. The first hypothesis suggested that all of the countries members of EU will achieve high coefficient of technical efficiency of public expenditure on higher education; the second hypothesis assumed that countries with greater public expenditures on higher education will achieve a higher coefficient of technical efficiency in using their resources. DEA is a non-parametric method that uses linear programming and combines multiple inputs and multiple outputs with ease, which makes it suitable for this research. DEA assigns weights to inputs and outputs and by doing so, after the scores are calculated, it ranks the decision making units by their efficiency coefficient. The analyzed models that were used in this paper were models of variable returns to scale with three inputs (number of students enrolled, academic staff and public expenditure on higher education per student in EUR) and two outputs (number of graduates and employability) in two years: 2013. (Model 1) and 2016. (Model 2). These models were, also, output oriented. We have reached the following conclusions: the countries of European Union, indeed achieve high coefficient of technical efficiency, however, second hypothesis showed that almost half of the countries in both observed models were in the bottom half when analyzing public expenditure on higher education per student (in EUR). In Model 1, 16 out of 28 countries managed to achieve maximum technical efficiency, while 15 countries achieved it in Model 2. The country with the weakest score from Model 1, Greece, was amongst the most efficient ones in Model 2, whilst certain countries that were efficient in Model 1, did not reach maximum coefficient in Model 2, although their score is, nevertheless, high. |
