The effects of Bologna Process on expenditure in HE systems of EU-15 countries 30 th Annual EAIR Forum (Copenhagen 2008) Tommaso Agasisti (Politecnico di Milano, Italy) Carmen Pérez Esparrells (UAM, Spain) Giusepe Catalano (Politecnico di Milano, Italy) Susana Morales Sequera (UAM, Spain)

Agenda Motivation Objectives Methodology and data Results Conclusions

The motivation Education => economic growth Patterns of HE expenditure Bologna process => process of convergence of educational policy (the objective: implementation of EHEA with common characteristics by the end of 2010)

The context Bologna Process: some characteristics The idea has fully overcome the first expectations. Why? Educational dimension => we are in the line of convergence Economic dimension => convergence? (our research question) Social dimension => the new challenge

Objectives Analysis of indicators of financial resources invested in HE (1998-2004) and verifying if is there a process of convergence in expenditure per student in HEIs. Estimating if the wealth of countries (measured by GDP per capita) has influenced the process of convergence.

Data Dependent variable: Expenditure per student Determinants: GDP per capita; % population who attained tertiary education; expenditure for HE as %GDP public funds to HE as a % of total HE expenditure “Bologna effect” (dummy) SOURCE: OECD data (Education at a Glance from 2002 to 2007)

Methodology Two approaches: Regression analysis: fixed-effects and random-effects to detect a “Bologna effect” Convergence analysis: Absolute convergence (β-convergence) Conditional β-convergence σ-convergence

σ-convergence The most important measure of cross-section analysis of dispersion that has been used: coefficient of variation (Barro & Salas-i-Martin). σ-convergence occurs if dispersion among countries falls in time. Coeficiente de variación: cociente entre la desviación típica y la media de la variable. Desviación típica: raiz cuadrada de la varianza Otros análisis utilizan la desviación típica de los logaritmos

β-convergence A process of absolute convergence (β-convergence) exists if countries with lower expenditure per student in HEIs levels have grown to higher rates than countries with better levels. β-convergence is calculated with the estimation of β in the following regression: β-convergence will exist if parameter β is positive and statistically significant

Conditional β-convergence In many situation an absolute convergence (β-convergence) cannot take place since there are different structural conditions between the different countries, so that they do not converge at a unique equilibrium point. There is absolute convergence if regions have the same starting level. In these cases, we use what Sala-i-Martin (1996), Barro and Sala-i-Marti (1992) and Mankiw, romer and Weil (1992) denominated conditional convergence (including an other explanatory variable).

Results of regression analysis Variable Model 1 _ Fixed effects Model 2 _ Fixed effects Model 1 _ Random effects Model 2 _ Random effects GDP per capita 0.224 0.156 0.253 0.191 % Population who attained tertiary education 91.860 61.804 62.311 54.232 Expenditure for tertiary education as %GDP 4,712.771 4,252.536 4,631.556 4,495.252 Public funds for tertiary education as %total 3.456 1.676 -3.593 -4.471 Bologna Process (Dummy) 756.482 630.207 Constant -4,100.000 -1,400.000 -5,200.000 -1,800.000 N 69 r2 0.613 0.658 Rmse 697.812 663.105 700.4 666.964 F 20.211 19.201

Results of σ-convergence Coeficiente de variación: cociente entre la desviación típica y la media de la variable. Desviación típica: raiz cuadrada de la varianza Source: authors’ elaboration

Results of β-convergence 1998-2004 1998-2001 2001-2004 β 0.046996* (1.842133) 0.028227 (0.560432) 0.082504** (2.563208) α 0.424430** (2.462573) 0.315045 (0.760094) 0.7131112** (3.092728) Adjusted R2 21.35% -5.33% 34.46% β(%) 4.7% 2.8% 8.2% Notes: t-Statistic in parenthesis. The coefficients are statistically significant with a confidence of 90%(*) or 95%(**).

Results of β-convergence in the entire period All the estimated parameters are statistically significant . β positive informs about absolute convergence. The goodness of fit is only 21.35%, which indicate scarce relation between both variables. However, in the period 2001-2004 is 34%.

Results of conditional β-convergence We also investigated whether convergence in the period 1998-2004 has been affected by national wealth, in this case, GDP per capita. In consequence, we ran a new estimation of the model including GDP per capita. The inclusion of GDPpc is only significant in the period 1998-2001 (after that period, the convergence is due to Bologna Process?)

Results of conditional β-convergence 1998-2001 Conditional convergence β 0.251728* (2.126098) α 1.238494*** (3.038887) λ 0.0000194*** (3.386701) Adjusted R2 43.75% Notes: t-Statistic in parenthesis. The coefficients are statistically significant with a confidence of 90%(*), 95%(**) or 99%.

Results of conditional β-convergence The introduction of GDPpc in the period 1998-2001 increases significantly the estimated parameters and increases the goodness of fit until 43.75%. In this period the growth rate of expenditure per student had been influenced by the level of GDPpc Countries have converged to different stationary states; that is, GDPpc variable explains cross-country patterns of growth in expenditure per capita in this period, before Bologna Process.

Conclusions The results of “financial” convergence for all the period (1998-2004) are clear and they are going in the “right” direction. We show evidence of an approaching process in the composition of HEIs expenditure per student in EU-15 countries. We have also identified that this convergence was more marked in the second part of the observation period, after Bologna process. New research: with data until 2010 what’s happen in the long term? Maybe, these difference in terms of expenditure per student will be less for a “natural” process 18

Conclusions The role of private sector in explaining the convergence in expenditure per student