Abstract | Keywords | Export | Availability | Bookmark

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis consists of four empirical studies that analyze the higher education sector from its two main dimensions: education and research. The first two essays model the demand for higher education and simulate the effects of tuition fee increases and reductions in supply diversity on demand on general welfare. The final two essays analyze the productivity determinants of academic scientists, looking at both the star scientists and the lower tail of the distribution. Essay 1 analyzes the determinants of participation (whether to study) and schooling (where and what to study) in a public system of higher education, based on a unique dataset of all eligible high school pupils in an essentially closed region (Flanders). We find that pupils perceive the available institutions and programs as close substitutes relative to the outside option, implying an ambiguous role for travel costs: they hardly affect the participation decision, but have a strong impact on the schooling decision. To illustrate how our empirical results can inform the debate on reforming public systems, we assess the effects of tuition fee increases. Uniform cost-based tuition fee increases achieve most of the welfare gains; the additional gains from fee differentiation are relatively unimportant. These welfare gains are quite large under conservative assumptions on the social cost of public funds, and there is a substantial redistribution from students to outsiders. Essay 2 studies the profit and welfare effects of reducing supply diversity, against the background of a funding system reform in Flanders (Belgium). We find that the social desirability of cutting programs at institutions is limited to less than 10% of the cases (first-year undergraduate education), due to the students' low willingness to travel and relatively limited variable and fixed cost savings. Furthermore, the originally proposed version of the new funding system would often miss its purpose. In general, it gives an incentive to cut the smaller programs. However, we find that for the programs where cuts are undesirable, the system nevertheless encourages to cut 30-60% of the cases. Furthermore, for the minority of cases where program cuts are actually desirable, we find it provides the wrong incentive for up to half of the cases. These findings emphasize the complexities in regulating the diversity of supply in higher education, and serve as a word of caution towards the various other measures to cut supply diversity that have recently been introduced. Essay 3 contributes to the debate on cumulative advantage effects in academic research, by examining top performance in research productivity and its persistence over time, using a panel dataset comprising the publications of biomedical and exact scientists at the KU Leuven in the period 1992-2001. The data set allows taking into account factors like gender, age, cohort, rank, promotion, seniority, teaching load and access to research funding. About one quarter of the scientists in the sample achieve top performance at least once in the observation period, with six out of a hundred scientists being persistently top. Analyzing the selection and hazard to first and subsequent top performance, shows support for an accumulative process with rank, hierarchical position, access to funding and past performance as highly significant explanatory factors. Also gender is a consistent factor in explaining both top performance and its persistency. Essay 4 uses a quantile regression approach to estimate the effects of age, gender, funding, teaching load and other observed characteristics of academic researchers on the full distribution of research output. We employ recent advances in quantile regression that allow its application to count data, i.e. numbers of publications and citations. We account for unobserved heterogeneity of researchers by estimating a random-effects model, exploiting the panel nature of our dataset. Estimation of the model for a panel of biomedical and exact scientists at the KU Leuven in the period 1992-2001 shows strong support for our quantile regression approach, revealing the differential impact of regressors along the distribution. We also find that variables like funding, teaching load and cohort have a different impact on research quantity than on research quality. This thesis consists of four empirical studies that analyze the higher education sector from its two main dimensions: education and research. The first two essays model the demand for (undergraduate) higher education and simulate the effects of tuition fee increases and reductions in supply diversity on demand and on general welfare. The final two essays analyze the productivity determinants of academic scientists, looking at both the star scientists and the lower tail of the productivity distribution. The first essay ties into the issue of higher education financing. More specifically, it contributes to the discussion whether public higher education systems may benefit from incorporating more market-oriented principles, such as increased private contributions from students. Using a discrete choice model, we analyze the determinants of participation (whether to study) and schooling (where and what to study) in the Flemish public system of higher education. Given the lack of variation in tuition fees, we focus on the impact of travel costs in pupils' decisions as a measure of their sensitivity to costs, whilst controlling for high school background and demographics. The dataset used to estimate the model comprises all high school pupils in Flanders who completed high school in 2001 and are thus eligible to enter higher education. Second, using the estimates from the model, we simulate the effects of increased tuition fees. We look at the impact on both demand and welfare. We find that fee increases have a small effect on overall participation. This holds for both uniform fee increases as well as increases differentiated by type or study field. Differentiated fee increases represent a strong incentive for students to substitute towards the cheaper programs. With respect to the welfare effects, we find that higher private contributions raise welfare. In particular, raising tuition fees uniformly across study programs to the level of the average variable cost of education achieves most of the welfare gain. The additional gains from differentiating fees are relatively unimportant. These welfare gains are quite large under conservative assumptions on the social cost of public funds, and there is a substantial redistribution from students to outsiders. While the first essay looks into the effects of higher private contributions by students, the focus in the second essay turns to diversity in higher education supply. First and foremost, we analyze whether the supply of study programs in Flemish higher education has proliferated beyond what is socially desirable. A second question is whether a funding system containing a simple financial incentive, aimed at encouraging institutions to cut part of their supply, is effective in increasing overall welfare. This analysis was carried out amidst a funding system reform in Flanders: we analyze one of the incentives that was originally proposed as a key component of the new funding system which will enter in vigor on 1 January 2009. The analysis is based on the estimation of a demand model for higher education, using the same data on observed higher education choices of first-year undergraduate students in Flanders. The model extends the one in the first essay using a more detailed definition of study alternatives, allowing for a richer specification. We find that the social desirability of cutting programs at institutions is limited to less than 10% of the cases, due to the students' low willingness to travel and relatively limited variable and fixed cost savings. So with respect to their schooling decision students are very sensitive to travel cost and any reputation effects that may convince students to travel further are not sufficient to offset the strong preference for nearby options. Further, we contrast these welfare effects with the profit incentives given by a funding system that contains a rudimentary financial incentive to cut supply. We find that such a system, which closely resembles the originally proposed version of the new funding system for Flemish higher education, would often be off-target. In general it gives an incentive to cut the smaller programs. However, we find that for the large part of supply where program cuts are undesirable, the system nevertheless encourages to cut at least one third of those programs. Furthermore, for the minority of cases where program cuts are actually desirable, we find it provides the wrong incentive for up to half of the cases. These findings emphasize the complexities in regulating the diversity of supply in higher education, and serve as a word of caution towards the various other measures that have recently been proposed. The focus in the third essay lies on the top end of the research productivity distribution: the star scientists. More specifically, our objective is to identify the factors that determine whether a researcher ever ends up in this "top league", and if she does, which variables explain whether she is able to repeat that top performance. By examining top performance in research productivity and its persistence over time, this paper contributes to the debate on cumulative advantage effects in academic research. The custom-built data set contains the publications of biomedical and exact scientists at the KU Leuven from 1992 till 2001 and allows taking into account factors like gender, age, cohort, rank, promotion, seniority, teaching load and access to research funding. We carry out a clustering of researchers' yearly publication records distinguishing between three productivity categories (top, medium, low) while controlling for discipline and temporary gaps in research output. We find that about one quarter of the scientists in the sample achieves top performance at least once in the observation period, with six out of a hundred scientists being persistently top. Using mobility matrices, we find that top productivity generally is persistent over time: previous top performers are more likely to reach top status in next periods. We use the entry of a researcher into the top performance category to define the event of interest in a duration analysis, addressing the aforementioned research questions. A hazard model predicting the time towards first top performance confirms the importance of gender, with females being significantly less likely to reach top performance. Age and seniority effects are not significant, but rank and hierarchical position, as well as access to excellence funding are important for explaining the hazard to first top performance. There is only limited evidence with respect to the substitution effect of teaching load on top research performance. Low previous performers are less likely to reach top status, confirming that first top performance is a gradual, accumulative process, as the Matthew effect or a learning perspective would predict. When analyzing subsequent top performances, we find strong support for the accumulative process, with the hazard to next top performance being significantly and increasingly positively affected by previous top performance. Rank is important not only in predicting first top performance, but also for persistency in top performance, supporting the accumulative effect. Also the gender bias remains significant in explaining subsequent top performance, but this time with the dependence on previous top performance in favor of females, suggesting that the gender effect is mainly a selection problem into first top. While funding and head of unit position are important for selecting into first top performance, they are less predictive for subsequent top performance. And finally, the correction for unobservable individual heterogeneity, like ability, is significant, suggesting that talent remains an integral part of the story of top performance and its persistence. Like the third essay, the starting point of the fourth essay is the skewness that characterizes scientific output, with a small number of scientists responsible for the lion's share of publications and citations. While the previous essay examined productivity determinants of (persistence in) scientific excellence, here we widen the scope and aim to characterize the whole productivity distribution. Although the skew distribution suggests substantial heterogeneity among researchers, previous work on scientific productivity has typically focused on explaining average productivity. This ignores the potentially strongly different impact of regressors at different points in the distribution. This essay examines whether such effects operate evenly across the whole distribution, in which case a focus on average productivity alone would be justified, or whether they impact some points in the distribution more than others. In addition, we contrast effects on research quantity (using publication counts) with the impact of the same variables on research quality (using citation counts). Our empirical approach uses quantile regressions to estimate the effects of age, gender, funding, teaching load and other observed characteristics of academic researchers on various locations in the distribution of publications and citations. We employ recent advances in quantile regressions that allow its application to count data and account for unobserved heterogeneity of researchers. Estimation of the model using the same dataset as in the previous essay shows strong support for the quantile regression approach, revealing the differential impact of various regressors along the distribution. Further, the results show that the magnitude of effects typically decreases towards the top of the distribution. This may be explained in terms of the predominance of talent as a key success factor at the high end of the distribution and/or may point to a progressive loss of incentive power with quantile for factors like funding, rank, etc. As far as the comparison of the quantity and quality distribution is concerned, we find that funding, teaching load and a scientist's entry cohort have a different impact on research quantity than on research quality. Although one must be careful generalizing the results based on a single university, we argue that the findings are informative with respect to the management of scientists. In particular, they may instill the right expectations in administrators who implement incentive programs (e.g. reduced teaching load initiatives) or who make funding decisions.