Updating Tertiary Education Expectations and Choices with Learning, In Press at Education Economics.

Published Version

CORE DP Version

Abstract:  This paper explores how expectations and post-secondary education (PSE) path disruption decisions are affected by a learning process that students experience once enrolled in a PSE program. An unexpected change in grades, between high school and first year PSE program, serves as an informative signal on how well their academic performance and preferences align with the academic requirements and difficulty of the program that they enrolled. I find that learning about the fit affects students’ immediate PSE decisions about dropping out, switching program of study and their expectations about the current educational level. The effects are heterogeneous by parental education.


Higher Education Value-added using Multiple Outcomes, with Ernesto San Martin and Sébastien Van Bellegem, Journal of Educational Measurement, Fall 2016, Vol. 53, No. 3, pp. 368–400, doi: 10.1111/jedm.12114.

Published Version

CORE DP Version

Abstract:  In this paper we develop a methodology for the joint value added analysis of multiple outcomes that takes into account the inherent correlation between them. This is especially crucial in the analysis of higher education institutions. We use a unique Colombian data base on universities, which contains scores in five domains tested in a standardized exit examination that is compulsory in order to graduate. We develop a new estimation procedure that accommodates any number of outcomes. Another novelty of our method is related to the structure of the random effect covariance matrix. Effects of the same school can be correlated and this correlation is allowed to vary among schools.


Grades, Aspirations and Post-secondary Education Outcomes, with Louis N. Christofides, Michael Hoy and Thanasis Stengos, Canadian Journal of Higher Education, 45(1), p. 48-82, 2015. 

Published Version

IZA DP Version

Abstract:  In this paper, we exploit a rich longitudinal data set to explore the forces that, during high school, shape the development of aspirations to attend university and achieve academic success. We then investigate how these aspirations, along with grades and other variables, impact educational outcomes such as going to university and graduating. It turns out that parental expectations and peer factors have direct and indirect effects on educational outcomes through their impact on both grades and aspirations. Policy measures that enlighten parents about the value of education may positively modify educational outcomes.



Submitted/Under Revision/Completed Papers

Teacher Incentives and Sorting: Performance Pay Incentives in Secondary Education

Abstract: This study evaluates the causal effect of an accountability and collective pay incentive program on school performance and disentangles the underlying mechanisms by using a sharp regression discontinuity design (RDD).  I analyze the effect of the program on the composition of incumbent and newly hired teachers to separate two potential mechanisms that generate the impact: teachers’ incentives and teacher sorting. The Chilean National Subsidized School Performance Evaluation System identifies the best performing schools, makes them public, and awards a monetary bonus that is distributed to the teachers of the winning schools. I exploit the universal coverage of this long-term program together with a unique administrative record on Chilean school teachers and student test scores. The longitudinal nature of the data allows for tracking the teachers’ mobility between schools and across time. I find strong effects of the program on the composition of teaching staff: the winning schools attract more experienced and qualified teachers. Furthermore, combining the RDD with the instrumental variable approach, I show that the effect of the program on school performance operates through both mechanisms. However, the estimated effect through teacher sorting is much higher than the effect through teacher incentives. The results have direct policy implications, since in the long run teacher sorting across schools may contribute to a widening inequality between schools, where the cost is borne by the students in non-winning schools.


Intergenerational Transmission of Higher Education: Nature vs. Nurture,  with Louis N. Christofides, Michael Hoy and Thanasis Stengos, 2016.

IZA DP Version

CORE DP Version

Abstract: A strong positive correlation between the educational attainment of parents and their children has been well documented. Determining whether this relationship is due to selection or causal factors is a challenging and important problem. We use the Vietnam Era Draft lottery as a natural experiment to address this issue. Substantially more men attended university in this time period, creating a cohort of men including many who would not normally have attended university. We find that the inter-generational transmission of higher education for this cohort was not less than that of other “normal” cohorts, suggesting a causal role for educational attainment.


Wage Premium to University Quality and the Penalty of Missing Out, 2016.

Abstract: In this study I estimate the wage returns to university quality 4–6 years after graduation using the Canadian Youth in Transition Survey. I distinguish between two distinct measures of university quality: a survey-based university reputation ranking, and a new university ranking index that I construct from several university characteristics in order to objectively reflect the university quality. Information in the data allows me to incorporate in the analysis the high school grades and first-year university grades as a measure for ability. Using a matching estimator I am able to separate the wage premium for the treated and for the control group, and thus to estimate (i) the effect of university quality on those that actually attended a high ranked university (Average Treatment Effect on the Treated), and (ii) what the effect would have been on those who actually attended a low ranked university (Average Treatment Effect on Control). I find that the two coefficients are consistently different, the former marginally significant and the later always statistically significant indicating a wage premium between 10–15 percent. In another set of results I use an instrumental variable technique to further account for sorting in universities of different quality based on unobservable characteristics. This approach delivers a local average treatment effect (LATE) that is almost twice as large as the matching estimates.



Work in Progress

The Caveat of Context Measures in Value Added Analysis, with Ernesto San Martín and Sébastien Van Bellegem.

Abstract: It has now become conventional practice in empirical papers estimating value added models (VAM) to make comparisons between the school/teacher value added estimates that include and those that exclude a variable that proxies for the classroom/school composition effect (or otherwise also known as peer effects). In principal the two models should be nested in order for such comparison to be reasonable. In this paper we show explicitly that the two models have to comply with very strong assumptions in order to be nested. These assumptions are not supported by the data in most of the papers that claim to be in pursuit of better VAMs. This paper provides a transparent discussion on why this comparison is misleading and not correct. This is of particular importance especially if the value added estimates are used in high-stakes decision-making that informs policy implementation.


Long-term Differential Impacts of a Universal Voucher System. 

Abstract: This paper studies the universal voucher system implemented in Chile since 1981. Program impacts are identified using a cross-cohort regression analysis, exploring the fact that exposure to the voucher program is determined solely by the students’ birth year. Twelve birth cohorts, already in secondary school in 1981, were exposed to the program at different stages of their educational path. I compare those to ten cohorts that were not exposed to the voucher, and to others that were exposed to the voucher since the start. This is an exceptional opportunity to tackle the question of when it is most effective and efficient in terms of long-term outcomes to introduce a voucher in a student’s educational path; a policy relevant question not yet investigated in the literature. I find strong impacts of the program on higher education attendance and graduation. The analysis shows an increase in attendance probability by 5%, and indicates that introducing the voucher at the beginning of middle school is more cost-efficient but equally effective to the current policy. I find that degree attainment gains are 3%. The results strongly suggest that the timing when the voucher is made available does not affect the impact size. In terms of efficiently using governmental resources, the optimal time of offering a voucher could be as late as the start of high school.


Value Added Analysis in Secondary Education when Social Skills Matter,  with Tania Cabello, Alejandro Carrasco, and Ernesto San Martín.

Abstract: Schools and teachers contribute to the human capital of the students by developing their skills in multiple domains. So far the literature has focused only on the use of students’ test scores to evaluate the performance of schools or teachers, upon which high-stakes decision-making is based. Examples include collective pay-incentive awards, and even hiring and firing decisions for the teachers or school closures depend on such evaluations. However, the conventional approach lacks a complete representation of the school or teacher contribution, and at the same time induces incentives for schools or teachers that can be detrimental to the social capital of the students in the long-run. As an example consider shifting time allocation from extracurricular and social activities to preparing for the test. In this paper we test whether including measures of students’ social skills along with other cognitive skill measures delivers a significantly different result in evaluation. At the same time, we are interested in investigating the consequences of excluding student social skill measures from teacher evaluations on students’ long-run outcomes (such as higher education attendance, performance and completion), which makes this a policy-relevant research question.



Other Publications

Analysis of Cointegration in Capital Markets of France, Germany, and United Kingdom, with Hande Erdinc, Economics & Business Journal: Inquiries & Perspectives, 2(1), p. 109-123, 2009.

Published Version