Tuition Fees and Students’ Effort at University(Work in progress)
Pilar Beneito, Jose E. Bosca and Javier FerriUniversity of Valencia
II Workshop AnaecoValencia
6-8 July 2016
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 1 / 34
Outline
Introduction: tuition fees rise.
Theoretical setting: a model of students’ effort.
Empirical setting:I Data: University of Valencia.I Identification strategy.I Estimation results.
Final comments.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 2 / 34
Introduction: tuition fees rise
Many developed countries have witnessed in the last decades a significant
expansion in the number of students participating in higher education.
Higher education systems vary across advanced countries in many aspects,
but most of them share a common feature: higher education is heavily
subsidised.
Although subsidising higher education is theoretically justified (positive
externalities of education) a debate is open as regards who should bear a
higher burden of the total cost of studying at university :
I Greenaway and Haynes (The Economic Journal, 2003): beneficiaries ofhigher education in the UK should make a greater contribution to thefinancing of the higher education system =⇒ in the UK tuition feesincreased from £1,000 per annum in 1998/99, to £3,000 in 2006/07and to £6,000-£9,000 in 2012/13 (accompanied by the possibility ofdefering the payment of fees by taking income contingent loans).
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 3 / 34
Introduction: tuition fees rise
In countries like Spain raising tuition fees has not been an issue under
political discussion for many years.
However, the devasting effects of the Great Recession in Spain made the
government to approve a reduction of the subsidy to higher education.
In Spain, Law 14/2012 seeks to ’rationalise’ public expenditure shifting a
higher part of the costs of education onto the beneficiaries, i.e. the students:
I From the 2012-13 academic year on, university tuition fees coverbetween 15 and 25 percent of the total cost of education in thefirst-time registration, between 30 and 40 in the second-timeregistration, between 65 and 75 if a student makes a third-timeregistration and between 90 and 100 percent in the fourth-time orsuccessive registrations.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 4 / 34
Introduction: tuition fees rise
UVEG: Cost per credit in 2011-12 (in euros) and increase in 2012-13 (in %)
1st Reg. 2nd Reg. 3rd Reg. 4th Reg.
Economics and Business 13.07 22.87 28.75 28.75
(% increase in 2012-13) (33.3%) (33.3%) (127.3%) (203.1%)
Medicine 18.48 32.34 40.66 40.66
(% increase in 2012-13) (33.3%) (33.3%) (127.3%) (203.1%)
The cost of a first or second registration for a module in the degrees of
Economics, Business or Medicine at the University of Valencia (UVEG)
increased more than 33 percent in 2012-13 with respect to previous
academic year.
More pronnounced was the fee rise for third or fourth registrations (127 and
203 percent, respectively).
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 5 / 34
Introduction: tuition fees rise
Some literature about the possible effects of increasing tuition fees:
I Tuition fees and enrollment: Hubner (Economics of Education Review,2012); Coelli (Canadian Journal of Economics, 2009).
I Tuition fees and drop-outs: Garibaldi, Giavazzi, Ichino and Rettore(The Review of Economics and Statistics, 2012); Bradley and Migali(WP, 2015, Lancaster University).
I Financial rewards and students achievement: Leuven, Oosterbeek andvan der Klaauw (Journal of the European Economic Association,2010); Camerer and Hogarth (Journal of Risk and Uncertainty, 1999).
I Tuition fees and student behaviour:F Bolli and Johnes (Journal of Education and Work, 2015): tuition fees
rise modifies students’ allocation of study time between classes andhome.
F Ketel et al. (The Economic Journal, 2016): psychological costs offailing their studies may be lower for students that pay low tuition feesfor their study.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 6 / 34
Introduction: tuition fees rise
What we do?
I We focus on the following question: do students ’smart-up’ whentuition fees rise?
I A positive answer implies a higher efficiency in the use of ’subsidised’education resources.
I To investigate the causal effect of tuition fees on students’ effort andacademic achievement:
F Theoretical setting: a simple theoretical model of students’ effort andtuition fees.
F Empirical setting: using data from the University of Valencia, we applydiff-in-diff (DID) regressions.
I (We don’t consider enrollment and/or drop-out decisions).
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 7 / 34
Theoretical setting
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 8 / 34
Theoretical setting
A representative student starts at university at period t0 where she is going
to be until t0 + nu =⇒ We do not consider enrolment decision.
Anyone entering university pays tuition fees (tf ) and gets a degree after nuperiods=⇒ We do not model drop-outs.
At the end of year t0 + nu the student takes an exam.I If she passes, then she gets a First Class (FC) certificate.I If she fails she pays a monetary penalty and obtains a Second Class
(SC) certificate.
Given inherent ability, the probability of getting a FC degree is a positive
function of the student’s effort, s.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 9 / 34
Theoretical setting
In order to choose her optimal effort level, our student maximizes the
expected present value of utility which depends on a constant flow of
consumption (c) and leisure (l).
Present value of consumption equalizes present value of income=⇒No
posibility for bequest.
There are two possible cases (which will depend on student’s effort, s):
I (1) the student gets a FC degree ⇒Probability Prob (FC)I (2) the student gets a SC degree ⇒Probability 1-Prob (FC)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 10 / 34
Theoretical setting
And the present value of consumption under these two cases:
I c1 = y0 + sβ + b− tf ⇒Probability Prob (FC)I c2 = y0 + sβ − (2tf + d)⇒Probability 1-Prob (FC)
where:
I y0 is the lifetime income if not-graduate;I sβ is the premium for getting a degree;I β measures the responsiveness of lifetime income to effort;I b is the premium lifetime income for a FC degree (=0 for a SC degree);I tf stands for the tuition fees involved in a FC degree ;I d is the ’extra’ penalty in fees for a SC certificate.
The monetary cost of getting a SC degree is more than double that of a FC
(the student has to pay again for the tuition fees plus a penalty for not
succeeding to get a FC certificate).
At university the student reveals her effort in obtaining the degree, which
remains constant for the whole working life (nw − nu periods).
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 11 / 34
Theoretical setting
The probability of succeding in getting a FC degree is assumed to follow a
sigmoid function (”S” shape curve) of effort, s.
In particular, we assume a generalized logistic function, bounded at 1,with
location parameter µ (and scale parameter assumed to be 1 and omitted in
notation):
Prob(FC) = F(µ, s) =1
1 + e−(s−µ)
Parameter µ is the midpoint of the function.
Changes in µ ⇒ shift the function. These are changes in ability and/or
inherent difficulty of studies.
Thus, µ can be interpreted as an ’ability/difficulty’ parameter: an increase in
ability reduces µ; an increase in difficulty of studies increases µ.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 12 / 34
Plot 1: Probability of getting a FC certificate for µ = 3, 5, 7.
0 1 2 3 4 5 6 7 8 9 10
Effort index
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pro
babili
ty o
f gettin
g a
FC
degre
e
P = [1 3]
P = [1 5]
P = [1 7]
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 13 / 34
Theoretical setting
Leisure is a negative function of effort⇒ l = h− s, where h is total time
available and split between leisure and time effort.
Preference for leisure (=1-preference for consumption): parameter 1− α.
Then the optimization problem of the student takes the form
maxs
E[U(c, l)] = F(µ, s)(
y0 + sβ + b− tf)α
(h− s)1−α +
(1− F(µ, s))(
y0 + sβ − (2tf + d))α
(h− s)1−α
s.t.
F(µ, s) =1
1 + e−(s−µ)
0 6 s 6 h
Parameters: µ, y0, β, b, h, α, d and tf .
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 14 / 34
Theoretical setting
The first order condition takes the form:
[Fs(µ, s)cα
1 + α cα−11 βsβ−1F(µ, s)
−Fs(µ, s)cα2 + α cα−1
2 βsβ−1 (1− F(µ, s))
](h− s)1−α
− (1− α) (h− s)−α [F(µ, s)cα1 + (1− F(µ, s)) cα
2 ] = 0
This equation does not admit an analytical solution.
Hence we rely on a numerical solution.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 15 / 34
Theoretical model: parameter set-up
First, we normalize total time available and backup lifetime income
h = y0 = 10.Next, we obtain tf and d consistent with y0, using information on the actual
cost of tuition fees at UVEG and assuming a temporal disccount rate
γ = 0.8.β is set such that a university degree implies a reward of 15% with respecto
to y0 (regardeless of it being a FC or a SC degree).
We assume that getting a First Class certificate almost doubles the income
with respect to a Second Class degree, and so we set b = y0.
Finally, we set α such that, for a student with µ = 6.5, the model solution
replicates the probability of passing in the first registration that we observe
in the data for Economics & Business at UVEG.
Table T1. Parameter values
y0 h β b α γ tf d µ f10 10 0.15 10 0.61 0.8 0.53 0.30 6.5 1
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 16 / 34
Theoretical model: solution
Table T2. Model solution
Student’s effort Prob FC degree
Benchmark 6.54 48.4y0 = 10→ 9 6.70 52.6β = 0.15→ 0 6.61 50.3b = 10→ 11 6.68 52.0α = 0.61→ 0.63 6.69 52.2tf = 0.53→ 0.7 6.62 50.5d = 0.3→ 0.4 6.57 49.3µ = 6.5→ 6.0 6.30 57.6
Higher effort and probability of getting a FC degree: the lower the
non-graduate income; the lower the income responsiveness to effort at
university; the higher the income premium for a FC degree; the lower the
preferences for leisure; the higher tuition fees; the higher the step penalty for
failing to obtain a FC certificate;
Lower effort and higher probability of getting a FC degree:I the higher innate ability.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 17 / 34
Theoretical model: solution
Plot 1: Heterogeneity in the response to tuition fees
(Different response of Prob(FC) to tuition fees’ rise depending on ability).
2.5 3 3.5 4 4.5 5.5 6 6.5 7 7.55
Ability parameter (µ)
0
1
2
3
4
5
6
Ch
an
ge
in
Pro
b o
f g
ett
ing
a F
C d
eg
ree
High ability Low ability
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 18 / 34
Empirical setting
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 19 / 34
Data: University of Valencia
Data: students of Economics, Business, and Medicine, at the University of
Valencia, from 2010 to 2014 (5 years).
Economics and Business students differ from Medicine students on: ability,
difficulty of studies, level of fees. Treated separately from each other in
estimation.
Information on the student’s academic ’performance’: entrance mark, grades
(per module, academic year, course), number of times registered in a
module.
Information on economic and educational levels of the family and the
student: mother’s and father’s education, mother’s and father’s labour
situation, student’s working situation.
Other: age, gender.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 20 / 34
Identification Strategy
Identification of causal effect of a ’policy intervention’: what the individual
would have done without intervention?: the ’counterfactual’.
Problem: the counterfactual is not observed.
Diff-in-diff approach (DID): how the difference between those treated
(individuals subject to the treatment) and a ’control’ group (those not
subject to the treatement) differs before and after the treatment.
DID approach: robust to idiosincratic permanent differences between the
treated and the control group.
In our setting:I Treated with the rise in tuition fees? All students with ’ordinary’
registration mode.I Control group: students exempt of paying tuition fees: in University of
Valencia → students of ’special numerous families’.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 21 / 34
Identification Strategy
Standard DID regression:
Yimt = α + αDDi + γPostt + γDPosttDi + γxit + cm + ai + uit
where subscripts i, m and t denote the student, module and year,
respectively.
Postt is a set of dummy step variable for all periods (years) after treatment.
Di is a dummy variable for ’treated’ students.
xit stands for a vector of control variables (educational level of mother and
father, economic situation of mother and father, type of module, age,
gender and working situation of the student).
cm stands for module-effects (we control for by including the average of the
dependent variable for each module).
ai stands for student-fixed effects (we estimate both by RE and FE).
uit: iid error
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 22 / 34
Identification Strategy
Standard DID regression:
Yimt = α + αDDi + γPostt + γDPosttDi + γxit + cm + ai + uit
Based on 2 asssumptions:I 1. ’Parallel path assumption’: treated and control group do not
differ in their trends before the treatment. A crucial assumption.I 2. Treatment effects (if any) are equal in all periods after treatment
(static effects)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 23 / 34
Identification Strategy
Fully flexible DID regression (Mora and Reggio, 2013): (i) it allows to
check for non-paralell pre-treatment trends; (ii) it allows for dynamic effects
after treatment.
Yimt = α + αDDi +T
∑τ=t0+1
βτIτ +T
∑τ=t0+1
βDτ IτDi + γxit + cm + ai + uit
I where Iτ is a set of dummy variables identifying years 2011, 2012,2013, 2014 (with 2010 absorbed in the constant)
With treatment ocurring at t∗ (sometime between 2011 and 2012):
If βDτ = 0 ∀τ < t∗ → parallel paths (pre-treatment) (1)
then βDτ ∀τ > t∗ → treatment effects (2)
I If pre-treatment trends are not parallel, take differences of thedependent variable and apply the DID regression: Diff-in-double-diff(DI2D)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 24 / 34
Identification Strategy
Yit = α + αDDi +T
∑τ=t0+1
βτIτ +T
∑τ=t0+1
βDτ IτDi + γxit + cm + ai + uit
We estimate the DID model for 3 alternative ’outcome’ variables Yit:I (1) Number of times the student has ’paid’ tuition fees to pass the
module (N. of registrations until pass);I (2) Probability of passing the module in the first registration;I (3) Grades.
We distinguish in estimation between:I ’Top’ Students: those with entrance grade above the upper 80
percentile.I ’Average’ Students: those with entrance grade below the 80 percentile.
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 25 / 34
Table D1. Descriptives
Economics & Business Medicine All Averg.
Stud. Top
Stud. All Averg.
Stud. Top. Stud.
Number Observations 143.976 119.726 24.250 39.579 8743 30.836
Control group (N. Obs) 1.739 1.328 411 1.002 824 178
N registrations until pass 1.64 1.71 1.40 1.20 1.23 1.10
Before fees rise (2010-2011) 1.57 1.65 1.31 1.16 1.19 1.08
After fees rise (2012-2014) 1.72 1.77 1.51 1.26 1.29 1.17
Prob. of pass in 1st registration 0.45 0.42 0.59 0.73 0.71 0.78
Before fees rise (2010-2011) 0.42 0.40 0.57 0.71 0.69 0.77
After fees rise (2012-2014) 0.48 0.46 0.59 0.74 0.73 0.80
Grades (if presented to the exam) 5.21 5.06 6.08 6.65 6.49 7.30
Before fees rise (2010-2011) 5.06 5.39 6.01 6.59 6.42 7.21
After fees rise (2012-2014) 5.39 5.02 6.20 6.76 6.61 7.33
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 26 / 34
Table R1. ECONOMICS+ BUSINESS Students Number of registrations to pass Flexible Diff-in-Diff estimation
RE
RE
RE
FE
FE
All Students Averg. Stud. Top Stud. Averg. Stud. Top Stud. Treated 0.153 0.258 0.032 - - (0.146) (0.185) (0.218) Year 2011 0.408*** 0.426*** 0.314** 0.463*** 0.286*** (0.091) (0.100) (0.124) (0.097) (0.109) Year 2012 0.837*** 0.887*** 0.746*** 0.938*** 0.720*** (0.106) (0.117) (0.163) (0.100) (0.131) Year 2013 1.201*** 1.423*** 0.761*** 1.504*** 0.760*** (0.090) (0.114) (0.115) (0.094) (0.085) Year 2014 1.630*** 1.782*** 1.164*** 1.868*** 1.150*** (0.087) (0.094) (0.161) (0.082) (0.107) Treated in Year 2011 -0.022 -0.015 -0.032 -0.027 0.009 (0.092) (0.101) (0.126) (0.098) (0.110) Treated in Year 2012 -0.071 -0.080 -0.153 -0.092 -0.109 (0.106) (0.117) (0.163) (0.101) (0.129) Treated in Year 2013 -0.138 -0.311*** 0.080 -0.338*** 0.104 (0.090) (0.114) (0.113) (0.094) (0.078) Treated in Year 2014 -0.225*** -0.320*** 0.011 -0.344*** 0.051 (0.087) (0.095) (0.152) (0.083) (0.089) Constant -1.924*** -2.196*** -0.204 0.014 0.239 (0.276) (0.350) (0.442) (0.206) (0.347) Observations 51,912 40,699 11,213 40,699 11,213 R-squared 0.274 0.301 0.213 0.416 0.278 Number of Students 3,502 2,806 826 2,806 826 Hausman fixed effecs 913.71 259.94 [p-value] 0.000 0.000
RE: random effects estimation; FE: fixed effects estimation. Robust errors in parenthesis (* signif. at 10%; **signif at 5%; ***signif at 1%)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 27 / 34
Table R2. MEDICINE Students Number of registrations to pass Flexible Diff-in-Diff estimation
RE
RE
RE
FE
FE
All Students Averg. Stud. Top Stud. Averg. Stud. Top Stud. Treated 0.323*** 0.362*** -0.058 - - (0.095) (0.095) (0.051) Year 2011 0.177*** 0.204*** 0.121*** 0.192*** 0.135*** (0.034) (0.043) (0.027) (0.043) (0.028) Year 2012 0.393*** 0.428*** 0.273*** 0.464*** 0.310*** (0.039) (0.053) (0.047) (0.070) (0.056) Year 2013 0.628*** 0.677*** 0.448*** 0.784*** 0.507*** (0.058) (0.076) (0.073) (0.090) (0.086) Year 2014 1.018*** 1.204*** 0.555*** 1.422*** 0.614*** (0.126) (0.187) (0.087) (0.138) (0.100) Treated in Year 2011 0.025 0.027 -0.001 0.075 0.012 (0.035) (0.044) (0.024) (0.047) (0.023) Treated in Year 2012 0.037 0.041 0.017 0.062 0.021 (0.036) (0.052) (0.025) (0.069) (0.031) Treated in Year 2013 0.035 0.040 0.035 0.021 0.023 (0.052) (0.072) (0.033) (0.084) (0.034) Treated in Year 2014 -0.211* -0.339* 0.058 -0.461*** 0.047 (0.123) (0.185) (0.050) (0.134) (0.052) Constant -1.708*** -2.132*** -1.053** -0.217 -0.090 (0.367) (0.448) (0.470) (0.230) (0.174) Observations 24,267 18,547 5,720 18,547 5,720 R-squared 0.242 0.257 0.244 0.224 0.128 Number of Students 1,326 1,018 343 1,018 343 Hausman fixed effecs 1151.39 343.02 [p-value] 0.000 0.000
RE: random effects estimation; FE: fixed effects estimation. Robust errors in parenthesis (* signif. at 10%; **signif at 5%; ***signif at 1%)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 28 / 34
Table R3. ECONOMICS+ BUSINESS Students Dependent variable: Probability of pass in 1st registration Linear Probability Model. Flexible Diff-in-Diff estimation
RE RE FE FE Averg. Stud. Top Stud. Averg. Stud. Top Stud. Treated -0.134*** -0.087 - - (0.048) (0.061) Year 2011 -0.063 -0.085 -0.040 -0.043 (0.046) (0.071) (0.045) (0.072) Year 2012 -0.058 -0.107 -0.020 -0.053 (0.050) (0.082) (0.051) (0.081) Year 2013 0.021 -0.053 0.042 -0.010 (0.063) (0.078) (0.066) (0.077) Year 2014 -0.092 -0.026 -0.068 0.029 (0.080) (0.148) (0.081) (0.146) Treated in Year 2011 0.061 0.033 0.045 -0.013 (0.046) (0.072) (0.045) (0.073) Treated in Year 2012 0.083* 0.068 0.066 0.017 (0.050) (0.083) (0.051) (0.082) Treated in Year 2013 0.052 0.046 0.059 0.005 (0.063) (0.078) (0.066) (0.077) Treated in Year 2014 0.212*** 0.015 0.224*** -0.035 (0.080) (0.149) (0.081) (0.146) Constant 0.834*** 1.135*** 0.732*** 1.070*** (0.065) (0.108) (0.096) (0.123) Observations 62,317 15,200 62,317 15,200 R-squared 0.426 0.438 0.049 0.035 Number of Students 2,848 827 2,848 827 Hausman test fixed effecs 336.96 73.46 [p-value] 0.000 0.000
RE: random effects estimation; FE: fixed effects estimation. Robust errors in parenthesis (* signif. at 10%; **signif at 5%; ***signif at 1%)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 29 / 34
Table R4. MEDICINE Students Dependent variable: Probability of passing in 1st registration
Linear Probability Model. Flexible Diff-in-Diff estimation RE RE FE FE Averg.
Stud. Top Stud. Averg.
Stud. Top. Stud.
Treated -0.050 0.081 - - (0.045) (0.082) Year 2011 -0.039 -0.000 -0.076* 0.006 (0.038) (0.062) (0.044) (0.063) Year 2012 -0.145*** -0.058 -0.143*** -0.037 (0.047) (0.074) (0.050) (0.076) Year 2013 -0.137** 0.119 -0.181** 0.121 (0.059) (0.116) (0.076) (0.118) Year 2014 -0.264** 0.001 -0.204 0.089 (0.106) (0.114) (0.237) (0.115) Treated in Year 2011 -0.010 -0.021 0.036 -0.028 (0.038) (0.063) (0.045) (0.064) Treated in Year 2012 0.093** 0.013 0.135*** 0.003 (0.047) (0.075) (0.051) (0.076) Treated in Year 2013 0.119** -0.139 0.217*** -0.120 (0.059) (0.116) (0.075) (0.118) Treated in Year 2014 0.340*** 0.028 0.408* 0.023 (0.107) (0.114) (0.237) (0.115) Constant 1.322*** 0.662*** 1.199*** 0.889*** (0.112) (0.210) (0.285) (0.058) Observations 22,643 6,750 22,643 6,750 R-squared 0.493 0.629 0.204 0.099 Number of Students 1,029 345 1,029 345 Hausman test fixed effecs 127.49 55.23 [p-value] 0.000 0.000
RE: random effects estimation; FE: fixed effects estimation. Robust errors in parenthesis (* signif. at 10%; **signif at 5%; ***signif at 1%)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 30 / 34
Table R5. ECONOMICS & BUSINESS Students Dependent variable: Grades Flexible Diff-in-Diff estimation
RE RE FE FE Averg.
Stud. Top Stud. Averg.
Stud. Top. Stud.
Treated -0.652** -0.775** - - (0.282) (0.391) Year 2011 -0.092 0.371 -0.040 0.721 (0.222) (0.791) (0.217) (0.811) Year 2012 0.250 0.124 0.335 0.579 (0.297) (1.166) (0.293) (1.239) Year 2013 0.173 0.081 0.246 0.526 (0.432) (0.433) (0.432) (0.401) Year 2014 -0.341 1.231 -0.176 1.848 (0.502) (1.156) (0.514) (1.296) Treated in Year 2011 0.294 -0.331 0.278 -0.656 (0.224) (0.797) (0.218) (0.816) Treated in Year 2012 0.192 0.087 0.173 -0.287 (0.299) (1.168) (0.295) (1.237) Treated in Year 2013 0.698 0.461 0.717* 0.105 (0.434) (0.443) (0.435) (0.402) Treated in Year 2014 1.436*** -0.740 1.392*** -1.230 (0.507) (1.162) (0.519) (1.295) Constant 5.666*** 7.590*** 4.745*** 6.995*** (0.418) (0.738) (0.580) (0.528) Observations 119,726 24,250 119,726 24,250 R-squared 0.100 0.077 0.039 0.031 Number of Students 3,170 935 3,170 935 Hausman test fixed effecs 1245.23 194.08 [p-value] 0.000 0.000
RE: random effects estimation; FE: fixed effects estimation. Robust errors in parenthesis (* signif. at 10%; **signif at 5%; ***signif at 1%)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 31 / 34
Table R6. MEDICINE Students Dependent variable: Grades Flexible Diff-in-Diff estimation
RE RE FE FE Averg.
Stud. Top Stud. Averg.
Stud. Top. Stud.
Treated -0.204 1.146 - - (0.498) (1.111) Year 2011 -0.429 0.318 -0.477 0.179 (0.350) (0.338) (0.481) (0.435) Year 2012 -0.518 -0.484 -0.474 -0.507 (0.326) (0.717) (0.432) (0.755) Year 2013 -0.435 0.228 -0.360 0.249 (0.551) (0.413) (0.627) (0.451) Year 2014 -0.945 0.951** -0.773 1.157** (1.218) (0.404) (1.225) (0.472) Treated in Year 2011 0.251 -0.498 0.381 -0.213 (0.354) (0.350) (0.488) (0.445) Treated in Year 2012 0.561* 0.177 0.686 0.415 (0.331) (0.723) (0.440) (0.758) Treated in Year 2013 0.758 -0.143 0.936 0.119 (0.557) (0.431) (0.637) (0.462) Treated in Year 2014 2.356* -0.269 2.551** -0.111 (1.226) (0.442) (1.236) (0.502) Constant 7.787*** 3.444* 5.808*** 5.703*** (0.914) (2.087) (1.098) (0.409) Observations 30,836 8,743 30,836 8,743 R-squared 0.100 0.077 0.061 0.033 Number of Students 1,081 364 1,081 364 Hausman test fixed effecs 564.30 151.12 [p-value] 0.000 0.000
RE: random effects estimation; FE: fixed effects estimation. Robust errors in parenthesis (* signif. at 10%; **signif at 5%; ***signif at 1%)
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 32 / 34
Final comments
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 33 / 34
Final comments
The effects of modifiying tuition fees is diverse.
Negative effects include possible reduction in enrollment rates and increase
in drop-outs, probably affecting poorer individuals more sharply . (In our
research agenda).
Tuition fees’, however, also act as an incentive mechanism that may enhance
students’ academic effort and performance.
We find positive effects on the ’efficiency’ in the use of education subsidised
resources. Also evidence on better students’ achievements (grades).
We also find that responses are heterogenous: in particular, we find that
average students respond more to tuition fees rise than top students.
The design of the higher education policy needs to be ’integral’: tuition fees
rise + loans (or grants).
II Workshop Anaeco Tuition fees & students’ effort 6-8 July 2016 34 / 34