Education Matter? Recent Evidence from International OECD data

advertisement
Does (Private) Education
Matter?
Recent Evidence from
International OECD data
Vincent Vandenberghe &
Stephane Robin
Alice Ballabio
1
Plan of presentation
Introduction
First
section: Theoretical framework
Second section: Data set and
estimation strategy
Third section: Results and Analysis
Conclusion
2
Introduction
The aim of this study is to estimate private vs. public
sector schools relating to the effect on academic
achievements.
In order to do that, the authors used a research of PISA,
which is a Programme for International Student
Assessment sponsored by OECD “Organization for
Economic Co-operation and Development”.
This survey, carried out in 2000, is aimed at testing the
competencies in Maths, Sciences and Reading of
representative samples of 15 year-old students across
OECD and non-OECD countries.
3
List of countries
Australia, Austria, Belgium (French-Speaking),
Belgium (Dutch-Speaking), Brazil, Canada,
China, Czech Republic, Denmark, Finland,
France, Germany, Greece, Hungary, Iceland,
Ireland, Italy, Japan, Hong Kong China, Korea,
Latvia, Luxembourg, Mexico, Netherlands, New
Zealand, Norway, Poland, Portugal, Russian
Federation, Spain, Sweden, Switzerland, United
Kingdom, United States.
4
Defining Public and Private School
Public school  is a school managed directly or
indirectly by a public education authority, a
government agency, or governing board appointed by
government.
Private school  is a school managed directly or
indirectly by a non-government organization (for
instance, a church, trade unions, businesses, or other
private institutions).
5
First Section: Theoretical
Framework
 Education
production function
 Different categories of variables and
biases that must be accounted for in order
to isolate a true private/public differential
effectiveness
 Empirical strategy used to estimate these
models
6
Education production function
The academic achievement A, at any time period t, is a
function of monetary school resources, of the
student’s peers and of the student’s individual
characteristics.
A = f (R, SC, P)
A = student’s achievement
R = school monetary resources or proxies
SC = student’s characteristics (including family and social background)
P = some characterization of the group of peers
7
Education production function
Student i’s achievement (Aij) in country j can be
explained by a linear model of the following form:
i = student index
j = country index
Aij = Rij j + SCij j + Pij j + j PRIVij + ij
8
Education production function
Where:
 = controls the effect of the monetary resources on the level of
A
 = controls the effect of the individual characteristics on A
 = controls the effect of the group of peers on A
The main important variable, which makes a difference is the
coefficient  which controls the effect on the variable PRIV
The estimated value (PRIV) for the coefficient  will
capture the differential effectiveness between private
and public schools
9
Sample Selection Bias of
Heckman
-
-
The sample is not randomly chosen.
There may be a number of unobserved variables or
variable that are imperfectly measured that could
determine the choice to attend or not to attend the
private school.
Because of motivation, ability, selection or selfselection are unobserved, the relative effectiveness of
private and public schools could be confused with the
background of their students. In this way, there is the
insertion of the BIAS.
10
“Treatment effect” model, inspired by
“Two-steps” correction model of Heckman
To eliminate Selection Bias:
- The first step, which indicates a choice within each j country, to attend or not to
attend the private school is given by;
PRIV* ij = Zijj + ij
Where:
PRIV* = indicates the choice to attend or not to attend the private school.
Z = represents the probability that a student decides to attend a private school on the
base of some controlled factors related on the private/public dummy variable,
which are: (you can see it looking at the database)
GIRL = gender
NSIB = number of siblings
BRTHORD = the birth order of student
MISCED = highest degree of mother
FISCED = highest degree of father
FATHIM = immigration status of father
HISEI = highest socio-economic index of the two parents
HEDRES = cultural resources
11
“Treatment effect” model, inspired by
“Two-steps” correction model of Heckman
The choice equation regresses the set of individual
characteristics on the private/public dummy variable (PRIV).
The choice to attend or not to attend the private school is made
according to the rule:
PRIV = 1 if PRIV* ij >0, PRIV = 0 otherwise
For the estimation of this equation they used the probit model
which hypothesizes that the choice between public or private
school is given by the propensity to attend private school or
the public one.
This model linked the linear expression Zijj + ij to the
probability of attending a private school.
12
“Treatment effect” model, inspired by
“Two-steps” correction model of Heckman
- The Second step, which indicates the achievement;
The achievement in private school (Aij  PRIVij = 1) is only observed when
the unobserved variable of the choice model is positive (PRIV*>0).
Algebraically, this means that:
E (Aij  PRIVij = 1) = E (Aij PRIV*ij >0)
= E (Aij ij > - Zij j)
= Rij j + SCij j + Pij j + j
+    ( Zijj) /  (Zijj) 
if (ij, ij)  bivariate normal 0,0,1, , 
Where:
 = is the probability density function
 = is the cumulative distribution function of the standard normal distribution
 = is the correlation coefficient between ij and ij
13
“Treatment effect” model, inspired by
“Two-steps” correction model of Heckman
The achievement equation regresses the whole set of variable on achievement
A, which are (looking at our database):
X=Z+
STRATIO = student/teacher ratio
SCHSIZE = size of the school (school total enrolment)
PRBUILD = school with poor buildings
LCTMAT = lack of teaching materials
PHISEI = socio-economic status of the peer group
The two equations contain the same set of individual variable;
Choice equation = Z
Achievement equation = X + Z (so X that contains also Z)
In order to avoid identification problems, the authors add to the equation a
variable that operates like an instrument which is not related to A, that is
PERCPUBL, which indicates the percentage of a school’s total resources
that are of a public origin
14
“Treatment effect” model, inspired by
“Two-steps” correction model of Heckman
Finally, the expected achievement of students who
choose public school is
E(Aij│PRIVij=0) = Rjjaj +SCijbj + Pjjcj + ρσε
[(Zijζj )/1-(Zijζj )]
And the difference between two types of schools in
country j is
E(Aij│PRIVij=1) - E(Aij│PRIVij=0) = dj + ρσε
[ (Zijζj )/1-(Zijζj )]
15
Second Section: Data set and
estimation strategy
Presentation
of the PISA programme
Presentation of the data set through
some main tables
Presentation of the estimation
strategy
16
What is PISA?
PISA is an internationally standardized assessment developed
jointly by the participating countries. PISA aims to assess
competences in Maths, Reading and Science, not so much to
evaluate the comprehension of the scholastic programme, but
rather to evaluate the capacity to develop these competences
also in the lifestyle.
In particular, they aim to evaluate the understanding of concepts
and the ability to apply them in different areas and situations.
The assessment is via written examination which can be through
multiple choice and through open questions.
Also the head teachers have to answer to a questionnaire in order
to provide information about the scholastic institution in which
students belong.
It is important to underline that the evaluation occurs every three17
years.
What is PISA?
Through this tool emerge:
- The outline of knowledge and skills of students
- Context indicators which relate the characteristics of schools and students
- Trend indicators that show changes in performance over time
Mathematical Literacy:
It implies the ability of an individual to identify and to apply the role of maths in the
real world. It is important that a student compare themselves with Maths in order to
satisfy needs that life as a citizen requires.
Reading Literacy:
It implies the ability of an individual to understand und utilize texts in order to reflect
on their content. In this way it is possible that a student develops own skills and
potentials and therefore they are able to perform an active role in the society.
Scientific Literacy:
It implies the ability to use scientific competences in order to understand the natural
world around them and the changes that science brings to human activity and
therefore makes decisions on them.
18
What is PISA?
Through this instrument emerges other context information
as:
 Family origin of students
 Socio-economic background of students and families
 Habits and attitudes of students in school and families
 Characteristics of schools as; quality of human resources and
materials, public or private funding, teacher recruitment.
 Educational context; institutional structures, class sizes,
parental involvement.
 Teaching strategies
 Learning methods
 Motivation and involvement of students
19
The data set
To carry out the analysis the authors selected countries for which
the number of students sampled and attending private school is
above a 10% threshold.
This lead to a subset of countries containing: Netherlands,
Belgium, Mexico, Ireland, Spain, France, Denmark, Austria
and Brazil.
The authors make a restriction to the states on which make
inquiries concerning the analysis. Justification for this
restriction are twofold:
1)It makes no sense, statistically speaking, to assess a private
school effect in a particular country using test scores of just of
few dozen students.
2)Policy makers who currently discuss the opportunity to expand
the private sector (for example using vouchers) are interested
in knowing whether private schools make a difference when
20
attended by a large and heterogeneous population.
Main Tables
21
Main Tables
22
Main Tables
23
Main Tables
24
Estimation strategy
Finally, I would like just to say some words about the estimation
strategy used by the authors in the regression; the authors
used scores that have been standardized by countries (for
instance, each country mean score = 0 and standard deviation
= 1) in order to measure gross differentials.
In this way they want to compare the average values of maths,
science and reading test scores of students for each type of
school. Then, they used the traditional OLS model to get a first
estimate of the net private school effect (for instance
controlling for level of resources, socio-economic status and
peer endowments). In the last step they used to estimate a real
“treatment effect” model and contrary to OLS approach, this
should produce unbiased estimates of private school effect,
controlling for selection bias
25
Third section: Results and Analysis

Reading the results using tables:
Looking at these tables we could see three results of
interest:
- the gross score differential between private and public
students
- the coefficient associated to the PRIV dummy () in an
OLS regression model without control for selection biases
- the coefficient associated to the PRIV dummy () in the
treatment model
26
Main Tables
27
Main Tables
28
Main Tables
29
Main Tables
30
Conclusion
-
-
Even when selection biases are taken into account, the type of school (private of public)
attended can have a significant effect on 15 year-olds' academic achievement.
We conclude to the absence of systematic advantage to private schools.
Yet the conclusion is slightly different if one looks at the results from a country perspective
How come that the advantage to private schools holds only for some countries? Conversely,
how can one explains that in some other countries privately run schools seem to deliver
less than public ones?
1) The “organizational” interpretation of achievement difference. Following this line of
reasoning, private school in Brazil, Ireland or Belgium could possibly perform better because
they are granted more autonomy. And maybe private schools have less autonomy than public
ones in the Netherlands or Austria.
2) The “cultural” interpretation of private/public school difference. Rather than talking about
“private schools” effects, it might make more sense – at least in some countries like Ireland or
Belgium - to talk about “religious” schools effect. Indeed, a majority of private schools are,
by religiously affiliated boards (the author focuses on the example of Catholic Schools).
According to this cultural interpretation, the (marginally) better training received in private
schools could be explained by religious values. In fact, the main religions existing today
enhance values such as hard work, effort and dedication to a task.
31
Download