Transcript lecture 7 - strategies in technology based industries

Changes in the extent of inequality in access to
university education in Italy
Giorgio Di Pietro
University of Westminster
London
11 December 2008, Zaragoza
Motivation

Several studies (Evans and Schwab, 1995; Nguyen and Taylor,
2003) show that people from poorer family backgrounds are less
likely to have access to university education than those from
richer family backgrounds.

Equality of educational opportunity has been for a long time at the
top of the policy agenda in Italy and several measures have been
adopted in an attempt to encourage participation of individuals
from less advantaged backgrounds.

However, in spite of all these provisions, there is evidence that
Italy shows a high degree of intergenerational transmission of
education (Comi, 2004).
Objective of the Paper

This paper attempts to look at changes in the university
enrolment gap between people from richer and poorer family
backgrounds in Italy between 1995 and 2001.

The period covered by this study is especially relevant given
that in 2001 the Italian university education underwent a
significant process of reform. One of the objectives of this
reform is to induce behavioural changes that may make people
from less privileged backgrounds more likely to enrol at
university.
Data

The data come from three waves (i.e. 1998, 2001 and 2004) of a
national cross-sectional survey (Percorsi di Studio e di Lavoro dei
Diplomati) carried out by the Italian National Statistical Institute.
Each wave consists of a representative sample of high school
leavers who are surveyed three years after their completing their
studies. Thus these data enable us to examine the post-high school
decisions made by three different cohorts of individuals who
completed their studies in 1995, 1998 and 2001.

We consider different individual characteristics that are likely to
affect the decision to enrol at university including age, gender,
high school final grade, type of high school. We also include
indicators for labour market conditions such as unemployment rate
and wage differential between university and high school
graduates.
Mean of the variables used in this study
1995 cohort
1998 cohort
2001 cohort
Enrolled at university
0.506
0.407
0.524
Family background
Parents' education- Highest educational attainment
Highest educational attainment is a university degree
Highest educational attainment is a high school diploma
Highest educational attainment is less than a high school diploma
0.108
0.359
0.533
0.088
0.363
0.549
0.116
0.430
0.454
0.549
0.546
0.535
0.092
0.592
0.316
0.046
0.642
0.312
0.041
0.656
0.303
0.393
0.403
0.204
0.419
0.382
0.199
0.354
0.390
0.256
School-characteristics
Type of upper secondary schools
General high school
Professional schools
Technical schools
Others
Private school
0.269
0.195
0.427
0.109
0.112
0.192
0.394
0.352
0.063
0.065
0.191
0.294
0.410
0.105
0.073
Individual ability
High school diploma classification
60-69
70-79
80-89
90-100
Repeted high school year
0.336
0.219
0.135
0.119
0.246
0.357
0.302
0.191
0.149
0.215
0.359
0.274
0.179
0.189
0.245
Labour market conditions
Unemployment rate (%)
Wage differential between university and high school graduates
0.364
1.526
0.347
1.469
0.270
1.455
17,969
19,558
19,699
Individual personal attributes
Female
Age
21 years old or less
22 years old
23 years old or more
Area of residence
South
North
Centre
N
Parental education as a proxy
for family background

In order to measure the difference in the propensity to enrol at
university among individuals from different family
backgrounds, we split them into three groups according to
their parents’ education.
1.
The first group refers to individuals who have at least one
of their parents with a university degree.
The second comprises individuals whose parents’ highest
educational attainment is a high school diploma.
The third group includes individuals whose parents’
highest educational attainment is less than a high school
diploma.
2.
3.
Three indicators for educational inequality

Given our choice of the proxy for family background, we use three different
indicators of the extent of educational inequality:
I1: the gap in university participation between individuals who have at least
one of their parents with a university degree and those whose parents’ highest
educational attainment is less than a high school diploma.
I2: the gap in university attendance between individuals who have at least one
of their parents with a university degree and those whose parents’ highest
educational attainment is a high school diploma.
I3: he gap in university enrolment between individuals whose parents’ highest
educational attainment is a high school diploma and those whose parents’
highest educational attainment is less than a high school diploma.
University enrolment and family background
1995 cohort
1998 cohort
2001 cohort
36.30
60.54
87.86
27.62
50.00
83.68
37.28
59.42
85.78
I1
51.56
56.06
48.50
I2
27.32
33.68
26.36
I3
24.24
22.38
22.14
University enrolment by parents' education (%)
parents' highest educational attainment is less than a high school diploma
parents' highest educational attainment is a high school diploma
parents' highest educational attainment is a university degree
Differences in university enrolment by parents' education (%)
University enrolment logits by cohort of high school
leavers- Marginal effects (S.E. in brackets)
1995 cohort
1998 cohort
2001 cohort
Parental education- Reference category is highest educational attainment is less
than a high school diploma
Highest educational
attainment is a
university degree
0.381
(0.021)
0.388
(0.020)
0.374
(0.018)
Highest educational
attainment is a high
school diploma
0.169
(0.010)
0.164
(0.009)
0.157
(0.009)
0.275
0.285
0.270
Pseudo R-squared
Controls include: constant, gender, age, area of residence, type of high school, high school grade, repeated high
school year, labour market conditions
Towards a decomposition analysis

The descriptive statistics show that in the 1995 cohort I1 is 51.6%. After
standardizing for observed characteristics in the logits, this enrolment gap
is reduced to 38.1%. This means that 13.5% out of the 51.6 % (i.e. 26.2%)
is explained by the difference in observed characteristics. The unexplained
residual (63.8%) is due to differences in unobserved traits of people from
different family backgrounds.

Nevertheless, this method of inferring the extent of the unexplained gap in
university enrolment probability is unsatisfactory as it assumes that people
from different family backgrounds behave in a similar way, apart from a
shift factor given by family background.

A more satisfactory technique is represented by the Oaxaca (1973)
decomposition method.
Decomposition- Method
-Perform logit estimates of the probability of enrolling at university by cohort of high school
leavers and family background
-Compute the predicted probability of enrolling at university for each typical individual
from a specific family backgrounds by cohort of high school leavers
-This allows us to compute the difference in these probabilities following our indicators
for educational inequality. For instance, as regards I1
1
1
P ( X t , t )  P ( X p ,  p ) 

( X t  t )
( X p  p )
1 e
1 e
subscripts t and p refer to those individuals who have at least one of their parents with a
university degree and individuals whose parents’ highest educational attainment
is less than a high school diploma, respectively. Xt and Xp indicate average characteristics
for these individuals, respectively.
Average predicted probabilites of enrolling at university by family background
1995 cohort
1998 cohort
2001 cohort
34.13
65.65
92.84
23.93
52.30
93.40
34.66
65.54
94.16
I1
58.71
69.47
59.50
I2
27.19
41.10
28.63
I3
31.52
28.37
30.88
Probabilites of enrolling at university by parents' education (%)
parents' highest educational attainment is less than a high school diploma
parents' highest educational attainment is a high school diploma
parents' highest educational attainment is a university degree
Differences in the probability of enrolling at university by parents' education (%)
Decomposition- Method-cont
P ( X t , t )  P ( X p ,  p ) 
1
1

1  e ( X t  t ) 1  e ( X p  p )
can be decomposed into:
=
P ( X , 
t
*
tp

 

)  P ( X p , tp )  P ( X t , t )  P ( X t , tp )  P ( X p , tp )  P ( X p ,  p )
*
*
*
*
Next define  tp the university enrolment structure that would prevail in
absence of unobserved family background differences.
Whilst the first term in brackets represents the explained part of I1, the
second and third terms together constitute the unexplained part of I1.
Decomposition of the family background gap in enrolment probability
1995 cohort
1998 cohort
2001 cohort
Total difference
58.71
69.47
59.50
Part explained by observed characteristics
Percentage of the total difference explained by observed characteristics
43.21
73.60
49.23
70.86
42.37
71.21
Part not explained by observed characteristics, i.e. residual
Percentage of the total difference not explained by observed characteristics
15.50
26.40
20.24
29.13
17.13
28.79
Total difference
27.19
41.10
28.63
Part explained by observed characteristics
Percentage of the total difference explained by observed characteristics
18.89
69.48
28.19
68.59
20.13
70.32
Part not explained by observed characteristics, i.e. residual
Percentage of the total difference not explained by observed characteristics
8.30
30.53
12.91
31.41
8.50
29.69
Total difference
31.52
28.37
30.88
Part explained by observed characteristics
Percentage of the total difference explained by observed characteristics
16.34
51.84
13.58
47.86
15.47
50.10
Part not explained by observed characteristics, i.e. residual
Percentage of the total difference not explained by observed characteristics
15.18
48.16
14.79
52.12
15.41
49.91
I1
I2
I3
Results from the decomposition analysis

The major part of the university attendance gap between people
from richer and poorer family backgrounds can be attributed to
differences in observed characteristics.

Whilst in the 1998-2001 period a reduction in the university
enrolment gap has been accompanied by a decrease in the
contribution of unquantifiable differences between people from
richer and poorer family backgrounds, this conclusion does not
hold if one compares results associated with the 1995 and 2001
cohorts.
Conclusions

This paper has attempted to examine changes in the university
participation gap between people from richer and poorer
family backgrounds. There are three main results:

First, while university attendance of people from richer family
backgrounds is pretty stable over time, university participation
of people from poorer family backgrounds exhibits significant
variations. Attendance from the former group of individuals
tends to rise over periods of high university participation rate,
while it falls over periods of low general university attendance.
.
Conclusions-cont

Second, educational inequality, measured as differential
university attendance rates between individuals from richer
and poorer backgrounds, has remained pretty stable between
1995 and 2001.

Third, our decomposition analysis indicates that there is no
consistent evidence that the 2001 university reform contributed
to a fall in educational inequality.