Session 6 c discussion of the edu s ti k for italy paper - iariw
1. Discussion of the paper:
Who benefits from public education
provision? Evidence from Italy
Authors: Francesco Andreoli, Giorgia Casalone and
Daniela Sonedda
Discussant: Gyorgy Gyomai
IARIW 2014, Rotterdam
2. Income and STiK for education by quintiles
(Italy 2008, euro billions)
18
15
12
9
6
3
0
600
500
400
300
200
100
0
Q1 Q2 Q3 Q4 Q5
Primary Income
Cash Disposable Income
Disposable Income
STIK Education (right scale)
3. The structure of the paper
1. presents a micro-model of two households to illustrate
utility implications of the existence of public education, and
conditions under which rich households opt for private
education while poor households stay with the public system
(Besley and Coate)
2. Describes datasets, estimations needed to complement the
datasets
3. Estimates variations of quantile treatment regressions and
discusses the results.
4. The model
• Besley and Coate 1991 AER
• Public provision of education
as STiK will lead to a welfare
loss, unless the provided
education quality is precisely
chosen
• The provided quality has to
be inferior to what the rich
households would choose,
to push this household to
purchase private education
(redistributive effect)
5. Datasets and preliminary calculations
• INVALSI – MIPA study provides educational STiK information
for 2003
• Survey on Household Income and Wealth for 2004 (trimmed)
• STiK for education received:
푘푐 = 휔 푔ℎ ∙ 퐴퐶 푟푐 , 푒푐 for 푐 ∈ ℎ
• ω is estimated using the ISTAT’s Multiscopo Survey from 2005
using region, age-group, occupational and education status of the parents
• Equivalisation
– ALL (Aaberge, Langloren and Lindgren 2013) for STiK
– EU scale for income (Oxford modified, OECD scale)
• Earning capacity - a proxy of primary income?
Estimated based on likelihood and maximum size of various tax
deductions by age group and gender
13. Comments
• On the angle chosen:
The redistribution – even within the relatively narrow scope of STiK for
education – is guaranteed by many channels (progressive taxation,
contribution of families with no children, public education quality that
leads to a separating equilibrium), and of these channels perhaps
choosing a quality level that leads to a separating equilibrium is the least
important for the policy maker.
• Can the model work with data imputations based mostly on
the variables that later on are used as explanatory or control
variables?
Editor's Notes
Equivalisation: Why? to put together the various education levels; but this may mask deliberately different quality choices of the government for the diffferent layers, and mixes therefore responses. The authors do a separate estimation on STiK by edu level for their main results - and analysis of that is preferable to the equivalised figures. The termi.nology should try to avoid refering to it as SNA
Isn’t there a problem of ‘multicollinearity of sorts’ in the control function variant?