Full-day vs half-day kindergarten Long run effects on test scores Christine Neill & Jean Eid WLU, Economics.
Download
Report
Transcript Full-day vs half-day kindergarten Long run effects on test scores Christine Neill & Jean Eid WLU, Economics.
Full-day vs half-day
kindergarten
Long run effects on test scores
Christine Neill & Jean Eid
WLU, Economics
Plan of talk
Benefits/costs of full-day kindergarten
How to estimate effects on test scores?
Difference-in-Differences / quasi-experiment
methodology
Some data (at the school board level)
Initial results
Little evidence of any effect on Grade 3 test
scores
Benefits of full-day kindergarten
– government reports
“Premier Dalton McGuinty says fullday kindergarten programs will be
available for every child by 2015, and
will increase students' chances of
completing university, attending postsecondary education and landing a
good job.”
• http://www.cbc.ca/canada/toronto/story/20
10/09/07/toronto-first-dayschool.html#ixzz13g20HZjd
Benefits of full-day
kindergarten
Parents are better able to work full time
Children learn more, are better
prepared for grade 1 and ….
This makes life easier for Grade 1
teachers!
Students perform better
Students learn more quickly
Peer effects?
Costs of full-day
kindergarten
Large financial cost to taxpayers
$1.5 billion a year
Does not include marginal cost of
public funds
Socialisation, or transition costs of
starting full-day kindergarten at a
young age
Baker and Milligan – less healthy;
more parent-reported temper tantrums
Previous studies
Studies of Perry Pre-school, or ABCDarian, etc, show
positive effects of targeted, intensive pre-school
programs on grades and later social and economic
outcomes of at-risk children
daCosta&Bell (2000; 2001)
Not universal - students in full-day were from inner
city low SES schools in Alberta; full day students
improved more over the kindergarten year
Herry, Maltais & Thompson (2007)
Full-day kindy in one French ON school board. 1 year
pre and post data. Teacher and parent-reports.
De Cicca (2007)
US: slight positive effect on test scores at start of year
1, nothing by end of year 1 (“long term”)
Methodology– less reliable
methods
Comparing one cohort of students who went to full-day
kindergarten with a group before who went to half-day
kindergarten
Something may have changed between the two years
Comparing a group of students who attend a full-day
program with another who attend a half-day program
The two groups of students could just be different
There could be non-random selection into the full-day
program
Only a problem if selection or differences are
unobservable
Methodology – “difference in
differences”
So look for a change in policy that
affected some students, but not others
“Difference in Differences” study
• Control for “fixed effects” at the school or
school board level, to deal with
unobservable fixed characteristics
• Control for “year effects” to deal with
unobservable time variation
Basic research plan
How did the switch from half to full-day
kindergarten in Ontario’s French language
school boards affect EQAO test scores?
We can use English language boards as a
control
Mobility between French and English boards
is less of a problem than mobility between
English language schools
We can look at longer-term outcomes
(though we miss very short-term outcomes)
A (very) little background
All French language boards have offered
full-day JK and SK programs since around
2001-02
Varying start dates (we use this in our
estimation)
“Educationally … initiated because children
in French schools obtained lower grades
than students in English schools on the
provincial reading, writing, and mathematics
exams taken by students in the third and
sixth grades.” (Herry, Maltais, Thompson,
2007)
Diff in diff – no effect
Performance
measure
Policy Change
English
French
Time
Diff-in-Diff – +ve effect
Performance
measure
Policy Change
French
English
Time
How do we estimate this?
Simple linear model, no break:
Ybt = a0 + a0Frenchb + b0Timet + ebt
Simple linear model, with break:
Ybt = a0 + a0Frenchb + b0Timet
+ gPostPolicyt*Frenchb + ebt
In actuality, use YearsFTKindybt as our key policy
variable (mostly 1999-2000; some later)
Add French*time to allow for different pre-trends in
French boards (as per graphs)
Standard errors clustered at school board level
The data
Currently: EQAO data at the school board
level (72 boards, 60 English and 12 French)
% achieving each level
number of students
%Female/%Male
% Second Language Program
% Special needs
In future: At the individual level
Allows us to examine whether there are
‘heterogeneous treatment effects’ (do kids
from low SES or non-French backgrounds
benefit more from full-day kindergarten?)
Grade 3 Reading
70
% of all students
65
English
60
55
50
French
45
40
35
30
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Test year
Grade 3 Writing
90
80
% of all students
French
70
60
English
50
40
30
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Test year
Grade 3 Maths
75
70
% of all students
English
65
60
French
55
50
45
40
35
30
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Test year
Grade 6 Reading
90
80
% of all students
French
70
60
English
50
40
30
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Test year
Grade 6 Writing
90
80
% of all students
French
70
60
English
50
40
30
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Test year
Grade 6 Maths
90
80
% of all students
French
70
60
English
50
40
30
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Test year
Numerical results
Reading
Writing
Math
58.97
86
36
63.44
88
24
Dependent variable: percentage achieving Level 3 or Level 4
English boards mean:
English boards max:
English boards min:
Years of full-day kindy
56.5
82
28
0.6
(1.22)
3.1
(6.00)**
4.5
(2.02)*
-2.7
(0.26)
16.4
(2.13)*
-1.1
(0.73)
0.567
(1.06)
4.6
(2.03)*
-3.4
(0.32)
16.1
(2.10)*
709
72
0.87
709
72
0.87
French Trend
Female/Male ratio
% Second language program
% SpecialNeeds
N
Clusters
R2
4.3
(6.34)**
5.6
(2.92)**
-8.1
(1.78)
15.7
(2.85)**
1.2
(1.27)
0.615
(1.96)
5.7
(3.05)**
-8.9
(2.15)*
15.4
(2.80)**
1.9
(0.76)
-3.8
(0.33)
24.5
(1.97)
3.6
(1.82)
0.238
(0.35)
2.0
(0.77)
-4.1
(0.35)
24.5
(1.97)
709
72
0.92
709
72
0.92
708
72
0.84
708
72
0.84
Note: Clustered t-statistics in parentheses. ** = significant at 1% level; * significant at 5% level.
Interpretation: percentage point increase in % achieving the level for each additional year of full-time kindergarten
Numerical results
Reading
Writing
Math
5.7
15
9.4
25
Dependent variable: percentage achieving Level 4
English boards mean:
English boards max:
Years of full day kindy
5.5
14
2.0
(6.02)**
2.1
(5.71)**
1.5
(1.74)
0.1
(0.03)
3.7
(1.52)
-0.7
(1.17)
0.934
(3.99)**
1.7
(2.10)*
-1.0
(0.38)
3.3
(1.33)
709
72
0.67
709
72
0.7
French Trend
Female/Male ratio
% Second language program
% SpecialNeeds
N
Clusters
R2
1.4
(3.33)**
1.5
(1.38)
-0.8
(0.41)
2.4
(1.02)
-0.9
(1.66)
1.003
(6.18)**
1.7
(1.85)
-2.0
(1.58)
2.0
(0.91)
2.3
(1.68)
-4.3
(0.73)
10.1
(1.60)
0.0
(0.03)
0.454
(1.29)
2.4
(1.74)
-4.8
(0.85)
10.0
(1.61)
709
72
0.89
709
72
0.9
708
72
0.76
708
72
0.77
Note: Clustered t-statistics in parentheses. ** = significant at 1% level; * significant at 5% level.
Interpretation: percentage point increase in % achieving the level for each additional year of full-time kindergarten
Concerns
Interpretation of EQAO scores is non-standard
ordered probit model, if use individual data?
Selection
Inadequate controls
Mobility across schools
Typically more than 80% of students had been in the
same French board for 3 or more years
Difficult to transfer into the French system from the
English system
Heterogeneous treatment effects
Need individual-level data – PEDAL at McMaster
Conclusions so far
French schools closed the gap with English
schools at Grade 3 in all three EQAO tests
But this mostly appears due to an upward
trend in French schools’ test scores, rather
than an obvious break for those who had
full-day kindergarten
Some positive effect on mathematics at the
provincial standard level?
Not likely that there is anything in the Grade
6 aggregate scores