Your Move: The Effect of Chess on Math Test Scores

Report
TrygFonden’s Centre for Child Research
AARHUS DECEMBER 12, 2014
Your Move:
The Effect of Chess on
Math Test Scores
Kamilla Gumede
Michael Rosholm
Outline
-very preliminary, comments welcome
1.
2.
3.
4.
5.
6.
7.
Intro: Background, Aim
Existing Evidence
The Intervention
Data
Methodology
Results
Conclusion
Intro: Background
• Costs of primary and lower secondary schooling in
Denmark are among the highest in OECD
• Yet, looking at cross-country data from the PISA studies,
Denmark is always ‘average’ in OECD
• Shift in focus from cognitive skills towards
personality/behavioral factors and lack of selfcontrol/grit/conscientiousness as explanations of school
failure
Intro: Background
• Teaching chess may affect cognitive skills (fluid
intelligence)
• directly
• indirectly through non-cognitive skill formation
Intro: Aim
Aim: Design intervention to help children improve in school,
specifically in mathematics, through teaching chess
Think of the present study as a pilot study preparing for a
larger scale randomized trial.
• As such, it is severely underpowered
• Nevertheless, interesting results emerge
Existing Evidence (Gobet & Campitallei 2006; Bart, 2014;
Boruch, 2011; Berkman, 2004; and many others)
Chess is a sequential game, where the players make moves
in turn with white and black pieces on the chess board with
the aim of capturing the opponent’s king
Very complex
•
•
•
•
•
•
calculation and planning ahead
the ability to concentrate
memorizing sequences of moves and resulting positions
rewards the ability to exert patience and self-control
a set of rules of conduct
may directly increase intelligence and problem solving abilities
Existing Evidence
• Trinchero (2013): Impact of chess instruction on PISA
math test scores. Non-randomized data on children aged
8-10 in Italy. Chess instruction in school improves problem
solving abilities
• Boruch (2011): Only sufficiently powered RCT to date. 33
Italian schools, 30 hours of chess instruction, 3rd grade
classes. Chess instruction increases math achievement
by a third of a standard deviation. Foreign born pupils
have better impacts.
The Intervention
•
•
•
•
•
•
•
Five schools in the City of Aarhus, 1st to 3rd grade
Starting in January 2013 and ending mid October 2013
1st-3rd grade classes participate in the study as either treated or
control classes.
No random allocation… however, almost always ‘A’ class is
control and the rest is treated.
Students in treated classes have one weekly math lecture
replaced by a weekly chess lecture
• NOTE: Control classes have more maths!
One teacher doing all the teaching
A book developed by Dansk Skoleskak: SKAK+MAT was used
The intervention
Figure 1. A typical chess exercise from the book used for chess instruction
Accompanying text:
“How many pieces can the knight
take? Write your answer on the
line below”
Data
Pre- and post math tests developed by City of Aarhus
• Calculation
• Problem solving
• Pattern recognition
Merged with register based info at TrygFondens Centre for
Child Research
Variable
Standardized pre-intervention test-score
Boy
Girl
Age
1st or 2nd generation immigrant
Days of school absence 2012
Grade 1
Grade 2
Grade 3
# siblings
Age of mother
Mother lower secondary school
Mother highschool
Mother vocational education
Mother short academic education
Mother medium academic education
Mother masters education or more
Mother’s average ann. earnings past 3 years, DKK
Mother not working 2011
Father present
Age of father
Father lower secondary school
Father highschool
Father vocational education
Father short academic education
Father medium academic education
Father masters education or more
Father’s average ann. earnings past 3 years, DKK
Father not working 2011
N
Treatment group
Comparison group
0.00
0.54
0.46
9.57
0.28
9.21
0.19
0.45
0.36
1.46
40.53
0.42
0.07
0.27
0.05
0.07
0.11
195,276
0,29
0.79
43.28
0.19
0.05
0.30
0.08
0.14
0.17
278,564
0.19
323
0.05
0.50
0.50
9.45
0.25
9.94
0.31
0.33
0.36
1.53
40.42
0.41
0.08
0.22
0.06
0.08
0.14
188,578
0,30
0.76
42.28
0.15
0.04
0.30
0.10
0.13
0.18
284,405
0.20
159
Data
0
.1
.2
.3
.4
Kernel density estimate, change in testscores
-4
-2
0
diffstdscore
Treatment
Comparison
kernel = epanechnikov, bandwidth = 0.2642
2
4
Methodology
Level model with control for pre-test scores, child’s
characteristics, parental characeteristics, and school fixed
effects
1 =  + 0 +  + 0 +  + 
(1)
‘Learning’ model, change in the test score is explained by the
same variables
1 − 0 =  +  + 0 +  + 
(2)
+ D-i-D PS mathching (Heckman, Ichimura, Smith, Todd, 1998)
Results
Table 3. Estimation results, change in standardized test-scores
Model
2: Only chess
dummy
3: 2+child and
school
characteristics
4: 3+mother’s and
father’s
characteristics
Impact estimate
0.18
Standard error
0.09
R-squared
0.01
0.16
0.09
0.04
0.17
0.09
0.08
Results summarized
• Positive effect for boys (0.22) but NS for girls
• Positive effect for native Danes (0.22) but not for 1st or 2nd
generation immigrants
• No interactions significant across grades, pre-intervention
test scores, or morther’s educational attainment
• Positive effects on pattern recognition (~fluid intelligence)
No impact on calculation or problem solving
• No impact on school absence after program (during?)
Conclusion
Are the impacts large?
Let’s do the RCT! Then we’ll conclude

similar documents