Barrow Evidence Computerized Algebra Instruction December 2007
description
Transcript of Barrow Evidence Computerized Algebra Instruction December 2007
Technology’s Edge: The Educational Benefits of Computer-Aided Instruction
Lisa BarrowFederal Reserve Bank of Chicago
Lisa MarkmanPrinceton University
Cecilia RousePrinceton University and NBER
Student achievement is critical for individuals and society
• U.S. math skills have been increasing (NAEP), but proficiency levels remain low.
• Math skills may explain a large portion of wage inequality (Grogger, 1996; Murnane, Willet, & Levy, 1995)
School districts are turning to advances in computers to…
– Reduce administrative burden;
– Compensate for poor teacher content knowledge (especially in districts that report difficulty recruiting and retaining teachers, particularly in math and science);
– Allow more individualized student attention; students can progress at own pace.
Evidence on effectiveness of CAI is surprisingly weak
– Poorly defined computer use.
– Little use of randomized controlled study design.
In a 2001 review of the literature, Larry Cuban (2001, p. 179) concludes,
“When it comes to higher teacher and student productivity and a transformation of teaching and learning … there is little
ambiguity. Both must be tagged as failures. Computers have been oversold
and underused, at least for now.”
Computer-Aided Instruction (CAI) Treatment:
• Typically used in large urban districts;• Software and hardware package designed
to deliver one-on-one instruction;• Software described as “meeting National
Council of Teachers of Mathematics standards”;
• Software may be configured to state standards;
• Includes classroom management tools.
The Program
• 5 components per lesson:– Pretest– Review– Lesson– Cumulative review– Comprehensive test
• Required to achieve certain degree of mastery before advancing;
• Teachers can monitor student progress.
The design of this experiment…
• Randomly assign 8th and 9th grade classes in three large urban districts to be taught using computer-aided algebra and pre-algebra instruction.
• Assess the impact on statewide tests and tests designed to target algebra and pre-algebra skills.
• Note that the computer use was well-defined.
Research Design: Within-school random assignment to
CAI or traditional instruction at the classroom level in three districts
Research design addresses two forms of selection bias:
• Non-random assignment of students to CAI;• Non-random assignment of teachers to CAI.
Table 1: Characteristics of the Districts in Our Study
Top 100 Districts
3 Districts Combined District 1 District 2 District 3
# Students 112,807 62,507 ~68,000 ~22,000 ~97,000
% Female 48.8 49.4 49.7 48.8 49.3
% Native American 0.6 0.5 0.1 0.1 1.0
% Asian 7.1 3.1 1.9 0.8 4.4
% African American 28.1 69.5 93.6 40.3 59.4
% Hispanic 34.1 16.2 1.1 54.3 18.0
Implementation of Random Assignment
• Schools provided us with class schedules of pre-algebra and algebra classes;
• Given option of eliminating particular teachers or classes from the experiment;
• Randomization information provided to schools after students’ classes were scheduled.
Teacher Period 1 Period 2 Period 3 Period 4
A Alg 1 Alg 1A Alg 1
B Alg 1 Alg 1A Alg 1A
C Alg 1A Alg I
D Alg I Alg 1 Alg 1A
Randomized Evaluation of a Computerized Math Curriculum
Current School Schedule
Teacher Period 1 Period 2 Period 3 Period 4
A Alg 1 Alg 1A Alg 1
B Alg 1 Alg 1A Alg 1A
C Alg 1A Alg I
D Alg I Alg 1 Alg 1A
Teacher Period 1 Period 2 Period 3 Period 4
A Regular Regular LAB
B LAB LAB Regular
C Regular Regular
D Regular Regular LAB
Randomized Evaluation of a Computerized Math Curriculum
Current School Schedule
Below is a sample schedule that would be returned to each school after random assignment
Table 2a: Schools and Students in Study – District 1Relevant Schools
Schools in Study
Students in Study
Number of Students 29,603 8,148 973
% Grade 8 19.3 16.8 40.4
% Grade 9 18.0 18.3 47.2
% Grade 10 15.1 17.8 4.4
% Female 50.5 49.0 52.0
% African American 94.2 97.2 87.8
% Hispanic 1.0 0.8 0.8
% White 2.6 0.4 0.1
% Native American <0.1 <0.1 0.0
% Asian 2.2 1.6 1.8
% Missing demographic data
9.6
Table 2b: Schools and Students in Study – District 2Relevant Schools
Schools in Study
Students in Study
Number of Students 5,270 4,476 412
% Grade 8 2.3 0.0 0.0
% Grade 9 38.0 40.0 52.7
% Grade 10 22.0 23.2 31.8
% Female 48.4 48.2 46.7
% African American 43.6 42.0 47.1
% Hispanic 50.1 51.2 44.7
% White 5.5 5.9 6.6
% Native American 0.2 0.1 0.2
% Asian 0.7 0.8 0.5
% Missing demographic data
0.2
Table 2c: Schools and Students in Study – District 3Relevant Schools
Schools in Study
Students in Study
Number of Students 27,572 3,540 200
% Grade 8 1.4 0.0 3.5
% Grade 9 35.6 40.0 91.5
% Grade 10 23.3 25.1 3.0
% Female 49.9 47.6 47.7
% African American 61.1 92.5 94.5
% Hispanic 15.2 1.2 0.5
% White 18.3 4.0 1.5
% Native American 1.1 0.4 0.0
% Asian 4.5 1.9 3.0
% Missing demographic data
0.5
Numbers of Schools Classes, Teachers, and Randomization Pools
Analysis Sample
Combined District 1 District 2 District 3
Number of schools 17 10 4 3
Number of randomization pools
60 31 19 10
Number of classes 141 74 44 23
Number of teachers 57 36 14 7
Number of students 1,585 973 412 200
Outcomes
• Algebra and pre-algebra tests by Northwest Evaluation Association (NWEA) to be consistent with state and district standards;
• State-wide administered math tests;
• District benchmark tests in pre-algebra.
Table 3: Randomization of Treatment and Control Using Full Sample
Random Assignment
TraditionalInstruction CAI
p-value ofdiff.
Full Sample
Baseline algebra test score
24.7 24.7 0.494
Female 47.2 47.1 0.637
African American 80.0 83.2 0.561
Hispanic 15.9 13.5 0.195
Class Size 25.8 25.7 0.860
# of Observations 1,133 1,145
Table 3 (cont.): Randomization of Treatment and Control Using Analysis Sample
Random Assignment
TraditionalInstruction CAI
p-value ofdiff.
Analysis Sample
Baseline algebra test score
24.7 24.8 0.304
Female 51.1 48.9 0.148
African American 81.9 84.0 0.060
Hispanic 13.8 12.1 0.061
Class Size 25.8 26.2 0.549
# of Observations 785 800
Intent-to-Treat Estimates of the Effect of CAI on Algebra Achievement
(with and without Teacher Fixed Effects)
0.17
0.37
0.17
0.28
0
0.1
0.2
0.3
0.4
0.5
0.6
Without Teacher Fixed Effects With Teacher Fixed Effects
Est
imat
ed In
tent
-to-T
reat
Effe
ct S
ize
Advantages/Disadvantages of the Intent-to-Treat Effect
• Represents the gains a policy maker might reasonably expect to observe.
• Does not necessarily represent the effectiveness of the program.
Amount of Time in the Computer Lab by the Random Assignment of the Student’s Class
80.1
10.3 109.1
70.3
5.6
52.7
14.8
64.5
33
55.3
83.8
0
10
20
30
40
50
60
70
80
90
Number oflessons studentsare expected to
complete
% of stud.completing nolessons in CAI
% of stud.completing >10lessons in CAI
% of stud.completing >20lessons in CAI
Nr. of lessonscompleted in CAI
Nr. of CAI lessonscompleted as a %
of courseexpecations
Traditional Instruction CAI
Intent-to-Treat and Treatment on the Treated Estimates of the Effect of CAI
(with and without teacher fixed effects)
0.173
0.283
0.249
0.417
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Without Teacher Fixed Effects With Teacher Fixed Effects Inte
nt-to
-Tre
at a
nd T
reat
men
t-on-
the-
Trea
ted
Effe
ct S
izes
Intent-to-Treat Estimates in District 1 Using Different Tests
0.26
0.604
0.381
0.226
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Algebra Scale Score 2nd Qtr Benchmark 3rd Qtr Benchmark State Math Test
Inte
nt-to
-Tre
at E
ffect
Siz
e
Intent-to-Treat Estimates in Districts 2 and 3 Using Different Tests
0.2
0.089
-0.124-0.062
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Algebra Scale Score State Math Test
Inte
nt-to
-Tre
at E
ffect
Siz
e
District 2 District 3
We might expect to see an advantage of CAI in…
• Classes where curriculum best suited to students;
• Larger classes;• Classes with more
disruptive students;• Classes with
heterogeneous students.
Differential Intent to Treat Effects of the Computerized Instruction on Pre-Algebra and
Algebra Achievement by Class Type
0.510.440.450.48
1.36
0.01 0.07 0.13
-0.31-0.23
-1.00
-0.50
0.00
0.50
1.00
1.50
2.00
2.50
3.00
All 3 districts Districts 1 & 2 District 1 District 2 District 3
Inte
nt-to
-Tre
at E
ffect
Siz
e
CAI Effect for Pre-Algebra CAI Effect for Algebra
Differential Intent to Treat Effects of CAI on Pre-Algebra and Algebra Achievement by
Baseline Test Score Quartile
0.2160.242
0.171 0.155
-0.1
0
0.1
0.2
0.3
0.4
0.5
Bottom baseline test scorequartile
2nd baseline test scorequartile
3rd baseline test scorequartile
Top baseline test scorequartile
Differential Intent to Treat Effects of CAI on Pre-Algebra and Algebra Achievement by
Individual Attendance Rates
0.439
-0.221 -0.051 -0.02
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Bottom baselineattendance quartile
2nd baseline attendancequartile
3rd baseline attendancequartile
Top baseline attendancequartile
Differential Intent-to-Treat Effects by Class Characteristic: Attendance, Class Size, and Class
Baseline Test Score S.D.
0.06
0.350.21
0.01
-1.1
0.412
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
ClassAvg.
Attend.Rate
Attnd.Rate OneSD below
Mean
Class size=25
Class size=15
Smallclasses
Largeclasses
Est
imat
ed In
tent
-to-T
reat
Effe
ct
Overall we find:
• On average, students in classes taught using CAI scored higher on algebra tests than students in traditionally-taught classes.
• The effects appear larger for students in larger classes (especially large, heterogenous classes), those with worse attendance rates, and those in classes with lower average attendance rates.
Cost-Benefit Analysis
Cost of CAI
# ClassesTotal
studentsClass size # periods
CAI labs needed
Cost/ student
School A 22 730 33.2 8 3.0 $218
School B 12 321 26.8 8 1.5 $245
District 1 74 1736 23.5 8 9.3 $279
Cost of reducing class size to 13
# ClassesTotal
studentsClass size # periods
# new teachers
Cost/ student
School A 22 730 33.2 6 5.7 $329
School B 12 321 26.8 6 2.1 $278
District 1 74 1736 23.5 6 9.9 $241