Income Effects of Education Interventions

	A	B	C	D	E	F	G	H	I	J
1	Study	Age of participants	SD test score gains	Years of fadeout	% Income Increase	% Income Increase per SD test score gain	Weight for line of evidence	Study Type	Source	Notes
2
3	Improving classrooms in the US
4	Grades 4-8 VAT	12	0.1	0	1.16	11.6		Natural experiment of entry of new teachers into a school.	https://www.nber.org/system/files/working_papers/w17699/w17699.pdf	Looks at value added of a classroom based on the outcomes of classmates. Effects fadeout to 1/3 after 3 years then persist.
5	Kindergarten Star	5	0.32	0	3	9.4		Random assignment of teachers and students.	http://www.equality-of-opportunity.org/assets/documents/star_paper.pdf	Public schools in Tennessee.
6	School Spending IV	6-18	0.49	0	7	14.4		IV	Jackson et al 2015	Separate studies looking at scores and incomes
7
8	Scholarships and Vouchers in LMICs
9	Colombia	15	0.2	1	8	30.0		RDD	Bettinger et al.	A separate paper administered tests 3 years after application. Therefore the education received was subject to on average 1 years of fadeout (with the most recent year having 0 fadeout, the previous 1 year, and the one before that 2 years). "The results from this survey show no significant enrollment differences between lottery losers and winners three years after application... We therefore administered achievement tests to a subset of the pupils surveyed. The test results suggest that, on average, lottery winners scored about .2 standard deviations higher than losers, a large but only marginally significant difference.
10	Chile	6-13	0.2	1	10	18.8		IV	Patrinos et al	The education received was subject to on average 1 years of fadeout (with the most recent year having 0 fadeout, the previous year having 1 year of adeout, and the one before that having 2 years)
11		14-18	0.2	1	0
12	Ghana	15	0.157	2.5	2.5	6.0		RCT	Duflo et al.	If the intervention had only effected the last year of high school, and the test was administered right after we would not need to account for fadeout. However, since the median year of high school was 1.5 years earlier than than the senior year, and the test was administered another year later, we assume that the effect faded out by 25%*2.5=62.5%" These results are based on oral tests administered as part of the 2013 in-person survey. Thus, these tests provide the effect after most study participants had completed or stopped going to senior high school but before participants had a chance to enroll in tertiary education. "
13
14	Preschool in the US
15	Perry	4	0.9	0.5	19	18.5		RCT	Heckman et al.	Effect size estimated in preschool tab. Fadeout based on a two year program, so a fadeout of 0 for the last year and 1 for the previous year, averaging 0.5 years.
16	Abecedarian	3	1.1	2	50	22.7		RCT	Campbell et al.	Fadeout based on a five year program, so a fadeout of 0 for the last year and 1 for the previous year, and one more for each previous year.
17	Headstart	4	0.3	0	5	16.7		RCT for scores, matched cohort for income	Thomson
18	CCPC	4	1	0.5	7	6.1		matched cohort	Reynolds et al.	Fadeout based on a two year program, so a fadeout of 0 for the last year and 1 for the previous year, averaging 0.5 years.
19
20	Cross-sectional regressions of individual cognitive ability
21	Hanushek individual income versus PIAAC score by country					25.0			Hanushek 2016	GiveWell and OP chose to use the Indonesia coefficient from this study because it was the closest in income to contexts where TaRL operates
22	Meta-analysis of correlation of cognitive ability to wages in LMICs					4.5			https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8573607/	Controls for years of schooling and sometimes industry. Many of our interventions will likely increase years of schooling as a result of boosting scores, therefore controlling for schooling makes this a very conservative estimate.
23	Income regressed against literacy scores					36.5			Evans 2017	Another study cited by Givewell. Table 6 looks at regressions in Bolivia, Colombia, Ghana, Kenya, Vietnam
24
25	Regressions of changes in GDP growth on changes in test scores		GDP change per stdev scores
26	Hanushek et al. 2012		2			50.0			Hanushek et al. 2012	Multiplied by 25 because growth assumed to continue during 25 years of program phase in
27	Goczek et al.		0.8			20.0			Goczek et al.
28	Komatsu and Rappleye		0.5			12.5			Komatsu and Rappleye
29
30	Averages for 5 lines of evidence
31	Improving Classrooms in the US					11.8	0.3	RCT, IV
32	Scholarships and vouchers in LMICs					18.2	0.35	RCT, IV and RDD
33	Pre School in the US					16.0	0.1	RCT
34	Cross-sectional regressions of individual cognitive ability					22.0	0.05	OLS	Cross Sectional Tab
35	Regressions of changes in GDP growth on changes in test scores					27.5	0.2	OLS and IV	Country Level Tab
36
37	Simple average across 5 lines of evidence					19.1
38
39	Weighted Average					18.1
40	Average excluding weakest studies and lines of evidence					17.5
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100