The impact of school quality, socioeconomic factors, and childhealth on students' academic performance: evidence from Sri
Lankan primary schools
Harsha Aturupane^ Paul Glewwe''* and Suzanne Wisniewski''
"The World Bank, Colombo, Sri Lanka; ''Department of Applied Economics, University ofMinnesota, Minnesota, USA; 'Department of Economics, University of St. Thomas,
Minnesota, USA
(Received 29 October 2009; final version received 26 Jtily 2010)
One of the eight Millennium Development Goals is that all ehildren in developingeountries should complete primary education. Mueh progress has been madetoward this goal, but eompleting primary sehool does not ensure that studentsattain basie literaey and numeraey skills. Indeed, there is ample evidenee thatmany children in developing eountries are not learning these basie skills. Thisraises the question: What ean sehools and eommunities do to inerease the learningthat takes plaee in sehools? Sri Lanka exemplifies these issues. It has aehieveduniversal primary eompletion, but many Sri Lankan primary sehool studentsperform poorly on aeademie tests. This paper uses unusually rich data from SriLanka to investigate the determinants of aeademie performanee, as measured byaehievement tests, of Grade 4 students. At the ehild and household level, edueatedparents, better nutrition, high daily attendanee, enrollment in private tutoringelasses, exereise books, eleetrie lighting, and ehildren's books at home all appearto inerease learning, while hearing problems have a strong negative effect. Amongsehool variables, principals' and teaehers' years of experienee, collaborating withother sehools in a 'school family,' and meetings between parents and teaehers allappear to have positive impacts on students' seores. Estimates that exelude someof the variables available in the unusually rieh data yield different results, whiehsuggests that results based on less eomplete data are likely to suffer from omittedvariable bias. A final seetion provides reeommendations for education policies inSri Lanka.
Keywords: education; health; Sri Lanka
1. Introduction
Academic economists and international development agencies claim that education isessential for economic growth and, more generally, a higher quality of life (Lucas1988; Barro 1991; Mankiw, Romer, and Weil 1992; World Bank 2001; UNDP 2003).One of the eight Millennium Development Goals is that all children in developingcountries should finish primary school. Yet developing country students who finishprimary school often have weak academic skills (Glewwe and Kremer 2006), whichjTiay sever the link between education and economic growth. This raises the question:What can developing countries do to promote leaming in their schools?
This paper investigates the determinants of leaming among Sri Lankan fourthgrade students. Sri Lanka has already attained universal primary completion (Bruns,Mingat, and Rakotomalala 2003), but many Sri Lankan students display weakacademic performance, and it is unclear what education policies would improve theirperformance. This paper uses unusually comprehensive and detailed data from SriLanka to study the impact of school quality, child health, and other factors on studentlearning in Sri Lanka.
The paper is organized as follows. Section 2 reviews the literature on leaming indeveloping countries. Section 3 describes Sri Lanka's educational system. Section 4discusses the data and methodological issues. Sections 5 and 6 present the results anddiscuss their policy implications, and Section 7 summarizes the findings andconcludes.
2. Previous research on academic performance in developing countries
Many studies have attempted to estimate the impact of school and teacher character-istics on student performance, yet most have serious estimation problems that castdoubt on their results (see Glewwe 2002; Glewwe and Kremer 2006). Most are 'retro-spective,' that is based on data collected from schools without attempting to alterthose schools' characteristics. Yet even the best retrospective studies offer onlylimited guidance due to their estimation problems, the most serious being omittedvariable bias and measurement error. This has led to wide variation in the estimatedimpacts of key variables. For example, of 30 developing country studies reviewed byHanushek (1995), 8 found significantly positive impacts of the pupil-^teacher ratio onstudent leaming, 8 found significantly negative impacts, and 14 found no impacts.
More recently, researchers have evaluated natural experiments and randomizedtrials. The former use 'natural' variation in a school characteristic that is unlikely tobe correlated with any other determinants of leaming. One example is assigningstudents by lottery to different schools. Three recent natural experiments suggest that:(1) increasing school resources (measured by student-teacher ratios) raises reading(but not math) skills among black South African students (Case and Deaton 1999); (2)vouchers that provide funds for Colombian secondary students to attend privateschools raise reading skills (Angrist et al. 2002); and (3) reducing class size raisesreading (and perhaps math) scores of Israeli students, but computers have no effect(Angrist and Lavy 1999, 2002).
Randomized trials have been implemented in several developing countries.Randomized studies in Kenya and Latin America show sizeable impacts on schoolenrollment from reducing schooling costs (including subsidies conditional on regularschool attendance). Regarding leaming outcomes, workbooks and radio instruction inNicaragua significantly increased pupils' math scores. Provision of textbooksincreased leaming in the Philippines, but in Kenya textbooks helped only the betterstudents. Evidence from Kenya also suggests little impact on test scores of reductionsin class size, flip charts, and deworming medicine, although school meals had positiveimpacts on test scores as long as teachers were well trained. A remedial educationprogram and computer-assisted leaming in India both appear to have increased leam-ing. For a comprehensive literature review, see Glewwe and Kremer (2006).
Research has shown that child health affects schooling outcomes. Such researchfaces serious econometric challenges, yet a few recent papers have used crediblemethods to quantify the impact of early childhood nutrition on schooling outcomes. A
4 H. Aturupane et al.
recent study found that providing dewonning medicine raises enrollment and dailyattendance in Kenya (Miguel and Kremer 2004). For a recent review, see Glewwe andMiguel (2008).
3. Primary education in Sri Lanka
This section reviews the education system and student academic performance in SriLanka, focusing on primary schools and the test perfonnance of Grade 4 students.
3.1. Sri Lanka's education system
Despite its low income ($1300 per capita in 2007), Sri Lanka has enrolled nearly allprimary aged (age 5-10) children in school; the net primary enrolment rate is 96%,and the primaiy completion rate is 95%. Gender equity also prevails; male and femaleenrollment rates are equal at all educational levels. These achievements reflect severalpolicies. First, in the 1950s Sri Lanka established an extensive network of free publicschools. Second, since the 1970s all students receive free textbooks and uniforms.Third, school enrollment has been compulsory since 1997 for children 6-14 years old(although families are rarely penalized for non-enrollment).
Sri Lanka's schools have two unusual features. First, private schools are rare (lessthan 5% of total enrollment); since the 1960s opening of new private schools has beenforbidden, although pre-existing private schools can continue operating. Second,schools typically offer both primary and secondary grades. More specifically,Sri Lanka has four types of schools: Type lAB schools offer Grades 1-13 and allthree curriculum streams (arts, commerce, and science); Type lC also offer Grades 1-13, but only two streams (arts and commerce); Type 2 offer only Grades 1-11; andType 3 offer only Grades 1-5 or 1-8. Almost all urban schools are Type lAB or lCand thus offer the full cycle (up to Grade 13). Many rural schools are also Type lABor lC, yet Type 2 (up to Grade 11) are also common. Other rural schools are Type 3and thus go up to only Grade 5 or 8. In the National Education Research and Evalua-tion Center (NEREC) data (described below), 46.3% of fourth graders attend schoolsoffering Grades 1-13, another 33.2% attend schools offering Grades 1-11, and 20.5%attend schools offering Grades 1-5 or 1-8.'
3.2. Performance of Grade 4 students
In 2002, Sri Lanka's NEREC collected data from a sample of 20 Grade 4 students ineach of 939 randomly selected public schools (NEREC 2004). To measure the learn-ing of students who had completed four years of school, these students were tested inmathematics, English, and 'first language' in March 2003, when almost all were start-ing Grade 5. 'First language' was Tamil for Tamil students and Sinhala for all otherstudents. Each test had 40 questions, almost all of which were multiple choice.
Table 1 shows mean test scores, by socioeconomic groups. Scores are normalizedto have 0 mean and a standard deviation of 1 (see the first row). On average, girlsoutperfomied boys on all tests (second and third rows). The next nine rows compareSri Lanka's provinces. Western Province, which contains the two wealthiest districts(Colombo and Gampaha), performed best in all three subjects. Northwestern andSouthern Provinces, which border Western Province, had the next highest scores.Northern and Eastern Provinces had the lowest scores, probably due to the impact of
Education Economics
Table I. Standardized NEREC test scores by geographic and socioeconomic groups.
All studentsBoys
Girls
Region
WesternCentral
Southern
Eastern
Northern
North-Central
North-Western
Sabaragamuwa
Uva
SinhaleseTamil
Moor/Malay
Burgher
Mother's education
None
Grades 1-5
Grades 6-10
0 level
A level
Postgraduation
Father's education
None
Grades 1—5
Grades 6-10
0 level
A level
Postgraduation
School type
Grades 1-13, three subjects
Grades 1-13, two subjects
Grades 1-11
Grades 1-5, 1-8
Expenditure quintile1
2
3
n
16,383
8299
8084
1842
1816
1808
1814
1857
1828
1820
1793
1805
10,999
3561
1715
17
1029
3970
6036
3452
1667
229
803
4461
6150
3226
1436
307
2673
4909
5448
3353
534
526
534
Math
Mean
0
-.09
.09
.29
-.12
.08
-.39
-.40
.09
.12
.02
-.14
.16
-.52
-.18
.46
-.69-.42
-.05
.37
.60
.62
-.71
-.38
.04
.38
.59
.64
.43
.03
-.25
-.03
-.39
-.22
-.06
SD
1
1.05
.94
.881.00
1.01
1.03
1.03
.93
.93
1.02
1.03
.95
1.03
.95
.89
1.01
1.00
.97
.82
.76
.77
1.01
1.01
.96
.83
.75
.79
.82
.97
1.01
1.04
1.04
1.02
.98
English
Mean
0
-.12
.12
.44
-.08
.01
-.34
-.34
-.09
-.04
-.05
-.19
.09
-.36
.01
.68
-.63
-.43
.13
.37
.79
.93
-.66
-.41
-.12
.42
.79
.95
.60
-.07
-.31
.06
-.41
-.33
-.13
SD
1
.99
.99
1.00
.96
1.04
.91
.91
.91
.93
.98
.94
1.01
.90
.96
.82
.73
.81
.90
.98
1.00
1.01
.69
.82
.91
.93
1.00
1.04
1.02
.93
.86
1.05
.86
.85
.91
First language
Mean
0
-.14
.15
.33-.11
.04
-.37
-.35
-.02
.10
.02
-.17
.15
-.49
-.18
.32
-.73
-.45
-.05
.39
.65
.66
-.73-.41
-.04
.41
.62
.67
.46
.03
-.28
-.02
-.41
-.26-.10
SD
1
1.03
.95
.88
1.01
1.05
1.02
.97
.92
.93
1.00
1.05
.97
1.00.91
.85
1.00
.99
.95
.82
.76
.72
.98
1.00
.95
.84
.76
.76
.82
.97
1.001.03
1.04
1.00
.98
H. Aturupane et al.
Table I. (Continued).
4
5
HAZ < -2- 2 > H A Z < - 1HAZ>-1WHZ < -2- 2 > WHZ<-1WHZ>-1
n
526
533
405
11541094348
1668637
Math
Mean
.08
.53
-.28-.10
.21
-.05.00
.03
SD
.94
.76
1.031.01.94
1.03.99
1.00
English
Mean
.11
.68
-.36-.10
.23-.14
.02
.02
SD
.96
1.00.89
.96
1.031.00.99
1.01
First language
Mean
.12
.58
-.32-.12
.24
-.09.01
.01
SD
.93
.72
1.061.01.91
1.011.00.98
20 years of secessionist conflict on their education systems. The next lowestperformer is Uva Province, a relatively poor, underdeveloped province.
Test scores vary widely by ethnicity. The tiny Burgher minority (descendents ofEuropean colonists) scored highest, followed by Sinhalese, Muslims (Moors), andTamils. Tamils' low scores reflect civil unrest in the Northeast, where most Tamilslive, and the low incomes of Tamil tea and rubber estate workers in Sri Lanka's centralmountains.
Students' scores are highly correlated with mothers' education, as expected.Children with uneducated mothers have the lowest scores, followed by children whosemothers have only primary schooling (one to five years). Additional mothers' school-ing is almost always associated with higher scores.
Given the discussion above, one would expect Type lAB schools to be the best,and indeed their students scored highest on all three exams. Type lC schools alsooffer 13 grades, and their students scored second highest on two of the three tests(math and first language). Surprisingly, student scores in Type 2 schools (up to Grade11) are lower than student scores in Type 3 schools, which are in remote or disadvan-taged areas.
Table 1 also examines test scores by household per capita expenditure and studenthealth status, from a subsample of the NEREC data with those variables (seeSection 4). Students from wealthier families score higher, as expected. The last threerows suggest a large role for health and nutrition. Height and weight data were usedto calculate stunting (low height-for-age) and wasting (low weight-for-height),expressed as Z-scores, which compare a child's height and weight with those from areference population of healthy children, whose Z-scores have a 0 mean and a stan-dard deviation of 1. Low height-for-age indicates slow physical growth due to poornutrition, diarrhea, or other illnesses in the first years of life. Weight-for-heightindicates recent malnutrition, diarrhea, or other illnesses.-* Stunted children (height-for-age Z-score < —2) scored about one-third of a standard deviation below the averagestudent. Marginally stunted children (Z-score from -2 to -1) scored slightly belowaverage, and children who were not stunted (Z-score > -1 ) scored above average. Thissuggests that early childhood nutrition influences academic skills. Yet the relationshipbetween current nutrition and academic performance is weak; the test scores of chil-dren with weight-for-height Z-scores below -2 ranged from -0.14 to -0.05, and thosewith higher Z-scores had scores either equal to or slightly above 0.00.
Education Economics 7
4. Data and methodology
This section describes the data available from Sri Lanka and the approach used toestimate the impacts of school quality, socioeconomic factors, and child health onleaming among Grade 4 students in that country.
4.1. Data
This paper uses three sources of data. The first is the NEREC survey of 16,383students in 939 randomly selected public schools (NEREC 2004).'* To measure theleaming of students who completed Grade 4 in Decejnber 2002, NEREC administeredrnath, English, and first language (Sinhalese or Tamil) tests to fifth graders near thebeginning (March) of the 2003 school year. (Students who repeated Grade 4 weretested, but not students who repeated Grade 5.) At the same time, NEREC also admin-istered questionnaires to students, parents, teachers, section heads, and principals.
The second source is the data collected by Sri Lanka's National EducationCommission (NEC) in the summer of 2003 from a random sub-sample of the NERECstudents: 2653 students in 140 schools (NEC 2004). The NEC collected three types ofdata. First, it used a household questionnaire to collect child and household data,including parents' reports on their child's health. Second, school questionnaires werecompleted for each school, focusing on Grade 4 teachers (including classroom obser-vation). Third, in mid-2003 medical staff recorded health and nutrition data for eachchild. The data from the NEREC and NEC questionnaires are summarized in AppendixTables A.I and A.2.
The third data source is the Sri Lanka Integrated Survey (SLIS), which sampled7500 households in all provinces from October 1999 to September 2000. It collecteda wide variety of data using household, community, and price questionnaires.
4.2. Key characteristics of education in Sri Lanka
Several aspects of education in Sri Lanka have implications for estimation. First,primary education is virtually universal (the SLIS data show a 97.0% enrollment ratefor children age 6-10), and this holds for all income groups (the rate for the poorest20% of the population is 94.7%). Thus, delayed entry into primary school is rare; if asubstantial fraction of children enter Grade 1 at age 7 or older, instead of age 5 or 6,the net enrollment rate would not be 97%. Therefore, selection bias from delayedenrollment or non-enrollment in primary school is very unlikely.
Another important aspect is low-grade repetition, which refieets an automaticpromotion policy. The primary repetition rate is 3% (World Bank 2005). Almost allfifth graders enroll on time and never repeat; 93% of the pupils in the NEC samplewere bom in the 12-month period (February 1993 to Januaiy 1994) corresponding toon-time enrollment and no repetition. Thus, repetition cannot cause serious selectionbias.
Finally, Sri Lanka has a low primary dropout rate. Only 1.4% of pupils who startGrade 1 fail to finish Grade 5, the last year of primary school (World Bank 2005).Overall, lack of delayed enrollment, repetition, and dropping out imply that both attri-tion and selection bias are unlikely; a sample of fifth grade pupils is a random sampleof all Sri Lankan children who were nine years old at the start of the school year.(Sri Lankan pupils start Grade 1 if they are five years old when the school year begins.)
8 H. Aturupane et al.
Another estimation concem is whether parents can choose from several nearbyschools and, if so, how frequently they do. Such choice is feasible; 66% of the NEChouseholds report being within one kilometer of the nearest school, and 86% reportbeing within two kilometers.' Yet most children attend the nearest school. The NECdata show that between 56% and 81% of Sri Lankan fourth graders attend the nearestschool (25% of the sample is missing the variable indicating attendance at the nearestschool). In contrast, only 15% report not attending the nearest school and doing sofor school quality concems. A far more common way for parents to address schoolquality worries is by enrolling their children in private tutoring ('tuition') classes thatoperate outside of school hours. The NEC data show that 74% of Grade 4 studentsattend such classes.
A related issue is whether parents alter the quality of their children's schools. Thisappears unlikely. Teachers are assigned at the national level. The NEC data suggestthat parents do not influence their children's schools. When asked 'how often do youparticipate in school activities related to your child?' 81% responded 'only if theschool requires it.' A more drastic way to alter school quality, moving to find a betterschool, is rare. About 72% of children live in their place of birth (SLIS data). Only29% of adults report living away from their birthplace, and the main reasons formoving are maniage, land availability, and work. Finally, parents rarely send theirchildren to boarding schools; 94% of children in the NEC sample lived with bothparents.
Overall, neither selection nor attrition bias is likely in Sri Lanka, and school qual-ity is unlikely to be endogenous due to parental actions. Moreover, the few studies thatcheck for such bias when estimating student leaming typically have found little or noevidence of it (e.g., Glewwe and Jacoby 1994; Glewwe et al. 1995).
4.3. Analytical framework
Estimating the impact of education policies on leaming requires a clear framework toguide, and interpret, the estimates. Assume that parents maximize life-cycle utility,which depends on goods and services in each time period, child health in each timeperiod, and each child's educational attainment and socioeeonomic success. Theconstraints are production functions for academic skills (learning) and child health,the impacts of years in school and skills on children's future incomes, a life-cyclebudget constraint, and (possibly) credit constraints.
The leaming production function, a stmctural relationship, is:^
A = a , (C, FS, MS, Q;^, H, El) (1)
where A is skills acquired ('achievement'); 'pf denotes production function; C is avector of fixed child characteristics (mainly 'innate ability' and motivation/prefer-ences); FS and MS are fathers' and mothers' schooling; Q ('quality') is a vector ofschool, teacher, and principal characteristics; S is the child's years in school, H is avector of child health variables; and El is all education 'inputs' under parental control(time children spend studying at home, purchased education materials, time in tuitionclasses, etc.).
Consider which variables are endogenous and exogenous in the sense that parentscontrol them.' The child characteristics in C, innate ability and motivation/tastes foreducation, are exogenous, as is parental schooling. Grade 4 students are young, so the
Education Economies 9
Q variables (school, teacher, or principal characteristics that affect leaming) arelargely time invariant. They are also arguably exogenous; most pupils attend the clos-est school and parents do little to alter school quality.
In most developing countries years of schooling (S) is endogenous since primaryschool children often start late, repeat, and/or drop out. Yet these are all rare forSri Lankan primary students, so S is exogenous. Indeed, S has no variation since allsampled children were in Grade 4 in 2002, so it is dropped from the analysis (and canbe dropped from Equation [1]).
The last two sets of variables are endogenous. Child health (H) can directly affectleaming (see Glewwe and Miguel (2008) for a recent review). Child health problemsinclude poor nutrition in early childhood, malnutrition while in school, parasitic infec-tions, vision and hearing problems, and micronutrient deficiencies. Educational inputs(El), which include daily attendance, enrollment in tuition (tutoring) classes, andpurchased textbooks and other educational materials, are also clearly endogenous.
Several variables that affect learning are excluded from Equation ( 1 ) because theireffects are only indirect^ they change A only by changing S, H, and El. They are schooland health care prices (P^ and P,,), dwelling and local environment conditions that affectchild health (DLE), household wealth (W), household productive assets that may affectchildren's time allocation (PA), and parental 'tastes' for education (T). Schooling prices(P^) include fees, school supply prices, tuition class fees, and travel time to school (whichmay affect daily attendance). Health care prices (P,,) include prices for health careservices and distances to health care facilities. Dwelling and local environment variablesinclude drinking water source, type of toilet, and local prevalence of infectious diseases.
Household wealth (W) indirectly affects student leaming through purchases ofeducational inputs. In theory, it can be endogenous: working children (who probablystudy less) raise household wealth. Yet Sri Lankan primary school age children rarelywork. Only 8% of children in the NEC data worked on family fanns or businesseswhen school was in session, and only 0.5% worked over six hours per week. Wagework is even rarer; only 2% report such work when school is in session, and only 0.2%work over six hours per week.
Sri Lankan households' main productive asset (PA) is land. Even after controllingfor household wealth, parents with land may expect their children to work on it, reduc-ing both time in school and time studying at home. Yet since few Sri Lankan childrenwork long hours when in school, the impact of productive assets on learning is likelyto be small.
Parents' choices regarding child health (H) and educational inputs (El) based onC, FS, MS, Q, Pg, P„, DLE, W, PA, and T can be expressed as:^
H = h(C, FS, MS, Q, P,, PH, DLE, W, PA, T) (2)
El = ei(C, FS, MS, Q, P , PH, DLE, W, PA, T) (3)
Inserting Equations (2) and (3) into Equation (1) gives the reduced form equation for A:
A = art(C, FS, MS, Q, P,, P», DLE, W, PA, T) (4)
Equation (4) is a causal relationship, but not a production function because it reflectshousehold preferences and includes prices as arguments. The 'rf (reduced form)subscript distinguishes it from the production function in Equation (1).
10 H. Aturupane et d\.
Policy-makers' primarily concem is the impact of education policies on years ofschooling (S) and academic achievement (A). Such policies include raising teacherquality, which affects Q, and changing school costs (Pg). Equation (4) shows howsuch changes affect A. Policy choices require cornparison ofthe costs of such changesto their benefits, measured by increases in A. Costs should also include costs borne byhouseholds, so changes in El, as given in Equation (3), and in household leisure mustbe included in the cost.
To clarify the difference between the production function in Equation ( 1 ) and thereduced form relationship in Equation (4), consider changing one element of Q, callit Q|. Equation (1) shows how changing Qj affects A holding constant all other vari-ables that directly affect learning, which is the partial derivative of A with respect toQj. In contrast. Equation (4) shows the impact on A after allowing H and El to changein response to changes in Q|, which is the total derivative of A with respect to Qj. Forexample, parents may respond to higher teacher quality by reducing education inputs.These two impacts of Q on A (partial and total derivatives) can differ; researchersshould indicate which relationship they are estimating. This paper estimates both.
When examining a policy's impact, should policy-makers use Equation (1) or (4)?The latter is useful because it shows what actually happens to A if Q changes. Incontrast, the former does not show this; it ignores changes in H and El due to changesin Q. Also, Equation (4) shows what happens after changing P^ and P^,, but Equation(1) cannot since P^ and P,, are excluded since they do not directly affect A. Yet thestructural impact in Equation (1) is useful: it may better capture overall welfareeffects. Intuitively, if increasing Q causes parents to reduce educational inputs El,they probably use the money saved to buy more consumer goods. The reduced formimpact in Equation (4) reflects the lower A from the drop in El, but ignores higherhousehold welfare from increased purchase of consumer goods. In contrast, the struc-tural impact in Equation ( 1 ) ignores the effect of Q on both El and consumer goods,and since these effects have opposite impacts on household welfare they tend to cancelout. Thus, welfare changes from increasing Q tend to be underestimated by thereduced form relationship in Equation (4), but this bias is absent from estimatedchanges in A in Equation (1). See Glewwe et al. (2004) for details.
4.4. Applying the framework to Sri Lanka
The objective of this paper is to estimate Equations (1) and (4) and use these estimatesto draw causal inferences regarding the determinants of leaming in Sri Lanka. Theseequations can be estimated if one has accurate data on every variable in them. Table 2shows the variables from the NEC and NEREC surveys. Some variables in Equations( 1 ) and (4) are in neither survey, and others may be measured with error. Both prob-lems lead to bias. This subsection explains the approach used to minimize bias. Whilewe cannot guarantee that our estimates are unbiased, the unusually rich data shouldminimize omitted variable bias and thus provide more accurate estimates than previousstudies have provided. Even so, the results presented in Sections 5 and 6 should beregarded as suggestive, rather than definitive.
To begin, consider the exogenous child variables, C, in Equation (1). While theyare exogenous in the sense that they are not chosen by households, they are endoge-nous in the (econometric) sense that they are correlated with unobserved variables.That is, much of the explanatory power of observed child variables such as sex, age,ethnicity, and birth order in that equation reflects their correlation with innate ability.
Education Economics 11
CO>
> • •
scCO
C L
XO+—O
co
cgCu
O
Q
coo -^ oct os
iri
^" Os ^O fN ^^• se ' ' '
tN
c^ c^ Ci
Os O — OC (NOS OC; rsi — poc' oc r-.: —
p
es) so (N(N r ' i <N
osoo<N (N (N
-O
S c w
ca ~
- c E. E ca
E c/3 ( -
'5.g
Èa ffl
-o x)2o
ca
JZ 00 ' ^ 60 ÛO
2o00
— O .S -3 .3 .s " .E
"ri C "ri . _ '7s "o Ts 'ñ
O
co cO
TO Ä
Q U Q
•^ E E EÜ E E E00 3 3 3< Q Q Q
t- x> îu -^ _ _c
g a é I I i 8 S>y 5 < U . . S c / 3 H S m
B.
c/) o
11-.s:
I
ca t io ^
•^ E
'S -ô¡3 "u 3> O
caEo
co
•t^llloc o
I ? Í
- o' "l/-) — —
c/î t/i ><U U u
aj "O "O —. 2 2 2 - 1g O Ü OCJ
o o
c^ vî /îcd CÖ rt
-oT3
C 00•c .E Î:
— "û CL.^ S: C'o. -a .2 •00 00 to
Oo -t:
> - '^
[/) '.¡3 ' ^ '.¡i co
23
t3 "O "OC C C
Xl -D X)ca ca ca
b«:§ Ei ^ .6 60
E E E ta
i § 1 ^ .Q Q Q £
co15o3
-O
Oc
x:0)
O o oE E E>. > >,
^ X) ^•o -a TD
<U u (D> > >u (U u
co co coo o o3 3 3
T3 TD TDtu (U (U
OJ U ÍDj= .x: j=00 00 00
î ï ;
3
-5co.X0)
p CL» ¿^
C - ^ CJ C L
O '¿] IL] ^
c oc o lo — _o c — so o
• 3 3 3 3 3C -73 -^ -a -OE. U <U U U
12 H. Aturupane et al.
•S
XI
2
ro -^ ' -
ro (N— O
e S"o -aII CJ
0 «
« 2bJO
II t?
- g.
1 o60(D
•a
-5 x:
ilX) x>-a -aQJ QJ> >(L) (U
ca cac cO o
ca a.3 3
•a -a
'—' t~^ro ^
r o r o r o r o r o
ca cao o3 3 <U
-O -a ^'OJ (L) • >
vî "tí r;Oj lU JJ
j= x: •£bO 60 o
X X
(Uaj >,
(U 2 II^ > II
•> r OJ OJ —C fe " "OJ x : ^ J OJ
II o o o -ao .. Il II OJ
„ w 11 II -s
^ - ° IM " S S 2II ^ >> >> bO
~ O II II slo — — — n."" ^ "ôÎ "ôÎ o"5 "3 > > u
^ ^ ^ _u _^ u•g 2 S _i J g5g Ü Ü o < -a^ u^ cß C/1 [/; (/
J— CÖ rt ni nï ^
B ^ s ^ s ^II ID OJ OJ Ï3 Ü<=> s s s -B •£c/î c 2 cS <i5 cS cSOJ>, >1 >^ >~> >-. > ,
¡p X> X5 X> X X3I t3 -o -u -a -O
•— OJ 0^ OJ Oj oc .OJ o OJ OJ (U.2 ^ ^ 2 IS ^+H o CJ o o oca ca ca ca ca ca3 c c c c c•o o o o o oOJ '^ '^ '*Z ". 'r^O ca ca ca ca ca_ CJ CJ CJ o CJ
c 3 3 3 3 3 (UVI T} T3 -O -O -O '^'ca 0> OJ fU OJ (U ' ^OJ
CÄ
O lyo — _C " VO C3 3 3 : :
" ^ " ^ "O ' ^ 'O "O
Q Q Q û Q Q
oD.3
— O x: I
x: x: j : j= ^ 5bo 60 bû M bO o
X ï S S î
os — fN MD (N00 00 00 ro —
(N oo ro I/-1 —
I I I
mo
f N
vOf N
. — 1
O
f N
rof N^Of N
ro
f N
ro
vof N
rovOf N
f N
f N
t—.f N
f N
ua.o a.oo
0)
f N
Ccax:
oEOJc/î
_ca^_,
'sE
oble
D.
C
C/3
cax:•uroü
x:
scor
e
SI
age
-sco
re
N
-age
: Z-s
coi
x:
-hei
j
OJ
c
.2OJwîca
T3ca
V
c/l
• a
• ch
i
oCoD.ïïV,c
o.c:
indi
cati
u
riab
l
N
§ §Q Q
X)o
-s: bO ÏÏJ 5¿ — C-•^ ^ : 2 ^ " bo
bb 60 bi) 53
Í I ä S
U
n.o
x:voro
3O
x:roO
3O
3 -C
S
il-.s:
S
•5
2 3.0 ^
8 §-3 ^
Education Economics 13
• c
n)
vO oo ONO^ <N "^
O O) — CN t--' r ' ^ r-) r-> —
o ONro —
— O t~-00 OO r-
ro• ^
(N
ro"O
roro^O
ONOS
( N
00
( N
rol/~j ly-)
( N
OOCN•O(N
ro"O
ro^O
roNOOl
OO
O l
roNO
Ol
roNOo l
c T3 -a ,s
•a3
-e-• o<u
itten
dla
ys a
Q
8CA
hild
n / bo
oib
rar;
_ i
trip
sdu
c. 1
tu
_oo1)
xerc
i.
u
_ooOJ
xerc
i.
M
_^O
se b
oxe
rci:
u
oo
resc
h
CL
CÜJ Oc ji;
it h o
rm
at
ngl.
£in
/Ta
fTi f / )
c
tuit
io
CL
c
tuit
io
C/1
o
C L
tù U
il
14 H. Aturupane et al.
CNO
3
Xl
2
•Hca
S.
CN lO 00 .—ro CN — —
oo r^ -— — CN 00oo ON o
CNO O ^ ) O(N CN (N (N (N
—
st / i(Ut/53• ^
eque
nt
ci:
her
U
B1-.
It th
ei
00
c ati
iden
ts s
tr f
or s
tiat
oi
u
blac
k
<u3
RE
C)
o(N
E
c
yes,
0 =
II
ISca
(Q)
ics
•2
a
-s:o
and
teac
he,
• a^^
stud
enqu
aad
e
C / j
)ol
hag
sehe
c
e in
dica
lva
riab
l
EE3
Q
Q
s"ooJZo
f
o
embe
r
Et / j
hooi
o
u
"f60C
e in
dica
lva
riab
l
1E3
Q
ol f
amil
y
o
Sel
hool
1S
ool
is a
for
sch
1E3Q
scho
ol
> ï
oCû
nool
ot/î
'Bbcat/)
"oo
for
sch
E"E3
Q
scho
olG
ir
(U<—
ca
prim
ai
coSE 33IA
J
rs o
f se
yea
—
o
h yr
s ex
p
Ü
Tec
feOca
ngs
hm
eeti
eu
nt-t
each
of p
are
ber
E3
Z
each
mtg
T
Par
ipal
dnci
o.co
ervi
ce
y^
year
s o
l's t
otal
co. G .
'5c£
: yrs
exp
Education Economics 15
tastes, and motivation. The latter are very difficult to measure, so these variables areincluded to 'control for' them. First-bom children may have higher ability, perhapsdue to better maternal nutrition before they are born.' Girls may have more innatereading talent, and boys may have more math talent, so a sex variable helps controlfor ability (OECD 2003). Age can also indicate ability; older children have had moretime to develop their innate ability.'" Another indicator of ability, and perhaps oftastes and motivation, is parents' education. Parents' ability, tastes, and motivationpresumably affected their schooling, and these are, in part, inherited by their children.Thus, parental schooling reflects all three.
Parental schooling (FS and MS) also enters Equation (1) directly. Thus, its esti-mated impact in Equation (4) reflects both direct effects and child ability, taste, andmotivation effects. Yet parents' education, while exogenous in the sense that it is nota choice variable, may be imprecisely measured. Fortunately, the NEC and NERECsurveys both have this variable, so one measurement can be used to instrument theother.
The NEREC and NEC data have many school, teacher, and principal variables(Q). School data include school type (see Subsection 3.1), student-teacher ratio,whether textbooks and teacher's guides arrived before the school year started, whetherstudent desks, blackboards, computers, and toilets are considered adequate, electricity,access to drinking water, and whether the school is for boys or girls only.
The NEC surveyed teachers who taught the sampled students in Grade 4. Mostschools had several Grade 4 teachers, who cannot be linked to specific students. Thus,Grade 4 teacher characteristics are school averages. They include sex, general educa-tion, teacher training, years of teaching experience, primary school teaching experi-ence, years at present school, days absent (reported by principal), in-service trainingsessions, visits from in-service advisors (school inspectors), divisional directors andeducation directors, parent-teacher meetings held, involvement in teaching studentsafter school, adequacy of equipment and materials received, and availability (forstudents) of textbooks, exercise books, workbooks, pencils, and other materials.Lastly, in 2003 trained observers visited classrooms to record teachers' preparation,pedagogical methods, use of English, use of learning materials, interactions withstudents, evaluation methods, and overall enthusiasm.
Principals have similar variables: gender, general education, teacher education,years of teaching experience, years of experience as a principal, experience as princi-pal in current school, frequency of supervising teachers and inspecting lesson plans,staff meetings on education matters, whether the school is in a 'school family'(explained below), parent awareness programs, and attendance at school family meet-ings, parent-teacher meetings, and educational reform programs at district, zone, anddivision levels.
The NEREC principal, section head, and teacher questionnaires collected similardata. Yet many NEREC variables had inconsistent or missing values, perhaps becauseschool personnel filled out these questionnaires without assistance. In contrast, theNEC used trained enumerators to interview school personnel, yielding data with fewerproblems. Unless otherwise noted, school and teacher variables are from the NECdata.
A final point regarding school and teacher variables is that one way to reduceomitted variable bias is to use school fixed effects estimation. This amounts to addinga dummy variable for each school, which accounts for variation in both observed andunobserved school characteristics. While this greatly reduces bias due to omitted
16 H. A turupane et al.
school and teacher variables, it comes at the cost of removing all school and teachervariables from the regression, so that one can estimate the impacts only of family- andstudent-level variables. The next section presents estimates both with and withoutschool fixed effects.
The NEC collected detailed child health (H) data. The household questionnairesolicited (from parents) child information on: illness in the past month, schoolabsence in the past year due to illness, any illness ever of more than two weeks, boutsof malaria (how many times in lifetime, in last year, and in last three months), wormsin stools (how many times in last three months and in last year), use of dewormingmedicine in past year, and the presence of vision problems, hearing problems, andphysical or mental disabilities. Data were also collected on children's diets and healthhabits.
The NEC employed trained medical personnel to conduct physical measurementsof 2459 (out the 2653) pupils. The data include height, weight, Bitot's spot in eyes (tocheck Vitamin A deficiency), goiter (iodine deficiency), visual acuity (with glasseson, using Snellen chart)," hemoglobin (iron) level in blood (finger prick), andpinworm, roundworm, whipworm, and hookworm (fecal samples).
The data indicate that several health problems are too rare to explain variation instudent leaming. Just eight students had Bitot spot. Very few had moderate round-worm (11 students), whipworm (10), or hookwonn (5) infections; none had heavyinfections. Only five had pinworms. This low incidence of helminths reflects the factthat 96% of parents report giving their children deworming medicine. Goiters (3% ofstudents) and malaria (only 2% of parents report their child had malaria in the pastyear) are also rare.
In contrast, anemia is common; 11.2% of children fall below the WHO standard(11.5 grams/deciliter), but only 0.1% have severe anemia (below 8.0 grams/deciliter).Stunting (height-for-age Z-score < -2) affects 16% of the sampled students, and 19%are wasted (weight-for-height Z-score < -2). Yet these variables measure nutritionproblems with eiTor; height and weight vary even among well-nourished children.This adds random measurement error to height-for-age and weight-for-heightZ-scores, leading to underestimated impacts. Consistent estimation requires instru-mental variables (IVs); possible instruments are the household's water source, type oftoilet, and current eating habits.
Lastly, consider vision and hearing. Vision was measured by medical personnel,and changes very slowly, which suggests little measurement error. Yet hearing is fromparental reports, which could have errors. There are no credible instruments for hear-ing. Yet hearing problems are rare; the parents of only 34 children report a problem,so while the lack of an instmment may lead to attenuation bias for the impact of hear-ing problems, it is unlikely to have a large effect on the estimated impacts of othervariables.
The last variables in Equation ( 1 ) are parent-provided educational inputs (El). TheNEREC and NEC data include: (1) how often different household members help thechild with schoolwork; (2) hours per week the child spends studying, attending tuitionclasses, and working; (3) access to textbooks, exercise books, and workbooks, bysubject; (4) children's books at home; (5) whether parents obtain library books for thechild; (6) school attendance (school records); (7) language spoken at home; and (8)parent provided 'educational trips.'
The NEC education input data may have measurement errors. Fortunately, theNEREC parent and child questionnaires also (imperfectly) measured most of them. If
Education Economies 17
these errors are random and uncon elated across surveys (NEREC questionnaires werefilled out by parents and children in March 2003, while trained interviewers filled outthe NEC questionnaire in the summer of 2003), the NEREC data can be used to instm-ment the NEC data. More specifically, education inputs from the NEC data wereinstrumented as follows. Hours per week studying and in tuition classes in the NECdata were instmmented by a dummy variable in the NEC indicating that the childattends tuition classes, a similar variable in the NEREC data, and two variables (seebelow) indicating parents' tastes for education. Children's books at home andpreschool attendance in the NEC data were both instrumented by sirnilar NEREC vari-ables. Finally, exercise books were in the NEC data instmmented by two NERECvariables: frequency (reported by the child) of teacher use of exercise books andblackboards.
Turn next to Equation (4); H and El in Equation (1) are replaced by P^, P,j, W,DLE, PA, and T. Sri Lanka's public schools are free and provide free textbooks anduniforms. Workbook and exercise book prices vary little by region, so the only schoolprice (P^) variables are tuition class fees, distance to the nearest primary school, anddistance (from school) to the nearest public library. The NEC data on pupils' hours in,and payments for, tuition classes were used to calculate school average tuition class'prices.' The NEC survey asked households the distance to the nearest primary school;this is a price since longer distances raise the opportunity cost of a day in school.Finally, the NEC school questionnaire indicates the distance from each school to thenearest public library.
Now consider health prices (P, ,). The NEC school questionnaire has distances tothe nearest hospitals and clinics. The SLIS collected local prices of medical services(registration fee, blood test, urine analysis, stool analysis, malaria test, and TB test).Mean prices were calculated for Sri Lanka's 25 districts.'^
Household wealth (W) is approximated by per capita expenditures from the NECdata.'"* Parents were asked for monthly expenditures on food and on 14 non-fooditems. They also reported monthly household income, in one of seven ranges, whichis used to instmment per capita expenditures to reduce attenuation (measurementerror) bias.
For DLE variables, dummy variables were created for toilet type and source ofdrinking water. A dummy variable indicates electric lighting, which presumably helpsstudents study at night. Regrettably, no data exist regarding local disease prevalence.
Land is Sri Lanka's main productive asset (PA); 27% of NEC households reportowning at least one acre. Few report owning other productive assets, such as boats orvehicles, so land is the only productive asset used.
The last variable in Equation (4) is parental tastes for education (T). Two variablesare used. The first, 'hope,' is parents' report of the highest degree they want for theirchild. The second, 'opinion,' is an index of parents' attitudes on education based ontheir agreement with eight statements such as 'It is a wise act to invest in education.'
5. Empirical results
This section presents estimates of Equations (1) and (4). The first subsection focuseson child and household variables, controlling for school characteristics using fixedeffects. The second adds school variables. The sample size drops from 2653 to lessthan 2450 because of missing data (mostly mother's age, use of libraries, and chil-dren's books at home). The IV estimates reduce the sample to slightly under 2400.
18 H. Aturupane et al.
5.7. School fixed effects
Columns (l)-(3) of Table 3 present OLS estimates, with school fixed effects, ofEquation (1): production functions for math, English, and first language skills.''* Thefirst four variables control for child's innate ability and motivation (C). Girls outper-fomi boys in each subject. Girls may have more reading talent than boys, but thiscannot explain the math score; perhaps girls are more motivated for all subjects. Agehas a significantly positive impact - presumably students' intellectual abilitiesincrease with age. First-bom children perform relatively well, perhaps reflectinghigher ability due to biological factors and/or greater attention from parents in the firstyears of life (relative to later-born children). Finally, mother's age raises test scores;early childbearing may have negative biological impacts on children's ability (Pevalin2003), but social factors may also play a role. A quadratic mother's age tenu (notshown here) had an insignificantly coefficient, revealing no negative impact of latechildbearing.
Next are three ethnic dummy variables. (Sinhalese is the omitted category; a fourthethnic dummy, 'other,' was never significant and was dropped.) Table 1 showed largeethnic differences in test scores. If the NEC and NEREC data include all variables inEquation (1), these dummy variables should be insignificant. For English, no ethnicvariable is significant, but in first language Tamils and Moors do worse, even after onecontrols for the language spoken at home, and Moors have lower math scores. Overall,most ethnic differences in Table 1 appear to be captured by other variables in Table 3.
The results suggest that both mothers' (MS) and fathers' (FS) schooling havelarge, statistically significant impacts on test scores. For each test, father's educationhas a larger estimated impact, a surprising result since mothers presumably spendmore time with their children; perhaps fathers have more say regarding children'seducation. Father's education may also reflect family income, which may be spent onunobserved educational inputs. Yet adding per capita expenditure as a regressor (notshown) does not change the father's education coefficient (nor any other). (The expen-diture variable was significant at the 5%, but not the 1%, level, revealing only weakevidence of omitted educational inputs.)
The next three variables measure child health status (H), and a fourth indicates thatdata are missing for child height (in which case height-for-age is replaced by itsmean).'^ Children who are not stunted (high height-for-age), and thus had better nutri-tion in their first years of life, scored higher on all three tests; a 1 standard deviationincrease raises each score by about 0.1 standard deviations. In contrast, weight-for-height, an indicator of current nutritional status, has a much smaller and statisticallyinsignificant impact. Moreover, the estimated impact of weight-for-height is small; aone standard deviation increase raises test scores by only 0.01-0.03 standard devia-tions. Overall, weight-for-height can be excluded from subsequent regressions.
Turning to other health variables, the few children with hearing problems (asreported by parents) have significantly lower test scores, 0.4-0.6 standard deviationsless. Children who have ever had a serious (>2 weeks) illness also had significantlylower scores on all three tests. Lastly, the goiter, malaria, and hemoglobin variableshad no explanatory power and so are excluded from all regressions.
Finally, consider educational inputs (El). Hours in tuition classes has a strong andsignificantly positive effect on all test scores, as does hours studying on math and firstlanguage scores. Hours working had a negative but insignificant effect, probablyreflecting Sri Lanka's low rate of child labor, so it was excluded from all regressions.
Education Economics 19
Table 3. Estimates of test score production functions using school fixed effects.
Sex
Age (months)
First-bom
Mother's age
Tamil
Moor
Burgher
Father yrs ed
Mother yrs ed
Height/age Z-sc
Haz dummy
Weight/height Z-sc
Hearing prob.
Severe ill
Hours tuit. class
Hours study
Father help
Mother help
Days attended
Children's books
OLS
(1)
Math
-0.134***
(0.037)
0.005***(0.001)
0.107***(0.032)0.010***
(0.003)
-0.206(0.162)
-0.303*(0.159)0.394
(0.600)0.039***
(0.006)
0.013**(0.006)0.086***
(0.019)-0.144**
(0.073)0.027
(0.020)
-0.636***
(0.165)-0.121**
(0.050)0.075***
(0.019)0.054**
(0.022)
0.025(0.022)
0.071***(0.022)
0.005***(0.001)
0.055***(0.020)
with fixed effects
(2)
English
-0.227***(0.038)
0.005***(0.001)
0.256***(0.035)0.014***
(0.003)0.044
(0.123)-0.002
(0.196)0.097
(0.745)
0.048***(0.007)0.015***
(0.005)0.109***
(0.020)-0.112*(0.064)
0.018(0.021)
-0.377**
(0.161)-0.081**(0.039)
0.076***(0.017)
0.021(0.022)
0.010
(0.025)0.094***
(0.023)0.005***
(0.001)
0.133***(0.022)
(3)
First lang.
-0.257***(0.040)
0.003**(0.001)0.179***
(0.034)0.010***
(0.003)-0.349**
(0.136)-0.276**
(0.138)0.374
(0.384)
0.034***
(0.006)0.017***
(0.006)0.106***
(0.018)-0.156**
(0.065)0.011
(0.018)-0.405**(0.172)
-0.138***(0.047)
0.088***
(0.017)0.064***
(0.019)
0.041*
(0.023)0.064***
(0.022)
0.006***(0.001)
0.095***(0.018)
Instrumental variable
(4)
Math
-0.124**(0.049)
0.005***(0.001)
0.033(0.050)0.005
(0.003)-0.078
(0.215)-0.196
(0.200)0.084
(0.455)
0.043***(0.012)
0.020
(0.015)0.088***
(0.022)
-0.069(0.089)
0.025(0.024)
-0.571***
(0.135)-0.094
(0.061)0.370***
(0.093)-0.105(0.191)
-0.005(0.029)
0.016(0.027)0.004***
(0.001)0.149
(0.097)
effects
(5)
English
-0.233***(0.050)
0.005***(0.001)
0.225***(0.046)0.009***
(0.003)0.224
(0.162)0.233
(0.240)0.089
(0.646)
0.057***(0.013)0.015
(0.016)0.096***
(0.024)
-0.002(0.079)
0.005(0.024)
-0.351**(0.150)
-0.048(0.056)
0.265**
(0.103)-0.009
(0.205)-0.054*(0.030)
0.035(0.030)0.004***
(0.001)0.517***
(0.111)
with fixed
(6)
First lang.
-0.253***(0.052)0.004***
(0.001)
0.136***(0.049)0.004
(0.003)-0.192
(0.160)
-0.047(0.140)0.456
(0.478)
0.025*(0.013)0.035**
(0.016)0.099***
(0.021)-0.091(0.079)
-0.007(0.021)
-0.376**(0.169)
-0.149**
(0.058)0.334***
(0.087)0.018
(0.199)
-0.011
(0.030)
-0.013(0.027)0.004***
(0.001)
0.303***(0.086)
20
Table 3. (Continued).
H. Aturupane et al.
Library books
Educ. trips
Exercise book
Preschool
Engl. at home
Sin/Tam at home
Constant
ObservationsR'
OLS
(1)
Math
0.162***(0.052)
0.094***
(0.035)0.432***
(0.042)
0.192***(0.072)
-3.862***(0.454)
24280.231
with tlxed effects
(2)
English
0.284***
(0.053)0.083**
(0.037)0.242***
(0.046)
0.096(0.069)0.148**
(0.061)
-3.831***(0.429)
24240.265
(3)
First lang.
0.150***
(0.052)
0.071**(0.034)
0.344***
(0.043)
0.203***(0.070)
0.282*(0.150)
-3.597***(0.469)
24240.273
Instrumental variable
(4)
Math
0.044
(0.073)0.033
(0.046)1.127**^
(0.190)
0.306(0.418)
2397
effects
(5)
English
0.055(0.084)
-0.023
(0.051)" 0.746***
(0.209)
-0.423(0.429)-0.081(0.077)
2393
with fixed
(6)
First lang.
-0.027
(0.078)-0.014(0.047)
1.085***(0.215)
0.234
(0.423)
0.338*(0.181)
2393
•Significant at 10%; **significant at 5%; ***significant at 1%.Robust standard errors, clustered at the school level, in parentheses.Instrumented variables (columns (4)-(6)): Father yrs ed. Mother yrs ed. Hours tuit. class. Hours study.Children's books, Exercise book, Preschool.Instruments: Dadedul5, DadeduólO, Dadeduol, Dadedual, Dadedupost, Mommedul5, MomeduólO,Momeduol, Momedual, Momedupost, Ptuitioii, Stuition, Hope, Opinion, Electric, Pbook, Tuseexer,Tuseblack, Nursery.
Mothers' time helping their children with schoolwork has a positive, statisticallysignificant impact on all test scores, but fathers' time has a (weakly) significant impactonly on first language. Interaction terms between parents' schooling and time helpingchildren were insignificant. The weak impact of fathers' time may reflect low varia-tion: 60% of mothers, but only 19% of fathers, report helping their children withschoolwork.
Daily attendance has a strong positive impact on all tests, as expected, as do chil-dren's books at home and borrowing library books. 'Educational trips' (to historic andcultural sites) also have positive and statistically significant effects. Finally, exercisebooks and preschool attendance have significantly positive effects.
The R^ coefficients indicate that these regressions 'explain' 23-27% of within-school variation (variation conditional on school fixed effects) of test scores. Since thetest scores probably contain substantial measurement errors, the amount of'true' vari-ation these regressions explain is probably rnuch higher.
Many explanatory variables in Table 3 may be measured with error, causingattenuation bias. Columns (4)-(6) of Table 3 report estimates (with school fixedeffects) that attempt to reduce bias by instrumenting seven variables: mother'sand father's education, hours in tuition classes, hours studying, children's books at
Education Economics 21
home, preschool attendance, and exercise books, using the instruments discussed inSubsection 4.4."'
The IV estimates pass several specification tests. First, overidentification tests donot reject the hypothesis that the mathematics and first language production functionresiduals are uncorrelated with the instmments, though it was rejected at the 5% (butnot the 1%) level for the English test. Second, checking for weak instruments, the(excluded) instmments have high predictive power for all but one of the instrumentedvariables, with F-tests from 6.64 (preschool) to 98.12 (father's years of education).The sole exception is hours studying; its F-test was 3.21. Third, Hausman testscomparing the fixed effects results in columns (l)-(3) with IV fixed effects results incolumns (4)-(6) decisively reject the null hypothesis of equal parameters (/;-values of0.0000 for each subject). Fourth, six other potentially endogenous variables (height-for-age, father helps child with schoolwork, mother helps child with schoolwork, useof library books, educational trips, and serious illness) passed the (joint) Hausmantest; their coefficients were unaffected by IV estimation.'^
Most uninstrumented variables have impacts similar to those in columns ( 1 )-(3)of Table 3, with four exceptions. First, the first-born coefficient is smaller and losessignificance in the math regression. Second, the effects of mother's age decline andlose significance for math and first language. Third, the ethnicity variables havesmaller impacts and lose statistical significance. Finally, the effects of mothers help-ing children, library books, and educational trips all fall and lose significance.
For the instrumented variables, the impacts of parental education are somewhatlarger, though in two of three cases mother's education loses significance due to lowerprecision. The impact of hours in tuition classes is four to five times higher and stillsignificant, but hours studying loses significance due to very high standard eiTors,perhaps reflecting weak instruments. The impacts of children's books are three to fourtimes higher and significant (although marginally for math). The exercise bookimpacts increase two to three fold (and remain significant), but the preschool effect isimprecisely estimated and thus insignificant.
Overall, IV results suggest that measurement errors generate serious attenuationbias in OLS estimates of the impacts of tuition classes, children's books at home, andexercise books. These estimates are large; raising time in tuition classes from one tothree hours per week, to four to six hours per week increases test scores by 0.27-0.38standard deviations, and exercise books increase test scores by 0.24-1.13 standarddeviations.
Recall that reduced form impacts of variables of interest may differ from theirproduction function impacts. Also, variables with only indirect effects on leaming areexcluded from the production function, yet their reduced foim (indirect) impacts maybe of interest. Table 4 presents such estimates, i.e. estimates of Equation (4); thehealth (H) and educational input (El) variables in Equation (1) are replaced by wealth(W), dwelling characteristics (DLE), productive assets (PA), and parental tastes foreducation (T).'**
Exogenous child characteristics (C) and parental education (FS and MS) appearin both the production function and the reduced fonti. The main changes in the OLSresults are that the impacts of child age, sex, first-bom status, and matemal age atbirth are smaller. This suggests that parents help 'less able' children, compensatingfor their lower innate ability. Also, mother's education has larger effects, whichsuggests that better educated mothers provide more education inputs and have health-ier children.
22 H. Aturupane et al.
Table 4. Estimates of test score reduced form equations with school fixed effects.
*Significant at 10%; **significant at 5%; ***significant at 1%.Robust standard errors, clustered at the school level, in parentheses.Instrumented (columns (4)-(6)): Father yrs ed. Mothers yrs ed, HH expenditure, Hope, Opinion.Instruments: Dadedul5, DadeduólO, Dadeduol, Dadedual, Dadedupost, Moniediil5, MomeduólO,Momeduol, Momedual, Momedupost, Hliinc, Nerechope, Nereeopin.
Education Economics 23
Regarding variables with only indirect effects, agricultural land has a positive butstatistically insignificant impact, indicating little role for child labor. Electric lightinghas a large, statistically significant impact, raising scores by 0.21-0.26 standard devi-ations; presumably it helps children study at night. Reduced form estimates supportthis; electricity raises hours studying, but this impact is imprecisely estimated (/-statisticof 1.60). Household expenditure per capita is also significant, but with a smaller impactthan electricity; a one standard deviation increase raises scores by 0.07-0.08 standarddeviations.
Two variables indicate parents' tastes for education. The one measuring hopes fortheir child's education is highly significant, while the other (general opinion on educa-tion) is less significant, and insignificant for math. Finally, children whose drinkingwater is from a river or stream have significantly lower math scores, and children inhouseholds with 'pit' latrine toilets do worse in English; these estimated effects prob-ably reflect lower child health.
Columns (4)-(6) of Table 4 examine whether the OLS results change when percapita expenditures, parental education, and parental tastes for education are instm-mented to reduce attenuation bias. The impact of per capita expenditures is too impre-cisely estimated to conclude anything. The impacts of parents' education and parentaltastes increase, suggesting attenuation bias in the OLS estimates." Other resultschange little, except that electricity has weaker effects.
5.2. School and teacher impacts
Table 5 replaces the school fixed effects in Table 3 with school and teacher charac-teristics from the NEC data. Columns (l)-(3) present OLS estimates, and columns(4)-(6) present IV estimates, instrumenting the same variables (with the same instru-ments) as in Table 3. The NEC data have many school and teacher variables, butwith just 140 schools the number of school variables one can use is limited. Vari-ables without explanatory power for any test are excluded. The child and householdvariable impacts are similar to those in Table 3, so the focus here is on school vari-ables. Yet one should keep in mind that, despite detailed school and teacher vari-ables, it is still possible that the following results suffer from omitted variable bias;the rich set of school and teacher characteristics used can reduce such bias, but theycannot eliminate it.
The only school physical facility or equipment variable with any significantimpact is desks. About 39% of principals report having insufficient desks. Sufficientdesks appear to increase all three test scores, but only the positive impact of 0.15 stan-dard deviations on math scores is statistically significant.
About 83% of Sri Lankan pupils attend schools associated with a 'school family,'a cluster of schools that collaborate and share resources. Teachers in a 'family' meetto discuss teaching methods, share solutions to problems, and exchange reading material,while principals discuss school organization and administrative issues. The OLS resultsin Table 5 indicate that joining a school family raises all test scores by about 0.2 standarddeviations, but this impact is halved and insignificant when IV estimation is used.
Consider next the single-sex school variables. Even after controlling for students'gender, boys enrolled in all-boy schools (14 of the 140 schools, 12 of which are inurban areas) do significantly worse on all tests, with impacts from -0.16 to -0.33. Incontrast, girls in the eight all-girl schools do much better in English, an impact of 0.46.The IV estimates show no impact of all-girl schools on English scores, although the
24 H. Aturupane et al.
Table 5. Estimates of test score production functions with school characteristics.
Sex
Age (months)
First-bom
Mother's age
Tamil
Moor
Burgher
Father yrs ed
Mother yrs ed
Height/age Z-sc
Haz dummy
Hearing prob.
Severe ill
Hours tuit. class
Hours study
Father help
Mother help
Days attended
Children's books
Library books
(1)
Math
-0.135***
(0.038)
0.006***
(0.001)
0.129***
(0.036)
0.015***
(0.003)-0.291***
(0.072)
-0.283***(0.101)
0.657
(0.546)
0.046***
(0.007)
0.016***
(0.006)
0.102***
(0.021)
-0.057
(0.084)
-0.592***
(0.179)
-0.100**
(0.049)
0.098***
(0.019)
0.066***
(0.021)
0.024
(0.024)
0.078***
(0.025)
0.005***
(0.001)
0.068**
(0.026)
0.146***
(0.052)
OLS
(2)
English
-0.236***
(0.039)
0.005***
(0.001)
0.308***
(0.039)
0.020***
(0.003)-0.079
(0.092)
0.012
(0.110)
0.502
(0.527)
0.058***
(0.008)
0.016***
(0.006)
0.133***
(0.023)-0.003
(0.083)
-0.320*
(0.168)
-0.049
(0.043)
0.074***
(0.020)
0.014
(0.023)
-0.015
(0.031)
0.113***
(0.026)
0.004***
(0.001)
0.133***
(0.032)
0.303***
(0.059)
(3)
First lang.
-0.268***
(0.044)
0.005***
(0.001)
0.189***
(0.038)
0.016***
(0.003)-0.237***
(0.072)
-0.223***(0.081)
0.640**
(0.247)
0.046***
(0.006)
0.021***
(0.006)
0.120***
(0.019)
-0.055
(0.076)
-0.340*
(0.173)
-0.094*
(0.048)0.109***
(0.019)
0.068***
(0.020)
0.023
(0.025)
0.072***
(0.024)
0.005***
(0.001)
0.113***
(0.021)
0.135**
(0.052)
Instrumental variable
(4)
Math
-0.184***
(0.049)
0.007***
(0.002)
0.069
(0.047)
0.008**
(0.004)
-0.423***
(0.116)
-0.355***
(0.126)
0.236
(0.462)
0.046***
(0.013)
0.022
(0.017)
0.092***
(0.026)-0.001
(0.105)
-0.492***(0.162)
-0.052
(0.066)
0.396***
(0.088)
-0.351*
(0.201)
-0.004
(0.034)
0.014
(0.032)
0.004***
(0.001)
0.166
(0.101)
0.055
(0.082)
(5)
English
-0.322***
(0.056)
0.004***
(0.002)
0.242***
(0.052)
0.012***
(0.004)-0.006
(0.139)
-0.031(0.159)
0.525
(0.530)
0.063***
(0.014)
0.004
(0.017)
0.109***
(0.027)0.096
(0.105)
-0.236
(0.163)
0.004
(0.061)
0.323***
(0.090)
-0.346
(0.222)
-0.093**
(0.040)0.049
(0.035)
0.004***
(0.001)
0.585***
(0.103)0.055
(0.100)
(6)
First lang.
-0.320***
(0.053)
0.006***
(0.002)
0.129***
(0.045)
0.009**
(0.003)-0.343***
(0.115)
-0.296**
(0.126)
0.421
(0.342)
0.039***
(0.012)
0.028*
(0.017)
0.103***
(0.023)-0.000
(0.091)
-0.255(0.172)
-0.072
(0.060)
0.396***
(0.086)
-0.264
(0.191)
-0.022
(0.033)
-0.003
(0.029)
0.004***
(0.001)
0.278***
(0.091)
-0.001
(0.085)
Education Economics 25
Table 5. {Continued).
Educ. trips
Exercise book
Preschool
Engl. at home
Sin/Tam at home
Desks
School family
Boys school
Girls school
Teacher yrs exp
Par-teach mtg
Princ yrs exp
Constant
Observations
R'
(1)
Math
0.152***
(0.041)
0.367***
(0.050)
0.148**
(0.067)
0.147**
(0.058)
0.220***
(0.067)
-0.262***
(0.085)
0.101
(0.107)
0.009**
(0.004)
0.032***
(0.011)
0.005
(0.005)
-5.065***
(0.457)
2305
0.37
OLS
(2)
English
0.129***
(0.048)
0.199***(0.049)
0.056
(0.063)
0.190***
(0.067)
0.070
(0.059)
0.205***
(0.071)
-0.155*
(0.078)
0.462***
(0.127)
0.010**
(0.005)
0.045***
(0.015)
0.007
(0.005)
-4.631***
(0.440)
2301
0.43
(3)
First lang.
0.112***
(0.040)
0.246***
(0.050)
0.153**
(0.069)
0.338**
(0.165)
0.084
(0.057)
0.197***
(0.069)
-0.131
(0.085)
0.145(0.094)
0.007*
(0.004)
0.030**
(0.012)
0.004
(0.005)
-4.895***
(0.516)2301
0.42
Instrumental variable
(4)
Math
0.110*
(0.056)
0.763***
(0.192)
-0.006(0.319)
0.130*
(0.074)
0.101
(0.107)
-0.330***
(0.106)
-0.033
(0.171)
0.003
(0.006)
0.018
(0.017)
0.008
(0.006)
-4.324***
(1.034)
2275
0.14
(5)
English
0.009
(0.061)
0.423**
(0.199)
-0.309
(0.334)
-0.101
(0.088)
0.044
(0.076)
0.066
(0.114)
-0.230**
(0.091)
0.128
(0.186)
0.007
(0.007)
0.030
(0.020)
0.013**
(0.006)
-3.863***
(1.167)2271
0.18
(6)
First lang.
0.046
(0.055)
0.676***
(0.192)
0.047
(0.295)
0.287
(0.219)0.069
(0.072)
0.101
(0.101)
-0.193**
(0.096)
-0.030
(0.155)
0.001
(0.006)0.017
(0.017)
0.007
(0.006)
-4.587***
(0.925)
2271
0.21
*Signiñcant at 10%; **significant at 5%; ***significant at 1%.Robust standard errors in parentheses.In.struniented (columns (4)-(6)): Father yrs ed. Mother yrs ed. Hours tuit. class. Hours study. Children'sbooks, Exercise book, Preschool.Instruments: Dadedul5, DadeduólO, Dadeduol, Dadedual, Dadedupost, Momedul5, MonieduólO,Momeduol, Momedual, Momedupost, Ptuition, Stuition, Hope, Opinion, Electric, Pbook, Tuseexer,Tuseblack, Nursery.
negative impacts of all-boy schools remain statistically significant. These resultsreflect some unobserved aspects of all-boy schools. Perhaps boys' behavior worsenswithout girls in the classroom, but this is speculative.
The last three variables pertain to teachers and principles. More experiencedteachers appear to increase students' scores on all three tests, but after instrumentingthis impact declines and loses significance. Similarly, teacher-parent meetings appear
26 H. Aturupane et al.
to increase all three test scores, but the estimated effects decline and lose significanceafter instmmenting. Finally, the coefficient on principals' years of experience (asprincipals) is positive and statistically significant effect on English, but only for IVestimates."°
In summary, two school and teacher characteristics appear to have significantlypositive impacts on leaming: student desks (math scores only) and principal years ofexperience (English only). One variable reduced all three scores: all-boy schools.OLS regressions explain much of the variance in test scores: 37-43%. Since testscores probably have random errors, the 'true' variation explained is probably muchhigher.
Table 6 shows the reduced form impacts of school variables. In principle, parents'responses to variation in school characteristics can yield different results in Tables 5and 6. Examining the IV estimates for math scores, the positive coefficient for desksis somewhat smaller and insignificant, while the coefficient for school family
Table 6. Estimates of test score reduced form equations with school eharacteristies.
Sex
Age (months)
First-born
Mother's age
Tamil
Moor
Burgher
Father yrs ed
Mother yrs ed
Engl. at home
Sin/Tam at home
Log of acres
Electricity home
(1)
Math
-0.166***(0.043)
0.003**(0.001)
0.213***(0.039)
0.017***(0.004)
-0.217***
(0.082)-0.342***(0.097)
0.456*(0.239)
0.050***(0.007)
0.028***(0.006)
0.007(0.012)
0.300***(0.058)
OLS
(2)
English
-0.272***(0.043)0.002*
(0.001)0.360***
(0.041)0.022***
(0.003)-0.055
(0.098)-0.109
(0.106)0.289
(0.311)0.060***
(0.008)0.025***
(0.006)0.222***
(0.069)
-0.002(0.008)0.330***
(0.059)
(3)
First lang.
-0.293***(0.047)0.001
(0.002)0.268***
(0.040)0.020***
(0.003)-0.181**(0.078)
-0.288***
(0.088)0.299***
(0.083)0.050***
(0.007)0.033***
(0.007)
0.379**
(0.170)0.004
(0.011)0.299***
(0.058)
Instrumental Variable
(4)
Math
-0.106*(0.057)
0.007***(0.002)0.162***
(0.049)0.013***
(0.005)-0.443***(0.149)
-0.646***(0.206)0.176
(0.336)0.053***
(0.015)0.055***
(0.018)
-0.012
(0.013)0.228***
(0.068)
(5)
English
-0.211***(0.065)
0.006**(0.003)0.265***
(0.058)0.017***
(0.005)-0.317*
(0.173)-0.518**(0.245)0.225
(0.375)
0.068***(0.016)0.044*
(0.023)-0.051(0.124)
-0.036**(0.016)0.223***
(0.084)
(6)
First lang.
-0.233***
(0.056)0.005**
(0.002)0.224***
(0.047)
0.015***(0.004)
-0.446***
(0.143)-0.626***(0.201)-0.049(0.152)0.049***
(0.014)
0.057***(0.018)
0.132
(0.237)-0.012(0.014)
0.226***(0.069)
Education Economics 27
Table 6. (Continued).
Hope
Opinion
H H expenditure
Drinks from river
Pit latrine
Desks
School family
Boys school
Girls school
Teach yrs exp
Par-teach mtg
Pr yrs exp
Constant
Observations
(1)
Math
0.048***(0.012)0.001
(0.008)0.131***
(0.038)
-0.411***(0.132)-0.111 *(0.063)0.087
(0.064)
0.160**(0.079)
-0.251**(0.102)0.079
(0.089)
0.007
(0.005)0.032**
(0.012)-0.001
(0.005)-3.657***
(0.566)23290.29
OLS
(2)
English
0.045***(0.012)0.012
(0.008)0.165***
(0.039)
-0.251(0.161)
-0.124**(0.061)0.025
(0.058)0.170**
(0.073)-0.140*(0.077)0.460***
(0.126)0.009*
(0.005)0.048***
(0.014)
0.001
(0.005)-4.325***(0.489)
23250.39
(3)
First lang.
0.045***(0.011)0.010
(0.008)0.145***
(0.039)-0.260**(0.111)
-0.153**(0.062)0.016
(0.064)
0.150*(0.077)-0.138(0.091)0.130
(0.083)
0.006(0.005)0.032**
(0.012)-0.002
(0.005)-3.857***
(0.629)23250.34
Instrumental Variable
(4)
Math
0.079
(0.108)0.192***
(0.073)-0.130(0.151)
-0.441**(0.174)
0.001(0.103)0.111
(0.088)0.246**
(0.109)
-0.472***(0.166)0.068
(0.193)
0.017**(0.008)0.066**
(0.028)
-0.005(0.008)-8.470***(2.094)
2252
(5)
English
0.026(0.131)0.302***
(0.089)0.087
(0.183)-0.291
(0.182)0.084
(0.133)0.044
(0.110)0.265*
(0.136)-0.464**(0.211)
0.353(0.261)
0.023**(0.010)0.086**
(0.037)
-0.003(0.009)
-12.913***(2.601)
2248
(6)
First lang.
0.126
(0.096)0.182**
(0.073)-0.152(0.142)-0.359**
(0.150)
-0.039(0.099)
0.043(0.085)0.247**
(0.112)
-0.320**(0.153)0.150
(0.166)0.016*
(0.008)0.070***
(0.026)
-0.007(0.007)-7.817***(2.084)
22480.04
*Signifieant at 10%; »»significant at 5%; »»»significant at 1%.Robust standard errors in parentheses.Instinmented (columns (4)-(6)); Father yrs ed. Father yrs ed, HH expenditure , Hope, Opinion.Instruments: Dadedul5, DadedtiólO, Dadeduol, Dadedual, Dadedupost, Momedul5, MonieduólO,Momeduol, Momedual, Momedupost, Hhinc, Erechope, Nereeopin.
increases and becomes significant. The negative infiuence of all-boy schools is largerand remains significant. The effect of parent-teacher meetings rises and acquiressignificance, while the principal's experience variable loses significance.'' ComparingTables 5 and 6, some school variables seem to substitute for parental inputs whileothers are complements. More specifically, parents appear to reduce educationalinputs when schools have more desks or more experienced principals, while schoolsbelonging to a school family and parent-teacher meetings seem to persuade parents toraise effort for their children's education.
28 H. Aturupane et al.
5.3. Is omitted variable bias a serious problem?
Tables 7 and 8 examine the sensitivity of the IV results in Tables 5 and 6 to omittedvariable bias. Columns (l)-(3) of Table 7 reproduce Table 5 IV estimates of theimpacts of school variables. Suppose no data exist for parent-provided educationalinputs (El). If parents provide more inputs in response to low quality schools, esti-mates from regression analysis will underestimate the impact of school quality vari-ables. Omitting El variables yields the results in columns (4)-(6) of Table 7. Theimpact of the school family variable increases and becomes (marginally) significant,as do the all-girl school variable (English only) and parent-teacher meetings. Overall,this suggests that higher quality schools generate more, not less, parentally providededucational inputs.
Table 8 repeats this exercise for the IV estimates in Table 6 of the (reduced form)impact of school variables on leaming, omitting the two parental tastes for educationvariables. The impact of all-boy schools falls by half, and the impact of parent-teacher meetings drops by more than half and loses some statistical significance.Admittedly, the intuition behind this result is unclear. Yet, overall. Tables 7 and 8suggest substantial omitted variable bias when estimating the impact of school char-acteristics on leaming without measures of parental attitudes or detailed data oneducational inputs.
Table 7. Production function, instrumental variable regression with school variables.
Desks
School family
Boys school
Girls school
Teach yrs exp
Par-teach mtg
Pr yrs exp
Constant
ObservationsR'
(1)
Math
0.130*(0.074)0.101
(0.107)
-0.330***(0.106)
-0.033
(0.171)0.003
(0.006)0.018
(0.017)
0.008(0.006)-4.324***
(1.034)22750.14
Table 5 results
(2)
English
0.044
(0.076)0.066
(0.114)
-0.230**(0.091)
0.128(0.186)0.007
(0.007)0.030
(0.020)
0.013**(0.006)
-3.863***(1.167)
22710.18
(3)
First lang.
0.069
(0.072)0.101
(0.101)-0.193**(0.096)
-0.030(0.155)0.001
(0.006)0.017
(0,017)
0.007(0.006)-4.587***
(0.925)22710.21
Table 5, dropping educational inputs
(4)
Math
0.125**(0.062)0.147**
(0.071)-0.312***
(0.100)0.037
(0.101)0.009**
(0.004)0.026**
(0.012)
-0.002(0.005)
-3.353***
(0.445)24300.24
(5)
English
0.068
(0.063)0.143*
(0.074)-0.198**
(0.078)
0.418***(0.131)0.010*
(0.005)0.040**
(0.016)-0.000(0.005)
-3.723***(0.421)
24260.33
(6)
First lang.
0.059
(0.063)0.130*
(0.074)-0.190**
(0.089)0.099
(0.093)0.007
(0.004)0.024*
(0.012)
-0.003(0.005)-3.194***
(0.516)24260.28
*Significant at 10%; **significant at 5%; ***significant at 1%.Robust standard errors in parentheses.
Education Economics
Table 8. Reduced form, instrumental variable regression with school variables.
29
Desks
School family
Boys school
Girls school
Teach yrs exp
Par-teach mtg
Pr yrs exp
Constant
ObservationsR'
(1)
Math
0.111(0.088)0.246**
(0.109)-0.472***
(0.166)0.068
(0.193)0.017**
(0.008)
0.066**(0.028)
-0.005(0.008)
-8.470***(2.094)
2252
Table 6 results
(2)
English
0.044(0.110)0.265*
(0.136)-0.464**
(0.211)
0.353(0.261)
0.023**(0.010)
0.086**(0.037)
-0.003(0.009)
-12.913***
(2.601)2248
(3)
First lang.
0.043(0.085)0.247**
(0.112)-0.320**
(0.153)0.150
(0.166)
0.016*(0.008)
0.070***(0.026)
-0.007(0.007)
-7.817***(2.084)
22480.04
Table 6, dropping taste for education
(4)
Math
0.082(0.062)0.128*
(0.071)-0.296***(0.102)
-0.006(0.100)
0.006(0.005)
0.020*(0.012)
-0.002
(0.005)-3.776***(0.622)
23370.25
(5)
English
0.016(0.060)0.109
(0.068)-0.202***
(0.076)0.336***
(0.128)0.008
(0.005)
0.031**(0.015)
-0.000(0.005)-5.066***(0.562)
23330.35
(6)
First lang.
0.014(0.062)0.112
(0.071)
-0.175*(0.093)0.053
(0.091)
0.005(0.005)
0.018(0.012)
-0.004
(0.005)-3.762***
(0.697)23330.30
*Significant at 10%; **sigiiificant at 5%; ***signiflcant at 1%.Robust standard errors in parentheses.
6. Policy implications
These findings suggest several policy initiatives to improve leaming outcomes inSri Lanka. First, grouping schools into school families appears to raise school quality,perhaps by providing opportunities for teachers and principals from different schoolsto leam from each others' experiences. Currently, school family networks in Sri Lankaare arranged infonnally. An official policy to set up school families could increaseleaming.
Second, teacher interactions with parents seem to improve leaming. More specif-ically, parent-teacher meetings may persuade parents to do more for their children'sschooling (this variable was insignificant in the production function estimates butsignificant in the reduced form estimates). As explained in Subsection 4.2, mostparents do little to change the quality of their local school. One policy worth explor-ing is to get parents more directly involved in raising school quality, and morespecifically in holding principals and teachers accountable for their children'sperformance.
Higher student attendance also increases leaming. This unsurprising finding high-lights the need for policies that raise attendance. Currently, Sri Lanka offers severalincentives for children to enroll, such as free tuition, textbooks, and uniforms. Subsi-dized transport and mid-day meals in poor areas also encourage daily attendance. Yetdaily attendance is often low, especially in poor areas. Cash transfers conditional on
30 H. Aturupane et al.
daily attendance, a successful policy in several Latin American countries (Damon andGlewwe 2009), should be considered in areas with low student attendance.
Students who have exercise books and attend schools with enough desks appear toleam more (although the effect ofthe latter variable is somewhat smaller and insignif-icant in the reduced fonn estimates, which may reflect a small reduction in parentalinputs in response to the provision of desks). This suggests that the government shouldequip all schools with basic learning equipment, such as pupil desks and chairs, andensure that all children have basic writing materials.
Children attending schools with experienced principals have higher scores, at leastin English. Principals' management skills and leadership are probably importantfactors affecting school perfonnancc. Sri Lanka has a training center that provides off-site training for principals. Yet no program provides on-site support for principals,which may be even more effective. Policy-makers should seriously consider providingon-site training to principals, especially less experienced principals.
Child health also contributes to leaming. Sri Lanka already has basic school healthprograms; health workers visit schools regularly to test children for several illnesses,and poor children receive free mid-day meals. Expanding school health programscould increase leaming. Yet the strong impacts of height-for-age suggest a need fornutrition programs directed toward infants and very young children. Two programscurrently exist: (1) the Thriposha program provides foodstuffs to pregnant and lactat-ing wotnen, to infants between 6 and 11 months old, and to older children whosegrowth falters (as certified by a medical health officer); and (2) the Samurdhi povertyreduction program transfers income to poor families, especially those with malnour-ished children. The impacts of these and other nutrition programs should be rigorouslyevaluated; those that are effective should be expanded.
Children enrolled in private tuition classes seem to leam more. While one couldargue that this may mostly reflect unobserved child motivation, which leads toincreased enrollment in these classes as well as increased leaming, for children whoare 10 or 11 years old the decision is primarily tnade by parents, and parental motiva-tion is controlled for, at least in part, by the variables that measure parental actions(e.g., time spent helping children with schoolwork). These classes have not only directcosts but also opportunity costs, as they crowd out extracurricular activities. Theycould create perverse incentives for teachers, who may reduce classroom teachingeffort to increase the demand for their tuition classes. Further research on thisphenomenon is needed to understand its costs and benefits.
7. Conclusion
This paper has used unusually rich data from Sri Lanka to investigate the detemiinantsof reading and math skills among fourth grade students. Several conclusions stand out.First, most ofthe differences in test scores across ethnic groups reflect differences inschool and (observed) family characteristics. Second, parents' education plays a largerole, but the mechanisms are unclear, especially for father's education. Third, consistentwith results from Pakistan (Alderman et al. 2001 ) and the Philippines (Glewwe, Jacoby,and King 2001), early childhood nutrition, measured by height-for-age, appears to havea sizeable impact on leaming. Hearing problems seem to have strong negative effects,but this applies to only 1-2% of students. Other health conditions, such as iron defi-ciencies and current nutritional status (measured by weight-for-height), had little
Education Economics 31
explanatory power. Fourth, the estimated effect of hours in tuition classes is large andsignificant. This suggests that, despite Sri Lanka's efforts to equalize access to educa-tion, better off students can 'buy' a higher quality education. Fifth, electric lighting athome appears to help students, presumably by increasing opportunities to study at night.
There are also useful results conceming school and teacher characteristics.Principals' and teachers' years of experience, grouping schools into 'school families,'and parent-teacher meetings all appear to increase leaming. For boys, the estimatedimpact of attending an all-boy school is negative, though the reasons why are unclear.
While these results provide useful policy guidance, many unanswered questionsremain. First, much more thinking is needed on the role of tuition classes. In essence,these classes constitute a partial privatization of education in Sri Lanka. Second, muchremains to be leamed about which school (and teacher) characteristics and policies aremost effective in promoting leaming. Given persistent econometric problems, a seriesof randomized interventions would probably provide the best evidence on the impactsof particular policies. Third, more information is needed on the role of child health,and on what policies (either in schools or in communities) best reduce children'shealth problems. While Sri Lanka's educational accomplishments are envied by manyother developing countries, much room remains for further progress.
Notes1. These percentages are almost exactly equal to those for Grade 1 pupils in Sri Lanka's 2003
school census.2. Sri Lanka's school year runs from January to December. The sample includes students who
repeated Grade 4, and thus were in Grade 4 in both 2002 and 2003. Note that the primaryrepetition rate is only 3%.
3. A third indicator of child nutritional status is weight-for-age, which reflects both malnutri-tion in early childhood and current malnutrition. In order to distinguish between these twotypes of malnutrition, this paper uses only height-for-age and weight-for-height.
4. NEREC originally planned to sample 1880 students from each of Sri Lanka's nine regions.In each region, about 100 schools were sampled, and 20 students were randomly sampledfrom each school. Only 94 schools with 20 students are needed for 1880 students, but a fewschools were added since only 10 students were sampled if schools had less than 20 fourthgraders.
5. These numbers imply that, for a given household, as long as there are other householdswithin one to two kilometers in two or more different directions, there are likely to be atleast two schools within two to three kilometers ofthat household.
6. For recent discussions of education production functions for developed and developingcountries, see Glewwe (2002), Todd and Wolpin (2003), Glewwe and Kremer (2006), andHanushek (2008).
7. Note that this distinction is not the same as that between exogeneity and endogeneity foreconometric estimation; the latter is concerned with whether a variable is correlated withthe error term in an econometric model, which is discussed in more detail in the nextsubsection. In this subsection, this distinction is made to clarify the distinction between theproduction function for cognitive skills and the reduced form equation for cognitive skills(the latter includes only exogenous variables).
8. In a more general setting, where years of schooling (S) varies and does so in part due tohousehold choices, that variable will also be a function of the explanatory variables inEquations (2) and (3), and substituting it out of Equation (1) will still yield Equation (4).
9. Later-born children may receive less prenatal nutrition if their mothers recently gave birthto their older sibling(s); see King (2003).
10. The idea here is that there is a fixed genetie endowment of innate ability, but it expressesitself in terms of abstract thinking ability over time as the child grows, so that ability interms of being able to leam in school depends not only on the genetic endowment but alsoon child age.
32 H. Aturupane et al.
11. There are no data on which students wear eyeglasses.12. The SLIS included no households from Mullaitivu or Kilinochchi districts; prices were
assigned from the districts with the longest borders along these districts.13. Household expenditure is closely related to household wealth, especially if households can
smooth consumption. See Dercon (2004) for a recent assessment of the extent of consump-tion smoothing in developing countries.
14. To see whether these estimates are biased by endogenous school choice within communi-ties, all estimates in Tables 3 and 4 were rerun using community (education zone) fixedeffects; the results are very similar. In another check for bias due to children not attendingthe nearest school, we re-estimated Tables 3, 4, 5, and 6, adding two dummy variables, oneindicating ehildren who were not attending the nearest school and the other indicating ehil-dren with missing data on whether they attended the nearest school; the results did notchange, and the dummy variables were never significant at the 5% level. Also, dropping allchildren not going to the nearest school from the sample had very little impact on theresults.
15. Preliminary regressions included worm egg counts and visual acuity. Egg counts wererarely significant, reflecting Sri Lanka's low incidence of helminth infections. Visual acuitywas never significant, probably because only 1% of the sampled pupils had serious visionproblems (Snellen ratio < 6/12 for both eyes). See Wisniewski (2010) for a recent analysisof child health and educational outcomes in Sri Lanka.
16. Recall that instruments are used only to avoid attenuation bias due to possible measurementerror in the observed variables. In particular, we do not claim that our instruments addressthe potential problem of omitted variable bias. As already explained (Subsection 4.4), ourmain approach for minimizing omitted variable bias is to use a large number of explanatoryvariables (and to use school fixed effects to avoid bias due to omitted school and teachervariables), but we recognize that some bias may remain in our results.
17. The Hausman test checks whether additional variables are endogenous, conditional on thefirst five being specified as endogenous (Davidson and MacKinnon 1993,241-2). AdditionalIVs were weight-for-age Z-score, parent is a library member, education trip expenditures,water source is tubewell, variables indicating how regutariy the child eats meals, distanceto nearest health clinic, missing school frequently in last year due to illness, parents helpchildren with schoolwork (NEREC), and whether parents arc alive.
18. Tests for the validity of the instrumental variables are mostly supportive. The overidentifi-cation test is not significant at the 5% level for the first language test and is barely signifi-cant at that level (p-value of 0.048) for the math test; however, the null hypothesis that theinstruments are not coiTelated with the error term is rejected at the I % level for the Englishtest (/)-value of 0.002). f-tests for weak instruments ranged from 4.88 (opinion) to 351.69(hope). Hausman tests that compare the uninstrumented results in columns (l)-(3) to theinstrument results in columns (4)-(6) decisively reject the (joint) null hypothesis that thecoefficients on the instrumented variables are the same.
19. For the parental taste variables, not only attenuation bias but also simultaneity bias is possi-ble; if students do well in school, parents may raise their opinions about the value of educa-tion. If this occurs, the instruments for these variables (which are second measurements)will be correlated with unobserved factors that increase test scores and so can yield incon-sistent results. Thus, even though the estimated impacts of parental taste variables are plau-sible, this estimation problem could lead to overestimation of their impacts.
20. A somewhat speculative explanation for this result is that well-off communities, which useEnglish in everyday life, are more attractive places for principals to work, and more expe-rienced principals are able to get transferred to those communities. If so, the direction ofcausality is from better English skills to more experienced principals.
21. One could argue that the single sex school variables are endogenous, since parents choosesuch schools for their children; estimates that exclude those variables yield similar resultsfor the other school variables.
ReferencesAlderman, H., J.R. Behrman, V. Lavy, and R. Menon. 2001. Child health and school enrollment:
A longitudinal analysis. Jotirnal ofHttman Resources 36, no. 1: 185-205.
Education Economics 33
Angrist, J., E. Bettinger, E. Bloom, E. King, and M. Kremer. 2002. Vouchers for privateschooling in Colombia: Evidence from a randomized natural experiment. AmericanEconomic Review 92, no. 5: 1535-58.
Angrist, J., and V. Lavy. 1999. Using Maimonides' rule to estimate the effect of class size onchildren's academic achievement. Quarterly Journal of Economics 114, no. 2: 533-76.
Angrist, J., and V. Lavy. 2002. New evidence on classroom computers and pupil learning.Economic Journal 112, no. 482: 735-86.
Barro, R. 1991. Economic growth in a cross-section of countries. Quarterly Journal ofEconomics \Q6, no. 2: AQl-Al,.
Bruns, B., A. Mingat, and R. Rakotomalala. 2003. Achieving universal primary education by2015: A chance for every child. Washington, DC: World Bank.
Case, A., and A. Deaton. 1999. School inputs and educational outcomes in South Africa.Quarterly Journal of Economics 114, no. 3: 1047-85.
Damon, A., and P, Glewwe. 2009. Tbree proposals to improve education in the LAC region:Estimates of the costs and benefits of each strategy. In Latin American development prior-ities: Costs and benefits, ed. B. Lomberg, 45-91. Cambrdige: Cambridge UniversityPress.
Davidson, R., and J.G. MacKinnon. 1993. Estimation and interference in econometrics. NewYork: Oxford University Press.
Dercon, S. 2004. Insurance against poverty. Oxford: Oxford University Press.Glewwe, P. 2002. Schools and skills in developing countries: Education policies and socio-
economic outcomes. Journal of Economic Literature 40, no, 2: 436—82,Glewwe, P,, M, Grosh, H. Jacoby, and M. Lockheed. 1995. An eclectic approach to estimating
the determinants of achievement in Jamaican primary education. World Bank EconomicReview 9, x\o. 2: 2'i\-5%.
Glewwe, P., and H. Jacoby. 1994. Student achievement and schooling choice in low-incomecountries: Evidence from Ghana. Journal of Human Resources 29, no. 3: 843-64.
Glewwe, P., H. Jacoby, and E. King. 2001. Early childhood nutrition and academic achieve-ment: A longitudinal analysis. Journal of Public Economics 81, no. 3: 345—68.
Glewwe, P., and M. Kremer. 2006. Schools, teachers and education outcomes in developingcountries. In Handbook of the economics of education, ed. E. Hanushek and F. Welch, vol.2, 945-1017. New York:'Elsevier.
Glewwe, P., M. Kremer, S. Moulin, and E. Zitzewitz. 2004. Retrospective vs. prospectiveanalyses of school inputs: The case of flip charts in Kenya. Journal of DevelopmentEconomics 74, no. 1: 251-68.
Glewwe, P., and E. Miguel. 2008. The impact of child health and nutrition on education indeveloping countries. In Handbook of development economics, ed. T.P. Schultz and J.Strauss, vol. 4, 3561-606. New York: Elsevier.
Hanushek, E. 1995. Interpreting recent research on schooling in developing countries. WorldBank Research Observer 10, no. 2: 227^6 .
Hanushek, E. 2008. Education production functions. In The new Palgrave dictionaiy of econom-ics, ed. S. Durlauf and L. Blume, 2nd ed. Basingstoke: Palgrave Macmillan. http://www.dictionaryofeconomics.com/art icle?id=pde2008_E00023 8.
King, J,C, 2003, The risk of matemal nutritional depiction and poor outcomes increases inearly or closely spaced pregnancies. Journal of Nutrition 133: 1732S-6S.
Lucas, R. 1988. On the mechanics of economic development. Journal of Monetary Economics22, no. 1: 3-42.
Mankiw, N.G., D. Romer, and D. Weil. 1992. A contribution to the empirics of economicgrowth. Quarterly Journal of Economics 107, no. 2: 407-37.
Miguel, E., and M. Kremer. 2004. Worms: Identifying impacts on education and health in thepresence of treatment externalities. Econometrica 72, no. 1: 159-217.
NEC (National Education Commission). 2004. An inter-sectoral study of education and healthfactors and student learning. Colombo: NEC.
NEREC (National Education Research and Evaluation Center). 2004. Achievement after fouryears of schooling: National report. Colombo: NEREC, University of Colombo.
OECD. 2003. Literacy skills for the world of tomorrow: Further results from the PISA 2000.Paris: OECD/UNESCO institute of Statistics.
34 H. Aturupane et al.
Pevalin, D. 2003. Outeomes in childhood and adulthood by mother's age at birth: Evidencefrom the 1970 British cohort study. Working Paper No. 2003-31, Institute for Social andEconomic Research (ISER), University of Essex.
Todd, P., and K. Wolpin. 2003. On the specification and estimation of the production functionfor cognitive achievement. Economic Journal 113, no. 485: F3-F33.
UNDP (United Nations Development Program). 2003. Human development report 2003:Millenium development goals: A compact among nations to end poverty. Oxford: OxfordUniversity Press.
Wisniewski, S. 2010. Child nutrition, health problems and school achievement in Sri Lanka.World Development 38, no. 3: 315-22.
World Bank. 2001. World development report 2000/2001 : Attaclcingpoverty. Washington, DG-World Bank.
World Bank. 2005. Treasures of the education system in Sri Lanka: Restoring performance,expanding opportunities and enhancing prospects. South Asia Region. Washington, DG:World Bank.
Table Al. Information collected in NEREC questionnaires.
/. Child questionnaire (completed by child at school)Age, sex, number of older brothers and sisters, and younger brothers and sistersLives with one or both parents, or somewhere elseLanguages spoken at homeSchool days without breakfast, and without meal after schoolAvailability of desk and/or chair at schoolMeans of transport to travel to school, and travel time requiredGrade first entered current school, and whether attended presehoolAccess to different types of textbooks and workbooksPerson at home who assists with schoolworkWhether attends private tuition (tutoring) classes, and attitudes on those classesDays absent from school, and reasons for absencesWatching TV and listening to radio, and if so favorite programsAvailability of newspapers at homeWhether teacher uses various books and/or visual aids in classAttitudes toward school subjects and other aspects of school
//. Parent questionnaire (completed by having parent come to the school)Relationship to the child (father, mother, other types of relatives, non-relativeAge (of respondent)Parents of child are still alive, and current marital status of parentsFather and mother live with child, or somewhere elseRace and religion of both parentsNumber of male and female children in family, and how many now in schoolWhere child lives while attending schoolType of dwelling, and whether child has bedroom, room for toys/books, gardenDwelling has water, electricity, toilets, telephone, various durable goodsEducation of mother, father, and older children, and occupation of each parentSources and amount (by six categories) of family incomeSchool expenses of different types, and who pays for those expensesWhether something interfered with child's education in 2002Languages used at homeChild's after school activitiesAvailability of newspapers in the household, and types of newspapersNumber of books in household, and whether use is made of librariesTV programs watched by child, and total hours per day child watches TVOther activities of child (e.g., cultural, religious, vacation)Discussions with child's teacher, with child; child participates in school activitiesWho helps child with schoolwork, hours per day child does school work at homeParticipation by child in tuition classes, including cost and hours per weekAny discussions on education with child's friends' parentsEducational aspirations for child, and opinions on child's education
Education Economics 35
Table AI. {Continued).
HI. Teacher questionnaireAge (10-year ranges) and sexType of living quarters, distance, and travel time from living quarters to schoolClass and grade of teacher training, highest educational and professional degreesYears experience as teacher, as primary teacher, and as teacher in current schoolDays trained in teaching of math, English, and local language, by training institutePossession of teacher handbook, skill list, class book, and workbook, by subjectMain sources of knowledge about teaching and leaming of Grade 4 studentsGrades taught in 2002Class size and prevalence of student absences, for Grade 4 in 2002Adequacy of 11 types of equipment (chairs, desks, blackboard, cupboard, etc.)Receipt of money for 'Quality Inputs' in 2002, and how many is spentOpinion of teaching environment (space, ventilation, noise, etc.)Who monitors teaching, and how often, and benefits provided by monitorsFraction of students reaching 'required level' in math, English, and local languageHow often leave is takenFrequency of using various teaching methods, and when began to prepare for 2002Number of students in special categories (dropout, repeat, orphan, disabled, etc.)Fraction of students who do not eat breakfast, do not have exercise books, pencilsOpinions about Grade 4 syllabus, by subject (math, English, local language)Opinions about Grade 4 'class books,' by subject (math, English, local language)Opinions about Grade 4 suggested activities, by subjectOpinions on 'list of essential learning skills,' and on how often to evaluate students
IV. Sectional head questionnaireSex, and current position/post heldYears of experience as deputy principal and sectional headHighest educational and professional degreesFacilities to conduct duties (room, storage area)Sufficiency of teacher handbooks, textbooks, and workbooks, by subjectWere 10 procedures followed in Grade 4 in 2002 (recordkeeping, discussions, etc.)Opinion of Grade 4 teachers' performance in 2002 in 13 different categories/tasksOpinion of Grade 4 teachers' monitoring of students perfonnance (nine activities)Opinion of Grade 4 teachers' ability/activities in influencing student learningOpinion on teacher evaluation methodsAgreement with 12 statements on improvements in the school and on monitoring
V. Principal questionnaire'Council,' age, and type of schoolTotal number of students and number of Grade 4 students and classes, all in 2002Age (10-year ranges), sex, and general education level of principalPrincipal's educational schooling and special trainingPrincipal's position (permanent or temporary), 'grade' and education service gradeYears of experience as principal, overall and in this schoolWhether principal lives in this schoolNumber of Grade 4 teachers, by sex and level of vocational/educational trainingFrequency with which Grade 4 teachers take leaveFrequency that Grade 4 teachers participate in voluntary trainingNumber of classes and classrooms, by grade, and numbers of other types of roomsOpinion on sufficiency of sports grounds, playground, and school gardenSufficiency of 12 types of facilities/equipment (e.g., water, electricity, telephones)Adequacy of toiletsDistances traveled by students and teachers to come to schoolDistance from school of seven types of public amenities (clinic, post office, bus, etc.)Frequency of bad behavior and thefts at school, and assessment of school securitySources of financial supportParticipation of principal in seven activities (teaching, supervision, meet parents, etc.)Opinion/satisfaction regarding 10 general issues
36 H. Aturupane et al.
Table A2. Information collected in NEC questionnaires.
/. Household questionnaire (administered by interviewer at the child's home)Child's age, sex, whether lives at home, and contact with home of not living thereBasic information on all household members (e.g., age, sex, education, occupation)Whether parents are alive, and whether they live with the childDwelling amenities for child (room, area for books/toys, garden, play area)Method child uses to go to sehool, and times he/she leaves home and returns homeHours per week in 11 after school activities (studying, working, playing, etc.)Frequency that child reads newspapersBooks at home that child can read, and child's access to library booksTypes of radio and television programs that child listens to or watchesCultural aetivities that family takes the child to (seven types)Parental contact with teachers and participation in school activitiesWho helps the child with schoolwork done at home, and how frequentlyParticipation in tuition (private tutoring classes), by academic subjectContact between child's parents and the parents of child's friendsParental expectations for child's education, and why parents value educationReasons ehild has missed school ¡n past yearAgreement with 17 statements regarding education and the child's schoolAbsence from sehool in past year due to health problemsHas child ever had a serious illness (one that lasted more than two weeks)Illnesses during the past month (diarrhea, cold, asthma, fever)Number of child bouts of malaria (ever, last three months)Number of times that child has had worms (ever, last three months, last year)Whether child has vision, hearing, or other disability/problem.Adequacy of child's diet, and events that lead to household food shortagesDetailed list of foods eaten by child in the previous dayChild health habits (drink boiled water, use of latrine, hours of sleep)Dwelling information (type, number of rooms, water source, toilet, energy source)Consumer durable (20 items) and productive assets (11 items) owned by householdDistance from dwelling to nearest, and most often used, of seven amenitiesHousehold monthly expenditures on food and 13 non-food itemsHousehold income in past year, by source of income, and self-perceived statusDetailed expenditure on health and education items, for child and for othersCredit and savings history in past five years, and eurrent savingsWhether household had a negative income shock, and what effect it hadHave any children dropped out of schoolHousehold participation in community organizations/societiesLanguages spoken at home, and family activities and discussion topicsNewspapers read, and books owned or borrowed, by family membersMethods used to discipline and reward the child
//. School questionnaire (administered at the child's school)Location and type of schoolWhere students eome from, and whether any reside at the sehoolDistance from school to closest of 14 amenities (clinic, market, govt. office)Student enrollment by grade and sexNumber of teachers, by sex and educational baekgroundNumber of Grade 4 teachers, by trainingLeave (absences) taken by Grade 4 teachers in 2004, by type of leaveAdequaey of supplies of 15 kinds of items (desks, blackboards, cupboards, etc.)Availability of larger items in 2002 (reading room, garden, first aid box, etc.)Whether textbooks were received on time in 2002, and if not how lateExistence of protective wall for sehool, and adequacy of toilets for ehildrenSex, educational background, and type of appointment of sehool principalYears of experience of school principalWhere principal lives, and distance and travel time to schoolLeave (absences) taken by principal in 2002
Education Economics 37
Table A2. (Continued).
Frequency with which principal supervises 'Key Stage 2 teachers'Frequency of school staff meetings to discuss education issuesQuestions on 'school family' activitiesNumber of PTA meetings and school developing society meetings held in 2002Number of times principal attended education refonn programs in 2002Parental awareness programs held in 2002, and reasons for themSpecial projects undertaken in 2002 by school development societySex, educational background, and type of appointment of all Grade 4 teachersYears of service and marital status of all Grade 4 teachersWhere each Grade 4 teacher lives, and distance and travel time to schoolIn service training and school inspectors visits for all Grade 4 teachersQuestions on 'school family' activities for all Grade 4 teachersHours Grade 4 teachers teach students 'after hours,' why and which subjectsMethods teacher uses to inform parents of children's progressAdequacy of equipment and materials of Grade 4 teachers in 2002Adequacy of Grade 4 student textbooks, class books, workbooks, and pens in 2002Child's sex, language spoken at home, and attendance of pre-schoolChild's favorite play aetivities, favorite school subjects, and activities after schoolChilds scores on 'Grade 4 learning competencies' (language, numeracy, environment)Student prizes/awards, and bad behaviorStudent attendance in 2002 (from school records), and date of birthClassroom observation of Grade 4 teachers (11 categories)
Copyright of Education Economics is the property of Routledge and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission. However, users
may print, download, or email articles for individual use.
