School attendance and student achievement: Evidence from rural Honduras

Economic Development and Cultural Change

Volumn 47, Issue 3, 1999, Pages 657-682

  Bedi, Arjun S ^a   Marshall, Jeffery H ^b  

^a UNIVERSITY OF BONN   (Germany)

^b U for the Msrmt of Educ Quality   (Honduras)

Author keywords

[No Author keywords available]

Indexed keywords

EDUCATIONAL ATTAINMENT; RETURNS TO EDUCATION; RURAL SOCIETY;

HONDURAS;

EID: 0032768672     PISSN: 00130079     EISSN: None     Source Type: Journal    
DOI: 10.1086/452423     Document Type: Article

References (47)

1. M. Lockheed, A. Verspoor, and Associates, Improving Primary Education in Developing Countries (London: Oxford University Press, 1991).
2. For an overview of Honduran educational coverage, see J. H. Y. Edwards, "The Status of Primary Education in Honduras" (Tegucigalpa, Honduras: USAID, 1995); J. H. Y. Edwards, B. Fuller, and S. Parandekar, "Primary Education Efficiency in Honduras: What Remains to Be Done?" (Tegucigalpa, Honduras: USAID, 1996); and N. Van Steenwyck and M. T. Mejia, Honduras 2000: Recursos humanos y perspectivas para el crecimiento económico (Honduras 2000: Human resources and perspectives on economic growth) (Tegucigalpa, Honduras: USAID, 1996).
3. For an overview of Honduran educational coverage, see J. H. Y. Edwards, "The Status of Primary Education in Honduras" (Tegucigalpa, Honduras: USAID, 1995); J. H. Y. Edwards, B. Fuller, and S. Parandekar, "Primary Education Efficiency in Honduras: What Remains to Be Done?" (Tegucigalpa, Honduras: USAID, 1996); and N. Van Steenwyck and M. T. Mejia, Honduras 2000: Recursos humanos y perspectivas para el crecimiento económico (Honduras 2000: Human resources and perspectives on economic growth) (Tegucigalpa, Honduras: USAID, 1996).
4. For an overview of Honduran educational coverage, see J. H. Y. Edwards, "The Status of Primary Education in Honduras" (Tegucigalpa, Honduras: USAID, 1995); J. H. Y. Edwards, B. Fuller, and S. Parandekar, "Primary Education Efficiency in Honduras: What Remains to Be Done?" (Tegucigalpa, Honduras: USAID, 1996); and N. Van Steenwyck and M. T. Mejia, Honduras 2000: Recursos humanos y perspectivas para el crecimiento económico (Honduras 2000: Human resources and perspectives on economic growth) (Tegucigalpa, Honduras: USAID, 1996).
5. Honduran Secretary of Public Education (SPE) data show that, for the entire country, official rates of grade repetition for 1996 were 19.4% in grade 1, 11.6% in grade 2, and 8.6% in grade 3. See N. Van Steenwyck, "Los rendimientos académicos y la reprobación de alumnos" (Academic achievement and repetition) (Tegucigalpa, Honduras: USAID/PEEP, 1997), p. 8. Dropout rates are much more difficult to calculate, but the World Bank has estimated that nationwide, 5% of Honduran primary school students desert annually. See "Staff Appraisal Report: Honduras Basic Education Project" (Washington, D.C.: World Bank, 1995), p. 9. Finally, in terms of academic achievement, a recent national application of criterion-referenced Spanish and mathematics exams resulted in national averages for grade 3 students of 39.7% and 35.9% in Spanish and mathematics, respectively. The criterion to pass the exam is a score of 60%, thus indicating that the average student (i.e., providing about two-fifths of the correct answers) needs to perform considerably better to pass the examinations. See UMCE, "Factores asociados con el rendimiento escolar en la escuela primaria Hondureña" (Factors associated with primary school academic achievement in Honduras) (Tegucigalpa, Honduras: Secretary of Public Education/World Bank, 1998).
6. Honduran Secretary of Public Education (SPE) data show that, for the entire country, official rates of grade repetition for 1996 were 19.4% in grade 1, 11.6% in grade 2, and 8.6% in grade 3. See N. Van Steenwyck, "Los rendimientos académicos y la reprobación de alumnos" (Academic achievement and repetition) (Tegucigalpa, Honduras: USAID/PEEP, 1997), p. 8. Dropout rates are much more difficult to calculate, but the World Bank has estimated that nationwide, 5% of Honduran primary school students desert annually. See "Staff Appraisal Report: Honduras Basic Education Project" (Washington, D.C.: World Bank, 1995), p. 9. Finally, in terms of academic achievement, a recent national application of criterion-referenced Spanish and mathematics exams resulted in national averages for grade 3 students of 39.7% and 35.9% in Spanish and mathematics, respectively. The criterion to pass the exam is a score of 60%, thus indicating that the average student (i.e., providing about two-fifths of the correct answers) needs to perform considerably better to pass the examinations. See UMCE, "Factores asociados con el rendimiento escolar en la escuela primaria Hondureña" (Factors associated with primary school academic achievement in Honduras) (Tegucigalpa, Honduras: Secretary of Public Education/World Bank, 1998).
7. Honduran Secretary of Public Education (SPE) data show that, for the entire country, official rates of grade repetition for 1996 were 19.4% in grade 1, 11.6% in grade 2, and 8.6% in grade 3. See N. Van Steenwyck, "Los rendimientos académicos y la reprobación de alumnos" (Academic achievement and repetition) (Tegucigalpa, Honduras: USAID/PEEP, 1997), p. 8. Dropout rates are much more difficult to calculate, but the World Bank has estimated that nationwide, 5% of Honduran primary school students desert annually. See "Staff Appraisal Report: Honduras Basic Education Project" (Washington, D.C.: World Bank, 1995), p. 9. Finally, in terms of academic achievement, a recent national application of criterion-referenced Spanish and mathematics exams resulted in national averages for grade 3 students of 39.7% and 35.9% in Spanish and mathematics, respectively. The criterion to pass the exam is a score of 60%, thus indicating that the average student (i.e., providing about two-fifths of the correct answers) needs to perform considerably better to pass the examinations. See UMCE, "Factores asociados con el rendimiento escolar en la escuela primaria Hondureña" (Factors associated with primary school academic achievement in Honduras) (Tegucigalpa, Honduras: Secretary of Public Education/World Bank, 1998).
8. This does not, of course, deny the importance of other factors such as direct costs of schooling (in some contexts) or malnutrition. However, the bulk of the literature has identified low school resources and opportunity costs as the two most important causes of wastage. See, e.g., J. B. Gomes-Neto and E. A. Hanushek, "The Causes and Consequences of Grade Repetition: Evidence from Brazil," Economic Development and Cultural Change 43, no. 1 (1994): 117-48.
9. World Bank, pp. 9-10.
10. H. Jacoby, "Borrowing Constraints and Progress through School: Evidence from Peru," Review of Economics and Statistics 76, no. 1 (1994): 151-60.
11. This is especially true in rural Honduras where children may drop out of school for several months during the harvest season only to return the next year. See World Bank, p. 9.
12. Of the 652 children in grade 1 in our sample, 15 have dropped out and 147 did not attend on the day the test was administered. For grades 2 and 3 there are 15 and 16 dropouts, respectively, and 153 and 164 no-shows. The apparently small number of dropouts in the sample is due to our inability to clearly distinguish between permanent and temporary dropouts. It is quite possible that some children who did not show up on the day the test was given have dropped out. However, this number is not likely to be large, and without additional information we simply cannot be sure.
13. See J. J. Heckman, "Sample Selection Bias as a Specification Error," Econometrica 47, no. 1 (1979): 153-62. The correction above does not account for selection bias associated with children who never registered in school. However, bias from this source is probably small, as almost all primary-school-age children in Honduras register for school initially.
14. Parental migration in response to school quality is quite unlikely in the Honduran context. Analysis of household survey data reveals that 90% of individuals migrate in order to find work, while the remaining migrate for family reasons. See A. S. Bedi, "The Importance of School Quality as a Determinant of Earnings in a Developing Country," International Journal of Educational Development 17, no. 4 (1997): 427-37. While differing by grade, in general the data indicate that areas with greater poverty (higher percentage of people below the poverty line) have higher quality teachers (according to our measure of teacher quality that will be discussed below). For instance, in grades 2 and 3, this correlation between poverty and quality of teachers is 0.2108 and 0.2496, respectively.
15. The standard errors are based on an error-components model (randomeffects model) that allows school- and individual-specific effects in the error term. Several examples of the extent of the downward bias in the presence of intragroup correlation are provided in B. R. Moulton, "Random Group Effects and the Precision of Regression Estimates," Journal of Econometrics 32, no. 2 (1986): 385-97. An alternative approach to capture school-specific effects is to allow the slope parameters to vary across schools and, in essence, to estimate school-specific educational production functions. While feasible, the small number of observations per school in our sample does not allow a pursuit of this approach.
16. Since the tests were criterion referenced, the results demonstrate the low academic achievement of students in rural Honduras. Few children in grades 2 and 3 were able to pass both the Spanish and mathematics exams (pass considered to be 60% and above). Another notable feature is the sharp decline in the average test scores between grade 1 and grades 2 and 3. While it is difficult to explain the decline, one possibility is the differential nature of the test application. Students in grade 1 were administered the test orally, i.e., students were read the questions and possible answers out loud and asked to mark the correct one on the answer sheet, while students in the other grades were administered a written examination. This differential application may have led to inflated and less accurate scores for grade 1 students.
17. For instance, the average is 26 in Ghana, 25 in northeast Brazil, and 21 for the United States. See P. Glewwe and H. Jacoby, "Student Achievement and Schooling Choice in Low-Income Countries: Evidence from Ghana," Journal of Human Resources 29, no. 3 (1993): 843-64; R. W. Harbison and E. A. Hanushek, Educational Performance of the Poor: Lessons from Rural Northeast Brazil (New York: Oxford University Press, 1992); and Unesco, Year Book of Education (Paris: Unesco, 1992).
18. For instance, the average is 26 in Ghana, 25 in northeast Brazil, and 21 for the United States. See P. Glewwe and H. Jacoby, "Student Achievement and Schooling Choice in Low-Income Countries: Evidence from Ghana," Journal of Human Resources 29, no. 3 (1993): 843-64; R. W. Harbison and E. A. Hanushek, Educational Performance of the Poor: Lessons from Rural Northeast Brazil (New York: Oxford University Press, 1992); and Unesco, Year Book of Education (Paris: Unesco, 1992).
19. For instance, the average is 26 in Ghana, 25 in northeast Brazil, and 21 for the United States. See P. Glewwe and H. Jacoby, "Student Achievement and Schooling Choice in Low-Income Countries: Evidence from Ghana," Journal of Human Resources 29, no. 3 (1993): 843-64; R. W. Harbison and E. A. Hanushek, Educational Performance of the Poor: Lessons from Rural Northeast Brazil (New York: Oxford University Press, 1992); and Unesco, Year Book of Education (Paris: Unesco, 1992).
20. For discussions about measuring school effectiveness, see B. Fuller and P. Clarke, "Raising School Effects While Ignoring Culture? Local Conditions and the Influences of Classroom Tools, Rules and Pedagogy," Review of Educational Research 64, no. 1 (1994): 119-57; B. Fuller and C. Snyder, "Vocal Teachers, Silent Pupils? Life in Botswana Classrooms," Comparative Education Review 35 (1991): 274-94; and A. R. Riddell, "Assessing Designs for School Effectiveness Research and School Improvement in Developing Countries," Comparative Education Review 41 (1997): 178-204.
21. For discussions about measuring school effectiveness, see B. Fuller and P. Clarke, "Raising School Effects While Ignoring Culture? Local Conditions and the Influences of Classroom Tools, Rules and Pedagogy," Review of Educational Research 64, no. 1 (1994): 119-57; B. Fuller and C. Snyder, "Vocal Teachers, Silent Pupils? Life in Botswana Classrooms," Comparative Education Review 35 (1991): 274-94; and A. R. Riddell, "Assessing Designs for School Effectiveness Research and School Improvement in Developing Countries," Comparative Education Review 41 (1997): 178-204.
22. For discussions about measuring school effectiveness, see B. Fuller and P. Clarke, "Raising School Effects While Ignoring Culture? Local Conditions and the Influences of Classroom Tools, Rules and Pedagogy," Review of Educational Research 64, no. 1 (1994): 119-57; B. Fuller and C. Snyder, "Vocal Teachers, Silent Pupils? Life in Botswana Classrooms," Comparative Education Review 35 (1991): 274-94; and A. R. Riddell, "Assessing Designs for School Effectiveness Research and School Improvement in Developing Countries," Comparative Education Review 41 (1997): 178-204.
23. E. A. Hanushek, "The Economics of Schooling: Production and Efficiency in Public Schools," Journal of Economic Literature 24, no. 3 (1986): 1141-77, claims that the failure of empirical studies to account for a preponderance of the variation in achievement between classrooms and schools is largely a result of inadequate measures of teacher quality, many of which yield inconsistent or statistically insignificant results. The problem is not that teachers do not affect learning outcomes but that the measures incorporated to date have not been able to effectively differentiate among teachers by skill.
24. For an overview of empirical analyses incorporating teacher quality measures, see Fuller and Clarke; and Harbison and Hanushek, chap. 2. The overwhelming majority of these studies incorporate teacher skill proxies, as opposed to explicit measures of teaching methodology. Nevertheless, some have included indicators of "active" teaching pedagogy or training. See M. Lockheed, J. Fonacier, and L. Bianchi, "Effective Science Teaching in the Philippines" (Washington, D.C.: World Bank, 1988); and G. Psacharopoulos, C. Rojas, and E. Velez, "Achievement Evaluation of Colombia's Escuela Nueva (New School): Is Multigrade the Answer?" Comparative Education Review 37 (1993): 263-76.
25. For an overview of empirical analyses incorporating teacher quality measures, see Fuller and Clarke; and Harbison and Hanushek, chap. 2. The overwhelming majority of these studies incorporate teacher skill proxies, as opposed to explicit measures of teaching methodology. Nevertheless, some have included indicators of "active" teaching pedagogy or training. See M. Lockheed, J. Fonacier, and L. Bianchi, "Effective Science Teaching in the Philippines" (Washington, D.C.: World Bank, 1988); and G. Psacharopoulos, C. Rojas, and E. Velez, "Achievement Evaluation of Colombia's Escuela Nueva (New School): Is Multigrade the Answer?" Comparative Education Review 37 (1993): 263-76.
26. The exam was designed by M. T. Mejia and R. Paz at USAID/Honduras.
27. Other studies have incorporated measures of teacher disposition toward innovation. See, e.g., J. P. Tan, J. Lane, and P. Coustère, "Putting Inputs to Work in Elementary Schools: What Can Be Done in the Philippines?" Economic Development and Cultural Change 45, no. 4 (1997): 857-79. The instrument used in this study was primarily designed to measure specific knowledge of the active and participative methodology rather than general attitudes toward teaching. Nevertheless, without classroom observations the interpretation of the test score must be limited to "disposition toward" or "knowledge of" active and participative teaching rather than "use of."
28. The World Bank (n. 3 above) estimates that in reality the number of school days in a year ranges from 100 to 120. While the use of this number will not alter our estimates (only a shift in the intercept), it does have an important policy implication. If low attendance is mainly a supply-side problem - that is, teachers are absent or classes are canceled - then a focus only on the demand side may obscure the true nature of the problem. While acknowledging this possibility, we retain a focus on the demand side and concentrate on the factors motivating school attendance on the days that school is offered.
29. 2 ranging from 0.56 to 0.91. The latter requirement is harder to satisfy. While the exclusion of some of the instruments may not be questioned (e.g., teacher age), the validity of excluding others (number of seminars, enrollment) may be questioned. The presence of overidentifying instruments allows us to conduct tests of overidentifying restrictions based on the null hypothesis that equation (3) is correctly specified and that the instruments are valid. Based on the mathematics test equations, these tests revealed that the null hypothesis is not rejected for grades 1 (P-value = .322) and 3 (P-value = .738) but is rejected for grade 2 (P-value = .028). Similar results hold for the Spanish test equations. Thus, the results for grade 2 must be interpreted with caution, as the tests suggest that the instruments are not valid. For a discussion of the test, see R. Davidson and J. G. MacKinnon, Estimation and Inference in Econometrics (New York: Oxford University Press, 1993), pp. 232-42.
30. Other studies have consistently shown that boys in Honduras are more likely to repeat grades and drop out of school than are girls. See Edwards, Fuller, and Parandekar (n. 2 above).
31. Although we recognize the possibility that attending preschool may be an endogenous decision, we do not have the required data to address this issue in detail. Similar to several other articles (e.g., Tan, Lane, and Coustère), we include this variable and assume that it is uncorrelated with the error term.
32. See R. Halpern and R. Myers, "Effects of Early Childhood Intervention on Primary School Progress and Performance in Developing Countries" (Washington, D.C.: USAID, 1985); and J. C. Palafox, J. Prawda, and E. Velez, "Primary School Quality in Mexico," Comparative Education Review 38 (1993): 167-80.
33. See R. Halpern and R. Myers, "Effects of Early Childhood Intervention on Primary School Progress and Performance in Developing Countries" (Washington, D.C.: USAID, 1985); and J. C. Palafox, J. Prawda, and E. Velez, "Primary School Quality in Mexico," Comparative Education Review 38 (1993): 167-80.
34. The need for greater community participation in Honduran education has been recognized by the World Bank, which has implemented a variety of programs designed to increase the links between parents, community leaders, and schools.
35. Teacher education does not appear in our regressions, since almost all teachers have the same educational qualification - a high school teaching certificate.
36. The program is based in part on the Escuela Nueva (new school) program in Colombia, which has had tremendous success in improving school quality in rural Colombia. See Psacharopoulos, Rojas, and Velez (n. 18 above); and E. Schiefelbein, In Search of the School of the 21st Century: Is the Colombian Escuela Nueva the Right Pathfinder? (Santiago: Unesco, 1991). For an overview of the Honduran program, see M. Lansing, "Investigación sobre el impacto de la metodologia participativa en las aulas de segundo y cuarto grado en 24 escuelas primarias de occidente" (An analysis of the impact of an active and participative teaching methodology in 24 second and fourth grade classrooms in western Honduras) (Tegucigalpa, Honduras: USAID, 1995).
37. The program is based in part on the Escuela Nueva (new school) program in Colombia, which has had tremendous success in improving school quality in rural Colombia. See Psacharopoulos, Rojas, and Velez (n. 18 above); and E. Schiefelbein, In Search of the School of the 21st Century: Is the Colombian Escuela Nueva the Right Pathfinder? (Santiago: Unesco, 1991). For an overview of the Honduran program, see M. Lansing, "Investigación sobre el impacto de la metodologia participativa en las aulas de segundo y cuarto grado en 24 escuelas primarias de occidente" (An analysis of the impact of an active and participative teaching methodology in 24 second and fourth grade classrooms in western Honduras) (Tegucigalpa, Honduras: USAID, 1995).
38. It is important to note that the teacher test score is primarily a result of the training received in the active teaching methodology. A regression of teacher test score on teacher characteristics such as experience, age, education, and the number of seminars attended, revealed that only the last variable was important. It was statistically significant (t-statistic = 4.1) and indicated that attending a seminar increased teacher test score by 6.22 points.
39. See Fuller and Clarke (n. 16 above).
40. The presence of heteroscedasticity will cause the tobit coefficients to be inconsistent and inefficient. However, the results from the tobit specification of the school attendance equation are very similar to the least-squares estimates. Detailed estimates are available. Also, noting the small differences between OLS and tobit estimates, we interpret the tobit coefficients as if they are marginal effects. We also estimated school attendance using an ordered probit model that split the sample into dropouts, part-time, and full-time school attendees. The results were similar and displayed the importance of age and the expected score.
41. The process of accumulating years of schooling in Honduras can be thought of as a race against time, given that many children are forced to leave when they reach adolescence. Edwards, Fuller, and Parendekar therefore suggest lowering the official primary school entrance age from seven to six. The data in this analysis support this recommendation, since age exerts a negative influence on attendance but has no effect on achievement.
42. These percentage increases are a lower bound (based on a school year of 160 days). If we were to work with the World Bank's estimate of the school year (100-120 days) these increases translate into a 3% to 7% increase in school attendance.
43. As argued earlier, school attendance affects test scores and consequently repetition and school desertion. This suggests the estimation of an education production function that includes predicted school attendance. This approach was precluded due to lack of identifying information. Additionally, lack of data on repetition and dropout rates prevents us from assessing the impact of school attendance on these outcomes. However, to provide some evidence of the importance of attendance, we estimated the impact of actual school attendance on test scores (see table Al). These estimates show that school attendance exerts considerable influence on mathematics and Spanish test scores, at least in grades 1 and 2. For instance, an increase in school attendance by 5 days increases grade 2 mathematics and Spanish scores by approximately 1.5 points. While the impact is large, it declines across grades. This decline may seem surprising, but it is important to interpret these results cautiously, as they are susceptible to endogeneity bias.
44. See, e.g., Harbison and Hanushek (n. 15 above), p. 132; and M. E. Lockheed and E. A. Hanushek, "Improving Educational Efficiency in Developing Countries: What Do We Know?" Compare 18 (1988): 21-38.
45. In 1996 the exchange rate was 12.9 lempiras per U.S. dollar.
46. R. J. Willis and S. Rosen, "Education and Self-Selection," Journal of Political Economy 87, no. 5 (1979): S1-S36.
47. See Jacoby, "Borrowing Constraints" (n. 6 above).