Cuba's academic performance in comparative perspective

Comparative Education Review

Volumn 49, Issue 2, 2005, Pages 230-261

  Carnoy, Martin ^a   Marshall, Jeffery ^a  

^a NONE

Author keywords

[No Author keywords available]

Indexed keywords

EID: 20744442862     PISSN: 00104086     EISSN: None     Source Type: Journal    
DOI: 10.1086/428721     Document Type: Review

References (68)

1. Evidence for Cuba's exceptional performance was given by UNESCO's Latin American Laboratory of Educational Evaluation (Laboratorio Latinoamericano de Evaluatión de la Calidad de la Educatión, or LLECE; hereafter, Laboratorio); see Primer estudio international comparativo sobre lenguaje, matemática y factores asociados en tercero y cuario grado (Santiago: UNESCO, 1998).
2. For earlier, qualitative research that focuses on Cuban education, see Martin Carnoy and Jorge Werthein, Cuba: Cambio económica y reforma educativa, 1955-1978 (Mexico City: Editorial Nueva Imagen, 1980);
3. L. Gasperini, The Cuban Education System: Lessons and Dilemas, World Bank Country Studies Education Reform and Management Publication Series, vol. 1, no. 5 (Washington, DC: World Bank, 2000).
4. J. S. Coleman, E. Q. Campbell, C. J. Hobson, J. McPartland, A. M. Mood, F. D. Weinfeld, and R. L. York, Equality of Educational Opportunity (Washington, DC: U.S. Department of Health, Education, and Welfare, Office of Education, 1966);
5. Marlaine Lockheed and Adriaan Verspoor, Improving Primary Education in Developing Countries (Washington, DC: Oxford University Press, for the World Bank, 1991);
6. Organization for Economic Cooperation and Development (OECD), Literacy Skills for the World of Tomorrow: Further Results from PISA 2000 (Paris: OECD, 2003);
7. Richard Rothstein, Class and Schools (New York: Teachers College Press, 2004).
8. James S. Coleman, "Social Capital in the Creation of Human Capital," American Journal of Sociology 94 (1988): S95-S120.
9. We are aware that "socioeconomic background/status" rather than social class emerged in the literature as a social functionalist terminology to supersede the historical discussion of the term "class" (see, e.g., Talcott Parsons, Social Systems and the Evolution of Action Theory [New York: Free Press, 1977]). Social class in Marxian theory is rooted in the social relations of production, so that class in capitalist societies is identified with the individual's or family's relation to ownership of the means of production (
10. Erik Olin Wright, Class Counts: Comparative Studies in Class Analysis [Cambridge: Cambridge University Press, 1977]). However, as workers in capitalist societies gained more political power, the argument goes, the Marxian definition of class became less relevant - in particular, it missed meaningful socioeconomic differences within the working class, differences that went beyond Marx's and Engels's notion of proletariat and lumpen-proletariat. Since, in Marxist theory, socialism eliminates social classes, this definition also missed socioeconomic differences in state socialist societies (see, e.g.
11. Rudolf Bahro, Die Alternative: Zur Kritik des real existierenden Sozialismus [Berlin: Tribüne, 1990]). Nevertheless, as we hope to make clear, we do not abandon the idea that socioeconomic background, or SES, has different meaning in societies with different social/economic organization.
12. Michael Brown, Martin Carnoy, Elliot Currie, Troy Duster, David B. Oppenheimer, Marjorie M. Schultz, and David Wellman, Whitewashing Race (Berkeley: University of California Press, 2003).
13. David Post, Children's Work, Schooling, and Welfare in Latin America (Boulder, CO: Westview, 2002).
14. Paul Willis, Learning to Labor (New York: Columbia University Press, 1981);
15. Richard Sennett and Jonathan Cobb, The Hidden Injuries of Class (New York: Vintage Books, 1973).
16. Rothstein, Class and Schools.
17. In a recent PhD dissertation, Uruguayan sociologist Tabaré Fernández develops a similar argument in the context of the relative reproduction of the class structure by the educational system in four Latin American countries: Argentina, Chile, Mexico, and Uruguay; see "Distribución del conocimiento escolar: Clases sociales, escuelas y sistema educativo en Latinamérica" (PhD dissertation, Colegio de México, 2004).
18. Coleman, "Social Capital," S98.
19. Coleman, "Social Capital," Ibid., S110.
20. For example, families that move often would, in this conception, have less social capital outside the family than those staying put. Other researchers have called these "neighborhood effects." See Christopher Jencks and Susan E. Mayer, "The Social Consequences of Growing Up in a Poor Neighborhood," in Inner-City Poverty in the United States, ed. Laurence Lynn and Michael McGeary (Washington, DC: National Academies Press, 1990), 111-86.
21. See also Robert D. Putnam, Making Democracy Work: Civic Traditions in Modern Italy (Princeton, NJ: Princeton University Press, 1993)
22. and Bowling Alone: The Collapse and Revival of American Community (New York: Simon & Schuster, 2000).
23. Putnam, Making Democracy Work and Bowling Alone.
24. See, e.g., Peter Evans, Embedded Autonomy: States and Industrial Transformation (Princeton, NJ: Princeton University Press, 1995);
25. Alice Amsden, Asia's Next Giant: South Korea and Late Industrialization (New York: Oxford University Press, 1989);
26. Alice Amsden and Wan-wen Chu, Beyond Late Development: Taiwan's Upgrading Policies (Cambridge, MA: MIT Press, 2003). We note that they were/are aided by (relatively authoritarian) family structures that place high value on academic success.
27. Theda Skocpol, States and Social Revolutions: A Comparative Analysis of France, Russia, and China (New York: Cambridge University Press, 1979);
28. Richard Fagen, The Transformation of Political Culture in Cuba (Stanford, CA: Stanford University Press, 1969).
29. Martin Carnoy and Joel Samoff, Education and Social Transformation in the Third World (Princeton, NJ: Princeton University Press, 1989).
30. Martin Carnoy and Joel Samoff, Education and Social Transformation in the Third World, 1989 Ibid.
31. This raises the important question of whether democratic capitalist societies that are highly regulative and stress socialist or communal values, such as the Scandinavian countries, have higher-quality education because of the relative social equality and emphasis on collective responsibility that characterizes them (see, e.g., Manuel Castells and Pekka Himanen, The Information Society and the Welfare State: The Finnish Model [Oxford: Oxford University Press, 2002]). The voters in those societies have voluntarily opted for a highly regulated social context. The OECD seems to take the position that greater equality in educational provision is an important explainer of higher performance on the PISA test
32. (see OECD, Literacy Skills).
33. There is evidence that schooling in ministate socialist conditions, such as those represented by military bases in the United States, seems to be much more effective for minority students than schools in civilian U.S. society (see Claire Smrekar, James Guthrie, Debra Owens, and Pearl Sims, March toward Excellence: School Success and Achievement in Department of Defense Schools [Nashville: Vanderbilt University Peabody Center for Educational Policy, 2001]).
34. Coleman finds significant negative effects on student performance of school changes; see "Social Capital."
35. See, e.g., Jeffery Marshall, "Build It and They Will Come" (PhD dissertation, Stanford University, School of Education, 2003).
36. Jencks and Mayer, "Social Consequences."
37. Fernśndez, "Distribución del conocimiento escolar."
38. See, e.g., J. Douglas Willms, "Patterns of Academic Achievement in Public and Private Schools: Implications for Public Policy and Future Research," in Comparing Public and Private Schools: School Achievement, ed. Edward Haertel, Thomas James, and Henry M. Levin (London: Falmer, 1987), 2:113-34;
39. Julian R. Betts, Andrew Zau, and Lorien Rice, Determinants of Student Achievement: New Evidence from San Diego (San Francisco: Public Policy Institute of California, 2003).
40. See, e.g., Luis Benveniste, Martin Carnoy, and Richard Rothstein, All Else Equal (New York: Routledge, 2002).
41. Marshall, "Build It and They Will Come."
42. Coleman et al., Equality of Educational Opportunity.
43. Eric Hanushek, "The Economics of Schooling: Production and Efficiency in Public Schools," Journal of Economic Literature 24, no. 3 (1986): 1141-77;
44. Alan Krueger, "An Economist's View of Class Size Research" (unpublished manuscript, Princeton University, 1999).
45. Cross-state or cross-province comparisons within the same country could capture such differences if there exists sufficient within-country variation and the educational and social service system is sufficiently decentralized. The United States, Brazil, and Germany are examples of countries with relatively decentralized education systems and some variation in the way states deal with social inequality.
46. John Knight and Richard Sabot, Education, Productivity, and Inequality (Washington, DC: Oxford University Press, for the World Bank, 1990).
47. Laboratorio, Primer estudio.
48. Not all the data sets are of equal quality. Countries often are missing considerable data for several variables, such that when we estimate the basic regression equations, the number of observations is much smaller than four thousand. It is reasonable to ask whether the very high test scores in Cuba are the result of picking a select group of schools or a select group of students in the schools surveyed. Laboratorio researchers returned to Cuba to retest students in five schools selected at random from the original 100 surveyed. They found no significant difference in results. However, this does not answer the question of whether the 100 schools were representative of Cuban schools. Our own classroom observations in 10 schools (rural, urban, and city periphery) suggested to us major differences in the level of performance of Cuban third graders compared with those in Brazilian and Chilean schools, but this, too, would not fully answer skeptics' doubts about randomness in the Cuban survey (and randomness of other countries' surveys as well).
49. A major study of rural schools in Brazil collected longitudinal data on students, schools, and teachers between 1982 and 1986 and identified a number of policy-relevant variables (such as learning materials and school hardware) that were not only significant determinants of achievement but also appeared to be cost-effective educational interventions; see Ralph A. Harbison and Eric A. Hanushek, Educational Performance of the Poor: Evidence from the Rural Northeast of Brazil (New York: Oxford University Press, 1992). In another ambitious study using Brazilian data, Bruce Fuller and a team of Brazilians measured test scores and conventional input variables and also observed classrooms, videotaped teachers teaching, and conducted extensive teacher interviews; see
50. Bruce Fuller et al., "How to Raise Children's Early Literacy: The Influence of Family, Teacher, and Classroom in Northeast Brazil," Comparative Education Review 43, no. 1 (1999): 1-35. Their findings showed that accounting for teaching methodology is important when considering the determinants of academic achievement.
51. Our estimation method is subject to the critique that with multilevel observations (clusters of students nested within schools) we should be using multilevel modeling, and that our OLS estimates will necessarily bias error terms (although not the regression coefficients themselves). By using robust standard errors that account for intracluster correlation we can deal with the standard error issue. Multilevel models are especially useful for studying interaction since they allow slopes to vary by school. But for the second phase of our analysis - the simulation of student achievement - we are primarily interested in the mean effect of each variable on achievement and are less concerned about variation in parameters within each country. For comparison's sake we estimated these models using the HLM program, and the results are not much different (see also J. Douglas Willms and Marie-Anne Somers, "Family, Classroom, and School Effects on Children's Educational Outcomes in Latin America," International Journal of School Effectiveness and Improvement 12, no. 4 [2001]: 409-45).
52. There is increasing evidence that primary school teachers in Latin America receive very little, if any, additional courses in mathematics and language in their preservice higher education (see OECD, Literacy Skills). Thus, if subject content knowledge is key to increasing student learning, a higher fraction of teachers with teacher education at university may not have the expected impact on how much students learn.
53. William H. Schmidt, Curtis C. McKnight, Richard T. Houang, HsingChi Wang, David E. Wiley, Leland S. Cogan, and Richard G. Wolfe, Schools Matter (San Francisco: Jossey-Bass, 2002).
54. Martin Carnoy, Jeffery Marshall, and Miguel Socias, "How Do School Inputs Influence Math Scores? A Comparative Approach" (unpublished manuscript, Stanford University School of Education, 2004).
55. The factor analysis was conducted using school averages for parental education and job prestige, and identified one factor that accounted for 61 percent of the variance. The loadings were 0.83 for parental education and 0.81 for job prestige.
56. Susanna Loeb, Bruce Fuller, Sharon Lynn Kagan, Bidemi Carroll, Judith Carroll, and Jan McCarthy, "Child Care in Poor Communities: Early Learning Effects of Type, Quality, and Stability," National Bureau of Economic Reasearch Working Paper no. W9954 (NBER, Cambridge, MA, September 2003).
57. Of course, working on the family farm is a common activity for rural young children even in the developed countries, but it still expresses family economic needs and a continuing culture of using child labor.
58. Post, Children's Work. The Laboratorio survey asks students whether they work rarely, sometimes, or frequently. The "free from work" variable asks them if they do what they choose to do when not in school. In order to maintain the number of cases in each country we recoded missing values to one (rarely work) for everyone. This makes the unrealistic argument that those who did not answer the question worked rarely. When we compare the characteristics of children who were missing these data in countries other than Cuba, they generally appeared to be from lower-SES families. Thus, in non-Cuba countries, those who did not answer are probably working more, not less. But missing values constitute a very small proportion of the total sample. In contrast, a relatively high percentage of Cuban children did not respond to the question about work outside the home, yet their SES characteristics are very similar to the mean of the sample. In Cuba, a tiny percentage of children reported working outside the home. We are therefore almost certainly on solid ground in receding the missing values as "rarely."
59. We sum the ordinal values of the individual responses to child labor outside the home (a proxy for family income), parental education, and books in home, then average this index across students in each sampled school to get a school average factor.
60. As in all our variables, the greater the variance in a particular variable, the larger we would expect the estimated coefficient of the variable to be, other factors equal. For example, we would expect a larger peer effect on individual student achievement when students are more divided up among different quality schools by student socioeconomic background characteristics. Consider the situation where students of different SES are distributed randomly among schools and classrooms. The peer coefficient would be zero. The greater the SES segregation across schools, the greater the peer effect, provided that students learn more when surrounded by higher-SES (and presumably more academically able) peers. If the coefficient of average SES is large and positive, a small increase in average school SES could have a large impact on average student performance. But if the coefficient is small, this means that how students are grouped in school has little effect on how much they learn. This may be the result of fairly equal social class distribution of pupils across schools, once rural/urban and private/public school differences in test scores are accounted for, or it may result from homogeneously high or low academic standards across schools, independent of average student social class.
61. For this methodology, see Ronald Oaxaca, "Male-Female Wage Differentials in Urban Labor Markets," International Economic Review 14, no. 3 (1973): 693-709;
62. and Martin Carnoy, Faded Dreams (New York: Cambridge University Press, 1994). Although both these studies simulate earnings, the methodology is applicable to other types of dependent variables, such as student achievement.
63. It is also possible to simulate the contribution of effectiveness of schooling across countries by comparing achievement results when the Cuban coefficients are applied to the average resources of each of the other countries. We made these estimates and can make them available to interested readers.
64. One of the mysteries of the Cuban results is the small difference between third- and fourth-grade test scores (on the same test but different students taking the test). One possible explanation is that the test was sufficiently easy for Cuban students that a high fraction of both third and fourth graders achieved perfect scores, so that it was difficult to achieve much higher average scores in the fourth grade. This is called "topping out." We observe the same phenomenon in Bolivia, but the topping out explanation is highly unlikely in that case.
65. We also estimate how much Cuban students' scores would change were Cuban student, school, and school social context characteristics equal to the means in other countries, but the "payoffs" (estimated coefficients) estimated for Cuba remained the same. However, we do not report these results because some of the more important Cuban coefficients, such as for work outside the home and student fights, are extremely large, mainly because working outside the home and classroom fights are so unusual in Cuba. Thus, they represent an effect of the variable in question at a very different point on the policy curve than in other countries. We felt that such results were difficult to interpret and possibly misleading. For the interested reader, they are available from the authors.
66. The detailed results of the simulations can be found in the online version of this article.
67. It is important to note, however, that all of these results may be subject to omitted variable bias. That is, Cuba may have made successful efforts to reduce child labor and may have also developed very good teachers, factors that were poorly measured by the Laboratorio survey. The observed correlation between low child labor and test scores may imply that this is a good policy (reducing child labor), when in reality this relation is largely spurious and instead driven by the good teacher effect that is also correlated with low levels of child labor. It may not be a lack of child labor per se that is driving higher test scores but instead a menu of things that we are not measuring well (especially dealing with teachers) that are also correlated (especially in Cuba but in other countries as well) with lower levels of child labor. In a structural sense our argument about Cuban social capital would remain correct: Cuba does a number of things that improve student achievement. Yet, as we move on to specific components, there is the question about whether or not these are the actual causal policy variables.
68. Marshall, "Build It and They Will Come."