DQ13

http://soe.sagepub.com/ Sociology of Education

http://soe.sagepub.com/content/79/2/153 The online version of this article can be found at:

DOI: 10.1177/003804070607900204

2006 79: 153Sociology of Education Maurice Garnier and Mark Schafer

Educational Model and Expansion of Enrollments in Sub-Saharan Africa

Published by:

http://www.sagepublications.com

On behalf of:

American Sociological Association

can be found at:Sociology of EducationAdditional services and information for

http://soe.sagepub.com/cgi/alertsEmail Alerts:

http://soe.sagepub.com/subscriptionsSubscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://soe.sagepub.com/content/79/2/153.refs.htmlCitations:

What is This?

- Apr 1, 2006Version of Record >>

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Educational Model and Expansion of Enrollments in

Sub-Saharan Africa

Maurice Garnier Indiana University

Mark Schafer Louisiana State University

This article relies on the concept of educational model to explain why African countries that were

colonized by Britain and France experienced significantly different forms of educational expansion

after independence. The authors argue that both models represented legitimate variants of world

institutional processes that drove the expansion of mass education in the sub-Saharan region. Each

model defined the nature of the relationship between the educational system and other institu-

tions differently, in particular what should happen in school, who may teach, and who is respon-

sible for the provision of education. Using panel data for 28 sub-Saharan nations between 1970

and 2000, the authors found strong model effects that persisted after they controlled for

resources, pupil-teacher ratio (an imperfect measure of quality), and percentage Muslim.

Specifically, the Francophone model contributed to lower mean enrollment levels, a slower expan-

sion of enrollments from 1970 to 1985, and (3) a more rapid expansion of enrollments from 1985

to 2000. Gender differences are also discussed. Institutional theorists have minimized the impor-

tance of resources, but the findings presented here indicate that resources and the characteristics

of the educational model matter for educational expansion.

Sociology of Education 2006, Vol. 79 (April): 153–176 153

Sub-Saharan African nations expandededucational opportunity substantially inthe decades following independence. In 1960, only one in three children in the region was enrolled in school, whereas by 2000, this fraction had risen to such a point that two out three children were enrolled. This expansion is particularly impressive in light of the region’s rapid population growth (the total number of children who were enrolled more than tripled between 1960 and 2000) and minimal economic development.

In this article, we compare the expansion of enrollment among two groups of sub-

Saharan nations: those with French versus those with British colonial histories. We argue that differences between French-based and English-based educational systems were char- acterized by fundamentally distinct models— institutional arrangements—that differentially shaped the expansion of enrollments both directly, through policies and the structures of the models, and indirectly, through interac- tions with religious traditions and gender norms. The institutional arrangements includ- ed both (1) the structural arrangements and mandates of educational institutions within the state and (2) the salient practices, proce-

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

154 Garnier and Schafer

dures, and precedents that these institutions adopted.

World institutional theory suggests that all modern states design and implement national educational systems, establish cur- ricula, train teachers, and stimulate the demand for schooling among their citizens in a more-or-less similar manner, regardless of the financial resources that are available (Benavot and Riddle 1988; Boli and Thomas 1997; Meyer, Ramirez, and Soysal 1992; Meyer et al. 1979). This body of thought fur- ther suggests that African nations, as do all developing nations, adopt similar strategies to expand enrollments primarily because (1) they seek to “signal” their desire to be con- sidered part of the modern world (Fuller 1991) and (2) intergovernmental and inter- national nongovernmental organizations (INGOs) design and promote standardized models that all developing nations are oblig- ed to replicate (Schafer 1999). Both factors contribute to isomorphism and convergence in educational systems and in the interna- tional system of nation-states, more general- ly.

This article’s objective is not to challenge or test the world institutional perspective specifically, but, rather, to explore the effects of two dominant institutional variants in sub- Saharan Africa, one Francophone and the other Anglophone. We describe the charac- teristics of each institutional arrangement that we believe has had broad implications for the development of national educational systems. We distinguish the effects of institu- tional arrangements from differences in resources because resources can be expected to influence the pace of expansion of enroll- ments (Carroll 1981; Fuller 1991; Nielsen and Hannan 1977). While critical, resources should not be so influential as to explain fully differences in the ways in which national edu- cational systems expanded in the sub- Saharan region. The basic premise of institu- tional theory is that educational expansion proceeded much more rapidly than available resources would have justified. All nations sought to expand schooling and deepen its effects on children to maintain legitimacy in the eyes of their citizens and the community of nations (Fuller 1991). The motivation to

expand an educational system may indeed stem from the increasing commitment to uni- versal values, but that expansion takes place within a specific institutional context that may shape the nature of that expansion.

African nations had two institutional vari- ants.1 Educational models with French or British roots were introduced to (imposed upon) nations that were typically composed of a multiplicity of societal groups, with sub- stantial intranational diversity. The vast litera- ture on African societal groupings suggests many potential cross-national comparisons of intranational diversity, but two societal factors seem particularly critical with regard to the expansion of mass education: religion and gender. Within the great diversity of religious traditions in sub-Saharan Africa, we focus on the role of Islam, primarily because modern schooling was introduced by Christian mis- sionaries. Consequently, Muslims tended to be wary of enrolling children in schools whose overt mission was religious conversion to Christianity. Norms regarding the proper role of girls and women also varied widely across African societies. Gender norms can be linked to religious institutions and to other social institutions, such as patrilineal and matrilineal rules of descent. Our exploration essentially attempts to tease out the effects of the models in consideration of these critical societal factors, but we also recognize and discuss the interaction between model and society.

Finally, we explore whether the effects of the models on the expansion of enrollments changed over time. More specifically, we examine zonal differences separately for the periods 1970–1985 and 1985–2000. The early stage of educational expansion, the late 1960s and 1970s, was a time of great enthu- siasm and optimism for the newly indepen- dent nations of sub-Saharan Africa. By the mid-1980s, however, that optimism had faded as states’ capacity to fund education declined considerably in the wake of eco- nomic stagnation, crushing external debt, and concomitant structural adjustment pro- grams. at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 155

EDUCATIONAL MODEL AND EDUCATIONAL EXPANSION

The colonial powers left a strong imprint on educational systems, not only because the colonial language became, and remains, the lingua franca in a particular country, but because when the time came to expand, the colonial power’s models were used as models and shaped the initial institutional arrange- ments. Thus, French-like schools are found in Francophone Africa, and English-like schools have prevailed in Anglophone Africa. Within each group of nations, differences have emerged in pedagogical practices, criteria for recruiting teachers, employment policies, and so forth. However, essential within-model similarities have remained. The Francophone model assigns greater decision-making power to the central state educational institution, the Ministry of Education. In Francophone nations, teachers were historically national- level civil servants who could be appointed to any school in the country;2 they were recruit- ed on the basis of an anonymous written examination, were similarly remunerated, and were responsible for teaching a national- ly determined curriculum at a prescribed pace. The local community was expected to contribute to the funding of some expenses, but it had little voice in how the money was spent. Schools were emanations of the central state, and its administrative leaders, including principals, had, in theory at least, little auton- omy. Principals moved from school to school as a result of administrative decisions with vir- tually no input from the community.

By contrast, local communities and indi- vidual schools in Anglophone countries enjoyed far more autonomy in recruitment, pedagogical practices, and general policies. In many cases, representatives of the commu- nity had a voice in the selection of principals, who were expected to exert significant autonomy in the administration of their schools. Historically, the differences between French and British educational systems were linked to different conceptions of the proper relationship between the central state and regional administrative units (Archer 1984). The strong, centralized French state con- trolled the pace of educational expansion

(Garnier, Hage, and Fuller 1989), while the more decentralized British state gave more power to local state actors to enable or con- strain the expansion of mass education (Archer 1984; Hage and Garnier 1990).

Newly independent African states assumed organizational structures that closely resem- bled their respective colonial rulers, and these structures influenced the pace of educational expansion. The Francophone educational model relied upon a centralized bureaucratic structure, and ministry officials ensured that national educational policy was carried out: One national curriculum was followed, teach- ers’ assignments were national, one language of instruction (French) was accepted, and lit- tle effort was made to include national lan- guages. The high degree of centralization that was characteristic of Francophone nations may have impeded the emergence of local practices and the possibilities of state- society partnerships.3

In contrast, the Anglophone model allowed more room for local practices to emerge, and these practices facilitated expan- sion.4 Self-help harambee schools in Kenya (Bradshaw 1993, Schafer 2004), distance education centers in Malawi, and Ujamaa schooling in Tanzania represented but a few examples of greater local control over educa- tion. Furthermore, the recruitment of teach- ers, the control of parents over teachers’ behavior, and the tradition of teachers’ autonomy were factors that encouraged par- ents in Anglophone nations to enroll their children in the decades following indepen- dence.

The two groups of nations differed with respect not only to their internal approach to education, but to their external relationships with their former colonizer and brethren nations. Francophone nations have main- tained supranational institutions that serve as agents of centralization and constraint, including two central banks (Collier 1991). The Francophone educational model is so strong that interstate agreements make it vir- tually automatic for students who have taken examinations in one Francophone country to enter another country’s educational system.5 These practices had financial consequences: By 1990, on average, teachers’ salaries in nine at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

156 Garnier and Schafer

Francophone nations were nearly eight times the per capita gross national product (GNP), while teachers in nine Anglophone nations earned only 31/2 times the per capita GNP, a statistically significant difference (p < .05) (Mingat and Suchaut 2000). Francophone nations also adhered to tighter achievement norms; examinations determined who could proceed to a higher grade, for example. The failure to receive a passing grade in a major subject led to grade repetition, a practice that could involve 30 percent of the children in the first grade (Association for the Development of Education in Africa, 1999). Such grade repetition, far less prevalent in English-speaking countries, increased oppor- tunity costs and made it difficult for the poor- er children to remain in school. Because of the higher professional standards, unit costs of primary and secondary education were substantially higher for a selected group of Francophone nations in 1980 (Mingat and Tan 1985).6

But African teachers in both zones did share common difficulties: They had to teach children whose native language was not usu- ally the language of instruction. Knowing the local language afforded a distinct advantage,7 but such knowledge was not as prevalent in French-speaking Africa as in Anglophone Africa owing to more centralized teachers’ assignments. Parents are more likely to com- municate easily with teachers and principals who speak the local language and hence are more likely to enroll their children. Such a sit- uation was more likely to occur in Anglophone than in Francophone Africa. In Anglophone Africa, parents mustered their resources and paid for supplies, while such an expense was often viewed as an undesirable “tax” in Francophone Africa, where it was assumed that the state should provide such resources. By contrast, the greater local autonomy that prevailed in Anglophone Africa led parents to accept the idea that they were responsible, to a significant extent, for providing some of the resources that were needed for their children’s schooling. Our assumption is that these powerful differences in the models hindered the initial expansion of enrollments in Francophone Africa because educational practices were heavily influenced

by past colonial and present international practices that require highly centralized and standardized administrative arrangements.

SAMPLE AND ENROLLMENT TRENDS: 1970–2000

To investigate whether differences in educa- tional models influenced the expansion of enrollment in sub-Saharan Africa, we con- structed panel data for 28 sub-Saharan African nations for which adequate data were available on enrollments and key indepen- dent variables of interest. For each nation, we compiled data on enrollments and key inde- pendent variables for seven five-year periods from 1970 to 2000 and used these quin- quennial data to explore how the education- al models influenced the expansion of enroll- ments over time. We begin with a descriptive discussion of trends in enrollment.

Dependent Variable

We chose to use net enrollment ratio, the per- centage of the official primary school age group that is enrolled in school, as our prima- ry measure of enrollment trends. The United Nations has more extensive data on gross enrollment ratios, that is, all enrolled children divided by age group. Gross enrollment ratios can be difficult to interpret, however, because they often exceed 100 percent in countries that are undergoing a rapid expansion in enrollment, since a large number of children outside the official age group enroll in school. Fortunately, many African nations reported both net and gross enrollments for several of the seven periods in the study. The availabili- ty of both statistics provides a simple method by which to impute missing data on net enrollment ratios. To complete our data set, we used officially published net enrollment ratios when they were available. When net enrollment data were not reported for a spe- cific country and period, we imputed net enrollment data if, and only if, two conditions were satisfied: Gross enrollment data were available for the specified period (panel), and both gross and net enrollments were avail- able for at least one of the other six panels. at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 157

These conditions enabled us to use average gross enrollment ratio/net enrollment ratio to interpolate missing “net” from available “gross” enrollment data for each nation. Approximately one third of the net enroll- ment observations (64 of 192 total observa- tions) were imputed in this way, while the remaining data were originally reported to the United Nations by each nation in the sample.

Table 1 lists the Francophone and Anglophone nations that were included in this study, along with decennial net primary school enrollment ratios from 1970 to 2000. In 1970, 45 percent of school-age children were enrolled in Anglophone Africa, com- pared to only 35 percent who were enrolled in Francophone Africa. Enrollments increased by 20 percentage points or more from 1970 to 2000 in both groups of nations, with con- siderable variation within the educational models.

The nations in Table 1 represent all nations for which we were able to construct data on net enrollment ratios (described in detail later). For a smaller sample of 24 nations, we were able to gather data on male and female enrollment ratios, separately, and to graph enrollment trends by educational model, reli- gion, and gender (see Figures 1–5).

Figure 1 demonstrates two distinct pat- terns of enrollment expansion from 1970 to 1985. In addition to having higher enroll- ments to begin the period, the Anglophone group realized a more rapid expansion of enrollments from 1970 to 1985 than did the Francophone group. After 1985, however, enrollment ratios stagnated and even declined slightly in Anglophone nations, from the peak of 72 percent enrolled in 1985 to 70 percent enrolled in 2000.8 By contrast, Francophone nations were better able to maintain their slower pace of expansion of enrollments throughout the 30-year period, regaining the ground lost in the 1970s in the 1990s.

The flattening of the Anglophone curve after 1985 is a particularly striking feature of Figure 1 and deserves more discussion. In the 1980s, the combination of economic stagna- tion and structural adjustment led to severely reduced educational spending. According to

a World Bank report, the total educational expenditure in Africa was cut by 30 percent in the first half of the 1980s (Fredrikson 1990), yet total enrollments in education increased throughout the continent. Figure 1 suggests that this reduced educational expenditure affected Anglophone and Francophone nations’ enrollment ratios differently. Diffusion theory would not predict this leveling off to occur at 70 percent of all children who were enrolled. We suggest that the different edu- cational models influenced this leveling off. The demand-driven Anglophone model was both better able to expand during the early period and more vulnerable to economic shock in the 1980s and 1990s, while the sup- ply-driven Francophone model was better able to withstand the early pressure for rapid expansion and better able to cope with eco- nomic hardships in the later period. Yet, there are other potential explanations for the observed differences in enrollment patterns between these two groups of nations.

Islam and Gender

The different patterns of expansion of enroll- ments between Anglophone and Francophone nations could be spurious, reflecting deep soci- etal chasms across linguistic zones. Religion is one key consideration because modern school- ing originated in the African continent with Christian missionaries. In many instances, Muslim parents who associated modern schools with Christianity were reluctant to enroll their children in modern schools.

Figure 2 compares the expansion of enroll- ment across nations with a Muslim majority versus those with a non-Muslim majority. Throughout the 30-year period, non-Muslim majority nations exhibited higher enrollment ratios than did Muslim-majority nations. However, the expansion of enrollments was greater in Muslim nations, increasing about 30 percentage points, from 22 percent who were enrolled in 1970 to 52 percent who were enrolled by 2000. In comparison, enroll- ment ratios increased only 18 percentage points in non-Muslim nations in this period, from 54 percent to 72 percent who were enrolled.9 Moreover, Muslim nations expand- ed enrollments substantially during the at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

158 Garnier and Schafer

1990s, while enrollments stagnated in non- Muslim nations. Clearly, religious differences could provide an alternative explanation for the differences in the models that are shown in Figure 1. Figure 3 combines the data on model and religion to show the separate

trends in enrollments for four groups of nations: Anglophone non-Muslim, Anglophone Muslim, Francophone non-Muslim, and Francophone Muslim. This graph is less smooth because of the small number of nations that are represented, but it suggests that both educa-

Table 1. Nations, Model Affiliation, and Decennial Net Enrollment Ratios, 1970–2000

Net Enrollment Ratio

Nation 1970 1980 1990 2000

Francophone 1. Benin 28 52 45 70 2. Burkina Faso 10 15 27 36 3. Burundi 23 20 51 54 4. Cameroon 64 71 76 66 5. Central African Republic 50 56 53 55 6. Chad 28 30 43 58 7. Comoros 24 60 53 56 8. Congo, Republic 67 95 90 65 9. Congo, Democratic Republic 68 70 54 33 10. Guinea 24 27 26 47 11. Mali 17 20 21 43 12. Mauritania 10 27 36 64 13. Niger 12 21 24 30 14. Rwanda 48 59 66 NA 15. Senegal 34 37 48 63 16. Togo 55 75 75 91

Francophone Average All (Figure 1 Only) 35 46 49 55

(33) (44) (48) (55)

Anglophone 1. Botswana 46 76 93 84 2. Gambia 21 50 52 69 3. Ghana 46 57 54 58 4. Kenya 47 91 77 69 5. Lesotho 66 67 73 78 6. Malawi 32 43 50 73 7. Sierra Leone 31 47 45 NA 8. Sudan 32 42 44 49 9. Swaziland 62 80 88 93 10. Tanzania 25 68 51 47 11. Zambia 68 77 86 66 12. Zimbabwe 62 72 98 80

Anglophone Average All Figure 1 Only 45 64 64 70

(48) (68) (72) (69)

Overall Average 39 53 57 61

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 159

tional model and majority religion influenced enrollment levels and expansion in sub- Saharan Africa from 1970 to 2000.

More specifically, Figure 3 shows that non- Muslim nations had higher enrollment levels than did Muslim nations, but irrespective of the religious majority, Anglophone nations had higher levels of enrollment than did Francophone nations. At the same time, 1980 enrollments were equivalent to 2000 enroll- ments for all Anglophone nations (73 percent enrolled for Anglophone non-Muslim and 63 percent enrolled for Anglophone Muslim, respectively. By contrast, Muslim nations saw a continued expansion of enrollments after 1980, with Francophone Muslim nations

expanding enrollments from 30 percent in 1980 to 50 percent by 2000).

Gender may also influence the expansion of primary-school enrollments. Figure 4 shows male and female enrollments for the Anglophone and Francophone nations from 1970 to 2000. The 1970 gender gap in Anglophone nations was small; male enroll- ments (51 percent) were only 5 percentage points higher than were female enrollments (46 percent). This gap closed, and, by 2000, male and female enrollments were virtually identical among Anglophone nations. In con- trast, the gender gap was large in Francophone Africa. In 1970, male enroll- ments in Francophone Africa exceeded

Figure 1. Enrollment, by Model

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

160 Garnier and Schafer

female enrollments by 18 percentage points (41 percent versus 23 percent enrolled). The gender gap in Francophone nations closed to about a 12 percentage-point difference by 2000, but even then, male Francophone enrollments, at 64 percent, lagged just behind Anglophone enrollments (both male and female), at 69 percent, while female Francophone enrollments remained much lower than all other enrollments, at 51 per- cent. This graph suggests that gender shapes the relationship between educational model and enrollment expansion.

Figure 5 includes all three factors—model, Islam, and gender—to provide the most con- ceptually disentangled picture of the expansion of enrollments in sub-Saharan Africa. In

Anglophone, non-Muslim nations, a higher percentage of both boys and girls were enrolled from 1970 to 2000, even though enrollments within this group of nations declined slightly after 1985. In Anglophone Muslim nations (only three nations), both male and female enrollments expanded rapidly from 1970 to 1980 and then declined (significantly in the case of male enrollments). Male and female enrollment ratios varied widely in Muslim and non-Muslim Franco-phone nations, providing a clear indication that gender differences may not be strictly tied to religion and should be considered independently. Moreover, these four groups within the Francophone model (male non-Muslim, male Muslim, female non- Muslim, and female Muslim) all showed steady

Figure 2. Enrollment, by Religion

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 161

increases throughout the 30-year period, and all with significant increases in enrollments in the late 1990s. Among all the non-Muslim nations in 2000, only 55 percent of the girls in Francophone nations were enrolled, compared to 71 percent of the boys in Francophone nations and 73 percent of all Anglophone non- Muslims, indicating a strong gender-model interaction.

The social norms leading to larger gender disparities in the Francophone region may partly reflect religious differences across zones, but there are also other structural (e.g., percentage rural and migration pat- terns) and cultural (lineage systems, marital norms, and rites-of-passage customs) factors that could potentially create gender differ-

ences in enrollments. Moreover, our discus- sion of differences in the models alludes to the strong possibility of interaction effects between the type of educational model and its ability to penetrate patriarchal and Muslim societies. To assess the relationship between educational model and enrollment expansion more systematically, we constructed an addi- tional quinquennial panel data set with a range of variables that could potentially have influenced the expansion of enrollments for the same period, 1970–2000.

Independent Variables

We used the dummy variable, Francophone, to test the effect of educational model. While

Figure 3. Enrollment, by Model and Religion

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

162 Garnier and Schafer

we acknowledge within-model variation in institutional arrangements, our focus is on the broad influence of the specific national edu- cational institution. Therefore, the dummy variable served as a global proxy for model effects. Data that would enable us to specify whether individual nations within groups conform to the different dimensions of the two models are not available.

We measured the availability of resources as the level of economic development, per cap- ital gross domestic product (GDP). We further defined economic growth as the change in the per capita GDP over each five-year period. In regression analyses, this variable is logged to correct for its skewed distribution.

We included an indicator of worldwide organizational and cultural resources, INGO memberships, a measure of world institution- alism. Institutional theorists have argued that INGOs contribute to global institutional stan- dardization and convergence (Boli & Thomas, 1997). Although it is also possible that the pluralism of local nongovernmental organiza- tions (NGOs) may further shape educational expansion and development, we did not find valid measures of local NGOs to include in our analysis. Because the INGO variable is highly skewed in the region, we took the nat- ural logarithm of the number of INGO mem- berships (Schafer 1999).10 The literature emphasizes the importance of INGO expan-

Figure 4. Enrollment, by Model, by Gender

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 163

sion in the wake of the declining state (Bratton 1989); therefore, we measured the expansion of INGO memberships (again using first differences).

We further distinguished between the availability of resources and investment by including a measure of investment and edu- cational expenditure, defined as the total pub- lic educational expenditure as a percentage of the GNP. Data on national expenditures for the region are limited. We obtained at least one panel of such data for each nation, but data for all panels were available for only 15 nations. This variable is “sluggish,” since edu- cational expenditure levels changed slowly over the 30-year period. The level of invest-

ment is more relevant than are short-term changes in state educational expenditures.

Finally, we included pupil-teacher ratio in our models. Pupil-teacher ratio is an imper- fect measure of class size, and higher pupil- teacher ratios are not necessarily linked to lower-quality schools. But large increases in pupil-teacher ratio may indicate that the expansion of enrollments stretched the state’s capacity to provide teachers. Fuller (1991), in particular, argued that deteriorating school quality that is due to the overcrowding of schools as a result of the rapid expansion of mass education is detrimental to the subse- quent expansion of enrollments because par- ents and communities see diminishing social

Figure 5. Enrollment, by Model and Religion, by Gender

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

164 Garnier and Schafer

and economic returns to schooling (see also Heyneman and Loxley 1983). Some variation in the number of students per teacher can be expected; therefore, we explored the effect of the increase in the pupil-teacher ratio over each five-year period.

Quality of the Data

All the independent variables just described were drawn from three international databas- es—the United Nations, World Bank, and Union of International Associations. All these variables have been used in cross-national studies, but quinquennial panel coverage for the nations of sub-Saharan Africa is incom- plete. To deal with missing values for some nations for some periods, we selected vari- ables for which there was adequate coverage for most nations in the sample for most of the five-year periods (described in Appendix Table A). An alternative strategy would have been to include more variables and to deal with missing data by imputing values. We opted not to impute missing values of inde- pendent variables. We imputed values of the dependent variable, net enrollment ratio, because gross enrollments were available for each time frame and for both net and gross enrollments for some periods. We did not have any similarly practical way to impute the values of the missing independent variables and, therefore, opted to include only vari- ables with sufficient coverage.

Bivariate Findings

Table 2 presents the bivariate relationships across the two sub-Saharan educational mod- els. The top four rows provide a test of whether the relationships depicted in Figure 1 are statistically significant. They explore dif- ferences between the models in enrollment levels, enrollment growth averaged over the 30-year span from 1970 to 2000, the early expansion of enrollments from 1970 to 1985, and later expansion of enrollments from 1985 to 2000, respectively. The intercept differ- ences that are depicted in Figure 1 are signif- icant, with average net enrollment ratios averaging 63 percent for Anglophone and 47 percent for Francophone nations, respective-

ly. The expansion of enrollments was roughly similar across groups for the entire 30-year period, with 5-year rates of enrollment growth averaging 3.5 in Francophone and 3.9 in Anglophone nations. The rapid early expansion in enrollments among the Anglophone group was much greater than among the Francophone group (with 5-year increases averaging 7.8 compared to 4.6, p = .06). After 1985, average enrollments fell in Anglophone Africa but continued to expand in Francophone Africa from 1985 to 2000, but the differences are not significant (p = .09). However, these early–late differences (also depicted in Figure 1) are not statistically significant at the p < .05 level. Nonetheless, the bivariate findings suggest the need for multivariate analyses to determine the extent of the effects of the models.

The bottom of Table 2 depicts differences between the models across a range of explanatory variables that are theorized to influence enrollments. The Anglophone nations had higher mean levels of economic development, INGO expansion, and educa- tional expenditure, but lower pupil-teacher ratios than did the Francophone nations. Moreover, the Anglophone nations exhibited greater within-charter variation in the level of economic development, INGO expansion, and educational expenditure, while there was more variation in pupil-teacher ratios among the members of the Francophone group.

In bivariate descriptive analyses, we exam- ined differences between the models in net enrollment levels and expansion, which we operationalized in terms of first differences (Net Enrollment at Time t – Net Enrollment at Time Time 5).11 In the next section, we fur- ther explore whether differences in the mod- els remain when resources, religion, and gen- der are taken into consideration.

First-Difference Model of Enrollment Expansion

To explore the effects of educational models in sub-Saharan Africa, we use a first-difference model of enrollment expansion. The first-dif- ference model provides a way of capitalizing on the strength of panel data, that is, the capacity to control for biases that are due to at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 165

omitted variables or unobserved, individual country effects (Beck 2001; Firebaugh and Beck 1994; Halaby 2004).

Although we focus on the effects of the educational model and a limited number of independent variables, we should not ignore the possibility that unmeasured national attrib- utes could have substantial effects on the expansion of enrollments. National attributes are enduring (do not change over time) and unique (vary across nations) characteristics of

individual nations. These are left out of quan- titative studies because they are difficult to measure meaningfully. Examples may include the cultural reception of the modern school, the history of mission schools, the viability of subsistence versus more modern lifestyles, the presence of alternatives to modern schools, state legitimacy and its relationship to school- ing, and the prevalence of war or unrest, among many others. Since national attributes are unmeasured, we cannot empirically deter-

Table 2. Independent Variable Means (Quinquennial Averages, 1970–2000), by Educational

Modela

Variable Francophone Anglophone Modelb Model t-statisticc

Net Enrollment 46.51 62.50 2.26* 1970–2000 (21.49) (19.24)

Net Enrollment Change 3.61 3.88 .20 1970–2000 (10.08) (8.65)

Net Enrollment Change 4.39 6.33 1.12 1970–1985 (7.59) (10.96)

Net Enrollment Change 3.12 -0.26 -1.67 1985–2000 (9.99) (7.24)

Level of Economic Development 409.62 609.20 2.73** (219.51) (698.93)

INGO Change 57.67 83.64 3.75** (37.67) (51.16)

Educational Expenditure 3.40 4.77 5.46** (1.38) (1.77)

Pupil-Teacher Ratio 49.86 38.53 -7.14** (11.35) (9.84)

Percentage Female 39.26 46.28 8.50** (5.67) (5.65)

*p < .05, **p < .01 (two-tailed tests).

a Figures in parentheses are standard deviations. b Figures standardized to 1995 prices. c H0: Anglophone mean - Francophone mean = 0.

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

166 Garnier and Schafer

mine whether they influence enrollment; we justify the first-difference model because we strongly suspect that national attributes have a substantial influence on the growth (and retardation) of enrollments in particular nations during a particular time frame. Therefore, following Firebaugh and Beck (1994), we used a first-difference model of enrollment growth as follows:

Enrollment Expansion = Net Enrollmentt – Net Enrollmentt-5 =

ß0 + ß1 (INGO Expansion = Log International

INGO Memberships*t – Log International INGO Memberships*t-5)

+ ß2 (Economic Growth = Log Level of Economic Development*t – Log Level of Economic Development*t-5)

+ ß3 (Pupil-Teacher Ratio Increase = Pupil-Teacher Ratio*t – Pupil-Teacher Ratio*t-5)

+ ß4 (Net Enrollment Level*t-5) + ß5 (Level of Economic Development* t-5)

+ ß6 (Francophone) + ß6 (Percentage Muslim) + ε

Both the dependent and independent vari- ables are first differenced, except in the cases of two time-invariant, dummy variables: the Francophone dummy variable and percent- age Muslim. We also include two “level” vari- ables: (1) net enrollment level, which controls for a ceiling effect because ratios cannot exceed 100 percent enrolled, and (2) level of economic development, which controls for a constant-effects assumption that the initial level of economic development (or availabili- ty of resources) causes the expansion of enrollments independent of shorter-term fluctuations in resource-generating capacity (economic growth). The panel model is esti- mated for three time frames: 1970–2000, 1970–1985, and 1985–2000 to explore early–late differences in charter effects on enrollment expansion.

Differencing not only improves estimation by effectively dealing with potential bias that is due to unobserved country effects, it also reduces estimation problems that are due to temporal autocorrelation and nonstationarity. Temporal autocorrelation (also serial autocor- relation or serial dependence) is expected in panel data; it is common for values for a par-

ticular country from one year to be associat- ed with values for the same country from other years. Using quinquennial data reduces the problem because five-year correlations are generally weaker than are one-year corre- lations. Differencing further reduces the prob- lem because the associations between changes in values are typically not as strong as are associations between levels of values from one time unit to the next. Nonetheless, in panel estimation, it is still necessary to test for first-order serial autocorrelation, which we did using a Wooldridge test (2002), imple- mented using the STATA command, XTSERI- AL. The tests (reported in Table 2) indicated first-order autocorrelation in most models. Therefore, we applied the Prais-Winston esti- mator with AR (1) disturbance terms to con- trol for first-order serial correlation in the first- difference model.

Differencing also helps address the issue of stationarity. Panel data analysis requires sta- tionary data—meaning that means, vari- ances, and autocovariances (at various lags) of data do not vary over time—otherwise there is a risk of a spurious regression problem (Granger and Newbold 1974). We tested for stationarity of all data using a Dickey-Fuller test: an ordinary least-squares (OLS) regres- sion of the first difference of a series on its lagged level. This test detected a unit root for one variable: level of economic development. Moreover, the logarithmic transformation did not eliminate the unit root, or problem of nonstationarity. Further tests confirmed that none of the differenced variables contained unit roots; all satisfied the condition of sta- tionarity that is required for panel analysis.

Panel analysis also requires that data be homoskedastic within each period or panel. We used panel-corrected standard errors (PCSE), as recommended by Beck and Katz (1995), adjusting standard errors in light of the panel structure of the data, assuming a common (as opposed to a unit-specific) autoregressive process. Since we detected autocorrelation in the model using the Wooldridge test, we used PCSEs with a Prais- Winston transformation that models first- order autocorrelation. We checked our results against two other procedures for adjusting standard errors in panel analysis: the Huber- at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 167

White robust-cluster error estimates (see Moller et al. 2003) and the Newey-West stan- dard error estimates (Halaby 2004). All the findings presented next were consistent across all three methods of correcting stan- dard errors.

In sum, first differencing helped to ensure that we would obtain unbiased estimates to determine whether associations between model and enrollment growth are genuine or spurious. But there is a drawback to using the first-difference model. First, differencing may contribute to an endogeneity problem because changes in some variables are often associated with changes in other variables. For example, our model suggests that eco- nomic growth affects enrollment growth, but, clearly, the reverse could also true (that enrollment expansion leads to economic growth). The same could be true for increas- es in the pupil-teacher ratio—they could be either a cause or an effect of enrollment growth. Although methods exist for address- ing endogeneity within the dynamic panel framework (see Halaby 2004 for an overview), we could not use such methods in this study because of the limited number of nations and time frames. However, our pri- mary interest is the model effects and whether such effects have changed over time. Therefore, we focus on establishing the rela- tionships in this section and return to the issue of causality in the Discussion.

FINDINGS

Table 3 presents Prais-Winston estimates with panel-corrected standard errors for the regression of the first-difference model. Model 1 presents the findings for the entire 30-year period, and Models 2 and 3 present the findings for the early and late periods (1970–1985 and 1985–2000), respectively. Model 1 reveals a strong relationship between economic growth and the expan- sion of enrollments. In addition, increases in pupil-teacher ratios were associated with enrollment expansion during the 30-year time frame. Finally, there is a ceiling effect because countries with higher enrollment ratios experienced less expansion, net of

other factors. The Wooldridge test showed positive autocorrelation in the model, and the Prais-Winston transformation reduced the extent of autocorrelation (e.g., increased the Durbin Watson statistic from 1.27 to 1.47). Model 1 does not indicate a model effect, but this should not be surprising because the Anglophone-Francophone gap in 1970 was about the same as in 2000 (see Figure 1).

The different form of expansion between the two models lies in the paths they each took to arrive at their millennial enrollment ratios, which are depicted in Models 2 and 3. The important finding depicted in Model 2 is the strong negative Francophone effect, indi- cating that the more centralized model restrained the expansion of enrollments in the early period, 1970–1985. Also during this time span, both the level of economic devel- opment and enrollment change were associ- ated with a greater expansion in enrollments. The Wooldridge test for serial autocorrelation is not significant in this model, yet the Durbin Watson statistic shows positive serial depen- dence, which is not improved by the Prais- Winston transformation. To explore the robustness of the model effect further, we reran the model with several different model designs.12

After 1985, the effect of educational model reversed itself (see Model 3), with Francophone nations achieving higher enroll- ment expansion than Anglophone nations from 1985 to 2000, net of all other factors. Increases in pupil-teacher ratios, not signifi- cant in the early period, were positively asso- ciated with more enrollment growth in the later period. Finally, the strong association between economic growth and enrollment growth remained. We tested for the signifi- cance of period differences (e.g., Is Model 2 significantly different from Model 3?) by using a dummy variable EARLY = 1 for all vari- ables years 1970–1985 and rerunning Model 1 with interaction terms with each right- hand-side variable. These tests confirmed the significance of period differences in educa- tional model.

The findings reported in Table 3 have sev- eral theoretical implications. First, they pro- vide empirical support for our contention that the educational model influenced the pattern at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

168 Garnier and Schafer

of enrollment expansion in sub-Saharan Africa after independence. The trend depict- ed in Figure 1 persisted after we controlled for resources, pupil-teacher ratio, percentage female, and percentage Muslim.13 The first-

difference model design reduces the likeli- hood that the Francophone dummy variable encapsulates other, unobserved, differences between the two groups of nations.

Second, the findings demonstrate the crit-

Table 3. Unstandardized Prais-Winston Regression Estimates with AR(1) Disturbances and Panel-Corrected Standard Errors

Dependent Variable: Net Enrollment Change

(1) (2) (3) Independent Variables 1970–2000 1970–1985 1985–2000

Economic Growth (logged) 17.69** 23.52*** 22.90** (5.65) (7.57) (10.21)

International NGO Expansion (logged) -2.51 -5.89 -11.59 (2.84) (4.05) (9.48)

Increase in Pupil-Teacher Ratio .18* .13 .22* (.09) (.10) (.11)

Percentage Female 1.09** .87 1.11 (.39) (.48) (.64)

Percentage Muslim -.05 -.06 -.03 (.03) (.04) (.03)

Francophone Dummy -1.74 -6.71*** 5.43* (2.27) (2.12) (2.75)

Floor Effect: Level of Economic Developmentt-5 3.09 7.42* .44 (1.79) (3.27) (1.50)

Ceiling Effect: Net Enrollment Ratio t-5 -.20* -.27 -.09

(.09) (.19) (.09)

R-squared .29 .34 .42

Rho .09 .05 -.02

Wooldridge Test 10.91** 2.75 15.14***

Durbin Watson Before 1.39 1.38 1.18

Durbin Watson After 1.43 1.21 1.28

Countries 27 25 27

Observations 143 70 73

*p < .05, **p < .01, ***p < .001 (two-tailed tests).

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 169

ical association between economic growth and the expansion of enrollments. In short, resources matter. The nations in the world’s poorest region may have shared the desire to expand educational opportunity, as world institutionalism suggests, but their capacity to achieve their educational expansion goals was strongly linked to their capacity to achieve economic growth. In testing for asso- ciations between economic growth and enrollment expansion over particular five-year time frames, our design does not indicate whether economic growth caused the expan- sion of enrollments. We return to this ques- tion in the Discussion.

Third, despite the depiction of large differ- ences in enrollment between Muslim and non-Muslim nations in Figure 2, the findings in Table 3 suggest that the percentage of Muslims in a nation did not significantly influ- ence the expansion of enrollments. Figure 2 was partially misleading in that we grouped nations with different proportions of Muslims into one of two groups (majority Muslim or majority non-Muslim). However, it is also pos- sible that the effect of a large Muslim popula- tion extends beyond the particular members of the faith to influence enrollments through- out the nation. If this were the case, it would be more appropriate to use a dummy variable than the percentages that we used in Table 3. We further explore the influence of religion in the next section. In addition, we explore gen- der differences, again differentiating between early and late periods of enrollment expan- sion.

Religion and Gender

To explore the effects of religion and gender, we used a reduced first-difference model in which enrollment expansion is regressed only on economic growth. We divided the Francophone dummy into two groups, one with five, mostly non-Muslim, nations and the other with seven, predominantly Muslim Francophone, nations. We kept the floor and ceiling effects in the model. Table 4 presents the results separately for the early and late peri- ods, consistent with the previous analysis (but excluding the model for the full 30-year peri- od).

The first model in Table 4 found an early- period negative effect of educational model for both the Muslim and non-Muslim nations. The Francophone educational model suppressed enrollment growth overall, as well as for boys and girls separately. The model effect is stronger (e.g., the size of the coefficient) for Muslim than for non-Muslim Francophone nations. Among non-Muslim nations, the neg- ative charter effect was much stronger for female enrollments than for male enrollments (-4.64 compared to -2.35).

Models 4, 5, and 6 test whether the positive, late-period Francophone effect that was shown in Table 3 held for both Muslim and non- Muslim nations and across gender lines. The signs of both the educational model and the religious majority dummy variables are reversed in all three equations, and four of the six model coefficients are statistically significant. The pos- itive charter effect in the later period is most apparent among non-Muslim nations, where Francophone systems experienced greater expansion than their Anglophone counterparts in overall, male, and female enrollments. The Francophone Muslim nations, however, real- ized a significantly greater growth in female enrollments than did the Anglophone nations, a finding that supports the implications of Figures 4 and 5 that both female and Muslim enrollment ratios increased substantially in the region over the course of the 1990s. Again, tests for period differences confirmed that the effects of educational model were significantly different for the early and late periods.

DISCUSSION

The findings just presented have considerable implications for theories of educational expan- sion. First, our contention that substantial dif- ferences in the educational models influenced the expansion of enrollments in the region is supported by this analysis. Centrally planned Francophone nations in sub-Saharan Africa failed to achieve the same rapid increases in enrollments as did their Anglophone counter- parts in the initial period following indepen- dence, but they maintained steady expansion during the period of severe economic difficulty and structural adjustment in the 1980s and at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

170 Garnier and Schafer

1990s. We offer a conceptual account of these differences in the models that focus on central- ization, international relations, and profession- alism. It may be useful to specify further the ele- ments of each model and variations within zones, so that the contribution of specific orga- nizational practices and institutional arrange- ments to educational expansion can be further understood. It is clear that organizational processes affect expansion, even if the overall

shape of that expansion reflects universal norms.

Second, resources matter. We present strong evidence that nations with higher rates of eco- nomic growth achieved a greater expansion of enrollments over the entire 30-year period. Furthermore, the level of economic develop- ment kick-started the enrollment-expansion process after independence, but became less important over time. We also note that the

Table 4. Unstandardized Prais-Winston Regression Estimates with AR(1) Disturbances and Panel- Corrected Errors

Dependent Variable: Net Change in Enrollment

Early: 1970–1985 Late: 1985–2000

(1) (2) (3) (4) (5) (6)

Independent Variables All Male Female All Male Female

Economic Growth (logged) 13.63*** 7.41 15.47*** 24.31*** 25.02*** 26.07*** (4.16) (6.20) (3.86) (6.35) (4.93) (6.73)

Francophone NonMuslim (5 countries) -3.94* -2.35* -4.64* 9.85*** 8.60*** 7.61***

(2.18) (1.32) (2.24) (1.31) (1.58) (2.71)

Francophone Muslim (7 countries) -9.21** -9.85* -9.33** 5.35 5.00 4.80*

(3.93) (4.70) (3.88) (3.56) (4.75) (3.02)

Floor Effect: Level of 4.47 4.72 4.65 1.51 1.42 .99 Economic Developmentt-5 (3.13) (3.77) (2.88) (1.69) (2.09) (1.59)

Ceiling Effect: Net Enrollment Ratio t-5 -.22 -.29 -.20 -.10 -.10 -.10 (.15) (.19) (.12) (.11) (.18) (.09)

R-squared .27 .23 .28 .41 .36 .44

Rho -.03 .24 -.04 -.18 -.18 -.13

Wooldridge Test 9.43* 38.66*** 8.59** 7.19* 3.18 5.04*

Durbin Watson Before 1.30 1.36 1.19 1.76 1.91 1.67

Durbin Watson After 1.21 1.34 1.19 1.46 1.49 1.53

Countries 22 20 20 22 22 22

Observations 63 60 60 63 63 63

*p < .05, **p < .01, ***p < .001 (two-tailed tests).

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 171

association between increases in pupil-teacher ratios and enrollment expansion can be inter- preted as indicating that the state’s capacity to produce teachers failed to keep pace with the popular demand for more and more schooling in the late 1980s and 1990s. Population growth and the demand for higher quality will, accord- ing to some, continue to place pressure on states to devote more scarce resources to schooling. These results contradict Fuller’s (1991) suggestion of a trade-off between the deterioration in school quality and enrollment expansion: Indeed, the national educational systems that allowed for substantial increases in pupil-teacher ratios also realized higher rates of expansion, net of other factors, in the same period. The extreme poverty and scarcity of resources in the regions means that sub- Saharan nations had to make difficult choices between expanding low-quality mass school- ing for more citizens versus maintaining higher- quality schooling for fewer citizens. It is proba- ble that the educational model played a signif- icant role in how these kinds of decisions were made.

Having said this, we note that our findings show associations between variables measured over the same five-year time frames, not causal associations that are associated with classical panel designs in which changes in enrollment are regressed on independent variables mea- sured at the earlier time frame. Although we presented strong methodological reasons for using the first-difference model, we also ran classic regression models to assess the causal relationship among resources, model, and enrollment expansion (see Table 5). This model confirms a strong, positive, causal and persis- tent relationship between the level of econom- ic development and the expansion of enroll- ments in sub-Saharan Africa. It also depicts neg- ative effects of educational model and Islam in the early period. However, it is prone to bias because it does not control for unmeasured national attributes. Third, the model effect was consistent for male and female enrollments and generally consis- tent across Muslim and non-Muslim popula- tions. It seems likely that these and other soci-

etal factors could indirectly affect the expansion of enrollments. Perhaps the Anglophone model was initially more adept at stimulating local demand for schooling, but later lost its legiti- macy in the wake of economic failure, structur- al adjustment, deteriorating school quality, and increased educational inequality within nations. By contrast, the Francophone model may have been more effective in resisting popular (and global) pressure to expand enrollments rapidly, opting in favor of greater quality14 and equali- ty of educational opportunity, so the Francophone nations were better able to sus- tain enrollment expansion over the longer term. When data become available, it should become possible to ascertain which specific organizational practices are associated with each model and how these organizational prac- tices affect the expansion of enrollments across the region.

Our findings are not fundamentally inconsis- tent with institutional theory’s arguments and the considerable evidence gathered to support it. Our findings do suggest that timing is impor- tant. It is the case that enrollments increased dramatically in sub-Saharan Africa, despite the general lack of resources. Moreover, the educa- tional models were no in way native to that continent. It happens that different models were brought to Africa and that these models affected the rate of expansion. It is worth not- ing that the rate of female attendance in Muslim areas is catching up, suggesting that the universal norm eventually spreads, despite local cultural practices. When data become available, it should become possible to deter- mine whether resources and model affect the rate of expansion. Enrollments should reach their maximum levels in areas that are relative- ly resource rich and that rely on local control over educational decisions. By contrast, enroll- ments should remain at their low level in areas that are resource poor and rely on central deci- sion making. Eventually, all children will attend school. However, which children will benefit first and which will benefit last are affected by resources and the prevalent educational model.

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

172 Garnier and Schafer

Table 5. Unstandardized Prais-Winston Regression Estimates with AR(1) Disturbances and Panel-Corrected Standard Errors

Dependent Variable: Net Enrollment Change

(1) (2) (3) Independent Variables 1970–2000 1970–1985 1985–2000

Net Enrollment Ratio t-5 -.30*** -.30 -.27*

.11 .19 .13

Log. Level of Economic Development t-5 5.66** 7.92** 4.64*

1.36 2.20 1.96

Log. International NGO Memberships t-5 .39 1.23 1.51

.75 1.18 1.06

Pupil-Teacher Ratio t-5 .07 .06 .01

.09 .13 .11

Percentage Female t-5 -.01 -.11 .00

.13 .13 .21

Percentage Muslim -.06* -.11** -.01 .03 .04 (.05)

Francophone Dummy -3.89** -7.43** 1.11 1.50 2.07 (4.39)

R-squared .17 .22 .21

Rho .06 .01 .07

Wooldridge Test 15.37** 4.90* 8.24**

Durbin Watson Before 1.34 1.22 1.17

Durbin Watson After 1.43 1.18 1.30

Countries 27 25 25

Observations 143 73 76

*p < .05, **p < .01, ***p < .001 (two-tailed tests).

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 173

NOTES

1. Actually, there are two more—Spanish and Portuguese—but France and Britain colo- nized many more countries.

2. This tradition is undergoing rapid change. In Senegal, for example, “volunteer” teachers are not civil servants, nor are “com- munity” teachers in Benin. These teachers now constitute about 50 percent of all teach- ers.

3. A number of development projects in Benin, Guinea, Mali, and Senegal, for exam- ple, seek to decentralize the administrative structure of the educational system and to foster the creation of state-society partner- ships.

4. Decentralization is now taking place in a number of French-speaking countries. For example, in Benin, some schools are hiring teachers who are paid by parents. These teachers have fewer educational credentials than do state-appointed teachers, are paid significantly less, and are under the control of parents who work in partnership with local educational authorities. It has been reported (Garnier 2003) that under these circum- stances, parents see to it that children attend school.

5. Another example is the CAMES. This organization groups French-speaking universi- ties, administers a highly selective examination (aggregation) in a number of disciplines, and has organized committees that are responsible

Appendix A. Definition, Scope, and Source of Variables

Variable Definition Scope Source

Dependent Variables Net Enrollment Percentage of 192 observations UNESCO Institute

primary school- (62 imputed values) of Statistics age group enrolled 164 lagged observations

28 nations

Net Enrollment, Male (and Female) Percentage of primary 154 observations UNESCO Institute school-age group 132 lagged observations of Statistics enrolled 22 nations

Independent Variables Level of Economic Gross domestic product 181 observations World Bank: World Development per capita (fixed 1995 154 observations Development

prices) 28 nations Indicators 3 with missing data (Comoros, Guinea, Tanzania)

International Memberships in 190 observations Union of Nongovernmental international 162 lagged observations International Organizations nongovernmental 28 nations Associations

organizations 1 with missing data (Comoros)

Pupil-Teacher Ratio Primary pupil-teacher 186 observations UNESCO Institute ratio 159 lagged observations of Statistics

28 nations 4 with missing data (C.A.R, Chad, Zaire, Zimbabwe)

Educational Expenditure Total public 160 observations UNESCO Institute educational 136 lagged observations of Statistics expenditure as a 28 nations percent of gross 14 with missing data national product

Percentage Muslim Percentage Muslim 28 nations CIA World Factbook

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

174 Garnier and Schafer

for the promotion of faculty members within their member countries’ universities. French academics also sit on these boards.

6. By specifying how much teachers should be paid as national civil servants, the model may have made rapid expansion diffi- cult because high salaries absorbed a large proportion of the resources that were allocat- ed to education. Whether higher salaries led to more effective schooling is unclear.

7. Even in French-speaking Africa, teachers often report resorting to the local language to help children learn. Of course, that kind of help is possible only when they know the local language.

8. Total enrollments continued to increase in all sub-Saharan nations from 1985 to 2000. Declining enrollment ratios mean that the growth of the population of the official school-age group outpaced the expansion of enrollments.

9. The graph depicts the mean enrollment ratios of majority Muslim and non-Muslim nations, respectively. The percentage of the population adhering to the Islamic faith varies within both the majority and minority groups.

10. In analyses not shown, we also explored the potential influence of official development assistance. We included the aid measure to distinguish between institutional effects of INGO memberships and access to external financial assistance and found no sig- nificant effects of aid in any of the models pre- sented here.

11. Since first differences are raw change scores and, therefore, more susceptible to the effects of outliers, we checked our results using two alternative ways of measuring enrollment expansion: (1) change score: (net enrollmentst – net enrollmentst-5)/net enroll- mentst-5 and (2) logged values: natural loga- rithm of net enrollmentt – natural logarithm of net enrollment t-5. All resulted in similar findings, so we report only the findings for raw change.

12. We reran the same model with OLS estimates and PCSEs, with Prais-Winston esti- mates and Huber-White robust estimates (with clustering), and with Newey-West stan- dard errors. All models showed essentially the same substantive results, particularly the neg-

ative effect of the Francophone dummy vari- able and the absence of a Muslim effect.

13. Furthermore, the findings regarding resource and charter effects are robust to alternative model specifications in which the following variables were included: percentage urban, educational expenditure, aid per capi- ta, gender parity, and fifth-grade students expressed as a percentage of all students in the first five grades.

14. It will be possible in the near future to test this possibility because achievement tests have been administered to children in both linguistic zones.

REFERENCES

Archer, Margaret Scotford. 1984. Social Origins of Educational Systems (abridged ed.). Beverly Hills, CA: Sage.

Association for the Development of Education in Africa. 1999. Statistical Profile of Education in Sub-Saharan Africa ’99. CD-ROM. Paris: Author.

Beck, Nathaniel. 2001. “Time-Series Cross-Section Data: What Have We Learned in the Past Few Years.” Annual Review of Political Science 4:271–293.

Beck, Nathaniel, and Jonathon Katz. 1995. “What to Do (And Not to Do) with Time-Series Cross- Section Data.” American Political Science Review 89:634–47.

Benavot, Aaron, and Phyllis Riddle. 1988. “National Estimates of the Expansion of Mass Education, 1870–1940.” Sociology of Education 61: 191–210.

Boli, John, and George M. Thomas. 1997. “World Culture in the World Polity.” American Sociological Review. 62:171–90.

Bradshaw, York W. 1993. “State Limitations, Self-Help Secondary Schooling, and Development in Kenya.” Social Forces 72:347–78.

Bratton, Michael. 1989. “The Politics of Government- NGO Relations in Africa.” World Development 4:569–87.

Carroll, Glen. 1981. “Dynamics of Organizational Expansion in National Systems of Education.” American Sociological Review 46:585–99.

Collier, Paul. 1991. “Africa’s External Economic Relations: 1960–1990.” African Affairs 90:339–56.

Firebaugh, Glenn, and Frank D. Beck. 1994. “Does Economic Growth Benefit the Masses? Growth, Dependence, and Welfare in the Third World.” American Sociological Review 59:631–53. at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from

Expansion of Enrollments in Sub-Saharan Africa 175

Fuller, Bruce. 1991. Growing Up Modern: The Western State Builds Third World Schools. New York: Routledge.

Fredrikson, Birger. 1990. “Increasing Foreign Aid for Primary Education: The Challenge to Donors.” Background paper, PHREE/90/30. World Bank, Education and Employment Division, Population and Human Resources Department. Washington, DC.

Garnier, Maurice. 2003, July. “Evaluation of Project PENGOP II.” Cotonou: U.S. AID Mission in Benin.

Garnier, Maurice, Jerald Hage, and Bruce Fuller. 1989. “The Strong State, Social Class, and Controlled School Expansion in France: 1881–1975.” American Journal of Sociology 95:279–306.

Granger, Clive W. J., and Paul Newbold. 1974. “Spurious Regressions in Econometrics.” Journal of Econometrics 2:111–20.

Hage, Jerald, and Maurice A. Garnier. 1990. “Social Class, the Hesitant State, and the Expansion of Secondary Schools in Great Britain: 1881–1975.” Research in Sociology of Education and Socialization 9:55–80.

Halaby, Charles N. 2004. “Panel Models in Sociological Research: Theory into Practice.” Annual Review of Sociology 30:507–44.

Heyneman, Stephen P., and William A. Loxley. 1983. “The Effect of Primary-School Quality on Academic Achievement Across Twenty-Nine High- and Low-Income Countries.” American Journal of Sociology 88:1162–94.

Meyer, John W. 1979. National Development and the World System: Educational, Economic, and Political Change, 1950–1970. University of Chicago Press.

Meyer, John W., Francisco O. Ramirez, and Yasmin

Nohoglu Soysal. 1992. “World Expansion of Mass Education, 1807–1980.” Sociology of Education. 65:128–49.

Meyer, John W., David Tyack, Joane Nagel, and Audry Gordonl. 1979. “Public Education as Nation Building in America: Enrollments and Bureaucratization in the American States, 1870–1930.” American Journal of Sociology 85:591–613.

Mingat, Alain, and Bruno Suchaut. 2000. Les Systèmes Educatifs Africains. Une Analyse Economique Comparative [African Educational Systems: A Comparative Economic Analysis]. Brussels: DeBoeck & Larcier.

Mingat, Alain, and Jee-Peng Tan. 1985. “On Equity in Education Again: An International Comparison.” Journal of Human Resources 20:298–308.

Moller, Stephanie, Evelyne Huber, John D Stephens, David Bradley, and François Nielsen. 2003. “Determinants of Relative Poverty in Advanced Capitalist Democracies.” American Sociological Review 68:22–51.

Nielsen, François, and Michael T. Hannan. 1977. “The Expansion of National Educational Systems: Tests of a Population Ecology Model.” American Sociological Review 42:479–90.

Schafer, Mark J. 1999. “International Non-govern- mental Organization and Thrid World Education in 1990: A Cross-National Study”. Sociology of Education 72:69–88.

—-. 2004. “Family School Enrollment in Rural Malawi and Kenya: A Regional Analysis.” Comparative Sociology 3:37–66.

Wooldridge, Jeffrey M. 2002. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press.

Maurice Garnier, Ph.D., is Professor, Department of Sociology, Indiana University, Bloomington. His main fields of interest are sociology of education (schools as organizations, expansion of edu- cational systems, and teachers in Africa). He is currently studying teachers’ recruitment, socializa- tion, and careers in French-speaking Africa and effective schools in French-speaking Africa.

Mark J. Schafer, Ph.D., is Assistant Professor, Department of Agricultural Economics and Agribusiness and Department of Sociology, Louisiana State University, Baton Rouge. His main fields of interest are education, comparative international development, and rural sociology. His current research examines the implications of the No Child Left Behind policies for rural and low-income schools in Louisiana and the United States. He is exploring the degree to which rural school strate- gies to meet NCLB requirements include increasing family and community involvement and whether these efforts have been effective. His other work also compares the implications of high-stakes test- ing for educational attainment and achievement across different state and local contexts.

Address correspondence to Maurice Garnier, Department of Sociology, Indiana University, 1210 East Kirkwood Avenue, Bloomington, IN 47405; e-mail: [email protected].

at LEHMAN COLLEGE LIBRARY on July 31, 2014soe.sagepub.comDownloaded from