Determinants of educational participation and gender differences in education in six Arab countries

Abstract

We study the determinants of educational participation and gender differences in education for young children in six Arab countries: Algeria, Egypt, Morocco, Syria, Tunisia and Yemen. Although these countries have made much progress in getting young children into school, school dropout after age 11 was still very high, and in the rural areas there were major gender differences in participation. In cities of most of these countries (except Yemen) gender differences have almost disappeared. Multivariate analyses show that similar household-level factors (e.g. wealth, education, number of siblings) as those in the West play a role, but that their importance relative to context factors is much less. For young rural girls, only 33 per cent of the variation in participation is explained by household-level factors. For older and urban girls and for boys this is more, but still substantially less than in the West. Strengthening the position of rural mothers and improving the educational infrastructure seem particularly important for reducing gender differences.

Keywords

Arab countries educational participation gender inequality multi-level urban versus rural

Introduction

The recent events in the Arab world have revealed a broadly felt discontent among the people of the Middle Eastern and North African (MENA) region. According to the Arab Human Development Reports (AHDR), this region distinguishes itself from other parts of the world in that it lacks freedom, democracy, education and gender equity (UNDP, 2002, 2003, 2006, 2009). This article focuses on the latter two aspects, the low level of education and the significant gender differences in education, which, according to the AHDR, have to be addressed if lasting change is to be brought to the region. An educated population is vital for constructing democratic societies, reducing poverty and inequality, and for building dynamic, competitive economies (Barro, 1999; Case, 2001; Sen, 1999; World Bank, 2006). A higher level of education of women has been associated with lower infant and child mortality, less malnutrition, reduced fertility and enhanced family welfare (EFA, 2011; King and Anne Hill, 1993; Lee and Mason, 2005; Smith and Haddad, 2000).

A well-educated female population is also essential for changing the traditional value patterns prevailing in many parts of the Arab world. Women with no education have little access to information. They are often unable to read and sometimes even do not speak the dominant language (Smits and Gündüz-Hoşgör, 2003). They depend heavily on people in their nearby environment for contact with and knowledge about the outside world. As a result, these women play an important role in the reproduction of existing values – including values which stress a subordinate position of women – to the next generation (Armstrong and Armstrong, 1994; Gündüz-Hoşgör and Smits, 2007).

In this article we take stock of the situation with regard to educational participation and gender differences in education in six Arab countries. We delve into the roots of under-enrolment in primary education, early school dropout in secondary education, and differences between boys and girls in these respects. The weak position of women in the Arab world is generally explained as the result of restrictive factors in the living environment (Inglehart and Norris, 2003; Moghadam, 1993, 2004). We therefore extend the classic status attainment model – which focuses on socio-economic and demographic characteristics of the family background – with cultural and economic context factors.

An important question that is addressed is whether or not the position of women in the Arab world is fundamentally different from that of women in Western countries, as has been argued by scholars such as Huntington (1996) and Inglehart and Norris (2003). If these authors are right, and gender differences are a fundamental characteristic of Arab culture, modernization of these countries might not be associated with increased female education, as it was in the West. If the differences are less fundamental, we would expect to see convergence to the more gender-equal pattern in Western countries, which only a few decades ago exhibited much higher gender differences in education (Breen et al., 2010; Buchman et al., 2008; McDaniel, 2010). By comparing areas that differ in level of development within and among Arab countries, we hope to provide more clarity regarding this issue.

For this study, we have built a multi-level database in which household-level information on 80,000 children aged 8–15 from six Arab countries is supplemented with relevant context information at the level of the nearby (village or neighbourhood) and further away (district) environment. We chose Algeria, Egypt, Morocco, Syria, Tunisia and Yemen because of their prominent roles in the recent Arab transformations and because excellent data are available for them. The data are analysed with multi-level logistic regression models, with educational participation of the children as dependent variable and socio-economic, demographic and cultural household characteristics and economic and cultural context characteristics as independent variables.

Because the rural areas of these countries are much less developed and more traditional than the urban areas, we test for variation between urban and rural areas in effects of risk factors and gender differences. Our analyses are conducted separately for children aged 8–11 and for children aged 12–15 who went to school before. The outcomes for the first group reveal the household and context determinants of going to school of young children. Those for the second group show which factors are most important in keeping children in school and preventing school dropout after age 11.

Six Arab countries

Table 1 presents background information on the educational situation and position of women in the six countries. Adult literacy rates ranged from about 53 per cent in Morocco and Yemen to 83 per cent in Syria. Female literacy rates were even lower, ranging from half of the percentage for men in Yemen to about 80 per cent in Syria and Tunisia. Hence many adult women in the region have no access to written information.

Table 1.

Characteristics of the countries.

	Algeria	Egypt	Morocco	Syria	Tunisia	Yemen
Educational structure and system
Adult literacy 2002¹	69.9	71.4	52.3	82.9	74.3	54.7
Female literacy rate as percentage of male 2002¹	76	71	60	82	78	48
Age compulsory education¹	6-14	6-13	6-14	6-11	6-16	6-14
Age primary education¹	6-12	6-11	6-12	6-12	6-12	6-12
Pupil–teacher ratio primary education 2000¹	28	22	28	24	22	22
Women’s position
Women’s non-agricultural employment 2000²	5.2	9.8	9.8	7.5	25.1	6.3
Gender Inequality Index 2005³	0.46	0.6	0.54	0.5	0.34	0.79
Women in Parliament female/male ratio 2005³	0.06	0.04	0.07	0.14	0.24	0.01

¹UNESCO (2012), ²Spierings et al. (2009), ³UNDP (2012).

The age of compulsory education is 6–16 in Tunisia, 6–13 in Egypt, 6–11 in Syria and 6–14 in the other three countries. The age of primary education varies less: 6–12 in all countries except Egypt, where it is 6–11. With regard to average class size, the countries do well. Even the countries performing worst in this respect (Algeria and Morocco) had no more than 28 pupils per teacher.

Regarding gender differences, Table 1 shows that female non-agricultural employment rates were extremely low (below 10 %) in all countries except Tunisia (25 %). Although in the rural areas many women are informally employed at the family farm, this finding indicates that empowerment of women through paid labour played a marginal role in these countries. Other gender-related indices, the Gender Inequality Index and the ratio of women in parliament, point towards a weak position of women too, with Yemen performing poorest, followed by Egypt. Tunisia, where a process of gender-related reforms has been underway since the 1950s (Moors, 1999), performs best on all indicators. It should be noted, however, that in most MENA countries there is a small group of highly educated women, mostly from middle class and higher background, that has a considerable share of high status professional occupations in law, medicine and academia (Gündüz-Hoşgör and Smits, 2008; Moghadam and Decker, 2010).

Theoretical background

According to human capital theory, participation in education is an investment in human capital made because of the expected returns later in life (Becker, 1964). In the case of young children, the investment decision is made by the parents or other caretakers. They are expected to weigh off the future benefits of sending their children to school against the immediate costs. Those benefits can be for the child, but also for the parents themselves, because, in the absence of pension systems, children may be their old-age security. The decisions made by the parents are influenced by how they perceive the world around them and may be coloured by cultural norms and values – which may legitimize existing inequalities.

Of the factors influencing a child’s educational chances in the Arab world, gender is one of the most important. This is particularly true in the rural areas. Although the situation has improved in recent decades, the Arab countries are still characterized by the world’s highest levels of gender disparity in education (EFA, 2011). The major reason is the weak position of women in the region, which is generally associated with the dominant patriarchal culture (Colclough et al., 2000; Gündüz-Hoşgör and Smits, 2007; Moghadam, 2004). The Middle Eastern countries are part of what has been called the ‘patriarchy belt’ (Caldwell, 1982), which includes – besides the Arab world – the Asian countries with a Hindu or Confucian background.

The form of patriarchy in this region is characterized by patrilocal extended households, where power is in the hands of the senior male, and property, residence and descend generally proceed through the male line (Moghadam, 2004). In this ‘classical patriarchal system’ (Kandiyoti, 1988), girls are given in marriage at a young age and then move to their husband’s household, where they are subordinated to all men and senior women. The major task of women is producing offspring, and the power they can eventually obtain is through their sons. The major task of the men is to provide income and security for their families. There is a strict separation between the male and female domains, with men operating in the public sphere and women in the private. Given the strong emphasis on the male breadwinner role, the high fertility level, and the restriction of women to the private domain, girls’ educational attainment and women’s employment in the formal economy are very low (Gündüz-Hoşgör and Smits, 2006; Spierings et al., 2010).

Since the onset of modernization, the system has come under pressure. A reduction in employment possibilities in rural areas and massive migration to the cities have eroded the extended family system and the male provider role. However, its ideology is still broadly upheld and its influence strongly present in the legal framework and institutions of most Arab countries (Moghadam, 2004).

Regarding cultural influences on education, Colclough et al. (2000) argue that poverty may be a major cause of under-enrolment, but that gender differences in enrolment are the product of cultural practices. Such differences are influenced not only by the values of the parents, but also by the ideas present in the environment in which the household lives, as it is difficult for parents to escape social pressures of neighbours, friends and extended family members (Webbink et al., 2012). Consequently, we expect the educational participation of girls to be lower if the context is more traditional.

In general, the influence of neighbourhoods should not be overestimated (Solon et al., 2000). In Western countries, most (80–90 %) of the variation in educational attainment between children is due to household-level factors (Breen and Jonsson, 2005); whether this is also true in Arab countries remains to be seen. In those countries, besides norms that restrict women’s movement in public space, there are also marriage traditions that may negatively affect the educational participation of girls. When girls are expected to marry into their husband’s families, the parents may be less willing to invest in their education, since returns to this investment go to the husband’s family (Smits and Gündüz-Hoşgör, 2006).

Besides cultural factors, economic and infrastructural factors might be important too. These include availability of schools, transportation and communication infrastructure. In more developed regions, facilities are often better, the influence of globalization stronger, and the idea that education is essential for both boys and girls more dominant (Huisman and Smits, 2009). This may be the case particularly in urban areas. For Arab countries it has been found that educational infrastructure and the influence of modern values are strongest in the cities and weakest in the countryside (UNDP, 2002). We therefore expect substantial differences in educational participation between urban and rural areas. In less developed and rural regions, schooling may entail higher costs because of the more limited availability and accessibility of schools (Hazarika, 2001; Mugisha, 2006). Returns to education are also higher in cities because the chances of finding a white collar job are better (Hazarika, 2001; Huisman and Smits, 2009).

Control factors

Our model contains a number of control factors that are known or expected to be related to educational enrolment. At the household level, parental education, father’s occupation, mother’s work status and household wealth are all important determinants of educational participation in both the developed and developing world (Blood and Wolfe, 1960; Coleman et al., 1966; Evangelista de Carvalho Filho, 2008; Glewwe and Jacoby, 2004; Jencks, 1972; Mingat, 2007; Rodman, 1972; Shavit and Blossfeld, 1993; Tansel, 2002). Demographic characteristics of the child – age, birth order and being a biological child (Chiswick and DebBurman, 2004; Dayıoğlu et al., 2009; Ejrnaes and Portner, 2004; Kirdar et al., 2007) – as well as of the household they live in – presence of the parents, extended or nuclear family, and number of siblings (Booth and Kee, 2005; Bradbury, 2007; Fafchamps and Wahba, 2006; Li et al., 2008; Wichman et al., 2006) – are also known to influence educational enrolment.

Data and methods

We use data from the Pan Arab Project for Family Health (PAPFAM) of the League of Arab States (www.papfam.org) and from the Demographic and Health Surveys (DHS, www.measuredhs.com). Data sets are available for Algeria 2002 (PAPFAM), Egypt 2005 (DHS), Morocco 2003 (combined PAPFAM and DHS), Syria 2001 (PAPFAM), Tunisia 2001 (PAPFAM) and Yemen 2003 (PAPFAM). Both PAPFAM and DHS use two-stage cluster samples, a cluster being a village or neighbourhood. Within each cluster, about 30 randomly sampled households obtain an oral interview in which basic information on all household members, including educational information, is collected. All women aged 16–49 are subsequently interviewed on demographic, health and socio-economic topics. The response rates are over 91 per cent in all countries.

Besides household-level data, we use context information at the cluster and district level which was aggregated from the household surveys. Our combined data set contains information on 79,846 children (39,047 girls and 40,799 boys) aged 8–15 living in 3,842 clusters and 107 districts within six countries. The distribution of districts and clusters over the countries is: Algeria 29, 510; Egypt 22, 1,359; Morocco 15, 480; Syria 14, 500; Tunisia 7, 344; Yemen 20, 649. Given that for our analyses relatively simple demographic and socio-economic variables are used, no serious comparability problems were encountered in creating the variables from different surveys.

Method and dependent variable

The data are analysed with multi-level regression models (also called mixed models or hierarchical linear models; compare Goldstein, 2011; Snijders and Bosker, 1999). Dependent variable is a dummy indicating whether (1) or not (0) a child was enrolled in education at the time of the survey. We apply four-level models because we use data on families nested within clusters, nested within districts, nested within countries. The models are estimated with MLWin, using second-order PQL, the recommended estimation technique for multi-level logistic regression analysis (Goldstein, 2011). To determine how much of the variance is explained by factors at the different levels, we compute the intra-class correlations rho (ρ) (also called Variance Partition Coefficient) using the Latent Variable Approach (compare Goldstein, 2011; Snijder and Bosker, 1999: 224). For this purpose, empty models were estimated separately for the different combinations of urbanization, age and gender (rural girls aged 8–11, urban girls aged 8–11, and so on).

Our analyses are restricted to children under age 16, because parental information is often not available for older children (who may have left their parental families because of early marriage). To make a division between primary and (lower) secondary education, the analyses are done separately for children aged 8–11 and children aged 12–15. The lower boundary of 8 is taken because in these countries many children start school at an older than legally required age. The analysis for children aged 12–15 is restricted to those who have been enrolled in school before, in order to focus on the determinants of staying in school. Since, for Tunisia, information about having been in school before was only available for children aged 14 and over, for this country the analysis of children aged 12–15 includes some children who never went to school.

To determine the degree to which the effects of our independent variables differ between boys and girls and between urban and rural areas, interactions between all independent variables and gender and living in a rural area were tested and included in the model if found significant. We tested both the two-way interactions with either gender or living in a rural area and the three-way interactions with both variables. To compute the interaction terms, centred or standardized versions of the variables were used. The main effects can therefore be interpreted as average effects. To be able to focus on the most important interaction effects, only significant interactions were included in the final model. In this way a parsimonious picture is obtained of the way in which the effects of the independent variables differ between boys and girls and urban and rural areas. To obtain all relevant coefficients with their standard errors for the interaction variables, the models were estimated several times with the categories of these variables exchanged. Since our analysis of the interaction effects is meant to be explorative in nature, no hypotheses have been formulated.

Independent variables

Fathers’ occupation is measured in three categories: (1) farm, (2) lower non-farm (sales, services, manual), (3) upper non-farm (professional, technical, managerial, clerical). Work status of the mother is a dummy indicating whether (1) or not (0) she was employed. Education of the mother and father is measured in years. Because income is lacking in the data sets, household wealth was measured by an index constructed on the basis of household assets. Using a method developed by Filmer and Pritchett (1999), all households within a country were ranked on the basis of these characteristics and divided into wealth index deciles.

Presence of the parents was measured by dummies indicating whether (1) or not (0) the mother or father was missing from the household. Missing fathers and mothers include parents who are divorced or separated, parents who live elsewhere because of work, as well as deceased parents. Birth order and number of sisters and brothers were measured with interval variables. Family structure was measured with a dummy indicating whether the household was a nuclear family (1), or, besides parents, also other adults were present (0). Being a biological child was measured using a dummy with categories (0) for foster, adopted or unrelated children and (1) for biological children. Age of the child was measured in years. Sex of the child was measured as (0) for boys and (1) for girls. As cultural indicators at the household level, we used the age difference and the difference in years of education between the parents. Traditionalism of the household was further indicated by a dummy indicating whether (1) or not (0) the mother had her first child under age 18. Although teenage pregnancies may have various reasons, they are often associated with arranged early marriage and hence may indicate a traditional family background. All interval variables at household and context level were standardized to make their coefficients comparable.

Children with a missing parent were given the mean score of the other children in the database on the variables indicating characteristics of the parents. Because we included dummies indicating whether (1) or not (0) the mother or father is missing, this procedure leads to better estimates of these variables (Allison, 2001: 87). For children with mothers younger than 16 or older than 49, information on occupation of the father, employment of the mother and the age at which the mother had her first child was not available in the data for Morocco and Egypt. To be able to include those children in the analyses, we gave them, on these variables, the average of the children for which information was available and we added a dummy to identify them in the analysis. To test whether this procedure might have influenced our results, two robustness tests were performed. In the first, the models were estimated again after removing (both separately and jointly) these variables from the model. In the second test, the models were estimated after removing the children with missing parents on these variables. Both tests showed that the way we handled these missing parents hardly influenced our results.

Context characteristics are degree of modernization, urbanization, educational facilities and patriarchy. District level of modernization was measured by an index constructed on the basis of four variables: percentages of households having a car, a fridge, a television and running water. Of these characteristics the mean was taken of the standardized values. For Syria, information on water was lacking, so the index is based on the ownership of the three household assets. Urbanization is measured with a variable indicating whether (0) or not (1) the household lived in a city. The availability of educational facilities in the nearby environment was indicated by the average number of years of education of adult males in the cluster (village or neighbourhood). Education of males was chosen because education of females is also strongly influenced by cultural factors.

The influence of traditionality and patriarchy was measured by the difference in average number of years of education between fathers and mothers in the cluster and by the percentage of households with grandfathers from father’s side in the district. The last variable is meant to indicate the tendency of girls to marry into the family of the husband. In cultures where sons are reckoned to look after their parents in old age, parents are more inclined to invest in their sons. This also means that in cultures where ‘a girl’s allegiance after marriage is mainly to her future husband’s family, the balance of perceived benefits to parents is likely to favour the education of sons over daughters’ (Colclough et al., 2000: 7).

Results

Figure 1 shows that at the time of interview (2001–2005) non-participation was lowest among children aged eight and nine (in Yemen aged 11). Non-participation rates for girls at that age were between 3 and 5 per cent in most countries, except for Morocco with 10 per cent and Yemen with 35 per cent. For boys, non-participation rates were between 1 and 3 per cent in most countries, 6 per cent in Morocco and 13 per cent in Yemen (at age 9). From age 11 onwards, participation decreased strongly in all countries. For girls, dropout was especially strong in Morocco, Syria and Yemen, with over half of 15-year-old girls out of school. For boys, dropout was strongest in Morocco and Syria (with over 40 per cent of 15-year-old boys out of school). Gender differences were most pronounced in Yemen, with 30 per cent of girls against 10 per cent of boys not going to school at age 11. The relatively high participation rates at a young age in most countries, together with the high dropout rates after age 11, indicate that the major challenge faced by these countries has shifted from getting children into school to keeping them there.

Figure 1.

Percentage of girls and boys not in school by age.

Our data further reveal that in all six countries educational participation was substantially lower in the countryside than in the cities (Appendix A1 and A2). For young girls, the differences were particularly large, with up to seven times higher non-participation rates in the countryside. The most extreme case is Yemen, with as many as 44 per cent of young rural girls against 17 per cent of young rural boys out of school. Morocco follows with 21 per cent of young rural girls and 12 per cent of young rural boys out of school. In the other countries, the situation was better, with non-participation rates of rural young children below 10 per cent.

In the cities, educational participation of young children was much higher than in the countryside. In all countries except Yemen, no more than 3 per cent of urban children aged 8–11 were out of school. In Yemen, 6 per cent of urban boys and 9 per cent of urban girls were out of school at this age. An important finding is that in the urban areas the difference in non-participation between boys and girls had almost disappeared and that this difference in Tunisia was even to the advantage of girls.

Among children aged 12–15 who went to school before, participation rates were substantially lower than among the younger ones, thus highlighting again the high dropout rates in these countries. Participation rates among older children were particularly low in rural areas, where also gender differences were larger. In Morocco and Yemen, less than 40 per cent of older rural girls were in school. Again we see that in all countries except Yemen gender differences had almost disappeared in the cities, where participation in several countries was higher among girls than among boys.

Factors influencing educational participation

Appendices A1 and A2 present the non-participation rates of younger and older boys and girls in the six countries according to characteristics of their family background. As expected, socio-economic characteristics of the parents play an important role. If the mother has at least some primary education or the father has more than primary education, the chances that a child is in school increase substantially. The same is true when the father is working in a non-farm occupation and even more so if he is working in an upper non-farm (white-collar) occupation. Household wealth is also an important predictor of school participation, which indicates that financial restrictions may still be important in these countries.

The effects of the demographic factors are less pronounced than those related to socio-economic factors. Most demographic factors show the expected link with going to school in most circumstances, but there are also exceptions. Several factors seem to be stronger for girls than for boys. This is for example the case for number of sisters and brothers and for absence of the mother.

Explained variance at household and context level

As mentioned before, in Western countries 80 to 90 per cent of the variation in school outcomes is estimated to be due to household-level factors (Breen and Jonsson, 2005). Our figures for the Arab countries are much lower. For young girls in rural areas, household-level factors explain only 33 per cent of school outcome variation. The figures are higher for young urban girls (53 %), young rural boys (51 %), young urban boys (58 %), older rural girls (55 %) and older urban girls (58 %). The role of the context is largest for older boys in rural (68 %) and urban (66 %) areas. Still, even for them it is substantially lower than the figures of Breen and Johnson (2005) for the West. Hence, although the models on which Breen and Johnson base their figures may not include exactly the same variables as ours, findings strongly suggest that the environment in which a child lives is much more important in MENA countries than it is in the West. Particularly for young rural girls, contextual factors such as culture and development to a very large extent determine their chances of going to school.

Multivariate analysis

Tables 2 and 3 present the coefficients of the multi-level logistic regression models. For factors that interact significantly with sex and/or urbanization, separate coefficients are presented for boys and girls, urban and rural areas, or both. Coefficients that do not differ according to sex or urbanization are presented in output column 1 (All); those that differ according to sex in columns 2 and 3; according to urbanization in columns 4 and 5; and according to both sex and urbanization in columns 6 to 9.

Table 2.

Exponential coefficients (odds ratios) of multi-level logistic regression analyses for children age 8–11 with the odds of being in school as dependent variable.

	All	Girls	Boys	Rural	Urban	Rural Girls	Rural Boys	Urban Girls	Urban Boys
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Household-level variables
Demographic factors
Age	0.47**
Age square	0.69**
Sex is girl				0.54**	0.78**
Mother missing				0.72**	0.47**
Father missing	0.96
Nuclear family	0.94
No. of sisters		0.91**	1.05
No. of brothers		0.91**	1.08*
Birth-order child		1.04	0.83**
Biological child	1.46**
Socio-economic factors
Occupation father
Farm	Ref.
Lower non-farm	1.24**
Upper non-farm	1.32**
Mother employed		0.82**	1.07
Education father (years)		1.4**	1.24*
Education mother (years)	1.22*
Household wealth	1.76**
Household wealth square	0.86**
Cultural factors
Age difference parents	1.01
Education difference parents	0.91
Mother had 1st child under 18	0.92
Contextual variables
Socio-economic indicators
District development index						1.1	0.96	0.85*	0.98
Living in rural area		1.04	1.48**
Education men in cluster						1.99**	1.64**	1.47**	1.21**
Cultural indicators
Gender diff. education cluster		0.75**	0.93
District % patriarchal households	0.97
N	39,393	19,263	20,130	23,246	16,147	11,238	12,008	8,025	8,122

Coefficients in Column 1 are for variables which do not differ significantly between sex and urbanization. Columns 2 and 3 contain coefficients that differ significantly between boys and girls. Columns 4 and 5 contain coefficients that differ significantly between urban and rural areas. Columns 6–9 contain coefficients that differ significantly both between boys and girls and between urban and rural areas.

*p<0.05; **p<0.01.

Table 3.

Exponential coefficients (odds ratios) of multi-level logistic regression analyses for children aged 12–15, with the odds of being in school as dependent variable.

	All	Girls	Boys	Rural	Urban	Rural Girls	Rural Boys	Urban Girls	Urban Boys
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Household-level variables
Demographic factors
Age	0.16**
Age square	1.31**
Sex is girl				0.48**	1.07
Mother missing		0.48**	0.85
Father missing		1.04	0.78**
Nuclear family	1.02
No. of sisters	0.97
No. of brothers						0.84**	0.97	0.77**	0.89**
Birth-order child		1.34**	1.19**
Birth-order child square	1.04**
Biological child				0.97	1.47*
Socio-economic factors
Occupation father
Farm	Ref.
Lower non-farm	1.05
Upper non-farm	1.24**
Mother employed	0.90
Education father (years)						1.41**	1.81**	1.52**	1.46**
Education father square	1.14**
Education mother (years)				1.26**	1.46**
Household wealth						1.18**	1.73**	1.37**	1.63**
Cultural factors
Age difference parents	0.98
Education difference parents	0.91
Mother had 1st child under 18				1.00	0.75**
Contextual variables
Socio-economic indicators
District development index	0.86
Living in rural area		0.73**	1.63**
Education men in cluster						2.34**	1.35**	1.50**	1.23**
Cultural indicators
Gender diff. Education cluster						0.61**	1.08	0.75**	0.96
District % patriarchal households						0.92	1.06	1.22**	1.40**
N	36,335	16,651	19,684	19,814	16,521	8,629	11,185	8,022	8,499

*p<0.05; **p<0.01.

The coefficients presented are exponential versions of the logistic regression coefficients, because these are more easily understood. For instance, the value of 1.22 for the effect of mother’s education on young children’s enrolment means that for each year increase in the mother’s education the odds of being enrolled are multiplied by a factor of 1.22 (or are 22 % higher). The value of 0.72 for the missing mother of young rural children indicates that the odds should be multiplied by a factor of 0.72 (or are 28 % lower).

Girls have substantially lower odds of being in school than boys. This difference is strongest in the rural areas, but for young girls also in the urban areas. Hence the near disappearance of urban gender differences observed in the bivariate figures (Appendix A1) for young children was probably due to a higher prevalence of factors positively related to girls’ school participation there (such as a more highly educated population, smaller family size and weaker influence of traditional values). For older urban children, the outcomes of the bivariate analyses are confirmed: In the cities, no significant multivariate gender differences are left.

The effect of age is non-linear for younger as well as older children. Studying the coefficients of this (standardized) variable further shows that for young children the odds of being in school first increase and then decrease, indicating that dropping out of school starts before age 11. For older children the dropout continues but with decreasing speed. When the mother is missing from the household, this has a negative effect on the likelihood of younger children and older girls being in school. For younger children, the presence of the mother is particularly important in the cities. These children may be pulled out of school to take over their mother’s duties. For older boys, a missing mother does not have a significant effect, but for them missing their father is significantly negative, indicating that they have to take over their fathers' tasks. Living in a nuclear or extended family does not make a difference for children’s schooling.

Our analyses reveal clear evidence of competition over scarce resources for education. Girls have less chance of being in primary school if they have more sisters and brothers. They also tend to drop out more if they have more brothers. Hence the burden of a large number of children is mostly carried by daughters. For young boys, having more brothers positively affects primary school participation, perhaps because there are more hands available for doing male family work. However, for older urban boys, having more brothers means a higher likelihood of school dropout, probably because cities offer more opportunities for (child) labour and the schooling costs of younger brothers are higher there. Birth order also plays an important role. Earlier-born boys are more likely to be in primary school than later-born boys, thus indicating that preference is given to the older sons. However, once in school, the earlier-born children, in particular girls, tend to drop out more. There also seems to be competition between biological and non-biological children, with non-biological children having a lower likelihood of being in primary school and, in urban areas, also of being in secondary school.

As expected, children living in households with more socio-economic resources are more likely to be in school. This is the case for children with parents with more education, a father with a higher level job, or wealthier parents. Only employment of the mother is an exception. Having a working mother decreases the chances of young girls being in school, indicating that these mothers work out of poverty and their young daughters have to take over their household tasks.

Our indicators for parents' traditionality show little effect on children’s chances of being in school. A larger difference in age or education between the parents has no effect at all, and having a mother who had her first child young only shows a significant negative effect for older urban children. The fact that this variable is only significant in urban areas might be because starting child-bearing young is less common there and women who do so may on average be more traditional.

The importance of education in the local context (as reflected by a higher educational level of adult men in the cluster) is strongly positively related to educational participation of all children and – more interestingly – the latter effect is particularly strong for girls in rural areas. Hence better facilities and a context where education is valued more have a positive effect on educational participation, and it is the girls in more difficult circumstances who profit most from this.

The, at first sight counterintuitive, finding that young girls are less in school if they live in more developed urban areas must probably be seen in light of this strong positive effect of the cluster educational level. When this effect is taken into account and the socio-economic and demographic factors at the household level are controlled for, development might for example mean more opportunities for poor mothers to work and hence more need for daughters to help at home.

Living in a more traditional context, as indicated by the gender difference in education in the cluster, has a significantly negative effect on the likelihood of girls being in school, whereas this makes no difference for boys. Hence, even after controlling for many other factors, traditional values still seem to have an independent negative effect on the educational participation of girls compared with boys. Contrary to expectations, living in a district where more households include grandparents from the father’s side positively affects older children in the cities staying in school. It is possible that these grandparents may create opportunities for their grandchildren to stay in school by taking over household tasks.

Conclusions

In recent decades the Arab countries studied in this article, Algeria, Egypt, Morocco, Syria, Tunisia and Yemen, have made substantial progress in getting young children into school. At the time of the survey, in four of the six countries (except Yemen and to a lesser extent Morocco) only a small percentage of children aged eight and nine were not in school. After age 11, however, participation rates decreased rapidly, thus indicating that the major challenge for policy-makers in these countries has shifted from getting children into school to keeping them there.

Another important problem is the low participation rates in rural areas, where girls in particular are kept at home. In those areas, between 6 and 44 per cent of younger girls and between 21 and 63 per cent of older girls were not enrolled at the time of the survey. This may have been due to the poor schooling and transport infrastructure and cultural restrictions on the freedom of movement of girls. The fact that for rural boys the situation was better (with non-participation percentages about half those of the girls) indicates that infrastructural restrictions are only part of the problem and that preferences of families for boys’ education over girls’ education play an important role.

In the cities, the situation is much better. Urban participation rates were substantially higher than rural ones, and urban gender differences had almost disappeared. In some cases, participation of urban girls was even higher than that of urban boys and, except for Yemen, nowhere was the difference more than 2 per cent to the disadvantage of girls. Hence, in the cities, the position of girls is considerably stronger than in the countryside, at least as far as participation percentages are concerned. Nevertheless, in the multivariate analyses, young urban girls were found to be significantly less in school than young urban boys. This indicates that for them the more favourable participation percentages were due to an over-representation of girl-friendly factors in the cities, which counteract discriminatory practices there.

For older urban girls, the situation seems better, as in the multivariate analysis no significant gender differences were found. However, a caveat is in place regarding this finding. The older girls in our analyses may be a selective group consisting of those who managed to continue their education in a girl-unfriendly environment. As these girls might be more highly motivated and talented than the ones who dropped out, our analyses might underestimate the true gender difference in the older group.

An important new finding of our study is that in the countries under study a large part of the variation in educational participation of children is explained by characteristics of the context in which they live. Whereas Breen and Jonsson (2005) estimated that in Western countries a major part (80–90 %) of the variation in educational attainment was due to factors at the household level, we found household factors explaining only 33 per cent of the variation in educational participation of young rural girls and between 51 and 58 per cent of older and urban girls and young boys. For older boys, the family background is relatively more important, explaining about 67 per cent of the variation in participation. Still, this is substantially less than the 80 to 90 per cent mentioned by Breen and Jonsson. Hence, in the Arab world, the environment in which children are born seems to influence their educational chances to a much greater extent than in Western countries.

With regard to the determinants of educational participation, the factors central in status attainment and social mobility models – father's and mother's education, household wealth and father’s occupation – were found to play an important role in these Arab countries as well. This might come as no big surprise, as these factors have consistently been found to profoundly affect educational attainment and achievement in other regions of the globe (e.g. Coleman et al., 1966; Evangelista de Carvalho Filho, 2008; Glewwe and Jacoby, 2004; Jencks, 1972; Mingat, 2007; Shavit and Blossfeld, 1993; Tansel, 2002). Nevertheless, it is important to establish that in these Arab countries similar relationships are found as in the West.

Besides socio-economic resources, demographic factors such as age, absence of parents, number of siblings, birth order, and being a biological child are all linked to educational participation. The likelihood of being in school first increases with age, but before age 11 starts to decrease. The coefficients of the number of brothers and sisters and birth order indicate that, in decisions about entering school, preference is given to earlier born sons. At the same time we found that earlier-born children (both boys and girls) are also the first to drop out, suggesting that they are reckoned upon to help at home.

Children in communities where men on average have higher levels of education are more likely to go to and stay in school. This effect is particularly strong for girls from rural areas. A higher current educational level of males indicates that, in the past, local educational facilities were good and that the area might be characterized by a positive attitude towards education. Our results indicate that these circumstances have beneficial effects on educational participation in general, but that rural girls may profit most.

The high dependency of Arab girls’ educational participation on the context in which they live, together with the more favourable situation for girls in urban areas and in (rural) areas with higher male education, contrasts with Huntington’s (1996) and Inglehart and Norris’s (2003) idea of a homogeneous Arab world that is fundamentally different from the West. Our analyses make clear that girls’ educational participation tends to rise and gender differences to decrease as soon as the opportunity structure becomes more favourable for girls, just as happened earlier in the West. Hence policies aimed at increasing girls’ education should first and foremost focus on the opportunity structure.

A major starting point in this respect is the educational infrastructure in rural areas, as rural girls seem to profit a lot from better facilities. However, improving facilities alone might not be very successful when traditional value patterns are left intact that prevent girls from entering school or staying in school after reaching puberty. Earlier research (Armstrong and Armstrong, 1994; Gündüz-Hoşgör and Smits, 2007) suggests that in this respect it is very important to reach the uneducated and often illiterate mothers of these girls because of the major role they play in transferring traditional values, including those that stress an inferior position of women, to the next generation.

Footnotes

Appendix A1

Percentage not in school of children aged 8–11 by characteristics of their family background and urbanization.

	Boys							Girls
	Algeria	Egypt	Morocco	Syria	Tunisia	Yemen	Total	Algeria	Egypt	Morocco	Syria	Tunisia	Yemen	Total
Mother present?
Yes	1.9	3.0	8.0	2.6	2.4	13.6	6.2	3.6	6.2	12.5	4.8	3.6	33.8	13.5
No	4.9	11.7	7.0	15.6	5.3	21.1	15.0	5.4	6.5	18.2	14.3	7.1	38.7	23.3
Father present?
Yes	1.8	3.2	8.0	2.7	2.2	14.8	6.4	3.7	6.2	13.0	4.9	3.6	35.4	13.8
No	3.0	3.5	7.9	3.5	5.5	10.0	6.9	3.6	6.3	11.7	5.5	4.3	27.1	14.4
Kind of family
Extended	2.5	4.3	8.1	4.5	2.7	12.0	6.9	4.0	9.2	13.4	6.3	4.3	31.7	15.2
Nuclear	1.4	2.4	7.8	1.9	2.3	16.3	6.2	3.4	4.2	12.0	4.2	3.1	36.6	12.8
Birth-order child
First-born child	1.0	1.6	5.5	1.9	1.7	12.1	4.4	1.5	2.4	10.0	3.7	1.1	32.0	10.2
2nd–4th child	1.9	2.9	7.6	1.8	1.3	14.5	5.9	4.1	5.2	12.2	4.7	2.8	34.0	12.5
5th or later child	2.3	5.2	10.1	3.7	4.1	14.4	8.0	4.2	11.7	15.4	5.4	5.9	34.8	17.0
No. of sisters
None	1.3	3.3	6.3	2.6	1.5	16.2	5.3	1.2	3.1	9.2	2.4	1.3	31.7	9.0
One or two	1.8	2.4	7.6	1.5	2.4	14.2	5.8	3.4	5.4	13.0	3.8	3.9	34.0	12.5
Three or more	2.4	6.3	10.6	4.4	3.6	13.6	8.4	5.3	12.0	15.2	6.8	4.8	34.8	18.3
No. of brothers
None	2.1	1.4	5.0	4.1	1.7	13.8	4.2	1.6	3.7	7.9	5.0	1.8	33.1	8.7
One or two	1.3	3.2	7.2	1.7	2.1	13.5	5.4	3.1	5.0	11.4	4.4	3.0	33.3	11.2
Three or more	2.7	4.8	12.2	3.6	3.8	14.6	8.8	5.3	12.6	20.1	5.4	6.2	34.8	19.4
Occupation father
Farm	5.4	6.2	15.6	4.7	2.9	23.7	12.4	13.0	15.3	25.0	9.4	5.6	58.1	27.1
Lower non-farm	1.0	2.8	4.6	2.7	2.6	14.9	6.1	2.5	4.7	7.3	4.5	4.0	35.2	13.2
Upper non-farm	0.4	0.7	1.3	1.2	1.0	6.2	2.2	0.8	0.8	4.4	2.3	1.1	15.3	5.0
Mother employed
No	2.0	3.0	7.4	2.3	2.6	13.8	6.2	3.8	5.5	11.1	4.2	3.8	32.9	13.0
Yes	0.5	2.4	8.9	3.9	1.1	12.8	5.9	2.0	7.7	14.1	7.7	1.8	39.2	15.4
Education father
None	3.3	7.2	11.1	6.2	3.0	19.6	11.9	7.1	15.6	18.3	10.7	7.1	46.9	25.9
Some primary	1.0	3.7	5.3	2.5	2.5	10.0	4.1	1.5	6.4	8.6	5.3	3.4	27.1	9.0
More than primary	0.5	0.8	0.8	0.9	1.0	4.3	1.3	0.9	1.0	0.8	0.9	1.0	10.0	2.5
Education mother
None	2.7	5.5	9.7	4.7	3.4	15.7	9.5	5.8	11.9	15.8	9.2	6.1	38.7	21.1
At least some primary	1.0	1.3	1.9	1.1	1.7	3.7	1.5	1.1	1.7	2.5	1.8	1.2	7.6	2.2
Household wealth
Lowest quintile	4.6	6.6	15.1	3.5	5.0	38.0	14.2	11.8	16.7	31.3	9.1	10.8	74.0	29.2
Middle quintiles	1.3	2.3	6.0	2.9	1.6	10.2	4.9	1.6	2.6	7.2	4.7	1.2	32.4	11.4
Upper quintile	0.6	0.4	1.5	1.2	1.0	2.0	1.2	0.8	0.1	2.3	1.1	0.6	4.3	2.0
Mother 1st child <18
No	1.8	2.5	7.8	2.5	2.4	14.9	5.6	3.3	5.6	11.9	4.7	3.5	36.4	12.0
Yes	2.9	4.8	8.7	2.7	3.2	11.7	8.0	7.1	8.6	14.9	4.9	6.3	30.0	18.6
Urbanization
Urban	1.0	1.6	2.9	2.5	1.7	6.2	2.7	2.1	2.0	3.0	3.5	1.1	8.9	3.6
Rural	3.1	4.1	12.0	2.9	3.4	16.9	9.1	5.8	8.8	21.0	6.3	6.8	43.6	21.2

Appendix A2

Percentage not in school of children aged 12–15 by characteristics of their family background and urbanization.

	Boys							Girls
	Algeria	Egypt	Morocco	Syria	Tunisia	Yemen	Total	Algeria	Egypt	Morocco	Syria	Tunisia	Yemen	Total
Mother present?
Yes	11.3	10.8	28.2	30.4	12.0	16.9	17.9	19.2	15.1	38.9	33.7	12.9	49.7	30.5
No	18.8	18.2	29.8	36.4	26.1	24.5	25.1	22.9	26.2	46.5	55.1	27.3	65.7	49.1
Father present?
Yes	11.2	10.5	28.0	30.1	11.9	17.9	18.0	19.4	14.6	39.8	34.1	13.3	51.0	31.0
No	14.2	16.0	29.4	35.9	15.3	15.4	20.6	19.0	23.0	38.8	37.7	11.1	52.5	36.1
Kind of family
Extended	12.1	13.8	29.5	32.8	14.2	15.4	18.9	20.5	18.9	40.9	37.9	14.9	47.6	32.9
Nuclear	10.7	8.3	26.4	28.6	9.9	20.4	17.5	17.6	12.6	37.5	31.0	11.1	56.5	30.1
Birth-order child
First-born child	10.2	7.4	26.1	26.8	7.6	22.5	16.9	14.6	12.8	39.6	32.5	8.5	58.4	30.1
2nd–4th child	13.1	12.0	29.1	32.2	13.1	19.1	19.3	21.3	15.3	38.5	36.0	13.2	57.3	32.5
5th or later child	10.8	13.3	29.1	31.0	14.4	13.7	18.0	19.8	20.1	41.4	33.9	15.7	42.0	31.5
No. of sisters
None	10.4	12.5	26.0	30.0	7.0	30.2	19.0	17.8	14.1	37.1	33.3	10.1	59.5	28.7
One or two	11.1	9.9	27.7	28.0	13.1	16.7	17.1	19.0	14.9	39.9	29.9	12.1	50.3	29.4
Three or more	12.5	13.5	31.4	33.9	13.3	16.0	19.8	20.3	19.1	40.9	39.2	16.6	50.3	36.2
No. of brothers
None	11.7	5.9	21.8	27.2	9.8	22.7	14.6	12.9	11.3	27.4	34.0	5.9	60.8	23.7
One or two	9.5	11.3	26.0	26.4	11.1	18.0	16.9	15.3	13.4	37.7	28.4	10.8	50.7	26.6
Three or more	13.5	14.4	37.0	35.0	15.7	16.6	20.9	24.7	24.9	50.9	39.6	19.8	50.7	39.7
Occupation father
Farm	21.4	16.8	44.5	34.8	19.0	26.2	27.9	44.0	29.1	62.7	48.3	25.9	70.3	51.3
Lower non-farm	10.4	10.8	20.7	35.5	13.6	18.0	18.1	18.0	13.6	29.1	35.6	13.6	50.9	30.6
Upper non-farm	2.8	1.7	9.0	19.2	4.3	6.6	7.8	6.3	2.5	12.0	21.5	5.6	28.2	13.9
Mother employed
No	11.6	10.8	26.4	31.9	11.8	16.8	17.9	19.7	14.5	37.4	33.8	13.4	48.2	30.4
Yes	6.0	9.6	30.8	22.8	14.0	17.5	16.5	11.5	13.3	36.8	32.9	9.2	57.6	30.0
Education father
None	16.1	19.7	36.7	47.1	20.2	23.4	25.8	29.2	28.3	49.3	57.3	23.5	59.9	46.6
Some primary	10.1	12.1	18.4	31.8	11.9	9.1	17.1	14.0	16.1	34.8	34.7	14.2	41.1	26.1
More than primary	2.7	2.3	6.8	13.3	2.8	3.6	4.9	6.7	3.8	8.1	14.3	1.3	17.9	8.2
Education mother
None	14.9	17.2	31.9	38.8	19.4	18.7	22.6	26.1	25.8	45.8	48.4	21.4	55.1	42.0
At least some primary	5.7	4.9	10.7	21.3	4.5	5.9	9.1	7.7	5.2	11.2	19.2	4.6	13.8	9.8
Household wealth
Lowest quintile	23.8	20.9	50.1	37.4	24.7	41.8	32.9	44.4	35.3	70.0	49.6	32.9	82.5	54.3
Middle quintiles	9.0	8.9	24.3	31.8	9.5	14.3	16.2	14.6	10.7	33.7	35.1	8.3	53.9	29.8
Upper quintile	4.5	1.8	5.1	19.1	2.2	6.0	6.6	4.3	1.4	9.8	14.1	0.9	16.8	9.5
Mother 1st child <18
No	11.4	10.4	29.0	28.8	11.9	19.0	18.1	18.9	13.9	40.7	33.1	12.7	52.3	29.6
Yes	12.4	13.2	28.7	37.7	18.5	14.2	18.6	25.9	21.5	40.1	36.4	17.3	46.6	36.7
Urbanization
Urban	8.7	8.9	14.5	32.1	7.1	11.2	14.0	9.8	8.1	16.0	28.2	4.3	20.4	14.7
Rural	15.0	12.4	40.9	29.3	18.9	20.0	21.4	31.4	20.5	61.2	39.9	24.2	62.9	44.0

Acknowledgements

We are grateful to the Pan Arab Project for Family Health of the League of Arab States and Dr. Ahmed Abdel Monem for making the Papfam data sets available for this project and to MEASURE DHS for providing the Demographic and Health Surveys.

Funding

This research was in part made possible by VIDI grant nr. 452-03-351 of the Netherlands Scientific Organization (NWO).

References

Allison

(2001) Missing Data. Thousand Oaks, CA: Sage.

Armstrong

(1994) Theorizing Women’s Work. Toronto: Garamond.

Barro

(1999) Determinants of democracy. Journal of Political Economy 107(S6): S158–S183.

Becker

(1964) Human Capital. New York: Columbia University Press.

Blood

Wolfe

(1960) Husbands and Wives. New York: Free Press.

Booth

Kee

(2005) Birth order matters. Discussion Paper No. 506, the Australian National University Centre for Economic Policy Research.

Bradbury

(2007) Why do children of young mothers have poorer outcomes? In: Longitudinal Study of Australian Children Conference, Melbourne, Australia, 3–4 December 2007. Available at: http://flosse.fahcsia.gov.au/fahcsiajspui/handle/10620/2577.

Breen

Jonsson

(2005) Inequality of opportunity in comparative perspective. Annual Review of Sociology 31: 223–243.

Breen

Luijkx

Muller

Pollak

(2010) Long-term trends in educational inequality in Europe. European Sociological Review 26(1): 31–48.

10.

Buchmann

DiPrete

McDaniel

(2008) Gender inequalities in education. Annual Review of Sociology 34: 319–337.

11.

Caldwell

(1982) Theory of Fertility Decline. New York: Academic Press.

12.

Case

(2001) The primacy of education. WWS Working Paper 203. Princeton, NJ: Princeton University.

13.

Chiswick

DebBurman

(2004) Pre-school enrollment: An analysis by immigrant generation. CReAM Discussion Paper No. 04/04, University College, London.

14.

Colclough

Rose

Tembon

(2000) Gender inequalities in primary schooling. International Journal of Educational Development 20(1): 5–27.

15.

Coleman

Campbell

Hobson

McPartland

Mood

Weinfeld

York

(1966) Equality of Educational Opportunity. Washington, D.C.: U.S. Government Printing Office.

16.

Dayıoğlu

Kirdar

Tansel

(2009) Impact of sibship size, birth order and sex composition on school enrolment in urban Turkey. Oxford Bulletin of Economics and Statistics 71(3): 399–426.

17.

EFA (2011) The hidden crisis: Armed conflict and education. Education For All Global Monitoring Report, 2011. Paris: UNESCO Publishing & Oxford University Press.

18.

Ejrnaes

Portner

(2004) Birth order and the intra-household allocation of time and education. Review of Economics and Statistics 86(4): 1008–1019.

19.

Evangelista de Carvalho Filho I (2008) Household income as a determinant of child labor and school enrollment in Brazil. 203, WP/08/241. Washington, DC: International Monetary Fund.

20.

Fafchamps

Wahba

(2006) Child labor, urban proximity, and household composition. Journal of Development Economics 79(2): 374–397.

21.

Filmer

Pritchett

(1999) The effect of household wealth on educational attainment: Evidence from 35 countries. Population and Development Review 25(1): 85–120.

22.

Glewwe

Jacoby

(2004) Economic growth and the demand for education. Journal of Development Economics 74(1): 33–51.

23.

Goldstein

(2011) Multilevel Statistical Models. Chichester: Wiley.

24.

Gündüz-Hoşgör

Smits

(2003) Linguistic capital: Language as a socio-economic resource among Kurdish and Arabic women in Turkey. Ethnic and Racial Studies 26(5): 829–853.

25.

Gündüz-Hoşgör

Smits

(2007) The status of rural women in Turkey: What is the role of regional differences. In: Moghadam

(ed.) Empowering Women: Participation, Rights and Women's Movements in the Middle East, North Africa, and South Asia. Syracuse, NY: Syracuse University Press, 180–202.

26.

Gündüz-Hoşgör

Smits

(2008) Variation in labor market participation of married women in Turkey. Women's Studies International Forum 31(2): 104–117.

27.

Hazarika

(2001) The sensitivity of primary school enrollment to the cost of post-primary schooling in rural Pakistan: A gender perspective. Education Economics 9(3): 237–244.

28.

Huisman

Smits

(2009) Effects of household- and district-level factors on primary school enrollment in 30 developing countries. World Development 37(1): 179–193.

29.

Huntington

(1996) The Clash of Civilizations and the Remaking of World Order. New York: Simon and Schuster.

30.

Inglehart

Norris

(2003) The true clash of civilizations. Foreign Policy Magazine 34(2): 67–74.

31.

Jencks

(1972) Inequality: A Reassessment of the Effect of Family and Schooling in America. New York: Basic Books.

32.

Kandiyoti

(1988) Bargaining with patriarchy. Gender and Society 2(3): 274−289.

33.

King

Anne Hill

(1993) Women’s Education in Developing Countries. Baltimore, MD: Johns Hopkins University Press.

34.

Kırdar

Dayıoğlu

Tansel

(2007) Impact of sibship size, birth order and sex composition on school enrollment in urban Turkey. MPRA Paper No. 2755. Munich: Munich Personal RePEc Archive.

35.

Lee

S-H

Mason

(2005) Mother’s education, learning-by-doing, and child health care in rural India. Comparative Education Review 49(4): 534–551.

36.

Zhang

Zhu

(2008) The quantity–quality trade-off of children in a developing country. Demography 45(1): 223–243.

37.

McDaniel

(2010) Cross-national gender gaps in educational expectations. Comparative Education Review 54(1): 27–50.

38.

Mingat

(2007) Social disparities in education in Sub-Saharan African countries. In: Teese

Lamb

Duru-Bellat

(eds) International Studies in Educational Inequality, Theory and Policy, vol. 1. Dordrecht: Springer, 223–255.

39.

Moghadam

(1993) Modernizing Women: Gender and Social Change in the Middle East. London: Lynne Rienner Publishers.

40.

Moghadam

(2004) Patriarchy in transition: Women and the changing family in the Middle East. Journal of Comparative Family Studies 35(2): 137−162.

41.

Moghadam

Decker

(2010) Social change in the Middle East. In: Lust

(ed.) The Middle East. Washington, D.C.: CQ Press, 65–98.

42.

Moors

(1999) Debating Islamic family law. In: Meriwether

Tucker

(eds) The Social History of Women and Gender in the Modern Middle East. Boulder, CO: Westview, 141–177.

43.

Mugisha

(2006) School enrollment among urban non-slum, slum and rural children in Kenya. International Journal of Educational Development 26(5): 471–482.

44.

Rodman

(1972) Marital power and the theory of resources in cultural context. Journal of Comparative Family Studies 3(1): 50–57.

45.

Sen

(1999) Development as Freedom. Oxford: Oxford University Press.

46.

Shavit

Blossfeld

(1993) Persistent Inequality: Changing Educational Attainment in Thirteen Countries. Boulder, CO: Westview.

47.

Smith

Haddad

(2000) Overcoming child malnutrition in developing countries. 2020 Vision Discussion Paper 30. Washington, D.C.: IFPRI.

48.

Smits

Gündüz-Hoşgör

(2006) Effects of family background characteristics on educational participation in Turkey. International Journal of Educational Development 26(5): 545–560.

49.

Snijders

Bosker

(1999) Multilevel Analysis. London: Sage.

50.

Solon

Page

Duncan

(2000) Correlations between neighboring children in their subsequent educational attainment. Review of Economics and Statistics 82(3): 383–392.

51.

Spierings

Smits

Verloo

(2009) On the compatibility of Islam and gender equality: Effects of modernization, state islamization, and democracy on women's labor market participation in 45 Muslim countries. Social Indicators Research 90(3): 503–521.

52.

Spierings

Smits

Verloo

(2010) Micro- and macrolevel determinants of women’s employment in six Arab countries. Journal of Marriage and the Family 72(5): 1391–1407.

53.

Tansel

(2002) Determinants of school attainment of boys and girls in Turkey. Economics of Education Review 21(5): 455–470.

54.

UNDP (2002, 2003, 2006, 2009) The Arab Human Development Reports 2002, 2003, 2005, 2009. United Nations Publications.

55.

UNDP (2012) International Human Development Indicators. Available at: http://hdrstats.undp.org (accessed 10 May 2012).

56.

UNESCO (2012) UIS data tables. Available at: http://uis.unesco.org (accessed 10 May 2012).

57.

Webbink

Smits

De Jong

(2012) Hidden child labour: Determinants of housework and family business work of children in 16 developing countries. World Development 40(3): 631–642.

58.

Wichman

Rodgers

MacCallum

(2006) A multilevel approach to the relationship between birth order and intelligence. Personality and Social Psychology Bulletin 32(1): 117–127.

59.

World Bank (2006) World Development Report 2007. Washington, D.C.: The World Bank.