Languages, Human Capital, and Well-Being in Sub-Saharan Africa

Similarly to the existing literature (Buzasi 2016; Lopes 1998), in this chapter language situation is generally applied as an “umbrella” concept that refers to all linguistic characteristics of a society including ethnolinguistic diversity, the prevalence of multilingualism, geographical distribution of languages as first and second, the legal and social status of languages, and the share of the population speaking officially recognized languages. However, our empirical analysis focuses only on four aspects, namely, linguistic diversity, the share of population speaking the former colonizer’s language, the share of population speaking the largest local language, and the intensity of local language use in education (ILLED).

The general description of the language situation in Sub-Saharan Africa is largely based on qualitative and quantitative information from Ethnologue (Lewis et al. 2015) and Albaugh (2014).

See, e.g., Verbeek (2008) for a short and accessible discussion of these techniques.

The status of languages is a multidimensional concept which refers to the official recognition as well as the social and communication importance of languages. The most complex language status indicator is the language vitality index designed by Lewis and Simons (2010) and published in Lewis et al. (2015). The vitality of languages is evaluated along five dimensions: identity function, official recognition, intergenerational transmission, literacy status (standardized orthography), and use among younger generations.

For surveys on human capital measurement, see Wößmann (2003) and Becker (2007).

The linguistic distance is based on the linguistic tree diagram of Ethnologue (Lewis et al. 2014) and is computed as proposed by Fearon (2003).

Although the majority of the literature argues that ethnolinguistic diversity has negative societal consequences, some studies find that the effect of diversity on socioeconomic outcomes is insignificant (Gerring et al. 2015), not convincing (Fish and Brooks 2004), even positive (Arcand and Grin 2013), or dependent on other factors (Easterly 2001; Collier 2000).

For a detailed description of the types and evolution of well-being measures, see Fleurbaey and Blanchet (2013: 1–8).

A major disadvantage of HDI, often neglected in the literature, is that it assigns arbitrary weights to its components. Bear in mind that the components of HDI (education, life expectancy, and per capita income) all contain the latent well-being factors and hence they each have a common piece of information (often referred to as commonality). This is the reason why we find a moderate to strong correlation between them. Aggregating them with equal weight necessarily means that the same factor is taken into account multiple times (double counting). The resulting variable will have higher variance than the latent well-being factor that it is supposed to represent.

Positive skew means that most individuals have lower than average income or in other words, low incomes are common, and high incomes are rare.

See Weil (2007).

The literacy rate reported by UNESCO is understood as the share of the population aged 15 or above who can, with understanding, read and write a short simple sentence on their everyday life. However, UNESCO data are based on various sources such as population censuses and the Demographic and Health Surveys (DHSs) which differ in terms of literacy measurement. While reading abilities of those with unfinished primary schooling are actually tested in the DHSs, literacy data in population censuses are usually based on self-report.

Standardization is the rescaling of a variable so that it has zero mean and unit standard deviation.

Communality measures the share of observed variance of a variable that is explained by the extracted factors. Since factors are common among all variables, this measure can be seen as an indicator of the fit of the underlying factor model.

Alesina et al. (2003) utilize the Encyclopedia Britannica 2001 that reports the share of languages spoken as mother tongue, generally based on national census data. Issues related to using national censuses and other surveys to describe a country’s linguistic situation as well as the discrepancy between linguistic terms used in linguistics and in national surveys are discussed in Buzasi (2016).

The proportion of local languages used in education can assume 0 (no local language used), 1 (classical Arabic), 2 (single minority language), 3 (few languages or one major language), 4 (several languages or one dominant language), and 5 (most languages or one overwhelmingly dominant language). The extent of local language use in primary education can assume 1 (experimental), 1.2 (moderate), 1.5 (extensive), 1.8 (generalized), and 2 (exclusive).

The linear correlation coefficient would require that the variables are at least on an interval scale, that is, an increase in the value from 0 to 1 and from 1 to 2 represents the same degree of change. This is obviously not true for the variable ILLED, since all we know is that, say, the value 2 represents a higher degree of local language use in education than the value of 1 or 0, but we cannot argue that the difference between the categories 1 and 0 is the same as between the categories 2 and 1. All we can assume is that the order of the assigned values are monotonic and increasing.

Gabon seems to behave as an outlier in the graphs, but removing it from our sample does not significantly change the results.

Sub-Saharan African countries in the sample are Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Central African Republic, Chad, Comoros, Democratic Republic of Congo, Equatorial Guinea, Eritrea, Ethiopia, Gabon, Ghana, Guinea, Guinea-Bissau, Ivory Coast, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Nigeria, Republic of the Congo, Rwanda, Sao Tome and Principe, Senegal, Seychelles, Sierra Leone, South Africa, Sudan, Swaziland, Tanzania, the Gambia, Togo, Uganda, Zambia, and Zimbabwe. Important data are not available for Djibouti, Somalia, and South Sudan.

Simultaneity means that the causality runs both ways between two variables. For example, more education may lead to higher income, but higher income may also result in a higher educational attainment.

Gross capital formation is the total value of investments in physical capital stock (equipment, machines, buildings, etc.). It consists of two components: one is the depreciation (the loss of capital stock as a result of wear) and the net change in capital stock (net capital formation).

The explanatory variable ILLED_1990 is treated as a variable measured on an interval scale. This common technique used in social sciences is chosen to limit the number of explanatory variables relative to the sample size.

Kamarck (1976) proposed that a country’s geographical location has a significant effect on its ability to develop. The tropics is characterized with climatic and biological conditions such as erratic patterns of rainfalls, soil of poor quality, lack of mineral resources, and a variety of weeds, fungi, insects, and other microbes, which affect crops and human life negatively. Due to the aforementioned factors, labor productivity and human and nonhuman capital accumulation tend to be low.

Note that the latent human capital variable is standardized and GNI per capita is used in the logarithmic form.

Although ILLED is a useful indicator and based on a thorough data collecting procedure, it has some shortcomings from the aspects of our study. The scores assigned according to the two pillars are quite arbitrary. For instance, score 2 in the first pillar is given to countries where there is only one (minority) language recognized in education covering up to 50% of the society. Score 3 is given if there are several minority languages or one dominant language is acknowledged covering between 50% and 70% of the population. For our research, a continuous or a more detailed measure would be more adequate.