The Impact of Educational Homogamy on Isolated Illiteracy Levels

The United Nations Educational, Scientific and Cultural Organization (UNESCO) defines the literacy rate in a given population as the percentage of adults (aged 15 and older) who are literate in that population.

In this context, an isolated illiterate will be an illiterate individual who marries another illiterate. Analyzing the literacy status of the other household members is an interesting but more complex topic that we will address in future research.

Because we aim to investigate the effect that homogamy levels have had on isolated illiteracy rates, it is natural that we focus exclusively on the population in union. However, one might wonder whether illiterates are discouraged from entering into the marriage market because of their lack of human capital, an issue that could eventually bias our results. In this respect, in 80 % of our samples, the percentage of illiterates who are in union (IU) is higher than the proportion of literates who are in union (LU). The average difference between IU and LU in those samples equals 8 percentage points. In the populations where the proportion LU is greater than the proportion IU, the corresponding literacy rates are higher than .95. For the samples with literacy rates above .95, the average difference between LU and IU equals 2 percentage points.

Recall that in this article, the partitioning of the education distribution into two groups (literates and illiterates) is very crude. Even if the educational distribution partition on which these notions are built is typically finer—for instance, defined over four or more educational groups—the meaning of the terms is nonetheless preserved.

Interestingly, Q is very similar to a conceptually related efficiency-loss indicator defined by Subramanian (2004:458). Substituting Eqs. (1), (8), and (9) into Eq. (10), it is straightforward to check that Q can be expressed as min{a,d} / (min{a,d} + min{b,c}).

The results are as follows: Brazil’s literacy rate in 1991 was equal to 0.876 and 0.856 according to IPUMS and DHS, respectively; Colombia’s literacy rate in 2005 was equal to 0.970 and 0.973 according to IPUMS and DHS, respectively; India’s literacy rate in 1992 was equal to 0.54 according to DHS, and in 1993, it was equal to 0.52 according to IPUMS; Tanzania’s literacy rate in 2003 was equal to 0.83 according to DHS, and in 2002, it was equal to 0.81 according to IPUMS.

When we averaged the results for the 64 DHS where information on both criteria was available at the same time, only 1.5 % of those who never attended school are able to read a sentence. Alternatively, the percentage of those who ever attended school and who are unable to read a sentence ( ) takes relatively high values (around 25 %) only in the few countries with extremely low literacy rates (around 0.2). As literacy rates increase, the values of decrease sharply: for those countries with literacy rates above 0.5, the values of are below 4 %. Although these discrepancies could slightly bias the estimation of our (il)literacy indicators, they do not invalidate the overall trends we report and the message the trends convey.

Education and homogamy effects are likely to be biased where illiteracy levels are spatially clustered in some particularly regions. This may be the case of large countries, such as Brazil or India. Despite the levels of illiteracy that show great variation across regions in these two countries, the number of illiterates in each region strongly correlates with the total population by region. Besides, time trends indicate that spatial inequalities remain fairly stable over time.

This article does not address the difference between married and cohabiting couples. Our analyses are based on all identifiable unions, regardless of their nature. However, we are taking into account only heterosexual unions because homosexual unions cannot be treated with the analytical tools presented here. In addition, it is generally not possible to identify homosexual unions with most of the existing census and household surveys employed in this article.

If we plot the values of couples’ literacy rates (on the horizontal axis) against the corresponding gender gap in couples’ literacy rates (i.e., married men’s literacy rate minus married women’s literacy rate), we observe an inverted U-shaped curve crossing the horizontal axis at 0 and 1, with its maximum around 0.5. The results are not presented here but are available upon request.

Consider a hypothetical stylized setting where (1) the literacy rate for men is higher than that for women; (2) there is complete tendency toward educational homogamy; and (3) literate males may, from necessity, accept illiterate women in union but not vice versa. Under conditions (1), (2), and (3), it can be shown that parameters a, b, c, and d from Table 1 would equal 1 – R _m , 0, R _m – R _w , and R _w , respectively (where R _m and R _w are, respectively, men’s and women’s literacy rates). Therefore, because a = 1 – R _m , one has that , so isolated illiteracy rates would be insensitive to marginal increases in women’s literacy rates. Even if assumptions (1), (2), and (3) only approximately match with real-world observations, the analytical result that is derived from them helps to explain why the initial stages of women’s education expansion are not accompanied by drastic reductions in isolated illiteracy rates (at least in the short run), as is suggested in Fig. 4.