A Teenager in Love: Multidimensional Human Capital and Teenage Pregnancy in Ghana

Abstract

I examine teenage pregnancy in Ghana, focusing on the role and interplay of Ghanaian and English reading skills, formal educational attainment, and adult literacy programme participation. Pursuing several alternative identification strategies three main results are established. First, I confirm the finding from previous studies that educational attainment is negatively related to teenage pregnancy. Second, however, once Ghanaian and English reading skills are introduced, the association between educational attainment and teenage pregnancy decreases or disappears altogether. Third, for the girls who have not completed primary school, adult literacy programme participation is associated with a much lower probability of experiencing a teenage pregnancy.

A bright future is the best contraceptive.

–Marian Wright Edelman

Acknowledgements

I thank David Bishai, Marco Gonzalez-Navarro, Anthony Keats, Melanie Khamis, Nishith Prakash, Claus Pörtner, Howard White, and participants at the Danish Academic Economists in North America Annual Meetings; the Economic Development in Africa conference (Centre for the Study of African Economies, Oxford), the Midwest International Economic Development Conference, the Population Association of America Annual Meetings, Southern Economic Association Annual Meetings, and seminar participants at Aarhus University, Ohio University, Wesleyan University, and Seattle University for helpful comments and suggestions. The comments and suggestions from two anonymous reviewers helped greatly improve this paper and are much appreciated. I gratefully acknowledge financial support from Washington and Lee’s Lenfest Summer Research Grant. Part of this research was carried out while visiting Princeton University’s Economics Department; the hospitality of the Department is gratefully acknowledged. The data were kindly provided by the Ghana Statistical Service. The findings and interpretations, however, are those of the author and should not be attributed to the Ghana Statistical Service. The data is proprietary to the Ghana Stata Statistical Service but can be obtained from the GSS website at http://www.statsghana.gov.gh/DataRequest.html. The Stata code used for the analysis in this paper is available upon request.

Disclosure statement

No potential conflict of interest was reported by the author.

Supplemental Material

Supplementary Material document is available at https://doi.org/10.1080/00220388.2017.1308486

Notes

1. Although it turns out that ultimately this does not appear necessary for the application considered here.

2. A caveat here is that a substantial literature exists on the related notions of ‘age at first birth’ or ‘first-birth timing’, though this literature seems to be predominantly Asian (for example, Basu, 1993; Bloom & Reddy, 1986; Hirschman & Rindfuss, 1980) – so that explicitly examining teenage pregnancy, per se, and especially doing so for the sub-Saharan African country of Ghana, is still a contribution to the literature in this dimension.

3. Similarly, several key reproductive health statistics in Ghana are comparatively low – including median age at first intercourse (18.4 years for women and 20.0 years for men), at first birth (20.7 years), and at first marriage (19.8 years for women and 25.9 for men) (Ghana Statistical Service [GSS], Ghana Health Service [GHS], and ICF Macro, 2009).

4. See Blunch and Pörtner (2005), Appendix B (whereupon much of the following discussion is also based) for additional details.

5. See, for example, Abadzi (1994) and Ortega and Rodríguez (2008).

6. Topics include family planning, teenage pregnancy, environmental hygiene, immunisation, HIV/AIDS, safe motherhood and child care, drug abuse, traditional medicine, and safe drinking water.

7. Topics include cocoa farming, maize cultivation, dry season farming, basket weaving, animal husbandry, bee-keeping, oil palm cultivation, borrowing money for work, hygienic way of preserving and selling fish, farm extension services, pottery, and soap making.

8. Topics include taxation, bushfires, interstate succession law, child labour, chieftaincy, community empowerment, and expensive funerals.

9. See Blunch (2013) for a cost-benefit analysis of adult literacy programmes in Ghana in the context of child mortality outcomes.

10. Again, most likely obtained from the formal education system and only to a lesser degree from participating in adult literacy programmes.

11. Descriptive statistics for the analysis samples are reported in Table 1.

12. Eight girls report having completed ‘other education’. These are dropped since it is not clear what ‘other education’ is.

13. In addition to having been born in one of the 10 different regions in Ghana, the variable also allows for being born abroad (three categories: Other ECOWAS, Other African [than ECOWAS], Outside Africa).

14. Though it is of course possible that individuals have attended the adult literacy programme in another region or community – though the inclusion of region of birth fixed effects in addition to either region of residence or cluster of residence fixed effects seem to address this to some extent, at least at the regional level.

15. One of the seminal papers here is Becker (1960).

16. Specifically examining the case of mother’s education and child health – though the concept translates to any type of health related behaviour.

17. I have witnessed this myself, on numerous occasions in both urban and rural areas in Ghana.

18. At a minimum, if these factors are not included, one may systematically over- or under-estimate the strength of the human capital-teenage pregnancy relationship.

19. As is well known, there may be some concern about using OLS, or, in effect, the linear probability model (LPM), when the dependent variable is binary. For example, predicted probabilities may fall outside the (0,1)-range and heteroskedasticity also is present by default. However, it can be argued that the LPM approximates the response probability well, especially if (1) the main purpose is to estimate the partial effect of a given regressor on the response probability, averaged across the distribution of the other regressors, (2) most of the regressors are discrete and take on only a few values and/or (3) heteroskedasticity-robust standard errors are used in place of regular standard errors (Wooldridge, 2010). All three factors seem to work in favour of the LPM for the purposes of the application here.

20. Due to space constraints, as well as since this strategy ultimately turns out not to be warranted for the analysis here (as revealed by the results from IV/2SLS specification tests), the details are omitted here (but are available upon request).

21. The variables used for the matching include all explanatory variables from the previous regressions except the possibly endogenous explanatory variables (formal educational attainment, adult literacy programme participation, and [English and Ghanaian] reading skills).

22. For details on Mahalanobis matching, see for example Rosenbaum and Rubin (1985).

23. The full set of results is shown in Tables A4 and A5 in the Supplementary Materials.

24. It should be remembered that if the instruments are valid then these tests will pass with high probability; therefore, the tests are necessary but not sufficient for the instruments to be valid.

25 Due to space constraints the results tables are not shown here but they are available upon request.

26. For both types of education treatments the variables used for the matching include all explanatory variables from the previous regressions except the possibly endogenous explanatory variables (formal educational attainment, adult literacy programme participation, and [English and Ghanaian] reading skills).

27. Due to space constraints the results tables for the sensitivity analyses are not shown here but they are available upon request.

Additional information

Funding

This work was supported by Washington and Lee University’s Lenfest Summer Research Grant. [NA].