The effect of the Universal Primary Education program on consumption and on the employment sector: Evidence from Tanzania
Introduction
Over the last decades, policy-makers have put education at the top of their agenda, and several governments of developing countries have implemented policies to universalize primary education. Yet, while the literature on the returns to human capital is extensive, little evidence exists on developing countries where the agricultural sector prevails (see Card (2001) for a literature review).
So far, Duflo, 2001, Maluccio, 1998 have examined the causal effect of education on earnings in developing countries, but both authors restrict their analysis to wage earners who constitute a very selected sample, preventing the extrapolation of the results to a broader population. To get around this issue, Duflo (2001) imputes a wage for self-employed workers, but this approach cannot be applied in agriculture-based countries, characterized by a small formal economic sector. In the nigerian context, Oyelere (2010) deals with the endogenous nature of education, but the author also limits her analysis to income earners.
Another strand of the literature estimates the returns to education among agricultural households by considering the agricultural production. Lockheed, Jamison, and Lau (1980) review papers estimating the impact of education on agricultural production and find very mixed results depending on the country and the specification of education. However, these studies do not consider the endogeneity of education and the estimations were prone to bias.
Using the extensive change in the access to education induced by the Tanzanian Universal Primary Education (UPE) program, this paper contributes to this existing literature by investigating the returns to education for the entire population. To include farmers and non-agricultural self-employed workers in the analysis, and to be able to compare the returns to education across the different sectors of activity, I estimate the effect of education of household heads on the predicted level of consumption of households. This predicted consumption is constructed by adopting a two-sample procedure from census data and household survey matching (Elbers et al., 2003, Tarozzi and Deaton, 2009).
The second contribution of this paper is to address the effect of education on the labor market participation, and more precisely, on the choice of the sector of activity. Education might not only increase earnings but also provide access to better-paid activities. Duflo et al., 2017, Ozier, 2015 investigate this question in African contexts, by focusing on secondary education. Duflo et al. (2017) find that an increase in secondary education in Ghana raises the probability of having a formal employment contract. Similarly, Ozier (2015) shows that in Kenya, children who complete secondary school are more likely to have a formal employment, but the effect is not always robust depending on the specification.
These questions are of public interest because even though the structural transformation is underway in rural Africa, agriculture remains the main source of employment, while agricultural productivity is still low and exposes population to poverty. The majority of field experiments in agriculture show that this low productivity is explained by a lack of information which prevents farmers from exploiting new technologies (De Janvry, Sadoulet, & Suri, 2017). Even when the information is available, farmers may fail to capture every aspect of the information (Duflo and Pande, 2007, Hanna et al., 2014). Social networks turn out to be useful in transmitting knowledge, but the available information is often incomplete (Cai, De Janvry, & Sadoulet, 2015), and the effectiveness is very context specific. Thus, this raises question of the relevance of investing in education. Is education, whose very essence is to transmit knowledge, effective in transmitting skills and in increasing household earnings in a rural environment ?
To assess the causal effect of education, I exploit the UPE program as a natural experiment. This program, implemented in Tanzania from 1974 to 1978, aimed at reducing disparities in access to education. In 1974, educational levels were low at the national level, with wide variations across regions arising from the activity during the colonial period. To lower these inequalities, the Tanzanian socialist government gave priority to disadvantaged areas, which led the latter to experience higher schooling expansion. The results were substantial: 3.3 million children aged 7 to 13 were enrolled in 1980, compared to 1.2 million in 1974 (Bonini, 2003). This achievement was made possible by the construction of new schools and by the villagization process which removed regional disparities of access to school by bringing together scattered homestead into community villages. The particularity of this UPE program was also to stand out against the elitist colonial education by providing a curriculum geared towards skills that are relevant for rural communities. In this respect, practical classes such as agricultural classes were incorporated.
The exposure to the UPE program of an individual depends on his age and on the regional educational attainment before the program was launched. On one side, the age informs whether the individual was young enough to go to school at the time of the program, and delineates the pre-treatment and the treatment age-cohorts. On the other side, the regional schooling supply illustrates the efforts that the government has made to remove disparities in access to schools. After controlling for region of birth and year of birth, the identification strategy combines information on the initial regional education level with information on the age of individuals to create a measure of the exposure to UPE program and to instrument education. A large literature exploits a similar identification strategy to study the effects of Free Primary Education programs on the education attainment (Al-Samarrai and Zaman, 2007, Hoogeveen and Rossi, 2013, Grogan, 2009), on the quality of education (Lucas & Mbiti, 2012), and on financial outcomes (Ajayi & Ross, 2018).
The first-stage results show that the education expansion increases with the exposure to the UPE program. This difference-in-difference estimate represents the effect of the UPE program on schooling, if there is no omitted regional time-varying characteristics correlated with the intensity of the UPE program.
To test the validity of the instrument, I attempt to identify the potential sources that violate the exclusion restriction assumption and I provide robustness checks. A major concern is that regions exposed to the program have specific characteristics that influence the education expansion. To account for these structural differences that can interfere with education, I control by a set of regional characteristics interacted with a time trend. The estimates may also be explained by a convergence phenomenon. To address this issue, I check the pre-trend to test whether the education expansion was higher in less educated regions than in the more educated ones before and after the introduction of the UPE program. Finally, I examine and discuss whether other programs occurring at the same time that the UPE program may be responsible of the results.
The 2SLS estimates show that the returns to education for the entire Tanzanian population are about 6 percent, which is close to what it is found for wage-workers in developing countries (Duflo, 2001, Maluccio, 1998). When I compare the returns to education across sectors of activity, I find that they are positive in every sector, including agriculture. I also find evidence that education increases the probability of working in farm activities for women. I explain this finding by the design of the program, whose aim was to provide agricultural classes and other practical classes in the curriculum. This paper demonstrates the importance of education and of the choice of the curriculum in helping households to increase their level of consumption in predominantly rural countries.
The remainder of the paper is organized as follows: Section 2 provides a broad picture of the Tanzanian context and describes the data as the main variables of the analysis. Section 3 outlines the estimation strategy and the effect of the UPE program on education. The results are introduced in Section 4 while Section 5 discusses the robustness of the results and Section 6 concludes.
Section snippets
Historical background and the UPE program
When colonization ended in 1961, access to education in Tanzania was very unequal between regions (Court & Kinyanjui, 1980).1 At this time, the purpose of primary education was to prepare for secondary education and to
Identification strategy
The instrumental variable approach that I use exploits the exposure to the UPE program. This assumes that being exposed to the program increases the probability of being enrolled in school but is orthogonal to unobserved household characteristics that determine labor market outcomes. To capture exposure to the UPE program, I adopt a difference-in-differences strategy based on variations in time and in space. It consists of comparing pretreatment cohorts () with treated cohorts () for
The results
This section presents the main results. The first subsection is devoted to the returns to education for the entire population, and by sectors of activity. Education may also have the benefit of increasing the probability to work in sectors that are better paid. Then, subSection 4.2 investigates whether education changes the labor distribution between the sectors of activity.
Discussion and robustness checks
To test whether 2SLS estimates are unbiased, I implement a series of robustness checks.
First, I test whether the introduction of individuals who were likely to be partially affected by the UPE program (individuals who were 13 years old between 1968 and 1974) change the results, but I do not find any significant difference (column 3 and 4 of Table A7).
Regarding the identification strategy, one of the main concerns is that the exclusion restriction is not satisfied. The villagization process,
Conclusion
This paper studies the benefits of education in Tanzania and considers two particular dimensions: household consumption and the choice of the sectors of activity. To address endogeneity issues, I instrument education of household heads by exploiting variation in time and in space of the exposition to the Universal Education Program.
I find that this massive primary education program contributed to a reduction in inequalities among regions. After this program ended, its effects persisted for the
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
I wish to thank Pierre André, Paul Glewwe, Flore Gubert, Arnaud Lefranc, William Parienté, Christopher Udry, Gonzague Vannoorenberghe, Philippe De Vreyer for their useful comments, as well as the participants at the following conferences and seminars: NBER Transforming Rural Africa, CSAE, SMYE, CFDS, and IRES.
References (41)
- et al.
Field experiments in developing country agriculture
The impact of education in rural ghana: Examining household labor allocation and returns on and off the farm
Journal of Development Economics
(2004)Quasi-maximum likelihood estimation and testing for nonlinear models with endogenous explanatory variables
Journal of Econometrics
(2014)- et al.
The effects of education on financial outcomes: Evidence from kenya
Economic Development and Cultural Change
(2018) - et al.
Abolishing school fees in malawi: the impact on education access and equity
Education Economics
(2007) - Appleton, S., & Balihuta, A., et al. (1996). Education and agricultural productivity: Evidence from uganda, University...
- et al.
Inference on treatment effects after selection among high-dimensional controls
The Review of Economic Studies
(2014) - et al.
How much should we trust differences-in-differences estimates?
The Quarterly Journal of Economics
(2004) Un siècle d’éducation scolaire en tanzanie
Cahiers d’études africaines
(2003)- et al.
Social networks and the decision to insure
American Economic Journal Applied Economics
(2015)
Bootstrap-based improvements for inference with clustered errors
The Review of Economics and Statistics
Estimating the return to schooling: Progress on some persistent econometric problems
Econometrica
Development policy and educational opportunity: the experience of kenya and tanzania, occasional Paper 33
Schooling and labor market consequences of school construction in indonesia: Evidence from an unusual policy experiment
American Economic Review
The impact of free secondary education: Experimental evidence from ghana
Micro-level estimation of poverty and inequality
Econometrica
Cited by (4)
Mothers and fathers: education, co-residence, and child health
2023, Journal of Population EconomicsBenefits of enhanced access to education in Tanzania
2023, Development Policy ReviewIs wealth found in the soil or in the brain? Investing in farm people in Malawi
2023, Review of Development EconomicsCommunity livelihood vulnerability to drought in semi-arid areas of northern Cameroon
2022, Discover Sustainability