Life expectancy and economic development: Evidence from microdata

• Published date: 01 August 2020
• Date: 01 August 2020
• Author: Belgi Turan
• DOI: http://doi.org/10.1111/rode.12665

Rev Dev Econ. 2020;24:949–972. wileyonlinelibrary.com/journal/rode

|

949

© 2020 John Wiley & Sons Ltd

Received: 3 August 2017

|

Revised: 16 March 2020

|

Accepted: 18 March 2020

DOI: 10.1111/rode.12665

REGULAR ARTICLE

Life expectancy and economic development:

Evidence from microdata

BelgiTuran

TOBB University of Economics and

Technology, Sogutozu, Ankara, Turkey

Correspondence

Belgi Turan, TOBB University of

Economics and Technology, Sogutozu,

Ankara, Turkey.

Email: belgituran@etu.edu.tr

Abstract

This study examines the effect of life expectancy on fer-

tility, education, and labor force participation. Using birth

and sibling histories from the Demographic Health Surveys

conducted in sub-Saharan Africa, I construct a time series

of age-specific birth rates and mortality rates at the country-

region level. I use these data to test the implications of a

general equilibrium model linking life expectancy to fertil-

ity, human capital, and labor supply. My results suggest that

increases in life expectancy reduce fertility, increase educa-

tion, and increase labor force participation. Overall, my em-

pirical results suggest that in sub-Saharan Africa, increases

in life expectancy will have a positive impact on growth

through fertility, education, and labor supply but that the

effect will be small. My results also rule out the possibility

that recent shocks to adult mortality in high HIV prevalence

countries will reduce fertility, increase labor productivity,

and lead to faster growth.

KEYWORDS

economic development, education, fertility, labor supply, life expectancy,

HIV/AIDS

JEL CLASSIFICATION

I25; J13; J22; O12

950

|

TURAN

1

|

INTRODUCTION

A large body of literature studies the effect of improving health and life expectancy on economic de-

velopment and growth. However, there is thus far little consensus on either the theoretical or the em-

pirical front. On the theory side, standard neoclassical model highlights the limits of improvements in

health and life expectancy. Increases in life expectancy increase the population, which reduces capital/

labor ratios and lowers per capita income. On the contrary, endogenous growth models in the tradition

of Becker and Barro (1988) suggest that human capital investment and fertility responses may offset

the pessimistic predictions of the neoclassical model. In particular, increased life expectancy could

lead to a quantity/quality trade-off where parents have fewer children but invest more in the education

of their children as well as their own education (e.g., Cervellati & Sunde, 2007; Kalemli-Ozcan, 2003;

Soares, 2005; Tamura, 2006). This finding suggests that behavioral responses in fertility and human

capital investment could offset the decline in capital/labor ratios and productivity.

Likewise, little agreement on the empirical front exists. While Bloom and Sachs (1998), Gallup,

Sachs, and Mellinger (1999), Bloom, Canning, and Sevilla (2002), and Lorentzen, McMillan, and

Wacziarg (2008) find large effects of increasing life expectancy on growth, Acemoglu and Johnson

(2007) find little effect. They instrument changes in life expectancy with dates of global health in-

terventions to combat 15 major diseases. Nonetheless, while health interventions increase life expec-

tancy, they find little impact on per capita GDP.1

In this study, I adopt a general equilibrium model linking life expectancy to important behavioral

variables such as fertility, human capital investment, and labor supply. I extend the models introduced

by Zhang and Zhang (2005), who model education and fertility but not labor supply, and Boucekkine,

Desbordes, and Latzer (2009), who model fertility and labor supply but not education. My model

incorporates all three factors, thereby considering both the partial and the general equilibrium ef-

fects. The model has two offsetting effects. On the one hand, an increase in life expectancy leads to a

quantity/quality trade-off. Individuals also work and save more since they have a higher probability

of surviving to old age. This is called the horizon effect. On the other hand, there is a countervailing

effect in which increased life expectancy expands the size of the adult population and lowers wages.

Under certain parameter configurations, this general equilibrium effect, which works through lower

wages, can increase fertility (since the opportunity cost of bearing and rearing children is lower) and

reduce education and labor supply, leading to overall ambiguous effects.

The ambiguous theoretical predictions suggest that a thorough understanding of the link between

life expectancy, on the one side, and fertility, education, and labor supply, on the other side, rests on

empirical work. However, empirical work has thus far been limited by a lack of data. Researchers

in this area have largely relied on country-level data (Acemoglu & Johnson, 2007; Lorentzen et al.,

2008; Shastry & Weil, 2003).2

Empirical research has also largely focused on per capita GDP as the

outcome variable.3

In this study, I bridge this gap by constructing a new panel data set from the Demographic Health

Surveys (DHSs) and testing the implications of endogenous growth model that relates life expectancy

to fertility, human capital, and labor supply. More specifically, I use 107 DHSs of 29 countries in

sub-Saharan Africa taken between 1987 and 2017.4

I calculate fertility rates from birth histories and

adult mortality rates and life expectancy from sibling histories. Based on this information, I construct

birth and death rates by region5

and year. There are several advantages to these data. First, by building

mortality rates from sibling histories, I obtain more accurate measures of mortality and life expec-

tancy for certain countries. Country-level life expectancy measures for developing countries, espe-

cially sub-Saharan Africa, are often inaccurate since reliable vital registration data are not available.

Second, since the data are at the individual level, I can construct age-specific birth and mortality rates