The Effects of Rural Electrification on Employment:
New Evidence from South Africa
Taryn Dinkelman
∗
Princeton University
This version: August 2010 First version: November 2007
Abstract
This paper estimates the impact of electrification on employment growth by analyzing South
Africa’s mass roll-out of electricity to rural households. Using several new data sources and two
different identification strategies (an instrumental variables strategy and a fixed effects approach),
I find that electrification significantly raises female employment within 5 years. This new infras-
tructure appears to increase hours of work for men and women, while reducing female wages and
increasing male earnings. Several pieces of evidence suggest that household electrification raises
employment by releasing women from home production and enabling micro-enterprizes. Migration
behavior may also be affected.
∗
Email: tdinkelm@princeton.edu. Electrification data and institutional background and interviews pro-
vided by Jenny Barnard, Sheila Brown, Ed Bunge, Trevor Gaunt (University of Cape Town), Innocent Nxele,
Gisela Prasad (Energy Research Development Council at UCT), Jim Stephenson, Steven Tait and Amos
Zuma, in May 2006 and 2007. Conversations with and comments from Lori Beaman, Diether Beuermann,
Matias Busso, Anne Case, HwaJung Choi, John DiNardo, Ann Ferris, Raymond Fisman, Ben Keyes, David
Lam, Jim Levinsohn, Claudia Martinez-A, Justin McCrary, Nancy Qian, Dori Posel, Samara Potter-Gunter,
Mark Rosenzweig, Jeff Smith, Gary Solon, James Tybout, Rebecca Thornton and Martin Wittenberg sub-
stantially improved the paper. I also thank seminar participants at the following universities, organizations
and conferences for their valuable input: Brown, California-Los Angeles, California-Riverside Dartmouth,
Duke, the Harris School, the Harvard School of Public Health, the IDB, the International Institute of Eco-
nomic Studies at Stockholm University, the Kennedy School, KwaZulu-Natal, London School of Economics,
Michigan, Michigan State, the Microeconomics of Growth Network, Northwestern, North Carolina-Chapel
Hill, Penn State, Princeton, Resources for the Future, Syracuse, Washington-St Louis, Western Ontario,
Virginia and Yale as well as suggestions of the editor and referees. Data collection was funded through a
National Science Foundation Doctoral Dissertation Grant and an Eva Mueller New Directions in Demogra-
phy research award. I also acknowledge support from the Fogarty International Center of the U.S. National
Institutes of Health (D43TW000657) and the Population Studies Center at the University of Michigan. All
remaining errors are my own.
1
Electricity is pervasive in all industrialized countries and largely absent in the developing
world: about 1.6 billion people world-wide lack access to electricity (Saghir, 2005). Even
though many would consider electricity to be a “marker” for development, and despite
several historical episodes of wide-spread electrification in developed countries (for example,
the rural electrification of America in the 1930s), we know little about the direct effects that
new access to modern energy infrastructure will have on the process of development.
The primary objective of this paper is to analyze the impact of new access to modern
energy on an outcome of considerable interest: the ability of the poor to use their labor
resources for market production. In this paper, I estimate the causal impact of household
electrification on employment growth in rural communities by analyzing rural electrification
roll-out in post-apartheid South Africa. As a second objective, I investigate the mechanisms
through which this new infrastructure affects rural labor markets. Since energy infrastruc-
ture is likely to expand in poor areas over the next few decades
1
, this analysis provides
important lessons for many countries as well as for researchers studying the changing nature
of developing country labor markets.
The roll-out of grid infrastructure in South Africa provides a particularly good opportu-
nity to evaluate the effects of electrification on market employment. It was rapid, extended
into rural areas and targeted low capacity household use rather than industrial users (Gaunt,
2003). In 1993, a year before the end of apartheid, over two-thirds of South African house-
holds were without electricity and more than 80% relied on wood for home production.
2
Following the new government’s commitment to universal electrification, 2 million house-
holds, or almost one quarter of all households across the country, were newly connected to
the grid by 2001. This is twice as many households as the number of US farms connected
during the first five years of Roosevelt’s Rural Electrification Act (Beall, 1940).
Evaluating the effects of this electrification, or of any infrastructure roll-out, is not
straightforward. A large literature on the relationship between infrastructure and economic
growth acknowledges that infrastructure could be targeted towards growing areas, or to-
wards politically important areas.
3
Such selection biases any comparison of electrified and
1
World Bank commitments to energy infrastructure in Africa rose from $447 million in 2001 to $790
million in 2007. The World Bank’s Lighting Africa initiative aims to provide 250 million Africans with
modern sources of energy by 2030.
2
Charmes (2005) and Saghir (2005) document the time intensity of home production in developing coun-
tries. South Africans (mainly women) spend on average two working days per week in fuel-wood collection
(Budlender et al, 2002) and rural households spend an average of 3 hours per day on food preparation (own
calculations, 1997 October Household Survey).
3
The tradition in the macroeconomics literature has been to estimate the effects of public infrastructure
on total factor productivity using time-series data. Aschauer (1989) is a classic reference; see Canning (1998)
for cross-country evidence and Bogotic and Fedderke (2006) for South African evidence. The World Bank
(1994) and Jimenez (1995) provide good overviews of the infrastructure literature relevant for developing
2
non-electrified areas, and in unpredictable ways. Confounding trends in the economy make
it even more difficult to tease out the effects of infrastructure on any economic outcomes.
In this paper, I use two empirical strategies to identify the impact of electricity, taking
into account endogenous project placement and confounding economic trends. In the main
approach, I estimate community-level employment growth rates in communities that do and
do not receive an electricity project between 1996 and 2001, instrumenting for project place-
ment. To do this, I collect and match administrative data on roll-out in rural KwaZulu-Natal
(KZN) with geographical data and two Census surveys. I use land gradient to generate ex-
ogenous variation in electricity project allocation to communities. Higher gradient raises the
average cost of a household connection, making gradient an important factor in prioritizing
areas for electrification. I argue and provide evidence from a placebo experiment that in the
case of rural KZN, an area with poor agricultural prospects, gradient is unlikely to directly
affect employment outcomes, conditional on covariates.
As a complement to the main analysis, I use a fixed effects strategy to estimate the
impact of electrification on a richer set of labor market outcomes: employment, hours of
work, wages and earnings. For this analysis, I construct a four-period panel of magisterial
districts (agglomerations of communities) from cross-sectional household survey data in 1995,
1997, 1999 and 2001 and address non-random project placement and confounding economic
trends by directly controlling for magisterial district fixed effects and trends, estimating the
labor market effects of electrification using only within-district variation in electrification.
Results from both analyses show that employment in rural KZN increases in the wake of
electrification. Female employment in the Census rises by a significant 9.5 percentage points
in the IV results, which translates into 15,000 more women participating in the labor force, or
0.75 percent of the estimated 2 million new jobs created across the country over the period
(Casale and Posel, 2004). The fixed effects analysis using household survey data largely
supports these female employment results, although precise inference is more difficult with
the small samples in this dataset. Electrification increases work on the intensive margin for
women: in districts with the average increase in electrification over the period (15 percent),
women work about 8.9 more hours per week, a 3.5 percent increase. Under both analyses,
male employment rises (insignificantly) in electrifying areas, although to a lesser extent than
for females.
Having established that household electrification increases employment in rural commu-
nities, I turn to investigating mechanisms in the second part of the paper. I first explore the
impact of electrification on home production activities and find that newly electrified com-
munities experience substantial shifts away from using wood at home, and toward electric
countries.
3
cooking and lighting. This suggests that household electrification operates as a labor-saving
technology shock to home production in rural areas, releasing female time from home to
market work.
Second, I rule out the possibility that household electrification stimulated large scale
rural industrialization and hence a shift in labor demand by showing the absence of cross-
community employment spill-overs. As further evidence that electricity stimulated a net
increase in labor supply to the market, the fixed effects analysis indicates that female wages
fall (albeit imprecisely) in districts where electrification is expanding more rapidly. This fact
is difficult to reconcile with electricity causing large net increases in labor demand.
More plausibly, electricity may have lowered the cost of producing new, home-based
services for the market, thereby presenting individuals with alternative ways to use their
labor time in self employment and micro enterprizes. I am unable to provide direct evidence
on these mechanisms, but I argue that since employment results for men and women are not
statistically different from each other, it seems likely that the South African electrification
did not exclusively affect rural labor markets through the channel of freeing time from home
production. Rather, the reduced-form market employment results capture a combination of
increased labor supplied to the market (via the home production channel) as well as increased
small-scale labor demand (via new opportunities for producing new goods and services for
the market).
A final channel that I investigate relates to migration. I discuss how differential in-
and out-migration affect interpretation of the employment results. I show that differential
in-migration cannot explain all of the employment effects of electrification, and explain
how differential out-migration, while substantial, is also unlikely to account for employment
effects, given the profile of out-migrants from rural areas documented in other datasets and
by other researchers. Rather, the migration analysis broadly suggests that people may be
induced to stay in or to move towards areas in which infrastructure is rolling out.
This paper contributes to two literatures. First, it adds to what we know about the mi-
croeconomic effects of physical infrastructure in developing countries, placing new emphasis
on labor market effects in an area that has recently focussed on poverty, health and educa-
tion outcomes.
4
The results here suggest that studies that ignore employment effects could
be missing important economic impacts, particularly when the infrastructure has a home
production bias. Second, the main result that female employment rises in electrifying areas
connects with a large literature on the effects of changing constraints on women’s work in
the process of economic development.
5
4
For example, see Cutler and Miller (2005), Loshkin and Yemtsov (2005), Akee(2006), Duflo and Pande
(2006), Banerjee et al (2007), Cattaneo et al (2007).
5
See, for example, Goldin (1994), Goldin and Katz (2000), Mammen and Paxson (2000), Greenwood et
4
The paper begins by discussing how household electrification may affect rural labor mar-
kets through home and market production. Sections 2, 3 and 4 describe the context of South
Africa’s electrification, data and empirical strategies. Section 5 presents the main results
while section 6 investigates the channels through which electrification affects employment.
Section 7 concludes.
1 Theoretical impacts of household electrification
New access to household electrification may change the nature of work in the home as well
as the amount and type of work that can be done in the market. Providing new public
infrastructure to a location may affect migration of employed and unemployed individuals.
Outlining the form each of these changes may take is important for interpreting the empirical
results in the paper.
To begin, home production activities are important in my study area. Figure 1 (a) and 1
(b) show the fraction of rural African households in KZN reporting different sources of fuel
for cooking and lighting in the 1996 and 2001 Census, separately for communities that get
new access to electricity or not during this period. Almost 80 percent of households cook
with wood and light their homes using candles in the mid-1990s. In electrified areas, the
fraction of households cooking with electricity increases almost three-fold in five years, while
the fraction of households using electric lighting more than triples.
The labor supply effect of such a shock to the technology of home production is, however,
ambiguous.
6
With this new technology, households become more productive in time-intensive
activities like food preparation and storage, and so may substitute more time towards these
home-based activities. The same shock also increases the length of the effective day, produc-
ing an endowment effect that increases the demand for all normal goods, leading households
to supply more labor to market. The more income-elastic the demand for market-intensive
goods is, the stronger this endowment effect will be in pushing households to supply more
labor to the market. Which effect dominates is theoretically ambiguous; however, the substi-
tution effect is likely to be smaller since the demand for home-produced goods (e.g. meals)
is bounded above. Therefore, we expect the advent of household electricity to change the
nature of home production and increase labor supplied to the market, particularly for indi-
viduals who specialize in home production (i.e. women).
7
al (2005), Bailey and Collins (2006) and Coen-Pirani et al (2008).
6
Becker (1965) and Gronau (1986) provide the canonical models of home production, within which the
labor supply effects of a shock to home production technology can be shown to be ambiguous.
7
Responses to the technology shock may also differ across households. If there is heterogeneity across
households in initial home production technologies, or in the degree of substitutability of home for market
5
