Andrew M. Jones 567to have been unanticipated, that the income transfers were large in magnitude
(affecting the real value of savings, collectively bargained wages and pay in gen-
eral) and that there was individual variation in the impact, with civil servants
experiencing an immediate effect. This variation is exploited in an econometric
framework that also allows for entry and attrition from the panel dataset and for
inherent individual heterogeneity. The analysis uses longitudinal data from the
German Socio-economic Panel (GSOEP) from 1984 to 2002 for West Germans and
from 1990 to 2002 for East Germans. Data for East Germans is not available prior
to reunification, so separate models are estimated for East and West Germans and
the natural experiment has to be used indirectly.
In 1996 a crisis in the public pension system in Russia meant that 14 million
of the 39 million state pensioners faced substantial arrears in their payments.
Jensen and Richter (2004) exploit this pensions crisis as a natural experiment.
Their findings show a doubling of poverty rates, significant declines in calorie
and protein intake, and reductions in the use of health services and medica-
tions. They also show evidence of attempts to mitigate the loss of pension income
through work, sales of assets, borrowing and private transfers. Data from the Rus-
sian Longitudinal Monitoring Survey (RLMS) for 1995 and 1996 is used to assess
the impact of the crisis. Identification stems from geographic variation in arrears,
which arises because decisions were regionally decentralized across administrative
areas (oblasts) and discretion was exercised within oblasts. The control group is
made up of households who continued to receive their pensions. Estimation uses a
difference-in-differences design, with the policy effect measured by an interaction
between the post-1996 period and whether an individual’s pension was in arrears.
The identification strategy relies on the assignment of arrears not being associated
with outcomes prior to the crisis. The paper attempts to assess the validity of this
assumption and presents evidence of a common trend for treated and controls
prior to the crisis.
In contrast to Jensen and Richter (2004), Duflo (2000) uses a positive economic
shock associated with public pensions as a source of exogenous variation in income
in her study of child health in South Africa. The end of the Apartheid era in the
early 1990s led to large increases in benefits for black Africans within the South
African Old Age Pension System. Duflo’s study uses cross-section data collected
during 1993 and faces a selection problem, as children living in households with
pension recipients are more likely to be disadvantaged and to live in rural areas.
Her identification strategy compares eligible and non-eligible households and those
children exposed to the increased household pension income for all of their lives
or for only a fraction of their lives. Outcomes are measured using height-for-age
z-scores and there is evidence of an effect on child health and nutrition. This effect
is entirely attributable to pensions received by women and the effect is strongest
for girls.
Chay and Greenstone (2003) bring together a comprehensive set of data sources
within a quasi-experimental research design to investigate the impact of atmo-
spheric pollution on infant health in the US. They find significant effects of total
suspended particles (TSPs) on infant mortality, mostly driven by deaths within one