568 Panel Data Methods
month of birth. There is heterogeneity in this effect, with the impact on infant
mortality rates being twice as large among blacks. Identification is based on geo-
graphic variation in the impact of the 1981–82 recession on levels of TSP, which
is treated as a source of random variation. County-level data from various sources
are merged for the period 1978–84. First differenced (fixed effects) models and,
unusually, differenced models with fixed effects for the trend (double differencing)
are used. The latter allows for heterogenity in trends. The analysis goes a step fur-
ther by selecting neighboring counties as controls. This uses non-manufacturing
counties that either are or are not neighbors to manufacturing counties to
try and isolate the effects of pollution from the socioeconomic effects of the
recession.
Lindahl (2005) shows how lottery winnings can provide one source of exogenous
variation in income, in an attempt to overcome the selection biases inherent in
disentangling the socioeconomic gradient in health. There is a statistically signifi-
cant effect of income on morbidity and mortality and the magnitude of this effect
is largely unchanged when lottery winnings are used as an instrument, although
the estimates are less precise. This income effect is not apparent for the sub-sample
aged over 60. Data from the Swedish Level of Living Surveys (SLLS) for 1968, 1974
and 1981 are matched with register data on income and deaths up to 1997. Mor-
bidity is measured by combining 48 symptoms into a standardized measure and
mortality is measured as death within five or ten years of the surveys. Lottery win-
nings are treated as a source of exogenous variation in income: assuming that the
variation is independent of health. Models are estimated with lottery winnings
included directly. Then OLS and instrumental variable estimates (using winnings
as the instrument) are compared for the sample of individuals who are identified
as “players.” The magnitudes of the income effects are similar although standard
errors are inflated when IV is used. A similar strategy is adopted by Gardner and
Oswald (2007). They use data on the General Health Questionnaire (GHQ-12) mea-
sure of psychological well-being from the British Household Panel Survey (BHPS)
for 1996–2003 and compare those who received lottery winnings of between £1,000
and £120,000 to two control groups, those with smaller wins and those with no
wins. The study finds a statistically significant effect of 1.4 GHQ-12 points after
two years (compared to the average drop of 5 points associated with widowhood).
An important caveat is the small number of treated cases: there are only 137
observations with large lottery wins.
In Van Den Berget al.(2006) the state of the economy during infancy is shown
to have long-term consequences for mortality rates in this inventive study of
those born in the nineteenth century in the Netherlands. The analysis finds a
significant effect of the stage of the business cycle (boom or bust) at the time of
birth on individuals’ subsequent age of death. Data from the Historical Sample
of the Netherlands (HSN), drawn from registers of births, marriages and deaths,
covers 14,000 individuals born between 1812 and 1912 with follow-up to 2000.
This data is merged with macroeconomic time series that are used to identify the
phases of the business cycle. Macroeconomic conditions early in life are used as