Evidence from the Chernobyl Nuclear Accident: the Effect on Health, Education, and
Labor Market Outcomes in Belarus
Maksim Yemelyanau1
Aliaksandr Amialchuk2
Mir M. Ali3
Abstract
The Chernobyl nuclear accident of 1986 had deleterious health consequences for the population
of Belarus, especially for those who were children at the time of the disaster. Using the 20032008 waves of the Belarusian Household Survey of Income and Expenditure (BHSIE), we
estimate the effect of radiation exposure on the health, education, and labor market outcomes
among cohorts and areas affected by the accident, utilizing the nuclear accident as a natural
experiment. We find that young individuals who came from the most contaminated areas had
worse health, were less likely to hold university degrees, were less likely to be employed, and had
lower wages compared to those who were older at the time of the accident and who came from
less contaminated areas.
JEL-code: I18 I20 Q53 J24
Keywords: Chernobyl, Belarus, health, education, wage, employment.
The authors would like to thank Christopher Jepsen, Štěpán Jurajda, Randall K. Filer, Mykhaylo Salnykov,
Kateryna Bornukova, Lauren Heller and the participants of the seminar at BEROC for their valuable
comments. The views expressed here are those of the authors and do not necessarily reflect the views of the
Food & Drug Administration.
1
CERGE-EI, P.O. Box 882, Politických vězňů 7, Prague, 111 21, Czech Republic. Tel.: (420) 224 005
205; Fax: (+420) 224 211 374, e-mail: [email protected]
2
Department of Economics, University of Toledo, Toledo, OH 43606-3390, USA, phone: (419) 530-5147,
fax: (419) 530-7844, e-mail: [email protected]
3
Department of Economics, University of Toledo & Office of Regulations, Policy and Social Science,
Food & Drug Administration, College Park, MD 20740, USA, phone: (301) 436-1784, fax: (301) 4362637, e-mail: [email protected]
1. Introduction
In less than twenty years of its modern history, the Republic of Belarus experienced
several major external shocks that had serious impacts on its economy. In addition to the current
world economic crisis, the Russian crisis of 1998, and the collapse of the Soviet Union in 1991,
the Chernobyl disaster, which happened even before Belarus regained its independence,
affected—and continues to affect—the Belarusian economy. Twenty-five years later, there is an
ongoing debate on the socio-economic consequences of this radioactive catastrophe.4 The goal of
this paper is to investigate its long-term impact on health, education, and labor market outcomes
in Belarus.
On April 26, 1986, reactor number four at the Chernobyl nuclear plant in Ukraine
exploded. A series of subsequent explosions and the resulting fire sent clouds of highly
radioactive particles into the atmosphere and resulted in radioactive fallout over an extensive
geographical area. Large areas of Belarus, northern Ukraine, and parts of the Russian Federation
were contaminated, and radioactive clouds later reached Scandinavia and western Europe.
According to various estimates, 60-70 percent of all radioactive substances fell on Belarus, and
almost a quarter of its territory was affected (Reiners et al., 2008). While the medical and
physical consequences of the Chernobyl accident have been widely documented in the literature,
the long-term socioeconomic consequences of the accident have, with a few exceptions, received
little attention (Almond et al., 2009; Lehmann and Wadsworth, 2009). Because many countries
increasingly rely on nuclear power as a source of energy and because the radioactive waste
storage continues to expand and age, it is important to understand the socioeconomic
consequences of potential radiation accidents, including explosions and radiation leaks. The
recent Fukushima Daiichi nuclear plant crisis in Japan and the Three Mile Island accident in the
US in 1979 also highlight the need to understand more completely the consequences of nuclear
4
See, for example, the discussion of the safety of using the contaminated land for agriculture in Belarus:
http://lenta.ru/articles/2010/07/22/chernobyl/, accessed on 05.25.2011.
2
radiation exposure. Investigating the consequences of these nuclear accidents will help illuminate
their total cost, which goes beyond high energy prices, nuclear waste disposal, and global
warming concerns to include the large amounts of state resources required to deal with long-term
health and socioeconomic consequences. Comprehensive knowledge about the relationships
between radiation exposure and public health and socioeconomic outcomes will allow states to
formulate more efficient policies for effective response to such accidents and to clarify the direct
and indirect costs of such accidents.
Using data from a nationally representative sample of the population in Belarus, this
paper addresses the relationship between exposure during childhood to radiation from the
Chernobyl accident and subsequent health, education, and labor market outcomes. We treat
radiation exposure as an exogenous shock because the radiation fallout from Chernobyl was
distributed randomly, conditional upon the distance to the reactor, the prevailing wind patterns,
the rainfall levels and the local topography (Lehmann and Wadsworth, 2009). We examine the
relationship between the extent of irradiation in the region of residence of individuals who were
between one and seven years old at the time of the accident, and their health, education, and labor
market outcomes seventeen and more years later. We use differences in radiation dosage by
region and by age at exposure, using a difference-in-differences estimation approach, in order to
identify the effect of the accident on the outcomes of interest. We also account for a range of
confounding factors and address the issue of endogenous migration.
Our results indicate that individuals who lived in the affected regions and those who were
young at the time of the Chernobyl accident had worse health, were less likely to hold university
degrees, were less likely to be employed, and had lower wages compared to those who were older
at the time of the accident and who came from the less contaminated regions. We also find that
the negative effects were concentrated among males. To the best of our knowledge, this is the
first study to quantify the impact of the Chernobyl disaster on the socioeconomic outcomes in
Belarus, the country most affected by the accident.
3
The rest of the paper is structured as follows: section 2 summarizes the literature on
radiation-induced health effects, section 3 describes the health implications of the Chernobyl
accident for Belarus, section 4 describes the data, section 5 discusses our methods and
identification strategy, section 6 presents the results, and section 7 offers a concluding discussion.
2. Literature review
Social scientists have long recognized a significant correlation between health and
socioeconomic outcomes like education, wages, and employment. In particular, the link between
poor health experienced during childhood and poor education and socioeconomic outcomes in
adulthood has been widely documented in the literature (Case et al., 2002 and 2005). However,
understanding the long-term consequences of a negative health shock has presented serious
challenges in empirical work because of the endogeneity concerns that arise from omitted
variables and measurement errors in health measures. In other words, unobservable factors that
lead to both poor health and poor economic performance result in a downward bias of the
estimates. While the majority of extant empirical studies report only correlational relationships
between health and socioeconomic outcomes (see literature surveys in Grossman and Kaestner,
1997; Strauss and Thomas, 1998), a growing body of literature has attempted to establish a causal
link between poor health and socioeconomic outcomes.
Empirical research that has attempted to estimate this causal relationship has, utilizing a
variety of approaches, generally found evidence of a negative impact of poor health on
educational and economic outcomes. For example, Kremer and Miguel (2004) used a randomized
assignment of a deworming treatment in primary schools in Kenya and found a significant
reduction in school absenteeism but no gains in academic performance after the treatment.
Bleakley (2007) used the differential timing of different birth cohorts’ exposure to a large-scale
intervention against hookworm in children in the American South circa 1910 and found
improvements in health, income, and schooling as a result of the treatment. Maccini and Yang
4
(2009) used geographical variations in rainfall around the time of the birth of Indonesian adults
and found that higher levels of rainfall in women’s early years had positive effects on their adult
socioeconomic status, with the benefits mediated by higher education attainment. Fletcher and
Lehrer (2008) used differences in genetic inheritance among children in the same family and
found a significant impact of poor mental health on academic achievement.
Several recent studies have used environmental shocks to determine the effect of health
on socioeconomic status. Almond (2006) exploited the 1918 influenza pandemic to find that
cohorts in utero during the pandemic had reduced educational attainment, lower income, and
lower socioeconomic status later in life. Meng and Qian (2006) used regional variation in famine
intensity in China to find that childhood exposure to famine had significant adverse effects on
adult health and work capacity.
In the context of the literature that relies on natural experiments, the Chernobyl accident
can be viewed as an exogenous environmental shock to health whose influence varies according
to the extent of irradiation. Almond et al. (2009), arguing that young children and those in utero
appeared to be more vulnerable to radiation, used variation in radiation across Swedish regions to
compare the performance of those in utero to the performance of those born before and after
1986. They found that the cohort in utero during the accident had worse educational outcomes.
However, the doses of radiation received by the affected cohorts in Sweden were significantly
lower than those received by comparable cohorts in Belarus. Lehmann and Wadsworth (2009)
used Ukrainian data to find a positive association between radiation dosage in local areas and the
individual’s perception of poor health status, but only a weak association between the level of
radiation and labor market performance measures.
In this paper, we argue that the radioactive fallout from the Chernobyl accident produced
disparities in health by age at exposure and by region of residence. The radioactive contamination
was higher in Belarus than in any other country because of the predominantly northwestward
winds at the time of the accident. The most heavily contaminated areas in Belarus were located in
5
the Gomel region, followed by the Mogilev and Brest regions and Minsk city. However, certain
parts of the country, notably the Vitebsk region and most of the Grodno and Minsk regions,
suffered very little (Figure 1). In addition, the effect of radiation was age-specific: children aged
0-4 years were especially vulnerable to radiation because of their higher rates of absorption of
radioactive iodine.5
2. Health implications of Chernobyl in Belarus
2.1. Thyroid malignancies
One of the principal radioactive elements released during the Chernobyl accident was
radioactive iodine-131 (¹³¹I), which was significant for the first few months after the accident.
The thyroid gland accumulates iodine from the blood stream as part of its normal metabolism.
Iodine-131 was incorporated in the thyroid glands of the vast majority of Belarus’ inhabitants
through inhalation and ingestion of food, especially milk, containing high levels of ¹³¹I, causing
internal irradiation. Even though the entire population of the Republic of Belarus was exposed to
radioactive ¹³¹I in 1986, the thyroid doses varied widely based on age, level to which the ground
was contaminated with ¹³¹I, and milk consumption rate. Individual thyroid doses of ¹³¹I reached
50 Gy,6 with average doses in contaminated areas between 0.03 Gy and a few Gy, depending on
the region where people lived and on their age. Among those exposed, children and adolescents
received maximal doses, primarily from drinking milk from cows that ate contaminated grass
5
Lehmann and Wadsworth (2009) also looked at the differences in irradiation by age at exposure.
However, their analysis looked at the differences at much older ages, and their treatment group was
comprised of children under age 13 at the time of the Chernobyl accident.
6
Because many organs and tissues were exposed by the Chernobyl accident, the concept of effective dose,
which characterizes the overall health risk that is due to any combination of radiation, has been commonly
used. The effective dose accounts for both absorbed energy and type of radiation and for susceptibility of
various organs and tissues to developing a severe radiation-induced cancer or genetic effect. For gamma
radiation, the unit of effective dose is the sievert (Sy). (One Gy corresponds to one Sv.) One Sy is a
comparatively large dose, so the millisievert or mSv (one one-thousandth of a Sv) is commonly used to
describe normal exposures. Annual natural background doses of humans worldwide were estimated to
average 2.4 mSv, with a typical range of 1–10 mSv (IAEA 2005).
6
immediately after the accident7 (Bespalchuk et al., 2007; Farahati et al., 2000; Reiners et al.,
2008; IAEA, 2005; Balonov, 2007; WHO, 2006; Lehmann and Wadsworth, 2009).
Childhood thyroid cancer caused by radioactive iodine was one of the main
manifestations of the health impacts of the accident since the thyroid gland is one of the organs
most susceptible to cancer from radiation. Children have been found to be the most vulnerable
population, and a substantial increase in thyroid cancer among those exposed as children was
recorded after the accident (IAEA, 2005). Thyroid cancer in children occurs rarely in most
countries, but Belarus is an exception (Bespalchuk et al., 2007). A strong age-dependent risk for
thyroid cancer was previously observed in the Japanese population after the atomic bomb
explosions, with the highest risk in the group of children below the age of ten. After the
Chernobyl accident, children from Belarus who lived in highly exposed regions received mean
thyroid doses by radioactive fallout approximately twice as high as did the survivors of the
atomic bomb explosions in Japan (Reiners et al., 2008).
The literature suggests little doubt that the Chernobyl accident was responsible for the
abnormal increase in thyroid cancer in the affected areas. According to Bespalchuk et al. (2007),
three major factors are indicative of the radiogenic nature of the thyroid cancer. The first is the
high prevalence of the disease; before the Chernobyl accident, the incidence of thyroid cancer
was very low in both children and adults. The second factor is geographical; the majority of
children and adolescents diagnosed for thyroid carcinoma lived in the Gomel region and the
southeastern parts of the Brest region. The third factor is the correlation between level of
radiation dosage in a region and the number of registered cancer cases. Farahati et al. (2000)
found in a cohort of 483 patients that younger age at the time of the Chernobyl accident was
associated with greater cancer complications: compared with patients who were 6.1-8 years old at
the time of the accident, patients who were younger than 2 years old had significantly more
7
The rural population of Belarus consumed locally produced milk, and the urban population consumed
milk from the numerous local milk factories that used primarily local milk for their production (Kruk
2004).
7
cancer complications. The rapid cellular growth that occurs in children under age two in all
likelihood facilitated quicker and broader development of the cancer (Farahati et al., 2000). By
2008 radiation was linked to approximately five thousand cases of thyroid cancer in children and
adolescents, with the highest incidence in the group of those who were 0-4 years old in 1986
(Balonov, 2007; Reiners et al., 2008).
In addition to cancer among the most heavily irradiated individuals, exposure to
Chernobyl's radioactive iodine caused a large majority of the exposed children to have more
antithyroid antibodies than other children, which may cause later development of
hypothyroidism8 (Pacini et al., 1998). Ostroumova et al. (2009) found a significant relationship
between the prevalence of hypothyroidism and individual ¹³¹I thyroid doses from the Chernobyl
accident. In several studies conducted in Belarus and Ukraine, a significant increase in the rates
of juvenile hypothyroidism was reported in children who lived in radionuclide-contaminated
areas compared with children from uncontaminated areas (Goldsmith et al., 1999; Quastel et al.,
1997; Vykhovanets et al., 1997).
2.2. Other health effects of Chernobyl
While thyroid cancer is the most demonstrable health impact of Chernobyl, a number of
other illnesses, including lung diseases, digestive and blood disorders, birth defects, fertility
problems, and immune deficiencies, are more prevalent in the radiation-affected areas than in
other areas (UNDP, 2002). Acute Radiation Syndrome was the most immediate effect of
irradiation from the release of radioactive particles, which primarily consisted of isotopes of
uranium and plutonium (the most toxic), iodine-131, strontium-90 and cesium-137. However,
Acute Radiation Syndrome affected relatively few people, limited primarily to those who directly
participated in the liquidation of the effects of the disaster on the power plant itself.
8
Hypothyroidism is the result of an underactive thyroid gland that cannot make enough thyroid hormone to
keep the body running normally. When thyroid hormone levels are too low, the body's cells cannot get
enough thyroid hormone, and the body's processes start slowing down. As the metabolism slows, the
patient tires easily, becomes forgetful and depressed, and gains weight (Source: American Thyroid
Association, http://www.thyroid.org/patients/patient_brochures/childhood.html, accessed on 05.25.2011).
8
Apart from the dramatic increase in thyroid cancer among those exposed at a young age,
there is no clearly demonstrated increase in the population of Belarus of solid cancers or leukemia
from radiation. Several studies have cited a notable increase in psychological problems among the
affected population, compounded by insufficient communication about the effects of radiation
and by the social disruption and economic depression that followed the break-up of the Soviet
Union (see review in Lehmann and Wadsworth, 2009). Some studies have cited "Chernobyl
AIDS," which is characterized by weakened immune system and susceptibility to cardiac
conditions and common infections. There have also been reports of the effects of Chernobyl on
mental health, manifested in anxiety, depression, and reduced cognitive function from prenatal
radiation in the high fallout areas of Ukraine, Belarus, and Russia (Nyahu et al., 1998;
Kolominsky et al., 1999; Loganovskaja and Loganovsky, 1999; Almond et al., 2009).
3. Data
Our empirical analysis is based on the 2003-2008 waves of the Belarusian Household
Survey of Income and Expenditure (BHSIE), called the Belarusian Household Budget Survey
(BHBS) in the literature. BHSIE, which was first fielded in 1995, incorporates interviews with a
random sample of approximately five thousand households every year. BHSIE is the most
reliable and comprehensive source of micro-data in Belarus. In addition to demographic and labor
force variables, it contains information on self-assessed health status, hospitalizations, and
medical visits.
Our sample consists of individuals who were born before the Chernobyl accident.
Because we do not observe the exact month of birth and cannot distinguish between the
individuals born in 1986 before and after April 26, we restrict the latest year of birth to 1985. In
addition, we restrict our analysis to individuals who were under age sixteen in 1991. There had
been a major break-up in the labor markets in 1991 because of the collapse of the Soviet Union,
so restricting the age in this way allows us to compare individuals who faced similar labor market
9
conditions.9 We use two age cohorts: individuals born in 1983, 1984, and 1985 (the treatment
cohort) and those born in 1979, 1980, and 1981 (the control cohort). Those in the treatment
cohort were age 18-25 in 2003-2008, and those in the control cohort were age 22-29 in 20032008.
Our sample consists of 6,504 individuals in six years of data (2003-2008), although data
for some of our variables for some individuals is missing (see Table 1). Information on current
employment is available for 5,392 individuals, and information on wages is available for 4,443
individuals. Because the first half of the 1980s was a time of high birth rates in Belarus, the
treatment cohort represents 52 percent of the data, versus 48 percent for the control cohort. The
geographic identifier available in the data refers to six regions (corresponding to “oblast” in
Russian) and the capital city of Minsk. Among the six regions and the city of Minsk, the most
affected region was the Gomel region, the less affected were the Mogilev and Brest regions and
Minsk city, and the least affected were the Grodno, Vitebsk, and Minsk (excluding Minsk city)
regions. The average thyroid dose received by the residents of Gomel region (0.3-0.35 Gy), the
most affected area, was over thirty times the dose in the least affected regions (Hrodno, Minsk,
and Vitebsk). The average thyroid dose in the less affected areas (the Mogilev and Brest regions
and Minsk city) was approximately six times the dose in the least affected regions (see Figure 3).
Sixteen percent of our sample came from the most radiation-contaminated region,48 percent
came from the less contaminated areas, and the remaining 36 percent came from the least
contaminated areas.
4. Methods
Our empirical strategy is based on observed geographic and age differences in the
amount of irradiation received. Children and adolescents in the Gomel region received the
9
This choice assumes that entry into the labor market entry starts at age 16. While most people at this age
are continuing their secondary education, some choose to join the labor force.
10
greatest amount of radiation, followed by those in the less affected area (the Brest and Mogilev
regions and Minsk city), while those who lived in the Vitebsk, Grodno, and Minsk regions
received relatively mild doses (Figures 2 and 3). Children under the age of four at the time of
exposure were especially affected by the radiation. We estimate the following empirical model:
yijt    1 (Youngerijt  Most_affected ijt )   2 (Youngerijt  Less_affected ijt )
  3 Youngerijt   4 Most_affected ijt   5 Less_affected ijt  X ijt   ijt ,
where yijt is the outcome of interest of an individual i living in region j and observed in year t .
Among the health, educational, and labor market outcomes that we consider, the variable
“medical visits” refers to the number of an individual’s reported medical visits in the last three
months. The variable “hospitalized” is a dummy variable that equals one if the individual
reported being hospitalized in the last twelve months and zero otherwise. The variable “not good
health” is a dummy variable that equals one if the individual rated his/her health as “not very
good” or “bad” and zero otherwise. The variable "university diploma" is a dummy variable that
equals one if the individual has any diploma of higher education and above and zero otherwise.
The variable “employed” is a dummy variable equal to one if the individual reported being
currently employed and zero otherwise. The variable “log wages” refers to the natural logarithm
of the current wages at the individual’s main job.
Youngerijt is a dummy variable equal to one if an individual was one, two, or three
years old in 1986 (the treatment cohort) and equal to zero if an individual was five, six, or seven
years old in 1986 (the control cohort). This dummy variable captures any fixed differences in the
outcome between the two cohorts, as well as general economy-wide time trends that drive
changes between the cohorts. Our geographic indicators are based on the region of current
residence. Most_affected ijt is a dummy variable indicating the most contaminated Gomel
region. Less_affected ijt is a dummy variable indicating the less contaminated areas (the Mogilev
and Brest regions and Minsk city). These geographical variables control for any fixed differences
11
between the affected and non-affected regions. X ijt is a vector of demographic and
socioeconomic controls including age (in years) and its square, a rural/urban dummy, a dummy
for living in a big city (>100,000 residents), and a dummy for male. In addition, we have
information on how long the family lived in the current home. We constructed two dummy
variables, one indicating whether the family moved in 1986 and the other indicating whether the
family moved at any time after 1986. In the absence of data on the region of birth, these dummy
variables are intended to control for the possibility that the families that migrated on or after 1986
may have different characteristics from the families that migrated before the accident or that did
not migrate. Finally, year dummies are included to control for the general economy-wide trends.
Our empirical model seeks to identify the coefficients 1 and  2 on interaction variables
Youngerijt  Most_affected ijt and Youngerijt  Less_affectedijt , respectively. We hypothesize that
both of these coefficients are negative and that 1   2 . In other words, people who were
younger at the time of the accident and who lived in the more contaminated areas had worse
health, education, and labor market outcomes, and the effect increased in magnitude with the
extent of area-level contamination.
5. Results
Main results
Table 1 reports descriptive statistics for the sample used in our analysis. As a first step in
our analysis, we plotted all of our outcome variables by area of exposure to radiation (most
affected, less affected, least affected) and by age at the time of the accident. Our research design
is based on the assumption that younger children (especially those age four and below at the time
of the Chernobyl accident) who lived in the more contaminated regions experienced worse
outcomes. Accordingly, for the negative outcomes (number of medical visits, being hospitalized,
not good health), we expect to see that individuals who were younger and who were from the
12
more contaminated areas experienced disproportionately more of these outcomes than individuals
who were older and from the more contaminated areas. On the other hand, for the positive
outcomes (having a university degree, being currently employed, monthly wage), we expect to
see that younger individuals from the more contaminated areas experienced disproportionately
less of these outcomes compared to older individuals from the more contaminated areas. In other
words, we expect a more flat (steep) age profile in the more contaminated areas relative to the
less contaminated areas for the negative (positive) outcomes. Conforming to our expectations, the
left-hand panel of Figure 4 suggests that the number of medical visits, being hospitalized and
reports of not being in good health all follow a flatter age profile for the more contaminated areas
relative to the less contaminated areas, while having a university degree and a monthly wage
follow a steeper age profile for the more contaminated areas.
We estimate our empirical specification using a linear regression for all of our outcome
variables except for the number of medical visits. Linear regression provides the best linear
approximation to the Condition Expectation Function even if the latter is non-linear (Angrist and
Pischke, 2009, p. 38).10 We estimated a negative binomial regression for the count variable
“medical visits” and presented the incidence rate ratios (exponentiated coefficients).
The hypothesized mechanism for the effect of radiation on education and labor market
outcomes is via the effect on health. Therefore, we start our multivariate analysis by exploring the
impact of radiation exposure on three self-reported health outcomes (Table 2). The number of
medical visits for individuals who were exposed to radiation in the most contaminated region
increased by 1.112 visits, and these individuals had a 7 percentage points higher chance of not
being in good health. While we do not find differences in the effect of radiation on the number of
medical visits between the less exposed regions and the least exposed areas, we do find that
10
We also estimated probit regressions for the binary variables “hospitalized,” “not good health,”
“university,” and “employed” and obtained similar results. In order to address endogenous selection into
employment, we estimated the wage equation using the Heckman (1979) selection model, where the
selection equation included the “not good health” dummy variable in addition to all of the main regression
equation variables.
13
individuals in the less contaminated regions have a 9 percentage points higher chance of not
being in good health. While the probability of not being in good health is slightly higher among
those living in the less exposed areas, compared to the most exposed region, the difference is not
statistically significant. We do not find that the probability of being hospitalized is affected by the
amount of radiation exposure, which may be due to the non-critical nature of the malignancies
caused by the radiation in most people and the possibility that cancer patients are hospitalized at
the time of the survey and so are unable to respond. Chernobyl’s lack of effect on the probability
of hospitalization is in line with the corresponding finding in Almond et al. (2009). However, the
radiation doses received by the treated group in that study were an order of magnitude smaller
than the doses received by individuals in the most contaminated areas of Belarus. On the other
hand, our findings of the negative effect of radiation on the other health measures are in
accordance with Lehmann and Wadsworth (2009), who found a significantly positive association
between a self-reported measure of poor health and residence in contaminated areas in 1986 for
all adults living in Ukraine in 2003 and 2004. 11
There is evidence of a negative effect of radiation on education in the most contaminated
region, which reduces the probability of having a college degree by 2.5 percentage points. While
a negative effect of radiation is also present in the less contaminated areas, this effect does not
attain statistical significance at the conventional levels. We also find a negative effect of radiation
on the labor market outcomes, as shown in the last two columns of Table 2. 12 The probability of
being currently employed reduces by 4.1 percentage points in the most exposed region. While
11
There is also a significant association between the dummy variable “influenced by Chernobyl” (=1 if a
person reported being seriously or partially influenced by Chernobyl and zero otherwise) and all health
measures in the years 1999-2002 (results not shown). This association gives further support to the
hypothesis that the reported health problems were related to Chernobyl. Unfortunately, the survey does not
elaborate on exactly how the person was influenced by Chernobyl, and this variable is not available in
BHSIE after 2002.
12
Education variables were not included in the wage regression because they constitute “bad controls,”
since they are “outcome variables in the notional experiment at hand” (Angrist and Pischke, 2009). In our
case, since all of the individuals are under age 30 and many are still enrolled in secondary education or
college, including education variables as controls could introduce selection bias. Furthermore, we are
interested in the reduced effect of Chernobyl on earnings, including the part of this effect that works
through education.
14
individuals living in the less contaminated regions have 5.7 percentage points reduction in
employment probability, this effect is only marginally statistically significant (at the 10% level).
We also observe a negative effect of radiation on wages for those living in the most contaminated
region since there is a 5.5 percent reduction in the current wage at the main job for these
individuals. The 2.9 percent reduction in wages for individuals living in the less contaminated
regions is not statistically significant. The coefficients on the indicators “moved in 1986” and
“moved after 1986” were insignificant in all regressions, supporting the hypothesis that migration
is not an important predictor of our outcomes and that migration from the contaminated regions is
unlikely to bias our results.
Among the other regressors, the male dummy is the only highly statistically significant
predictor of every outcome in our analysis, which could in part reflect the different labor market
experiences of males and females. The industrial composition of employment in Belarus is very
gender-specific, with males dominating the manufacturing and construction industries and
females dominating healthcare, education, and service industries. Furthermore, the gender wage
gap, which has been persistently high, more than doubled in Belarus between 1996 and 2004
(Pastore and Verashchagina, 2007).
We explore the differences in the effect of radiation on health, education, and labor
market outcomes by gender in Table 3. The estimates reveal some differences by gender in the
negative effects of radiation on the outcomes. For example, the incidence rate of medical visits is
higher in the most contaminated areas only for males, but both males and females have a higher
probability of not being in good health in both the most and the less contaminated areas (the
differences in the effects are not statistically significant between the genders in each type of area).
Both males and females in the most contaminated region have a lower probability of having a
university degree, with the effect being more pronounced among females. The largest differences
in the effect of radiation are found in the labor market outcomes. While males are almost 6
percentage points less likely to be currently employed in the most contaminated region, the
15
corresponding effect among females is only 2 percentage points (the latter effect is only
marginally statistically significant). The negative effect of radiation on wage appears to be
present only among males in the most contaminated region, who suffer almost 10 percent
reduction in their wages.
Robustness of the results
If different regions experienced pre-existing trends in the socio-economic development
across age cohorts, our interaction variables would pick up these trends, and our results would
reflect these trends rather than the effect of the Chernobyl accident. Even though Belarus has a
relatively homogenous population, some regions may be historically less developed than others.
We test for the possibility that differential trends drive our results using a falsification test where
we obtained the main results using two unaffected groups: those who were five, six, or seven
years old in 1986 (the pseudo-treatment group) vs. those who were nine, ten, and eleven years old
in 1986 (another control group) (Table A1). Estimates on the interaction variables are not
statistically significant except for the estimates for the outcome “not good health” and
“employed,” which are significant only at the 5 percent level and have an unexpected sign. This
result supports the validity of our identification strategy based on comparisons across age cohorts
and regions.
There is also a possibility of endogenous migration in response to the Chernobyl
accident. If migrants had different characteristics than those of non-migrants, then the estimated
coefficients on the interaction terms would pick up the effect of selective migration, thus biasing
the effect of radiation on the outcomes. However, migration in the USSR, especially in the rural
areas, was difficult because of the institution of “propiska,” which required individuals to be
registered at their address of residence. In addition, there was a lack of public safety information
on the adverse health effects of radiation after the accident.13 In an attempt to evaluate the
13
Notably, a May 1st parade was held as usual in Kiev, right after the catastrophe on April 26, 1986, only
90 km (about 56 miles) away from the burning power plant. Medvedev (1990) documented the lack of
16
influence of selective migration on our results, we used the 1999 Belarus Census data, which has
information on migration, and compared the characteristics of migrants from their areas of birth
with different levels of irradiation (Table A2). We did not find that migrants from the most
contaminated region or from the less contaminated areas who moved after 1986 were
systematically different from migrants from the least contaminated regions or from individuals
who migrated before 1986 in terms of such characteristics as attainment of a university degree,
being currently employed, having a professional occupation, having a white collar occupation,
being married, and number of children. We also examined whether the population size of the
regions changed in response to the disaster, but as Figure A1 shows, there is no evidence that the
population of the most contaminated Gomel region declined in the decade from 1979 to 1989,
which ends three years after the accident. In fact, it increased by a greater percentage than in any
other region during this decade. The bottom panel of Figure A1, which plots population sizes by
the areas that differed in their levels of contamination, reveals no evidence that the population
trends after 1986 were related to the area-level contamination.
6. Conclusion
The Chernobyl nuclear accident of 1986 had deleterious health consequences for the
population of Belarus, especially for children under four years of age at the time of the disaster.
This paper utilizes the natural experiment generated by this nuclear accident, which produced
sizable increases in radiation levels in many areas of Belarus. Using the 2003-2008 waves of
BHSIE, we estimate the effect of radiation on health, education and labor market outcomes by
relying on differences in the amount of irradiation across age cohorts and across areas of Belarus.
We find that children who were younger at the time of the accident and who came from the most
information and slow response after the catastrophe, pointing out, for example, that “the population of
Pripyat was not warned about the accident, nor were the civil defense headquarters informed. Since the
civil defense staff had no information about the situation, they took no measures. As a result, the usual
rhythm of life on Saturday proceeded …. The evacuation began only 36 hours later and was conducted
according to a newly worked-out plan.”
17
contaminated Gomel region have significantly more medical visits, have a higher probability of
not being in good health, and have a lower probability of having a university degree. They are
also less successful in the labor market, where they have a lower probability of being employed
and earn a lower wage. In addition, we find some evidence that the effect of radiation is not
restricted to the most contaminated region: Younger children who lived in the less contaminated
areas have a significantly higher incidence of not being in good health than those who lived in the
least contaminated regions. Separate analysis by gender reveals that a higher incidence of medical
visits resulting from radiation exposure is present only for males, who also appear to have lower
rates of employment and to earn a lower wage than males from the least irradiated areas. Taken
together, our results support the hypothesis that the effect of radiation on education and labor
market outcomes can be attributed to the effect of radiation on health.
There are several potential caveats to our analysis. First, a potential problem with the
self-reported health measures is that individuals tend to assess their health by comparison to those
in their immediate surroundings (e.g., parents, siblings, and friends). Therefore, children who
come from affected families are likely to answer that their health is not bad by comparison.
Furthermore, while reported poor health status may be due to the real impacts of exposure, it may
also be a self-fulfilling prophecy since those who were exposed expect to have their health suffer
and may become highly sensitive to any health issue. Our results on health measures are also
consistent with the alternative hypothesis of psychological effect and parental stress, which may
affect younger children more than older ones (Lehmann and Wadsworth, 2009).
The current analysis does not consider the possibility of differential effects on child
mortality by level of exposure, which would introduce “survivor bias” in our estimates. One
example of such an effect occurs when wealthier people live farther from the plant so their
children are more likely to survive. In addition, higher mortality among the affected population
would tend to bias our estimates toward zero. However, according to the literature (Bespalchuk et
al., 2007; Farahati et al., 2000; Balonov, 2007; Reiners et al., 2008), only a small fraction of the
18
population has developed thyroid cancer (only about five thousand cancer cases for a population
of about 10 million by 2008), and very few of these patients have died. Most other affected
people develop intermediate stages of thyroid disease, which do not significantly impact mortality
rates but show up only in higher morbidity rates.
Another caveat to our analysis is that parents and the government may have invested
more resources in health and education in the more affected areas in order to offset the effect of
irradiation (Almond et al., 2009).14 If unaccounted for, these investments would tend to bias the
estimated effect of irradiation toward zero. Finally, while it is not possible to determine
individuals’ place of birth and residence during childhood using the BHSIE data in order to
address the issue of endogenous migration directly, the dataset does provide a unique opportunity
to look at various health and socio-economic measures in the country that suffered most from the
Chernobyl accident.
14
The government of Belarus created special privileges for applicants to secondary professional and higher
educational institutions who were from the contaminated regions. For example, Law #9-3 of January 6,
2009, “on social protection of citizens who suffered from the Chernobyl and other radioactive disasters”
(article 18 clause 7, article 21 clause 1.3, article 22 clause 1.3, article 23 clause 1.2) provides that people
who participated in the liquidation of the consequences of the Chernobyl accident or who lived in the
contaminated regions are entitled to preferential rights to obtain secondary, specialized, or higher
education.
19
References
Almond, D. (2006). "Is the 1918 Influenza Pandemic Over? Long-Term Effects of In Utero
Influenza Exposure in the Post-1940 U.S. Population". Journal of Political Economy,
2006, Vol. 114, No. 4, pp. 672--712.
Almond, D., Edlund, L., & Palme, M. (2009). "Chernobyl's Subclinical Legacy: Prenatal
Exposure to Radioactive Fallout and School Outcomes in Sweden". Quarterly Journal of
Economics, Vol. 124, No. 4, pp. 1729--1772.
Angrist, J.D., & Pischke, J.-S. (2009). "Mostly Harmless Econometrics: An Empiricist's
Companion". Princeton University Press, 2009. 392 p.
Balonov, M.I. (2007). "The Chernobyl Forum: major findings and recommendations". Journal of
Environmental Radioactivity, Vol. 96, No. 1-3, pp. 6--12.
Bespalchuk, P.I., Demidchik, Y.E., Demidchik, E.P., Gedrevich, Z.E., Dubovskaya, A.P.,
Saenko, V.A., & Yamashita, S. (2007). "Thyroid cancer in Belarus after Chernobyl".
International Congress Series, No. 1299, pp. 27--31.
Bleakley, H.C. (2007). "Disease and Development: Evidence from Hookworm Eradication in the
American South". Quarterly Journal of Economics, Vol. 122, No. 1, pp. 73--117.
Case, A., Lubotsky, D., & Paxson, C. (2002). "Economic Status and Health in Childhood: The
Origins of the Gradient". American Economic Review, Vol. 92, No. 5, pp. 1308--1334.
Case, A., Fertig, A., & Paxson, C. (2005). "The Lasting Impact of Childhood Health and
Circumstance". Journal of Health Economics, Vol. 24, pp. 365--389.
Farahati, J., Demidchik, E.P., Biko, J., & Reiners, C. (2000). "Inverse association between age at
the time of radiation exposure and extent of disease in cases of radiation-induced
childhood thyroid carcinoma in Belarus". Cancer, Vol. 88, No. 6, pp. 1470--1476.
Fletcher, J.M., & Lehrer, S.F. (2008). "Using Genetic Lotteries within Families to Examine the
Causal Impact of Poor Health on Academic Achievement". Yale University working
paper.
Grossman, M., & Kaestner, R. (1997). "Effects of Education on Health", in J. R. Behrman and N.
Stacey (eds.), The Social Benefits of Education, University of Michigan Press, Ann
Arbor.
Goldsmith, J.R., Grossman, C.M., Morton, W.E., Nussbaum, R.H., Kordysh, E.A., Quastel, M.R.,
Sobel, R.B., & Nussbaum, F.D. (1999). "Juvenile hypothyroidism among two populations
exposed to radioiodine". Environmental Health Perspectives, Vol. 107, No. 4, pp. 303-308.
Heckman, J. (1979). “Sample selection bias as a specification error”. Econometrica 47 (1): 153–
61.
IAEA (2005). Chernobyl Forum, 2003-2005 (online). http://www-ns.iaea.org/meetings/rwsummaries/chernobyl_forum.htm. Chernobyl Forum, 2003-2005.
Kling, J., Liebman, J., & Katz, L., (2007). "Experimental Analysis of Neighborhood Effects",
Econometrica, Vol. 75, No. 1 (January, 2007), pp. 83--119.
Kolominsky, Y., Igumnov, S., & Drozdovitch, V. (1999). "The psychological development of
children from Belarus exposed in the prenatal period to radiation from the Chernobyl
atomic power plant". Journal of Child Psychology and Psychiatry, Vol. 40, No. 2, pp.
299--305.
20
Kremer, M., & Miguel, E. (2004). "Worms: Identifying Impacts on Education and Health in the
Presence of Treatment Externalities". Econometrica, Vol. 72, pp. 159--217.
Kruk, J.E., Pröhl, G., & Kenigsberg, J.I. (2004). "A radioecological model for thyroid dose
reconstruction of the Belarus population following the Chernobyl accident". Radiation
and Environmental Biophysics, Vol. 43, No. 2, pp. 101--110.
Lehmann, H., & Wadsworth, J. (2009). "The Impact of Chernobyl on Health and Labour Market
Performance in the Ukraine", ESCIRRU Working Paper, No 12, DIW Berlin, German
Institute for Economic Research.
Loganovskaja, T.N., & Loganovsky, K.N. (1999). "EEG, cognitive and psychopathological
abnormalities in children irradiated in utero". International Journal of Psychophysiology,
Vol. 34, pp. 213--224.
Loughran, D., Datar, A., & Kilburn, M. (2008). "The response of household parental investment
to child endowments". Review of Economics of the Household, Vol. 6, No. 3 (Sep.,
2008), pp. 223--242.
Maccini, S., & Yang, D. (2009). "Under the Weather: Health, Schooling, and Economic
Consequences of Early-Life Rainfall". American Economic Review, Vol. 99, No. 3.
Medvedev A. Zhores. The Legacy of Chernobyl. Norton, New York, 1990. xiv, 352 pp., illus.
Meng, X., & Qian, N. (2006) "The Long Run Health and Economic Consequences of Famine on
Survivors: Evidence from China's Great Famine". IZA Discussion Paper No. 2471.
Nyahu, A.I., Loganovsky, K.N., & Loganovskaja, T.K. (1998). "Psychophysiologic aftereffects
of prenatal irradiation". International Journal of Psychophysiology, Vol. 30, pp. 303--311.
Ostroumova, E., Brenner, A., Oliynyk, V., McConnell, R., Robbins, J., Terekhova, G., Zablotska,
L., Likhtarev, I., Bouville, A., Shpak, V., Markov, V., Masnyk, I., Ron, E., Tronko, M., &
Hatch, M. (2009). "Subclinical hypothyroidism after radioiodine exposure: UkrainianAmerican cohort study of thyroid cancer and other thyroid diseases after the Chornobyl
accident (1998-2000)". Environmental Health Perspectives, Vol. 117, No. 5, pp. 745-750.
Pacini, F., Vorontsova, T., Molinaro, E., Kuchinskaya, E., Agate, L., Shavrova, E., Astachova, L.,
Chiovato, L., & Pinchera, A. (1998). "Prevalence of thyroid autoantibodies in children
and adolescents from Belarus exposed to the Chernobyl radioactive fallout". Lancet, Vol.
352, No. 9130, pp. 763--766.
Pastore, F. and Verashchagina, A. (2007) When Does Transition Increase the Gender Wage Gap?
An Application to Belarus. IZA Discussion Paper No. 2796.
Quastel, M.R., Goldsmith, J.R., Mirkin, L., Poljak, S., Barki, Y., Levy, J., & Gorodischer, R.
(1997). "Thyroid-stimulating hormone levels in children from Chernobyl".
Environmental Health Perspectives, Vol. 105, Suppl. 6, pp. 1497--1498.
Reiners, C., Demidchik, Y.E., Drozd, V.M., & Biko, J. (2008). "Thyroid cancer in infants and
adolescents after Chernobyl". Minerva Endocrinologica, Vol. 33, No. 4, pp. 381--395.
Remennick, L.I. (2002). "Immigrants from Chernobyl-affected areas in Israel: the link between
health and social adjustment". Social Science & Medicine, Vol. 54, No. 2 (Jan. 2002), pp.
309--317.
Strauss, J., & Thomas, D. (1998). "Health, Nutrition, and Economic Development". Journal of
Economic Literature, Vol. 36, No. 2, pp. 766--817.
21
UNDP (2002). "The Human Consequences of the Chernobyl Nuclear Accident: A Strategy for
Recovery". A Report Commissioned by UNDP and UNICEF with the support of OCHA
and WHO, 25 Jan. 2002.
Vykhovanets, E.V., Chernyshov, V.P., Slukvin, I.I., Antipkin, Y.G., Vasyuk, A.N., Klimenko,
H.F., & Strauss, K.W. (1997). "¹³¹I dose-dependent thyroid autoimmune disorders in
children living around Chernobyl". Clinical Immunology and Immunopathology, Vol. 84,
No 3, pp. 251--259.
WHO (2006). "Health effects of the Chernobyl accident and special health care programmes".
WHO, Geneva, ISBN 978 92 4 159417 2.
22
Table 1. Descriptive statistics.
Variable
N
Mean
SD
Min
Max
Young
Most affected
Less affected
YoungXMost affected
YoungXLess affected
6504
6504
6504
6504
6504
0.52
0.16
0.48
0.08
0.25
0.5
0.36
0.5
0.27
0.44
0
0
0
0
0
1
1
1
1
1
Year
Male
Age
Age square
Year born
Rural
Big city
Moved in 1986
Moved after 1986
6504
6504
6504
6504
6504
6504
6504
6504
6504
2005.51
0.49
23.43
556.36
1982.09
0.21
0.56
0.07
0.54
1.72
0.5
2.76
129.64
2.15
0.41
0.5
0.26
0.5
2003
0
18
324
1979
0
0
0
0
2008
1
29
841
1985
1
1
1
1
Number of medical visits
Hospitalized
Not good health
University degree
Currently employed
Monthly wage in 2008$
6504
6504
6504
6504
5392
4443
0.67
0.09
0.47
0.14
0.84
175.3
1.35
0.28
0.5
0.35
0.37
138.63
0
0
0
0
0
0
20
1
1
1
1
1391.43
Note: Authors’ calculations based on the BHBS. The estimates were adjusted using the BHBS sampling
weights.
23
Table 2. The effect of radiation on health, education, and labor market outcomes.
Medical visits
Hospitalized
Not good
health
University
Employed
Log Wages
YoungXMost affected
1.112***
(0.039)
-0.002
(0.004)
0.070**
(0.024)
-0.025***
(0.006)
-0.041***
(0.003)
-0.055***
(0.014)
YoungXLess affected
1.036
(0.068)
-0.006
(0.007)
0.092**
(0.026)
-0.107
(0.061)
-0.057*
(0.028)
-0.029
(0.029)
Young
1.049
(0.145)
-0.001
(0.012)
-0.037
(0.034)
0.033
(0.050)
0.053**
(0.016)
0.128*
(0.074)
Most affected
1.236
(0.163)
0.013
(0.013)
0.107
(0.063)
0.033**
(0.013)
0.006
(0.013)
-0.056
(0.081)
Less affected
1.188
(0.150)
0.002
(0.015)
0.082
(0.062)
0.090
(0.053)
0.015
(0.022)
0.046
(0.143)
Age
0.768
(0.125)
0.002
(0.022)
-0.058
(0.032)
-0.028
(0.015)
-0.092
(0.048)
0.930***
(0.122)
Age square
1.006
(0.004)
-0.000
(0.000)
0.001
(0.001)
0.001**
(0.000)
0.002**
(0.001)
-0.017***
(0.003)
Rural
0.801***
(0.049)
-0.006
(0.009)
-0.017
(0.037)
-0.051**
(0.015)
-0.019
(0.011)
-0.299***
(0.071)
Big city
1.191*
(0.123)
-0.013
(0.011)
0.041
(0.050)
0.065***
(0.016)
-0.008
(0.015)
0.100
(0.079)
Male
0.576***
(0.023)
-0.055***
(0.010)
-0.043***
(0.011)
-0.045***
(0.009)
0.030***
(0.007)
0.407***
(0.028)
Moved in 1986
1.159
(0.120)
-0.011
(0.019)
-0.002
(0.020)
-0.014
(0.013)
0.016
(0.031)
-0.002
(0.132)
Moved after 1986
1.069
(0.080)
0.006
(0.010)
-0.012
(0.013)
0.012
(0.012)
0.029
(0.017)
0.101
(0.063)
Observations
R-squared
Pseudo R-squared
6504
6504
0.012
6504
0.031
6504
0.157
5392
0.039
4443
0.193
0.011
Note: Authors’ calculations based on the BHBS. Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01.
Standard errors in parentheses are adjusted for arbitrary heteroskedasticity and clustered by region. The
estimates were adjusted using the BHBS sampling weights. Incidence rate ratios from negative binomial
regression are reported for the outcome “Medical visits”, OLS coefficients are reported for the other
outcomes. Estimates for the outcome “Log Wages” were obtained from the Heckman (1979) selection
24
model. Actual number of non-missing wage observations is shown. Year dummy variables are included in
all regressions. See the text for the definition of all the variables.
Table 3. The effect of radiation on health, education, and labor market outcomes by gender.
Medical
visits
Hospitalized
Not good
health
University
Employed
Log
Wages
YoungXMost affected
1.272***
(0.082)
-0.011*
(0.005)
0.045***
(0.009)
-0.010**
(0.004)
-0.059***
(0.003)
-0.096***
(0.019)
YoungXLess affected
0.950
(0.089)
-0.012
(0.009)
0.074***
(0.012)
-0.076
(0.064)
-0.085*
(0.035)
-0.040
(0.032)
Observations
3174
3174
3174
3174
2644
2339
YoungXMost affected
1.014
(0.055)
0.006
(0.012)
0.097**
(0.039)
-0.033***
(0.009)
-0.021*
(0.009)
-0.002
(0.050)
YoungXLess affected
1.094
(0.085)
0.001
(0.021)
0.108*
(0.045)
-0.135*
(0.058)
-0.029
(0.029)
-0.012
(0.061)
Observations
3330
3330
3330
3330
2748
2104
A. Estimates for males
B. Estimates for females
Note: Authors’ calculations based on the BHBS. Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01.
Only main coefficients are displayed. However, all models also include all of the variables from Table 3
and year dummies. Standard errors in parentheses are adjusted for arbitrary heteroskedasticity and clustered
by region. The estimates were adjusted using the BHBS sampling weights. Incidence rate ratios from
negative binomial regression are reported for the outcome “Medical visits”, OLS coefficients are reported
for the other outcomes. Estimates for the outcome “Log Wages” were obtained from the Heckman (1979)
selection model. Actual number of non-missing wage observations is shown. See the text for the definition
of all the variables.
25
Figure 1. Degree of radioactive contamination and number of post-Chernobyl thyroid cancer
cases in children and adolescents in the regions of Belarus.
Source: Pacini et al. (1998).
Note: The numbers in the figure refer to the number of thyroid carcinoma cases registered in each region of
Belarus between May 1986 and December 1995 among individuals aged less than 21 years old. These
numbers represent 97.7% of all thyroid carcinoma cases. The percentages of cancer cases in each region in
the total number of cancer cases are indicated in parentheses.
Figure 2. Estimated average thyroid doses for the age group 0–18 years in districts of Belarus.
Source: Kruk et al. (2004), Figure 13.
Note: The estimates are derived from a radioecological model based on cesium-137 local area depositions
during the Chernobyl accident, the rainfall data for different regions of Belarus, and the iodine-131/cesium137 ratio in the deposit at the start of the grazing period in Belarus in April/May 1986.
26
Figure 3. Estimated average thyroid doses for the age group 0–18 years in all Belarus regions and
in the cities of Gomel and Minsk.
Source: Kruk et al. (2004), Figure 14.
Note: The estimates are derived from a radioecological model based on cesium-137 local area depositions
during the Chernobyl accident, the rainfall data for different regions of Belarus, and the iodine-131/cesium137 ratio in the deposit at the start of the grazing period in Belarus in April/May 1986.
27
Figure 4. Trends in the outcomes across single-year age in 1986, by area of exposure to
radiation.
Average percent with university degree
0
.5
.1
University degree
.2
.3
Number of medical visits
.6
.7
.8
.9
1
.4
Average number of medical visits
0
1
2
3
4
5
6
Most affected
7
8
9
Age in 1986
10
11
Less affected
12
13
14
15
0
1
2
Least affected
3
4
5
6
Most affected
10
11
12
Less affected
13
14
15
Least affected
Average probability of being employed
.06
.6
.08
Hospitalized
.1
.12
Currently employed
.7
.8
.14
.9
.16
Average probability of being hospitalized
7
8
9
Age in 1986
0
1
2
3
4
5
6
Most affected
7
8
9
Age in 1986
10
11
Less affected
12
13
14
15
0
1
2
Least affected
3
4
5
6
Most affected
10
11
12
Less affected
13
14
15
Least affected
Average monthly wage in USD (base year 2000)
.3
100
Not good health
.4
.5
Monthly wage in 2008$
150
200
.6
250
Average probability of not being in good health
7
8
9
Age in 1986
0
1
2
3
4
Most affected
5
6
7
8
9
Age in 1986
Less affected
10
11
12
13
14
15
Least affected
0
1
2
3
4
Most affected
5
6
7
8
9
Age in 1986
10
11
12
Less affected
Note: Authors’ calculations based on the BHBS. “Most affected” area includes Gomel region; “Less
affected” area includes Brest, Mogilev regions and Minsk city; and “Least affected” area includes Vitebsk,
Grodno, and Minsk regions.
28
13
14
Least affected
15
Appendix
Table A1. Falsification test of the effect of radiation using two older age groups.
Medical
visits
Hospitalized
Not good
health
University
Employed
Log
Wages
YoungXMost affected
0.835
(0.131)
-0.005
(0.022)
-0.073**
(0.037)
0.028
(0.030)
0.024
(0.025)
-0.075
(0.069)
YoungXLess affected
1.188
(0.142)
0.004
(0.016)
-0.036
(0.028)
0.015
(0.024)
0.038**
(0.019)
-0.010
(0.065)
Observations
R-squared
Pseudo R-squared
6263
6263
0.015
6263
0.030
6263
0.079
6036
0.009
4891
0.166
0.012
Note: Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01. Only main coefficients are displayed.
However, all models also include all of the variables from Table 3 and year dummies. Standard errors in
parentheses are adjusted for arbitrary heteroskedasticity and clustered by region. The estimates were
adjusted using the BHBS sampling weights. Incidence rate ratios from negative binomial regression are
reported for the outcome “Medical visits”, OLS coefficients are reported for the other outcomes. Estimates
for the outcome “Log Wages” were obtained from the Heckman (1979) selection model. Actual number of
non-missing wage observations is shown. Year dummy variables are included in all regressions. See the
text for the definition of all the variables.
Table A2. Differences in the outcomes of migrants between affected and unaffected regions.
Born in Gomel and
out-migrated from
Gomel after 1986
Born in Brest,
Mogilev, or Minsk
city and out-migrated
after 1986
Observations
R-squared
University
degree
Employed
Professional
occupation
White
collar
occupation
Married
Number
of
children
-0.045
0.006
-0.007
-0.001
0.033
-0.029
(0.027)
(0.031)
(0.030)
(0.001)
(0.023)
(0.071)
-0.029
-0.082*
-0.029
-0.002
-0.055
-0.225**
(0.019)
(0.037)
(0.027)
(0.001)
(0.029)
(0.085)
63572
0.010
63572
0.006
63572
0.006
63572
0.001
63572
0.018
36329
0.005
Note: Authors’ calculations based on the Belarusian Population Census of 1999. Significance levels: * p <
0.1, ** p < 0.05, *** p < 0.01. Standard errors in parentheses are adjusted for arbitrary heteroskedasticity
and clustered by region. The estimates are obtained from linear regressions, where other regressors include
dummies “Born in Gomel and out-migrated from Gomel”, “Born in Brest, Mogilev, or Minsk and outmigrated” and “Migrated after 1986”. Estimation sample includes only individuals who out-migrated from
their region of birth. Question about the number of children was only asked to females.
29
Figure A1. Population trends by region and by area of exposure to radiation, 1969-2009.
Note: Data for 1969-1999 is taken from the Statistical Yearbook of the Republic of Belarus, ed. 2002. Data
for 2009 is taken from the National Statistical Committee of the Republic of Belarus web-site
http://belstat.gov.by/homep/en/main.html, accessed on 05.25.2011. The vertical dashed line marks year
1986.
30