## Abstract

We examine the effect of radiation exposure in utero, resulting from nuclear weapon testing in the 1950s and early 1960s, on long-run outcomes of Norwegian children. Exposure to low-dose radiation, specifically during months 3 and 4 in utero, leads to lower IQ scores for men and lower education attainment and earnings among men and women. Children of persons affected in utero also have lower cognitive scores, suggesting a persistent intergenerational effect of the shock to endowments. Given the lack of awareness about nuclear testing in Norway at this time, our estimates are likely unaffected by avoidance behavior or stress effects.

## I.  Introduction

A large literature documents substantial persistence in early childhood endowments. Increasingly, the evidence shows that differences in endowments at birth need not be genetic but instead are influenced by environmental factors while the fetus is in the womb, and these differences can persist into adult life.1 However, it is often difficult to identify the effect of exposure to environmental factors on individuals across the income spectrum, as these factors tend to disproportionately affect those at the bottom of the income distribution who are unable to move to avoid them. There is also little compelling evidence on the importance of the specific time during the pregnancy the mother is exposed, which provides valuable information on the mechanisms that underlie these effects. Finally, there is very little evidence on the longer-run intergenerational consequences of environmentally induced shocks to endowments.2

In this paper, we address these gaps in understanding by studying the effects of one such environmental factor—exposure to radiation—on both the long-run outcomes of exposed children as well as the IQ scores of their adult offspring. To do so, we use variation in radioactive exposure throughout Norway in the 1950s and early 1960s resulting from the extensive nuclear testing during that period. Norway provides an ideal laboratory for this type of analysis; because of its geographical location and topography, with high precipitation in coastal areas, Norway received considerable radioactive fallout from Russian atmospheric nuclear weapons tests in the 1950s and 1960s (Hvinden & Lillegraven, 1961). Unlike other pollutants that disproportionately affect one part of society, nuclear exposure affected members of all socioeconomic groups. Regional fallout was determined by many factors, including wind, rainfall, and topography; we use the variation in fallout across Norway and over time for identification.

We build on previous work that examines the effect of exposure to pollution in utero on the long-run outcomes of children. Most closely related is work by Almond, Edlund, and Palme (2009), who find that low-dose exposure in utero leads to lower grades in school.3 Our paper adds to this research on a number of dimensions, including studying effects on completed education and adult earnings. Also, because we incorporate both cross-sectional and time-series variation in exposure over a longer period of time, we are able to verify the conclusions drawn by the medical literature (with much smaller samples) by documenting that it is at months 3 and 4 of pregnancy, when brain development is most sensitive to radiation, that fetuses are most affected by exposure. Finally, we also add to the literature by studying adult cognitive scores of the second generation (the children of the generation affected in utero), thus identifying the intergenerational effects of exposure to radiation.

We find that exposure to nuclear fallout in the air or on the ground, even in low doses, leads to a decline in men's IQ scores at age 18, completed years of education, and earnings at age 35. Among women, radiation exposure leads to declines in educational attainment and high school completion and lower earnings at age 35. Additionally, there is evidence that the children of the affected generation also have lower cognitive scores, and we estimate the degree of intergenerational transmission. These results are robust to the choice of specification, tests of selection, and the inclusion of sibling fixed effects.

Unlike the nuclear bombings of Hiroshima and Nagasaki in 1945, Chernobyl, and the accident in the nuclear power plant in Fukushima in 2012, there was very little public awareness in Norway of the exposure to nuclear fallout resulting from nuclear testing taking place in foreign countries.4 Moreover, the first medical studies analyzing the effect of nuclear fallout on cognitive achievement were published only in the 1980s (Otake & Schull, 1984). Therefore, there is no reason to expect that avoidance behavior is important. As a result, we can study whether, conditional on exposure, better-educated parents are better able to mitigate the negative effects of exposure. Additionally, it implies that the effects we find cannot be explained as resulting from stress due to worry about the effects of radiation.5

The paper unfolds as follows. Section II describes the relevant history of nuclear testing affecting Norway. Section III describes our empirical strategy, and section IV describes our data. Section V presents the results for the effects of exposure on children in utero, along with a variety of robustness checks, and section VI presents results for the second generation—the children of these children. Section VII concludes.

## II.  Background

### A.  Nuclear Testing

There was intensive nuclear weapon testing worldwide in the periods 1952 to 1954, 1957 to 1958, and 1961 to 1962, with deposition rates peaking in 1963.6 An atmospheric nuclear weapon test produces about 150 fission products with half-lives long enough to contribute to radioactive fallout. In general, the fallout can be divided into three components: (a) large particles that are deposited from the atmosphere within hours of the test, (b) smaller particles that remain in the troposphere only a few days, and (c) longer-lived particles such as cesium (CS-137), strontium (Str-90), rubidium (Ru-103), xenon (Xe-133), iodine (I-131), and barium (Ba-140), which are injected into the stratosphere (Bergan, 2002). Radioactive debris injected into the stratosphere—so-called global fallout—trickles down slowly to the troposphere; from there, debris is deposited on the ground mainly through precipitation. Differences in the rate of deposition across locations can thus partly be explained by temporal and spatial variation in precipitation.

Immediately following a nuclear explosion, the activity of short-lived radionuclides is much greater than that of long-lived radionuclides. However, the short-lived radionuclides such as iodine-131, with a half-life of eight days, decay substantially during the time it takes the fallout cloud to reach distant locations like Norway, and the more long-lived radionuclides, such as ruthenium-103 and Zirconium-95, with half-lives of several months, become relatively more important. In the polar region (which includes most of Norway), radionuclides remain in the stratosphere on average from three to twelve months (UNSCEAR, 1982). Bergan (2002) estimates the average age of the fallout in Norway to be between three and five months during the intensive testing periods.

The western Norwegian coastline was particularly exposed to atomic fallout coming from nuclear testing taking place in Novaya Zemlya in the Russian arctic archipelago, one of the most intense test regions between 1955 and 1962. The macro weather system is the primary force that moved long-lived radionuclides from Russian test stations to their ultimate deposition along the Norwegian coast: cold air over the poles creates high-pressure zones, taking the air to lower latitudes.7

Figure 1 shows estimates of the in situ total beta fallout in each municipality in Norway in 1958, 1960, 1962, and 1964.8 The activity of fallout in the air or on the ground or other surfaces is measured in becquerels (Bq), which is defined as the number of radioactive disintegrations per second. The fallout varies significantly by municipality (partly due to the much higher rainfall levels in western compared to eastern Norway) and also over time. There was an international moratorium on nuclear testing from November 1958 to September 1961, so Norway received almost no fallout in the second half of 1959, in 1960, and throughout most of 1961. The partial test ban treaty in October 1963 led to very little fallout in 1964 and subsequent years. However, there is significant fallout in 1957 and 1958 and, even more so, in 1962 and 1963 because the explosions after the expiration of the moratorium were much larger than before. This results in substantial time series variation in addition to that across municipalities.

Figure 1.

Total Beta Fallout in Situ per Community in 1958, 1960, 1962, and 1964

In situ total beta fallout is measured in pCi/m$2$. The figures reflect the testing activity. In 1960, a moratorium on nuclear testing honored by the United States, United Kingdom, and Soviet Union was in place. The moratorium broke down in 1961 and testing resumed. After October 1963 a partial test ban treaty prohibited all but underground detonations.

Figure 1.

Total Beta Fallout in Situ per Community in 1958, 1960, 1962, and 1964

In situ total beta fallout is measured in pCi/m$2$. The figures reflect the testing activity. In 1960, a moratorium on nuclear testing honored by the United States, United Kingdom, and Soviet Union was in place. The moratorium broke down in 1961 and testing resumed. After October 1963 a partial test ban treaty prohibited all but underground detonations.

### B.  Prenatal Radiation Exposure and Cognitive Damage

Following the deposition of fallout into the air and on the ground, there are different ways in which people absorb radiation. Radiation might come from penetrating gamma rays emitted by particles in the air and on the ground. In this case, simply staying inside a building reduces exposure. Moreover, people inhale fallout or absorb it through skin. A further source is the consumption of contaminated food. Vegetation can be contaminated when fallout is directly deposited on the surface of plants, or when it is deposited on the ground and plants absorb it through their roots. People can also be exposed when they eat meat and drink milk from animals grazing on contaminated vegetation or if they drink contaminated water.

It is well established that ionizing radiation can lead to molecular, cellular, and tissue damage in utero (Hall, 2009). While formation of most human organs is largely complete by the eighth week after conception, the development of the cerebral cortex occurs rapidly from weeks 8 to 15 postconception, and medical evidence suggests prenatal exposure to ionizing radiation is particularly harmful to the brain if it occurs during this two-month period (Otake & Schull, 1984). At higher radiation doses, however, prenatal exposure to ionizing radiation during weeks 16 to 25 postconception have also been found to be harmful (Otake & Schull, 1998; Nowakowski & Hayes, 2008; Donnelly et al., 2011). After the 25th week, the central nervous system becomes quite resistant to radiation, and major fetal brain damage becomes unlikely (Otake & Schull, 1998).9

The first studies indicating that iodizing radiation causes cognitive abnormalities were analyses of individuals exposed in utero to diagnostic X-ray procedures in the 1980s (Brent, 1989). However, most early evidence on the effects of acute exposure to ionizing radiation has been obtained from studies on the survivors of the atomic bombs at Hiroshima and Nagasaki. These survivor studies are limited in that they analyze the effects of a single, relatively high dose and not of small, intermittent, or continual doses typical of medical, professional, or environmental exposure. Studies evaluating the impact of smaller doses of radiation, such as those experienced in Sweden after the reactor incident in Chernobyl, on early health outcomes such as spontaneous abortion, stillbirth, length of gestation, birth weight, and neonatal mortality, are not conclusive. Some find effects of prenatal exposure, while others do not (Lüning et al., 1989; Sperling et al., 1994; Auvinen et al., 2001). However, studies focusing on cerebral dysfunctions do suggest that the prenatal exposure to radioactive fallout after Chernobyl resulted in detectable brain damage or lower schooling performance and fetal death (Almond et al., 2009, Halla & Zweimüller, 2014). Importantly, unlike most of the literature, we can confirm that even with low-dose exposure, radiation in months 3 and 4 of the pregnancy is particularly relevant.10

## III.  Empirical Strategy

We use a similar approach to that in the Chernobyl study of Almond et al. (2009) but incorporate the fact that we have variation over a relatively long period of time as well as across space. The amount of fallout experienced by any individual depends on that person's month of birth, year of birth, and municipality of birth.

### A.  Basic Specification

We estimate the following equation:
$Hict=α0+α1Fct+βXit+γt+λc+εict.$
(1)
$Hict$ represents outcomes such as education, IQ score, height, and earnings for child $i$ born in municipality $c$ at time $t$. We use the same specification in the intergenerational analysis where $Hict$ then represents the outcomes of the offspring of the exposed children. $F$ is a vector indicating fallout in each month of pregnancy, beginning three months before conception and ending three months postbirth. $X$ is a vector of controls that includes parental education, the county-level unemployment rate when the child was in utero, and birth order indicators (family size at birth of child). We also include controls for year of birth by month-of-birth indicators ($γt)$ and municipality fixed effects ($λc)$. We are therefore comparing individuals born within the same municipality but born in a different month or year of birth (and thus exposed to different levels of radiation in utero). We use OLS estimation. In the case of high school completion, we estimate a linear probability model.

### B.  Specification with Municipality-Specific Trends

One concern might be that our results are driven by different trends in municipalities that are exposed to high doses of radiation relative to those that are not. While we found no evidence of differential trends in the observable control variables during this time period, we also report estimates from a specification that allows for municipality-specific linear trends. These trends are included in addition to the year of birth by month-of-birth indicators.

### C.  Specification with Interactions

As a further robustness check, we also estimate a richer model that adds interactions of the municipality dummies with month of birth (to allow for seasonal factors that differ by area) and interactions of the municipality dummies with year of birth (to allow cohort effects to differ by municipality). Note that we cannot include the interaction of year of birth by month of birth by municipality, as that is our identifying variation. Letting $y$ denote year of birth and $m$ denote month of birth, we estimate
$Hict=α0+α1Fct+βXit+γt+λcy+μcm+εict.$
(2)
This model is still identified, as there is much variation in fallout over the course of any particular year that is not seasonal but determined by the timing of nuclear tests in the Soviet Union.

### D.  Sibling Fixed-Effects Model

While exposure is arguably exogenous to family and neighborhood characteristics within municipalities, one might still worry that nonrandom migration might change the composition of people in the municipality over time. Furthermore, the composition of the sample could be correlated with the fallout if there are changes over time and region in the types of people who give birth and these are, by chance, correlated with fallout levels. While we have no evidence that this is the case, we also estimate a specification that includes sibling fixed effects. Identifying variation is then differences in exposure within families across children, thereby differencing out anything that is constant within families, such as socioeconomic status.

### E.  Multiple Hypothesis Testing

In each regression, we estimate coefficients on the level of fallout by month from three months before conception to three months after birth. Therefore, we estimate sixteen separate coefficients relating to the timing of fallout. This makes it likely that one or more of these estimates will be statistically significant by chance alone. However, as by Romano, Shaikh, and Wolf (2010) noted, this is not a problem if one is focusing on a particular hypothesis a priori. In this case, the decision can still be based on the corresponding marginal $p$-value. A problem arises only if one searches the list of $p$-values for significant results after the fact. Our situation is somewhat intermediate. We know from the medical literature that effects are most likely from exposure in months 3 and 4, so we are not searching for significant results a posteriori. However, we are also open to the possibility that exposure in other months in utero could have effects on child outcomes. To be conservative, in addition to standard hypothesis testing, we also report whether the statistical significance of our coefficients survives a correction for multiple hypothesis testing.11 The method we use is the Romano-Wolf step-down method (Romano & Wolf, 2005), which maintains the probability of making a type 1 error at approximately the designated significance level. We provide the details about this approach in the online appendix.

## IV.  Data

Our primary data source is the Norwegian Registry Data, a linked administrative data set that covers the population of Norwegians up to 2009.12 We include data for cohorts born 1956 to 1966. Using month and year of birth and assuming pregnancy lasts 266 days, we can identify the months of pregnancy (ranging from one to nine). We allocate a municipality to each child born between 1956 and 1964 using the 1960 census by assuming that the municipality during pregnancy is the mother's municipality of residence in 1960. For individuals born in 1965 and 1966, we can use register data on the exact municipality where the mother lived when the child was born.

### A.  Data on Outcomes

The IQ score and height data are taken from Norwegian military records. In Norway, military service is compulsory for every male; as a result, we have military data for men only, and measurement is at around age 18 or 19. We measure educational attainment in 2009 and use two measures of education achievement. High school graduation is an indicator equal to 1 if the child obtained a three-year high school diploma. We also consider the years of education completed by the individual. In the online appendix, we show that estimates are similar when we study two other variables: whether the individual has attended high school or college. Earnings are measured as gross annual earnings for taxable income as reported in the tax registry when the individual is aged 35 (Bhuller, Mogstad, & Salvanes, 2017, show that earnings in early to mid-30s predicts lifetime income well in Norway). These are not top-coded and include labor earnings, taxable sick benefits, unemployment benefits, parental leave payments, and pensions.

### B.  Data on Nuclear Fallout

In the period from 1956 to 1984, the Norwegian Defense Research Establishment (FFI) monitored radioactivity in the air and on the ground at thirteen stations across Norway.13 They collected two primary measures of radiation: (a) a measure of the total beta radiation in the air expressed as Bq/m$3$ and (b) a measure of the total beta radiation in situ (ion the ground) expressed in Bq/m$2$.14 These two measures of deposition (air and ground) have a correlation coefficient of 0.75, implying that they are highly—but far from perfectly—correlated. Figures 2a and 2b show the two measures for Oslo and Bergen. One can see that the temporal pattern differs for the two measures. This is not surprising, as ground deposition is largely determined by rainfall, while fallout in the air is more related to the presence of centers of high air pressure as well as influxes of warm subtropical air (Bergan & Steenhuisen, 2012).

Figure 2.

Monthly Total Beta Fallout in Situ and Air Exposure

In situ total beta fallout is measured in Bq/m$2$ (left $y$-axis). Total beta fallout in the air is measured in Bq/m$3$ (right $y$-axis). The figures reflect the testing activity: in 1960, a moratorium on nuclear testing honored by the United States, United Kingdom, and Soviet Union is in place. The moratorium broke down in 1961 and testing resumed. After October 1963, a partial test ban treaty prohibited all but underground detonations.

Figure 2.

Monthly Total Beta Fallout in Situ and Air Exposure

In situ total beta fallout is measured in Bq/m$2$ (left $y$-axis). Total beta fallout in the air is measured in Bq/m$3$ (right $y$-axis). The figures reflect the testing activity: in 1960, a moratorium on nuclear testing honored by the United States, United Kingdom, and Soviet Union is in place. The moratorium broke down in 1961 and testing resumed. After October 1963, a partial test ban treaty prohibited all but underground detonations.

To minimize the measurement error in our measure of nuclear fallout, we limit our sample to municipalities within 20 kilometers of a test station (46 municipalities). We have tested the sensitivity of our results to different distance cutoffs (15, 25, and 30 kilometers) and find the results are quite insensitive to this choice (estimates are reported in the online appendix).

For radiation in the air, we estimate the fallout for each municipality in our sample in each month by using the fallout at the geographically closest measuring station. For radiation on the ground, we estimate the fallout for each municipality in each month by using the fallout at the geographically closest measuring station and then weight that by the precipitation in that month in the municipality relative to the precipitation in that month at the measuring station. This is equivalent to
$Fct=FstPctPst,$
(3)
where $Fct$ measures the nuclear fallout in municipality $c$ at time $t$ and $Fst$ represents the nuclear fallout at the closest test station $s$ at time $t$. $Pit$ measures the precipitation in month $t$ in municipality $c$ or $s$. The reweighting implies that there will be more fallout in areas of relatively heavier rain and less in areas of relatively less rain.15

Summary statistics for our first-generation sample are presented in table 1a, along with descriptive statistics for the whole country. Because our sample is disproportionately urban, education levels are higher in our sample than in the country as a whole. Table 1b presents summary statistics for the sons of the exposed generation (second generation) and describes the sample we use to analyze whether our findings persist into the next generation.

Table 1.
Summary Statistics
A. Summary Statistics, Exposed Generation
Men 20 Kilometer SampleMen AllWomen 20 Kilometer SampleWomen All
MeanStandard DeviationMeanStandard DeviationMeanStandard DeviationMeanStandard Deviation
Control variables
Father high school completed 0.424 0.495 0.301 0.458 0.421 0.494 0.296 0.457
Mother high school completed 0.328 0.469 0.229 0.420 0.321 0.467 0.225 0.418
Unemployment rate at birth 0.010 0.008 0.013 0.008 0.010 0.008 0.013 0.009
Birth year 1961 2.944 1961 3.152 1961 2.989 1961 3.187
Radioactive fallout (months 3 and 4)
Mean monthly; total beta air (Bq/m$3)$ 0.042 0.059   0.042 0.059
Total month; total beta ground (Bq/m$2)$ 2,532 3,789   2,537 3,810
Outcome variables
IQ at age 18 (scale: 1–9) 5.260 1.996 5.010 1.998
Height at age 18 in cm 179.7 6.379 179.4 6.387
Years of education 12.334 2.603 12.11 2.482 12.37 2.658 12.15 2.591
High school completed 0.732 0.443 0.714 0.452 0.688 0.463 0.653 0.476
Earnings at age 35 (in NOK) 150,614 110,081 140,258 102,672 84,048 60,226 78,658 55,350
Observations 95,280 297,889 96,288 305,252
B. Summary Statistics, Exposed Generation's Sons
Fathers 20 Kilometer Sample Women 20 Kilometer Sample
Mean Standard Deviation Mean Standard Deviation
Control variables
Grandfather high school completed 0.420 0.494 0.406 0.491
Grandmother high school completed 0.321 0.467 0.305 0.461
Unemployment rate at birth of father 0.010 0.009 0.010 0.009
Birth year of father 1961 2.871 1961 2.879
Birth year 1985 3.191 1984 3.710
Number of siblings 1.684 0.996 1.663 0.976
Father or mothers's exposure to radioactive fallout (months 3 and 4)
Mean monthly; total beta air (Bq/m$3)$ 0.043 0.059 0.042 0.055
Total month; total beta ground (kBq/m$2)$ 2.937 3.743 2.882 3.353
Father's IQ
IQ at age 18 (scale: 1–9) 5.760 1.809
Outcome variable
IQ at age 18 (scale: 1–9) 5.014 1.682 5.054 1.690
Observations 24,281 36,947
A. Summary Statistics, Exposed Generation
Men 20 Kilometer SampleMen AllWomen 20 Kilometer SampleWomen All
MeanStandard DeviationMeanStandard DeviationMeanStandard DeviationMeanStandard Deviation
Control variables
Father high school completed 0.424 0.495 0.301 0.458 0.421 0.494 0.296 0.457
Mother high school completed 0.328 0.469 0.229 0.420 0.321 0.467 0.225 0.418
Unemployment rate at birth 0.010 0.008 0.013 0.008 0.010 0.008 0.013 0.009
Birth year 1961 2.944 1961 3.152 1961 2.989 1961 3.187
Radioactive fallout (months 3 and 4)
Mean monthly; total beta air (Bq/m$3)$ 0.042 0.059   0.042 0.059
Total month; total beta ground (Bq/m$2)$ 2,532 3,789   2,537 3,810
Outcome variables
IQ at age 18 (scale: 1–9) 5.260 1.996 5.010 1.998
Height at age 18 in cm 179.7 6.379 179.4 6.387
Years of education 12.334 2.603 12.11 2.482 12.37 2.658 12.15 2.591
High school completed 0.732 0.443 0.714 0.452 0.688 0.463 0.653 0.476
Earnings at age 35 (in NOK) 150,614 110,081 140,258 102,672 84,048 60,226 78,658 55,350
Observations 95,280 297,889 96,288 305,252
B. Summary Statistics, Exposed Generation's Sons
Fathers 20 Kilometer Sample Women 20 Kilometer Sample
Mean Standard Deviation Mean Standard Deviation
Control variables
Grandfather high school completed 0.420 0.494 0.406 0.491
Grandmother high school completed 0.321 0.467 0.305 0.461
Unemployment rate at birth of father 0.010 0.009 0.010 0.009
Birth year of father 1961 2.871 1961 2.879
Birth year 1985 3.191 1984 3.710
Number of siblings 1.684 0.996 1.663 0.976
Father or mothers's exposure to radioactive fallout (months 3 and 4)
Mean monthly; total beta air (Bq/m$3)$ 0.043 0.059 0.042 0.055
Total month; total beta ground (kBq/m$2)$ 2.937 3.743 2.882 3.353
Father's IQ
IQ at age 18 (scale: 1–9) 5.760 1.809
Outcome variable
IQ at age 18 (scale: 1–9) 5.014 1.682 5.054 1.690
Observations 24,281 36,947

## V.  Results-First Generation

We first present the results for IQ scores for men using the two different measures of radiation exposure (in separate regressions), the beta radiation from the air and the in situ, or ground, radiation. For ease of interpretation, when included in regressions, both measures of radioactive exposure are standardized to have mean 0 and variance 1.

Table 2 presents the results for men when the IQ score is the outcome. Each column is a separate regression that includes the standardized measure of exposure in each month of pregnancy, in the three months before conception and the three months after birth. The first four columns present the results using in situ exposure, and the second four columns use air exposure. As the IQ score is taken from the Norwegian military records and military service is compulsory only for men, this analysis is restricted exclusively to men. We cluster the standard errors by municipality and so allow arbitrary correlations of the error terms for people born in the same municipality. However, we also tested the sensitivity of our conclusions to various assumptions about the standard errors and found them to be quite robust (these estimates are reported in the online appendix). We verified as well that the results are robust to the exclusion of any individual measuring station, a potential concern given the existence of only thirteen stations in our sample.

Table 2.
Effect of Fallout by Month on Men's IQ: Total Beta Fallout in Situ and in the Air
In SituAir
BaselineMunicipality-Specific TrendsFully SaturatedSibling Fixed EffectsBaselineMunicipality-Specific TrendsFully SaturatedSibling Fixed Effects
Three months prior to 0.003 0.003 0.001 0.002 −0.015 −0.013 −0.014 −0.014*
pregnancy (0.006) (0.015) (0.005) (0.006) (0.019) (0.019) (0.011) (0.008)
Two months prior to −0.012 −0.011 −0.010 0.001 −0.027* −0.035 −0.014 −0.025
pregnancy (0.008) (0.009) (0.009) (0.006) (0.014) (0.024) (0.016) (0.020)
One month prior to −0.015 −0.014 −0.014 −0.016* −0.063 −0.060 −0.052* −0.055
pregnancy (0.016) (0.018) (0.016) (0.008) (0.052) (0.062) (0.031) (0.046)
Pregnancy −0.018* −0.016* −0.021* −0.012 −0.038 −0.040 −0.055 −0.050
month 1 (0.010) (0.009) (0.012) (0.010) (0.028) (0.038) (0.042) (0.032)
Pregnancy −0.020* −0.018* −0.016 −0.030 −0.078* −0.063* −0.074* −0.061*
month 2 (0.011) (0.010) (0.010) (0.019) (0.046) (0.033) (0.044) (0.035)
Pregnancy −0.039*** −0.038*** −0.054*** −0.044*** −0.127*** −0.142*** −0.142*** −0.109**
month 3 (0.009) (0.008) (0.015) (0.010) (0.052) (0.069) (0.061) (0.050)
Pregnancy −0.043*** −0.041*** −0.058*** −0.035*** −0.093** −0.112*** −0.125*** −0.102*
month 4 (0.011) (0.010) (0.016) (0.011) (0.047) (0.058) (0.050) (0.051)
Pregnancy 0.005 0.008 −0.014** −0.010** −0.046* −0.052 −0.021 −0.051
month 5 (0.005) (0.005) (0.007) (0.005) (0.027) (0.038) (0.029) (0.031)
Pregnancy 0.010 0.012* 0.011 0.014* −0.083 −0.076* −0.063 −0.05
month 6 (0.006) (0.006) (0.008) (0.008) (0.064) (0.040) (0.098) (0.032)
Pregnancy 0.013 0.007 0.015** 0.001 −0.050 −0.053 0.063 0.6
month 7 (0.015) (0.005) (0.006) (0.009) (0.067) (0.036) (0.048) (0.041)
Pregnancy −0.014 −0.010 −0.014 −0.002 −0.043 −0.051 −0.042 −0.046
month 8 (0.009) (0.008) (0.011) (0.009) (0.031) (0.032) (0.040) (0.051)
Pregnancy −0.007 −0.003 −0.003 −0.002 −0.019 −0.017 −0.026 −0.029
month 9 (0.007) (0.027) (0.008) (0.002) (0.013) (0.019) (0.019) (0.023)
Month of birth 0.005 0.012* 0.008 0.003 −0.057* 0.061 0.052* 0.032
(0.004) (0.007) (0.005) (0.003) (0.030) (0.041) (0.031) (0.022)
After pregnancy 1 0.007 −0.001 −0.004 −0.015* 0.037 0.035 0.027 0.023
(0.006) (0.007) (0.007) (0.009) (0.029) (0.027) (0.019) (0.020)
After pregnancy 2 −0.012* −0.006 −0.009 −0.000 −0.035 0.030* −0.042 0.031
(0.007) (0.007) (0.007) (0.006) (0.038) (0.017) (0.031) (0.022)
After pregnancy 3 −0.009 −0.003 −0.005 −0.007 −0.021 −0.027 −0.020 −0.022
(0.006) (0.006) (0.008) (0.009) (0.078) (0.019) (0.018) (0.019)
Observations 89,892 89,892 89,892 54,164 88,446 88,446 88,446 53,292
In SituAir
BaselineMunicipality-Specific TrendsFully SaturatedSibling Fixed EffectsBaselineMunicipality-Specific TrendsFully SaturatedSibling Fixed Effects
Three months prior to 0.003 0.003 0.001 0.002 −0.015 −0.013 −0.014 −0.014*
pregnancy (0.006) (0.015) (0.005) (0.006) (0.019) (0.019) (0.011) (0.008)
Two months prior to −0.012 −0.011 −0.010 0.001 −0.027* −0.035 −0.014 −0.025
pregnancy (0.008) (0.009) (0.009) (0.006) (0.014) (0.024) (0.016) (0.020)
One month prior to −0.015 −0.014 −0.014 −0.016* −0.063 −0.060 −0.052* −0.055
pregnancy (0.016) (0.018) (0.016) (0.008) (0.052) (0.062) (0.031) (0.046)
Pregnancy −0.018* −0.016* −0.021* −0.012 −0.038 −0.040 −0.055 −0.050
month 1 (0.010) (0.009) (0.012) (0.010) (0.028) (0.038) (0.042) (0.032)
Pregnancy −0.020* −0.018* −0.016 −0.030 −0.078* −0.063* −0.074* −0.061*
month 2 (0.011) (0.010) (0.010) (0.019) (0.046) (0.033) (0.044) (0.035)
Pregnancy −0.039*** −0.038*** −0.054*** −0.044*** −0.127*** −0.142*** −0.142*** −0.109**
month 3 (0.009) (0.008) (0.015) (0.010) (0.052) (0.069) (0.061) (0.050)
Pregnancy −0.043*** −0.041*** −0.058*** −0.035*** −0.093** −0.112*** −0.125*** −0.102*
month 4 (0.011) (0.010) (0.016) (0.011) (0.047) (0.058) (0.050) (0.051)
Pregnancy 0.005 0.008 −0.014** −0.010** −0.046* −0.052 −0.021 −0.051
month 5 (0.005) (0.005) (0.007) (0.005) (0.027) (0.038) (0.029) (0.031)
Pregnancy 0.010 0.012* 0.011 0.014* −0.083 −0.076* −0.063 −0.05
month 6 (0.006) (0.006) (0.008) (0.008) (0.064) (0.040) (0.098) (0.032)
Pregnancy 0.013 0.007 0.015** 0.001 −0.050 −0.053 0.063 0.6
month 7 (0.015) (0.005) (0.006) (0.009) (0.067) (0.036) (0.048) (0.041)
Pregnancy −0.014 −0.010 −0.014 −0.002 −0.043 −0.051 −0.042 −0.046
month 8 (0.009) (0.008) (0.011) (0.009) (0.031) (0.032) (0.040) (0.051)
Pregnancy −0.007 −0.003 −0.003 −0.002 −0.019 −0.017 −0.026 −0.029
month 9 (0.007) (0.027) (0.008) (0.002) (0.013) (0.019) (0.019) (0.023)
Month of birth 0.005 0.012* 0.008 0.003 −0.057* 0.061 0.052* 0.032
(0.004) (0.007) (0.005) (0.003) (0.030) (0.041) (0.031) (0.022)
After pregnancy 1 0.007 −0.001 −0.004 −0.015* 0.037 0.035 0.027 0.023
(0.006) (0.007) (0.007) (0.009) (0.029) (0.027) (0.019) (0.020)
After pregnancy 2 −0.012* −0.006 −0.009 −0.000 −0.035 0.030* −0.042 0.031
(0.007) (0.007) (0.007) (0.006) (0.038) (0.017) (0.031) (0.022)
After pregnancy 3 −0.009 −0.003 −0.005 −0.007 −0.021 −0.027 −0.020 −0.022
(0.006) (0.006) (0.008) (0.009) (0.078) (0.019) (0.018) (0.019)
Observations 89,892 89,892 89,892 54,164 88,446 88,446 88,446 53,292

The sample includes persons born between 1956 and 1966 and includes municipalities within a radius of 20 kilometers of the test stations. “Total beta in air” refers to air deposition measured in kBq/m$3$. “Total beta in situ” refers to ground deposition measured in kBq/m$2$. The fallout measures are standardized to mean 0 and standard deviation 1. Each column represents a separate regression that includes controls for municipality dummies and year of birth by month of birth dummies. Also included in each specification are controls for parental education, birth order, and the municipality unemployment rate. Standard errors are clustered at the municipality level. Significant at $***$1%, $**$5%, $*$10%. Coefficients in italics survive a correction for multiple hypothesis testing on a 10% significance level.

Columns 1 and 5 present results from our basic specification that controls for municipality and month of birth by year-of-birth fixed effects. Columns 2 and 6 then show the results when we add municipality-specific time trends. Columns 3 and 7 present the results from the most saturated model, including municipality-specific month of birth and municipality-specific year of birth controls. In all specifications, we find that radioactive exposure in months 3 and 4 of pregnancy, even the relatively small doses experienced in Norway from the Russian nuclear testing in the 1950s and 1960s, appears to have a significant negative effect on the IQ score of exposed males. This is true regardless of the measure of exposure that we use. Conditional on ground radiation in other months, a 1 standard deviation increase in ground exposure in month 3 or 4 leads to a decline in the IQ score of about 0.04. Given the standard deviation of the IQ score is about 2, this is an effect size of about 0.02 of a standard deviation. The effect of air exposure is larger, with a 1 standard deviation increase in exposure in month 3 or 4 leading to about 0.06 of a standard deviation fall in the IQ score. This is equivalent to about 0.8 of an IQ point on a standard IQ scale.16

To get a better sense of the magnitudes of these effects, one can compare the magnitude of fallout in Bergen, a heavily exposed city, to that in Oslo, a less affected city. When comparing in situ exposure, the difference in the mean monthly fallout is 2.9 standard deviations, and the difference in median monthly fallout is 2.6 standard deviations. (The difference in the maximum monthly fallout is 6.4 standard deviations.) For air exposure, the differences are even bigger, with the difference in mean monthly fallout at 4.5 standard deviations and the difference in median monthly fallout at 4.8 standard deviations. (The difference in the maximum monthly fallout is 5.3 standard deviations.) Thus, the findings imply that conditional on fallout in other months, moving from being exposed to the mean monthly ground fallout in month 3 or month 4 in Oslo to that in Bergen would reduce the expected IQ score by about 0.06 of a standard deviation. For fallout in air, the equivalent effect would be about 0.25 of a standard deviation. Of course, the effect of moving from average Bergen to average Oslo exposure in both months 3 and 4 would be predicted to be approximately twice this. So our estimates imply reasonably large effects when placed in the context of the variation in fallout in the data.

The other key finding from this table is that there is no systematic evidence of effects of radiation exposure in any other pregnancy month. One partial exception to this is that there is some evidence for a smaller adverse effect from radiation in month 5. However, in general, the findings are consistent with adverse effects of radiation being confined to months 3 and 4.17 In the online appendix, we show that this remains true even when we include fallout in the nine months before and after pregnancy.

### A.  Sibling Fixed Effects

We also estimate a specification that includes sibling fixed effects, restricting the sample to families in which there are at least two children born during the period in the same municipality. The control variables are birth order, the unemployment rate, and year of birth by month of birth dummies. These results are presented in columns 4 and 8 of table 2. The sibling fixed effects results are similar to our earlier findings, with IQ score at age 18 significantly affected by exposure in months 3 and 4 but with little evidence of any effects in other months.

### B.  Multiple Hypotheses Tests

As discussed earlier, we have implemented the Romano-Wolf stepdown approach and report estimates in italics in the table if they remain statistically significant at the 10% level following this correction. Our finding that fallout in months 3 and 4 affects IQ scores remains, with fourteen of the sixteen coefficients in table 2 that pertain to months 3 and 4 remaining statistically significant under the correction. Interestingly, while several of the coefficients relating to other months are individually statistically significant at the 10% level in table 2, none of these remains statistically significant under the correction. Our interpretation is that many of these may have been significant by chance due to the multiplicity of coefficients being estimated.

### C.  Other Outcomes

Given the robustness of our IQ score results to our choice of specification, the remaining tables present the results from the primary specification that includes both year by month of birth fixed effects and municipality fixed effects. We show the robustness of the estimates for these outcomes to using the other specifications in the online appendix.

Tables 3A and 3B present the results for the other outcomes for both men and women with the two different measures of exposure: table 3A uses in situ exposure and table 3B uses air exposure. As women born in the 1950s and 1960s have different labor force attachment than men, we first present the results separately for men and women and then test whether the effects differ by gender (we report pooled estimates in the online appendix). Importantly, it is clear again that it is months 3 and 4 in utero when exposure has a significant effect on education and earnings. When we look at the results for educational attainment and high school completion, we find that radioactive exposure seems to have a negative and statistically significant effect on education among men. Similarly, there is a significantly negative effect of exposure on the educational attainment of women; again, this is robust to the measure of exposure used. The magnitudes suggest that a 1 standard deviation increase in ground exposure during months 3 or 4 reduces educational attainment by 0.08 years for men and 0.1 years for women, with effects on high school completion of less than 1 percentage point for men and about 1 percentage point for women. We also find negative effects on earnings at age 35 for both men and women; however, these coefficients are not always statistically significant at the 5% level.18 As with the IQ score, the coefficients for months 3 and 4 for education and earnings mostly survive the correction for multiple hypothesis tests, but very few of the coefficients from the other months remain significant.

Table 3.
Total Beta Fallout by Month
A. Fallout in Situ
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Three months prior to 0.003 0.026 0.008 −0.001 0.001 0.005 −0.002 0.006
pregnancy (0.006) (0.025) (0.011) (0.002) (0.003) (0.009) (0.002) (0.005)
Two months prior to −0.012 −0.003 0.016 0.002 −0.003 −0.029** −0.003 −0.007
pregnancy (0.008) (0.025) (0.010) (0.002) (0.003) (0.012) (0.002) (0.005)
One months prior to −0.015 −0.010 0.028 0.003 0.002 0.014 0.001 0.002
pregnancy (0.016) (0.030) (0.020) (0.002) (0.003) (0.016) (0.002) (0.005)
Pregnancy −0.018* 0.064* −0.006 0.001 −0.000 −0.005 0.005 0.003
month 1 (0.010) (0.035) (0.012) (0.001) (0.005) (0.011) (0.003) (0.004)
Pregnancy −0.020* −0.056* −0.017 −0.003 −0.000 −0.052* −0.006*** 0.005
month 2 (0.011) (0.030) (0.015) (0.002) (0.007) (0.030) (0.002) (0.005)
Pregnancy −0.039*** −0.068* −0.075*** −0.002 −0.007** −0.100*** −0.009** −0.006
month 3 (0.009) (0.036) (0.020) (0.002) (0.003) (0.033) (0.005) (0.004)
Pregnancy −0.043*** −0.093*** −0.082*** −0.008*** −0.006 −0.107*** −0.011*** −0.011**
month 4 (0.011) (0.020) (0.022) (0.003) (0.004) (0.033) (0.003) (0.006)
Pregnancy 0.005 0.018 −0.062*** 0.002 −0.005 0.022 −0.003 0.006
month 5 (0.005) (0.026) (0.017) (0.002) (0.005) (0.019) (0.002) (0.006)
Pregnancy 0.010 0.092* −0.001 −0.002 0.002 0.000 0.004* 0.002
month 6 (0.006) (0.051) (0.008) (0.002) (0.008) (0.012) (0.002) (0.004)
Pregnancy 0.013 −0.020 0.027 0.005 0.001 −0.000 −0.004 −0.010*
month 7 (0.015) (0.031) (0.017) (0.003) (0.005) (0.012) (0.004) (0.005)
Pregnancy −0.014 −0.053** −0.003 −0.004 −0.006 −0.030 0.006 0.016
month 8 (0.009) (0.024) (0.014) (0.003) (0.004) (0.019) (0.004) (0.015)
Pregnancy −0.007 −0.018 −0.003 −0.003 0.001 0.018* 0.002 −0.004
month 9 (0.007) (0.038) (0.015) (0.003) (0.003) (0.010) (0.002) (0.004)
Month of birth 0.005 −0.021 0.003 0.006 −0.006* −0.005 −0.003 −0.011
(0.004) (0.033) (0.013) (0.004) (0.003) (0.014) (0.002) (0.007)
After pregnancy 1 month 0.007 0.051* 0.006 −0.002 0.006 −0.019 −0.001 −0.017
(0.006) (0.030) (0.013) (0.002) (0.005) (0.012) (0.002) (0.015)
After pregnancy 2 months −0.012* 0.021 −0.012 −0.000 0.006 −0.023* −0.003 −0.002
(0.007) (0.023) (0.010) (0.002) (0.004) (0.013) (0.002) (0.005)
After pregnancy 3 months −0.009 0.033 −0.014 −0.005* 0.001 −0.001 −0.002 −0.012*
(0.006) (0.023) (0.011) (0.003) (0.003) (0.015) (0.002) (0.007)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
B. Fallout in Air
Men Women
IQ Height Years of Education High School Completed Log Earnings at Age 35 Years of Education High School Completed Log Earnings at Age 35
Three months prior to −0.015 0.031 −0.039* −0.004* −0.005 −0.007 0.000 0.003
pregnancy (0.019) (0.057) (0.022) (0.002) (0.003) (0.011) (0.002) (0.003)
Two months prior to −0.027* −0.055 −0.015 0.001 0.003 0.003 −0.001 −0.007
pregnancy (0.014) (0.037) (0.021) (0.003) (0.003) (0.012) (0.002) (0.006)
One month prior to −0.063 0.004 −0.034 −0.003 0.004 0.000 −0.000 0.003
pregnancy (0.052) (0.043) (0.025) (0.002) (0.005) (0.012) (0.002) (0.006)
Pregnancy −0.038 0.006 −0.020 −0.004 −0.004 −0.009 −0.002 0.004
month 1 (0.028) (0.046) (0.012) (0.003) (0.004) (0.015) (0.003) (0.008)
Pregnancy −0.078* −0.077** −0.033** −0.001 0.006 −0.029* −0.006** −0.006
month 2 (0.046) (0.037) (0.015) (0.003) (0.006) (0.017) (0.002) (0.006)
Pregnancy −0.127*** −0.011 −0.190*** −0.017*** −0.006 −0.118*** −0.011*** −0.008
month 3 (0.052) (0.034) (0.035) (0.002) (0.004) (0.032) (0.004) (0.005)
Pregnancy −0.093** −0.029 −0.172*** −0.014*** −0.008** −0.135*** −0.010*** −0.010**
month 4 (0.047) (0.039) (0.035) (0.003) (0.004) (0.033) (0.003) (0.005)
Pregnancy −0.046* −0.004 −0.011 0.000 0.004 −0.015 −0.001 −0.004
month 5 (0.027) (0.004) (0.021) (0.003) (0.003) (0.018) (0.003) (0.006)
Pregnancy −0.083 0.079* −0.041* −0.007* 0.001 −0.023* −0.003 −0.006
month 6 (0.064) (0.045) (0.024) (0.004) (0.004) (0.013) (0.002) (0.005)
Pregnancy −0.050 −0.042 −0.015 0.002 0.001 −0.030** −0.003 −0.001
month 7 (0.067) (0.029) (0.013) (0.002) (0.006) (0.011) (0.003) (0.007)
Pregnancy −0.043 −0.059 0.004 −0.003 0.007 0.010 −0.002 −0.001
month 8 (0.031) (0.036) (0.016) (0.002) (0.006) (0.014) (0.003) (0.006)
Pregnancy −0.019 0.005 −0.019 −0.005 0.003 −0.006 −0.004 −0.002
month 9 (0.013) (0.036) (0.014) (0.003) (0.004) (0.018) (0.005) (0.006)
Month of birth −0.057* −0.003 0.005 −0.002 0.010* −0.003 −0.001 −0.007
(0.030) (0.043) (0.014) (0.003) (0.006) (0.015) (0.002) (0.005)
After pregnancy 1 0.037 −0.012 0.002 0.001 0.002 −0.018 −0.002 −0.011
(0.029) (0.059) (0.014) (0.002) (0.004) (0.013) (0.002) (0.007)
After pregnancy 2 −0.035 −0.018 −0.033* −0.001 −0.000 −0.026* −0.005 0.010
(0.038) (0.039) (0.019) (0.003) (0.006) (0.014) (0.003) (0.006)
After pregnancy 3 −0.021 −0.060 −0.017 −0.001 0.006 −0.019 −0.006 −0.005
(0.078) (0.038) (0.013) (0.002) (0.004) (0.012) (0.004) (0.006)
Observations 88,446 92,793 93,275 93,723 86,544 94,018 94,511 81,984
A. Fallout in Situ
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Three months prior to 0.003 0.026 0.008 −0.001 0.001 0.005 −0.002 0.006
pregnancy (0.006) (0.025) (0.011) (0.002) (0.003) (0.009) (0.002) (0.005)
Two months prior to −0.012 −0.003 0.016 0.002 −0.003 −0.029** −0.003 −0.007
pregnancy (0.008) (0.025) (0.010) (0.002) (0.003) (0.012) (0.002) (0.005)
One months prior to −0.015 −0.010 0.028 0.003 0.002 0.014 0.001 0.002
pregnancy (0.016) (0.030) (0.020) (0.002) (0.003) (0.016) (0.002) (0.005)
Pregnancy −0.018* 0.064* −0.006 0.001 −0.000 −0.005 0.005 0.003
month 1 (0.010) (0.035) (0.012) (0.001) (0.005) (0.011) (0.003) (0.004)
Pregnancy −0.020* −0.056* −0.017 −0.003 −0.000 −0.052* −0.006*** 0.005
month 2 (0.011) (0.030) (0.015) (0.002) (0.007) (0.030) (0.002) (0.005)
Pregnancy −0.039*** −0.068* −0.075*** −0.002 −0.007** −0.100*** −0.009** −0.006
month 3 (0.009) (0.036) (0.020) (0.002) (0.003) (0.033) (0.005) (0.004)
Pregnancy −0.043*** −0.093*** −0.082*** −0.008*** −0.006 −0.107*** −0.011*** −0.011**
month 4 (0.011) (0.020) (0.022) (0.003) (0.004) (0.033) (0.003) (0.006)
Pregnancy 0.005 0.018 −0.062*** 0.002 −0.005 0.022 −0.003 0.006
month 5 (0.005) (0.026) (0.017) (0.002) (0.005) (0.019) (0.002) (0.006)
Pregnancy 0.010 0.092* −0.001 −0.002 0.002 0.000 0.004* 0.002
month 6 (0.006) (0.051) (0.008) (0.002) (0.008) (0.012) (0.002) (0.004)
Pregnancy 0.013 −0.020 0.027 0.005 0.001 −0.000 −0.004 −0.010*
month 7 (0.015) (0.031) (0.017) (0.003) (0.005) (0.012) (0.004) (0.005)
Pregnancy −0.014 −0.053** −0.003 −0.004 −0.006 −0.030 0.006 0.016
month 8 (0.009) (0.024) (0.014) (0.003) (0.004) (0.019) (0.004) (0.015)
Pregnancy −0.007 −0.018 −0.003 −0.003 0.001 0.018* 0.002 −0.004
month 9 (0.007) (0.038) (0.015) (0.003) (0.003) (0.010) (0.002) (0.004)
Month of birth 0.005 −0.021 0.003 0.006 −0.006* −0.005 −0.003 −0.011
(0.004) (0.033) (0.013) (0.004) (0.003) (0.014) (0.002) (0.007)
After pregnancy 1 month 0.007 0.051* 0.006 −0.002 0.006 −0.019 −0.001 −0.017
(0.006) (0.030) (0.013) (0.002) (0.005) (0.012) (0.002) (0.015)
After pregnancy 2 months −0.012* 0.021 −0.012 −0.000 0.006 −0.023* −0.003 −0.002
(0.007) (0.023) (0.010) (0.002) (0.004) (0.013) (0.002) (0.005)
After pregnancy 3 months −0.009 0.033 −0.014 −0.005* 0.001 −0.001 −0.002 −0.012*
(0.006) (0.023) (0.011) (0.003) (0.003) (0.015) (0.002) (0.007)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
B. Fallout in Air
Men Women
IQ Height Years of Education High School Completed Log Earnings at Age 35 Years of Education High School Completed Log Earnings at Age 35
Three months prior to −0.015 0.031 −0.039* −0.004* −0.005 −0.007 0.000 0.003
pregnancy (0.019) (0.057) (0.022) (0.002) (0.003) (0.011) (0.002) (0.003)
Two months prior to −0.027* −0.055 −0.015 0.001 0.003 0.003 −0.001 −0.007
pregnancy (0.014) (0.037) (0.021) (0.003) (0.003) (0.012) (0.002) (0.006)
One month prior to −0.063 0.004 −0.034 −0.003 0.004 0.000 −0.000 0.003
pregnancy (0.052) (0.043) (0.025) (0.002) (0.005) (0.012) (0.002) (0.006)
Pregnancy −0.038 0.006 −0.020 −0.004 −0.004 −0.009 −0.002 0.004
month 1 (0.028) (0.046) (0.012) (0.003) (0.004) (0.015) (0.003) (0.008)
Pregnancy −0.078* −0.077** −0.033** −0.001 0.006 −0.029* −0.006** −0.006
month 2 (0.046) (0.037) (0.015) (0.003) (0.006) (0.017) (0.002) (0.006)
Pregnancy −0.127*** −0.011 −0.190*** −0.017*** −0.006 −0.118*** −0.011*** −0.008
month 3 (0.052) (0.034) (0.035) (0.002) (0.004) (0.032) (0.004) (0.005)
Pregnancy −0.093** −0.029 −0.172*** −0.014*** −0.008** −0.135*** −0.010*** −0.010**
month 4 (0.047) (0.039) (0.035) (0.003) (0.004) (0.033) (0.003) (0.005)
Pregnancy −0.046* −0.004 −0.011 0.000 0.004 −0.015 −0.001 −0.004
month 5 (0.027) (0.004) (0.021) (0.003) (0.003) (0.018) (0.003) (0.006)
Pregnancy −0.083 0.079* −0.041* −0.007* 0.001 −0.023* −0.003 −0.006
month 6 (0.064) (0.045) (0.024) (0.004) (0.004) (0.013) (0.002) (0.005)
Pregnancy −0.050 −0.042 −0.015 0.002 0.001 −0.030** −0.003 −0.001
month 7 (0.067) (0.029) (0.013) (0.002) (0.006) (0.011) (0.003) (0.007)
Pregnancy −0.043 −0.059 0.004 −0.003 0.007 0.010 −0.002 −0.001
month 8 (0.031) (0.036) (0.016) (0.002) (0.006) (0.014) (0.003) (0.006)
Pregnancy −0.019 0.005 −0.019 −0.005 0.003 −0.006 −0.004 −0.002
month 9 (0.013) (0.036) (0.014) (0.003) (0.004) (0.018) (0.005) (0.006)
Month of birth −0.057* −0.003 0.005 −0.002 0.010* −0.003 −0.001 −0.007
(0.030) (0.043) (0.014) (0.003) (0.006) (0.015) (0.002) (0.005)
After pregnancy 1 0.037 −0.012 0.002 0.001 0.002 −0.018 −0.002 −0.011
(0.029) (0.059) (0.014) (0.002) (0.004) (0.013) (0.002) (0.007)
After pregnancy 2 −0.035 −0.018 −0.033* −0.001 −0.000 −0.026* −0.005 0.010
(0.038) (0.039) (0.019) (0.003) (0.006) (0.014) (0.003) (0.006)
After pregnancy 3 −0.021 −0.060 −0.017 −0.001 0.006 −0.019 −0.006 −0.005
(0.078) (0.038) (0.013) (0.002) (0.004) (0.012) (0.004) (0.006)
Observations 88,446 92,793 93,275 93,723 86,544 94,018 94,511 81,984

The sample includes persons born between 1956 and 1966 and includes municipalities within a radius of 20 kilometers of the test stations. “Fallout in Situ” refers to ground deposition measured in kBq/m$2$. “Fallout in Air” refers to air deposition measured in kBq/m$3$. The fallout measures are standardized to mean 0 and standard deviation 1. Each column represents a separate regression that includes controls for municipality dummies and year of birth by month of birth dummies. Also included in each specification are controls for parental education, birth order, and the municipality unemployment rate. Standard errors are clustered at the municipality level. Significant at $***$1%, $**$5%, $*$10%. Coefficients in italics survive a correction for multiple hypothesis testing on a 10% significance level.

For boys, we can also study height at around age 18. The evidence for adverse effects on height is much weaker, with little evidence of a consistent pattern. These weaker results for height are unsurprising as the scientific research speaks to the effects of radiation on cognitive rather than physical development.

We next examine whether there might be nonlinearities in the effects by splitting fallout levels into quintiles. To limit the number of coefficients, we average exposure over two-month periods going from two months before pregnancy to two months after birth and include indicator variables for quintiles for each two-month period. For tractability, we report only estimates for months 3 and 4 in table 4 (the estimates for quintiles for all months are in the online appendix) and use the original specification (with municipality dummies and year of birth by month of birth effects). These estimates provide further evidence that fallout in months 3 and 4 matters, and we see that the estimates are monotonically increasing in magnitude with quintile. For both men and women, the effects of being exposed in months 3 and 4 to in situ fallout from the fifth quintile compared to the first quintile is a predicted reduction in years of education of 0.17 years, a 2 percentage point reduction in the probability of completing high school, and a 2% decrease in earnings at age 35. The effects of fallout in air are similar. Thus, the magnitudes we find are of some economic significance.19

Table 4.
Quintile of Fallout, in Situ and in Air
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Total beta in situ
Quintile 2 −0.102* 0.012 −0.065*** −0.006 −0.005 −0.035 0.006 −0.006
(0.059) (0.051) (0.022) (0.004) (0.003) (0.022) (0.004) (0.004)
Quintile 3 −0.106* −0.056 −0.077*** −0.008 −0.006 −0.079*** −0.008** −0.010**
(0.058) (0.041) (0.027) (0.005) (0.004) (0.022) (0.004) (0.005)
Quintile 4 −0.146*** −0.064 −0.111*** −0.012** −0.013*** −0.100*** −0.015*** −0.016***
(0.055) (0.044) (0.022) (0.005) (0.004) (0.023) (0.003) (0.005)
Quintile 5 −0.214*** −0.058 −0.169*** −0.021*** −0.018*** −0.167*** −0.023*** −0.020***
(0.062) (0.038) (0.025) (0.005) (0.003) (0.020) (0.004) (0.005)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
Total beta in air
Quintile 2 −0.102* 0.024 −0.030 −0.002 −0.006 −0.017 −0.005 −0.004
(0.059) (0.062) (0.028) (0.004) (0.005) (0.012) (0.004) (0.004)
Quintile 3 −0.071 −0.050 −0.065* −0.005 −0.006 −0.062*** −0.006 −0.008
(0.052) (0.039) (0.036) (0.004) (0.004) (0.022) (0.004) (0.005)
Quintile 4 −0.095** 0.023 −0.174*** −0.011** −0.012** −0.114*** −0.012*** −0.012**
(0.047) (0.040) (0.035) (0.005) (0.005) (0.023) (0.004) (0.005)
Quintile 5 −0.156** 0.016 −0.210*** −0.023*** −0.017*** −0.177*** −0.018*** −0.017***
(0.062) (0.047) (0.035) (0.005) (0.003) (0.020) (0.004) (0.006)
Observations 94,649 92,793 93,275 93,723 86,544 94,018 94,511 81,984
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Total beta in situ
Quintile 2 −0.102* 0.012 −0.065*** −0.006 −0.005 −0.035 0.006 −0.006
(0.059) (0.051) (0.022) (0.004) (0.003) (0.022) (0.004) (0.004)
Quintile 3 −0.106* −0.056 −0.077*** −0.008 −0.006 −0.079*** −0.008** −0.010**
(0.058) (0.041) (0.027) (0.005) (0.004) (0.022) (0.004) (0.005)
Quintile 4 −0.146*** −0.064 −0.111*** −0.012** −0.013*** −0.100*** −0.015*** −0.016***
(0.055) (0.044) (0.022) (0.005) (0.004) (0.023) (0.003) (0.005)
Quintile 5 −0.214*** −0.058 −0.169*** −0.021*** −0.018*** −0.167*** −0.023*** −0.020***
(0.062) (0.038) (0.025) (0.005) (0.003) (0.020) (0.004) (0.005)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
Total beta in air
Quintile 2 −0.102* 0.024 −0.030 −0.002 −0.006 −0.017 −0.005 −0.004
(0.059) (0.062) (0.028) (0.004) (0.005) (0.012) (0.004) (0.004)
Quintile 3 −0.071 −0.050 −0.065* −0.005 −0.006 −0.062*** −0.006 −0.008
(0.052) (0.039) (0.036) (0.004) (0.004) (0.022) (0.004) (0.005)
Quintile 4 −0.095** 0.023 −0.174*** −0.011** −0.012** −0.114*** −0.012*** −0.012**
(0.047) (0.040) (0.035) (0.005) (0.005) (0.023) (0.004) (0.005)
Quintile 5 −0.156** 0.016 −0.210*** −0.023*** −0.017*** −0.177*** −0.018*** −0.017***
(0.062) (0.047) (0.035) (0.005) (0.003) (0.020) (0.004) (0.006)
Observations 94,649 92,793 93,275 93,723 86,544 94,018 94,511 81,984

The sample includes persons born between 1956 and 1966 and includes municipalities within a radius of 20 kilometers of the test stations. “Total beta in air” refers to air deposition measured in kBq/m$3$ during months 3 and 4 in utero (the average value over the two months). “Total beta in situ” refers to ground deposition measured in kBq/m$2$ during months 3 and 4 in utero (the average value over the two months). The fallout measures are standardized to mean 0 and standard deviation 1. Each set of quintile estimates comes from a separate regression with controls for municipality dummies and year of birth by month of birth dummies. Also included in each specification are controls for parental education, birth order, and the municipality unemployment rate, and quintile dummies for fallout in months from two months before pregnancy to two months after birth. The full set of quintile coefficients is in the online appendix. Standard errors are clustered at the municipality level. Significant at $***$1%, $**$5%, $*$10%.

### D.  Heterogeneous Effects

#### Varying intensity of exposure.

One might expect effects to be larger in months with more sunlight when individuals are more likely to be outside. As another check, we also estimate specifications where we include an interaction indicating whether the exposure (during months 3 and 4 in utero) occurred during the spring or summer months (defined as month 3 occurring April to September). Table 5 presents these results. We find statistically significant interaction effects for both ground and air fallout, suggesting that exposure is more harmful during spring and summer months.

Table 5.
Interaction with Season of Exposure: Summer (April–September)
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Total beta in situ
Total beta in situ −0.040*** 0.021 −0.094*** −0.005*** −0.007** −0.118*** −0.010*** −0.011***
(0.012) (0.047) (0.034) (0.002) (0.003) (0.017) (0.003) (0.004)
Summer 0.203* 0.066 0.065 0.012 0.017 −0.076 0.014 0.032
(0.112) (0.152) (0.084) (0.012) (0.015) (0.086) (0.023) (0.020)
Interaction term −0.093*** −0.102 −0.230*** −0.022*** −0.017 −0.062*** −0.016*** −0.020*
(0.023) (0.068) (0.035) (0.007) (0.011) (0.017) (0.006) (0.012)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
Total beta in air
Total beta in air −0.113*** 0.001 −0.234*** −0.018*** 0.002 −0.220*** −0.019*** −0.021**
(0.037) (0.045) (0.065) (0.006) (0.006) (0.077) (0.007) (0.009)
Summer 0.140 0.062 −0.291 −0.039 0.026 −0.030 −0.003 −0.017
(0.111) (0.521) (0.252) (0.048) (0.047) (0.206) (0.021) (0.060)
Interaction −0.238*** −0.283** −0.794*** −0.055*** −0.065*** −0.814*** −0.077*** −0.043**
(0.068) (0.118) (0.173) (0.014) (0.021) (0.149) (0.009) (0.021)
Observations 88,446 92,793 93,275 93,723 86,544 94,018 94,511 81,984
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Total beta in situ
Total beta in situ −0.040*** 0.021 −0.094*** −0.005*** −0.007** −0.118*** −0.010*** −0.011***
(0.012) (0.047) (0.034) (0.002) (0.003) (0.017) (0.003) (0.004)
Summer 0.203* 0.066 0.065 0.012 0.017 −0.076 0.014 0.032
(0.112) (0.152) (0.084) (0.012) (0.015) (0.086) (0.023) (0.020)
Interaction term −0.093*** −0.102 −0.230*** −0.022*** −0.017 −0.062*** −0.016*** −0.020*
(0.023) (0.068) (0.035) (0.007) (0.011) (0.017) (0.006) (0.012)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
Total beta in air
Total beta in air −0.113*** 0.001 −0.234*** −0.018*** 0.002 −0.220*** −0.019*** −0.021**
(0.037) (0.045) (0.065) (0.006) (0.006) (0.077) (0.007) (0.009)
Summer 0.140 0.062 −0.291 −0.039 0.026 −0.030 −0.003 −0.017
(0.111) (0.521) (0.252) (0.048) (0.047) (0.206) (0.021) (0.060)
Interaction −0.238*** −0.283** −0.794*** −0.055*** −0.065*** −0.814*** −0.077*** −0.043**
(0.068) (0.118) (0.173) (0.014) (0.021) (0.149) (0.009) (0.021)
Observations 88,446 92,793 93,275 93,723 86,544 94,018 94,511 81,984

The sample includes persons born between 1956 and 1966 and includes municipalities within a radius of 20 kilometers of the test stations. “Total beta in air” refers to air deposition measured in kBq/m$3$ during months 3 and 4 in utero (the average value over the two months). “Total beta in situ” refers to ground deposition measured in kBq/m$2$ during months 3 and 4 in utero (the average value over the two months). The fallout measures are standardized to mean 0 and standard deviation 1. Each set of estimates comes from a separate regression with controls for municipality dummies and year of birth by month of birth dummies. Also included in each specification are controls for parental education, birth order, and the municipality unemployment rate. Standard errors are clustered at the municipality level. Significant at $***$1%, $**$5%, $*$10%.

#### Family background.

Finally, the negative effect of poor childhood health on human capital accumulation is often found to be stronger for individuals growing up in a less-educated or low-income family (Almond & Currie, 2011). Although there was little public awareness of the exposure to nuclear fallout and scientific papers on the health effects of nuclear fallout on unborn children were published only in the 1980s, well-educated parents might have had access to more information and therefore engaged in avoidance behavior (Aizer & Stroud, 2011). When we interact the nuclear fallout measures with an indicator variable equal to 1 if the individual's mother had completed high school or more, we find that the interaction term is not statistically significant in most cases and the coefficient on the level effect of exposure is quite similar to the earlier estimates (see table 6). Interestingly, the effect of exposure is actually greater for individuals born to more highly educated parents when we look at years of education for both men and women. This is contrary to what the literature would suggest, but given the general insignificance of the interaction terms, we do not put too much weight on this finding.

Table 6.
Interaction with Mother's Education
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Total beta in situ
Total beta in situ −0.046*** 0.035 −0.203*** −0.007*** −0.008** −0.214*** −0.013*** −0.018***
(0.012) (0.028) (0.032) (0.002) (0.004) (0.041) (0.005) (0.005)
Mother completed high school 1.103*** 0.992*** 1.513*** 0.175*** 0.144*** 1.338*** 0.182*** 0.193***
(0.018) (0.041) (0.042) (0.008) (0.008) (0.038) (0.009) (0.009)
Interaction term −0.009 −0.027 −0.084*** −0.002 −0.004 −0.038 0.001 −0.003
(0.011) (0.064) (0.027) (0.003) (0.006) (0.026) (0.003) (0.005)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
Total beta in air
Total beta in air −0.126*** −0.048 −0.349*** −0.021*** −0.009* −0.285*** −0.030*** −0.024***
(0.032) (0.041) (0.054) (0.006) (0.005) (0.052) (0.009) (0.011)
Mother has completed high school 1.248*** 1.010*** 1.438*** 0.154*** 0.123*** 1.418*** 0.169*** 0.202***
(0.023) (0.045) (0.052) (0.005) (0.005) (0.042) (0.007) (0.008)
Interaction term −0.005 −0.034 −0.067** 0.002 −0.007* −0.019 0.003 −0.002
(0.015) (0.041) (0.031) (0.004) (0.005) (0.021) (0.002) (0.007)
Observations 88,446 92,793 93,275 93,723 86,544 94,018 94,511 81,984
MenWomen
IQHeightYears of EducationHigh School CompletedLog Earnings at Age 35Years of EducationHigh School CompletedLog Earnings at Age 35
Total beta in situ
Total beta in situ −0.046*** 0.035 −0.203*** −0.007*** −0.008** −0.214*** −0.013*** −0.018***
(0.012) (0.028) (0.032) (0.002) (0.004) (0.041) (0.005) (0.005)
Mother completed high school 1.103*** 0.992*** 1.513*** 0.175*** 0.144*** 1.338*** 0.182*** 0.193***
(0.018) (0.041) (0.042) (0.008) (0.008) (0.038) (0.009) (0.009)
Interaction term −0.009 −0.027 −0.084*** −0.002 −0.004 −0.038 0.001 −0.003
(0.011) (0.064) (0.027) (0.003) (0.006) (0.026) (0.003) (0.005)
Observations 89,892 94,339 94,827 95,280 88,024 95,781 96,288 83,509
Total beta in air
Total beta in air −0.126*** −0.048 −0.349*** −0.021*** −0.009* −0.285*** −0.030*** −0.024***
(0.032) (0.041) (0.054) (0.006) (0.005) (0.052) (0.009) (0.011)
Mother has completed high school 1.248*** 1.010*** 1.438*** 0.154*** 0.123*** 1.418*** 0.169*** 0.202***
(0.023) (0.045) (0.052) (0.005) (0.005) (0.042) (0.007) (0.008)
Interaction term −0.005 −0.034 −0.067** 0.002 −0.007* −0.019 0.003 −0.002
(0.015) (0.041) (0.031) (0.004) (0.005) (0.021) (0.002) (0.007)
Observations 88,446 92,793 93,275 93,723 86,544 94,018 94,511 81,984

The sample includes persons born between 1956 and 1966 and includes municipalities within a radius of 20 kilometers of the test stations. “Total beta in air” refers to air deposition measured in kBq/m$3$ during months 3 and 4 in utero (the average value over the two months). “Total beta in situ” refers to ground deposition measured in kBq/m$2$ during months 3 and 4 in utero (the average value over the two months). The fallout measures are standardized to mean 0 and standard deviation 1. Each set of estimates comes from a separate regression with controls for municipality dummies and year of birth by month of birth dummies. Also included in each specification are controls for parental education, birth order, and the municipality unemployment rate. Standard errors are clustered at the municipality level. Significant at $***$1%, $**$5%, $*$10%.

Differences in coefficient values between OLS and sibling fixed-effects regressions can also provide information on parental responses. Compensatory behavior within families would lead to fixed-effects estimates being smaller; reinforcing responses would have the opposite effect. Table 2 includes sibling fixed-effects estimates for IQ scores, and analogous estimates for the other outcomes are in the online appendix. In general, sibling fixed-effects estimates are a little smaller than OLS, but the differences are not large and provide little evidence of parental response.

## VI.  Results—Second Generation

Our rich data set allows us to link across generations and provides a rare opportunity to examine the effects of in utero exposure to a pollutant on the children of those exposed in utero. As noted earlier, there is little known about the intergenerational consequences of shocks in utero, with the most compelling work examining the effects of a shock in utero on the birth outcomes of offspring. We can add to this literature by examining the effects on the IQ score of the offspring at age 18.21

Exposure could be transmitted across generations through biological or socioeconomic channels. As we discuss in the online appendix, several biological pathways are plausible and could result in effects on children even for parents who were exposed outside of months 3 and 4 in utero. The socioeconomic channel is that parents exposed in utero have worse socioeconomic outcomes, leading to worse outcomes for their children. If this is the case, we would expect that exposure of the parent during months 3 and 4 in utero will have the strongest effect on their offspring.

To avoid confusion, we refer to persons in utero between 1956 and 1966 as the first generation and refer to their children as the second generation. We have data on IQ test scores for men at age 18 or 19 up through 2010. This allows us to study the effect of nuclear fallout on IQ scores of second-generation men (i.e., for sons of the first generation but not for their daughters).

Of our sample of people born between 1956 and 1966, 23% of men and 30% of women have at least one son who had taken the military tests by 2010. A likely explanation for this disparity is that women have children at a younger age than men do, increasing their likelihood of having sons who are old enough to be in our sample.22

We estimate the intergenerational effects of exposure to nuclear fallout using two specifications (a full set of robustness checks is in the online appendix). The first includes the same control variables that we used when studying outcomes for the first generation that were predetermined at the time of the fallout. The second specification includes additional controls for child-specific factors that are likely to have direct effects on birth outcomes and IQ scores (family size, birth order, year of birth). While we have found no systematic effects of fallout on fertility behavior of the first generation, there may still be some variation in these control variables that is correlated with exposure of the first generation to radiation. For comparability of the estimates across generations, the regressions are weighted such that each first-generation parent gets equal weight. We run separate regressions for the sons of first-generation men and women and then provide estimates when both are combined.

These intergenerational results are in table 7a (for in situ radiation) and table 7b (for radiation in the air). Columns 1 and 4 have results with the same controls as in the regression for the first generation for father's and mother's exposure respectively, while columns 2 and 5 add the additional second-generation controls (year of birth dummies, birth order dummies, and family size). For in situ radiation, we find statistically significant negative effects of exposure of first-generation men in months 3 and 4 in utero on IQ test scores of the second generation. For exposed women, the effects are also negative and statistically significant for month 4 but not month 3.23 The inclusion of the extra controls in columns 2 and 5 makes very little difference to the estimates. For exposure through air, the magnitudes are similar, but the standard errors are higher, so fewer of the coefficients for months 3 and 4 are statistically significant. Although findings are weaker than for the first generation, they still imply that exposure in months 3 and 4 is particularly important.

Table 7.
Effect of Fallout by Month on the Second Generation, Male Offspring
A. In Situ
Father and Mother Exposed
Father ExposedMother ExposedSon's IQ
Son's IQ (1)Son's IQ (2)Father's IQ (3)Son's IQ (4)Son's IQ (5)Father's Exposure (6a)Mother's Exposure (6b)
Three months prior to −0.002 0.001 0.010 −0.012 −0.017 −0.002 −0.001
pregnancy (0.022) (0.023) (0.023) (0.017) (0.017) (0.026) (0.003)
Two months prior to 0.016 0.015 −0.005 0.001 0.000 0.016 0.001
pregnancy (0.022) (0.021) (0.022) (0.021) (0.021) (0.020) (0.003)
One month prior to 0.018 0.027 0.021** −0.015 −0.012 0.031 −0.001
pregnancy (0.019) (0.020) (0.009) (0.019) (0.019) (0.023) (0.004)
Pregnancy 0.015 0.007 −0.045 0.025* 0.021 0.009 0.015
month 1 (0.018) (0.018) (0.029) (0.013) (0.013) (0.021) (0.050)
Pregnancy 0.004 0.006 0.009 0.005 0.005 0.001 0.027
month 2 (0.030) (0.030) (0.025) (0.008) (0.008) (0.033) (0.052)
Pregnancy −0.024*** −0.028*** −0.042** −0.023 −0.024 −0.018** −0.013**
month 3 (0.007) (0.008) (0.018) (0.017) (0.016) (0.007) (0.005)
Pregnancy −0.027** −0.030** −0.043* −0.030** −0.030** −0.022** −0.013
month 4 (0.013) (0.013) (0.026) (0.013) (0.013) (0.011) (0.009)
Pregnancy −0.039* −0.029 0.012 −0.017 −0.013 −0.031 −0.011
month 5 (0.021) (0.022) (0.025) (0.015) (0.015) (0.024) (0.026)
Pregnancy 0.043** 0.038* 0.019 0.000 −0.000 0.044 −0.000
month 6 (0.021) (0.023) (0.023) (0.011) (0.011) (0.028) (0.005)
Pregnancy 0.015 0.014 0.046* −0.027 −0.029 −0.016 −0.006
month 7 (0.020) (0.022) (0.026) (0.021) (0.021) (0.028) (0.012)
Pregnancy 0.027 0.034 −0.005 0.038** 0.037** 0.032* −0.001**
month 8 (0.023) (0.022) (0.023) (0.017) (0.016) (0.019) (0.000)
Pregnancy −0.033 −0.031 −0.027 −0.012 −0.010 −0.028 −0.011
month 9 (0.024) (0.023) (0.023) (0.014) (0.013) (0.022) (0.007)
Month of birth −0.012 −0.006 0.011 −0.013 −0.014 −0.034 −0.015
(0.019) (0.020) (0.025) (0.016) (0.016) (0.026) (0.017)
After pregnancy 1 −0.001 0.001 0.001 0.029 0.031 0.022 0.033
(0.020) (0.019) (0.026) (0.027) (0.028) (0.020) (0.023)
After pregnancy 2 0.008 0.008 0.004 0.032*** 0.033*** 0.022 0.014*
(0.020) (0.019) (0.014) (0.008) (0.008) (0.022) (0.008)
After pregnancy 3 0.014 0.011 0.001 0.003 0.004 0.004 0.004
(0.015) (0.014) (0.026) (0.012) (0.012) (0.014) (0.011)
Post-1960 controls
Observations 24,281 24,281 19,079 36,947 36,947 20,773
B. In Air
Father and Mother Exposed
Father Exposed Mother Exposed Son's IQ
Son's IQ (1) Son's IQ (2) Father's IQ (3) Son's IQ (4) Son's IQ (5) Father's Exposure (6a) Mother's Exposure (6b)
Three months prior to −0.030 −0.022 0.006 −0.028 −0.031 0.034 −0.022
pregnancy (0.030) (0.034) (0.040) (0.020) (0.020) (0.035) (0.031)
Two months prior to −0.000 0.002 −0.012 0.007 0.009 0.031 −0.025
pregnancy (0.025) (0.027) (0.031) (0.020) (0.020) (0.027) (0.026)
One month prior to 0.014 0.020 0.009 0.015 0.021 −0.006 0.010
pregnancy (0.033) (0.033) (0.019) (0.019) (0.019) (0.034) (0.014)
Pregnancy −0.014 −0.015 −0.011 −0.034 −0.029 0.026 −0.009
month 1 (0.039) (0.037) (0.024) (0.022) (0.023) (0.029) (0.015)
Pregnancy 0.014** 0.014** −0.008 −0.031 −0.030 0.013* −0.030
month 2 (0.006) (0.005) (0.026) (0.025) (0.025) (0.007) (0.023)
Pregnancy −0.030 −0.036 −0.045* −0.027*** −0.028*** −0.043* −0.032**
month 3 (0.027) (0.027) (0.025) (0.010) (0.010) (0.022) (0.015)
Pregnancy −0.077* −0.077* −0.042 −0.026 −0.026 −0.081** −0.031*
month 4 (0.041) (0.041) (0.040) (0.017) (0.017) (0.039) (0.017)
Pregnancy 0.009 0.013 −0.030 0.023 0.026 −0.061* −0.015
month 5 (0.037) (0.034) (0.034) (0.024) (0.023) (0.036) (0.026)
Pregnancy −0.013 −0.007 0.005 0.024 0.027 0.009 0.018
month 6 (0.053) (0.051) (0.027) (0.030) (0.029) (0.026) (0.025)
Pregnancy 0.008 0.009 −0.020 0.006 0.006 −0.003 −0.006
month 7 (0.031) (0.029) (0.027) (0.019) (0.017) (0.025) (0.021)
Pregnancy 0.026 0.027 0.007 0.008 0.008 0.013 −0.006
month 8 (0.028) (0.026) (0.021) (0.021) (0.020) (0.034) (0.032)
Pregnancy 0.016 0.010 0.015 −0.007 −0.001 0.018 −0.019
month 9 (0.028) (0.030) (0.028) (0.023) (0.022) (0.022) (0.022)
Month of birth 0.054 0.053 −0.004 0.001 0.004 0.036 −0.024
(0.038) (0.037) (0.021) (0.023) (0.023) (0.039) (0.027)
After pregnancy 1 −0.010 −0.009 0.028 0.012 0.012 −0.019 0.008
(0.029) (0.030) (0.028) (0.031) (0.030) (0.041) (0.014)
After pregnancy 2 0.015 0.015 −0.005 −0.024 −0.019 0.007 −0.033
(0.035) (0.033) (0.025) (0.024) (0.024) (0.038) (0.026)
After pregnancy 3 0.013 0.012 −0.008 0.007 0.009 0.034 0.017
(0.028) (0.028) (0.026) (0.023) (0.023) (0.041) (0.013)
Post 1960 Controls
Observations 23,378 23,378 18,412 35,745 35,745 11,919
A. In Situ
Father and Mother Exposed
Father ExposedMother ExposedSon's IQ
Son's IQ (1)Son's IQ (2)Father's IQ (3)Son's IQ (4)Son's IQ (5)Father's Exposure (6a)Mother's Exposure (6b)
Three months prior to −0.002 0.001 0.010 −0.012 −0.017 −0.002 −0.001
pregnancy (0.022) (0.023) (0.023) (0.017) (0.017) (0.026) (0.003)
Two months prior to 0.016 0.015 −0.005 0.001 0.000 0.016 0.001
pregnancy (0.022) (0.021) (0.022) (0.021) (0.021) (0.020) (0.003)
One month prior to 0.018 0.027 0.021** −0.015 −0.012 0.031 −0.001
pregnancy (0.019) (0.020) (0.009) (0.019) (0.019) (0.023) (0.004)
Pregnancy 0.015 0.007 −0.045 0.025* 0.021 0.009 0.015
month 1 (0.018) (0.018) (0.029) (0.013) (0.013) (0.021) (0.050)
Pregnancy 0.004 0.006 0.009 0.005 0.005 0.001 0.027
month 2 (0.030) (0.030) (0.025) (0.008) (0.008) (0.033) (0.052)
Pregnancy −0.024*** −0.028*** −0.042** −0.023 −0.024 −0.018** −0.013**
month 3 (0.007) (0.008) (0.018) (0.017) (0.016) (0.007) (0.005)
Pregnancy −0.027** −0.030** −0.043* −0.030** −0.030** −0.022** −0.013
month 4 (0.013) (0.013) (0.026) (0.013) (0.013) (0.011) (0.009)
Pregnancy −0.039* −0.029 0.012 −0.017 −0.013 −0.031 −0.011
month 5 (0.021) (0.022) (0.025) (0.015) (0.015) (0.024) (0.026)
Pregnancy 0.043** 0.038* 0.019 0.000 −0.000 0.044 −0.000
month 6 (0.021) (0.023) (0.023) (0.011) (0.011) (0.028) (0.005)
Pregnancy 0.015 0.014 0.046* −0.027 −0.029 −0.016 −0.006
month 7 (0.020) (0.022) (0.026) (0.021) (0.021) (0.028) (0.012)
Pregnancy 0.027 0.034 −0.005 0.038** 0.037** 0.032* −0.001**
month 8 (0.023) (0.022) (0.023) (0.017) (0.016) (0.019) (0.000)
Pregnancy −0.033 −0.031 −0.027 −0.012 −0.010 −0.028 −0.011
month 9 (0.024) (0.023) (0.023) (0.014) (0.013) (0.022) (0.007)
Month of birth −0.012 −0.006 0.011 −0.013 −0.014 −0.034 −0.015
(0.019) (0.020) (0.025) (0.016) (0.016) (0.026) (0.017)
After pregnancy 1 −0.001 0.001 0.001 0.029 0.031 0.022 0.033
(0.020) (0.019) (0.026) (0.027) (0.028) (0.020) (0.023)
After pregnancy 2 0.008 0.008 0.004 0.032*** 0.033*** 0.022 0.014*
(0.020) (0.019) (0.014) (0.008) (0.008) (0.022) (0.008)
After pregnancy 3 0.014 0.011 0.001 0.003 0.004 0.004 0.004
(0.015) (0.014) (0.026) (0.012) (0.012) (0.014) (0.011)
Post-1960 controls
Observations 24,281 24,281 19,079 36,947 36,947 20,773
B. In Air
Father and Mother Exposed
Father Exposed Mother Exposed Son's IQ
Son's IQ (1) Son's IQ (2) Father's IQ (3) Son's IQ (4) Son's IQ (5) Father's Exposure (6a) Mother's Exposure (6b)
Three months prior to −0.030 −0.022 0.006 −0.028 −0.031 0.034 −0.022
pregnancy (0.030) (0.034) (0.040) (0.020) (0.020) (0.035) (0.031)
Two months prior to −0.000 0.002 −0.012 0.007 0.009 0.031 −0.025
pregnancy (0.025) (0.027) (0.031) (0.020) (0.020) (0.027) (0.026)
One month prior to 0.014 0.020 0.009 0.015 0.021 −0.006 0.010
pregnancy (0.033) (0.033) (0.019) (0.019) (0.019) (0.034) (0.014)
Pregnancy −0.014 −0.015 −0.011 −0.034 −0.029 0.026 −0.009
month 1 (0.039) (0.037) (0.024) (0.022) (0.023) (0.029) (0.015)
Pregnancy 0.014** 0.014** −0.008 −0.031 −0.030 0.013* −0.030
month 2 (0.006) (0.005) (0.026) (0.025) (0.025) (0.007) (0.023)
Pregnancy −0.030 −0.036 −0.045* −0.027*** −0.028*** −0.043* −0.032**
month 3 (0.027) (0.027) (0.025) (0.010) (0.010) (0.022) (0.015)
Pregnancy −0.077* −0.077* −0.042 −0.026 −0.026 −0.081** −0.031*
month 4 (0.041) (0.041) (0.040) (0.017) (0.017) (0.039) (0.017)
Pregnancy 0.009 0.013 −0.030 0.023 0.026 −0.061* −0.015
month 5 (0.037) (0.034) (0.034) (0.024) (0.023) (0.036) (0.026)
Pregnancy −0.013 −0.007 0.005 0.024 0.027 0.009 0.018
month 6 (0.053) (0.051) (0.027) (0.030) (0.029) (0.026) (0.025)
Pregnancy 0.008 0.009 −0.020 0.006 0.006 −0.003 −0.006
month 7 (0.031) (0.029) (0.027) (0.019) (0.017) (0.025) (0.021)
Pregnancy 0.026 0.027 0.007 0.008 0.008 0.013 −0.006
month 8 (0.028) (0.026) (0.021) (0.021) (0.020) (0.034) (0.032)
Pregnancy 0.016 0.010 0.015 −0.007 −0.001 0.018 −0.019
month 9 (0.028) (0.030) (0.028) (0.023) (0.022) (0.022) (0.022)
Month of birth 0.054 0.053 −0.004 0.001 0.004 0.036 −0.024
(0.038) (0.037) (0.021) (0.023) (0.023) (0.039) (0.027)
After pregnancy 1 −0.010 −0.009 0.028 0.012 0.012 −0.019 0.008
(0.029) (0.030) (0.028) (0.031) (0.030) (0.041) (0.014)
After pregnancy 2 0.015 0.015 −0.005 −0.024 −0.019 0.007 −0.033
(0.035) (0.033) (0.025) (0.024) (0.024) (0.038) (0.026)
After pregnancy 3 0.013 0.012 −0.008 0.007 0.009 0.034 0.017
(0.028) (0.028) (0.026) (0.023) (0.023) (0.041) (0.013)
Post 1960 Controls
Observations 23,378 23,378 18,412 35,745 35,745 11,919

The sample includes parents born between 1956 and 1966 in municipalities within a radius of 20 kilometers of the test stations. All regressions are weighted so each parent gets equal weight. Each column represents a separate regression that includes controls for municipality dummies and year of birth by month of birth, controls for parental education, birth order, and the municipality unemployment rate. Standard errors are clustered at the municipality level. Significant at $***$1%, $**$5%, $*$10%. Coefficients in italics survive a correction for multiple hypothesis testing on a 10% significance level. Columns 1 and 4: only pre-1960 controls included. Columns 2 and 5: also includes controls for the second generation. Column 3: Comparative estimates for first-generation men who have at least one son in the sample. Columns 6a/6b: includes only pre-1960 controls for both mothers and fathers.

It is of some interest to contrast the effect of fallout on IQ scores of the first and second generations. To the extent that radiation exposure mostly affects cognitive function (as is suggested by the medical literature), the ratio of the two effects will approximate the intergenerational transmission coefficient for IQ. There is a small but growing literature in economics on intergenerational correlations of IQ scores (Black, Devereux, & Salvanes, 2009, for estimates for Norway, and Black & Devereux, 2011, for an international review). However, little is known about causal intergenerational effects of increasing (or reducing) cognitive abilities in one generation. Because not all first-generation men have sons in our sample, column 3 provides first-generation estimates of the effect of exposure on IQ when the sample includes only men who also have sons in our sample. Given the imprecision of the estimates for air exposure, we focus on the estimates for in situ radiation in table 7a. The second-generation estimates are about $-$0.025; the first generation ones are about $-$0.042. Taken together, these suggest an intergenerational transmission coefficient of about 0.6.24 Importantly, this suggests that a large proportion of the adverse cognitive effects of radiation exposure is passed on from fathers to sons.

For a subsample of sons whose father and mother are both born between 1956 and 1966 in a municipality within 20 kilometers of a measuring station, we can observe both mother's and father's in utero exposure to nuclear fallout. Tables 7a and 7, columns 6a and 6b, present the results from a specification where both are included in the same regression. There is a positive correlation in fallout across parents (the correlation is 0.065 for in situ fallout and 0.018 for fallout in the air when both parents are born between 1956 and 1966), and this can be seen in the reduction in the parental coefficients in columns 6a and 6b of table 7a, relative to the size of these coefficients when exposure of the other parent is excluded. Importantly, we still observe that for both ground and air fallout, exposure of mothers and fathers during months 3 and 4 in utero has long-run effects on the next generation.

## VII.  Conclusion

A large literature has shown that shocks in utero can have lasting effects on children. In this paper, we study one such environmental factor, exposure to radiation, that affects members of all socioeconomic groups. We use variation in radioactive fallout that was generated by nuclear weapons testing in the Northern Hemisphere to identify the effects of exposure in utero on children's long-run outcomes and the long-run outcomes of their children. We find negative long-run effects of exposure to nuclear fallout on cognitive scores, education, and earnings at age 35. While the existing literature has suggested that there are effects of low-dose radiation on early cognitive development, we are the first to show that there are other, persistent, effects on children's outcomes. Importantly, our data also allow us to verify the hypotheses from the medical literature that exposure in months 3 and 4 postconception is particularly harmful in terms of cognitive development.

Another important contribution we make is showing that there are intergenerational effects on cognitive scores for the second generation (the children of people exposed in utero). Importantly, the initial shock to IQ of men exposed to fallout in utero is passed along to their sons, and the transmission is approximately 0.6, suggesting very high persistence. As far as we are aware, this is the first causal evidence on the intergenerational transmission of IQ scores.

Given the lack of knowledge about the fallout in Norway at the time, our estimates are likely unaffected by avoidance behavior or by maternal stress. Interestingly, and contrary to much existing literature, we find no evidence that high-SES families offset these negative effects.

While our findings will not necessarily carry over to other pollutants that potentially have very different biological effects, they speak to the effects of radiation exposure, an increasingly important feature of the modern world. Nowakowski and Hayes (2008) state, “Exposure to ionising radiation is now a part of our everyday lives, coming not only from man-made sources such as medical treatments, colour television, smoke detectors and industrial accidents, but more commonly from environmental background sources, building materials and airplane travel. With plans to expand the use of nuclear power plants and the possible threat of “dirty bombs,” the issue of radiation exposure in either high or low doses becomes even more relevant and part of the public dialogue.” Indeed, a large amount of radioactivity was discharged after damage to the cooling systems of several reactors in the Fukushima nuclear power plant in March 2011. We find adverse effects on children in a period where pregnant mothers did not have the required information to enable them to take actions to avoid exposure to radiation. This suggests an important role for information campaigns targeted to pregnant women in situations where they are at risk of radiation exposure.

## Notes

1

See Almond, Currie, and Duque (2018) and Almond and Currie (2011) for reviews of this vast literature.

2

Some of the most convincing work shows that exposure to famine while in utero has effects on the birth weight and sex composition of the offspring of the exposed children (Almond et al., 2010; Painter et al., 2008).

3

More recently, Halla and Zweimüller (2014) study the effect of Chernobyl on Austrians who were in utero and find evidence of adverse effects; they also examine whether parents change their investment behavior and find some evidence of compensating parental responses. There has also been much work in the medical literature studying the effect of nuclear fallout on the health and cognitive ability of children in utero during the bombings of Hiroshima and Nagasaki (see Otake & Schull, 1984, 1998; Donnelly et al., 2011).

4

It was not the Norwegian government that established test stations in Norway during the 1950s but the U.S. and British military, who were interested in collecting information about test activity in the Soviet Union. We checked the electronic newspaper archive of the Norwegian National Library for articles referring to nuclear weapon testing in the Soviet Union between 1950 and 1975. There are no articles in Norwegian newspapers describing the proximity to nuclear weapon tests in Novaya Zemlya or nuclear fallout measures in Norway.

5

Stress during pregnancy has been linked to poor infant health outcomes (Black, Devereux, & Salvanes, 2016; Persson & Rossin-Slater, 2018).

6

On October 10, 1963, a partial test ban treaty came into force, banning nuclear tests in the atmosphere, underwater, and in space. The treaty was not signed by France and China; as a result, the last atmospheric explosion was performed by China as late as October 1980.

7

Due to the Coriolis forces, the cold dry air moves away from the pole twisting westward, resulting in the so-called polar easterlies. These winds carry air from northern Russia southwest toward the Norwegian Sea and Iceland. At around 60 degrees north, the airstream enters the low-pressure zone and the air is brought eastward again toward the Norwegian coast.

8

We use the phrase in situ to denote nuclear fallout that has been deposited to the ground (as distinct from being suspended in the air). Total beta fallout measures beta decay, a radioactive process in which an electron is emitted from the nucleus of a radioactive atom. Beta particles are able to penetrate living matter and can cause spontaneous DNA mutation.

9

No evidence of a radiation effect has been seen among children exposed prior to the 8th week or subsequent to the 25th week after conception (Otake & Schull, 1984). Moreover, Otake and Schull (1998) report that the risk of severe mental retardation was five times greater for persons exposed during weeks 8 to 15 postconception than for individuals exposed during weeks 16 to 25 postconception.

10

Oftedal (1989) evaluates the effect of radiation exposure on scholastic achievement of the 1965 cohort in Norway by hypothesizing that children from the west of Norway should have been more exposed than those from the east and the degree of exposure should vary by season of birth. He compares school test grades from a 10% sample of seventh graders in the two regions and finds deviations by region that differ by month of birth. He concludes that scholastic achievement is reduced in children exposed in utero to radiation. Unfortunately, the work is limited in that he studies only one cohort born two years after the test ban treaty, and he has no measures of geographic dispersion of radiation.

11

Note that this is particularly conservative; among these sixteen coefficients, we have strong priors that exposure prior to conception or after birth is unlikely to affect later outcomes.

12

More detail about the data is in the online appendix.

13

The locations of measurement stations for radioactivity are (from north to South in Norway): Vadsø, Tromsø, Bardufoss, Bodø, Værnes (close to Trondheim in mid-Norway), Røros, Ålesund, Bergen, Finse, Sola (close to Stavanger), Gardermoen (close to Oslo), Kjeller (also close to Oslo), and Kjevik (close to Kristiansand).

14

We obtained the data for air and ground deposition measured in picoCurie/m$3$ and picoCurie/m$2$, respectively. Bergan digitalized the original protocols to obtain the radiation data (Bergan, 2002, 2010; Bergan & Steenhuisen, 2012).

15

Bergan (2002) states that “the fallout is correlated to the amount of precipitation and concentration in air, and the deposited radioactivity is proportional to monthly precipitation” (p. 206). We also used the in situ total beta directly without weighting by the relative rainfall and obtained very similar results. This is unsurprising as we include only municipalities that are within 20 kilometers of a test station.

16

We have also rescaled the IQ score so that in each test year, it has mean 100 and standard deviation 15. When we use this as the dependent variable, we find a slightly larger effect of about 0.9 IQ point. These estimates are in the online appendix.

17

We have tested for equality of parameter values across different months. For in situ fallout, in the main specification, all coefficients are significantly different from months 3 and 4 (at the 5% level) except for one month prior to pregnancy, pregnancy month 1, and pregnancy month 2. For fallout in air, there are fewer significant differences across months.

18

The effects of exposure during months 3 and 4 on educational attainment and earnings do not significantly differ by gender.

19

We report robustness checks in the online appendix. These include controlling for rainfall, using different income measures, and including both measures of fallout (air and ground) in the same regression. We also test for selection and find no evidence that radiation affected fertility rates or the ratio of still to live births. While our estimates are not directly comparable to those of Almond et al. (2009), as their main specifications use discrete measures of the degree of regional exposure and they use different measures of radioactive fallout, it is still useful to try to get a sense of relative magnitudes. In the online appendix, we provide a detailed account of how we do this. We find that our estimates for education are close to theirs, but our estimates for the cognitive score are about 10 times smaller than theirs. However, our results are much more precisely estimated.

20

Based on much smaller samples of observations, previous medical studies find that in utero exposure to iodizing radiation higher than 50 to 100 mSv during weeks 8 to 15 postconception may lead to detectable brain damage (Donnelly et al., 2011). However, Nowakowski and Hayes (2008) suggest that even smaller doses of radiation may cause DNA damage and thereby subclinical cognitive deficits.

21

Almond et al. (2010) estimate what they describe as “echo effects” of the 1959–1961 Chinese famine on birth weight and sex composition of babies born to women who were pregnant during the famine. Concurrent with our work, Tan, Tan, and Zhang (2015) examine cognitive scores of children born to parents who were exposed to the Chinese famine early in life and find that cognitive abilities of both girls and boys born to exposed fathers were affected. However, children born to exposed mothers were not affected.

22

In our sample, the mean age at first child is 28.8 for men and 25.9 for women. We show in the online appendix that selection is unlikely to be a problem as exposure in pregnancy months 3 and 4 has no effect on the probability of having a son that has taken the military exams by 2010.

23

While some of our intergenerational coefficients do not survive the correction for multiple hypothesis testing, we do not consider this a serious problem; given the results for the first generation, we have a strong prior that exposure in months 3 and 4 will affect on the outcomes of the next generation.

24

We know from our earlier work that the intergenerational correlation in the IQ score in Norway is about 0.4, so our finding of an implied estimate of about 0.6 is larger than anticipated. There is a substantial confidence interval around this estimate.

## REFERENCES

Aizer
,
A.
, and
L.
Stroud
, “
Education, Knowledge and the Evolution of Disparities in Health
,”
Brown University working paper
(
2011
).
Almond
,
D.
, and
J.
Currie
, “Human Capital Development before Age Five” (pp.
1315
1486
), in
D.
Card
and
O.
Ashenfelter
, eds.,
Handbook of Labor Economics
(
Amsterdam
:
Elsevier
,
2011
).
Almond
,
D.
,
J.
Currie
, and
V.
Duque
, “
Childhood Circumstances and Adult Outcomes: Act II
,”
Journal of Economic Literature
56
:
4
(
2018
),
1360
1446
.
Almond
,
D.
,
L.
Edlund
, and
M.
Palme
, “
Chernobyl's Subclinical Legacy: Prenatal Exposure to Radioactive Fallout and School Outcomes in Sweden,
Quarterly Journal of Economics
124
(
2009
),
1729
1772
.
Almond
,
D.
,
L.
Edlund
,
H.
Li
, and
J.
Zhang
, “
Long-Term Effects of the 1959–61 China Famine: Mainland China and Hong Kong
” (pp.
321
350
), in
T.
Ito
and
A.
Rose
, eds.,
The Economic Consequences of Demographic Change in East Asia
(
Chicago
:
University of Chicago Press
,
2010
).
Auvinen
,
A.
,
M.
Vahteristo
,
H.
Arvela
,
M.
Suomela
,
T.
Rahola
,
M.
Hakama
, and
T.
Rytömaa
, “
Chernobyl Fallout and Outcome of Pregnancy in Finland,
Environmental Health Perspectives
109
(
2001
),
179
185
.
Bergan
,
T. D.
, “
Radioactive Fallout in Norway from Atmospheric Nuclear Weapons Tests,
60
(
2002
),
189
208
.
Bergan
,
T. D.
Oversikt over total beta-målinger i luft og nedbør utført ved Forsvarets forskningsinstitutt i perioden 1956–1982,
memo, Statens strålevern
(
2010
).
Bergan
,
T. D.
, and
F.
Steenhuisen
, “
Deposition of Radioactive Fallout in Norway due to Atmospheric Nuclear Weapons Testing,
memo, Norwegian Directorate of Civil Protection and Emergency Planning
(
2012
).
Bhuller
,
M.
,
M.
, and
K. G.
Salvanes
, “
Life Cycle Earnings, Education Premiums and Internal Rates of Return,
Journal of Labor Economics
35
(
2017
),
993
1030
.
Black
,
S. E.
, and
P. J.
Devereux
, “Recent Developments in Intergenerational Mobility” (pp.
1487
1541
), in
D.
Card
and
O.
Ashenfelter
, eds.,
Handbook of Labor Economics
(
Amsterdam
:
Elsevier
,
2011
).
Black
,
S. E.
,
P. J.
Devereux
, and
K. G.
Salvanes
, “
Like Father, Like Son? A Note on the Intergenerational Transmission of IQ Scores,
Economics Letters
105
(
2009
),
138
140
.
Black
,
S. E.
,
P. J.
Devereux
, and
K. G.
Salvanes
Does Grief Transfer across Generations? Bereavements during Pregnancy and Child Outcomes,
American Economic Journal: Applied Economics
8
(
2016
),
193
223
.
Brent
,
R. L.
, “
The Effect of Embryonic and Fetal Exposure to X-Ray, Microwaves, and Ultrasound: Counseling the Pregnant and Nonpregnant Patient about These Risks,
Seminars in Oncology
16
(
1989
),
347
368
.
Donnelly
,
E. H.
,
J. M.
Smith
,
E. B.
Farfan
, and
I.
Ozcan
, “
Prenatal Radiation Exposure: Background Material for Counseling Pregnant Patients following Exposure to Radiation,
Disaster Medicine and Public Health Preparedness
5
(
2011
),
62
68
.
Hall
,
E. J.
, “
39
(
2009
),
57
64
.
Halla
,
M.
, and
M.
Zweimüller
, “
Parental Response to Early Human Capital Shocks: Evidence from the Chernobyl Accident
,”
IZA discussion paper
7968
(
2014
).
Hvinden
,
T.
, and
A.
Lillegraven
, “
Cæsium-137 and Strontium-90 in Precipitation, Soil and Animals in Norway,
Nature
192
(
1961
),
1144
1146
.
Lüning
,
G.
,
J.
Scheer
,
M.
Schmidt
, and
H.
Ziggel
, “
Early Infant Mortality in West Germany before and after Chernobyl,
Lancet
334
(
1989
),
1081
1083
.
Nowakowski
,
R. S.
, and
N. L.
Hayes
, “
Radiation, Retardation and the Developing Brain: Time Is the Crucial Variable,
Acta Paediatrica
97
(
2008
),
527
531
.
Oftedal
,
P.
, “Scholastic Achievement in Relation to Fetal Exposure to Radioactive Fallout in Norway” (pp.
323
353
), in
K. F.
Baverstock
and
J. W.
Stather
, eds.,
Low Dose Radiation: Biological Bases of Risk Assessment
(
London
:
Taylor & Francis
,
1989
).
Otake
,
M.
, and
W. J.
Schull
, “
In Utero Exposure to A-Bomb Radiation and Mental Retardation: A Reassessment,
57
(
1984
),
409
414
.
Otake
,
M.
, and
W. J.
Schull
Review: Radiation-Related Brain Damage and Growth Retardation among the Prenatally Exposed Atomic Bomb Survivors,
74
(
1998
),
159
171
.
Painter
,
R. C.
,
C.
Osmond
,
P.
Gluckman
,
M.
Hanson
,
D. I. W.
Phillips
, and
T. J.
Roseboom
, “
Transgenerational Effects of Prenatal Exposure to the Dutch Famine on Neonatal Adiposity and Health in Later Life,
BJOG: An International Journal of Obstetrics and Gynaecology
115
(
2008
),
1243
1249
.
,
P.
, and
M.
Rossin-Slater
, “
Family Ruptures, Stress, and the Mental Health of the Next Generation,
American Economic Review
108
(
2018
),
1214
1252
.
Romano
,
J. P.
,
A.
Shaikh
, and
M.
Wolf
, “
Hypothesis Testing in Econometrics,
Annual Review of Economics
2
(
2010
),
75
104
.
Romano
,
J. P.
, and
M.
Wolf
, “
Exact and Approximate Stepdown Methods for Multiple Hypothesis Testing,
Journal of the American Statistical Association
100
(
2005
),
94
108
.
Sperling
,
K.
,
J.
Pelz
,
R.-D.
Wegner
,
A.
Dorries
,
A.
Gruters
, and
M.
Mikkelsen
, “
Significant Increase in Trisomy 21 in Berlin Nine Months after the Chernobyl Reactor Accident: Temporal Correlation or Causal Relation?
British Medical Journal
309
(
1994
),
158
162
.
Tan
,
C. M.
,
Z.
Tan
, and
X.
Zhang
, “
Sins of the Father: The Intergenerational Legacy of the 1959–61 Great Chinese Famine on Children's Cognitive Development,
Rimini Centre for Economic Analysis, working paper 08_14
(
2014
).
UNSCEAR
,
Ionizing Radiation: Sources and Biological Effects
(
New York
:
United Nations
,
1982
).

## Author notes

We thank Tone D. Bergan at the Norwegian Radiation Protection Authority for providing us with the data she digitalized from the original archives at the Norwegian Defense Research Establishment (FFI), and for her help in clarifying measurement issues. This work was partially supported by the Research Council of Norway through its Centres of Excellence Scheme, FAIR project 262675. We appreciate comments from presentations at the NBER Summer Institute, Bocconi, Columbia University, MIT, Northwestern University, Boston College, National University of Ireland Maynooth, Oxford University, University of Essex, University of Toronto, Simon Fraser University, Harrison School Chicago University, Northwestern University, Statistics Norway, Stockholm School of Economics, University of Bern, University of British Columbia, University of California Santa Barbara, University of California San Diego, University of St. Gallen, Trinity College Dublin, and a workshop at the Norwegian School of Economics.