Climate Change and Birth Weight
University of California, Santa
Barbara and NBER
Department of Economics
2127 North Hall
University of California
Santa Barbara, CA 93106
MIT and NBER and
Department of Economics
50 Memorial Drive, E52-359
Cambridge, MA 02142-1347
(p) 617- 452-4127
University of Chicago and
University of Chicago Booth
School of Business
5807 S. Woodlawn Avenue
Chicago, IL 60637
Session Title: The Economic Impacts of Climate Change
Presiding: Michael Greenstone, Massachusetts Institute of Technology
Discussant: Douglas Almond
Climate Change and Birth Weight
By OLIVIER DESCHÊNES, MICHAEL GREENSTONE, and JONATHAN GURYAN∗
There is a growing consensus that emissions of greenhouse gases due to human activity
will alter the earth’s climate, most notably by causing temperatures, precipitation levels, and
weather variability to increase. The design of optimal climate change mitigation policies
requires estimates of the health and other benefits of reductions in greenhouse gases; current
evidence on the magnitudes of the direct and indirect impacts, however, is considered
insufficient for reliable conclusions (A.J. McMichael et al, 2003).1
In addition to the overall predicted warming trend, one important feature of many global
climate change forecasts is an increased incidence of very high and low temperatures. To
provide evidence of the potential benefits of greenhouse gas reductions, this paper documents
whether the temperature variation predicted to be part of climate change, including extreme high
and low temperatures, historically have had negative health consequences through their effect on
babies while in utero.
Using individual-level data on more than 37.1 million births, we find that exposure to
extreme hot temperatures during pregnancy leads to lower birth weights. We combine this
finding with estimates of the distribution of future daily temperatures from state of the art
climate change predictions. We then use these predictions of the effect of climate change on the
∗ Department of Economics, 2127 North Hall, University of California, Santa Barbara, CA 93106, and NBER
(email: firstname.lastname@example.org); MIT, Department of Economics, 50 Memorial Drive, E52-359, Cambridge, MA
02142, and Brookings Institution and NBER (email: email@example.com); University of Chicago Booth School of
Business, 5807 S. Woodlawn Avenue, Chicago, IL 60637 (email: firstname.lastname@example.org). We thank Douglas
Almond for his especially insightful discussion. Henry Swift provided outstanding research assistance. We
gratefully acknowledge support from the University of Chicago’s Chicago Energy Initiative.
1 See Richard S.J. Tol (2002a, b) for overall estimates of the costs of climate change, which are obtained by
summing costs over multiple areas including human health, agriculture, forestry, species/ecosystems, and sea level
rise. Olivier Deschenes and Michael Greenstone (2007) provides evidence on the impacts on the U.S. agricultures
sector. Deschenes and Greenstone (2008) and Robin Burgess et al (2009) estimate the mortality impacts of climate
change in the U.S. and India, respectively.
distribution of daily temperatures to estimate the predicted effect of global climate change on
future birth weights by the end of the century. These estimates imply that mean birth weights
will decrease on average by 0.22 percent (7.5 grams) among whites and by 0.36 percent (11.5
grams) for blacks by the end of the century. Further, the impact is not spread evenly through the
birth weight distribution. We find an estimated 5.9% increase in the probability of a low birth
weight birth (defined as less than 2,500 grams) for whites and a 5.0% increase for blacks.2
A small set of studies attempt to explain seasonality in birth weight with exposure to
extreme temperatures in utero (L. J. Murray et al, 2000; D.A. Lawlor, D.A. Leon, and G.D.
Smith, 2005). Although there is some evidence of a negative relationship between mean
temperatures and birth weight, the absence of a cohesive framework, limited statistical models,
and relatively small samples have limited the development of clear stylized facts. Further, no
consensus has emerged concerning the physiological mechanisms linking heat and cold stress in
utero, although fetal nutrient intake is often mentioned without direct evidence.
Our detailed historical temperature data allow us to date specifically each shock to the
conditions experienced in utero. We are thus able to estimate not only the effects of exposure to
extreme temperature while in utero, but to explore how this effect varies according to the timing
during gestation. This allows us to shed light on the mechanisms at work in fetal programming.
Of particular interest, we find that as much as 95 percent of the effect of in utero exposure to
very hot ambient temperatures occurs for exposure during the second and third trimesters.
I. Data Sources and Summary Statistics
2 This study of the impact of extreme temperature exposure on birth weight is related to the fetal origins hypothesis
(David J. P. Barker, 1994). In its extended form, it states that economic and environmental conditions during
pregnancy may have long-lasting impacts on long-term health and socioeconomic status (see e.g., Douglas Almond,
2006; Sandra Black et al, 2007; Heather Royer 2009). A key variable in this literature is birth weight, which is
considered by some to be a summary measure of the fetal experience. Future work will investigate whether
exposure to extreme temperatures affects long-run economic and health outcomes, whether through its effect on
birth weight or on latent factors that do not affect birth weight directly.
To implement the analysis, we collected the most detailed and comprehensive data
available on infant birth outcomes, weather, and predicted climate change. This section
describes these data and reports summary statistics. See Olivier Deschenes and Michael
Greenstone (2008) for further details.
A. Data Sources
Birth Weight Data. Microdata on birth weight and exact date of birth are taken from the
1972-1988 National Center for Health Statistics Natality Detail Files (NDF). The data include
exact date of birth, race, sex, birth weight, county of birth, mother’s marital status, mother’s
education, and the exact date of the mother’s last menstrual period. From the latter we infer the
date of ‘conception’ and measure the trimesters of the pregnancy relative to that date. This is
preferable to dating gestation age relative to the date of birth, since gestation length itself may be
an outcome of exposure to extreme temperatures. Observations are dropped from the analysis
sample if either race, sex, birth weight or date of last menstrual period is missing. The sample
includes 37.1 million singleton births to mothers aged 16-45 in the continental 48 states and D.C.
Weather Data. The weather data are drawn from the National Climatic Data Center
Summary of the Day Data (File TD-3200). The key variables for our analysis are the daily
maximum and minimum temperature, which we average to construct daily average temperature.
Station-level weather data are aggregated at the county level by taking an inverse-distance
weighted average of all the valid measurements from stations that are located within a 200 km
radius of each county’s centroid.
Climate Change Prediction Data. Climate predictions are taken from the application of
the A2 scenario to the National Center for Atmospheric Research’s Community Climate System
Model (CCSM) 3, which is a state of the art coupled atmospheric-ocean general circulation
model (National Center for Atmospheric Research, 2007). We use the CCSM 3 A2 daily
temperature predictions for grid points throughout the continental U.S. for the years 2070-2099.
From the CCSM predictions, we assign each county a daily weather realization, aggregating in
the same way we aggregate the contemporaneous temperatures. Each county’s end of century
predicted climate is the simple average of the predicted weather realizations for the 2070-2099
B. Descriptive Statistics
Birth Outcomes Statistics. The main analysis will examine the effect of extreme
temperatures on birth weights, by race. There are substantial differences in birth outcomes
between blacks and whites. Mean birth weights are higher for whites (3,417 grams), by about
235 grams. The percent of ‘low birth weight’ births is more than twice as large for blacks than
whites (9.3% compared to 4.1%), while the black gestational period is 5 days shorter on average.
For both races there are marked seasonal differences in birth outcomes, including birth weight
and the number of births, across quarters of conception.
If differences in fertility by date of conception reflect differences in socioeconomic,
health, or overall fertility attributes, associations between temperature exposure and pregnancy
outcomes may be confounded. We address the potential confounding of socioeconomic and
health attributes with the timing of conception by utilizing county-level inter-annual variation in
temperature and controlling for a smooth function of the date of conception.
Weather and Climate Change Statistics. The regressors of interest in the analysis are the
number of days during the gestation period in which a county’s daily average (mean of the
minimum and maximum) temperature falls into each of five temperature bins (< 25° F, 25° - 45°
F, 45° - 65° F, 65° - 85° F, > 85° F). In Figure 1, the white bars depict the average distribution
of daily mean temperature exposure over the course of the typical pregnancy in the U.S. between
1972 and 1988. In the 270 days after conception, the average pregnancy in the U.S. is exposed
to 13.4 days with a temperature less than 25° F, 252.8 days in the 25° - 85° F range, and 3.8 days
greater than 85° F. In the subsequent analysis, these bins are calculated separately by trimester
of the pregnancy so to allow for substantial flexibility in the semi-parametric modeling of the
effect of temperature on birth weight. The average pregnancy is exposed to 13.4 days lower than
25° F and 3.8 days warmer than 85° F.
After we estimate the effect of temperature exposure in utero on birth weight, we will
estimate the implied effect on birth weight of the 2070-2099 CCSM 3 A2 temperature
predictions.3 The black bars in Figure 1 reports on the predicted change in the number of days in
each of the temperature bins, assuming the geographic distribution of births remains constant.
The CCSM 3 A2 predictions are that the typical pregnancy will be exposed to 7.8 fewer days
below 25° F per year and 29.9 more days in excess of 85° F. The majority of this temperature
increase comes from a reduction of 24.1 days in the 45° - 65° F bin.4
II. Econometric Strategy
The subsequent results are based on the fitting of the following equation:
where the i denotes the individual birth, c the county, t the year, g the demographic group
(defined by sex and race), and d represents the ‘day of the year’ of conception, which ranges
from 1-365. Therefore, the indices t and d fully determine the exact date of conception.
++++++= ∑∑∑ )( 1 332211
3 For comparability, we follow much of the previous literature on climate change and focus on the temperatures
predicted to prevail at the end of the century. We were unable to obtain CCSM 3 A2 predictions for the past, so the
magnitude of any model errors is unknown and we cannot adjust the predicted temperature changes for such errors.
4 We focus on averages although there is substantial heterogeneity in historical temperatures and climate change
The two dependent variables are log birth weight and an indicator for low birth weight
(i.e., less than 2,500 grams). Individual level controls for mother’s age, fertility history, marital
status, and educational attainment are included in the vector Xi. The county-year-demographic
group fixed effects, ctg
, account for cross-sectional variation in temperature and birth weight
and a quartic in conception date ranges (ranging from 1-365), f(dg), accounts for demographic
group-specific seasonality. These variables are included to capture the effect of any secular
difference in birth outcomes within year that is independent of temperature exposure. The last
term in the equation is the stochastic error term, i
. We report standard errors that are clustered
at the county-by-demographic group level. We also weight by the ‘record weight’ variable
contained in the NDF.
The variables of interest, , , and , are the measures of
temperature exposure during each trimester of the gestational period. The j index refers to the
five daily mean temperature bins described above. The choice of five temperature bins is
motivated by an effort to allow the data, rather than parametric assumptions, to determine the
birth weight-temperature exposure relationship, while also obtaining estimates that have
5 By conditioning on county-year-demographic group effects, the associated
parameters are identified from county-specific deviations in weather about the county averages,
after controlling for county-specific annual shocks. Due to the unpredictability of temperature
fluctuations, we presume that this variation is orthogonal to unobserved determinants of birth
Figure 2 plots the estimated regression coefficients associated with each temperature bin,
5 A similar methodology is used in Deschenes and Greenstone (2007).
by trimester, from a model that pools all births and includes gender-by-race indicators. The
panels A, B, and C correspond to the three trimesters. In each case the reference bin is 45°-65°
F, so each coefficient measures the impact of an additional day in a given bin in a given trimester
on the log birth weight, relative to the impact of a day between 45° and 65° F. The figure also
plots the 95% confidence interval associated with the estimates.
The figures show a strong negative relationship between birth weight and temperature. In
each of the three trimesters, the response function is downward-sloping, with the relationship
more pronounced for the second and third trimesters. For all three trimesters, exposure to hot
days is associated with a statistically significant decline in birth weight, which ranges in
magnitude from 0.003% to 0.009% per such day, relative to a day in the reference category. The
third trimester effect is likely an underestimate because many of the births are delivered less than
270 days after conception.
Table 1 presents calculations of the predicted impacts of climate change on birth weight
outcomes, based on CCSM 3 A2 end of century predictions. Panel A pertains to whites and
panel B to blacks. To calculate the implied impact of climate change on birth weight for each
county on each conception date, we multiply the estimated θ’s for each trimester by the predicted
change in days in each temperature bin for each trimester. We then report the weighted average
across counties and over the calendar year. It is straightforward to calculate the standard error of
this prediction (reported below the estimate in parentheses), since the estimated mortality change
is a linear function of the parameters.
Columns (1) – (3) report this calculation for gestational trimesters 1-3 while column (4)
sums the impact across trimesters to obtain the full effect on log birth weights. For each panel
we report three temperature categories (i.e., < 25° F, 25° - 85° F, and > 85° F) along with the
‘total impact’, which sums across the whole daily temperature distribution. Taken as a whole,
the CCSM 3 A2 results suggest that climate change would lead to a reduction in average birth
weights of 0.22 percent (whites) to 0.36 percent (blacks). In terms of actual weights, these
reductions correspond to losses of 7.5 and 11.5 grams respectively. The null hypothesis of a zero
effect is easily rejected at conventional significance levels.
Colum (5) repeats this exercise, except the dependent variable is now an indicator for
whether the birth is designated as low birth weight. The estimated impacts for this outcome are
substantially larger. In particular, climate change is predicted to increase the fraction of low
birth weight births by roughly 5.9 percent for whites and 5.0 percent for blacks. It is evident that
the impacts of temperature on weight are not equal through the birth weight distribution.
Two further findings are noteworthy. First, the birth weight reduction predicted by
climate change models is almost entirely attributable to the predicted increase in exposure to
‘hot’ temperatures. Across both races and outcome variables, 79%- 94% of the estimated change
is due to the increased number of days exceeding 85° F. Second, the impacts are concentrated in
the second and third trimesters. For example, between 91% and 97% of the full gestational
impact of climate change on birth weights comes from the second and third trimesters.
IV. Conclusions and Future Directions
In addition to a predicted increase in average temperature, many global climate change
models contain the oft-overlooked prediction that there will be a large increase in the number of
very hot days. Using inter-annual variation from long-run county-level temperature
distributions, our estimates imply that exposure to such extreme ambient temperatures have
deleterious effects on fetal health, causing a decrease in birth weight and an increase in the
probability of low birth weight. The sensitivity of birth weight to temperature is concentrated
almost completely in the second and third trimesters of the pregnancy, though it is not yet clear
whether this is due to a compensatory mechanism that operates during the first trimester or
because negative shocks to fetal health in the first trimester show up in indicators other than birth
weight. Our analysis also does not identify the effect of temperature exposure that works
through nutrient accumulation separately from the direct effect on gestation length.
There are a few important caveats to these calculations and, more generally, to the
analysis. The estimated impacts likely overstate the impacts, because the analysis relies on inter-
annual variation in weather, and a greater array of adaptations (e.g., migration) will be available
in response to permanent climate change. Additionally, an effort to project outcomes at the end
of the century requires a number of strong assumptions, including that the climate change
predictions are correct, relative prices will remain constant, the same medical technologies will
prevail, and the demographic characteristics of parents and their geographical distribution will
remain unchanged. The benefit of these strong assumptions is that they allow for a transparent
analysis that isolates the impact of temperature on birth weight and is based on data rather than
on unverifiable assumptions.
The important question left unanswered by this paper is whether these changes in birth
weight are related to welfare losses; for example, through changes in health, human capital,
income, or neonatal mortality. If there is indeed a relationship between in utero exposure to high
temperature and welfare measures, then these impacts should become part of the calculations of
the benefits of greenhouse gas reductions. This is an important topic for future research.
Almond, Douglas. 2006. “Is the 1918 Influenza Pandemic Over? Long-Term Effects of In Utero
Influenza Exposure in the Post-1940 U.S. Population.” Journal of Political Economy, 114(4):
Barker, David J. P. 1994. Mothers, Babies, and Disease in Later Life. London: BMJ Publishing
Black, Sandra E., Paul J. Devereux, and Kjell G. Salvanes. 2007. “From the Cradle to the Labor
Market? The Effect of Birth Weight on Adult Outcomes.” The Quarterly Journal of Economics,
Burgess, Robin, Olivier Deschenes, Dave Donaldson, and Michael Greenstone. 2009. “Climate
Change, Consumption Smoothing, and Mortality in India.” Mimeo, MIT.
Deschênes, Olivier, and Michael Greenstone. 2007. “The Economic Impacts of Climate Change:
Evidence from Agricultural Profits and Random Fluctuations in Weather.” American Economic
Review, 97(1): 354-385.
Deschênes, Olivier, and Michael Greenstone. 2008. “Climate Change, Mortality, and Adaptation:
Evidence from Annual Fluctuations in Weather in the US.” MIT Working Paper Series 07-19.
Lawlor D.A., D.A. Leon, and G.D. Smith. 2005. “The Association of Ambient Outdoor
Temperature Throughout Pregnancy and Offspring Birthweight.” BJOG: An International
Journal of Obstetrics & Gynaecology, 112(5): 647-657.
McMichael, A.J., D.H. Campbell-Lendrum, C.F. Corvalan, K.L. Ebi, A. Githelo, J.D. Scheraga,
and A. Woodward (Eds.). 2003. Climate Change and Human Health – Risks and Responses.
Geneva: World Health Organization.
Murray, L. J., D.P. O`Reilly, N. Betts, C.C. Patterson, G. Davey Smith, and A.E. Evans. 2000.
“Season and Outdoor Ambient Temperature: Effects on Birth Weight.” Obstetrics and
Gynecology, 96(5): 689-695.
National Center for Atmospheric Research. 2007. “Community Climate System Model.”
http://www.cgd.ucar.edu/csm/ (accessed May, 2007).
Royer, Heather. 2009. “Separated at Girth: US Twin Estimates of the Effects of Birth Weight.”
American Economic Journal - Applied Economics, 1(1): 49–85.
Tol, Richard S.J. 2002a. “Estimates of the Damage Costs of Climate Change, Part I: Benchmark
Estimates.” Environmental and Resource Economics, 21(1): 47-73.
Tol, Richard S.J. 2002b. “Estimates of the Damage Costs of Climate Change, Part II: Dynamic
Estimates.” Environmental and Resource Economics, 21(2): 135-160.
Table 1: Estimated Impact of Climate Change on Birth Weights, Based on CCSM 3 A2 Predictions
Impact of change in temperature exposure during: Full Gestational Period Full Gestational Period
1st Trimester 2nd Trimester 3rd Trimester (birth weight <= 2500 gr)
(1) (2) (3) (4) (5)
Total impact (percent) -0.0002 -0.0010 -0.0010 -0.0022 0.0585
(0.0002) (0.0001) (0.0001) (0.0004) (0.0098)
1. Days Less Than 25F 0.0001 -0.0001 -0.0001 0.0000 0.0049
(0.0000) (0.0000) (0.0000) (0.0001) (0.0024)
2. Days Between 25F and 85F 0.0000 -0.0002 -0.0001 -0.0003 0.0073
(0.0000) (0.0000) (0.0000) (0.0000) (0.0000)
3. Days More Than 85F 0.0003 -0.0007 -0.0009 -0.0019 0.0463
(0.0001) (0.0001) (0.0002) (0.0003) (0.0073)
Total impact (percent) -0.0001 -0.0018 -0.0017 -0.0036 0.0495
(0.0006) (0.0004) (0.0004) (0.0010) (0.0129)
1. Days Less Than 25F 0.0000 -0.0001 -0.0001 -0.0002 0.0032
(0.0000) (0.0000) (0.0000) (0.0001) (0.0022)
2. Days Between 25F and 85F 0.0000 -0.0001 0.0000 -0.0001 0.0022
(0.0001) (0.0000) (0.0000) (0.0000) (0.0011)
3. Days More Than 85F -0.0002 -0.0016 -0.0016 -0.0034 0.0441
(0.0005) (0.0004) (0.0004) (0.0009) (0.0118)
Notes: The estimates are from fixed effect regressions models estimated separately by race. Each model includes county-by-year
fixed effects, a quartic polynomial in date of conception, an indicator for the baby’s sex, and controls for mother characteristics.
The dependent variables are the log birth weight and an indicator for low birth weight status (i.e., < 2,500 grams). Standard
errors are clustered at the county-by-race level. See the text for further details
Figure 1: Distribution of Historical and Predicted Exposure Days in 5 Daily Temperature Bins
<25F 25-45 45-65 65-85 >85F
Exposure Days in 5 Temperat ure Bins, for 270 Day Gestation
1972-1988 Average Predicted Change, CCSM 3 A2
Figure 2: Impact of Temperature Exposure During Gestational Period on Birth Weight
(A) First Trimester
<25F 25-45 45-65 65-85 >85F
Impact of Temperatur e Exposure During Fi rst Trimester on Log Bir th Weight
-2 st d err Percent Change in Birt h Weight +2 std err
(B) Second Trimester
<25F 25-45 45-65 65-85 >85F
Impact of Temperatur e Exposure During Second Tr imester on Log Birt h Weight
-2 std err Perc ent Change in Birt h Weight +2 st d err
(C) Third Trimester
<25F 25-45 45-65 65-85 >85F
Impact of Temperature Exposure Duri ng Third Trime ster on Log Birt h Weight
-2 std err Per cent Change in Bi rth Wei ght +2 st d err