Heat, Infant Mortality and Adaptation: Evidence from India · The authors also thank Panchajanya Banerjee and Aniruddha Ghosh for helping us with the weather and lights data. All

Heat, Infant Mortality and Adaptation: Evidence from India

Rakesh Banerjee and Riddhi Bhowmick

University of Southern California

August 2016

Abstract

Predicted rise in global temperatures and more frequent extreme weather events are likely tohave significant impact on human health. This paper asks if higher temperatures during pregnancyincrease infant mortality and explores if two public policies, an employment guarantee program(NREGA)- and a community health care worker program (ASHA)- act as adaptation strategies.We construct birth records, using the pregnancy history of mothers from District Level House-hold Survey (DLHS) of India, and combine it with daily level temperature data. We disentanglethe effect of season of birth from temperature faced during pregnancy by comparing individualsborn in the same district quarter in different years. The results show higher temperatures duringpregnancy increase infant mortality in rural areas, while no such effects are observed in urbanareas. There can be two plausible mechanisms at play behind this effect a) higher temperaturescould lead to reduced agricultural output and increased food prices and in absence of consumptionsmoothing may lead to reduced fetal nutrition b) higher temperatures could directly impact moth-ers physiologically or indirectly through increased disease prevalence. An employment guaranteeprogram can help households to smooth consumption and a community health worker can helppregnant mothers in dealing with heat related physiological stress. We use the phased introduc-tion of NREGA and a partial introduction of ASHA to measure the effectiveness of NREGAand ASHA as adaptation strategies. This paper shows NREGA and ASHA separately act as ef-fective adaptation measures in reducing the effects of higher temperatures during pregnancy oninfant mortality. The results indicate a significant impact on child health due to a predicted risein global temperatures and highlight the importance of public policy as adaptation strategies.

Keywords: Adaptation, Climate Change, Infant Mortality, Temperature

JEL Classification: Q50, Q54, Q56, Q58, I15, I18, O10

Preliminary Draft. Please do not cite or circulate without authors’ permission.

Email all correspondence to Rakesh Banerjee at [email protected] authors thank John Strauss, Jeff Nugent, Anant Nyshadham, Titus Galama, Yu-Wei Hsieh, Yilmaz Ko-cer, Jinkook Lee, Reed Walker, Tushar Bharati, Nazmul Ahsan and participants at the Western EconomicAssociation 2015 for helpful comments and suggestions. The authors also thank Panchajanya Banerjee andAniruddha Ghosh for helping us with the weather and lights data. All errors are ours.

1

Banerjee and Bhowmick (2016) Heat, Infant Mortality and Adaptation

1 Introduction

This paper explores if a large scale public workfare program and a community health care worker program

help in reducing the effects of exposure to high temperatures during pregnancy on infant mortality. We use

the District Level Household Survey (DLHS) from India to obtain birth and death records from mother’s

pregnancy history and combine it with daily level temperature data. Season of birth and temperature has

been shown to have effects on health and is related with fertility and parental characteristics (Lam and Miron,

1996; Banegas, Rodriguez-Artalejo, Graciani and De La Cruz, 2001; Buckles and Hungerman, 2013; Wilde,

Apouey, Jung et al., 2014). Since season of birth is related with several unobserved parental characteristics

and is also associated with temperature, estimating the effect of temperature during pregnancy on infant

mortality is challenging. In this paper, we disentangle the effect of season of birth from temperature, after

controlling for district-quarter of birth fixed effects in our estimation. Our results show, higher temperatures

in the second and third trimester and in the birth month increases infant mortality in rural India. In urban

India, we find no effect of high temperatures during pregnancy on the consequent chances of the newborn

dying as an infant. There could be two mechanisms behind this observed effect. First, higher temperatures

could affect agricultural productivity and food prices. In absence of consumption smoothing this can effect

pre-natal nutrition. This could lead to poor fetal health and cause increased chances of infant mortality.

Second, higher temperatures could have a direct physiological impact on mother’s health or an indirect

effect through increased disease prevalence. For instance, higher temperatures can increase blood pressure

or can cause dehydration in pregnant mothers. This can also contribute to poor fetal health and may elevate

the risk of the newborn dying as an infant. In this paper we explore two plausible adaptation strategies to

deal with these mechanisms , in the context of rural India.

In 2005 Government of India introduced the National Rural Employment Guarantee Act (NREGA), which

provides a guarantee of 100 days of employment to rural households at a fixed minimum wage. Higher than

usual temperatures in a year decreases agricultural yields and real wages (Burgess, Deschenes, Donaldson

and Greenstone, 2013) and rural households may not be able to completely smooth consumption which may

effect fetal health. This may be a plausible reason for increased infant mortality. In such circumstances,

access to NREGA employment may help households to smooth consumption. The policy was introduced in

a phased manner — it was first implemented in 200 districts, followed by 130 districts in the next year and

the remaining districts receiving it in the following year. We use this phased introduction, to estimate the

effect of NREGA in reducing the effects of higher temperatures during pregnancy. Results show access to

Page 2


NREGA during third trimester and month of birth reduces the effect of higher temperatures during pregnancy

on infant mortality.

We also explore if access to a community health care worker program reduces the effect of heat during

pregnancy on infant mortality. Community health care workers can help in dealing with any physiological

stress caused by heat or in treating mothers suffering from any heat related illness. For instance, they can

provide treatment to pregnant mothers suffering from dehydration caused by high temperatures. Government

of India introduced a community health care worker program, Accredited Social Health Activists (ASHA),

in 18 high focus states in India in 2005. Their main tasks include paying home visits to pregnant mothers

for ante-natal care, registering pregnant mothers, accompanying pregnant mothers to the health facility for

delivery, promoting general awareness about health and sanitation. We find access to ASHA during preg-

nancy reduces all the effects of heat on infant mortality, particularly during the second trimester and birth

month.

This paper primarily relates to the literature dealing with effects of climate change on human health

and possible adaptations. Global climate change is expected to increase both average temperatures and

make incidence of extreme temperatures more frequent (Zivin and Shrader, 2016). This is likely to effect

human health through a variety of channels - effecting disease patterns, water and food insecurity, migration

and population growth (Costello, Abbas, Allen, Ball, Bell, Bellamy, Friel, Groce, Johnson, Kett et al., 2009).

Developing countries, which depend more on weather dependent production process and have weaker health

infrastructure are more likely to be affected. Children and pregnant women are also particularly susceptible

to higher temperatures (Edwards, Saunders and Shiota, 2003). Recent literature have documented the effects

of extreme temperatures on mortality, birth weight and later life earnings (Deschenes and Greenstone, 2011;

Deschenes, Greenstone and Guryan, 2009; Isen, Rossin-Slater and Walker, 2015). However, far less has been

explored on possible adaptation strategies and even less from developing countries. Some recent studies have

documented that the spread of air-conditioning in the US had significant impacts in reducing the effects of

heat on mortality (Barreca, Clay, Deschenes, Greenstone and Shapiro, 2016). The use of air conditioning in

rural India during our sample period was less than 5 percent and even in 2012 it was less than 6 percent.1

More than 50 percent of rural households did not possess an electric fan during our sample period (Refer to

Figure 1 and Figure 2). Burgess et al. (2013) shows that access to credit in rural India reduces the effect of

high temperatures on mortality. However, more research is required in studying other plausible adaptation

measures (Deschenes, 2014).

1Our sample is births from 1998 to 2007

Page 3


This paper makes two key contributions. Firstly, it provides evidence that higher temperatures during

pregnancy have important health implications in rural India. More specifically, it increases chances of dying

as an infant. The result is important in measuring the costs of global climate change on human health.

Secondly, it explores two public policy measures which can help in reducing the effect of temperatures

during pregnancy. It shows a public workfare program and a community health care worker program can

serve as effective adaptation strategies in dealing with the health impacts of global climate change.

The paper is organized as follows, section 2 provides a literature review, section 3 provides the policy

description, section 4 discusses the data, section 5 discusses the empirical strategy, section 6 discusses the

results and section 7 concludes.

2 Literature Review

Some recent studies have focused on the effects of extreme temperatures in utero. These studies generally

find, extreme hot temperatures during pregnancy negatively affects several birth-outcomes like birth weight.

Most of these studies also carefully control for geographic fixed effects or geography-season specific fixed

effects to control for unobservables related to both temperature and birth outcomes. For instance, using data

on 37.1 million births from United States, Deschenes et al. (2009) finds that exposure to extreme temperature

during pregnancy leads to lower birth weight. They combine their estimates with future predictions of

climate change and estimates mean birth weight will reduce by 0.22 percent for whites and 0.36 percent

for blacks by end of the century. In another study, Andalon, Rodriguez-Castelan, Sanfelice, Azevedo and

Valderrama (2014) uses data from Colombian national registry and shows that heat waves during pregnancy

leads to reduced chances of being born as full term and being classified as healthy. They also carefully

control for municipality of birth fixed effects to account for geographic level unobservables and also include

month of birth fixed effects to control for seasonality. In another recent study, Molina and Saldarriaga (2016)

uses birth data from the Demographic Health Surveys (DHS) in Bolivia, Colombia and Peru and estimates

that exposure to a temperature which is one standard deviation from municipality’s long term mean during

pregnancy reduces birth weight by 20 grams. They also carefully control for municipality-month of birth

fixed effects to account for seasonality. The literature has also recently focused on long-term effects of

in utero exposure to extreme temperatures. Longer term effects on human capital accumulation can have

important consequences for economic growth. In a recent study, Isen et al. (2015) combines administrative

earnings records in United States with weather data and finds an extra day above 32C in utero causes a 0.2

Page 4


percent reduction in annual earnings. In another study in Ecuador, Carrillo, Fishman, Russ et al. (2015)

uses administrative earnings data and finds, in utero higher temperatures leads to lower earning for women.

These studies highlight the negative effects of the predicted rise in global temperatures on future economic

growth and well being.

Recent research have also documented the effects of contemporaneous exposure to extreme temper-

atures. Several studies have documented that extreme temperatures increase mortality in United States

(Deschenes and Moretti, 2009; Deschenes and Greenstone, 2011; Barreca, 2012). (Burgess et al., 2013) in-

vestigates the effect of high temperatures on mortality in India. They carefully construct vital statistics data

from 1957-2000 and show one standard deviation increase in high temperature days increases annual mor-

tality in rural India by 7.3 percent. They find no effect of high temperatures in urban areas. They also show

that real wages and agricultural yields in rural areas decrease with high temperature days, while they find no

such effect in urban areas. They point out that inability of households to smooth consumption in response

to income shocks in rural areas are a reason for differential effects of temperature on mortality in rural and

urban areas. Sudarshan, Somanathan, Somanathan, Tewari et al. (2015); Deryugina and Hsiang (2014) have

focused on the effect of high temperatures on labour productivity and manufacturing output. They use data

from manufacturing firms in India and show that output in labour intensive production process decrease at

high temperatures by 3 percent per degree Celsius. In another study, using income data from United States

Deryugina and Hsiang (2014) shows high temperature days have negative effect on individual productivity

and economic performance. They estimate that a weekday above 30C costs an average county 20 dollars per

person. They combine their estimates with climate change projections and find warmer daily temperatures

would lower annual growth by 0.06-0.16 percentage points in the United States. Zivin and Neidell (2014)

have focused on effects of high temperatures on time allocation and substitution between labour and leisure

in United States. Individuals might reallocate their time between labour and leisure to minimize the effect of

high temperatures and this may lead to significant welfare losses for households. They find at the high end of

temperature distribution an increase in temperature reduces hours worked in industries with high exposure

to climate and also non-employed individuals reduce time allocated to outdoor leisure. These studies point

out the different effects of a predicted rise in global temperatures with effects on mortality, productivity and

employment.

However very little evidence exists about the role of public policy in adaptation. In United States, the

spread of air-conditioning has been shown to reduce the effects of heat on mortality and income (Barreca

et al., 2016; Isen et al., 2015). These papers show that the spread of residential air-conditioning in the United

Page 5


States played an important role in reducing the effect of temperature on mortality and in reducing the the

effect of in utero exposure to high temperatures on income. However, the spread of air-conditioning in de-

veloping countries is very limited. Moreover, temperature fluctuations and weather shocks also have large

effect on income in developing countries, where large proportions of the work force are employed in the

agricultural sector. Moreover, in many developing access to credit, especially in times of aggregate weather

shocks is limited. Thus, individuals often cannot smooth consumption. In India Burgess et al. (2013) ex-

plores the effect of availability of credit on reducing the effect of temperature on mortality. The find that

expansion of bank branches in rural areas reduces the effect of high temperatures on mortality. Adhvaryu,

Kala and Nyshadham (2014) explores the effect of use of LED lights on productivity in garments facto-

ries in India. They find a negative relationship between temperature and production.They then show LED

lights which emits less heat and reduces the temperature on factory floors raises productivity, particularly in

hot days. These studies explores the effectiveness several adaptation methods which are likely to become

increasingly important with the increase in global temperatures.

We contribute to the growing literature on impacts of being exposed to high temperatures in utero.

We show high temperatures in utero is likely to increase to the chances of the newborn dying as an infant.

We also explore the role of public policy in reducing the effect of high temperatures. We show that a large

scale public workfare program which provides guaranteed employment and a community health care worker

program reduces the effect of high temperatures on infant mortality.

3 Adaptation Programs

3.1 National Rural Employment Gurantee Act

The National Rural Employment Guarantee Act was passed by the Government of India in 2005. The main

purpose of the act is to provide a legal guarantee of 100 days of employment to every rural household at a

fixed minimum wage. The act stipulates the wage should be same for both males and females and at least

one third of the employment is mandated to be for females. The household is required to apply for a job card

with the local village council and is typically expected to provide manual labor on projects like road con-

struction, earthwork on irrigation canal or other maintenance work. The local village council is responsible

for processing and issuing the job card. The government is mandated to provide employment within 15 days

of application, failing which it has to provide unemployment insurance. The federal government bears all

the cost of labor and 75 percent of the material costs and state government bears the rest. A public workfare

Page 6


program like NREGA has several advantages. Since those who are willing to work at the minimum wage are

only likely to apply, it will self select people into the program. Since NREGA increases labour demand , it

is also likely to increase private sector wages (Imbert and Papp, 2015). Particularly, in times of low demand

for labor, NREGA is likely to provide a guarantee of employment and hence smooth consumption (Klonner

and Oldiges, 2014). In our context, NREGA is likely to act as a coping mechanism if the effects of expo-

sure to extreme temperature on fetal health is through reduced income, arising from reduced employment

opportunities or reduced agricultural production during times of bad weather .

NREGA was introduced in a phased manner. The act was passed in 2005 and 200 districts got access to

the program in February 2006. In April 2007, 130 districts were added and the remaining districts got access

in 2008. The explicit goal of the phased roll out was to target the poorer districts first. It also guaranteed

each state would have atleast one district in the first phase of the program. We use this phased roll out to

identify the effects NREGA on reducing the impacts of extreme temperature on infant mortality.

3.2 Accridited Social Health Activists (ASHA)

. In 2005, the government of India introduced a community health worker program, called ASHA, with a

particular focus on 18 states in India.2 ASHA is one of the key strategies of National Rural Health Mission

(NHRM) of Government of India.3. One of the primary responsibilities of an ASHA worker is to pay

regular visits to pregnant mothers, registering pregnant mothers, accompanying them to the health facility.

They are also tasked with counselling women on needs of immunization, organize and mobilize people for

immunization camps, spread general awareness about health, sanitation and nutrition. The prescribed policy

of the Government of India requires ASHA workers to be educated, with atleast upto 8th standard, and

should belong to the age group of 25 to 40 years. They are also required to receive training in the sub center.

The ASHA worker is not a paid worker and receives small sums of money for each task performed.4.

Community health care workers and trained local midwifes have been successful in many parts of world

in providing maternal care. Government of India introduced a community health care worker program in

1977. These health care workers later came to be known as Village Health Guides. They were mostly

males whose main targets were males for family planning program. This program was not very successful

2The 18 high focus states are Uttar Pradesh, Bihar, Rajasthan, Madhya Pradesh, Orissa, Uttaranchal, Jharkhand, Chhat-tisgarh, Assam, Sikkim, Arunachal Pradesh, Manipur, Meghalaya, Tripura, Nagaland, Mizoram, Himachal Pradesh andJammu and Kashmir. After 2009 the program was also introduced in other parts of India.

3Though there were several other policies of NHRM apart from ASHA, Table 10 shows both high focus states andnon-high focus states had similar access to other programs except, ASHA

4For more details about the ASHA program refer to Rao (2014) and Bhowmick (2016)

Page 7


and was later discontinued (Bhowmick (2016)). In the recent years there have been a renewed interest in

community health care workers. They are particularly important in remote rural locations where it is hard to

have trained doctors and nurses. In India the health care set up in rural India is three-tiered, with health Sub

Centers (SCs) at the primary level, followed by Primary Health Centers (PHC), Community Health Centers

(CHC) and district hospitals. Health SCs are single room facilities run by Auxiliary Nurse Midwife (ANM)

who is responsible for a population of 4, 500 individuals. The ANM worker caters to too many individuals

to provide regular home visits. ASHA workers provide the necessary bridge between th health care system

and households (Rao (2014)).

ASHA workers have already been shown to be effective in improving maternal and child care. For

instance, Bhowmick (2016) finds presence of ASHA workers in villages leads to increased use of antenatal

care services, especially for mothers from the backward sections of the community. In a survey in Uttar

Pradesh, a state in North India, Jain, Srivastava, Khan, Dhar, Manon, Adhish and Nandan (2008) finds

that 70 percent of institutional deliveries conducted are motivated by ASHA. Rao (2014) uses the partial

introduction of ASHA and finds ASHA workers lead to an increased vaccination coverage. We also use

this partial introduction of ASHA to identify the role of ASHA in reducing the effects of extreme heat

on infant mortality. In our context, It is expected that the physiological stress created by extreme heat

during pregnancy can be reduced by better access to prenatal care, increased awareness and counseling by

community health care workers. Thus, ASHA workers can be effective in reducing the effects of extreme

heat during pregnancy on infant mortality.

4 Data

4.1 District Level Household Survey

This paper uses data from District Level Household Survey (DLHS)-II and District Level Household Survey

(DLHS)-III. DLHS-II was conducted by International Institute of Population Sciences (IIPS), Mumbai be-

tween 2002 and 2004 across 504 Districts in India. It surveyed 620, 107 households and 507, 622 married

women across India. Similarly, DLHS-III was conducted by IIPS, Mumbai between 2007-2008 across 601

districts in India and collected data from 720, 320 households and 643, 944 ever married woman. The main

focus of the survey was reproductive health of woman, while it also collected information on general house-

hold characteristics. We combine DLHS-II and DLHS-III and use pregnancy history of married woman to

create records of born children by using their month and year of birth and if died the month and year of

Page 8


death. We limit the recall period to 5 years only to limit any errors in recall.5. Unfortunately, we do not

have information on districts of birth so we proxy the location of birth of the child with the district in which

the mother is interviewed. Though we do not have data on migration in the survey, we believe district of

interview is a good proxy for district of birth as we are limiting our recall period to 5 years. In the paper,

we refer the district of birth of the child as the district where mother was interviewed. We exclude from our

analysis all union territories. 6 The survey also provides information on several mother level characteristics

which we use as control variables.

4.2 Temperature Data

This paper uses NCEP/NCAR Reanalysis 1 daily temperature data from National Oceanic and Atmospheric

Administration (NOAA). 7. The data is available as daily mean from 1948 till present over a 2.5 X 2.5

latitude longitude global grid. We match the latitude longitude of a district with the all the grids within a

distance of 250 kilometers. To calculate the daily temperature of a district, a weighted mean of temperature

of all the grids within 250 kilometers of a district is calculated, with the weights being inverse of the distance

from the district to the respective latitude-longitude grid. There is a large system of weather station data

available in India. However, a primary problem with using weather station data is the placement of a

weather station is not exogenous. Weather stations might be placed in less remote areas and the consequent

measurement error in extrapolating the data from these stations across space might create non-classical

measurement error in the explanatory variable, in this case temperature. Moreover, the data from the weather

stations are not consistently available over time, creating a problem of missing data. Re-analysis data uses

data from several sources and combines it with global climate models to produce consistent estimates over

time and space. One advantage of using re-analysis data is it is available uniformly over time and space.

The measurement error in that case will not be related with district level characteristics. These type of data

have been increasingly used in recent times in the economics literature. (Burgess et al. (2013))

5For DLHS-III the recall is restricted to 3-4 as the pregnancy history is only till 2004. For some districts, DLHS-II wasconducted in 2004, for them we have used a 7 year recall period to have a continuous sample from all districts from 1998

6As several new districts were created by the time DLHS-III was conducted, we map the new districts in to theircorresponding old districts.

7The following link provides more details http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.surface.html

Page 9

http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.surface.html

http://www.esrl.noaa.gov/psd/data/gridded/data.ncep.reanalysis.surface.html


4.3 Rainfall Data

This paper uses monthly mean rainfall data from Center for Climatic Research, University of Delaware.8

The data contains mean monthly rainfall at a 0.5X 0.5 latitude longitude grid, from 1900 to 2010. We

match the latitude longitude of a district with the all the grids within a distance of 250 kilometers. To

calculate the mean rainfall of a month in a district , a weighted mean of rainfall of all the grids within 250

kilometers of a district is calculated, with the weights being inverse of the distance from the district to the

respective latitude-longitude grid. We control for rainfall as this may be correlated with temperature and

infant mortality.

5 Empirical Strategy

5.1 Main Specification

One of the key challenge in estimating the effect of high temperatures during pregnancy on infant mortality

is spatial distribution of temperature is correlated with several unobservables, like quality of available health

care, which can have an effect on maternal and child health. Moreover, high temperatures are correlated

with seasons and season and month of birth is correlated is several unobservable parental and household

characteristics. Thus to estimate the effect of temperature, seasonal and spatial unobservables have to be

accounted for.

In this paper we use a fixed effect strategy to identify the effect of in utero exposure to high temperature

on infant mortality. It estimates the following equation.

Oidyq = α+ β0 (CDD > 90F in Birth Month) +

3∑j=1

βj (CDD > 90F in Trimesterj)+

χ′idyqγ + δdq + θqy + γm + εidq

where, Oidyq indicates the mortality status of child i born in quarter q, district d. and year y . It is a

dummy variable and it takes the value thousand if the child died as an infant and zero otherwise. (CDD >

90F in Birth Month) refers to number of Cumulative Degree Days exceeding 90 degrees Fahrenheit in

8The following link provides more details http://climate.geog.udel.edu/˜climate/html_pages/README.lw.html

Page 10

http://climate.geog.udel.edu/~climate/html_pages/README.lw.html

http://climate.geog.udel.edu/~climate/html_pages/README.lw.html


the month of birth of individual i in district d in the year y and quarter q .9 (CDD > 90F in Trimesterj)

refers to Cumulative Degree Days exceeding 90 degrees Fahrenheit in Trimester j of individual i in district

d in the year y and quarter q. χidyq includes covariates like parental years of schooling, gender of the child,

mean rainfall in the each of the trimester and month of birth, night time lights of the district-year of birth

and of the year prior to birth. δdq are the district by quarter of birth dummies and θqy are quarter of birth by

year of birth dummies. γm is the month of birth dummy, εidyq is the error term.

The above regression controls for all spatial and seasonal unobservables which are constant within a

district over time. The district by quarter dummies controls for all seasonal unobservables at the district and

quarter of birth level. For example, consider the district Delhi. If it happens more unhealthy mothers give

birth during the summer quarter in Delhi and the children of unhealthy mothers are more likely to die as an

infant, then if we do not control for district-quarter of birth fixed effects, then our estimates would be upward

biased. In particular, the district-quarter of birth fixed effects controls for all seasonal unobservables which

are correlated with temperature and also affect infant mortality for Delhi. Similarly, the quarter of birth by

year of birth dummies controls for all unobservables at the quarter-year level which effect infant mortality

and are also correlated with temperature. For example, if the summer quarter of 2005 was a particularly hot

summer and it was also hit by a swine flu epidemic across the country which affected pregnant mothers and

increased the chances of newborns dying as an infant, then if we do not control for quarter of birth-year of

birth dummies, it would misattribute the effect of swine flu epidemic on high temperatures. In the regression

we also control for rainfall, since rainfall is often correlated with temperature and it can independently effect

pregnant mothers either through effects on disease environment or indirectly through reduced agricultural

wages and production. We also control for night time lights in the district of birth for the year of birth

and the preceding year. This will help to control for broad economic changes and changes in infrastructure

like electrification. The identifying assumption is if we compare individuals born in a district-season over

different years, the event of extreme heat one faces in utero is exogenous. Thus β’s measure the effect of

having one extra Cumulative Degree Day above 90F in different stages of pregnancy on the chances of a

child dying as an infant.

9Cumulative Degree Days> 90 F is illustrated by the following example, if, say, in district Delhi in July 2005, there aretwo days where temperature exceeded 90F. On one day it was 95F and on the other it was 100F then Cumulative DegreeDays > 90 F = (100− 90) + (95− 90) = 15

Page 11


5.2 NREGA

We use the phased introduction of NREGA, in order to estimate the effect NREGA may have on reducing

the effects of high temperatures in utero on infant mortality. Since NREGA was partially introduced in some

districts in 2006, no districts in our sample were exposed to NREGA prior to 2006. NREGA was introduced

in 200 districts in 2006, and in 2007 it was introduced in 130 districts. Thus it creates four groups of mothers

within a district-quarter a) mothers who were exposed to high temperatures during pregnancy but did not

have access to NREGA b) mothers who were not exposed to high temperatures during pregnancy and also

did not have access to NREGA c) mothers who were exposed to high temperatures during pregnancy and

had access to NREGA d) mothers who were not exposed to high temperatures during pregnancy but had

access to NREGA. We use comparison across these groups to identify the effects of NREGA in reducing

the effects of high temperatures during pregnancy on infant mortality.

In particular, we estimate the following equation to identify, if the effects of extreme heat in utero are

reduced by the availability of public works program, namely NREGA.

Oidyq = α+ β0 (CDD > 90F in Birth Month) +

3∑j=1

βj (CDD > 90F in Trimesterj)+

σ0 (CDD > 90F in Birth Month) ∗ (Months till birth NREGA)idy+

3∑j=1

σj (CDD > 90F in Trimesterj ∗ (Months till T rimesterj NREGA)idy)+

π ∗ (Months till birth NREGA)idy +

3∑j=1

ηj ∗ (Months till trimesterj NREGA)idy+

χ′idyqγ + δdq + θqy + γm + εidq

where, as in the previous section Oidyq indicates the mortality status of child i born in quarter q,

district d. and year y . It is a dummy variable and it takes the value thousand if the child died as an

infant and zero otherwise. Similarly, (CDD > 90F in Birth Month) refers to number of Cumulative

Degree Days exceeding 90 degrees Fahrenheit in the month of birth of individual i in district d in the year

y and quarter q and (CDD > 90F in Trimesterj) refers to Cumulative Degree Days exceeding 90

degrees Fahrenheit in Trimester j of individual i in district d in the year y and quarter q. In this equation,

Months till birth NREGA captures number of months NREGA is present in the district of interview,

till the time of the birth. Months till trimesterj NRGEGA captures number of months the NREGA

Page 12


is present in the district of interview, till trimester j. For individuals born in 2005 and before this variables

takes the value 0. Like in the previous section, we also control for mean rainfall in birth month and each

trimester and also control for covariates like parental years of schooling, gender of the child, night time lights

of the district-year of birth and of the year prior to birth. We also control for district-quarter of birth fixed

effects and quarter-year of birth fixed effects along with month of birth. Like in the previous equation, β’s

measure the effect of having one extra Cumulative Degree Day above 90F in different stages of pregnancy

on the chances of a child dying as an infant and σ’s measure how much of that effect can be reduced by

having being exposed to one extra month in pregnancy to the NREGA program. π and η’s measures for any

direct effect that exposure to NREGA in utero may have on infant mortality.

5.3 ASHA

We use the partial introduction of ASHA to see if a community health worker program like ASHA can be

effective in reducing the effects of high temperatures during pregnancy on infant mortality. We use the fact

that ASHA was introduced from 2006 onwards in only 18 states in India. Since our sample is from 1998

to 2007, we again have four groups of mothers a) mothers who were exposed to high temperatures during

pregnancy but did not have access to ASHA workers b) mothers who were not exposed to high temperatures

during pregnancy and also did not have access to ASHA workers c) mothers who were exposed to high

temperatures during pregnancy and had access to ASHA workers d) mothers who were not exposed to high

temperatures during pregnancy but had access to ASHA workers. We use comparison across these groups

to identify the effects of ASHA in reducing the effects of high temperatures during pregnancy on infant

mortality.

Similarly, we estimate the following equation to identify, if the effects of extreme heat in utero are

reduced by the availability of a community health care worker program, namely ASHA.

Oidyq = α+ β0 (CDD > 90F in Birth Month)

+

3∑j=1

βj (CDD > 90F in Trimesterj)

+σ0 (CDD > 90F in Birth Month) ∗ASHAidy+

3∑j=1

σj (CDD > 90F in Trimesterj ∗ASHAidy)+

+πASHAidy + χ′idyqγ + δdq + θqy + γm + εidq

Page 13


where, ASHAidy= I(ASHAStates) * I(Y ear > 2005), where I(ASHAStates) indicates if the

mother is interviewed in a state where ASHA is implemented and I(Y ear > 2005) indicates if the child

is born in after 2005 ( when ASHA was implemented). As in the previous sections we control for mean

rainfall in the month of birth and each of the trimester, night time lights in the district-year of birth and in

the previous year, mother’s age and education, district-quarter of birth fixed effects, quarter-year of birth

fixed effects and month of birth fixed effects. Here β’s measure the effect of having one extra Cumulative

Degree Day above 90F in different stages of pregnancy on the chances of a child dying as an infant and

σ’s measure how much of that effect can be reduced by having being born in a state which had the ASHA

program in its year of birth. π measures for any direct effect that being born in a state which had the ASHA

program in its year of birth may have on infant mortality.

6 Results

6.1 Effect on Infant Mortality

Table 1 shows the results of the effect of high temperatures in utero on infant mortality. Panel A shows

the results for households interviewed in the rural areas and Panel B shows the results for households

interviewed in the urban areas. Column 1a is our preferred specification, which controls for district-quarter

of birth fixed effects and hence compares individuals born in the same district-quarter across different years.

This is our preferred specification because standard errors are the least amongst comparable specification.

The results show higher temperatures in utero increases the chances of a child dying as an infant, with the

second and third trimester as well as the birth month being particularly critical. To interpret the numbers,

consider the following example — if there are 10 days in the birth month when the average temperature was

95F and for the rest of the 20 days the average temperature was less than 90F, then Cumulative Degree Days

(CDD) exceeding 90F would be 50. This would increase infant mortality by 4.1 per 1000 births. A similar

situation in the second trimester and third trimester would increase infant mortality by 1.5 and 1.1 per 1000

births.

The results show that higher temperatures in the first trimester do not have any effect on infant mortality.

Higher temperature in utero can lead to both scarring and culling of the fetus. Though we cannot explicitly

test the hypothesis, it is plausible that higher temperatures early in pregnancy leads to culling and later in

the pregnancy causes scarring, explaining why we do not observe any significant effect in first trimester.

Column 1b and 1c present the results for alternate specifications. In column 1b, instead of a district-

Page 14


quarter fixed effect we add district-month fixed effects. The results are quite similar. In Column 1c, we

add state-quarter of birth specific quadratic trend — though we lose significance in some coefficients, the

direction and the magnitudes are comparable to column 1a.

Panel B presents the results for households interviewed in the urban areas. The results indicate higher

temperatures in utero do not have any effect on infant mortality. Though we cannot explicitly test, there are

several plausible reasons. Firstly, adaptation methods are perhaps better available in urban areas compared

to rural areas. For instance, electrification and air conditioning are more available in urban areas. Health

facilities to treat the illness caused by high temperatures are more readily available in urban areas. Second,

income and agricultural yields in rural areas are more susceptible to temperatures shocks than in urban areas

(Burgess et al., 2013).

6.2 Adaptation by NREGA

Table 2 shows that NREGA helps in reducing the effects of being exposed to high temperatures in utero

on infant mortality. Since NREGA is only implemented in rural areas, we can show the results only for

rural areas. As before, column 1a is our preferred specification which controls for district-quarter of birth

fixed effect and compares individuals born in same district-quarter over different years. The results show

that the effects of being exposed to high temperatures in utero -particularly in the birth month and third

trimester-on chances of dying as an infant is reduced by having access to NREGA program while the mother

is pregnant. To interpret the numbers, compare children born when there was no NREGA during their

mother’s pregnancy with children born in the same district-quarter when there was NREGA during their

mother’s pregnancy for all 10 months including the birth month. Assume, in both cases their mothers faced

10 days in third trimester when the average temperature was 95F and for the rest of the days it was less than

90F. For children who did not have NREGA at all during the pregnancy, infant mortality would increase

by 1.17 per 1000 born, while those who had NREGA for all 10 months there would be no effect of high

temperatures on infant mortality. 10

Column 1b and 1c show the results in alternative specification. In Column 1b instead of district-quarter

fixed effect we add a district-month fixed effect. The results compared to column 1a are very similar. In

Column 1c we add a state-quarter of birth specific quadratic trend. We lose significance in some results,

though the directions of the results are similar compared to column 1a.

10The NREGA coefficients in itself are insignificant, except for first trimester.

Page 15


6.3 Adaptation by ASHA

Table 3 shows results that a community health care worker program (ASHA) introduced in some select states

helps in adapting to the effects of being exposed to extreme heat in utero on infant mortality. The community

health care workers program was introduced only in rural areas, so we present the results only for rural areas.

Column 1a is our preferred specification, which compares individuals born in the same district-quarter over

different years. The results indicate ASHA workers helps in reducing the effects of high temperatures in

utero-particularly in second trimester and birth month- on infant mortality. For instance like in the previous

example if we compare mothers who faced 10 days in the second trimester when the average temperature was

95F and had ASHA during her pregnancy with a mother who did not have ASHA during her pregnancy but

faced the same temperature profile during her second trimester, in absence of ASHA infant mortality would

increase by 1.51 per 1000 births but in the presence of ASHA there will be no effect of high temperatures

in the second trimester. 11. Similarly the effects of high temperatures in the birth month and third trimester

are also reduced by the presence of ASHA worker.

Column 1b and 1c show the results in different specifications. In column 1b we control for district-

month fixed effects instead of district-quarter fixed effect. The results are both similar in magnitude and

direction when compared to column 1a. In column 1c we add state-quarter of birth specific linear and

quadratic trend. We lose significance in some results, but the direction of the results are similar to column

1a.

6.4 Robustness Check

Table 4 and Table 5 show the robustness of the of the effects of temperature in utero with different levels

of temperature values. Table 4 shows the results with CDD 95F and CDD 92F and Table 5 shows the

results with CDD 87F and CDD 85F. If we compare the results from Table 1 with Table 4 and Table 5, the

coefficients are in same direction and have similar magnitude. For instance, if there are 10 days in the birth

month, when the average temperature was 95F and for the rest of the month it was below 85F, coefficients

from table 1 would imply infant mortality will increase by 4.1 per 1000 live births in rural areas, and panel

A (bottom panel) in table 4 would imply infant mortality would increase by 3.8 per 1000 births, panel A (top

panel) in table 5 would imply infant mortality would increase by 4.2 per 1000 live births, panel A (bottom

panel) in table 5 would imply infant mortality would increase by 4.1 per 1000 live births.

11The main ASHA coefficient is insignificant.

Page 16


In Table 6, we show the robustness of the results of Table 1 by including the temperatures of 13th

month to 24th month after birth in the regression. The results show a) the coefficients of temperatures of

13th month to 24th month is insignificant. In principle, temperatures from 13th to 24th month after birth

should have no effect on infant mortality. If the specifications of Table 1 would not have controlled for

seasonality of births, then perhaps the coefficients of temperature of 13th to 24th month after birth could

have been significant. The insignificant results shows the specifications of Table 1 adequately control for

seasonality of births. b) The results of temperature in utero is unaffected by the including the temperatures

of 13th to 24th month. This shows after controlling for relevant fixed effects, temperatures faced in different

points in time are unrelated. This indicates the temperatures are more of shocks to the mothers.

In Table 7 we show temperatures faced during pregnancy is unrelated with mother’s years of schooling

and mother’s age at the time of interview. Though we control explicitly for mother’s years of schooling

and mother’s age in all the regressions and also include district-quarter specific fixed effects to control for

any seasonality associated with parental characteristics, the results in the Table 7 assure that there are no

associations of mother’s characteristics and temperature during pregnancy. Results in column 1a indicates

neither mother’s age at the time of interview nor mother’s years of schooling is associated with temperatures

at the time of pregnancy. Column 2a shows the same results for urban areas.

In Table 8 we test if number of live births at a district-month-year level is associated with temperatures

10 months before birth, i.e at the time of conception. If temperatures at conception alters fecundity of

mothers, then results of Table 1 could be biased estimates. In column 1a and 2a we do find some evidence

of number of live births being associated with temperatures at conception, but when we control for district-

month of birth fixed effects in columns 1b and 2b, there is no evidence of number of live births being

associated with temperature at conception. The results remain the same when we add state-quarter specific

quadratic trend.

7 Conclusion

Average global temperatures are predicted to rise. Extreme weather events are also likely to become more

frequent. This changed weather pattern is predicted to affect human health on several dimension and also

through several channels. Among other things, it is predicted to change disease patterns, affect agricultural

production and reduce water availability. Populations in rural parts of developing countries who rely on

weather dependent production process for food and livelihood are likely to be affected the most. These

Page 17


areas also lack reliable health infrastructure to deal with heat and other weather related illness, making them

perhaps most vulnerable to weather changes. Aggressive mitigation efforts in terms of reducing greenhouse

gas emissions is required prevent further global warming. However, mitigation efforts are costly to the cur-

rent population as it would involve investments in cleaner technology. To determine the optimal amount of

investment in mitigation efforts it is important to measure the different costs from predicted global warming

. Consequently, it is important to measure the costs on human health of global climate change, especially

for populations living in rural parts of developing countries. In this context, this paper asks what is the effect

of high temperatures during pregnancy on infant mortality in India. We use data from pregnancy history of

mothers from a household survey and combine it with daily level temperature data. It explicitly disentangles

the effect of season of birth from temperature during pregnancy and finds higher temperatures during preg-

nancy increases chances of the newborn dying as an infant in rural India. There is no such effect in urban

India. The results of this paper points out the expected cost of predicted rise in global temperatures.

However, achieving a global agreement of an aggressive mitigation strategy is difficult. Hence, some

effects of global climate change are considered unavoidable. Thus adaptation strategies, which are actions

to reduce the adverse effects of climate change are crucial. Moreover in case of failure of agreement across

countries, individual nations are to rely on adaptation strategies only. In developing countries the role of

public policies are critical. So at the same time, it is also important to explore possible policy measures

which can serve as adaptation strategies to deal with the impacts of global climate change on human health.

In this context, this paper explores if a public workfare program which provides a guaranteed employment

to households in rural areas and a community health care worker program serve as plausible adaptation

strategies. A public workfare program which provides employment guarantee can help households to smooth

consumption in times weather induced shocks on agricultural production. A community health worker

can treat pregnant mothers from weather induced diseases. This paper finds the public workfare program

(NREGA), which provides employment guarantee for rural households- and a community health care worker

program (ASHA)- helps in reducing the effects of higher temperatures during pregnancy on infant mortality.

The results of this paper points out the importance of public policy as adaptation strategies.

Page 18


References

Adhvaryu, Achyuta, Namrata Kala, and Anant Nyshadham (2014) “The light and the heat: Productivity

co-benefits of energy-saving technology.”

Andalon, Mabel, Carlos Rodriguez-Castelan, Viviane Sanfelice, Joao Pedro Azevedo, and Daniel Valder-

rama (2014) “Weather shocks and health at birth in Colombia,” World Bank Policy Research Working

Paper.

Banegas, JR, F Rodriguez-Artalejo, A Graciani, and JJ De La Cruz (2001) “Month of birth and height of

Spanish middle-aged men,” Annals of human biology, Vol. 28, pp. 15–20.

Barreca, Alan, Karen Clay, Olivier Deschenes, Michael Greenstone, and Joseph S Shapiro (2016) “Adapting

to Climate Change: The Remarkable Decline in the US Temperature-Mortality Relationship over the

Twentieth Century,” Journal of Political Economy, Vol. 124, pp. 105–159.

Barreca, Alan I (2012) “Climate change, humidity, and mortality in the United States,” Journal of Environ-

mental Economics and Management, Vol. 63, pp. 19–34.

Bhowmick, Riddhi (2016) “Evaluating Impact of Community Health Workers on Maternal and Child Health

Behavior in India.”

Buckles, Kasey S and Daniel M Hungerman (2013) “Season of birth and later outcomes: Old questions,

new answers,” Review of Economics and Statistics, Vol. 95, pp. 711–724.

Burgess, Robin, Olivier Deschenes, Dave Donaldson, and Michael Greenstone (2013) “The unequal effects

of weather and climate change: Evidence from mortality in india,” Unpublished working paper.

Carrillo, Paul E, Ram Fishman, Jason Russ et al. (2015) “Long-Term Impacts of High Temperatures on

Economic Productivity,”Technical report.

Costello, Anthony, Mustafa Abbas, Adriana Allen, Sarah Ball, Sarah Bell, Richard Bellamy, Sharon Friel,

Nora Groce, Anne Johnson, Maria Kett et al. (2009) “Managing the health effects of climate change,”

The Lancet, Vol. 373, pp. 1693–1733.

Deryugina, Tatyana and Solomon M Hsiang (2014) “Does the environment still matter? Daily temperature

and income in the United States,”Technical report, National Bureau of Economic Research.

Page 19


Deschenes, Olivier (2014) “Temperature, human health, and adaptation: A review of the empirical litera-

ture,” Energy Economics, Vol. 46, pp. 606–619.

Deschenes, Olivier and Michael Greenstone (2011) “Climate Change, Mortality, and Adaptation: Evidence

from Annual Fluctuations in Weather in the US,” American Economic Journal: Applied Economics, Vol.

3, pp. 152–85.

Deschenes, Olivier, Michael Greenstone, and Jonathan Guryan (2009) “Climate change and birth weight,”

The American Economic Review, Vol. 99, pp. 211–217.

Deschenes, Olivier and Enrico Moretti (2009) “Extreme weather events, mortality, and migration,” The

Review of Economics and Statistics, Vol. 91, pp. 659–681.

Edwards, MJ, RD Saunders, and K Shiota (2003) “Effects of heat on embryos and foetuses,” International

journal of hyperthermia, Vol. 19, pp. 295–324.

Imbert, Clement and John Papp (2015) “Labor market effects of social programs: Evidence from india’s

employment guarantee,” American Economic Journal: Applied Economics, Vol. 7, pp. 233–263.

Isen, Adam, Maya Rossin-Slater, and R Walker (2015) “Heat and long-run human capital formation,” Un-

published Manuscript.

Jain, Neera, NK Srivastava, AM Khan, Neera Dhar, S Manon, V Adhish, and Deoki Nandan (2008) “As-

sessment of functioning of ASHA under NRHM in Uttar Pradesh,” Health and Population: Perspectives

and Issues, Vol. 31, pp. 132–140.

Klonner, Stefan and Christian Oldiges (2014) “Safety Net for India’s Poor or Waste of Public Funds? Poverty

and Welfare in the Wake of the World’s Largest Job Guarantee Program.”

Lam, David A and Jeffrey A Miron (1996) “The effects of temperature on human fertility,” Demography,

Vol. 33, pp. 291–305.

Molina, Oswaldo and Victor Saldarriaga (2016) “The Perils of Climate Change: In Utero Exposure to

Temperature Variability and Birth Outcomes in the Andean Region.”

Rao, Tanvi (2014) “The Impact of a Community Health Worker Program on Childhood Immunization:

Evidence from India’s’ ASHA’Workers,” Available at SSRN 2444391.

Page 20


Sudarshan, Anant, E Somanathan, Rohini Somanathan, Meenu Tewari et al. (2015) “The Impact Of Tem-

perature On Productivity And Labor Supply-Evidence From Indian Manufacturing,”Technical report.

Wilde, Joshua, Benedicte Apouey, Toni Jung et al. (2014) “Heat waves at conception and later life out-

comes,” Unpublished Manuscript.

Zivin, Joshua Graff and Matthew Neidell (2014) “Temperature and the allocation of time: Implications for

climate change,” Journal of Labor Economics, Vol. 32, pp. 1–26.

Zivin, Joshua Graff and Jeffrey Shrader (2016) “Temperature Extremes, Health, and Human Capital,” The

Future of Children, pp. 31–50.

Page 21


8 Tables and Figures

Figure 1: Calculated from the National Sample Survey’s (NSS) consumption survey, using the 55,61, 66 and 68 round.

Page 22


Figure 2: Calculated from the National Sample Survey’s (NSS) consumption survey, using the55,61,66 and 68 round.

Page 23


Table 1: Effect of Temperature in utero on Infant Mortality

Panel A: Rural Panel B: Urban(1a) (1b) (1c) (2a) (2b) (2c)

CDD exceeding 90F Birth Month 0.0843*** 0.0650* 0.0380 -0.00181 0.0409 0.0323(0.0141) (0.0327) (0.0322) (0.0358) (0.0646) (0.0744)

CDD exceeding 90F Trimester 1 -4.45e-05 -0.0127 -0.00803 0.00743 0.00117 0.000959(0.0147) (0.0353) (0.0318) (0.0216) (0.0419) (0.0443)

CDD exceeding 90F Trimester 2 0.0318** 0.0615*** 0.0416** -0.0176 -0.0318 -0.0220(0.0126) (0.0194) (0.0172) (0.0163) (0.0351) (0.0348)

CDD exceeding 90F Trimester 3 0.0232* 0.0687 0.0455 -0.00750 0.0396 0.0491*(0.0133) (0.0452) (0.0408) (0.0159) (0.0276) (0.0282)

Mean of Dependent Variable 54 54 54 43 43 43N 395,070 395,070 395,070 116,198 116,198 116,198p value for Joint Test 0.0312 0.0246 0.0246 0.6584 0.392 0.3501

Controls Y Y Y Y Y YDistrict-Quarter FE Y YQuarter-Year FE Y Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise. The controls in the regressions include,

mean rainfall in the district of interview in the month of birth, and in each of the three trimesters, DMSP mean night time lights in the year

of birth and the preceding year in the district of interview, DLHS survey round fixed effect, mothers age in four categories, mothers years of

schooling in categories and a dummy for religion of the household head. The sample is births from 1998 to 2007. Cumulative Degree Days

(CDD) > 90 F is illustrated by the following example, if, say, in district Delhi in July 2005, there are two days where temperature exceeded

90F. On one day it was 95F and on the other it was 100F then Cumulative Degree Days > 90 F = (100− 90) + (95− 90) = 15. All standard

errors are clustered at the state level.

Page 24


Table 2: Effect of Temperature in utero on Infant Mortality and adaptation by NREGA

Panel A: Rural(1a) (1b) (1c)

Months NREGA Till Birth Month*CDD exceeding 90F Birth Month -0.0155 -0.0193* -0.00637(0.00926) (0.00962) (0.00811)

Months NREGA Till Trimester 1*CDD exceeding 90F Trimester 1 -0.00964* -0.00994 -0.000552(0.00508) (0.00613) (0.00492)

Months NREGA Till Trimester 2*CDD exceeding 90F Trimester 2 0.000157 0.000445 0.00339(0.00318) (0.00294) (0.00442)

Months NREGA Till Trimester 3*CDD exceeding 90F Trimester 3 -0.00758** -0.00775** -0.00437(0.00321) (0.00366) (0.00543)

CDD exceeding 90F Birth Month 0.0872*** 0.0703** 0.0396(0.0144) (0.0316) (0.0317)

CDD exceeding 90F Trimester 1 0.00128 -0.0125 -0.00950(0.0141) (0.0359) (0.0328)

CDD exceeding 90F Trimester 2 0.0295** 0.0560*** 0.0370**(0.0126) (0.0189) (0.0176)

CDD exceeding 90F Trimester 3 0.0234* 0.0693 0.0457(0.0136) (0.0447) (0.0397)

N 394,746 394,746 394,746p value for Joint Test 0.0821 0.0754 0.3417

Controls Y Y YDistrict-Quarter FE YQuarter-Year FE Y Y YBirth Month FE YDistrict-Month FE Y YState-Quarter specific Quadratic Trend YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise. The controls in the regressions

include, mean rainfall in the district of interview in the month of birth, and in each of the three trimesters, DMSP mean night time

lights in the year of birth and the preceding year in the district of interview, DLHS survey round fixed effect, mothers age in four

categories, mothers years of schooling in categories and a dummy for religion of the household head. The sample is births from 1998

to 2007. Cumulative Degree Days > 90 F is illustrated by the following example, if, say, in district Delhi in July 2005, there are two

days where temperature exceeded 90F. On one day it was 95F and on the other it was 100F then Cumulative Degree Days > 90 F

= (100−90)+(95−90) = 15. Months NREGA Till Trimester i is number of months NREGA was present in the district of interview

till trimester i. All standard errors are clustered at the state level.Page 25


Table 3: Effect of Temperature in utero on Infant Mortality and adaptation by ASHA

Panel A: Rural(1a) (1b) (1c)

ASHA*CDD exceeding 90F Birth Month -0.0731* -0.0643 0.00288(0.0385) (0.0449) (0.0268)

ASHA*CDD exceeding 90F Trimester 1 -0.0694*** -0.0693** -0.0405**(0.0241) (0.0268) (0.0149)

ASHA*CDD exceeding 90F Trimester 2 -0.0455* -0.0429* -0.0239***(0.0230) (0.0228) (0.00575)

ASHA*CDD exceeding 90F Trimester 3 -0.0297 -0.0200 0.0371*(0.0270) (0.0223) (0.0194)

CDD exceeding 90F Birth Month 0.0959*** 0.0714** 0.0369(0.0177) (0.0336) (0.0338)

CDD exceeding 90F Trimester 1 0.00845 -0.00964 -0.00570(0.0157) (0.0354) (0.0313)

CDD exceeding 90F Trimester 2 0.0302** 0.0494** 0.0458**(0.0127) (0.0193) (0.0167)

CDD exceeding 90F Trimester 3 0.0195 0.0561 0.0503(0.0137) (0.0460) (0.0411)

N 395,070 395,070 395,070p value for Joint Test 0.0322 0.015 0.0009

Controls Y Y YDistrict-Quarter FE YQuarter-Year FE Y Y YBirth Month FE YDistrict-Month FE Y YState-Quarter specific Quadratic Trend YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise.

The controls in the regressions include, mean rainfall in the district of interview in the month of

birth, and in each of the three trimesters, DMSP mean night time lights in the year of birth and the

preceding year in the district of interview, DLHS survey round fixed effect, mothers age in four cat-

egories, mothers years of schooling in categories and a dummy for religion of the household head.

The sample is births from 1998 to 2007. Cumulative Degree Days (CDD) > 90 F is illustrated by

the following example, if, say, in district Delhi in July 2005, there are two days where temperature

exceeded 90F. On one day it was 95F and on the other it was 100F then Cumulative Degree Days

> 90 F = (100− 90)+ (95− 90) = 15. ASHA takes the value one if a) the mother is interviewed

in a state where ASHA is implemented and b) the child is born after 2005. All standard errors are

clustered at the state level.

Page 26


Table 4: Effect of Temperature in utero on Infant Mortality (Robustness Check)


CDD exceeding 95F Birth Month 0.257*** 0.232*** 0.173*** 0.0293 0.322*** 0.296**(0.0334) (0.0476) (0.0386) (0.0842) (0.0935) (0.106)

CDD exceeding 95F Trimester 1 0.0460 0.0868 0.0696 0.0772 0.0417 0.0396(0.0511) (0.109) (0.116) (0.0556) (0.0758) (0.0844)

CDD exceeding 95F Trimester 2 0.106*** 0.104** 0.0543 -0.0234 -0.0337 -0.0200(0.0320) (0.0459) (0.0365) (0.0590) (0.0789) (0.0557)

CDD exceeding 95F Trimester 3 0.0454 0.0812 0.0353 -0.0503 0.0347 0.0313(0.0312) (0.0888) (0.0821) (0.0784) (0.0917) (0.106)

N 395,070 395,070 395,070 116,198 116,198 116,198p value for Joint Test 0.0166 0.0601 0.3651 0.4293 0.8086 0.9007


CDD exceeding 92F Birth Month 0.126*** 0.108** 0.0803** 0.00911 0.130* 0.119(0.0166) (0.0401) (0.0372) (0.0408) (0.0760) (0.0903)

CDD exceeding 92F Trimester 1 0.00449 0.000427 0.00628 0.0250 0.0159 0.0191(0.0230) (0.0550) (0.0514) (0.0295) (0.0477) (0.0521)

CDD exceeding 92F Trimester 2 0.0520*** 0.0811*** 0.0562** -0.0210 -0.0329 -0.0235(0.0179) (0.0273) (0.0218) (0.0218) (0.0428) (0.0379)

CDD exceeding 92F Trimester 3 0.0330 0.0847 0.0593 -0.0171 0.0342 0.0435(0.0198) (0.0576) (0.0544) (0.0260) (0.0404) (0.0443)


Controls Y Y Y Y Y YDistrict-Quarter FE Y YQuarter-Year FE Y Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise. The controls in the regressions include,







Page 27




CDD exceeding 87F Birth Month 0.0523*** 0.0233 0.00941 -0.000517 -0.0119 -0.0171(0.0143) (0.0268) (0.0290) (0.0138) (0.0276) (0.0590)

CDD exceeding 87F Trimester 1 -0.00243 -0.0275 -0.0229 -0.00760 -0.00455 -0.0120(0.00899) (0.0211) (0.0164) (0.0125) (0.0312) (0.0306)

CDD exceeding 87F Trimester 2 0.0154* 0.0471*** 0.0288** -0.000598 0.0354 0.00111(0.00815) (0.0127) (0.0123) (0.0111) (0.0227) (0.0313)

CDD exceeding 87F Trimester 3 0.0129 0.0419 0.0225 -0.00213 -0.00354 0.0454*(0.00822) (0.0301) (0.0275) (0.0293) (0.0508) (0.0225)



CDD exceeding 85F Birth Month 0.0405*** 0.0142 0.00491 -0.00154 -0.0215 -0.0356(0.0133) (0.0226) (0.0244) (0.0260) (0.0456) (0.0521)

CDD exceeding 85F Trimester 1 -0.00279 -0.0318* -0.0274** -0.00328 -0.0211 -0.0218(0.00732) (0.0161) (0.0114) (0.0106) (0.0219) (0.0242)

CDD exceeding 85F Trimester 2 0.00935 0.0389*** 0.0250** -0.00428 0.00951 0.0154(0.00650) (0.00978) (0.00959) (0.00992) (0.0251) (0.0256)

CDD exceeding 85F Trimester 3 0.00849 0.0266 0.0115 0.000780 0.0329* 0.0435**(0.00630) (0.0246) (0.0223) (0.00902) (0.0190) (0.0187)


Controls Y Y Y Y Y YDistrict-Quarter FE Y Y YQuarter-Year FE Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise. The controls in the regressions include,







Page 28




CDD exceeding 90F- Month13 to Month 24 0.00508 0.00880 -0.00653 0.0299 0.0299 0.0111(0.0117) (0.0108) (0.00902) (0.0350) (0.0337) (0.0312)

CDD exceeding 90F Birth Month 0.0842*** 0.0591* 0.0381 -0.00191 0.0435 0.0318(0.0140) (0.0324) (0.0324) (0.0360) (0.0632) (0.0748)

CDD exceeding 90F Trimester 1 1.41e-05 -0.0144 -0.00796 0.00815 -0.000194 0.000998(0.0147) (0.0357) (0.0319) (0.0209) (0.0390) (0.0441)

CDD exceeding 90F Trimester 2 0.0320** 0.0651*** 0.0413** -0.0161 -0.0270 -0.0217(0.0126) (0.0198) (0.0173) (0.0164) (0.0377) (0.0353)

CDD exceeding 90F Trimester 3 0.0233* 0.0701 0.0452 -0.00636 0.0400 0.0494*(0.0133) (0.0438) (0.0402) (0.0161) (0.0270) (0.0279)


Controls Y Y Y Y Y YDistrict-Quarter FE Y YQuarter-Year FE Y Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise. The controls in the regressions include, mean

rainfall in the district of interview in the month of birth, and in each of the three trimesters, DMSP mean night time lights in the year of birth and the

preceding year in the district of interview, DLHS survey round fixed effect, mothers age in four categories, mothers years of schooling in categories

and a dummy for religion of the household head. The sample is births from 1998 to 2007. Cumulative Degree Days (CDD) > 90 F is illustrated by the

following example, if, say, in district Delhi in July 2005, there are two days where temperature exceeded 90F. On one day it was 95F and on the other

it was 100F then Cumulative Degree Days > 90 F = (100− 90) + (95− 90) = 15. All standard errors are clustered at the state level.

Page 29


Table 7: Prenatal Temperature and Mother’s Charachteristics


Mother Years of Schooling

CDD exceeding 90F Birth Month 3.24e-05 -0.000433 4.10e-05 -0.000357 -0.000591 -0.000482(0.000838) (0.00123) (0.00113) (0.00110) (0.00112) (0.000874)

CDD exceeding 90F Trimester 1 7.61e-05 0.000301 0.000393 0.000574 0.000786 0.000502(0.000270) (0.000373) (0.000379) (0.000373) (0.000925) (0.00100)

CDD exceeding 90F Trimester 2 -0.000323 0.000146 6.60e-06 0.000813* 0.00272*** 0.00234**(0.000334) (0.000779) (0.000709) (0.000418) (0.000635) (0.000848)

CDD exceeding 90F Trimester 3 0.000195 -0.000510 -0.000403 -9.41e-05 -0.00148 -0.00182(0.000329) (0.000469) (0.000418) (0.000506) (0.00136) (0.00141)


Mother Age (in Years)

CDD exceeding 90F Birth Month 0.000635 0.00200* 0.00167 0.000899 5.56e-05 -0.00144(0.000829) (0.00101) (0.00113) (0.000768) (0.00160) (0.00164)

CDD exceeding 90F Trimester 1 0.000815** -0.000288 -7.26e-05 1.45e-05 -0.000199 0.000183(0.000341) (0.000462) (0.000429) (0.000636) (0.000740) (0.000769)

CDD exceeding 90F Trimester 2 0.000498 0.000744 0.000469 4.75e-05 5.45e-05 -0.000142(0.000418) (0.000476) (0.000502) (0.000900) (0.000840) (0.000794)

CDD exceeding 90F Trimester 3 0.000130 0.000759 0.000377 0.000931 0.00101 9.60e-05(0.000301) (0.000561) (0.000699) (0.00103) (0.00133) (0.00131)


Controls Y Y Y Y Y YDistrict-Quarter FE Y YQuarter-Year FE Y Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe sample for mother years of schooling includes all mothers who can read and write if interviewed in DLHS 2 and all mothers who have ever attended

school if interviewed in DLHS 3. Mother age is mother’s age at the time of interview. The controls in the regressions include, mean rainfall in the

district of interview in the month of birth, and in each of the three trimesters, DMSP mean night time lights in the year of birth and the preceding year

in the district of interview, DLHS survey round fixed effect and a dummy. The sample is births from 1998 to 2007. Cumulative Degree Days (CDD) >

90 F is illustrated by the following example, if, say, in district Delhi in July 2005, there are two days where temperature exceeded 90F. On one day it

was 95F and on the other it was 100F then Cumulative Degree Days > 90 F = (100− 90) + (95− 90) = 15. All standard errors are clustered at the

state level.Page 30


Table 8: Temperature and Number of Births


CDD exceeding 90F Birth Month 6.90e-05 -0.000739 -0.000770 0.000852 -0.00159 -0.00251(0.00261) (0.00215) (0.00169) (0.00335) (0.00629) (0.00556)

CDD exceeding 90F at Conception -0.00690** -4.04e-05 0.00196 -0.00676** -0.00338 -0.00526(0.00321) (0.00395) (0.00456) (0.00289) (0.00855) (0.00948)

CDD exceeding 90F Trimester 1 -0.00106 -0.000921 0.00170 -0.00231 -0.00150 -0.00166(0.00238) (0.00204) (0.00237) (0.00231) (0.00344) (0.00341)

CDD exceeding 90F Trimester 2 0.00307* 0.000624 0.00180 0.00378 0.000354 0.000258(0.00150) (0.00359) (0.00331) (0.00315) (0.00516) (0.00645)

CDD exceeding 90F Trimester 3 0.00657** 0.000707 0.00152 0.00596** -0.000382 -0.000893(0.00239) (0.00252) (0.00311) (0.00228) (0.00409) (0.00413)


District-Quarter FE Y YQuarter-Year FE Y Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe dependent variable is number of births at the district-month-year level. At conception means 10 months before the month of birth, trimester 1 is 9,

8, 7 months before birth, trimester 2 is 6, 5, 4 months before birth, trimester 3 is 3, 2, 1 months before birth. All standard errors are clustered at district

level. The regressions are weighted by sum of sample weights in a district. The regression specification also controls for DLHS wave fixed effects,

rainfall in all birth trimesters and month of conception and birth as well as DMSP mean night time lights in the year of birth and the preceding year

in the district of interview. The sample is births from 1998 to 2008. Cumulative Degree Days (CDD) > 90 F is illustrated by the following example,

if, say, in district Delhi in July 2005, there are two days where temperature exceeded 90F. On one day it was 95F and on the other it was 100F then

Cumulative Degree Days > 90 F = (100− 90) + (95− 90) = 15. All standard errors are clustered at the state level.

Page 31


9 Appendix A

Table 9 presents results of the effects of textitin utero exposure to high temperatures on neonatal mortality.

The results show the effects of in utero exposure to high temperatures on neonatal mortality are similar to

the effects of high temperatures on infant mortality as presented in Table 1. For rural areas, second and third

trimester and the birth month are particularly important. High temperatures increases chances of a new born

dying within one month. The magnitudes of the effects are also similar to that of Table 1. For instance,

using the same example, if mothers faces 10 days in the birth month when the temperature was 95F and for

the rest of the month it below 90F, then neonatal mortality would increase by 2.7 per 1000 live births, which

is a 7.5 percent increase over mean. A similar situation would increase infant mortality by 4.1 per 1000 live

births, which is also a 7.6 percent increase over mean.

Page 32


Table 9: Effect of Temperature in utero on Neonatal Mortality


CDD exceeding 90F Birth Month 0.0548** 0.0717*** 0.0545*** 0.0229 0.111** 0.112**(0.0206) (0.0208) (0.0172) (0.0295) (0.0505) (0.0522)

CDD exceeding 90F Trimester 1 -0.00718 -0.0140 -0.00858 0.0303* 0.0412 0.0482(0.00783) (0.0234) (0.0235) (0.0165) (0.0374) (0.0390)

CDD exceeding 90F Trimester 2 0.0172* 0.0196 0.0157 0.0196 0.0130 0.0254(0.00934) (0.0159) (0.0142) (0.0156) (0.0303) (0.0319)

CDD exceeding 90F Trimester 3 0.0203* 0.0552*** 0.0388** 0.0231 0.0114 0.0208(0.0114) (0.0173) (0.0153) (0.0143) (0.0301) (0.0316)

Mean of Dependent Variable 36 36 36 29 29 29N 509,096 509,096 509,096 146,482 146,482 146,482p value for Joint Test 0.0193 0.0144 0.0961 0.0143 0.4746 0.2345

Controls Y Y Y Y Y YDistrict-Quarter FE Y YQuarter-Year FE Y Y Y Y Y YBirth Month FE Y YDistrict-Month FE Y Y Y YState-Quarter specific Quadratic Trend Y YThe dependent variable in the regression is 1000 if the born child died as an infant and 0 otherwise. The controls in the regressions include, mean

rainfall in the district of interview in the month of birth, and in each of the three trimesters, DMSP mean night time lights in the year of birth

and the preceding year in the district of interview, DLHS survey round fixed effect, mothers age in four categories, mothers years of schooling

in categories and a dummy for religion of the household head. The sample is births from 1998 to 2007. Cumulative Degree Days(CDD) > 90

F is illustrated by the following example, if, say, in district Delhi in July 2005, there are two days where temperature exceeded 90F. On one

day it was 95F and on the other it was 100F then Cumulative Degree Days > 90 F = (100 − 90) + (95 − 90) = 15. All standard errors are

clustered at the state level.

Page 33


Table 10: Different NRHM provisions in High and Non-Focus states (Reproduced from Rao (2014))

High-Focus Non-Focus

% of villages with an ASHA worker 74.02 33.04Average number of ASHA workers who have completed first round of training in Sub-Center area 4.6 1.66% of Sub-Centers that have received untied funds 76.26 89.14% of Sub-Centers that have fully utilized untied funds 28.92 50.13% of Sub-Centre areas that have a Village Health & Sanitation Committee 63.96 84.64% of Primary Health Centers that have received untied funds 65.34 87.2% of Primary Health Centers that have fully utilized untied funds 24.26 48.02% of Community Health Centers that have received untied funds 81.82 90.09% of Community Health Centers that have fully utilized untied funds 32.62 46.06% of District Hospitals with a Rogi Kalyan Samiti (RKS) 86.61 93.43

Page 34


10 Appendix B

In this section we show the effect of temperature on yield of several important crops. We use data on

agricultural yields , production and area cropped from Directorate of Economics and Statistics, Ministry of

Agriculture. It contains information on these variables for each district, crop, year and season. We estimate

the following regression for major kharif crops only.

Ydy = α+ βCDD > 90 During Cropping Season+ γMean Rainfall During Cropping Season

+δd + πy + epsiliondy

In the above equation, Ydy is the log(yield) of a particular crop during Kharif season in a district d in

the year y and δd is district fixed effects and πy is the year fixed effects.

The results are in table 11. The results show high temperatures clearly reduce yields of several major

crops. The yields of rice, bajra and jowar are particularly important as they are food crops. This result hints

at the mechanism through which high temperatures might effect fetal health.

Page 35


Table 11: Effect of Temperature on Yeild (Kharif Season only)

(1) (2) (3)Rice Bajra Cotton

CDD >90F -.0003** -.0012*** -.0013***During Sowing and Harvesting

(.0001) (.0003) (.0003)

Mean Rainfall .0004*** .0002*** .0005***During Sowing and Harvesting

.0000 .0000 .0001N 4809 3214 2611

(4) (5) (6)Soyabean Groundnut Jowar

CDD >90F -.0009*** -.0004** -.0010***During Sowing and Harvesting

(.0003) (.0002) (.0004)

Mean Rainfall .0001*** .0004*** .0002***During Sowing and Harvesting

.0000 .0000 .0000N 2041 3615 3319

The dependent variable is log yield of each of these crops. All the regressions contain a district fixed effects and a year

fixed effects. The sample is from 1998 to 2010. The season of sowing and harvesting kharif crops is considered from May

to October. Robust Standard Errors clustered at the district level.

Page 36

Heat, Infant Mortality and Adaptation: Evidence from India · The authors also thank Panchajanya Banerjee and Aniruddha Ghosh for helping us with the weather and lights data. All

Documents