ArticlePDF Available

Water is Life, Clean Water Means Health: The Effect of Early-Life Exposure to the City-Wide Water Filtration on Old-Age Mortality

November 2024
American Journal of Health Economics

DOI:10.1086/734081

Authors:

Hamid Noghanibehambari

Austin Peay State University

Exploring Mechanism Channels

…

Figures - uploaded by Hamid Noghanibehambari

Content may be subject to copyright.

Content uploaded by Hamid Noghanibehambari

Content may be subject to copyright.

Water is Life, Clean Water Means Health: The Effect of Early-Life

Exposure to City-Wide Water Filtration on Old-Age Male Mortality*†

Hamid Noghanibehambari‡

Jason Fletcher§

Abstract

This study examines the impact of water purification on long-run old-age mortality. We

examine the effects of early-life and childhood exposure to improvements in water quality

due to city-wide water filtration programs in 25 major American cities on later-life old-age

longevity of male individuals. We employ data from Social Security Administration death

records linked with the 1940 census. The difference-in-difference regressions suggest an

improvement in male longevity of about 3.2 months. A series of balancing tests do not

reveal evidence that changes in sociodemographic and socioeconomic characteristics of

individuals confound the estimates. We also implement a full battery of sensitivity analyses

and show that the effect is robust across specifications, subsamples, and functional form

checks. Analyses using 1950-1970 censuses suggest that a portion of the long-term links

can be explained by improvements in individuals’ education and income as a result of

early-life exposure to water filtration. We also show that treated cohorts reveal

improvements in height and cognitive scores during early adulthood.

Keywords: Mortality, Longevity, Public Health, Clean Water, Water Filtration, In-Utero

Exposures, Early-Life Exposures, Historical Data

JEL Codes: I18, J18, N51, N52, O13, Q28

*The authors claim that they have no conflict of interest to report. The authors would like to acknowledge financial

support from NIA grants (R01AG060109, R01AG076830) and the Center for Demography of Health and Aging

(CDHA) at the University of Wisconsin-Madison under NIA core grant P30 AG17266.

† The phrase "Water is life, clean water means health" is inspired by a quote from Audrey Hepburn, who was a

passionate advocate for clean water access.

‡ Corresponding Author, College of Business, Austin Peay State University, Marion St, Clarksville, TN 37040, USA

Email: noghanih@apsu.edu, ORCID ID: https://orcid.org/0000-0001-7868-2900

§La Follette School of Public Affairs, University of Wisconsin-Madison, 1225 Observatory Drive, Madison, WI

53706-1211, USA

Email: jason.fletcher@wisc.edu

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

1. Introduction

Contaminated drinking water has been the source of various diseases and a serious threat

to public health for centuries. As of 2019, about 2.2 billion people worldwide still lack access to

safely managed drinking water (WHO 2019). The health impacts of poor water quality are

disproportionate across age distribution, with higher impacts among infants and children. This is

more evident as death rates due to waterborne diseases such as typhoid, cholera, and diarrhea have

been historically higher among infants and children (Armstrong, Conn, and Pinner 1999). In the

US during the early 20th century, there have been substantial improvements in water quality

through a series of state-wide and city-wide campaigns to raise expenditures toward public health

infrastructures. These campaigns were successful in raising the access and quality of water and

contributed to reductions in urban mortality during the first decades of the 20th century (Anderson,

Charles, & Rees, 2022; Beach et al., 2016; Cutler & Miller, 2005; Troesken, 2004) Specifically,

studies suggest that clean water benefits infants’ health more than other age groups (Anderson,

Charles, and Rees 2022). Among many public health interventions, improvements in water

technology, specifically water filtration, have affected infant mortality rates most (Costa 2015).

Studies that use more recent data from developing countries that experience rapid industrialization

suggest the importance of water quality for infants’ health outcomes (Zhang and Xu 2016;

Greenstone and Hanna 2014; Mettetal 2019).

The potential effects of water quality on infants could have long-lasting consequences. A

growing literature suggests that life-cycle outcomes could partly reflect conditions in early-life

(Almond and Currie 2011b; 2011a; Almond, Currie, and Duque 2018; Almond, Currie, and

Herrmann 2012; Aizer et al. 2016; Barker 1997; 1994; 1995; 2004). A strand of this literature

explores the sources of disparate longevity across individuals and documents the significant effects

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

of early-life shocks on old-age health and mortality (Bailey et al., 2016; Hayward & Gorman,

2004; Montez et al., 2014; Lindeboom et al., 2010; Scholte et al., 2015; Smith, 2009; Van Den

Berg et al., 2006, 2011). For instance, studies show that in-utero and early-life disease environment

is associated with adulthood education and income and old-age cognitive functioning and

longevity (Blackwell, Hayward, and Crimmins 2001; Crimmins and Finch 2006; Finch and

Crimmins 2004; Moore et al. 2006; Venkataramani 2012; Bleakley 2007; Case and Paxson 2009).

Therefore, the effects of water quality in early life can also be detected in life-cycle outcomes.

However, while the evidence on its short-term effects is abundant, very few studies have explored

the long-term effects, specifically on old-age health and mortality (Grossman and Slusky 2019;

Hafeman et al. 2007; He and Perloff 2016; Jones 2019; Beland and Oloomi 2019). To fill this void

in the literature, this paper examines the effects of water filtration across US cities on old-age

mortality.

Water filtration involves passing water through a series of physical or chemical barriers or

employing other biological mechanisms to remove contaminants, impurities, pollutants,

pathogens, and other harmful microorganisms, resulting in cleaner, safer water. Such filtration and

purification systems are crucial in preventing waterborne diseases, such as cholera, typhoid fever,

and dysentery.

These diseases are caused by bacteria, viruses, and protozoa in the contaminated

water. While these pathogens are harmful to the whole population, the impacts are considerably

stronger in critical stages of development, especially during in-utero, early life, and childhood

(Kunitz 1984; Condran and Crimmins-Gardner 1978; Beach et al. 2016). Moreover, these diseases

Several bacterial diseases (e.g., Salmonella infections, Cholera caused by Vibrio cholerae, Typhoid Fever caused by

Salmonella typhi, Dysentery caused by Shigella bacteria, Legionellosis caused by Legionella pneumophila, and certain

strains of E. coli infection), parasitic diseases (e.g., Giardiasis caused by Giardia lamblia, Cryptosporidiosis caused

by Cryptosporidium, and amoebic dysentery caused by Entamoeba histolytica), and viral diseases (e.g., Hepatitis A,

Norovirus infection, and Polio) can be transmitted through contaminated water.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

have spillover impacts on survivors and make them vulnerable to other non-waterborne diseases.

For instance, based on the hypothesis of Mills–Reincke phenomenon, the prevention of typhoid

fever death resulting from water filtration and purification is associated with a threefold reduction

in deaths from other non-waterborne diseases (McGee 1920; Friedrich 1912). Moreover, the low

case-fatality rate of waterborne diseases such as typhoid increases the vulnerability of survivors to

later-life diseases such as tuberculosis, pneumonia, and kidney failure (Ferrie and Troesken 2008).

Therefore, one might expect to detect the benefits of exposure to water filtration during infancy

and childhood for later-life and adulthood outcomes.

This paper employs data from Social Security Administration death records of male

individuals over the years 1975-2005 linked with the full-count 1940 census. We focus on male

individuals because alternative data containing female death records cover a relatively narrower

window of deaths. Furthermore, since women often change their names after marriage and data

linkage between death records and census records is primarily based on name similarities, women

are underrepresented in the linked data. We exploit cross-census linkages to find individual records

in historical full-count censuses 1900-1930 in order to infer their city-of-birth/childhood. We

implement difference-in-difference regressions to compare the longevity of individuals who were

exposed to city-specific water filtration projects across different ages. We show that city-wide

improvements in water quality during in-utero and early-life is associated with about 3.2 months

higher age-at-death during old ages.

The main assumption in our identification strategy is that the longevity of individuals in

cities that implemented water filtration projects would have followed the same path and been

influenced by the same factors in the absence of any water filtration project. We provide empirical

evidence to support the exogeneity of the treatment. We show that the observed effect is not an

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

artifact of changes in the sociodemographic composition of the final sample due to differential

survival into adulthood, changes in the socioeconomic and sociodemographic composition of cities

following public health reforms, changes in public spending as well as implementation of other

public health interventions, changes in other city-level and state-level policies and reforms,

endogenous merging across censuses and death records, and endogenous fertility. A series of

sensitivity analyses suggest that the effect is robust across a wide range of alternative

specifications, functional forms, and subsamples.

To explore mechanism channels, we use census data for the years 1950-1970, a period

when cohorts under study are experiencing their adulthood years. We show that exposure to water

quality improvements in the year-of-birth is associated with higher schooling, higher income, and

increased socioeconomic scores. Several studies document the association between

education/income profile and later-life longevity (Chetty et al., 2016; Cristia, 2009; Cutler et al.,

2006; Fletcher et al., 2021; Fletcher, 2015; Kinge et al., 2019; Lleras-Muney, 2022). Therefore,

we argue that increases in education and improvements in labor market outcomes could be

potential mechanism channels.

This paper contributes to two strands of literature. First, to our knowledge, this is the first

study to examine the long-run mortality effects of water quality in early-life. While several studies

document the relevance of in-utero and early-life water quality exposure on education and labor

market outcomes, no study has explored its effects on old-age health and mortality (Beach et al.,

2016; Smith et al., 2012; Zaveri et al., 2019; Zhang & Xu, 2016). More specifically, this study is

the first to examine the long-term effects of water quality improvements in the US’s first decades

of the 20th century on later-life health outcomes. Studies that examine health effects of public

health interventions during this era have focused mostly on short-run outcomes (Anderson,

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Charles, and Rees 2022; Anderson et al. 2022; 2019; D. Cutler and Miller 2005). Improving

drinking water quality reduces the disease burden of early-life environments. Therefore, our study

also evaluates the impact of reducing early-life disease burden on later-life health. Hence, the

second contribution of our paper is to add to the growing literature on the relevance of early-life

disease environment and later-life mortality (Case and Paxson 2009; Case, Fertig, and Paxson

2005; Bleakley 2007; Crimmins and Finch 2006; Almond, Currie, and Herrmann 2012; Cormack,

Lazuka, and Quaranta 2024).

The rest of the paper is organized as follows. Section 2 reviews the literature. Section 3

introduces data sources and sample selections. Section 4 discusses the econometric method.

Section 5 reviews the main results. Section 6 explores potential mechanism channels. Finally, we

depart some concluding remarks in section 7.

2. A Brief Literature Review

Several studies document the health benefits of public health infrastructures in American

cities during the early decades of the 20th century. Cutler & Miller (2005) employ data from 13

major American cities and explore the role of improvements in clean water technologies in

reducing urban mortality rates. They find significant reductions in total mortality rates. Further,

they show that infant mortality drops by about 35% in the years following water disinfection.

Anderson, Charles, & Rees (2022) revisit their analysis and make some corrections to their data

caused by transcription errors. Their reevaluation suggests smaller but significant impacts of water

filtration. They document that post-water filtration infant mortality drops by about 11%. Although

they also explore the impacts of other public health interventions, including water projects, water

chlorination, water filtration, and sewage treatment, they fail to find statistically significant effects

of any of these interventions. Anderson et al. (2019) examine the effect of tuberculosis movements

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

in the US between 1900-1917 on TB mortality rates. They find modest and mostly insignificant

effects of various anti-TB measures on mortality. However, they show that establishing state-run

sanatoriums resulted in about 4 percent reductions in pulmonary TB mortality. Anderson, Charles,

McKelligott, et al. (2022) explore the effect of milk inspections in major American cities during

1880-1910 on infants’ and children’s mortality and find small and insignificant effects.

A strand of research employs more recent data and explores the association between water

quality and infants’ and children’s health outcomes (Apergis et al., 2019; Currie et al., 2017;

Greenstone & Hanna, 2014; Hafeman et al., 2007; He & Perloff, 2016; Hill & Ma, 2017). For

instance, Clay et al. (2014) examine the effects of exposure to lead in drinking water across cities

in the US between 1900-1920. They exploit the fact that lead is more likely to leach into drinking

water if the water is more acidic. They document that going from the city with high exposure to

the city with low exposure in their sample is associated with a 7-33 percent reduction in infant

mortality rates.

Grossman & Slusky (2019) examine the impact of the Flint water crisis, in which the city

of Flint, Michigan, changed its water supply resulting in sharp increases of contaminants in the

water supply, on fertility and birth outcomes. They find that rises in drinking water contaminants,

including lead, resulted in a 12 percent decrease in fertility rate and a 5.4 percent reduction in birth

weight. Currie et al. (2013) examine the impact of water quality on birth outcomes using data from

New Jersey between 1997-2007. They compare birth outcomes of infants from the same mother

who were exposed to differential contaminations in drinking water. They find negative and

significant effects on birth weight and the gestational length of infants of low-educated mothers.

Brainerd & Menon (2014) investigate the effect of water pollution due to seasonal changes in

fertilizer agrichemical use on infants’ and children’s health outcomes in India. They show that

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

children exposed to higher levels of agricultural water pollution exposure reveal higher mortality

rates and lower height and weight-for-age scores. Hill & Ma (2022) and Hill (2018) provide

evidence that shale gas development during the recent fracking boom in the US resulted in higher

water pollution and negative impacts on infants’ health outcomes. Jones (2019) examines the

impact of microcystin in drinking water, a potent toxin produced by cyanobacteria in freshwater

algal blooms, on infants’ health outcomes. He uses data from Michigan and exploits a one-time

municipal attempt to improve water quality and remove algae. He finds that the intervention

increased to 17 grams in birth weight and 3.2 days additional gestational age.

Therefore, one expects to observe positive impacts of improvements in water quality on

infants’ health. Healthier infants are more likely to experience a healthier childhood, develop more

cognitive and noncognitive skills, attain higher levels of human capital, reveal better labor market

outcomes, and generally have a healthier adulthood (Behrman & Rosenzweig, 2004; Black et al.,

2007; Cook & Fletcher, 2015; Fletcher, 2011; Maruyama & Heinesen, 2020; Royer, 2009; Shenkin

et al., 2009). A narrow strand of research examines the direct link between in-utero and early-life

exposure to a change in water quality and later-life outcomes (Smith et al., 2006, 2012; Zaveri et

al., 2019). For instance, Beach et al. (2016) argue that water purification technologies significantly

reduced typhoid mortality rates. They proxy water quality with city-level typhoid mortality for the

period 1900-1940 and examine the impacts of improvements in water quality in early-life on

adulthood education and earnings. They find that water improvements resulted in an increase of

about 9 percent higher adulthood income and 0.7 years additional years of schooling. Zhang & Xu

(2016) examine the impact of a major water treatment program in rural China. They find that those

who benefited from the program in early-life attain roughly 1 additional year of schooling during

adulthood.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

3. Data and Sample

The primary data source of this study comes from Death Master Files (DMF) of Social

Security Administration death records extracted from the Censoc Project (Goldstein et al. 2021;

Breen, Osborne, and Goldstein 2023; Breen and Osborne 2022). DMF data covers deaths that

occurred among male individuals born between 1975 and 2005.

It contains information on dates

of birth, death, limited demographic characteristics, and an identifier to link the individual with

the full-count 1940 census. DMF data has two advantages that make it superior to alternative data

sources in examining long-term associations. First, the raw data contains millions of observations,

allowing for various types of heterogeneity analyses. Second, the data contains all familial and

geographic information available in the 1940 census. Specifically, we have information on the

below-state place of residence in both 1935 and 1940, as reported in the 1940 census. Moreover,

the existence of cross-census linking rules allow researchers to link the full-count 1940 census to

historical censuses which enables them to observe place-of-residence of individuals in earlier

decades. We merge the DMF data with the 1940 census extracted from Ruggles et al. (2020).

We use city-level water filtration project data from Anderson, Charles, & Rees (2022). It

provides a city-by-year panel of 25 major cities between 1900-1940 and reports whether a city has

a water filtration system each year. Figure 1 shows the geographic distribution of cities in the final

sample and the year each city implemented water filtration. Error! Reference source not found.

provides a list of these cities with the year of water filtration in each city.

The DMF data reports deaths to male individuals only. We use the Berkeley Unified Numident Mortality Database

(BUNMD) to examine the effects across both genders. The disadvantage of BUNMD data is that its death coverage

is more comprehensive post-1988 years and that it is not linked to the 1940 census. Hence, we implement the main

analysis of the paper using DMF data. In Error! Reference source not found., we show that using BUNMD data we

observe almost identical coefficients to those of the main results if restrict the sample to male individuals only.

However, among female individuals, the coefficients are considerably smaller in magnitude and statistically

insignificant.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Since this study focuses on assessing early-life exposures, we need to assign city-level

water filtration status based on individuals’ year-of-birth and city-of-birth/childhood. However,

the 1940 census does not report the city-of-birth/childhood. One idea is to use the city-of-residence

in 1935 and 1940 as a proxy for place-of-birth. However, individuals may migrate from their

birthplace, and this migration could be a response to city-level improvements in public health,

hence being endogenous. To mitigate this issue and infer city-of-birth/childhood, we start with

DMF-census-linked data and merge the records with historical censuses 1900-1930 using cross-

census linkage datasets extracted from Abramitzky et al. (2020). We then use the geographic

information provided by the census in which a person appears for the first time in any census as

the place of birth/childhood.

We then merge the DMF-census data with the water filtration data

based on birth/childhood-city and birth-year.

In our regressions, we also control for a series of city-by-year covariates. These covariates

are constructed using full-count decennial censuses 1900-1940 and linearly interpolated for inter-

decennial years. Section 6 also uses censuses from 1950-1970 to explore potential pathways. These

census data are extracted from Ruggles et al. (2020). Finally, in section 5.3, we employ natality

and mortality data at the city level extracted from Bailey, Clay, et al. (2016).

Our sample consists of a relatively long birth window (i.e., 41 years) and a relatively

limited death window (i.e., 31 years). One concern in our sample selection is that our method of

comparison between early versus late filtration adoption could reflect the longevity differences

To the extent that migration is correlated with childhood exposure to water filtration, measurement errors induced

by migration in city of birth/childhood assignment may confound our estimates. In Error! Reference source not

found., we argue that although between 20-40% of the linked samples (from the full count 1940 census to 1910-1930

census) moved across cities, such migration patterns do not correlate with exposure to water filtration after accounting

for fixed effects and covariates.

Error! Reference source not found. discusses cross-census linking procedure and steps of sample construction in

more detail.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

between earlier cohorts versus later cohorts. This problem is more evident given the sharp rises in

life expectancy at birth for these cohorts (Smith & Bradshaw, 2006). To mitigate this issue, we

restrict cohorts to those born 15 years before and after the city-specific year of water filtration.

The final sample includes 338,758 observations born between 1900-1940 and who died

between 1975-2005. Summary statistics of the final sample are reported in Table 1. The average

age-at-death is 867.2 months or 72.3 years but it varies between 35-104 years.

Roughly 98

percent of the observations are white. This overrepresentation of whites in the final sample results

from a higher match rate among whites both in the DMF-1940-census match and in linking

historical censuses. In section 5.3, we empirically test whether the observed match rules are

endogenous, i.e., they are correlated with city-level water filtration. Moreover, Breen & Osborne

(2022) argue that while certain groups are underrepresented in the Censoc-linked death records,

they represent their original population of 1940 records in terms of socioeconomic and education.

About 90 percent of mothers and fathers are literate. The parental information is also extracted

from the earliest census each individual appears. Therefore, they reveal parental covariates during

individuals’ birth/childhood. Roughly 95 percent of fathers are active in the labor force at the time

of the 1940 census enumeration. Our primary independent variable is the share of childhood years

In Error! Reference source not found., we argue that this selection is not critical for the main findings. Removing

this restriction or making a stricter balancing window of selection around water filtration reforms only slightly changes

the coefficient size.

We do not restrict the sample based on age at death. One concern is that individuals who were born earlier in the

sample and were exposed to earlier filtration reforms must have lived into much older ages to be in the final sample

than those later cohorts. Moreover, the longevity of earlier cohorts for inclusion in the final sample is beyond the life

expectancy of cohorts in the early 20th century. Therefore, these cohorts could possibly contain quiet different

characteristics than the later cohorts to have lived beyond life expectancy of their cohorts. Such differential longevity

of earlier versus later cohorts might bring causes of concern related to endogenous survival. In Error! Reference

source not found., we restrict the sample to individuals survived up to ages 50, 55, 60, 65, and 70 and replicate the

main results. We observe coefficients that are about 30% smaller than the main results. This fact implies two scenarios.

One, the benefits of filtration appear to be larger for younger ages at death and that expanding death window to cover

deaths prior to 1975 might increase coefficient sizes. Second, possible survival selection of earlier cohorts might bias

coefficients downward and that the main results underestimate the true effects.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

(ages 0-15) that the individual is exposed to water filtration. The average share of exposure is 0.16

with a standard deviation of 0.36.

4. Econometric Method

To operationalize the long-run effects of water filtration on mortality, we implement a

difference-in-difference framework to compare the age-at-death of individuals who were exposed

to the adoption of the water filtration system to those in cities without a filtration system across

different ages during their childhood. In other words, we examine the impacts across different ages

at exposure. Specifically, we implement regressions of the following forms:

    󰇛  󰇜

󰇟󰇠   

(1)

Where  is age-at-death (in months) of individual  who was born in city  in census region

 and belonged to birth-year . 󰇛󰇜 is a unit function that equals one if the inside argument is true.

 represents city-specific year of water filtration. Therefore, the parameters  represent impacts

across various age-at-exposure . For instance,  is the coefficient of exposure for cohorts who

turn age 5 at the time of water filtration. Similarly,  is the coefficient of exposure of cohorts who

are born 5 years post-waterwork, hence a full in-utero and childhood exposure. We eliminate the

coefficient of 13-15-year-old individuals (i.e., =[-15,-13]) to compare all coefficients to the values

of the oldest cohorts in our sample.

In matrix , we include individual and family covariates, including indicators of race,

ethnicity, maternal literacy, paternal literacy, and paternal occupational income score. In , we

include city-level controls including share of married, labor force participation rate, share of people

in different occupations, share of homeowners, share of children, and average socioeconomic

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

score. The parameter  represents city fixed effects that account for time-invariant city-level

confounders in longevity. Therefore, we rely on cross-cohort differences post-versus-pre water

filtration to eliminate the unobserved heterogeneity across cities. The parameter  represents birth-

cohort by birth-region fixed effects to absorb unobserved temporal heterogeneity across cohorts

that are specific to each census region. Finally,  is a disturbance term. We cluster standard errors

at the city level. Since the final sample includes only 25 cities, we have very few clusters to rely

on inference based on city-level clustering. Therefore, in visual representations of equation 1, we

illustrate confidence intervals extracted from the wild bootstrap procedure.

We further examine the impacts in a difference-in-difference framework and assign the

exposure measure based on the share of childhood years (up to age 15) that an individual was

exposed to water filtration. Specifically, we implement regressions of the following forms:

     

(2)

In this formulation, the variable  is the share of childhood between ages 0-15 that the

individual was exposed to water filtration. For instance, Baltimore, MD initiated water filtration

in 1915. An individual born in 1910 is potentially exposed to water filtration for ages 6-onward.

Hence, the average share of exposure is 0.6.

Therefore,  is the parameter of interest that

captures the association between full exposure to water filtration (versus no exposure) and old-age

longevity. All other parameters are similar to those of equation 1. In all regressions, we report P-

values based on wild bootstrap procedures.

Recall that we limit the sample to those born 15 years pre- and post-waterwork. In Error! Reference source not

found., we show the results for other threshold ages. Specifically, we assign exposure measure up to age 1, 5, 10, and

14. Since all these age groups are treated in the final sample, as we limit exposure ages, we expect to observe smaller

coefficients as those ages join the reference group.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

5. Results

5.1. Age-at-Exposure Analysis

The main results of equation 1 are reported in the top panel of Figure 2. Compared to

cohorts 13-15 years old at the time of water filtration, we observe very small changes for

individuals 11-12 years old. For cohorts between ages 5-10, we observe a positive average effect

that is small in magnitude. The effects start to rise for those 3-4 years old and younger. For

instance, the coefficients imply an increase of 0.8, 3, and 1.5 months of additional longevity for

age of exposure of [5,6], [3,4], and [1,2], respectively. For exposure during year of birth and

between 1-7 years prior to birth, the coefficients suggest increases in longevity in the range of 1.9

– 2.5. Overall, four out of seven coefficients for the age group of 3-4-and-younger are statistically

significant at the 5 percent.

Difference-in-Difference Bias. The literature suggests that OLS-produced difference-in-

difference estimates in a staggered adoption setting where different units receive treatments in

different periods could produce biased estimates (Callaway and Sant’Anna 2021; Borusyak,

Jaravel, and Spiess 2021; Sun and Abraham 2021; Goodman-Bacon 2021). To explore this

potential bias, we implement an alternative difference-in-difference method developed by Sun &

Abraham (2021) and replicate the regressions of equation 1. The results are depicted in the bottom

panel of Figure 2. Compared with the OLS estimations of the top panel, we observe a very similar

pattern in estimated effects, suggesting little bias in the OLS estimates.

In Error! Reference source not found., we group different ages at exposure to observe and compare coefficient

sizes across different ages. Specifically, we examine exposure during in utero and age 0, ages 1-4, and ages 5-9. We

find a monotonic pattern: the earlier the exposure, the larger the coefficient. For instance, for in utero and early-life,

exposure is associated with 3.3 months higher longevity while for exposure during ages 1-4 the coefficient implies a

2.6-month rise in longevity.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

5.2. Main Results

The main results of equation 2 for the cumulative childhood exposure are reported in Table

2. We start with a parsimonious model that only includes city fixed effects, birth-year fixed effects,

and individual and family controls (column 1). Then, we add city controls in column 2 and birth-

region-by-birth-year fixed effects in column 3. The estimated effects are quite stable across

specifications. Based on the fully parametrized specification of column 3, exposure to filtered

water throughout childhood is associated with about 3.2 months higher longevity.

This finding is in line with a narrow literature that shows removing contaminants in

drinking water, e.g., arsenic, during in-utero and early-life could reduce mortality in young adults

(Smith et al., 2006, 2012). This is also consistent with studies that suggest improvements in early-

life water quality increase later-life human capital (Zhang and Xu 2016; Beach et al. 2016).

We can compare this effect with the impacts of other early-life exposures to gauge its

economic magnitude. Aizer et al. (2016) examine the impacts of the Mothers’ Pension (MP)

program, a cash transfer needs-based program to help poor families, on later-life education and

longevity. They find that MP receipt during childhood is associated with 0.6 years of more

schooling and almost 1 year of additional longevity. MP transfers accounted for about 12-25

percent of family income and lasted for three years, resulting in a cumulated income shock of about

50 percent of family income. Therefore, the intent-to-treat effect of Table 2 is equivalent to an

increase of 13.3 percent in family income. Halpern-Manners et al. (2020) employ twin fixed effect

strategy and explore the effect of education on longevity. They find that an increase of 1 year of

schooling is associated with about 4 months higher longevity. Therefore, our estimated effect is

equivalent to about 0.8 additional years of schooling. Noghanibehambari & Fletcher (2023b)

explore the impact of in-utero exposure to state-level and federal alcohol prohibition during the

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

early decades of the 20th century on old-age longevity. They find that prohibition is associated with

a treatment-on-treated increase of 1.7 years in longevity. The estimated impact of Table 2 suggests

that unfiltered water has at least about 16 percent of the effect of alcohol consumption during

pregnancy. Fletcher & Noghanibehambari (2024) examine the in-utero and early-life exposure to

agrichemical pesticide exposure on old-age longevity. They use the emergence of cyclical cicadas

in eastern states that raises the pesticide use in tree croplands as a natural experiment. They employ

DMF data and show that exposed cohorts reveal 2.2 months lower longevity. They argue that

contaminating drinking water is a likely channel of exposure of infants and mothers to pesticide

use. Their estimated effect is about 69 percent of the benefit of water filtration in the current study.

5.3. Endogeneity Concerns

This section discusses several potential sources of endogeneity and selection concerns. We

list these concerns below and attempt to empirically test them using available data.

Balancing Tests. One concern is that the final sample is unbalanced and represents certain

sociodemographic populations more than others. If this over/under-representation is correlated

with water filtration even after controlling for fixed effects and covariates, then the regressions

produce biased estimates. For instance, assume that following the public health improvements that

resulted in water filtration, cities observe a sharp inflow of migrants from neighboring cities and

counties. Also, assume that there are more whites and people of better socioeconomic conditions

among these migrants. In this case, the regressions overstate the true effects and capture the higher

longevity of these subpopulations of migrants rather than the effect of water filtration. Another

source of an unbalanced sample is differential mortality and survival into adulthood and old age.

If childhood and middle age mortality is affected by early-life water quality and the effects vary

by sociodemographic characteristics, then the observed effects on longevity could reflect the

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

differential mortality patterns of survivors rather than the direct impact of water quality. We can

directly examine these sources of endogeneity by implementing a series of balancing tests. In so

doing, we use individual and family characteristics as the outcomes and implement regressions of

equation 1, conditional on city and region-by-cohort fixed effects. We then depict the coefficients

dummies for each outcome in different panels of Figure 3 through Figure 5. To facilitate

comparison across panels and figures, we standardize each outcome with respect to the mean and

standard deviation of the sample. We do not observe a discernible effect on the likelihood of being

white, black, or Hispanic across ages of exposure to the water filtration reforms (top-left, top-right,

and bottom-left panels of Figure 3). The coefficients are economically small and statistically

insignificant at 95 percent level.

However, we observe negative and significant coefficients for the outcome of mother

literate for younger children (bottom-right panel of Figure 3). Since mother education is shown to

have a positive impact on infants’ and children’s health, these reductions suggest that our results

may underestimate the true effects (Lundborg, Nilsson, and Rooth 2014; Huebener 2020; 2019;

Noghanibehambari, Salari, and Tavassoli 2022).

We do not observe any discernible change across ages following the reform for mother

labor force status, father literacy, and missing indicators of parental literacy (Figure 4). We also

observe no cross-age trend in the father’s socioeconomic and occupational income scores (top-

right and bottom-left panels of Figure 5, respectively).

We further examine potential changes in fertility following water filtration. These results are discussed in Error!

Reference source not found.. Although consistent with prior research in this area we find reductions in infant

mortality rates, we do not observe significant changes in birth rate (Anderson, Charles, & Rees, 2022; Cutler & Miller,

2005).

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

A similar concern relates to the fact that water filtration may be preceded by other state-

level policy changes or city-level socioeconomic and sociodemographic changes. Thus, to the

extent that such place-specific policy evolution and compositional changes correlate with

longevity, the impacts might pick up on these confounders rather than water filtration. In Error!

Reference source not found., we implement event studies to examine the evolution of state-level

and city-level outcomes in different years relative to water filtration years, conditional on city and

region-by-year fixed effects. The evidence does not provide empirical support for this concern.

Specifically, we do not observe any association between the water filtration and prohibition

movement, suffrage movement, tax policies, birth registration laws, child labor laws, compulsory

attendance laws, sociodemographic composition, and a battery of socioeconomic outcomes.

Similarly, one might argue that water filtration projects are one step out a larger set of

staggered piecemeal development plans that expand the general provision of public goods. In that

case, the effects pick up on the benefits of other public projects and social spending. In Error!

Reference source not found., we empirically investigate such concerns, and, through a series of

event studies, show that water filtration projects do not correlate with spending on public

education, per capita doctors as a measure of healthcare access, water chlorination projects, and

several other public health intervention projects.

Endogenous Merging with Censuses. Selection from the original population to the final

sample caused by data linking may generate bias in our estimations if the selection procedure is

correlated with water filtration projects (see section 3 and Error! Reference source not found.).

To examine this source of selection-induced endogeneity, we assess the association between

The policies mentioned here and explored in Error! Reference source not found. are documented to influence

later-life mortality and longevity (Lleras-Muney 2005; Noghanibehambari and Fletcher 2023a; Noghanibehambari

and Noghani 2023; Noghanibehambari and Fletcher 2023b).

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

successful survival from the original population to the final sample with the water filtration

exposure measure. In so doing, we start with the universe of cohorts born between 1900-1940 and

who reside in the final sample’s cities. We merge this population with those in the final sample

and generate a successful merger dummy if the merging is successful. We then implement

regressions that include city and cohort-by-region fixed effects, similar to equation 2. The results

are reported in Table 3. We show the estimated associations between successful merger and water

filtration for the full sample, sample of whites, and sample of nonwhites in columns 1-3,

respectively. The coefficients suggest insignificant associations. Moreover, the magnitude of the

effects is small. For instance, exposure to water filtration is associated with an insignificant 2.6

basis-points increase in the probability of merging, equivalent to about a 0.8 percent change from

the outcome mean. These results do not provide evidence for the endogeneity caused by cross-

census and DMF-census linking and selection.

5.4. Robustness Checks

In Table 4, we explore the sensitivity of the results to alternative model specifications.

Column 1 replicates the results of column 3 of Table 2 to provide a benchmark comparison. All

other columns include all covariates and fixed effects used in column 1. In column 2, we interact

birth-state by 1940-state fixed effects to control for the influence of early- adulthood cross-state

migration on the water-longevity relationship. The estimated effect is quite similar to the main

results.

In columns 3-4, we interact city fixed effects with individual and family dummies. Thus,

we allow for time-invariant unobserved factors of each city to have a differential impact on health

and longevity across people of different sociodemographic and socioeconomic backgrounds. The

estimated effects are comparable to that of column 1.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Studies suggest that the season of birth is associated with infants’ health and later-life

mortality (Doblhammer and Vaupel 2001; Currie and Schwandt 2013). Moreover, several causes

of death reveal a seasonality pattern (Seretakis et al. 1997; Falagas et al. 2009). To account for

seasonality-related confounders, we add birth-month and death-month fixed effects to the

regressions. The result, reported in column 5, is almost identical to that of column 1.

In column 6, we add a wide range of additional city-level controls, including share of

people in different demographic groups, share of people in different age groups, average

socioeconomic score, female labor force participation rate, male labor force participation rate,

female literacy rate, male literacy rate, and population. The estimated effect becomes only slightly

smaller than the main results and remains statistically significant.

In column 7, we implement an alternative standard error correction method. Instead of

clustering, we use Huber-White robust standard error. The estimated effect remains statistically

significant at 95% level.

Another concern is regarding the functional form of the regressions. In column 8, we

replace the outcome with the log of age-at-death, hence estimating a semi-log specification. We

observe an effect of a 0.38 percent rise in longevity. This is very similar to the 0.37 percent rise

with respect to the outcome mean, implied by column 1. In column 9, we replace the outcome with

a dummy variable that equals one if the individual’s age-at-death is more than 75 years. Early-life

water filtration is associated with a 1.5 percentage-point rise in the probability of living beyond 75

years, off a mean of 0.38.

The main regressions of the paper do not incorporate any weighting method. In column 10,

we assign higher weights to more populated cities by weighting the regressions using the average

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

city population. The estimated effect increases by roughly 19% and remains statistically

significant.

Several studies suggest the long-term effects of early-life exposure to the Spanish Flu of

1918-1919 (Almond & Mazumder, 2005; Cook et al., 2019; Fletcher, 2018a, 2018b; Myrskylä et

al., 2013). In column 11, we remove cohorts born between 1918-1919 who were likely affected by

the pandemic. The estimated effect is quite similar to column 1.

The Great Depression induced unprecedented economic hardship among families that

could affect the children’s long-term outcomes (Cutler et al., 2007; Noghanibehambari et al., 2024;

Van Den Berg et al., 2006, 2009). Moreover, studies point to the benefits of New Deal relief

programs during this period for later-life outcomes (Noghanibehambari and Engelman 2022). Both

economic conditions and social spending could confound our estimates if they are correlated with

water filtration exposure. In column 12, we remove cohorts born between 1930-1940 who were

probably impacted by the Great Depression and New Deal social spending. The estimated effect

is quite similar to that of column 1.

6. Mechanisms

Improvements in health accumulation during infancy and childhood could lead to higher

longevity through several mediatory channels, including better human capital accumulation, better

mental health, better physical health, lower obesity, higher probability of family formation, better

spousal attributes, and higher socioeconomic index (Benítez-Silva & Ni, 2008; Cutler et al., 2006;

Diener & Chan, 2011; Gardner & Oswald, 2004; Lleras-Muney et al., 2022; Lleras-Muney &

Moreau, 2022; Noghanibehambari & Fletcher, 2023b, 2023c, 2023d; Preston, 2005; Van Den Berg

et al., 2015). In this section, we examine two important channels, education and socioeconomic

measures during adulthood. Since many cohorts have not yet completed their education in the 1940

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

census or entered the labor force, we turn to decennial censuses 1950-1970. We use information

about the city of residence in these censuses to proxy for city-of-birth. We restrict the sample to

male individuals born between 1900-1940 aged 25-55 and merge the data by water filtration

database based on census-city and year-of-birth. We assess the association between water filtration

and education/socioeconomic outcomes by implementing regressions that include individual

covariates, census year fixed effects, birth-year-by-birth-region fixed effects, and city fixed effects.

The results are reported in Table 5. Early-life and childhood water exposure are associated with

roughly 3.5 additional months of schooling (column 1). This effect is similar to the OLS findings

of Beach et al. (2016), who examine the effect of water purification in early-life on later-life

education. Water filtration also leads to about 3.7 percentage-points reductions in the likelihood

of less than high school education, off a mean of 0.16 (column 2).

Water quality exposure in early-life is also linked with socioeconomic measures. Exposure

to water filtration during childhood results in a 1.8-unit increase in the socioeconomic index,

equivalent to a 4.8 percent rise from the mean of the outcomes (column 4). We also observe

positive impacts on family income although the point estimates are noisy and limit interpretation

(columns 5-6).

If improvements in human capital and measures of socioeconomic status are mechanism

channels, then one would expect that these pathways follow similar heterogeneous variations as

those of longevity. In Error! Reference source not found., we provide evidence of significant

and sizable reductions in infant mortality following water filtration. If such improvements are the

results of water filtration and improvements in initial health capital, we may observe larger impacts

on longevity outcomes in areas with higher initial infant mortality rates. In Error! Reference

source not found., we examine this source of heterogeneity and show that the effects are

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

considerably larger for the subsample of high infant mortality rate cities. Moreover, we show that

the effects on schooling and socioeconomic index are primarily driven by the high infant mortality

subsample, supporting the role of human capital as mechanism channels between exposure to

water filtration and later life mortality.

The next question is to what extent improvements in these outcomes can explain the link

between water filtration and longevity. Chetty et al. (2016) examine the association between

household income and individual longevity in the US between 2001-2014 using individual tax

returns linked to the mortality database. They find that for each 5-percentile increase in income,

longevity increases by 0.7-0.9 years. For a household in the median of the sample, this means an

increase of about $40K (in 2020 dollars). Therefore, an increase of $2,603 (induced by water

filtration, column 5 of Table 5) is associated with about 0.62 months higher longevity. This is

about 20 percent of the reduced-form effect of Table 2.

Halpern-Manners et al. (2020) and Cutler & Lleras-Muney (2006) estimate that an increase

of 1 year in schooling is associated with 0.34 and 0.6 years higher longevity. Combining these

estimates with the estimated effect of column 1 of Table 5, one can deduce that the water-filtration-

induced rise in schooling leads to 1.2-2.1 months higher longevity. These effects are equivalent to

If the population of infants that survived as a result of water filtration is weaker (who would have died for their

weakness of other reasons in the absence of water filtration), then the overall benefits on longevity underestimate the

true effects. On the other hand, if water filtration brings health benefits and the results illustrate the improvements in

infants’ health capital, then the observed impacts on infant mortality are indeed the primary mechanism channel. We

can do a back-of-an-envelope calculation to examine this. Using Social Security Administration cohort life tables, we

estimate that the difference between post-infancy life expectancy and life expectancy at birth increased by about 4.5

years between the years 1900-1940 (SSA 2020). Based on aggregate vital statistics death records, infant mortality

rates decreased from around 150 to 47 infant deaths per 100K births (CDC 2015). The results of Error! Reference

source not found. implies a reduction of about 6 infants per 100K. Assuming that the difference in life expectancy at

age 1 and 0 can be solely attributed to reductions in infant mortality, the reduction of 6 infants per 100K imply roughly

3 months increases in life expectancy, a number that is quite similar to our main results.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

about 37-65 percent of the effect of column 3 of Table 2. Therefore, improvements in

education/income can explain about 20-65 percent of the observed long-term links.

To complement the mechanism channel analysis, we employ World War II enlistment data

linked with the 1940 census and DMF, extracted from Goldstein et al. (2021). This data is a subset

of DMF data in our main analysis for individuals who were enlisted for World War II. We explore

the effects on two measures of human capital and health capital. First, we focus on the Army

General Classification Test (AGCT) score. The AGCT was designed to capture the learning and

intellectual abilities of enlistees during World War II in order to assign them to different military

tasks and jobs (Potter, Helms, and Plassman 2008). Second, we explore the effects on height as

reported by enlistment enumerators. Height is an indicator of general health and is correlated with

other economic and health outcomes (Deaton and Arora 2009; Bozzoli, Deaton, and Quintana-

Domeque 2009; Deaton 2007). Specifically, some studies link height to old-age health and

longevity (Jousilahti et al. 2000; Spijker, Cámara, and Blanes 2012; Wilson 2019). We implement

the same sample construction and empirical method as the main results. The results are reported

in columns 7-8 of Table 5. We find a positive link between childhood exposure to water filtration

and AGCT score as well as height. Full exposure to water filtration during childhood is associated

with a 1.5% higher AGCT score and 1.7% increase in height. The wild bootstrap p-values for the

estimated coefficients of AGCT score and height are 0.17 and 0.15, respectively.

7. Conclusion

In the early 20th century, state and local authorities initiated a series of improvements in

public health infrastructure, including drinking water filtration and purification. In later decades,

many state and federal laws, including the Safe Drinking Water Act, Water Pollution Control Act,

and Clean Water Act, attempted to elevate drinking water quality further. Although there have

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

been substantial improvements in the quality of drinking water, there remain communities with

access to unsafe water (Mueller and Gasteyer 2021). There are also instances of temporary water

pollution with considerable negative health consequences (Grossman and Slusky 2019; Jones

2019). The situation is far worse in the rest of the world, specifically in poorer countries. About 1

in 4 people lack safely managed drinking water at home (WHO 2019). According to UNICEF

estimates, billions of people will lose access to safely managed drinking water by 2030 (UNICEF

2021). Therefore, it is important and policy-relevant to document the short-run and long-run effects

of water quality on human health outcomes.

This paper explored the long-run effects of in-utero and early-life exposure to water

filtration on old-age longevity. We exploited city-wide public health efforts to initiate a water

filtering system to purify water across 25 major American cities in the early 20th century. Our

results suggested a benefit of 3.2 months of additional longevity. We implemented a wide array of

tests to argue against endogeneity issues. We provided empirical evidence that changes in

sociodemographic and socioeconomic characteristics do not confound the estimates. We found no

evidence that these public health interventions coincide with any other city, county, or state-level

policy changes. Finally, we showed that the results are robust to a wide array of specification

checks, subsamples, and functional form checks.

Life expectancy at birth among male Americans increased from 46.3 to 60.8 years between

1900-1940, an increase of roughly 174 months of additional longevity. Based on our estimated

effect of Table 2, exposure to cleaner water as a result of water filtration can account for about 2

percent of the overall improvements in longevity for cohorts born between 1900-1940.

In our final sample, about 15.6% of observations are fully exposed to water filtration during

their childhood. Using the intent-to-treat estimate of the main results combined with the number

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

of fully exposed individuals in the final sample, one can calculate 14.1 thousand life years gained

due to childhood exposure to improvements in water quality as a result of water filtration.

Further, we can monetize this number using the Value of Statistical Life (VSL) estimates. Studies

suggest a VSL of about $10 million for the case of the United States (Kniesner and Viscusi 2019;

Viscusi 2018). Given the average longevity in the final sample of 72.3 years, one can roughly

calculate an annual VSL of $138.3 thousand. Therefore, the overall improvement in longevity of

the exposed cohorts in the final sample due to childhood exposure to water filtration is equivalent

to roughly $2 billion.

This is calculated using the number of fully exposed individuals in the final sample (15.6% of 338,758

observations), multiplying by the estimated effect of 3.2 months of additional longevity from Table 2, and

converting this value into years:      

 

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

References

Abramitzky, Ran, Leah Boustan, and Myera Rashid. 2020. “Census Linking Project: Version 1.0

[Dataset].” https://doi.org/https://censuslinkingproject.org.

Acemoglu, Daron, and Joshua Angrist. 2000. “How Large Are Human-Capital Externalities?

Evidence from Compulsory Schooling Laws.” NBER Macroeconomics Annual 15 (June):9–

59. https://doi.org/10.1086/654403.

Aizer, Anna, Shari Eli, Joseph Ferrie, and Adriana Lleras Muney. 2016. “The Long-Run Impact

of Cash Transfers to Poor Families.” American Economic Review 106 (4): 935–71.

https://doi.org/10.1257/AER.20140529.

Almond, Douglas, and Janet Currie. 2011a. Human Capital Development before Age Five.

Handbook of Labor Economics. Vol. 4. Elsevier. https://doi.org/10.1016/S0169-

7218(11)02413-0.

———. 2011b. “Killing Me Softly: The Fetal Origins Hypothesis.” Journal of Economic

Perspectives 25 (3): 153–72. https://doi.org/10.1257/JEP.25.3.153.

Almond, Douglas, Janet Currie, and Valentina Duque. 2018. “Childhood Circumstances and

Adult Outcomes: Act II.” Journal of Economic Literature 56 (4): 1360–1446.

Almond, Douglas, Janet Currie, and Mariesa Herrmann. 2012. “From Infant to Mother: Early

Disease Environment and Future Maternal Health.” Labour Economics 19 (4): 475–83.

https://doi.org/10.1016/J.LABECO.2012.05.015.

Almond, Douglas, and Bhashkar Mazumder. 2005. “The 1918 Influenza Pandemic and

Subsequent Health Outcomes: {An} Analysis of {SIPP} Data.” American Economic Review

95 (2): 258–62.

Anderson, D. Mark, Kerwin Kofi Charles, Michael McKelligott, and Daniel I. Rees. 2022.

“Estimating the Effects of Milk Inspections on Infant and Child Mortality, 1880−1910.”

AEA Papers and Proceedings 112 (May):188–92.

https://doi.org/10.1257/PANDP.20221066.

Anderson, D. Mark, Kerwin Kofi Charles, Claudio Las Heras Olivares, and Daniel I. Rees. 2019.

“Was the First Public Health Campaign Successful?” American Economic Journal: Applied

Economics 11 (2): 143–75. https://doi.org/10.1257/APP.20170411.

Anderson, D. Mark, Kerwin Kofi Charles, and Daniel I. Rees. 2022. “Reexamining the

Contribution of Public Health Efforts to the Decline in Urban Mortality.” American

Economic Journal: Applied Economics 14 (2): 126–57.

https://doi.org/10.1257/APP.20190034.

Apergis, Nicholas, Tasawar Hayat, and Tareq Saeed. 2019. “Fracking and Infant Mortality:

Fresh Evidence from Oklahoma.” Environmental Science and Pollution Research 26 (31):

32360–67. https://doi.org/10.1007/S11356-019-06478-Z/TABLES/5.

Armstrong, Gregory L., Laura A. Conn, and Robert W. Pinner. 1999. “Trends in Infectious

Disease Mortality in the United States During the 20th Century.” JAMA 281 (1): 61–66.

https://doi.org/10.1001/JAMA.281.1.61.

Bailey, Martha, Karen Clay, Price Fishback, Michael Haines, Shawn Kantor, Edson Severnini,

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

and Anna Wentz. 2016. “U.S. County-Level Natality and Mortality Data, 1915-2007.”

Inter-University Consortium for Political and Social Research.

https://doi.org/https://doi.org/10.3886/E100229V4.

Barker, D. J.P. 1994. Mothers, Babies, and Disease in Later Life. BMJ publishing group London.

———. 1995. “Fetal Origins of Coronary Heart Disease.” BMJ 311 (6998): 171–74.

https://doi.org/10.1136/BMJ.311.6998.171.

———. 1997. “Maternal Nutrition, Fetal Nutrition, and Disease in Later Life.” Nutrition 13 (9):

807–13. https://doi.org/10.1016/S0899-9007(97)00193-7.

———. 2004. “The Developmental Origins of Adult Disease.” Journal of the American College

of Nutrition 23 (December):588S-595S. https://doi.org/10.1080/07315724.2004.10719428.

Beach, Brian, Joseph Ferrie, Martin Saavedra, and Werner Troesken. 2016. “Typhoid Fever,

Water Quality, and Human Capital Formation.” The Journal of Economic History 76 (1):

41–75. https://doi.org/10.1017/S0022050716000413.

Behrman, Jere R, and Mark R Rosenzweig. 2004. “Returns to Birthweight.” Review of

Economics and Statistics. https://doi.org/10.1162/003465304323031139.

Beland, Louis Philippe, and Sara Oloomi. 2019. “Environmental Disaster, Pollution and Infant

Health: Evidence from the Deepwater Horizon Oil Spill.” Journal of Environmental

Economics and Management 98 (November):102265.

https://doi.org/10.1016/J.JEEM.2019.102265.

Benítez-Silva, Hugo, and Huan Ni. 2008. “Health Status and Health Dynamics in an Empirical

Model of Expected Longevity.” Journal of Health Economics 27 (3): 564–84.

https://doi.org/10.1016/J.JHEALECO.2007.09.008.

Berg, Gerard J. Van Den, Gabriele Doblhammer-Reiter, Kaare Christensen, Gerard J den Berg,

Gabriele Doblhammer-Reiter, Kaare Christensen, Gerard J van den Berg, et al. 2011.

“Being Born under Adverse Economic Conditions Leads to a Higher Cardiovascular

Mortality Rate Later in Life: Evidence Based on Individuals Born at Different Stages of the

Business Cycle.” Demography 48 (2): 507–30. https://doi.org/10.1007/s13524-011-0021-8.

Berg, Gerard J. Van Den, Gabriele Doblhammer, and Kaare Christensen. 2009. “Exogenous

Determinants of Early-Life Conditions, and Mortality Later in Life.” Social Science &

Medicine 68 (9): 1591–98. https://doi.org/10.1016/J.SOCSCIMED.2009.02.007.

Berg, Gerard J. Van Den, Sumedha Gupta, Gerard J van den Berg, and Sumedha Gupta. 2015.

“The Role of Marriage in the Causal Pathway from Economic Conditions Early in Life to

Mortality.” Journal of Health Economics 40:141–58.

https://doi.org/10.1016/j.jhealeco.2014.02.004.

Berg, Gerard J. Van Den, Maarten Lindeboom, France Portrait, Gerard J Van Den Berg, Maarten

Lindeboom, France Portrait, Gerard J den Berg, Maarten Lindeboom, and France Portrait.

2006. “Economic Conditions Early in Life and Individual Mortality.” American Economic

Review 96 (1): 290–302. https://doi.org/10.1257/000282806776157740.

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 122 (1): 409–39. https://doi.org/10.1162/qjec.122.1.409.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Blackwell, Debra L., Mark D. Hayward, and Eileen M. Crimmins. 2001. “Does Childhood

Health Affect Chronic Morbidity in Later Life?” Social Science & Medicine 52 (8): 1269–

84. https://doi.org/10.1016/S0277-9536(00)00230-6.

Bleakley, Hoyt. 2007. “Disease and Development: Evidence from Hookworm Eradication in the

American South.” The Quarterly Journal of Economics 122 (1): 73–117.

https://doi.org/10.1162/QJEC.121.1.73.

Borusyak, Kirill, Xavier Jaravel, and Jann Spiess. 2021. “Revisiting Event Study Designs:

Robust and Efficient Estimation,” August. https://arxiv.org/abs/2108.12419v1.

Bozzoli, Carlos, Angus Deaton, and Climent Quintana-Domeque. 2009. “Adult Height and

Childhood Disease.” Demography 46 (4): 647–69. https://doi.org/10.1353/DEM.0.0079.

Brainerd, Elizabeth, and Nidhiya Menon. 2014. “Seasonal Effects of Water Quality: The Hidden

Costs of the Green Revolution to Infant and Child Health in India.” Journal of Development

Economics 107 (March):49–64. https://doi.org/10.1016/J.JDEVECO.2013.11.004.

Breen, Casey F., and Maria Osborne. 2022. “An Assessment of CenSoc Match Quality,” June.

https://doi.org/10.31235/OSF.IO/BJ5MD.

Breen, Casey F., Maria Osborne, and Joshua R. Goldstein. 2023. “CenSoc: Public Linked

Administrative Mortality Records for Individual-Level Research.” Scientific Data 2023

10:1 10 (1): 1–12. https://doi.org/10.1038/s41597-023-02713-y.

Callaway, Brantly, and Pedro H.C. Sant’Anna. 2021. “Difference-in-Differences with Multiple

Time Periods.” Journal of Econometrics 225 (2): 200–230.

https://doi.org/10.1016/J.JECONOM.2020.12.001.

Case, Anne, Angela Fertig, and Christina Paxson. 2005. “The Lasting Impact of Childhood

Health and Circumstance.” Journal of Health Economics 24 (2): 365–89.

https://doi.org/10.1016/J.JHEALECO.2004.09.008.

Case, Anne, and Christina Paxson. 2009. “Early Life Health and Cognitive Function in Old

Age.” American Economic Review 99 (2): 104–9. https://doi.org/10.1257/AER.99.2.104.

CDC. 2015. “Vital Statistics of the US 1890-1938.” 2015.

https://www.cdc.gov/nchs/products/vsus/vsus_1890_1938.htm.

Chetty, Raj, Michael Stepner, Sarah Abraham, Shelby Lin, Benjamin Scuderi, Nicholas Turner,

Augustin Bergeron, and David Cutler. 2016. “The Association Between Income and Life

Expectancy in the United States, 2001-2014.” JAMA 315 (16): 1750–66.

https://doi.org/10.1001/JAMA.2016.4226.

Chou, Shin Yi, Michael Grossman, and Henry Saffer. 2006. “Reply to Jonathan Gruber and

Michael Frakes.” Journal of Health Economics 25 (2): 389–93.

https://doi.org/10.1016/J.JHEALECO.2005.12.004.

Clay, Karen, Werner Troesken, and Michael Haines. 2014. “Lead and Mortality.” The Review of

Economics and Statistics 96 (3): 458–70. https://doi.org/10.1162/REST_A_00396.

Condran, Gretchen A., and Eileen Crimmins-Gardner. 1978. “Public Health Measures and

Mortality in U.S. Cities in the Late Nineteenth Century.” Human Ecology 6 (1): 27–54.

https://doi.org/10.1007/BF00888565/METRICS.

Cook, C. Justin, Jason M. Fletcher, and Angela Forgues. 2019. “Multigenerational Effects of

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Early-Life Health Shocks.” Demography 56 (5): 1855–74. https://doi.org/10.1007/S13524-

019-00804-3.

Cook, C Justin, and Jason M Fletcher. 2015. “Understanding Heterogeneity in the Effects of

Birth Weight on Adult Cognition and Wages.” Journal of Health Economics 41:107–16.

https://doi.org/10.1016/j.jhealeco.2015.01.005.

Cormack, Louise, Volha Lazuka, and Luciana Quaranta. 2024. “Early-Life Disease Exposure

and Its Heterogeneous Effects on Mortality Throughout Life: Sweden, 1905–2016.”

Demography 61 (4): 1187–1210. https://doi.org/10.1215/00703370-11466677.

Costa, Dora L. 2015. “Health and the Economy in the United States from 1750 to the Present.”

Journal of Economic Literature. American Economic Association.

https://doi.org/10.1257/jel.53.3.503.

Crimmins, Eileen M., and Caleb E. Finch. 2006. “Infection, Inflammation, Height, and

Longevity.” Proceedings of the National Academy of Sciences of the United States of

America 103 (2): 498–503.

https://doi.org/10.1073/PNAS.0501470103/SUPPL_FILE/01470FIG4.PDF.

Cristia, Julian P. 2009. “Rising Mortality and Life Expectancy Differentials by Lifetime

Earnings in the United States.” Journal of Health Economics 28 (5): 984–95.

https://doi.org/10.1016/J.JHEALECO.2009.06.003.

Currie, Janet, Joshua Graff Zivin, Katherine Meckel, Matthew Neidell, and Wolfram Schlenker.

2013. “Something in the Water: Contaminated Drinking Water and Infant Health.”

Canadian Journal of Economics/Revue Canadienne d’économique 46 (3): 791–810.

https://doi.org/10.1111/CAJE.12039.

Currie, Janet, Michael Greenstone, and Katherine Meckel. 2017. “Hydraulic Fracturing and

Infant Health: New Evidence from Pennsylvania.” Science Advances 3 (12).

https://doi.org/10.1126/SCIADV.1603021/SUPPL_FILE/1603021_SM.PDF.

Currie, Janet, and Hannes Schwandt. 2013. “Within-Mother Analysis of Seasonal Patterns in

Health at Birth.” Proceedings of the National Academy of Sciences of the United States of

America 110 (30): 12265–70.

https://doi.org/10.1073/PNAS.1307582110/SUPPL_FILE/PNAS.201307582SI.PDF.

Cutler, David, Angus Deaton, and Adriana Lleras-Muney. 2006. “The Determinants of

Mortality.” Journal of Economic Perspectives 20 (3): 97–120.

https://doi.org/10.1257/JEP.20.3.97.

Cutler, David M., and Adriana Lleras-Muney. 2006. “Education and Health: Evaluating Theories

and Evidence.” National Bureau of Economic Research, July, 37.

https://doi.org/10.3386/W12352.

Cutler, David M, Grant Miller, and Douglas M Norton. 2007. “Evidence on Early-Life Income

and Late-Life Health from America’s Dust Bowl Era.” Proceedings of the National

Academy of Sciences 104 (33): 13244–49.

Cutler, David, and Grant Miller. 2005. “The Role of Public Health Improvements in Health

Advances: The Twentieth-Century United States.” Demography 2005 42:1 42 (1): 1–22.

https://doi.org/10.1353/DEM.2005.0002.

Deaton, Angus. 2007. “Height, Health, and Development.” Proceedings of the National

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Academy of Sciences 104 (33): 13232–37. https://doi.org/10.1073/PNAS.0611500104.

Deaton, Angus, and Raksha Arora. 2009. “Life at the Top: The Benefits of Height.” Economics

and Human Biology 7 (2): 133–36. https://doi.org/10.1016/j.ehb.2009.06.001.

Diener, Ed, and Micaela Y. Chan. 2011. “Happy People Live Longer: Subjective Well-Being

Contributes to Health and Longevity.” Applied Psychology: Health and Well-Being 3 (1):

1–43. https://doi.org/10.1111/J.1758-0854.2010.01045.X.

Doblhammer, G., and J. W. Vaupel. 2001. “Lifespan Depends on Month of Birth.” Proceedings

of the National Academy of Sciences of the United States of America 98 (5): 2934–39.

https://doi.org/10.1073/PNAS.041431898/SUPPL_FILE/4318FIG5.PDF.

Falagas, Matthew E., Drosos E. Karageorgopoulos, Lambros I. Moraitis, Evridiki K.

Vouloumanou, Nikos Roussos, George Peppas, and Petros I. Rafailidis. 2009. “Seasonality

of Mortality: The September Phenomenon in Mediterranean Countries.” Canadian Medical

Association Journal 181 (8): 484–86. https://doi.org/10.1503/CMAJ.090694.

Ferrie, Joseph P, and Werner Troesken. 2008. “Water and Chicago’s Mortality Transition, 1850-

-1925.” Explorations in Economic History 45 (1): 1–16.

Finch, Caleb E., and Eileen M. Crimmins. 2004. “Inflammatory Exposure and Historical

Changes in Human Life-Spans.” Science 305 (5691): 1736–39.

https://doi.org/10.1126/SCIENCE.1092556/ASSET/02F255E8-C8D1-46C1-B84F-

0E0CD47778A7/ASSETS/GRAPHIC/ZSE0360428410001.JPEG.

Fletcher, Jason M. 2018a. “Environmental Bottlenecks in Children’s Genetic Potential for Adult

Socio-Economic Attainments: Evidence from a Health Shock.” Population Studies 73 (1):

139–48. https://doi.org/10.1080/00324728.2018.1498533.

———. 2018b. “Examining the Long-Term Mortality Effects of Early Health Shocks.” Applied

Economics Letters 26 (11): 902–8. https://doi.org/10.1080/13504851.2018.1520960.

Fletcher, Jason M. 2011. “The Medium Term Schooling and Health Effects of Low Birth

Weight: Evidence from Siblings.” Economics of Education Review 30 (3): 517–27.

https://doi.org/10.1016/j.econedurev.2010.12.012.

———. 2015. “New Evidence of the Effects of Education on Health in the US: Compulsory

Schooling Laws Revisited.” Social Science & Medicine 127 (February):101–7.

https://doi.org/10.1016/J.SOCSCIMED.2014.09.052.

Fletcher, Jason, and Hamid Noghanibehambari. 2024. “The Siren Song of Cicadas: Early-Life

Pesticide Exposure and Later-Life Male Mortality.” Journal of Environmental Economics

and Management 123 (January):102903. https://doi.org/10.1016/J.JEEM.2023.102903.

Fletcher, Jason, Michael Topping, Fengyi Zheng, and Qiongshi Lu. 2021. “The Effects of

Education on Cognition in Older Age: Evidence from Genotyped Siblings.” Social Science

& Medicine 280 (July):114044. https://doi.org/10.1016/J.SOCSCIMED.2021.114044.

Friedrich, M. 1912. “The Mills--Reincke Phenomenon.” Ohio State Medical Journal 20:514–17.

Gardner, Jonathan, and Andrew Oswald. 2004. “How Is Mortality Affected by Money, Marriage,

and Stress?” Journal of Health Economics 23 (6): 1181–1207.

https://doi.org/10.1016/J.JHEALECO.2004.03.002.

Goldstein, Joshua R, Monica Alexander, Casey Breen, Andrea Miranda González, Felipe

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Menares, Maria Osborne, Mallika Snyder, and Ugur Yildirim. 2021. “Censoc Project.”

CenSoc Mortality File: Version 2.0. Berkeley: University of California.

https://censoc.berkeley.edu/data/.

Goodman-Bacon, Andrew. 2021. “Difference-in-Differences with Variation in Treatment

Timing.” Journal of Econometrics, June. https://doi.org/10.1016/J.JECONOM.2021.03.014.

Greenstone, Michael, and Rema Hanna. 2014. “Environmental Regulations, Air and Water

Pollution, and Infant Mortality in India.” American Economic Review. American Economic

Association. https://doi.org/10.1257/aer.104.10.3038.

Grossman, Daniel S., and David J.G. Slusky. 2019. “The Impact of the Flint Water Crisis on

Fertility.” Demography 56 (6): 2005–31. https://doi.org/10.1007/S13524-019-00831-0.

Gruber, Jonathan, and Michael Frakes. 2006. “Does Falling Smoking Lead to Rising Obesity?”

Journal of Health Economics 25 (2): 183–97.

https://doi.org/10.1016/J.JHEALECO.2005.07.005.

Hafeman, Danella, Pam Factor-Litvak, Zhonggi Cheng, Alexander van Geen, and Habibul

Ahsan. 2007. “Association between Manganese Exposure through Drinking Water and

Infant Mortality in Bangladesh.” Environmental Health Perspectives 115 (7): 1107–12.

https://doi.org/10.1289/EHP.10051.

Halpern-Manners, Andrew, Jonas Helgertz, John Robert Warren, and Evan Roberts. 2020. “The

Effects of Education on Mortality: Evidence From Linked U.S. Census and Administrative

Mortality Data.” Demography 57 (4): 1513–41. https://doi.org/10.1007/S13524-020-00892-

Hayward, Mark D, and Bridget K Gorman. 2004. “The Long Arm of Childhood: The Influence

of Early-Life Social Conditions on Men’s Mortality.” Demography 2004 41:1 41 (1): 87–

107. https://doi.org/10.1353/DEM.2004.0005.

He, Guojun, and Jeffrey M Perloff. 2016. “Surface Water Quality and Infant Mortality in China.”

Economic Development and Cultural Change 65 (1): 119–39.

https://doi.org/10.1086/687603.

Hill, Elaine L., and Lala Ma. 2022. “Drinking Water, Fracking, and Infant Health.” Journal of

Health Economics 82 (March):102595. https://doi.org/10.1016/J.JHEALECO.2022.102595.

Hill, Elaine L. 2018. “Shale Gas Development and Infant Health: Evidence from Pennsylvania.”

Journal of Health Economics 61:134–50. https://doi.org/10.1016/j.jhealeco.2018.07.004.

Huebener, Mathias. 2019. “Life Expectancy and Parental Education.” Social Science & Medicine

232 (July):351–65. https://doi.org/10.1016/J.SOCSCIMED.2019.04.034.

———. 2020. “Parental Education and Children’s Health throughout Life.” The Economics of

Education: A Comprehensive Overview, January, 91–102. https://doi.org/10.1016/B978-0-

12-815391-8.00007-0.

Jones, Benjamin A. 2019. “Infant Health Impacts of Freshwater Algal Blooms: Evidence from an

Invasive Species Natural Experiment.” Journal of Environmental Economics and

Management 96 (July):36–59. https://doi.org/10.1016/J.JEEM.2019.05.002.

Jousilahti, Pekka, Jaakko Tuomilehto, Erkki Vartiainen, Johan Eriksson, and Pekka Puska. 2000.

“Relation of Adult Height to Cause-Specific and Total Mortality: A Prospective Follow-up

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Study of 31, 199 Middle-Aged Men and Women in Finland.” American Journal of

Epidemiology 151 (11): 1112–20.

https://doi.org/10.1093/OXFORDJOURNALS.AJE.A010155.

Kinge, Jonas Minet, Jørgen Heibø Modalsli, Simon Øverland, Håkon Kristian Gjessing, Mette

Christophersen Tollånes, Ann Kristin Knudsen, Vegard Skirbekk, Bjørn Heine Strand, Siri

Eldevik Håberg, and Stein Emil Vollset. 2019. “Association of Household Income With

Life Expectancy and Cause-Specific Mortality in Norway, 2005-2015.” JAMA 321 (19):

1916–25. https://doi.org/10.1001/JAMA.2019.4329.

Kniesner, Thomas J., and W. Kip Viscusi. 2019. “The Value of a Statistical Life.” Oxford

Research Encyclopedia of Economics and Finance, July.

https://doi.org/10.1093/ACREFORE/9780190625979.013.138.

Kose, Esra, Elira Kuka, and Na’ama Shenhav. 2021. “Women’s Suffrage and Children’s

Education.” American Economic Journal: Economic Policy 13 (3): 374–405.

https://doi.org/10.1257/POL.20180677.

Kunitz, Stephan J. 1984. “Mortality Change in America, 1620-1920.” Human Biology, 559–82.

Lee, Jin Young, and Gary Solon. 2011. “The Fragility of Estimated Effects of Unilateral Divorce

Laws on Divorce Rates.” B.E. Journal of Economic Analysis and Policy 11 (1).

https://doi.org/10.2202/1935-1682.2994/MACHINEREADABLECITATION/RIS.

Lindeboom, Maarten, France Portrait, and Gerard J. Van Den Berg. 2010. “Long-Run Effects on

Longevity of a Nutritional Shock Early in Life: The Dutch Potato Famine of 1846–1847.”

Journal of Health Economics 29 (5): 617–29.

https://doi.org/10.1016/J.JHEALECO.2010.06.001.

Lleras-Muney, Adriana. 2005. “The Relationship Between Education and Adult Mortality in the

United States.” The Review of Economic Studies 72 (1): 189–221.

https://doi.org/10.1111/0034-6527.00329.

———. 2022. “Education and Income Gradients in Longevity: The Role of Policy.” Canadian

Journal of Economics/Revue Canadienne d’économique 55 (1): 5–37.

https://doi.org/10.1111/CAJE.12582.

Lleras-Muney, Adriana, and Flavien Moreau. 2022. “A Unified Model of Cohort Mortality.”

Demography 59 (6): 2109–34. https://doi.org/10.1215/00703370-10286336.

Lleras-Muney, Adriana, Joseph Price, and Dahai Yue. 2022. “The Association between

Educational Attainment and Longevity Using Individual-Level Data from the 1940

Census.” Journal of Health Economics 84 (July):102649.

https://doi.org/10.1016/J.JHEALECO.2022.102649.

Lundborg, Petter, Anton Nilsson, and Dan-Olof Rooth. 2014. “Parental Education and Offspring

Outcomes: Evidence from the Swedish Compulsory School Reform.” American Economic

Journal: Applied Economics 6 (1): 253–78. https://doi.org/10.1257/APP.6.1.253.

Maruyama, Shiko, and Eskil Heinesen. 2020. “Another Look at Returns to Birthweight.” Journal

of Health Economics 70 (March):102269. https://doi.org/10.1016/j.jhealeco.2019.102269.

McGee, Harold G. 1920. “Mills-Reincke Phenomenon and Typhoid Control by Vaccine.”

American Journal of Public Health 10 (7): 585–87.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Meer, Jonathan, and Jeremy West. 2016. “Effects of the Minimum Wage on Employment

Dynamics.” Journal of Human Resources 51 (2): 500–522.

https://doi.org/10.3368/JHR.51.2.0414-6298R1.

Mettetal, Elizabeth. 2019. “Irrigation Dams, Water and Infant Mortality: Evidence from South

Africa.” Journal of Development Economics 138 (May):17–40.

https://doi.org/10.1016/J.JDEVECO.2018.11.002.

Montez, Jennifer, and Mark D Hayward. 2014. “Cumulative Childhood Adversity, Educational

Attainment, and Active Life Expectancy Among U.S. Adults.” Demography 51 (2): 413–35.

https://doi.org/10.1007/S13524-013-0261-X.

Moore, Sophie E., Andrew C. Collinson, Pa Tamba N’Gom, Richard Aspinall, and Andrew M.

Prentice. 2006. “Early Immunological Development and Mortality from Infectious Disease

in Later Life.” Proceedings of the Nutrition Society 65 (3): 311–18.

https://doi.org/10.1079/PNS2006503.

Mueller, J. Tom, and Stephen Gasteyer. 2021. “The Widespread and Unjust Drinking Water and

Clean Water Crisis in the United States.” Nature Communications 2021 12:1 12 (1): 1–8.

https://doi.org/10.1038/s41467-021-23898-z.

Myrskylä, Mikko, Neil K. Mehta, and Virginia W. Chang. 2013. “Early Life Exposure to the

1918 Influenza Pandemic and Old-Age Mortality by Cause of Death.” American Journal of

Public Health 103 (7). https://doi.org/10.2105/AJPH.2012.301060.

Nadimpalli, Maya L., Val F. Lanza, Maria Camila Montealegre, Sonia Sultana, Erica R.

Fuhrmeister, Colin J. Worby, Lisa Teichmann, et al. 2022. “Drinking Water Chlorination

Has Minor Effects on the Intestinal Flora and Resistomes of Bangladeshi Children.” Nature

Microbiology 2022 7:5 7 (5): 620–29. https://doi.org/10.1038/s41564-022-01101-3.

Neumark, David, J. M.Ian Salas, and William Wascher. 2014. “Revisiting the Minimum Wage—

Employment Debate: Throwing Out the Baby with the Bathwater?:” ILR Review 67

(SUPPL): 608–48. https://doi.org/10.1177/00197939140670S307.

Noghanibehambari, Hamid, and Michal Engelman. 2022. “Social Insurance Programs and Later-

Life Mortality: Evidence from New Deal Relief Spending.” Journal of Health Economics

86 (December). https://doi.org/10.1016/J.JHEALECO.2022.102690.

Noghanibehambari, Hamid, and Jason Fletcher. 2023a. “Childhood Exposure to Birth

Registration Laws and Old-Age Mortality.” Health Economics 32 (3): 735–43.

https://doi.org/10.1002/HEC.4643.

———. 2023b. “In Utero and Childhood Exposure to Alcohol and Old Age Mortality: Evidence

from the Temperance Movement in the US.” Economics & Human Biology 50

(August):101276. https://doi.org/10.1016/J.EHB.2023.101276.

———. 2023c. “Long-Term Health Benefits of Occupational Licensing: Evidence from

Midwifery Laws.” Journal of Health Economics 92 (December):102807.

https://doi.org/10.1016/J.JHEALECO.2023.102807.

Noghanibehambari, Hamid, and Jason M. Fletcher. 2023d. “Dust to Feed, Dust to Grey: The

Effect of In-Utero Exposure to the Dust Bowl on Old-Age Longevity.” Demography,

October. https://doi.org/10.3386/W30531.

Noghanibehambari, Hamid, Jason Fletcher, Lauren Schmitz, Valentina Duque, and Vikas Gawai.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

2024. “Early-Life Economic Conditions and Old-Age Male Mortality: Evidence from

Historical County-Level Bank Deposit Data.” Journal of Population Economics 37 (1): 1–

33. https://doi.org/10.1007/S00148-024-01007-W/TABLES/7.

Noghanibehambari, Hamid, and Farzaneh Noghani. 2023. “Long-Run Intergenerational Health

Benefits of Women Empowerment: Evidence from Suffrage Movements in the US.” Health

Economics, July. https://doi.org/10.1002/HEC.4744.

Noghanibehambari, Hamid, Mahmoud Salari, and Nahid Tavassoli. 2022. “Maternal Human

Capital and Infants’ Health Outcomes: Evidence from Minimum Dropout Age Policies in

the US.” SSM - Population Health 19 (September):101163.

https://doi.org/10.1016/J.SSMPH.2022.101163.

Palloni, Alberto, and Hantamala Rafalimanana. 1999. “The Effects of Infant Mortality on

Fertility Revisited: New Evidence from Latin America.” Demography 1999 36:1 36 (1):

41–58. https://doi.org/10.2307/2648133.

Potter, Guy G., Michael J. Helms, and Brenda L. Plassman. 2008. “Associations of Job Demands

and Intelligence with Cognitive Performance among Men in Late Life.” Neurology 70 (19

PART 2): 1803–8.

https://doi.org/10.1212/01.WNL.0000295506.58497.7E/ASSET/621CF9B5-EF81-4C8C-

8B52-D156DD04EBBD/ASSETS/GRAPHIC/12FSM1.GIF.

Preston, Samuel H. 2005. “Deadweight? — The Influence of Obesity on Longevity.” The New

England Journal of Medicine 352 (11): 1135–37. https://doi.org/10.1056/NEJME058009.

Royer, Heather. 2009. “Separated at Girth: US Twin Estimates of the Effects of Birth Weight.”

American Economic Journal: Applied Economics 1 (1): 49–85.

https://doi.org/10.1257/app.1.1.49.

Ruggles, Steven, Sarah Flood, Ronald Goeken, Josiah Grover, and Erin Meyer. 2020. “IPUMS

USA: Version 10.0 [Dataset].” Minneapolis, MN: IPUMS.

https://doi.org/10.18128/D010.V10.0.

Sandberg, John. 2016. “Infant Mortality, Social Networks, and Subsequent Fertility:” American

Sociological Review 71 (2): 288–309. https://doi.org/10.1177/000312240607100206.

Scholte, Robert S., Gerard J. Van Den Berg, and Maarten Lindeboom. 2015. “Long-Run Effects

of Gestation during the Dutch Hunger Winter Famine on Labor Market and Hospitalization

Outcomes.” Journal of Health Economics 39 (January):17–30.

https://doi.org/10.1016/J.JHEALECO.2014.10.002.

Seretakis, Dimitrios, Pagona Lagiou, Loren Lipworth, Lisa B. Signorello, Kenneth J. Rothman,

and Dimitrios Trichopoulos. 1997. “Changing Seasonality of Mortality From Coronary

Heart Disease.” JAMA 278 (12): 1012–14.

https://doi.org/10.1001/JAMA.1997.03550120072036.

Shenkin, Susan D., Ian J. Deary, and John M. Starr. 2009. “Birth Parameters and Cognitive

Ability in Older Age: A Follow-Up Study of People Born 1921–1926.” Gerontology 55 (1):

92–98. https://doi.org/10.1159/000163444.

Smith, Allan H., Guillermo Marshall, Jane Liaw, Yan Yuan, Catterina Ferreccio, and Craig

Steinmaus. 2012. “Mortality in Young Adults Following in Utero and Childhood Exposure

to Arsenic in Drinking Water.” Environmental Health Perspectives 120 (11): 1527–31.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

https://doi.org/10.1289/EHP.1104867.

Smith, Allan H., Guillermo Marshall, Yan Yuan, Catterina Ferreccio, Jane Liaw, Ondine von

Ehrenstein, Craig Steinmaus, Michael N. Bates, and Steve Selvin. 2006. “Increased

Mortality from Lung Cancer and Bronchiectasis in Young Adults after Exposure to Arsenic

in Utero and in Early Childhood.” Environmental Health Perspectives 114 (8): 1293–96.

https://doi.org/10.1289/EHP.8832.

Smith, David W., and Benjamin S. Bradshaw. 2006. “Variation in Life Expectancy during the

Twentieth Century in The United States.” Demography 2006 43:4 43 (4): 647–57.

https://doi.org/10.1353/DEM.2006.0039.

Smith, James P. 2009. “The Impact of Childhood Health on Adult Labor Market Outcomes.” The

Review of Economics and Statistics 91 (3): 478–89. https://doi.org/10.1162/REST.91.3.478.

Spijker, Jeroen J.A., Antonio D. Cámara, and Amand Blanes. 2012. “The Health Transition and

Biological Living Standards: Adult Height and Mortality in 20th-Century Spain.”

Economics & Human Biology 10 (3): 276–88. https://doi.org/10.1016/J.EHB.2011.08.001.

SSA. 2020. “Social Security Program Data.” 2020.

https://www.ssa.gov/oact/HistEst/CohLifeTables/2020/CohLifeTables2020.html.

Sun, Liyang, and Sarah Abraham. 2021. “Estimating Dynamic Treatment Effects in Event

Studies with Heterogeneous Treatment Effects.” Journal of Econometrics 225 (2): 175–99.

https://doi.org/10.1016/J.JECONOM.2020.09.006.

Troesken, Werner. 2004. Race, Water, and Disease. Cambridge: MIT Press.

UNICEF. 2021. “Progress on Household Drinking Water, Sanitation and Hygiene, 2000-2020:

Five Years into the SDGs - UNICEF DATA.” https://data.unicef.org/resources/progress-on-

household-drinking-water-sanitation-and-hygiene-2000-2020/.

Venkataramani, Atheendar S. 2012. “Early Life Exposure to Malaria and Cognition in

Adulthood: Evidence from Mexico.” Journal of Health Economics 31 (5): 767–80.

https://doi.org/10.1016/J.JHEALECO.2012.06.003.

Viscusi, W. Kip. 2018. “Best Estimate Selection Bias in the Value of a Statistical Life.” Journal

of Benefit-Cost Analysis 9 (2): 205–46. https://doi.org/10.1017/BCA.2017.21.

WHO. 2019. Progress on Household Drinking Water, Sanitation and Hygiene 2000-2017:

Special Focus on Inequalities. World Health Organization.

Wilson, Sven E. 2019. “Does Adult Height Predict Later Mortality?: Comparative Evidence

from the Early Indicators Samples in the United States.” Economics & Human Biology 34

(August):274–85. https://doi.org/10.1016/J.EHB.2019.05.004.

Zaveri, Esha, Jason Russ, Sebastien Desbureaux, Richard Damania, Aude-Sophie Rodella, and

Giovanna Ribeiro. 2019. “The Nitrogen Legacy : The Long-Term Effects of Water

Pollution on Human Capital.” The Nitrogen Legacy, December.

https://doi.org/10.1596/33073.

Zhang, Jing, and Lixin Colin Xu. 2016. “The Long-Run Effects of Treated Water on Education:

The Rural Drinking Water Program in China.” Journal of Development Economics 122

(September):1–15. https://doi.org/10.1016/J.JDEVECO.2016.04.004.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Tables

Table 1 - Summary Statistics

Variable

Mean

Std. Dev.

Min

Max

Death Age (Months)

867.238

123.454

420

1249

Birth Year

1919.480

8.438

1900

1940

Death Year

1991.749

8.619

1975

2005

White

.975

.154

Black

.024

.154

Hispanic

.008

.089

Mother Literate

.909

.287

Mother Literacy Missing

.017

.129

Mother in Labor Force

.058

.235

Father Literate

.900

.299

Father Literacy Missing

.037

.189

Father in Labor Force

.945

.226

Father Occupational Income Score

28.407

9.174

Father Occupational Income Score

Missing

.084

.278

Water Filtration

.156

.362

Chlorination of Water

.753

.422

Bacteriological Standard for Milk

.767

.415

Sewage Treatment or Diversion

.333

.468

Observations

338,758

Census 1960-1980 Data:

Year of Schooling

10.693

3.684

Education less than High School

.118

.323

Education < 12

.501

.500

Socioeconomic Index

38.282

23.373

Total Family Income

72303.89

45351.124

43.718

874361.5

Ln Total Family Income

11.001

.668

3.777

13.681

Year of birth

1920.487

10.869

1900

1940

White

.833

.372

Black

.163

.369

Hispanic

.009

.095

Observations

362,167

WWII Enlistment Data:

Army General Classification Test

(AGCT) score

140.517

12.919

160

Log AGCT

4.940

.112

.693

5.075

Height

2.722

.414

.034

4.5

Log Height

.988

.171

-3.367

1.504

Nonwhite

.019

.139

Observations

33,139

Birth/Death Data:

Infant Mortality Rate per 100,000

66.344

21.642

28.059

148.979

Log Infant Mortality Rate per

4.141

.328

3.334

5.003

Birth per 1,000 Women

37.875

8.238

22.906

69.283

Log Birth Rate

3.639

.204

3.174

4.252

Observations

559

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Table 2 - Main Results: Childhood Exposure to Water Filtration and Later Life Longevity

Outcomes: Age at Death (Months)

(1)

(2)

(3)

Exposure to Water

Filtration

3.68*

3.23**

3.24**

(1.76)

(1.51)

(1.11)

Observations

338758

338742

R-squared

.35

Mean DV

867.23

867.24

P-Value

0.07

0.05

0.03

City FE



Birth Year FE



Individual Controls



Family Controls



City Controls



Region-by-Cohort FE



Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering. Individual controls include dummies for race and ethnicity. Family controls

include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor force

status dummy, paternal socioeconomic score dummies, and a series of missing indicators for missing values of each

variable. City controls include average share of homeowners, average occupational income score, share of white-

collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Table 3 - Exploring Endogenous Merging

Outcome: Successful Merging between the Final Sample and the Original

Population in 1940

Full Sample

Whites

Nonwhites

(1)

(2)

(3)

Exposure to Water

Filtration

-.0002

.0001

.0003

(.0077)

(.0078)

(.0082)

Observations

7218487

6844592

373895

R-squared

.0057

.0058

.004

Mean DV

0.033

0.034

0.015

P-Value

0.978

0.987

0.966

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering. Regressions include city and region-by-year fixed effects. Regressions also

include city covariates. City controls include average share of homeowners, average occupational income score,

share of white-collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Table 4 - Robustness Checks

Outcome: Age at Death (Months)

Column 3 of Table 2

Adding Birth-State by

1940-State FE

Adding City-by-

Individual Covariates

Interaction

Adding City-by-

Parental Covariates

Interaction

Adding Birth-Month

and Death-Month FE

Adding

more City

Controls

(1)

(2)

(3)

(4)

(5)

(6)

Exposure to

Water

Filtration

3.24**

3.23**

3.20**

3.62**

3.25***

2.90*

(1.11)

(1.14)

(1.09)

(1.25)

(1.10)

(1.31)

Observations

338758

310097

338758

306759

R-squared

.35

.35625

.35655

.35687

.22901

Mean DV

867.23

867.17

867.23

883.28

P-Value

0.02

0.03

0.01

0.09

Using

Heteroscedasticity

Robust SE

Outcome: Log Age at

Death

Outcome: Age at

Death>75

Weighted by City

Population

Dropping Cohorts of

1918-1919

Dropping

Cohorts of

1930-1940

(7)

(8)

(9)

(10)

(11)

(12)

Exposure to

Water

Filtration

3.24**

.003***

.014**

3.85***

3.46***

3.10**

(1.55)

(.001)

(.006)

(.75)

(1.06)

(1.19)

Observations

338758

307954

306185

R-squared

.35

.36

.18

.38

.37

.22

Mean DV

867.23

6.75

0.38

848.53

865.46

884.48

P-Value

0.03

0.01

0.05

0.006

0.01

0.05

Notes. Standard errors, clustered at the city level (except column 7), are in parentheses. P-values are extracted from the wild bootstrap procedure with city-

level clustering. All regressions include city and birth-region-by-birth-year fixed effects. Individual controls include dummies for race and ethnicity. Family

controls include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor force status dummy, paternal

socioeconomic score dummies, and a series of missing indicators for missing values of each variable. City controls include average share of homeowners,

average occupational income score, share of white-collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Table 5 - Exploring Mechanism Channels

Outcomes and Samples:

1950-70 Census Data

DMF-Enlistment Data

Years of

Schooling

Education

 High

School

Education

 12 years

Socioeconomic

Score

Total

Family

Income

Log Total

Family

Income

Log AGCT

Log Height

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Exposure to

Water

Filtration

.29***

-.037**

-.026

1.84***

2603.0

.03

.014

.016

(.10)

(.013)

(.018)

(.65)

(1580.3)

(.02)

(.008)

Observations

245879

234773

235625

233920

33139

R-squared

.27

.20

.11

.12

.09

.12

.02

.85

Mean DV

10.41

0.16

0.56

38.03

69308.46

10.97

4.94

0.98

P-Value

0.01

0.04

0.19

0.007

0.15

0.13

0.17

0.15

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap procedure with city-level

clustering. All regressions include city fixed effects, birth-region-by-birth-year fixed effects, and city covariates. Regressions of columns 1-6

also include census year fixed effects. Individual controls include dummies for race and ethnicity. City controls include average share of

homeowners, average occupational income score, share of white-collar occupation, share of farmers, share of other occupation, literacy rate,

and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American Journal of Health Economics, published by The

University of Chicago Press on behalf of the American Society of Health Economics.

Figures

Figure 1 - Water Filtration Year across Cities in the Final Sample

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-level

clustering, are reported. All regressions include city and birth-region-by-birth-year fixed effects. Individual

controls include dummies for race and ethnicity. Family controls include maternal literacy dummy, paternal literacy

dummy, maternal labor force status dummy, paternal labor force status dummy, paternal socioeconomic score

dummies, and a series of missing indicator for missing values of each variable. City controls include average share

of homeowners, average occupational income score, share of white-collar occupation, share of farmers, share of

other occupation, literacy rate, and share of married.

Figure 2 - Exposure to Water Filtration across Different Ages and Later-Life Longevity

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-level

clustering, are reported. All regressions include city and birth-region-by-birth-year fixed effects.

Figure 3 - Exposure to Water Filtration across Different Ages and Observable Individual/Family

Characteristics

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-level

clustering, are reported. All regressions include city and birth-region-by-birth-year fixed effects.

Figure 4 - Exposure to Water Filtration across Different Ages and Observable Individual/Family

Characteristics

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-level

clustering, are reported. All regressions include city and birth-region-by-birth-year fixed effects.

Figure 5 - Exposure to Water Filtration across Different Ages and Observable Individual/Family

Characteristics

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix

Table of Contents

Appendix A ........................................................................................................................... 48

Appendix B ........................................................................................................................... 49

Appendix C ........................................................................................................................... 51

Appendix D ........................................................................................................................... 61

Appendix E ........................................................................................................................... 63

Appendix F............................................................................................................................ 65

Appendix G ........................................................................................................................... 67

Appendix H ........................................................................................................................... 69

Appendix I ............................................................................................................................ 71

Appendix J ............................................................................................................................ 73

Appendix K ........................................................................................................................... 80

Appendix L ........................................................................................................................... 83

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix A

Appendix Table A-1 - List of Cities and Year of Water Filtration

City

Year of Water Filtration

Providence, RI

1904

Indianapolis, IN

1904

Washington, DC

1905

Philadelphia, PA

1906

Cincinnati, OH

1907

Pittsburgh, PA

1908

Louisville, KY

1909

New Orleans, LA

1909

Minneapolis, MN

1913

Baltimore, MD

1915

St. Louis, MO

1915

Cleveland, OH

1918

St. Paul, MN

1923

Detroit, MI

1923

Buffalo, NY

1926

Kansas City, MO

1928

Milwaukee, WI

1939

Rochester, NY

Post-1940

Memphis, TN

Post-1940

Chicago, IL

Post-1940

San Francisco, CA

Post-1940

Boston, MA

Post-1940

Newark, NJ

Post-1940

New York, NY

Post-1940

Jersey City, NJ

Post-1940

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix B

In the main results, we evaluated the effects of the share of childhood exposure up to age

15. Recall that we limit the sample to those born 15 years pre- and post-waterwork. Therefore, our

childhood ages end at 15 years. In Appendix Table B-1, we show the effects of the share of

exposure between birth and age 𝑧, where 𝑧 ∈{1,5,10,14}. We observe increases in magnitude as

we include more childhood ages. This is expected as all ages are treated although the effects are

more pronounced for earlier ages of life.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table B-1 - Exploring the Robustness across Age Thresholds for Childhood Exposure

Outcomes: Age at Death (Months)

𝑍=1 Year

𝑍=5 Years

𝑍=10 Years

𝑍=14 Years

(1)

(2)

(3)

(4)

Share of Childhood

Up to Age 𝑍

Exposed to Water

Filtration

1.87**

2.35**

3.16***

3.45***

(.53)

(.68)

(.61)

(.75)

Observations

338742

R-squared

.38

Mean DV

848.53

P-Value

0.04

0.009

0.008

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap procedure with city-level clustering. All regressions include city

fixed effects, birth-year-by-birth-region fixed effects, individual controls, family controls, and city-level covariates. Individual controls include dummies for race and

ethnicity. Family controls include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor force status dummy, paternal

socioeconomic score dummies, and a series of missing indicators for missing values of each variable. City controls include average share of homeowners, average

occupational income score, share of white-collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

University of Chicago Press on behalf of the American Society of Health Economics.

Appendix C

Ambitious public projects such as establishing water filtration stations could mirror

changes in political environment and economic conditions and may be followed by changes in

other public expenditures. For instance, public legislatures could allocate funds for public

education in addition to establishing water treatment facilities. In that case, the results might partly

incorporate the influences of improvements in public education, considering the literature that

documents the education longevity relationship. Similarly, public authorities might consider public

health interventions as substitutes and reallocate funds from other health expenditures toward

improving water quality.

For a subset of the sample years, we have information on state-level education expenditure

per capita and the number of doctors per capita (both extracted from Kose et al. (2021)). We then

examine changes in these outcomes in different years relative to the city-specific year of water

filtration, conditional on fixed effects and covariates. To ease interpretation and cross-panel

comparison, we standardize these outcomes with respect to their mean and standard deviation over

the sample period. These results are reported in the two panels of Appendix Figure C-1. The results

do not reveal any significant changes in many years prior to and after water filtration.

Another concern is that the public health interventions related to water quality were

accompanied by other interventions, such as chlorination of water and sewage treatment. In

Appendix Figure C-2 through Appendix Figure C-6, we show the year of different public health

interventions relative to the year of water filtration across cities in the final sample. In most cases,

water filtration occurs after water chlorination. Therefore, one argument is that the positive impacts

we observe in the paper are due to the combined benefits of filtration and chlorination. However,

we do not observe a significant association of water filtration status with chlorination of water

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

once we examine their correlation using an event study framework with city and region-by-year

fixed effects (top left panel of Appendix Figure C-7). This is also true for sewage treatment and

implementation of bacteriological standards for milk (top right and bottom left panels of Appendix

Figure C-7). We do observe a lower probability of any major water project after water filtration

(bottom right panel of Appendix Figure C-7).

Further, we examine the influence of these other public health interventions on longevity.

In so doing, we generate dummies that equal one if a city has initiated any of these interventions.

We then merge with the final sample based on year and city of birth and implement regressions

similar to equation 1. The results are reported in columns 1-3 of Appendix Table C-1. We observe

a 0.9-month rise in longevity due to early-life exposure to water chlorination. However, the

estimated effect is insignificant at 10 percent level (column 1). For the sewage treatment, we

observe a significant increase in longevity of about 2 months (column 2). However, when we

include these interventions in the presence of water filtration (column 4), we observe a larger

coefficient for water filtration. The respective coefficient of other interventions becomes

statistically insignificant, suggesting that the main benefits of waterworks arise from water

filtration. We should note that previous studies suggest that among several public health

interventions during the early 20th century in the US, water filtration was the most successful, with

significant health benefits (Anderson, Charles, & Rees, 2022; Costa, 2015; Cutler & Miller, 2005).

Despite the evidence in column 4, we should acknowledge that our sample covers only 25 cities

in a specific timeframe in the US. The US currently has 150,000 public water systems. Therefore,

our sample may not fully reveal the benefits of other interventions including chlorination of water.

Specifically, other interventions such as chlorination of water have been documented to be quite

beneficial for health outcomes in other settings (Nadimpalli et al. 2022).

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from

wild bootstrap procedure with city-level clustering, are reported. Regressions

include city and region-by-year fixed effects. Regressions are weighted

using city-level population.

Appendix Figure C-1 - Event-Study Tests to Examine the Evolution of

Public Expenditure pre/post Water Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Figure C-2 - The Evolution of Water Filtration along with Other Public Health Interventions in the

Cities in the Final Sample

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Figure C-3 - The Evolution of Water Filtration along with Other Public Health Interventions in the

Cities in the Final Sample

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Figure C-4 - The Evolution of Water Filtration along with Other Public Health Interventions in the

Cities in the Final Sample

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Figure C-5 - The Evolution of Water Filtration along with Other Public Health Interventions in the

Cities in the Final Sample

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Figure C-6 - The Evolution of Water Filtration along with Other Public Health Interventions in the

Cities in the Final Sample

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-

level clustering, are reported. Regressions include city and region-by-year fixed effects. Regressions are

weighted using city-level population.

Appendix Figure C-2 - Event-Study Tests to Examine the Evolution of City-Level Public Health Interventions

pre/post Water Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table C-1 - Examining Other Public Health Interventions

Outcome: Age at Death (Months)

(1)

(2)

(3)

(4)

Water Filtration

2.57

(1.22)

[0.15]

Chlorination of

Water

.85

.23

(.74)

(.72)

[0.31]

[0.77]

Sewage Treatment

or Diversion

2.02

1.69

(1.10)

(1.22)

[0.14]

[0.31]

Bacteriological

Standard for Milk

.74

.56

(.68)

[0.34]

[0.44]

Observations

338742

R-squared

.35

Mean DV

867.24

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering and are reported in brackets. All regressions include city fixed effects, birth-

year-by-birth-region fixed effects, individual controls, family controls, and city-level covariates. Individual controls

include dummies for race and ethnicity. Family controls include maternal literacy dummy, paternal literacy dummy,

maternal labor force status dummy, paternal labor force status dummy, paternal socioeconomic score dummies, and

a series of missing indicators for missing values of each variable. City controls include average share of

homeowners, average occupational income score, share of white-collar occupation, share of farmers, share of other

occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix D

The final sample of the paper, constructed from the DMF data, exclusively focused on male

individuals. In this appendix, we employ the Berkeley Unified Numident Mortality Database

(BUNMD) from the Censoc project to examine the effects across both genders. The advantage of

the BUNMD data is that it covers both genders. Although this data covers a small portion of pre-

1975 deaths and ends in 2007 (hence more death years compared with the DMF 1975-2005), the

death coverage is relatively thin and unreliable for the years prior to 1988. Moreover, the data is

not linked to the 1940 census. On the other hand, the BUNMD data reports county/city-of-birth

directly and relieves us from the measurement errors caused by cross-census linking. We replicate

the main results using BUNMD data and report them in Appendix Table D-1. In column 1, we

observe an insignificant increase in longevity of about 1.6 months. When we focus on male

individuals in column 2, we observe a significant change of about 3.1 months, an effect size that

is quite comparable to the main results of the paper. For the female subsample in column 3, we

observe a relatively small and insignificant coefficient. Therefore, we argue that the main benefits

appear to be for male individuals only.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table D-1 - Replicating the Main Results Using BUNMD Data

Outcomes: Age at Death (Months)

Full Sample

Males

Females

(1)

(2)

(3)

Exposure to Water

Filtration

1.61

3.05

.68

(1.47)

(1.65)

(1.71)

Observations

3480529

1702463

1777755

R-squared

.40

.32

.44

Mean DV

928.75

906.001

950.55

P-Value

0.35

0.18

0.74

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering. All regressions include city fixed effects, birth-year-by-birth-region fixed

effects, individual controls, family controls, and city-level covariates. Individual controls include dummies for race

and ethnicity. Family controls include maternal literacy dummy, paternal literacy dummy, maternal labor force

status dummy, paternal labor force status dummy, paternal socioeconomic score dummies, and a series of missing

indicators for missing values of each variable. City controls include average share of homeowners, average

occupational income score, share of white-collar occupation, share of farmers, share of other occupation, literacy

rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix E

In the main results of the paper, we restrict the sample to cohorts that were born 15 years

before and 15 years after city-specific water filtration year. In panel A of Appendix Table E-1, we

remove this restriction and replicate the main results. The effect size of column 3 remains quite

comparable to the main results of Table 2. In panel B, we make the stricter balancing window

restriction, i.e., restricting to cohorts born 12 years before and after city-specific water filtration

year. The fully parameterized regression of column 3 suggests a slightly larger effect size.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table E-1 - Sensitivity of the Results to the Balancing Window Restriction

Outcomes: Age at Death (Months)

(1)

(2)

(3)

Panel A. No Balancing Window Restriction

Exposure to Water

Filtration

2.40

2.35

2.92***

(1.45)

(1.34)

(.82)

Observations

396339

396330

R-squared

.35

Mean DV

870.39

P-Value

0.13

0.10

0.004

Panel B. 12-Years Balancing Window Restriction

Exposure to Water

Filtration

3.90**

3.50*

4.04**

(1.60)

(1.53)

(1.11)

Observations

318636

318620

R-squared

.35

Mean DV

869.38

P-Value

0.03

0.05

0.02

City FE

✓

Birth Year FE

✓

Individual Controls

✓

Family Controls

✓

City Controls

✓

Region-by-Cohort FE

✓

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering. Individual controls include dummies for race and ethnicity. Family controls

include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor force

status dummy, paternal socioeconomic score dummies, and a series of missing indicators for missing values of each

variable. City controls include average share of homeowners, average occupational income score, share of white-

collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix F

The age profile of individuals in the final sample varies between 35 and 104. In this

appendix, we replicate the analysis using the sample of individuals who survived up to ages 50,

55, 60, 65, and 70. These estimates are reported in Appendix Table F-1. Although the estimated

coefficient sizes are smaller than that of the main results, they are fairly robust across different

subsamples in consecutive columns. We should note that older individuals in the subsamples

represent early treated cities. The fact that the inclusion of individuals who died earlier (before age

50) boosts the magnitude of the coefficients may imply that the effects are slightly larger for later

cohorts and that survival of earlier cohorts beyond the life expectancy of those cohorts only pushes

the coefficients downward.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table F-1 - Replicating the Main Results Using Different Subsamples based on Survival Age

Outcomes: Age at Death (Months), Conditional on Survival up to Age:

(1)

(2)

(3)

(4)

(5)

Exposure to Water

Filtration

2.28

2.36*

1.80

2.35*

2.29

(1.11)

(.96)

(1.11)

(1.38)

(1.15)

Observations

328859

316598

292413

249347

192493

R-squared

.28

.25

.21

.17

Mean DV

876.70

885.62

900.56

923.95

953.91

P-Value

0.10

0.07

0.18

0.05

0.11

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap procedure with city-level clustering. All regressions

include city fixed effects, birth-year-by-birth-region fixed effects, individual controls, family controls, and city-level covariates. Individual controls include

dummies for race and ethnicity. Family controls include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor

force status dummy, paternal socioeconomic score dummies, and a series of missing indicators for missing values of each variable. City controls include average

share of homeowners, average occupational income score, share of white-collar occupation, share of farmers, share of other occupation, literacy rate, and share

of married.

*** p<0.01, ** p<0.05, * p<0.1

University of Chicago Press on behalf of the American Society of Health Economics.

Appendix G

One potential heterogeneity in the results is that improvements in water quality could be

more beneficial in areas with a more severe disease environment. Infant mortality rates (IMR) are

highly correlated with the availability of sanitation, water quality, healthcare access, and general

disease environment. Indeed, several studies use IMR as a proxy for the general disease

environment (Case and Paxson 2009; Almond, Currie, and Herrmann 2012). In columns 1 and 2

of Appendix Table G-1, we examine the source of heterogeneity by replicating the main results in

the subsamples based on city-cohort-specific IMR. The estimated coefficient of the high IMR

subsample is about twice the size of the low IMR subsample.

In section 6, we argued that human capital and socioeconomic status are potential pathways

between early life exposure to water quality and later life longevity. Therefore, one could expect

to observe larger impacts on the same mediatory outcomes in high IMR versus low IMR

subsamples. Using the same sample and method as in section 6, we replicate the results on years

of schooling and socioeconomic index for high and low IMR subsamples and report them in

columns 3-6 of Appendix Table G-1. Relative to the low IMR subsample, the high IMR subsample

reveals a slightly larger and statistically significant impact on years of schooling. While we observe

positive, large, and significant impacts on the socioeconomic index for the high IMR subsample,

the coefficient of the low IMR subsample points to negative and insignificant effects on the

socioeconomic index. Overall, the results of this appendix support the notion that improvements

in human capital and socioeconomic status during adulthood are pathways of the main findings of

the paper.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table G-1 - Heterogeneity in the Results Based on Birth-City-Level Infant Mortality Rates

Data, Outcome, and Subsample:

DMF

Age-at-Death

High IMR

DMF

Age-at-Death

Low IMR

Census 1950-1970

Schooling

High IMR

Census 1950-1970

Schooling

Low IMR

Census 1950-1970

SEI

High IMR

Census 1950-1970

SEI

Low IMR

(1)

(2)

(3)

(4)

(5)

(6)

Exposure to Water

Filtration

3.76

1.93

.25***

.21

1.85***

-1.37

(.71)

(1.27)

(.08)

(.33)

(.52)

(1.29)

Observations

170909

167794

158445

87434

151631

83142

R-squared

.38

.39

.31

.17

.12

Mean DV

844.02

854.21

10.40

10.43

38.48

37.21

P-Value

0.16

0.41

0.00

0.65

0.00

0.49

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap procedure with city-level clustering. Regressions of columns 1-2 include city

fixed effects, birth-year-by-birth-region fixed effects, individual controls, family controls, and city-level covariates. Individual controls include dummies for race and ethnicity. Family

controls include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor force status dummy, paternal socioeconomic score dummies, and

a series of missing indicators for missing values of each variable. City controls include average share of homeowners, average occupational income score, share of white-collar occupation,

share of farmers, share of other occupation, literacy rate, and share of married. Regressions of columns 3-6 include city fixed effects, birth-year-by-birth-region fixed effects, individual

controls, and city-level covariates. IMR stands for infant mortality rate and is calculated based on birth-city-birth-year-level rate of infant mortality. SEI stands for socioeconomic index.

*** p<0.01, ** p<0.05, * p<0.1

University of Chicago Press on behalf of the American Society of Health Economics.

Appendix H

One concern in the main results is the migration of individuals post-water filtration. For

these migrations to affect our estimates, they should correlate with exposure to water filtration.

We can empirically examine the association between migration and exposure to water filtration

using the full count 1940 census. We link the full count census individuals in the cities covered in

our final sample and who were born between 1900-1940 to 1910, 1920, and 1930 full count

censuses. For the subsample of linked individuals, we can observe whether they changed city (or

state) between each census year (1910-1930) and 1940. We then use the migration status as the

outcome in regressions similar to equation 1. The results are reported in Appendix Table H-1. We

do not observe a significant association between exposure to water filtration during childhood and

the probability of migration from 1910-city, 1920-city, and 1930-city to 1940-city (columns 1-3).

We do observe a significant coefficient for across-state migration between 1920 and 1940.

However, this is not consistent for across states migration between 1910 to 1940 and 1930 to 1940

years.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table H-1 - Exposure to Water Filtration and Cross-Census Migration

Outcomes:

Changed city

between 1910 and

1940

Changed city

between 1920 and

1940

Changed city

between 1930 and

1940

Changed state

between 1910 and

1940

Changed state

between 1920 and

1940

Changed state

between 1930

and 1940

(1)

(2)

(3)

(4)

(5)

(6)

Exposure to Water

Filtration

.05827

.0009

-.00575

.03258

.03744***

.01535

(.04778)

(.0253)

(.02397)

(.02841)

(.01156)

(.01529)

Observations

69114

265064

393467

69114

265064

393467

R-squared

.04241

.03674

.03363

.02303

.02856

.02787

Mean DV

0.396

0.315

0.207

0.184

0.134

0.079

P-Value

0.185

0.975

0.847

0.242

0.001

0.354

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap procedure with city-level clustering. All regressions

include city fixed effects, birth-year-by-birth-region fixed effects, individual controls, family controls, and city-level covariates. Individual controls include

dummies for race and ethnicity. Family controls include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor

force status dummy, paternal socioeconomic score dummies, and a series of missing indicators for missing values of each variable. City controls include average

share of homeowners, average occupational income score, share of white-collar occupation, share of farmers, share of other occupation, literacy rate, and share

of married.

*** p<0.01, ** p<0.05, * p<0.1

University of Chicago Press on behalf of the American Society of Health Economics.

Appendix I

The main difference-in-difference coefficient of the main results aggregates the impacts

across childhood ages. The general pattern of Figure 2 suggests larger impacts for in-utero and

early-life exposures. In this appendix, we disaggregate the exposure measure across different ages

to be able to better isolate critical ages. Specifically, we are low for different ages at exposure to

compete with each other. In so doing, we define dummy variables capturing exposure during in

utero and ages 0, between ages 1-4, and between ages 5-9. The age group 10-15 (and those in

treated cities) serves as the contrast group. The results are reported in Appendix Table I-1. In

column 3, we observe a monotonic pattern across coefficients: the earlier in life the exposure, the

higher the magnitude of the impact. Further, the effects become comparably small in magnitude

and statistically insignificant for the age group 5-9.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table I-1 - Exploring the Heterogeneity in the Effects of Exposure across Different Ages

Outcomes: Age at Death (Months)

(1)

(2)

(3)

Exposure to Water

Filtration in-Utero and

Age 0

1.674

2.606*

3.334**

(.8025)

(1.093)

(.836)

{0.150}

{0.070}

{0.012}

Exposure to Water

Filtration Ages 1-4

1.966

2.628***

(.992)

(.732)

{0.116}

{0.002}

Exposure to Water

Filtration Ages 5-9

1.001

(.836)

{0.422}

Observations

338742

R-squared

.356

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering and reported in curly bracket. All regressions include city fixed effects, birth-

year-by-birth-region fixed effects, individual controls, family controls, and city-level covariates. Individual controls

include dummies for race and ethnicity. Family controls include maternal literacy dummy, paternal literacy dummy,

maternal labor force status dummy, paternal labor force status dummy, paternal socioeconomic score dummies,

and a series of missing indicators for missing values of each variable. City controls include average share of

homeowners, average occupational income score, share of white-collar occupation, share of farmers, share of other

occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix J

One concern in the main results is the potential endogenous association of water filtration

with other state/city-level policy changes and sociodemographic transitions. In so doing, we use a

panel of city-by-year covariates and use city-level characteristics and policy changes as the

outcome. We execute event studies similar to equation 2 conditional on city and region-by-year

fixed effects. In these regressions, we use city-level population as weights and cluster standard

errors at the city level. We report confidence intervals based on wild bootstrap procedures with

city-level clustering. To enable comparison across regressions and figures, we standardize all

outcomes with respect to the variables’ mean and standard deviations in the final sample. The

results are reported in Appendix Figure J-1 through Appendix Figure J-5.

We do not find a robust and statistically significant association between water

implementation and state-wide prohibition reforms, share of dry counties in each state, suffrage

reform, poll tax policy change, state-level implementation of birth registration laws, state entrance

into birth registration area, and the presence of child labor and compulsory attendance laws

(Appendix Figure J-1 and Appendix Figure J-2). Almost all pre-trend and post-trend coefficients

are small in magnitude and statistically insignificant.

In the next set of figures, we explore differences in city-level sociodemographic and

socioeconomic characteristics across treated-control groups and over different years relative to the

public health reforms. We do not observe a discernible pre-trend and post-trend in various

outcomes, including the share of whites, blacks, people of other races, and immigrants (Appendix

Compulsory Attendance (CA) is a measure of state-imposed mandatory years of schooling and is calculated as the

largest of required years of schooling before dropping out and the difference between the minimum school-leaving

age and the maximum age at enrollment. Child Labor (CL) index measures the enforcement of age limitation for a

work permit and is the largest of years of education required for a work permit and the difference between the

minimum age for a work permit and the maximum age allowed for school enrollment. These measures are extracted

from Acemoglu & Angrist (2000) and are used in Appendix Figure J-2.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Figure J-3); share of females, children less than 5 years old, married women, and literate people

(Appendix Figure J-4); average occupational income score, the share of homeowners, blue-collar

workers, and farmers (Appendix Figure J-5). These tests fail to provide robust, consistent, and

significant evidence that changes in the demographic and socioeconomic characteristics of the

cities could confound the estimates.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-

level clustering, are reported. Regressions include city and region-by-year fixed effects. Regressions are

weighted using city-level population.

Appendix Figure J-1 - Event-Study Tests to Examine the Evolution of City Observables pre/post Water

Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-

level clustering, are reported. Regressions include city and region-by-year fixed effects. Regressions are

weighted using city-level population.

Appendix Figure J-2 - Event-Study Tests to Examine the Evolution of City Observables pre/post Water

Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-

level clustering, are reported. Regressions include city and region-by-year fixed effects. Regressions are

weighted using city-level population.

Appendix Figure J-3 - Event-Study Tests to Examine the Evolution of City Observables pre/post Water

Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-

level clustering, are reported. Regressions include city and region-by-year fixed effects. Regressions are

weighted using city-level population.

Appendix Figure J-4 - Event-Study Tests to Examine the Evolution of City Observables pre/post Water

Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Notes. Point estimates and 95 percent confidence intervals, extracted from wild bootstrap procedure with city-

level clustering, are reported. Regressions include city and region-by-year fixed effects. Regressions are

weighted using city-level population.

Appendix Figure J-5 - Event-Study Tests to Examine the Evolution of City Observables pre/post Water

Filtration

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix K

In this appendix, we explain the procedure of cross-census linking and construction of the

final sample. We start with the full count 1940 census and link the records to DMF death records.

This leaves us with an initial linked sample size of roughly 7.7 million observations. We then

restrict the sample to individuals born between 1900-1940, reducing the sample to about 6.6

million observations. Next, we use cross-census linking rules to link individuals across historical

censuses 1900-1930. For cohorts born between 1900-1905, we use the following information from

the 1900 census: city, state, and parental information. Similarly, for cohorts born between 1906-

1910, 1911-1920, and 1921-1930 we use information from 1910, 1920, and 1930 censuses. In the

1940 census, we have information on county of residence in 1935. If the household did not move

from 1935 to 1940, the 1935 county is the same as the 1940 county. For cohorts born between

1931-1935, we use the information of 1935 county (and state) to assign the city of birth/childhood.

This is possible because for 24 cities out of 25 cities of the final sample, there is a 1-to-1 link

between city and county. Further, several counties within New York City can be mapped only to

New York City (the 25th city). Finally, for cohorts born between 1936-1940, we use information

from the 1940 census. We drop all individuals who are not linked and for whom we cannot infer

the city of birth/childhood as well as parental information.

These selections leave us with a sample size of about 2.4 million observations. The sample

contains 1,037 cities. Restricting the sample to 25 cities in the final sample for which we have

information on water filtration reduces its size by about 84%. We further restrict the sample to

cohorts born 15 years before and after the city-specific water filtration year (only for treated cities).

The final sample size covers roughly 338 thousand observations.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Although the cross-census linking considerably reduces the sample size, the process

arguably limits measurement error in assigning the city of birth/childhood. In Appendix Table K-1,

we use the 1940 city as the city of birth/childhood and replicate the main results. We observe point

estimates that are 30 percent smaller in size than the main results, suggesting that measurement

errors likely result in coefficients that underestimate the true impacts.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table K-1 - Replicating the Main Results Using the 1940 City as City of Birth/Childhood

Outcomes: Age at Death (Months)

(1)

(2)

(3)

Exposure to Water

Filtration

1.44

1.40

2.18

(1.22)

(1.30)

(1.70)

Observations

996533

R-squared

.36

Mean DV

887.05

P-Value

0.24

0.33

0.28

City FE

✓

Birth Year FE

✓

Individual Controls

✓

Family Controls

✓

City Controls

✓

Region-by-Cohort FE

✓

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering. Individual controls include dummies for race and ethnicity. Family controls

include maternal literacy dummy, paternal literacy dummy, maternal labor force status dummy, paternal labor force

status dummy, paternal socioeconomic score dummies, and a series of missing indicators for missing values of each

variable. City controls include average share of homeowners, average occupational income score, share of white-

collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix L

One concern regarding the exogeneity assumption pertains to parental fertility response to

city-wide public health infrastructure improvements. For instance, if water filtration had an evident

impact on infant mortality, parents may observe improvements in their children's survival and be

encouraged to increase fertility. Similarly, they may reduce their fertility as the target number of

children could be attained by fewer births in the presence of lower child mortality rates (Palloni

and Rafalimanana 1999; Sandberg 2016). Moreover, the fertility decision could also be correlated

with other sociodemographic characteristics of parents, which are among the determinants of their

children’s later-life longevity.

To address these concerns, we use limited natality and mortality data available for a subset

of cities and counties in the data for the years 1915-1940 extracted from Bailey et al. (2016). The

data reports the infant mortality and birth rates at the city-year level. We merge this data with water

filtration data and implement regressions that include city and year fixed effects. First, we explore

the effects on infant mortality rates. These results are reported in column 1 of Appendix Table L-1.

Water filtration results in roughly 6.3 fewer infant deaths per 100,000 live births, equivalent to a

9.8 percent reduction from the outcome’s mean.

However, this finding is sensitive to the

functional form and becomes insignificant when we use the log infant mortality rate as the outcome

(column 2). Next, we assess the associations with the birth rate per 1,000 women and log birth rate

There are two reasons that our findings on mortality rate is different than those reported by Anderson, Charles, &

Rees (2022), i.e., roughly 11% reduction. First, they include a city-specific time trend while we do not. Including unit-

specific trends may over-control for time-varying treatment effects and has been a controversial issue in the literature

(Lee and Solon 2011; Meer and West 2016; Neumark, Salas, and Wascher 2014; Gruber and Frakes 2006; Chou,

Grossman, and Saffer 2006; Goodman-Bacon 2021). Second, since the main purpose of this section is to evaluate

fertility response, we use natality records from a data source that contains birth rate information starting from 1915.

To have a similar panel with natality records, we also use mortality during the same period (i.e., 1915-1940). Indeed,

when we use the replication data of Anderson, Charles, & Rees (2022), limit the sample to 1915-1940 years, and

remove the city-trend, we reach almost identical effects as columns 1-2 of Appendix Table L-1, both for level and log

of infant mortality rate.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

(columns 3-4). The estimated coefficient suggests small and statistically insignificant reductions

in the birth rate, which limits further interpretations. Overall, while we find some evidence for

improvements in infants’ health, we fail to observe a discernible fertility response.

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

Appendix Table L-1 - Effects on Infant Mortality and Fertility Rates

Outcomes:

Infant Mortality

Rate

Log Infant

Mortality Rate

Birth Rate

Log Birth Rate

(1)

(2)

(3)

(4)

Exposure to Water

Filtration

-6.389**

-.032

-2.539

-.024

(2.414)

(.041)

(1.859)

(.032)

Observations

559

R-squared

.976

.977

.969

.979

Mean DV

63.997

4.103

37.580

3.630

P-Value

0.024

0.396

0.154

0.447

Notes. Standard errors, clustered on city, are in parentheses. P-values are extracted from the wild bootstrap

procedure with city-level clustering. Regressions include city and region-by-year fixed effects. Regressions also

include city covariates. City controls include average share of homeowners, average occupational income score,

share of white-collar occupation, share of farmers, share of other occupation, literacy rate, and share of married.

*** p<0.01, ** p<0.05, * p<0.1

This is the author's accepted manuscript without copyediting, formatting, or final corrections. It will be published in its final form in an upcoming issue of American

Journal of Health Economics, published by The University of Chicago Press on behalf of the American Society of Health Economics.

ResearchGate has not been able to resolve any citations for this publication.

The Siren Song of Cicadas: Early-Life Pesticide Exposure and Later-Life Male Mortality

Article

Full-text available

Nov 2023
J ENVIRON ECON MANAG

CenSoc: Public Linked Administrative Mortality Records for Individual-level Research

Article

Full-text available

Nov 2023

In the United States, much has been learned about the determinants of longevity from survey data and aggregated tabulations. However, the lack of large-scale, individual-level administrative mortality records has proven to be a barrier to further progress. We introduce the CenSoc datasets, which link the complete-count 1940 U.S. Census to Social Security mortality records. These datasets—CenSoc-DMF (N = 4.7 million) and CenSoc-Numident (N = 7.0 million)—primarily cover deaths among individuals aged 65 and older. The size and richness of CenSoc allows investigators to make new discoveries into geographic, racial, and class-based disparities in old-age mortality in the United States. This article gives an overview of the technical steps taken to construct these datasets, validates them using external aggregate mortality data, and discusses best practices for working with these datasets. The CenSoc datasets are publicly available, enabling new avenues of research into the determinants of mortality disparities in the United States.

Long‐run intergenerational health benefits of women empowerment: Evidence from suffrage movements in the US

Article

Full-text available

Jul 2023
HEALTH ECON

An ongoing body of research documents that women empowerment is associated with improved outcomes for children. However, little is known about the long‐run effects on health outcomes. This paper adds to this literature and studies the association between maternal exposure to suffrage reforms and children's old‐age longevity. We utilize changes in suffrage laws across US states and over time as a source of incentivizing maternal investment in children's health and education. Using the universe of death records in the US over the years 1979–2020 and implementing a difference‐in‐difference econometric framework, we find that cohorts exposed to suffrage throughout their childhood live 0.6 years longer than unexposed cohorts. Furthermore, we show that these effects are not driven by preexisting trends in longevity, endogenous migration, selective fertility, and changes in the demographic composition of the sample. Additional analysis reveals that improvements in education and income are candidate mechanisms. Moreover, we find substantial improvements in early‐adulthood socioeconomic standing, height, and height‐for‐age outcomes due to childhood exposure to suffrage movements. A series of state‐level analyses suggest reductions in infant and child mortality following suffrage law change. We also find evidence that counties in states that passed the law experienced new openings of County Health Departments and increases in physicians per capita.

Social Insurance Programs and Later-Life Mortality: Evidence from New Deal Relief Spending*

Article

Full-text available

Oct 2022
J HEALTH ECON

A growing body of research explores the long-run effects of social programs and welfare spending. However, evidence linking welfare support in early life with longevity is limited. We add to this literature by evaluating the effect of in-utero and early-life exposure to the largest increases in welfare spending in the US history under the New Deal programs. Using Social Security Administration death records linked with the 1940-census and spending data for 115 major cities, we show that the spending is correlated with improvements in old-age longevity. A treatment-on-treated calculation focused on a period when spending rose by approximately 1900 percent finds that a 100 percent rise in municipal spending in the year of birth is associated with roughly 3.5 months higher longevity. We show that these effects are not driven by endogenous selection of births, selective fertility, endogenous migration, and sample selection caused by endogenous data linking. Additional analysis suggests that rises in education and socioeconomic status are likely channels of impact.

Maternal human capital and infants’ health outcomes: Evidence from minimum dropout age policies in the US

Article

Full-text available

Sep 2022

The purpose of this cross-sectional study is to examine the causal relationship of maternal education and infants' health outcomes. Using birth certificate data over the years 1970–2004 and exploiting the space-time variation in Minimum Dropout Age laws to solve the endogeneity of education, we find a sizeable effect of mothers' education on their birth outcomes. An additional year of maternal education is associated with a reduction in incidences of low birth weight and preterm birth by 15.2 and 12.7 percent, respectively. The estimates are robust across various specifications and even when allowing mothers’ cohort-of-birth to vary across regions. The results suggest that the candidate mechanisms of impact include improvements in timing, quantity, and quality of prenatal care, lower negative health behavior during pregnancy such as smoking and drinking, and higher spousal education.

Drinking water chlorination has minor effects on the intestinal flora and resistomes of Bangladeshi children

Article

Full-text available

May 2022

Healthy development of the gut microbiome provides long-term health benefits. Children raised in countries with high infectious disease burdens are frequently exposed to diarrhoeal pathogens and antibiotics, which perturb gut microbiome assembly. A recent cluster-randomized trial leveraging >4,000 child observations in Dhaka, Bangladesh, found that automated water chlorination of shared taps effectively reduced child diarrhoea and antibiotic use. In this substudy, we leveraged stool samples collected from 130 children 1 year after chlorine doser installation to examine differences between treatment and control children’s gut microbiota. Water chlorination was associated with increased abundance of several bacterial genera previously linked to improved gut health; however, we observed no effects on the overall richness or diversity of taxa. Several clinically relevant antibiotic resistance genes were relatively more abundant in the gut microbiome of treatment children, possibly due to increases in Enterobacteriaceae. While further studies on the long-term health impacts of drinking chlorinated water would be valuable, we conclude that access to chlorinated water did not substantially impact child gut microbiome development in this setting, supporting the use of chlorination to increase global access to safe drinking water. A substudy nested within a double-blind cluster-randomized controlled trial in Bangladesh shows that drinking chlorinated water had relatively minor impacts on children’s gut microbiome development in this setting.

Early-Life Disease Exposure and Its Heterogeneous Effects on Mortality Throughout Life: Sweden, 1905-2016

Article

Jul 2024
DEMOGRAPHY

Exposure to infectious diseases in early life has been linked to increased mortality risk in later life in high-disease settings, such as eighteenth- and nineteenth-century Europe. Less is known about the long-term effects of early-life disease exposure in milder disease environments. This study estimates heterogeneous effects from disease exposure in infancy on later-life mortality in twentieth-century Sweden, by socioeconomic status at birth and sex. Using historical population data for southern Sweden, we study 11,515 individuals who were born in 1905–1929 from age 1 until age 85. We measure exposure to disease using the local post–early neonatal mortality rate in the first 12 months after birth and apply flexible parametric survival models. For females, we find a negative effect on life expectancy (scarring) at ages 1–85 following high disease exposure in infancy, particularly for those born to unskilled workers. For males, we find no negative effect on later-life survival, likely because stronger mortality selection in infancy outweighs scarring. Thus, even as the incidence of infectious diseases declined at the start of the twentieth century, early-life disease exposure generated long-lasting negative but heterogeneous population health effects.

Revisiting Event Study Designs: Robust and Efficient Estimation

Article

Feb 2024

We develop a framework for difference-in-differences designs with staggered treatment adoption and heterogeneous causal effects. We show that conventional regression-based estimators fail to provide unbiased estimates of relevant estimands absent strong restrictions on treatment-effect homogeneity. We then derive the efficient estimator addressing this challenge, which takes an intuitive “imputation” form when treatment-effect heterogeneity is unrestricted. We characterize the asymptotic behavior of the estimator, propose tools for inference, and develop tests for identifying assumptions. Our method applies with time-varying controls, in triple-difference designs, and with certain non-binary treatments. We show the practical relevance of our results in a simulation study and an application. Studying the consumption response to tax rebates in the United States, we find that the notional marginal propensity to consume is between 8 and 11 percent in the first quarter — about half as large as benchmark estimates used to calibrate macroeconomic models — and predominantly occurs in the first month after the rebate.

An Assessment of CenSoc Match Quality

Preprint

Jun 2022

The CenSoc datasets link individual-level 1940 Census records to Social Security death records using deterministic record linkage algorithms. In this technical report, we describe our record linkage methodology and assess the accuracy and representativeness of the CenSoc Version 2.1 matches.

Estimating the Effects of Milk Inspections on Infant and Child Mortality, 1880–1910

Article

May 2022

In the mid-nineteenth century, the urban milk supply in the United States was regularly skimmed or diluted with water, reducing its nutritional value. At the urging of public health experts, cities across the country hired milk inspectors, who were tasked with collecting and analyzing milk samples with the goal of preventing adulteration and skimming. Using city-level data for the period 1880-1910, we explore the effects of milk inspections on infant mortality and mortality among children under the age of five. Event study estimates are small and statistically insignificant, providing little evidence of post-treatment reductions in either infant or child mortality.

Water is Life, Clean Water Means Health: The Effect of Early-Life Exposure to the City-Wide Water Filtration on Old-Age Mortality

Figures

Recommended publications

Long-Term Health Benefits of Occupational Licensing: Evidence from Midwifery Laws

Unequal before death: The effect of paternal education on children's old-age mortality in the United...

Childhood exposure to birth registration laws and old‐age mortality

Long‐run intergenerational health benefits of women empowerment: Evidence from suffrage movements in...