Can Disaster Preparedness Change the Game? Mitigating the Health Impact of Disease Outbreaks

Abstract

In times of epidemics and pandemics, depletion or diversion of health system resources from routine health care is common, posing serious threats to primary care. This paper estimates the contemporaneous and long-run effects of health disasters on maternal and child mortality in 111 countries during 2000-2019 using two-way fixed-effects and two-step system general method of moments frameworks. We also provide evidence that indicates how health system, macroeconomic, institutional, and structural characteristics can mitigate disaster effects. In low- and middle-income countries, health disasters increase maternal, under-5, and neonatal mortalities by 0.3%, 0.3%, and 0.2% instantaneously and by 35%, 80%, and 26% after one year, respectively. Our estimates show that disaster preparedness can prevent these effects. However, other mitigators, namely health emergency finance, universal health coverage, education, gender equality, and water, sanitation, and hygiene coverage, have greater impact.

Full text

ERF Working PaPers series
Can Disaster Preparedness Change
the Game? Mitigating the Health
Impact of Disease Outbreaks
Amira El-Shal, Mahmoud Mohieldin and Eman Moustafa
Working Paper No. 1466
March 2021
2021

CAN DISASTER PREPAREDNESS CHANGE THE GAME?
MITIGATING THE HEALTH IMPACT OF DISEASE
OUTBREAKS
Amira El-Shal, Mahmoud Mohieldin
1
and Eman Moustafa
2
Working Paper No. 1466
March 2021
Send correspondence to:
Amira El-Shal
Cairo University
amira.elshal@feps.edu.eg
1
Faculty of Economics and Political Science, Cairo University; 1, Gamaa Street, Giza, 12613, Egypt.
United Nations; 405 East 42nd Street, New York, NY, 10017, USA. mahmoud.mohieldin@feps.edu.eg
2
General Authority for Investment & Free Zones; 3, Salah Salem Street, Cairo, 11562, Egypt.
e.fawzy@gafinet.org.eg

First published in 2021 by
The Economic Research Forum (ERF)
21 Al-Sad Al-Aaly Street
Dokki, Giza
Egypt
www.erf.org.eg
Copyright © The Economic Research Forum, 2021
All rights reserved. No part of this publication may be reproduced in any form or by any electronic or
mechanical means, including information storage and retrieval systems, without permission in writing
from the publisher.
The findings, interpretations and conclusions expressed in this publication are entirely those of the
author(s) and should not be attributed to the Economic Research Forum, members of its Board of
Trustees, or its donors.

Abstract
In times of epidemics and pandemics, depletion or diversion of health system resources from
routine health care is common, posing serious threats to primary care. This paper estimates the
contemporaneous and long-run effects of health disasters on maternal and child mortality in
111 countries during 2000-2019 using two-way fixed-effects and two-step system general
method of moments frameworks. We also provide evidence that indicates how health system,
macroeconomic, institutional, and structural characteristics can mitigate disaster effects. In
low- and middle-income countries, health disasters increase maternal, under-5, and neonatal
mortalities by 0.3%, 0.3%, and 0.2% instantaneously and by 35%, 80%, and 26% after one
year, respectively. Our estimates show that disaster preparedness can prevent these effects.
However, other mitigators, namely health emergency finance, universal health coverage,
education, gender equality, and water, sanitation, and hygiene coverage, have greater impact.
Keywords: Preparedness; mitigation; health emergency; epidemic; maternal mortality; child
mortality.
JEL Classifications: I1; I18; H5; H51.
1

1. Introduction and background
During disease outbreaks, surging demand for health care to care for the affected, deaths and
illness among health providers, destruction of health facilities, and disruption of electricity,
water, and sanitation, and supply chains inevitably jeopardize the provision of routine health
services (Kruk et al., 2016). Such testing of the resilience of a nation’s health system
disproportionately affects pregnant women and children during and in the aftermath of disasters
(Harville et al., 2010).
The COVID-19 pandemic began at a time when low- and middle-income countries (LMICs)
had achieved improved maternal and child health, with a 38% drop in maternal mortality and
a halved child mortality from 2000 to 2018 (World Health Organization, 2020). Yet, with less
than a decade left to achieve the United Nations’ Sustainable Development Goals (SDGs),
progress toward SDG3 “Ensure healthy lives and promote well-being for all at all ages” has
been uneven. Strengthening preparedness for health emergencies becomes more urgent as
health disasters continue to erode recent gains. The United Nations Independent Accountability
Panel (2020) released concerning figures about the impact of the pandemic in January-April
2020. First, women, children, and adolescents lost access to 20% of health and social services
as a result of the COVID-19 pandemic. Second, about 13.5 million children missed life-saving
vaccinations over the first four months of 2020; such skipping is particularly dangerous in
LMICs. Third, before the pandemic, approximately 295,000 women died during or shortly after
pregnancy in 2017. These and other statistics show that addressing avoidable maternal and
child mortality is paramount in disaster mitigation.
Empirical research on maternal and child health during disease outbreaks is still in its infancy,
with very few quantitative studies examining the various facets of disease outbreak impacts.
Conceptually and statistically, research has examined the effect of disease outbreaks on
maternal and child mortality in single LMICs. However, the scale of nationwide averages
ignores more micro-level impacts at the local level and the effects of outbreaks on neighboring
countries. Evidence from the 2014 Ebola outbreak in Guinea, Liberia, and Sierra Leone reveals
a deterioration in the uptake and provision of maternal and child health services (Delamou et
al., 2017; Iyengar et al., 2015; Sochas et al., 2017). Studies reported significant reductions in
facility-based deliveries and utilization across the three countries, compared to pre-outbreak
metrics. Estimating the indirect effect of COVID-19 on maternal and child mortality in LMICs,
Roberton et al. (2020) modeled three possible scenarios of reduction in the coverage of
essential maternal and child health services over 3, 6, and 12 months, using the Lives Saved
Tool (LiST). They estimated, for example, that 1,157,000 additional child deaths and 56,700
additional maternal deaths occurred across the 118 countries included in the analysis due to
reductions of around 45% for 6 months were estimated to result in. Using a similar LiST model,
Menendez et al. (2020) estimated the indirect impact of the COVID-19 pandemic on maternal
and newborn health in India, Indonesia, Nigeria, and Pakistan over 12 months. Their results
suggest that the pandemic will yield an estimated 766,180 additional maternal and newborn
deaths and stillbirths, which corresponds to a 31% increase in mortality.
2

Most recent empirical literature on the indirect effects of health disasters and disaster
mitigation measures focuses on clinical factors associated with maternal and child mortality in
individual countries and uses facility-level data. However, an array of social and economic
macro-structural determinants, ranging from social structures to the functioning and
responsiveness of health systems, affects maternal and child health. A careful understanding
of these determinants and how they operate during public health emergencies is central to
sustain achieved health gains, reach the women and children still excluded from recent
improvements, and mitigate the negative effects of health disasters.
This paper addresses the gap in the literature by estimating both the short- and long-run effects
of epidemic and pandemic disasters on maternal and child health across different groups of
countries and by providing evidence of how governments can mitigate the negative effects.
Specifically, we seek to answer three questions: (1) What is the indirect short- and long-run
impact of disease outbreaks on maternal, under-5, and neonatal mortality? (2) What role does
health emergency preparedness (including surveillance, response, preparedness, and risk
communication) play in mitigating the effects of outbreaks? (3) Which of these factors mitigate
health disaster effect: health service coverage; health emergency finance; water, sanitation, and
hygiene (WASH) coverage; education; and gender equality? To answer these questions, we
estimate two-way fixed-effects and two-step “system” generalized methods of moments
(GMM) models based on the available data for 111 countries, of which we consider 93 to be
LMICs, between 2000 and 2019.
The novelty of this study is twofold. First, to our knowledge, this is the first empirical study to
provide macroeconomic evidence on the indirect impact of disease outbreaks on maternal and
child mortality and to quantify this impact in both the short and long runs. Similar studies use
facility-level data, which, while being well positioned to capture one-time drops in health
service utilization, is not appropriate to capture the effect of system preparedness. Importantly,
we estimate this impact within a holistic framework that accommodates the well-established
determinants of maternal and child mortality in the spirit of McCarthy and Maine (1992) and
Mosley and Chen (1984), respectively. Second, this study is the first to weigh the significance
of health emergency preparedness in mitigating the negative effects of disease outbreaks
against other mitigation factors. Our findings provide timely insights on how we can efficiently
mitigate the effects of health disasters to prevent maternal and child deaths as well as how to
improve preparedness for future disasters.
2. Data and conceptualization
To answer our research questions, we construct a panel dataset by merging maternal and child
health indicators (dependent variables) with the occurrence of and estimated people affected
by epidemic and pandemic disasters (main explanatory variable); established determinants of
maternal and child health; and potential mitigators of health disasters (explanatory variables).
The analysis is conducted at the country level across all world regions over the period 2000-
2019 to address the need for global evidence. A total of 111 countries are covered, out of which
3

23 are low-income, 70 are middle-income, and 18 are high-income countries (HICs), allowing
us to compare the (mitigated) health effects in LMICs and in HICs. A full list of covered
countries is provided in Appendix A.
2.1. Health indicators
Three health indicators available in the United Nations Global SDG Indicators Database
constitute our dependent variables: maternal mortality ratio (MMR, modeled estimate, per
100,000 live births), neonatal mortality rate (NMR, per 1,000 live births), and under-5 mortality
rate (U5MR, per 1,000 live births). MMR is a key impact indicator for women’s health and
well-being, which reflects the health system’s capacity to provide effective health care in
preventing and addressing the complications that occur during pregnancy and childbirth. NMR
and U5MR are key impact indicators for child health and well-being, which reflect the access
of children and communities to basic health interventions (e.g., vaccination, medical treatment
of infectious diseases, etc.).
3
2.2. Health disasters
Our two main explanatory variables are (1) the occurrence of and (2) estimated people affected
by epidemic and pandemic disasters. Information on health disasters and their human impacts
is extracted from the Emergency Events Database (EM-DAT), which is a service of the Centre
for Research on the Epidemiology of Disasters (CRED).
4
The EM-DAT reports the number of
people killed or injured or rendered homeless as a result of disease outbreaks. A disaster is
defined as an incident meeting any of the following criteria: (1) 10 or more people reported
killed; (2) 100 people reported affected; (3) declaration of a state of emergency; or (4) call for
international assistance. The EM-DAT categorizes disease outbreaks, including pandemics and
epidemics, under the biological subgroup (epidemic, pandemic, insect infection, and animal
accident) of natural disasters.
As we presume that the impact of disease outbreaks on maternal and child health depends on
the magnitude of disasters, we standardize our disaster measure (2) of the estimated number of
people affected by epidemic and pandemic disasters. Since the disaster itself affects the current
year’s population, we divide the measures of the number of people affected by the population
size in the year prior to the disaster (Przylnski & Hallegatte, 2010; Noy, 2009; Raddatz, 2007).
To verify that the way we construct the disaster measure does not raise any endogeneity
concerns, we also estimate our exact model using our disaster measure (1) as a binary dummy
indicator of disaster occurrence.
3
The definitions of and rationale for MMR, NMR, and U5MR are provided online at
https://unstats.un.org/wiki/display/SDGeHandbook/Home
4
Established in 1973 as a non-profit institution, CRED is based at the Catholic University of Louvain in Belgium
and publicly available on CRED’s website at www.cred.be.
4

2.3 Determinants of maternal mortality
Selection of the determinants of MMR is based on robust evidence from both country- and
individual-level studies in the spirit of the McCarthy-Maine (1992) framework for analyzing
the determinants of maternal mortality. Analyzing the cultural, social, economic, behavioral,
and biological factors influencing maternal mortality, they concluded that all determinants of
maternal mortality are associated with pregnancy and pregnancy-related complications. These
two outcomes are directly influenced by five sets of woman-related intermediate determinants:
health status, reproductive status, access to health services, health care behavior, and other
unknown or unpredicted factors. In turn, three sets of socioeconomic and cultural factors
(distant determinants) influence the intermediate determinants: women’s status in family and
community, family’s status in community, and community’s wealth. We include MMR
country-level determinants previously explored in the literature and variables identified as
MMR determinants at the individual or household level and that can be extrapolated from
the micro-level to the macro-scale to suit our country level analysis.
5
We include a set of variables to reflect four general categories of MMR determinants. The first
pertains to economic status (Anand & Bärnighausen, 2004; Ensor et al., 2010), and is captured
by one variable: gross domestic product (GDP) per capita (constant 2010 US$). This variable
is a proxy for community’s aggregate wealth or, generally, community’s status (McCarthy &
Maine, 1992).
The second pertains to health, and is captured by five variables: physician density (per 1,000
people), a measure of community’s health resources or, generally, community’s status (Anand
& Bärnighausen, 2004; McCarthy & Maine, 1992); adolescent fertility rate (births per 1,000
women ages 15-19), a measure of reproductive health (Godefay et al., 2015; Makinson, 1985;
McCarthy & Maine, 1992); births attended by skilled health staff (%), a measure of the use of
modern care for delivery or, generally, utilization of health services and health care behavior
(Berhan & Berhan, 2014; McCarthy & Maine, 1992);
6
prevalence of anemia among women of
reproductive age (% of women ages 15-49), a measure of nutritional status or, generally, health
status (Brabin et al., 2001; Daru et al., 2018; McCarthy & Maine, 1992); and incidence of
tuberculosis (per 100,000 people), a measure of prevalence of infectious diseases or, generally,
health status (McCarthy & Maine, 1992; Moran & Moodley, 2012; Zaba et al., 2013).
The third category pertains to education (Anand & Bärnighausen, 2004), and is captured by
one variable: female primary completion rate (% of relevant age group). This is a measure of
women’s education and, generally, a proxy for women’s status in family and community
(McCarthy & Maine, 1992).
5
One example is mother’s education that can be captured at the country level by female educational attainment.
6
Skilled-assisted delivery is of particular relevance to MMR as it can help stop around 16%-33% of maternal
deaths through the primary or secondary prevention of four life-threatening obstetric complications (Graham et
al., 2001).
5

The fourth category pertains to gender equality (Chirowa et al., 2013), and is captured by one
variable: the gender parity index for gross enrollment ratio in primary education, defined as the
ratio of girls to boys enrolled at primary level in public and private schools.
7
This is a proxy
measure for women’s social autonomy or women’s status in family and community in the broad
sense (McCarthy & Maine, 1992).
We also control for the type of residence or, at the country scale, the degree of urbanization
using one variable urban population (% of total population). The significance of this factor
remains controversial in the literature (Matthews et al., 2010).
2.4. Determinants of child mortality
Selection of the determinants of U5MR, including NMR, is based on previous evidence from
country- and individual-level studies in the spirit of Mosley and Chen’s (1984) analytical
framework for the study of child survival. They theorize that the relationship between income
(per capita) and child mortality is mediated through underlying socioeconomic status, which
manifests in proximate determinants influencing the risk of disease that, in turn, links to the
probability of death. We draw from both U5MR country- and individual- or household-level
determinants explored in the empirical literature.
We include a set of variables to reflect four main channels through which socioeconomic status
can influence the risk of childhood disease and thus U5MR, in harmony with the earlier
classification of MMR determinants. The first channel pertains to economic status (Farahani et
al., 2009; O’Hare et al., 2013), and is captured by GDP per capita (constant 2010 US$). In a
sense this variable reflects financial access to health care.
The second channel pertains to health, and we capture it four variables. The first is physician
density (per 1,000 people), a proxy for health system resources in general (Anand &
Bärnighausen, 2004; Hanmer et al., 2003; Jamison et al., 2016). The second is DPT
immunization (% of children ages 12-23 months),
8
a proxy for vaccination coverage and
utilization of health care services in the broad sense (Chowdhury et al., 2020; Hanmer et al.,
2003; Jamison et al., 2016)
9
that is irrelevant to NMR. The third is the percentage of the
population using at least basic drinking water services (Hanmer et al., 2003; Kayode et al.,
2012; Kipp et al., 2016). The fourth is the percentage of the population using at least basic
sanitation services (Hanmer et al., 2003; Kayode et al., 2012; Tagoe et al., 2020).
7
Further details on the gender parity index are provided online at
http://mdgs.un.org/unsd/mi/wiki/3-1-Ratios-of-girls-to-boys-in-primary-secondary-and-tertiary-education.ashx
8
DPT: diphtheria, tetanus toxoids, and pertussis.
9
Jamison et al. (2016) describe vaccination coverage, which some studies consider a health policy measure, as a
potential determinant of technical progress as it provides a natural indicator of the extent to which a country’s
health services adopt powerful mortality-reducing technologies.
6

The third channel pertains to education, specifically mother’s education (Anand &
Bärnighausen, 2004; Hanmer et al., 2003; Jamison et al., 2016). It is captured by the percentage
of women in the relevant age group who have completed primary education, a proxy for
mothers’ knowledge of health care practices.
The fourth channel pertains to gender equality (Hanmer et al., 2003). It is captured by the
gender parity index for gross enrollment ratio in primary education.
Moreover, we include one variable to reflect the type of residence or, at the country scale, the
degree of urbanization: urban population (% of total population).
10
The United Nations Global SDG Indicators Database provides the maternal and child mortality
determinants used here.
11
2.5. Disaster mitigating factors
Women and children suffer from various inequities and discrimination based on, for instance,
gender, income, age, place of residence, education, as well as health inequities, resulting in
worse health outcomes (Temmerman et al., 2015). Thus, we construct a comprehensive dataset
of macro-level mediating factors that measure these inequities and discrimination forms. We
group the included factors into five categories: (1) health service coverage, (2) public health
emergency management, (3) health emergency finance, (4) WASH coverage, and (5) education
and gender equality. We hypothesize that these factors can mitigate (or exacerbate) the negative
impact of health disasters on maternal and child mortality.
Health service coverage is proxied by the universal health coverage (UHC) service coverage
index and physician density (per 1,000 people). The former is defined as the average coverage
of essential services pertaining to reproductive, maternal, newborn and child health; infectious
diseases; non-communicable diseases; and service capacity and access.
12
UHC advances
maternal and child health in several ways, most importantly through promoting the removal of
point-of-care fees for essential services and, therefore, reducing the vulnerabilities of the most
marginalized groups during health disasters (Temmerman et al., 2015). In parallel, a high
density of physicians can help absorb the increased demand for health care that outbreaks bring.
10
Some studies suggest urban-rural residence, or the ecological setting in the broad sense, is a differentiating
factor for U5MR (Chowdhury et al., 2020; Hanmer et al., 2003; Jamison et al., 2016; Kayode et al., 2012; Kipp
et al., 2016; Van Malderen et al., 2019; Yaya et al., 2018), but other studies present more mixed findings (e.g.,
Matthews et al., 2010).
11
We ideally would have included estimates for health expenditure (% of GDP; Kipp et al., 2016), poverty
headcount ratios (Anand & Bärnighausen, 2004), modern contraceptive prevalence rates (Aheto, 2019; Hanmer
et al., 2003; Haroun et al., 2007; Kayode et al., 2012; Kumar et al., 2000; Tagoe et al., 2020), antenatal care
coverage (4+visits) (Kayode et al., 2012), exclusive breastfeeding (% of children under 6 months; Kayode et al.,
2012; Mani et al., 2012), and prevalence of underweight (% of children under 5; Hanmer et al., 2003; Haroun et
al., 2007; Tagoe et al., 2020; Yaya et al., 2018), but this data is too sparse to be used at the country level.
12
A detailed description of the UHC service coverage index is provided online at
https://www.who.int/data/gho/indicator-metadata-registry/imr-details/4834
7

Stronger preparedness for public health emergencies is likely to reduce the risks and the impact
of all-hazards emergencies on population health (Khan et al., 2018). To reflect the overall
health system capacity for public health emergency management, we use the International
Health Regulations core capacity index, which is the average of attributes of 13 core capacities:
national legislation, policy, and financing; coordination and national focal point
communications; surveillance; response; preparedness; risk communication; human resources;
laboratory; points of entry; zoonotic events; food safety; chemical events; and radio-nuclear
emergencies. While all these capacities are necessary for the effective management of health
disasters, we separately report the mitigation effects of four highly relevant capacities:
surveillance, response, preparedness, and risk communication.
13
Most maternal and child deaths occur in LMICs whose limited resources have historically
hindered their ability to provide adequate quality of care even in the absence of health disasters.
Health emergency finance can play a pivotal role in channeling resources to impacted countries
whose health systems are already financially pressured. In the broad sense, impacted countries
and individuals can draw from different financial sources to mitigate the disaster effect; five
variables are included to represent these sources: total official development assistance (ODA)
to medical research and basic health sectors, personal remittances received (% of GDP), GDP
per capita (constant 2010 US$), adjusted net savings (% of gross national income), as well as
commercial bank branches (per 100,000). The latter reflects financial inclusion. These sources
increase the fiscal space of LMICs and the purchasing power of their populations for more
health investments. Financial inclusion is associated with individuals’ ability to access timely
and appropriate health services during health disaster episodes. Impacted individuals typically
suffer from income loss because of their inability to work and their need for medical treatment
and to meet the incurred payments.
Access to WASH services is essential to population health in normal times. It becomes even
more important during disease outbreaks, as WASH interventions have consistently reduced
both the risk of disease and the risk of transmission in a variety of settings. Such initiatives
involve, variously, handwashing and hygiene education, greater access to drinking water, and
improved toilets.
14
Vulnerable women and children exhibit the greatest benefits from such
interventions (Akseer et al., 2020). The three variables we use to capture WASH coverage as
a measure of disaster preparedness are also used to measure socioeconomic determinants of
child mortality: use of drinking water services and safe sanitation services and access to basic
handwashing facilities.
Finally, the analysis includes three measures of education and gender equality: female primary
completion rate (% of relevant age group); the gender parity index for gross enrollment
13
A comprehensive description of the data coverage, reporting, aggregation, and anonymization of the
International Health Regulations core capacity index is provided online at
https://www.who.int/health-topics/international-health-regulations#tab=tab_1
14
A detailed description of WASH interventions and data is provided online at
https://data.unicef.org/resources/wash-water-supply-sanitation-hygiene/
8

ratio in primary education; and the women, business, and the law index. The latter is consists
of eight indicators reflecting women’s interactions with the law in the eight areas of mobility,
workplace, pay, marriage, parenthood, entrepreneurship, assets, and pension
15
and is a proxy
measure for women’s social and legal autonomy. Reports indicate education was associated
with higher levels of resilience after a natural disaster over the long term (Frankenberg et al.,
2013). Several studies have recognized the effect of gender inequality on disease transmission
(e.g., Ehrhardt et al. (2009) on gender inequality, women empowerment, and HIV/AIDS).
The United Nations Global SDG Indicators Database provided all of the disaster mitigating
factor data except the public health emergency management measures, which we obtained from
the World Health Organization.
3. Methodology
3.1. Estimating health disaster impacts
3.1.1. Static longitudinal specification
Drawing from Farahani et al. (2009) from the health economics literature and Noy (2009) from
the natural disasters literature, we propose a two-way fixed-effects model to estimate the causal
impact of epidemic and pandemic disasters on maternal and child mortality. For each country
!
at year
"
, the following parsimonious specification is estimated three times, once for each of
our health impact indicators of interest:
#$%& ' ()$ * (+,$%& * (-.$%& * /$* 0&* 1$&
2
(1)
The dependent variable,
#$%&
, is either MMR, U5MR, or NMR for country
!
at year
"
.
,$%&
denotes the four vectors of determinants pertinent to economic status, health, education, and
gender equality risk factors for the three health outcomes.
.$%&
is the epidemic and/or pandemic
disaster variable. To investigate whether the way we constructed the damage variable (
.$%&
)
could have created any endogeneity problem, we convert the continuous disaster measure of
the number of people affected into a binary indicator for the occurrence of a disaster
(1=disaster, 0=no disaster) and examine whether this changes our results
16
.
/$
and
0&
are sets
of country- and year-fixed effects, respectively.
()$
is a country-specific intercept and
1$&
is a
random and normally distributed disturbance term.
By including country fixed effects, we eliminate any confounding from country characteristics,
whether observed or unobserved, that are constant over time within each country. We also
include period dummies to reflect the global level of health technology available at time
"
and
account for the general improvement in maternal and child health over time. The counterfactual
15
Further details on the women, business, and the law index are provided online at
https://openknowledge.worldbank.org/handle/10986/32639#
16
Because the binary approach masks the distinctions between the magnitudes of different disasters, we only
record (binary variable=1) those disasters whose magnitude is bigger than the mean for that type of disaster data.
9

of an affected country, then, is the same country without the outbreak effect. If outbreaks
increase maternal and/or child mortality, we should observe an increase relative to the
country’s average levels in the indicator during the outbreak or in the period following the
outbreak.
Estimates of the effect of health disasters on maternal and child mortality exploiting cross-
sectional variability are likely to be biased upward (in absolute value). Empirically, the
magnitude of health disaster effect is larger among poor countries, ceteris paribus. We expect
the effect to be less or insignificant in HICs due to the greater amount of resources allocated to
prevention and mitigation efforts. Hence, we reestimate equation (1) by income level and
disaggregate countries to LMICs and HICs to attenuate this omitted variable bias.
We are also confident that our fixed-effects model additionally over protects against omitted-
variable bias. In particular, the effect of health disasters on the countries that have consistently
experienced epidemics and/or pandemics over our estimated time period is under-estimated as
pandemics and epidemics are largely part of the “fixed effect” of these countries. Since these
countries are also likely to be the most severely affected, the fixed-effects model may yield too
conservative estimates. This is accentuated by our use of a relatively short period framework.
Moreover, some countries may be poor at the start of our data series because of the disasters
they have experienced up to then. Ignoring this effect implies that our conservative estimates
are more likely not to detect an effect of disasters. But in fact our model produces a substantial
detrimental effect of disasters, especially as our time series is extended.
3.1.2. Long-term dynamic specification
Our specification in equation (1) assumes that the impact of health disasters on maternal and
child mortality is instantaneous. While most empirical studies make this assumption, we
hypothesize that the long-run effects of disasters are not to be underestimated. To allow for the
possibility of long-run effects of health disasters on maternal and child mortality, we propose
a long-term dynamic specification, where the lagged disasters (
.$%&3+
) and the lagged
determinants (
,$%&3+
), as well as current disasters (
.$%&
) and other determinants (
,$%&
), affect the
log of MMR, U5MR, and NMR. The lags of the logs of MMR, U5MR, and NMR (
#$%&3+
) are
included as explanatory variables to measure the persistence in the three health indicators. Our
model specification becomes as follows:
#$%& ' ()$ * 4#$%&3+ * (+,$%& * 5+,$%&3+ * (-.$%& * 5-.$%&3+ * /$* 0&* 1$&
2
(2)
According to this framework,
(-
captures the immediate short-run impact of a health disaster
on MMR, U5MR, and NMR. The long-run impact begins after a one-year lag and is given by
(-* 5-
6 7 4 2%
10

where
4
captures the persistence of the adjustment process, specifically the total adjustment of
the mortality rate following a health disaster.
While the lagged dependent variable is endogenous and typically correlates with the lagged
error term, requiring the residuals to sum to zero within countries implies that the errors are
correlated. Therefore, the obtained fixed-effects estimates of equation (2) will be biased and
inconsistent, especially with few time periods (Blundell et al., 2000; Cameron & Trivedi, 2005;
Wooldridge, 2002). Hence, we estimate equation (2) by the two-step Arellano-Bond GMM
estimator, first outlined by Arellano & Bover (1995) and developed by Blundell and Bond
(1998). Our two-step “system” GMM estimator has superior finite sample properties to handle
the issues of endogeneity of contemporaneous changes in the independent variables and the
endogeneity of the lagged level of MMR, U5MR, or NMR in the dynamic specification. The
two-step estimator combines the regression equation in differences and the regression equation
in levels into one system, where the lagged values of the explanatory variables are used as
instruments. It is properly designed for dynamic panels that may contain fixed effects and, in
addition to these fixed effects, idiosyncratic errors that are heteroskedastic and correlated
within but not across countries, as Roodman (2009) argued. Taking the residuals from equation
(2), we have the following moment conditions:
89:1$%& 7 1$%&3+;,$%&3<= ' >%222891$%&:,$%&3< 7 ,$%&3<3+;= ' >
(3)
Moreover, our estimator is properly designed for panel data with large number of cross sections
and a relatively short time series, which is the case in this study. To establish our moment
conditions, we assume that the disaster and time dummies are strictly exogenous and, hence,
serve as standard instrumental variables (IVs). The use of such IVs also helps reduce the
incidence of bias due to potential mis-measurement in our exogenous disaster variables. We
assume the rest of the current and lagged explanatory variables in equation (2) is potentially
endogenous; we construct our moment conditions for each of these variables for each lag length
from two and higher.
Diagnostic tests. As the GMM method we employ is, in its essence, an IV one, we consider the
Hansen test of exogeneity of instrument subsets to verify that our instruments are valid. We
also consider the Arellano-Bond autoregressive (AR) test for autocorrelation of the residuals
to verify that the differenced residuals do not exhibit significant AR(2) behavior. The former
test has low power if the number of moment conditions is large. So, we follow Roodman (2009)
and reduce the instrument count by specifying the use of only two lags in constructing the
GMM instruments. Limiting the number of lags prevents possible loss of efficiency that an
unrestricted set of lags, potentially introducing a huge number of instruments, can cause.
3.2. Estimating disaster mitigation effects
In additional model specifications, we examine if health system, macroeconomic, institutional,
and structural characteristics of countries struck by disease outbreaks have any bearing on the
11

magnitude of the health outcomes trajectory that typically follows. Specifically, we explore the
effectiveness of several factors in determining countries’ ability to mitigate the impacts of such
disasters. To do so, we extend equation (1) to include two more coefficients and examine the
significance of the coefficient on each mitigating factor (health service coverage, public health
emergency management, health emergency finance, WASH coverage, education, and gender
equality), denoted by
?$%&
, of countries struck by outbreaks. The following specification is
estimated:
#$%& ' ()$ * (+,$%& * (-.$%& * (@
A
.$%&2B ?$%&
C
* (D?$%& * /$* 0&* 1$&
2
(4)
The coefficient on the interaction term of disease outbreaks and mitigating factor,
(@
, is our
coefficient of interest as it measures the effect of each respective factor on the magnitude of
the change in maternal and child mortality associated with an outbreak.
We also account for the direct effect of the mitigating factors,
(D
, to verify that the significance
of the interaction coefficient is not a result of the direct correlation between these factors and
health indicators.
4. Results and discussion
4.1. Estimated short-run impact of health disasters
We first investigate the impacts of disease outbreaks on maternal and child health indicators
using our static model specification. Table 1 shows the results from the fixed-effects model
estimation in equation (1) based on the two reported measures of disease outbreaks: occurrence
of disasters and estimated number of people affected. The impacts of health disasters are
estimated on MMR, U5MR, and NMR. Our estimates indicate a strong detrimental effect of
health disasters. Columns 1, 3, and 5 indicate a significant positive impact of epidemics and
pandemics occurrence on increasing maternal, under-5, and neonatal mortalities by ~11%, 2%,
and 1%, respectively. Using the number of people affected measure of health disasters,
columns 2, 4, and 6 confirm the same significant positive impact on increasing maternal, under-
5, and neonatal mortalities by ~13%, 1%, and 0.1%, respectively.
Compared to the coefficients of the other economic, health, and education determinants in
Table 1, the magnitude of the negative impact of disease outbreaks on MMR is the highest and
the most significant. However, it is GDP per capita, which reflects the economic status of each
country, that has the highest impact on under-5 and neonatal mortalities. Globally, this provides
novel evidence that, over the last two decades, disease outbreaks can be considered the main
determinant of the surge or the weakened improvement in maternal mortality and one of the
effective determinants of child mortality.
12

Table 1. Estimated short-run impact of health disasters on MMR, U5MR, and NMR
(2000-2019) Dependent variables: MMR, U5MR, and NMR
MMR
U5MR
NMR
.$&=
Disaster
dummy
.$&
= Affected
people
.$&=
Disaster
dummy
.$&=
Affected
people
.$&=
Disaster
dummy
.$&=
Affected
people
(1)
(2)
(3)
(4)
(5)
(6)
Health disasters
Disaster dummy
10.743**
1.707**
0.616***
(4.534)
(0.826)
(0.165)
Affected people
13.445***
1.125***
0.137*
(2.568)
(0.410)
(0.083)
Other determinants
Economic status
GDP per capita
13.123
-14.634
-7.106**
-7.586**
-5.766***
-5.780***
(16.103)
(16.376)
(2.995)
(2.996)
(0.597)
(0.604)
Health
Physician density
2.847
-0.822
-1.903**
-1.993**
-0.346*
-0.342*
(5.927)
(5.700)
(0.925)
(0.923)
(0.184)
(0.186)
Adolescent fertility
2.304***
2.559***
(0.372)
(0.361)
Skilled-assisted delivery
-2.884***
-2.231***
(0.346)
(0.361)
Education
Primary education completion (F)
-1.376***
-1.297***
-1.903**
-1.993**
-0.346*
-0.342*
(0.290)
(0.279)
(0.925)
(0.923)
(0.184)
(0.186)
Constant
304.699**
450.892***
175.876***
179.649***
77.099***
77.527***
(128.491)
(125.870)
(23.591)
(23.576)
(4.702)
(4.751)
Country effects
Yes
Yes
Yes
Yes
Yes
Yes
Year effects
Yes
Yes
Yes
Yes
Yes
Yes
E-
0.748
0.767
0.683
0.684
0.748
0.743
Number of countries
81
81
111
111
111
111
Number of observations
360
360
794
794
794
794
Each column represents a separate regression. Standard errors are reported in parentheses. *, **, and *** denote
statistical significance at the 10%, 5%, and 1% levels, respectively. F: female.
Our further investigation in Table 2 differentiates between the impact of health disasters in
LMICs and in HICs. The estimates indicate that the observed global adverse impact of disasters
on maternal and child mortality is basically originating from LMICs rather than HICs. In HICs,
epidemics and pandemics appear to have no impact on MMR, U5MR, or NMR.
Next, we examine the extended model specifications that incorporate additional determinants
of maternal and child mortality that are commonly used in the empirical literature of maternal
and child health. These cover a set of health (anemia prevalence among women of reproductive
age, tuberculosis prevalence, DPT immunization, access to drinking water services, and access
to sanitation services), gender equality (the gender parity index), and ecological (urbanization
degree) determinants. We report the estimates from this exercise in Table A.2 (Appendix A).
The results align with our initial hypothesis that, in the time of health disasters, epidemics
and/or pandemics become the dominant determinants indirectly affecting maternal and child
mortality. Gender inequality comes second, and is particularly relevant in the case of MMR
and U5MR.
13

Table 2. Estimated short-run impact of health disasters on MMR, U5MR, and NMR in
LMICs versus HICs (2000-2019) Dependent variables: MMR, U5MR, and NMR
MMR
U5MR
NMR
LMICs
HICs
LMICs
HICs
LMICs
HICs
(1)
(2)
(3)
(4)
(5)
(6)
Health disasters
Disaster dummy
14.129**
-2.379
1.600*
0.231
0.588***
0.179
(6.142)
(1.972)
(0.959)
(0.412)
(0.192)
(0.223)
Other determinants
Economic status
GDP per capita
30.843
-15.274***
-1.606
-7.364***
-5.469***
-3.795***
(26.042)
(4.373)
(3.894)
(0.876)
(0.778)
(0.475)
Health
Physician density
1.981
0.828
-1.910*
2.701***
-0.362*
1.251***
(7.384)
(3.945)
(1.019)
(0.737)
(0.204)
(0.399)
Adolescent fertility
2.185***
-1.142***
(0.475)
(0.350)
Skilled-assisted delivery
-2.749***
-0.569
(0.434)
(0.594)
Education
Primary education completion (F)
-1.378***
-0.056
-0.662***
-0.019
-0.118***
-0.022*
(0.380)
(0.141)
(0.049)
(0.024)
(0.010)
(0.013)
Constant
187.231
278.026***
135.489***
83.398***
74.000***
44.766***
(197.314)
(68.843)
(28.435)
(8.238)
(5.683)
(4.461)
Country effects
Yes
Yes
Yes
Yes
Yes
Yes
Year effects
Yes
Yes
Yes
Yes
Yes
Yes
E-
0.765
0.656
0.738
0.703
0.785
0.684
Number of countries
69
12
93
18
93
18
Number of observations
259
101
587
207
587
207
Each column represents a separate regression. Standard errors are reported in parentheses. *, **, and *** denote
statistical significance at the 10%, 5%, and 1% levels, respectively. F: female.
4.2. Estimated long-run impact of health disasters
Table 3 lists the results of estimating our long-run dynamic specification in equation (2) for
LMICs; these are the two-step system GMM estimates. We include lags of both dependent and
independent variables, precisely GDP per capita, physician density, and urbanization degree
(for MMR), as instruments. Year dummies are included (but not reported) in all specifications
to control for period fixed effects. In Table 4, we report the estimated long-run effects of
changes in our explanatory variables of interest on MMR, U5MR, and NMR, indicating how
each parameter is calculated.
We find a significant effect of lagged maternal, child, and neonatal mortalities on each current
respective mortality. The large coefficients on the lags of MMR, U5MR, and NMR (>0.900)
suggest that maternal and child mortality is slow to adjust to occurrence of health disasters in
LMICs and tends to persist close to its previous levels (Table 3).
14

Table 3. Estimated dynamic impact of health disasters on MMR, U5MR, and NMR in
LMICs (2000-2019) Dependent variables: MMR, U5MR, and NMR
Parameter
GMM estimate
MMR
U5MR
NMR
(1)
(2)
(3)
Lag log mortality
4
0.981***
0.993***
0.934***
(0.001)
(0.003)
(0.003)
Health disasters
Disaster dummy
(-
0.003***
0.003***
0.002***
(0.001)
(0.000)
(0.001)
Lag disaster dummy
5-
0.004***
0.002***
0.015***
(0.002)
(0.001)
(0.001)
Other determinants
Economic status
Log GDP per capita
(++
-0.265***
-0.247***
-0.217***
(0.021)
(0.014)
(0.009)
Lag log GDP per capita
5++
0.253***
0.241***
0.186***
(0.021)
(0.014)
(0.010)
Health
Physician density
(+-
-0.027***
-0.007***
-0.034***
(0.002)
(0.001)
(0.001)
Lag physician density
5+-
-0.006***
-0.001**
0.002***
(0.001)
(0.000)
(0.000)
Constant
()$
0.139***
0.026
0.422***
(0.019)
(0.019)
(0.021)
Year effects
Yes
Yes
Yes
Arellano-Bond test for AR(2) in 1st differences
z-statistic
0.07
0.94
0.12
Pr > z =
0.941
0.346
0.902
Hansen difference test
F-
44.68
46.07
39.94
Pr > F- =
0.320
0.428
0.686
Number of countries
85
88
88
Number of observations
550
577
577
Columns (1), (2), and (3) represent separate regressions. Estimates are two-step system GMM ones.
Robust standard errors are reported in parentheses. *, **, and *** denote statistical significance at
the 10%, 5%, and 1% levels, respectively.
Table 4. Calculated long-run impact of health disasters on MMR, U5MR, and NMR in
LMICs from a dynamic model specification
Explanatory variable
Parameter calculation
Long-run effect estimate
MMR
U5MR
NMR
Long-run disaster dummy
:(-* 5-; :6 7 4
G;
0.353***
0.803**
0.258***
(0.116)
(0.320)
(0.020)
Long-run log GDP per capita
:(++ * 5++; :6 7 4
G;
-0.598***
-0.934***
-0.468***
(0.147)
(0.203)
(0.019)
Long-run physician density
:(+- * 5+-; :6 7 4
G;
-1.712***
-1.264***
-0.486***
(0.098)
(0.442)
(0.029)
Robust standard errors are reported in parentheses. *, **, and *** denote statistical significance at the
10%, 5%, and 1% levels, respectively.
Significant contemporaneous and long-run effects of health disasters are observed for all three
of our health indicators. Specifically, we estimate that occurrence of health disasters increases
MMR, U5MR, and NMR, respectively, by about 0.3%, 0.3%, and 0.2% immediately in LMICs
(Table 3). While these effects are significant, the magnitude of the long-run effects is much
larger, and about 35%, 80%, and 26%, respectively, after a year (Table 4). The lagged values
of the disaster dummy are highly significant for all three mortality ratios (Table 3), confirming
15

an appropriate long-term dynamic specification (Table 4). As anticipated, both GDP per capita
and physician density have significant short and long-run effects on maternal and child
mortality in LMICs (Tables 3 and 4).
The speed of convergence to the long-run steady state, given by (
6 7 4
), is around 1.9%, 0.7%,
and 6.6% for maternal mortality, under-5 mortality, and neonatal mortality, respectively, over
a one-year period. This reflects how health disasters are hindering LMICs’ efforts in meeting
international adequate levels of maternal and child mortalities.
As the employed GMM method relies on IVs, we test the validity of the used instruments. For
this purpose, the Hansen test for over-identified restrictions is reported in Table 3, with the null
hypothesis being that the used instruments are valid. As shown in Table 3, we fail to reject the
null hypothesis, verifying that our instruments are valid. We also report the Arellano-Bond
autoregressive (AR) test for autocorrelation of the residuals in Table 3. Failing to reject the null
hypothesis suggests that the differenced residuals do not exhibit significant AR(2) behavior.
4.3. Estimated effects of disaster mitigating factors
The results presented in Table 5 provide evidence on the determinants of the magnitude of
previously identified impact of health disasters on maternal and child mortality in LMICs.
Specifically, we estimate equation (4) to test if UHC, public health emergency management,
health emergency finance, WASH coverage, education, and gender equality in the countries
struck by disasters have any bearing on the magnitude of maternal and child mortality surge.
The reported coefficient on the interaction of the disaster measure and the mitigation variable
in equation (4) defines the effect of a mitigating factor on the magnitude of the mortality impact
indicated in Table 1.
Estimating the role of public health emergency preparedness in mitigating disaster effects, we
find that the health system capacity to detect, assess, notify, report, and respond to public health
emergencies, summarized by the health system capacity variable, significantly mitigates health
disaster effects, especially on MMR and U5MR. Preparedness is the core system capacity with
the highest effect. In LMICs, an increase in the preparedness index by 1 unit is associated with
0.3, 0.1, and 0.01 decrements in the impact of a health disaster on maternal, under-5, and
neonatal mortality, respectively. Risk communication comes second, with an increase in its
index by 1 unit being associated with 0.2 and 0.05 decrements in the impact of a health disaster
on MMR and U5MR, respectively. These results suggest that health emergency management
(surveillance, response, preparedness, risk communication, etc.) policies can partially absorb
the negative effects of health disasters on maternal and child mortality in LMICs. Epidemic
and pandemic preparedness especially can have substantially high returns through developing
public health emergency response plans.
Ordering the mitigating effects of various factors by effectiveness, GDP per capita comes first
as the main mitigator of disaster effects, surpassing all sources of domestic and international
health emergency finance in LMICs (ODA, personal remittances, and national savings). A 1%
16

increase in a country’s GDP per capita can absorb the disaster effect on maternal, under-5, and
neonatal mortalities by 26%, 5%, and 0.4%, respectively. Countries with lower GDP per capita
typically face greater difficulty in dealing with disasters due to budgetary restrictions; a
government that is cash strapped can hardly mobilize resources to respond to disease outbreaks.
Two results are worth noting in this regard: financial inclusion significantly mitigates the
disaster effect on MMR and U5MR; and channeling resources through remittances
significantly mitigates the disaster effect on U5MR. The latter result suggests that reducing
remittance transaction costs as well as strengthening the financial infrastructure that supports
remittances can be of particular relevance to mitigating a disaster effect on child mortality.
Health system related factors come second after GDP per capita as key mitigators of disaster
effects. An intuitive result is the significant mitigating effect of physician density on mortality
in all settings. In LMICs, an increase in the UHC index by 1 unit is associated with a 4.2 and
0.7 decrements in the health disaster effect on maternal and under-5 mortality, respectively.
The reported results suggest that creating UHC systems is a priority investment during COVID-
19 and beyond. Moving toward UHC can address barriers to health care access and alleviate
any undue financial burden, with spillover effects on health system resilience to health
disasters.
Education and gender equality come in third place. An increase in the gender parity index by
1 unit is associated with 2, 0.5, and 0.02 decrements in the impact of a health disaster on
maternal, under-5, and neonatal mortality, respectively. Pointing in the same direction, an
increase in the women, business, and the law index by 1 unit is associated with 1, 0.2, and 0.02
decrements in the impact of a health disaster on maternal, under-5, and neonatal mortality,
respectively. Moreover, an increase in female primary education completion rate by 1% is
associated with 1% and 0.03% decrements in the impact of a health disaster on maternal, under-
5, and neonatal mortality, respectively. These results suggest that improving female education
helps make people resilient to shocks and better cope with stresses these shocks bring; this
strengthened resilience lessens the direct and indirect impact of disasters on health and
wellbeing. We also conclude that gender inequality is an important factor of disaster mitigation
efforts. More unequal societies tend to have fewer resources allocated to mitigation as they are
unable to resolve the collective action problem of implementing mitigating measures.
Estimating the disaster mitigating effects of WASH coverage, we find that increasing access to
safe drinking water, safely managed sanitation, and basic handwashing facilities by 1% absorbs
the adverse effect of disasters on MMR by almost 1%. Drinking water and sanitation appear to
have a significant mitigating effect on U5MR, while handwashing has a greater effect on NMR.
These results suggest that the heterogeneity in health infrastructure between countries matters
for controlling and mitigating the effect of disease outbreaks. In the aftermath of disasters,
especially in LMIC settings, breakdowns in infrastructure, and specifically failures of basic
hygiene, sanitation, and/or running water systems, leave populations susceptible to elevated
health risks from disease outbreaks.
17

Comparing the disaster mitigating effects of several factors, we confirm that while public
health emergency management plays a significant role, but it is secondary to the rest of the
explored factors.
We reestimate equation (4) exploiting our pooled dataset including HICs and report the results
using a disaster occurrence dummy and estimated total affected, respectively, in Tables A.3
and A.4 (Appendix A). The estimates verify our main findings that strengthening public health
emergency management, raising GDP per capita, increasing physician density, investing in
UHC, promoting gender equality, improving female education, and expanding WASH
coverage can significantly mitigate the negative impact of health disasters. Such factors can
make countries better able to withstand an initial disaster shock and, importantly, prevent
further spillovers on maternal and child health. The impacts are stronger for LMICs than HICs.
18

Table 5. Mediating factors of health disaster effects on MMR, U5MR, and NMR in
LMICs (2000-2019) Dependent variables: MMR, U5MR, and NMR
.$&= Disaster dummy
MMR
U5MR
NMR
Health service access/coverage
.$&* UHC
-4.155*** (0.844)
-0.689*** (0.150)
-0.056 (0.034)
UHC
1.980 (2.158)
0.350 (0.382)
0.059 (0.087)
.$&
2
198.927*** (40.951)
32.019*** (7.258)
3.222* (1.655)
.$&* Physician density
-14.217*** (3.517)
-2.944*** (0.628)
-0.086 (0.128)
Physician density
-14.916** (5.780)
-1.604 (0.999)
-0.353* (0.204)
.$&
-7.375 (7.256)
-2.619** (1.300)
0.465* (0.266)
Public health emergency management
.$&* Health system capacity
-0.261* (0.152)
-0.084*** (0.032)
-0.005 (0.008)
Health system capacity
0.114 (0.134)
0.010 (0.026)
0.003 (0.007)
.$&
11.442 (9.621)
4.830** (1.984)
0.345 (0.507)
.$&* Surveillance
-0.269 (0.162)
-0.061* (0.034)
-0.004 (0.008)
Surveillance
0.001 (0.101)
0.013 (0.021)
0.004 (0.005)
.$&
17.064 (13.145)
4.552 (2.790)
0.396 (0.675)
.$&* Response
-0.211 (0.138)
-0.077*** (0.029)
-0.011 (0.007)
Response
0.222** (0.097)
0.044** (0.020)
0.011** (0.005)
.$&
12.104 (10.225)
5.339** (2.146)
0.960* (0.519)
.$&* Preparedness
-0.300*** (0.105)
-0.085*** (0.022)
-0.009* (0.005)
Preparedness
0.114 (0.072)
0.025* (0.015)
0.007* (0.004)
.$&
11.604* (6.425)
4.195*** (1.325)
0.587* (0.331)
.$&* Risk communication
-0.197* (0.108)
-0.048** (0.023)
-0.008 (0.006)
Risk communication
-0.025 (0.083)
0.004 (0.018)
0.005 (0.004)
.$&
8.290 (7.822)
2.817* (1.661)
0.629 (0.403)
Health emergency finance
.$&* ODA-health sector
-2.833 (2.962)
-0.057 (0.526)
0.140 (0.107)
ODA-health sector
-6.923** (2.795)
-1.903*** (0.493)
0.010 (0.100)
.$&
21.769** (10.687)
1.791 (1.904)
0.149 (0.386)
.$&* Remittances
-1.249 (0.912)
-0.360** (0.167)
-0.025 (0.033)
Remittances
0.468 (0.739)
-0.020 (0.135)
-0.066** (0.027)
.$&
15.247** (6.619)
2.907** (1.207)
0.782*** (0.242)
.$&* GDP per capita
-25.651*** (5.103)
-4.540*** (0.916)
-0.442** (0.187)
GDP per capita
-40.870* (21.988)
-0.831 (3.803)
-5.394*** (0.775)
.$&
198.233*** (37.239)
34.369*** (6.681)
3.782*** (1.361)
.$&* Adjusted net savings
0.117 (0.306)
-0.023 (0.066)
0.029** (0.014)
Adjusted net savings
0.209 (0.355)
0.026 (0.075)
-0.037** (0.016)
.$&
9.956** (4.933)
0.736 (1.061)
0.489** (0.220)
.$&
* Financial inclusion
-1.450* (0.746)
-0.393*** (0.130)
-0.002 (0.005)
Financial inclusion
2.513** (1.097)
0.555*** (0.185)
-0.039** (0.016)
.$&
17.195** (6.665)
4.469*** (1.158)
0.107 (0.113)
WASH coverage
.$&* Access to drinking water
-1.023*** (0.253)
-0.210*** (0.045)
-0.024 (0.028)
Access to drinking water
-3.038*** (0.695)
-0.608*** (0.124)
-0.006 (0.040)
.$&
86.067*** (18.339)
16.510*** (3.285)
0.409 (0.253)
.$&* Access to safely managed sanitation
-1.207*** (0.412)
-0.401*** (0.116)
-0.005 (0.009)
Access to safely managed sanitation
-1.146 (0.713)
0.341* (0.200)
-0.120*** (0.025)
.$&
43.368*** (14.114)
12.191*** (3.960)
1.005 (0.665)
.$&* Access to handwashing facilities
-0.549* (0.279)
-0.069 (0.043)
-0.058*** (0.018)
Access to handwashing facilities
0.218 (0.595)
0.180** (0.091)
0.030 (0.031)
.$&
33.796*** (10.610)
5.098*** (1.621)
1.641*** (0.623)
Education and gender equality
.$&* Primary education completion (F)
-1.146*** (0.203)
-0.252*** (0.036)
-0.006 (0.007)
Primary education completion (F)
-3.046*** (0.288)
-0.535*** (0.050)
-0.001 (0.015)
.$&
99.085*** (16.150)
20.511*** (2.841)
0.712** (0.273)
.$&* Gender parity index (education)
-1.886*** (0.605)
-0.546*** (0.107)
-0.017** (0.007)
Gender parity index (education)
-3.722*** (0.785)
-0.588*** (0.136)
-0.109*** (0.010)
.$&
186.593*** (57.237)
52.260*** (10.082)
1.864*** (0.594)
.$&* Women, business, & the law index
-1.164*** (0.414)
-0.158** (0.075)
0.023 (0.021)
Women, business, & the law index
-0.378 (0.499)
-0.048 (0.089)
-0.235*** (0.027)
.$&
90.762*** (27.827)
12.167** (5.041)
-1.811 (1.983)
Each row represents a separate regression. Standard errors are reported in parentheses. *, **, and *** denote statistical
significance at the 10%, 5%, and 1% levels, respectively. GDP per capita, physician density, primary education
completion (F), and country and year effects are included in all estimations. F: female.
19

5. Conclusion
This study provides novel, global evidence on the short- and long-run impacts of epidemic and
pandemic disease outbreaks on maternal and child mortality across 111 countries over the
period 2000-2019 in HICs and LMICs. The evidence provides crucial knowledge that should
guide future interventions.
Our fixed-effects results show that health disasters have a significant impact on maternal,
under-5, and neonatal mortality worldwide, that the impact is significant in LMICs, and that
maternal mortality is the most impacted. We hypothesize that this health disaster effect takes
place through repeated depletion or diversion of resources providing routine primary care,
which distort the overall efficiency of the health system and pose serious threats to health
service utilization. Such depletion and diversion occurred during the Ebola outbreak in West
Africa (2014-2016); the current COVID-19 pandemic is likely to lead to similar effects (2019-
present). Our GMM results show that, in LMICs, health disasters increase maternal, under-5,
and neonatal mortalities by 0.3%, 0.3%, and 0.2% instantaneously and by 35%, 80%, and 26%
after one year, respectively.
Exploring various mitigating factors of disaster effect, our findings provide pertinent evidence
on how LMICs can strengthen their resilience to disease outbreaks by investing in the most
effective disaster mitigation strategies. We confirm that public health emergency management,
especially preparedness and risk communication, can play a significant role in preventing
negative spillover effects from health disasters to maternal and child health. However, its role
is secondary to the rest of the factors, namely health emergency finance, UHC, education and
gender equality, and WASH coverage. The obtained evidence is stronger for LMICs.
Our findings of the instrumental role of epidemic and pandemic preparedness have direct
policy implications. They are consistent with the view that much health spending in LMICs is
poorly targeted or otherwise ineffective. Hence, we construct evidence-based policy toolkits
for maternal and child health. To effectively absorb disaster effects on maternal mortality,
LMICs governments should invest more to ensure UHC, promote equal education
opportunities, increase women’s access to quality and affordable financial services, close
financing gaps to achieve universal access to safely managed sanitation, drinking water, and
handwashing facilities, and improve women’s economic opportunities and empowerment. As
for child health, the discussed policy toolkit should be augmented by personnel remittances as
an effective mitigating factor, suggesting that reducing remittance transaction costs as well as
strengthening the financial infrastructure that supports remittances is key to mitigating disaster
effects on child mortality. In conclusion, our findings suggest that governments, donors, and
development partners should target broad development goals within their public health
emergency response plans.
20

References
Akseer, N., Kandru, G., Keats, E. C., & Bhutta, Z. A. (2020). COVID-19 pandemic and
mitigation strategies: Implications for maternal and child health and nutrition. The
American Journal of Clinical Nutrition. https://doi.org/10.1093/ajcn/nqaa171
Anand, S., & Bärnighausen, T. (2004). Human resources and health outcomes: Cross-country
econometric study. The Lancet, 364(9445), 1603–1609. https://doi.org/10.1016/s0140-
6736(04)17313-3
Arellano, M., & Bover, O. (1995). Another look at the instrumental variable estimation of
error-components models. Journal of Econometrics, 68(1), 29–51.
https://doi.org/10.1016/0304-4076(94)01642-d
Berhan, Y., & Berhan, A. (2014). Skilled health personnel attended delivery as a proxy
indicator for maternal and perinatal mortality: A systematic review. Ethiopian Journal
of Health Sciences, 24(0), 69. https://doi.org/10.4314/ejhs.v24i0.7s
Blundell, R., & Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel
data models. Journal of Econometrics, 87(1), 115–143. https://doi.org/10.1016/s0304-
4076(98)00009-8
Blundell, R., Bond, S., & Windmeijer, F. (Eds.). (2000). Estimation in dynamic panel data
models: Improving on the performance of the standard GMM estimator. Amsterdam
New York and Tokyo: Elsevier Science, JAI.
Brabin, B. J., Hakimi, M., & Pelletier, D. (2001). An analysis of anemia and pregnancy-related
maternal mortality. The Journal of Nutrition, 131(2), 604S-615S.
https://doi.org/10.1093/jn/131.2.604s
Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: Methods and applications.
Cambridge: Cambridge University Press.
Chirowa, F., Atwood, S., & Van der Putten, M. (2013). Gender inequality, health expenditure
and maternal mortality in sub-Saharan Africa: A secondary data analysis. African
Journal of Primary Health Care & Family Medicine, 5(1).
https://doi.org/10.4102/phcfm.v5i1.471
Chowdhury, A. H., Hanifi, S. M. A., & Bhuiya, A. (2020). Social determinants of under-five
mortality in urban Bangladesh. Journal of Population Research, 37(2), 161–179.
https://doi.org/10.1007/s12546-019-09240-x
Daru, J., Zamora, J., Fernández-Félix, B. M., Vogel, J., Oladapo, O. T., Morisaki, N., … Khan,
K. S. (2018). Risk of maternal mortality in women with severe anaemia during
pregnancy and post partum: A multilevel analysis. The Lancet Global Health, 6(5),
e548–e554. https://doi.org/10.1016/S2214-109X(18)30078-0
Delamou A., El Ayadi AM., Sidibe S., Delvaux T., Camara BS., Sandouno SD., 2017. Effect
of Ebola virus disease on maternal and child health services in Guinea: A retrospective
observational cohort study. Lancet Glob Health. 5:e448-57.
Ehrhardt, A. A., Sawires, S., McGovern, T., Peacock, D., & Weston, M. (2009). Gender,
empowerment, and health: What is it? How does it work? JAIDS Journal of Acquired
Immune Deficiency Syndromes, 51(supplement 3), S96–S105.
https://doi.org/10.1097/qai.0b013e3181aafd54
21

Ensor, T., Cooper, S., Davidson, L., Fitzmaurice, A., & Graham, W. J. (2010). The impact of
economic recession on maternal and infant mortality: Lessons from history. BMC
Public Health, 10(1). https://doi.org/10.1186/1471-2458-10-727
Farahani, M., Subramanian, S. V., & Canning, D. (2009). The effect of changes in health sector
resources on infant mortality in the short-run and the long-run: A longitudinal
econometric analysis. Social Science & Medicine, 68(11), 1918–1925.
https://doi.org/10.1016/j.socscimed.2009.03.023
Frankenberg, E., Sikoki, B., Sumantri, C., Suriastini, W., & Thomas, D. (2013). Education,
vulnerability, and resilience after a natural disaster. Ecology and Society, 18(2).
https://doi.org/10.5751/es-05377-180216
Godefay, H., Byass, P., Graham, W. J., Kinsman, J., & Mulugeta, A. (2015). Risk factors for
maternal mortality in rural Tigray, Northern Ethiopia: A case-control study. PLOS
ONE, 10(12), e0144975. https://doi.org/10.1371/journal.pone.0144975
Hanmer, L., Lensink, R., & White, H. (2003). Infant and child mortality in developing
countries: Analysing the data for robust determinants. Journal of Development
Studies, 40(1), 101–118. https://doi.org/10.1080/00220380412331293687
Harville, E., Xiong, X., & Buekens, P. (2010). Disasters and perinatal health: A systematic
review. Obstetrical & Gynecological Survey, 65(11), 713–728.
https://doi.org/10.1097/ogx.0b013e31820eddbe
Henry, M. W., & Chen, L. C. (1984). An analytical framework for the study of child survival
in developing countries. Population and Development Review, 10, 25–45.
https://doi.org/10.2307/2807954
Independent Accountability Panel for Every Women Every Child Every Adolescent (IAP).
Caught in the COVID-19 storm: Women’s, children’s, and adolescents’ health in the
context of UHC and the SDGs. Geneva: IAP, 2020. https://iapewec.org/reports/annual-
reports/iap-2020-report.
Iyengar, P., Kerber, K., Howe, C. J., & Dahn, B. (2015). Services for mothers and newborns
during the Ebola outbreak in Liberia: The need for improvement in emergencies. PLoS
Currents.
https://doi.org/10.1371/currents.outbreaks.4ba318308719ac86fbef91f8e56cb66f
Jamison, D. T., Murphy, S. M., & Sandbu, M. E. (2016). Why has under-5 mortality decreased
at such different rates in different countries? Journal of Health Economics, 48, 16–25.
https://doi.org/10.1016/j.jhealeco.2016.03.002
Khan, Y., O’Sullivan, T., Brown, A., Tracey, S., Gibson, J., Généreux, M., … Schwartz, B.
(2018). Public health emergency preparedness: A framework to promote
resilience. BMC Public Health, 18(1). https://doi.org/10.1186/s12889-018-6250-7
Kruk, M. E., Kujawski, S., Moyer, C. A., Adanu, R. M., Afsana, K., Cohen, J., …Yamey, G.
(2016). Next generation maternal health: External shocks and health-system
innovations. The Lancet, 388(10057), 2296–2306. https://doi.org/10.1016/s0140-
6736(16)31395-2.
Makinson, C. (1985). The health consequences of teenage fertility. Family Planning
Perspectives, 17(3), 132. https://doi.org/10.2307/2135024
22

Matthews, Z., Channon, A., Neal, S., Osrin, D., Madise, N., & Stones, W. (2010). Examining
the “urban advantage” in maternal health care in developing countries. PLoS
Medicine, 7(9), e1000327. https://doi.org/10.1371/journal.pmed.1000327
McCarthy, J., & Maine, D. (1992). A framework for analyzing the determinants of maternal
mortality. Studies in Family Planning, 23(1), 23. https://doi.org/10.2307/1966825
Menendez, C., Gonzalez, R., Donnay, F., & Leke, R. G. F. (2020). Avoiding indirect effects
of COVID-19 on maternal and child health. The Lancet Global Health, 8(7), e863–
e864. https://doi.org/10.1016/s2214-109x(20)30239-4
Moran, N. F., & Moodley, J. (2012). The effect of HIV infection on maternal health and
mortality. International Journal of Gynecology & Obstetrics, 119, S26–S29.
https://doi.org/10.1016/j.ijgo.2012.03.011
Noy, I. (2009). The macroeconomic consequences of disasters. Journal of Development
Economics, 88(2), 221–231. https://doi.org/10.1016/j.jdeveco.2008.02.005
O’Hare, B., Makuta, I., Chiwaula, L., & Bar-Zeev, N. (2013). Income and child mortality in
developing countries: A systematic review and meta-analysis. Journal of the Royal
Society of Medicine, 106(10), 408–414. https://doi.org/10.1177/0141076813489680
Przyluski, V., & Hallegatte, S. (2010). Climate change adaptation, development, and
international financial support: Lessons from EU pre-accession and solidarity funds.
SSRN Electronic Journal. https://doi.org/10.2139/ssrn.1710476
Raddatz, C. (2007). Are external shocks responsible for the instability of output in low-income
countries? Journal of Development Economics, 84(1), 155–187.
https://doi.org/10.1016/j.jdeveco.2006.11.001
Roberton, T., Carter, E. D., Chou, V. B., Stegmuller, A. R., Jackson, B. D., Tam, Y., … Walker,
N. (2020). Early estimates of the indirect effects of the COVID-19 pandemic on
maternal and child mortality in low-income and middle-income countries: A modelling
study. The Lancet Global Health, 8(7), e901–e908. https://doi.org/10.1016/s2214-
109x(20)30229-1
Roodman, D. (2009). How to do Xtabond2: An introduction to difference and system GMM in
Stata. The Stata Journal: Promoting Communications on Statistics and Stata, 9(1), 86–
136. https://doi.org/10.1177/1536867x0900900106
Sochas, L., Channon, A. A., & Nam, S. (2017). Counting indirect crisis-related deaths in the
context of a low-resilience health system: The case of maternal and neonatal health
during the Ebola epidemic in Sierra Leone. Health Policy and Planning, 32(suppl_3),
iii32–iii39. https://doi.org/10.1093/heapol/czx108
Temmerman, M., Khosla, R., Bhutta, Z. A., & Bustreo, F. (2015). Towards a new global
strategy for women’s, children’s and adolescents’ health. BMJ, h4414.
https://doi.org/10.1136/bmj.h4414
Wooldridge, J. M. (2002). Econometric analysis of cross section and panel data. Cambridge,
MA: MIT Press.
World Health Organization. World Health Statistics 2020. May 2020.
https://www.who.int/gho/publications/world_health_statistics/en/.
Zaba, B., Calvert, C., Marston, M., Isingo, R., Nakiyingi-Miiro, J., Lutalo, T., … Ronsmans,
C. (2013). Effect of HIV infection on pregnancy-related mortality in sub-Saharan
23

Africa: Secondary analyses of pooled community-based data from the network for
Analysing Longitudinal Population-based HIV/AIDS data on Africa (ALPHA). The
Lancet, 381(9879), 1763–1771. https://doi.org/10.1016/s0140-6736(13)60803-x
24

Appendix A
Table A.1. List of included countries
Low income
Middle income
High income
Afghanistan
Albania
Malaysia
Germany
Burkina Faso
Angola
Maldives
Israel
Burundi
Argentina
Marshall Islands
Italy
Central African Republic
Bangladesh
Mauritania
Korea, Rep.
Chad
Benin
Mexico
Kuwait
Congo, Dem. Rep.
Bolivia
Mongolia
Latvia
Ethiopia
Botswana
Morocco
Mauritius
Gambia, The
Cabo Verde
Myanmar
Palau
Guinea
Cambodia
Namibia
Panama
Guinea-Bissau
Cameroon
Nepal
Romania
Liberia
China
Nicaragua
Saudi Arabia
Madagascar
Colombia
Nigeria
Seychelles
Malawi
Comoros
North Macedonia
Singapore
Mali
Congo, Rep.
Pakistan
Spain
Mozambique
Costa Rica
Papua New Guinea
Sweden
Niger
Cote d’Ivoire
Paraguay
Switzerland
Rwanda
Dominican Republic
Peru
United Kingdom
Sierra Leone
Ecuador
Philippines
United States
Sudan
Egypt, Arab Rep.
Russian Federation
Tajikistan
El Salvador
Samoa
Togo
Eswatini
Sao Tome and Principe
Uganda
Fiji
Senegal
Yemen, Rep.
Ghana
Solomon Islands
Guatemala
South Africa
Honduras
Sri Lanka
India
Tanzania
Indonesia
Thailand
Iran, Islamic Rep.
Timor-Leste
Iraq
Tonga
Jamaica
Turkey
Kazakhstan
Vanuatu
Kenya
Venezuela, RB
Kyrgyz Republic
Vietnam
Lao PDR
Zambia
Lesotho
Zimbabwe
25

Table A.2. Estimated short-run impact of health disasters on MMR, U5MR, and NMR
(2000-2019)
Dependent variables: MMR, U5MR, and NMR
MMR
U5MR
NMR
.$&=
Disaster
dummy
.$&=
Affected
people
.$&=
Disaster
dummy
.$&
=
Affected
people
.$&=
Disaster
dummy
.$&=
Affected
people
Health disasters
Disaster dummy
8.777**
1.793**
0.495***
(3.997)
(0.730)
(0.146)
Affected people
6.955***
0.759**
0.060
(2.451)
(0.353)
(0.071)
Other determinants
Economic status
GDP per capita
7.770
-3.155
-1.651
-1.833
-4.070***
-4.032***
(14.279)
(14.794)
(2.948)
(2.957)
(0.591)
(0.597)
Health
Physician density
6.737
3.539
-1.270
-1.272
-0.230
-0.208
(5.509)
(5.528)
(0.829)
(0.831)
(0.166)
(0.167)
Adolescent fertility
1.584***
1.761***
(0.366)
(0.373)
Skilled-assisted delivery
-2.040***
-1.768***
(0.307)
(0.327)
Anemia prevalence (women)
2.711***
1.996**
(0.789)
(0.817)
Incidence of tuberculosis
0.309***
0.296***
(0.040)
(0.040)
Immunization, DPT
-0.353***
-0.351***
(0.053)
(0.053)
Access to drinking water
-0.951***
-0.927***
-0.166***
-0.159***
(0.106)
(0.106)
(0.021)
(0.021)
Access to sanitation
0.288***
0.261***
-0.044**
-0.051***
(0.085)
(0.085)
(0.017)
(0.017)
Education
Primary education completion (F)
-0.850***
-0.848***
-0.422***
-0.423***
-0.060***
-0.060***
(0.283)
(0.281)
(0.044)
(0.044)
(0.009)
(0.009)
Gender equality
Gender parity index (education)
-3.103***
-3.151***
-0.735***
-0.737***
-0.215***
-0.219***
(0.688)
(0.671)
(0.107)
(0.107)
(0.020)
(0.020)
Ecological setting
Urbanization
1.800**
1.572*
0.255
0.256
0.100***
0.103***
(0.890)
(0.890)
(0.174)
(0.174)
(0.035)
(0.035)
Constant
343.060**
439.014***
247.197***
249.016***
88.479***
88.647***
(146.613)
(145.967)
(24.392)
(24.431)
(4.882)
(4.926)
E-
0.792
0.795
0.771
0.770
0.809
0.806
Number of countries
81
81
110
110
110
110
Number of observations
351
351
764
764
764
764
Each column represents a separate regression. Standard errors are reported in parentheses. *, **, and *** denote statistical
significance at the 10%, 5%, and 1% levels, respectively.2.$&2is the EM-DAT health-disaster. Country and year effects
are included in all estimations. F: female.
26

Table A.3. Mediating factors of health disaster effects on MMR, U5MR, and NMR (2000-
2019) Dependent variables: MMR, U5MR, and NMR
.$&= Disaster dummy
MMR
U5MR
NMR
Health service access/coverage
.$&
* UHC
-3.113*** (0.639)
-0.528*** (0.117)
-0.052** (0.026)
UHC
0.818 (1.527)
-0.147 (0.280)
-0.080 (0.062)
.$&
2
166.484*** (33.448)
27.286*** (6.138)
3.114** (1.365)
.$&* Physician density
-13.150*** (2.683)
-2.831*** (0.504)
-0.164 (0.103)
Physician density
-14.591*** (4.973)
-1.558* (0.906)
-0.326* (0.184)
.$&
-2.549 (5.258)
-1.526 (0.992)
0.428** (0.202)
Public health emergency management
.$&* Health system capacity
-0.311** (0.135)
-0.096*** (0.030)
-0.009 (0.008)
Health system capacity
0.155 (0.107)
0.024 (0.023)
0.008 (0.006)
.$&
16.384* (8.538)
6.102*** (1.868)
0.752 (0.496)
.$&* Surveillance
-0.266* (0.144)
-0.061* (0.032)
-0.004 (0.008)
Surveillance
-0.072 (0.079)
-0.006 (0.018)
-0.000 (0.005)
.$&
18.913 (11.744)
5.087* (2.647)
0.585 (0.676)
.$&* Response
-0.244** (0.122)
-0.089*** (0.027)
-0.014** (0.007)
Response
0.226*** (0.079)
0.050*** (0.017)
0.013*** (0.004)
.$&
15.759* (9.020)
6.582*** (2.001)
1.276** (0.510)
.$&* Preparedness
-0.337*** (0.093)
-0.096*** (0.020)
-0.013** (0.005)
Preparedness
0.101* (0.056)
0.023* (0.012)
0.007** (0.003)
.$&
15.381*** (5.687)
5.249*** (1.245)
0.907*** (0.327)
.$&* Risk communication
-0.266*** (0.096)
-0.065*** (0.021)
-0.012** (0.005)
Risk communication
0.044 (0.066)
0.018 (0.015)
0.007* (0.004)
.$&
15.026** (6.944)
4.446*** (1.554)
1.014** (0.395)
Health emergency finance
.$&
* ODA-health sector
-1.946 (2.731)
0.029 (0.491)
0.115 (0.099)
ODA-health sector
-5.890** (2.572)
-1.531*** (0.458)
0.048 (0.093)
.$&
18.991* (9.811)
1.613 (1.768)
0.274 (0.358)
.$&* Remittances
-0.925 (0.774)
-0.255* (0.149)
-0.010 (0.030)
Remittances
0.334 (0.640)
-0.085 (0.123)
-0.075*** (0.025)
.$&
12.876** (5.238)
2.295** (1.007)
0.701*** (0.202)
.$&* GDP per capita
-15.938*** (3.130)
-3.097*** (0.592)
-0.361*** (0.120)
GDP per capita
-36.434** (16.090)
-5.678* (2.949)
-5.599*** (0.596)
.$&
133.243*** (24.124)
25.171*** (4.555)
3.348*** (0.920)
.$&* Adjusted net savings
0.117 (0.255)
-0.025 (0.058)
0.024** (0.012)
Adjusted net savings
0.294 (0.278)
0.085 (0.062)
-0.019 (0.013)
.$&
9.551** (4.055)
0.885 (0.915)
0.502*** (0.189)
.$&* Financial inclusion
-0.425 (0.492)
-0.097 (0.093)
0.025 (0.019)
Financial inclusion
-0.702* (0.411)
-0.241*** (0.077)
-0.069*** (0.016)
.$&
11.803** (5.571)
2.921*** (1.046)
0.200 (0.219)
WASH coverage
.$&* Access to drinking water
-0.948*** (0.196)
-0.196*** (0.036)
-0.007 (0.007)
Access to drinking water
-3.924*** (0.556)
-0.905*** (0.103)
-0.174*** (0.021)
.$&
83.757*** (14.919)
16.363*** (2.771)
1.237** (0.558)
.$&
* Access to safely managed sanitation
-0.410*** (0.150)
-0.124*** (0.042)
-0.018** (0.007)
Access to safely managed sanitation
-1.202** (0.478)
-0.024 (0.135)
-0.047* (0.024)
.$&
25.891*** (8.342)
6.374*** (2.353)
0.831** (0.416)
.$&* Access to handwashing facilities
-0.549* (0.279)
-0.069 (0.043)
-0.006 (0.007)
Access to handwashing facilities
0.218 (0.595)
0.180** (0.091)
-0.001 (0.015)
.$&
33.796*** (10.610)
5.098*** (1.621)
0.712** (0.273)
Education and gender equality
.$&* Primary education completion (F)
-1.168*** (0.167)
-0.257*** (0.031)
-0.021*** (0.006)
Primary education completion (F)
-3.230*** (0.229)
-0.627*** (0.042)
-0.124*** (0.009)
.$&
103.695*** (13.712)
21.759*** (2.548)
2.231*** (0.529)
.$&* Gender parity index (education)
-1.940*** (0.506)
-0.554*** (0.094)
0.010 (0.019)
Gender parity index (education)
-3.816*** (0.648)
-0.653*** (0.118)
-0.224*** (0.023)
.$&
192.777*** (48.259)
53.487*** (8.956)
-0.517 (1.781)
.$&* Women, business, & the law index
-1.104*** (0.316)
-0.167*** (0.060)
-0.028** (0.012)
Women, business, & the law index
-0.045 (0.379)
0.002 (0.071)
0.044*** (0.014)
.$&
87.046*** (21.770)
13.019*** (4.147)
2.480*** (0.823)
Each row represents a separate regression. Standard errors are reported in parentheses. *, **, and *** denote statistical
significance at the 10%, 5%, and 1% levels, respectively. GDP per capita, physician density, primary education
completion (F), and country and year effects are included in all estimations. F: female.
27

Table A.4. Mediating factors of health disaster effects on MMR, U5MR, and NMR (2000-
2019) Dependent variables: MMR, U5MR, and NMR
.$&= Affected people
MMR
U5MR
NMR
Health service access/coverage
.$&
* UHC
-1.474 (0.928)
-0.321* (0.167)
-0.007 (0.036)
UHC
0.808 (1.611)
-0.126 (0.290)
-0.087 (0.063)
.$&
2
44.569** (21.469)
9.197** (3.870)
0.332 (0.844)
.$&* Physician density
-0.900 (2.251)
-0.150 (0.428)
0.046 (0.086)
Physician density
-16.755** (5.045)
-2.001** (0.924)
-0.340* (0.186)
.$&
5.127 (6.224)
0.735 (1.184)
0.255 (0.238)
Public health emergency management
.$&* Health system capacity
1.160* (0.700)
0.202 (0.161)
0.044 (0.042)
Health system capacity
0.037 (0.100)
-0.008 (0.022)
0.004 (0.006)
.$&
-63.690 (43.013)
-10.344 (9.907)
-2.043 (2.572)
.$&* Surveillance
1.876 (1.212)
0.354 (0.273)
0.057 (0.070)
Surveillance
-0.149** (0.072)
-0.024 (0.016)
-0.002 (0.004)
.$&
-154.477 (105.033)
-28.409 (23.690)
-4.099 (6.024)
.$&* Response
1.113* (0.654)
0.170 (0.148)
0.013 (0.037)
Response
0.117* (0.069)
0.013 (0.016)
0.007* (0.004)
.$&
-82.269 (52.234)
-11.544 (11.859)
-0.269 (2.987)
.$&* Preparedness
0.514 (0.755)
0.189 (0.169)
0.028 (0.043)
Preparedness
0.005 (0.054)
-0.005 (0.012)
0.003 (0.003)
.$&
-16.313 (34.277)
-6.251 (7.657)
-0.457 (1.929)
.$&* Risk communication
0.244 (0.498)
0.069 (0.112)
-0.015 (0.028)
Risk communication
-0.061 (0.064)
-0.008 (0.014)
0.004 (0.004)
.$&
-9.569 (33.985)
-2.525 (7.617)
1.726 (1.919)
Health emergency finance
.$&
* ODA-health sector
-3.187** (1.508)
-0.404 (0.273)
-0.077 (0.056)
ODA-health sector
-5.580** (2.504)
-1.430*** (0.447)
0.090 (0.092)
.$&
16.668*** (5.218)
2.162** (0.944)
0.344* (0.193)
.$&* Remittances
0.867 (1.300)
0.160 (0.251)
0.049 (0.051)
Remittances
0.265 (0.641)
-0.101 (0.123)
-0.077*** (0.025)
.$&
-1.506 (5.223)
-0.376 (1.008)
-0.122 (0.204)
.$&* GDP per capita
-8.762** (3.443)
-1.416** (0.655)
-0.013 (0.133)
GDP per capita
-49.925*** (16.293)
-7.998*** (2.994)
-5.784*** (0.605)
.$&
60.349*** (20.894)
9.673** (3.977)
0.214 (0.804)
.$&* Adjusted net savings
-0.274*** (0.102)
-0.045* (0.023)
-0.001 (0.005)
Adjusted net savings
0.134 (0.269)
0.062 (0.060)
-0.022* (0.013)
.$&
4.190* (2.135)
0.593 (0.488)
0.131 (0.103)
.$&* Financial inclusion
0.127 (0.312)
0.032 (0.059)
0.014 (0.012)
Financial inclusion
-0.791* (0.411)
-0.263*** (0.077)
-0.072*** (0.016)
.$&
0.198 (2.497)
-0.016 (0.473)
-0.048 (0.099)
WASH coverage
.$&* Access to drinking water
-0.337* (0.186)
-0.056 (0.035)
-0.000 (0.007)
Access to drinking water
-4.106*** (0.560)
-0.950*** (0.104)
-0.171*** (0.021)
.$&
26.264** (10.548)
4.253** (1.968)
0.161 (0.395)
.$&
* Access to safely managed sanitation
-11.006*** (2.460)
-2.802*** (0.697)
-0.352*** (0.124)
Access to safely managed sanitation
-1.182** (0.469)
-0.034 (0.133)
-0.050** (0.024)
.$&
470.440*** (105.696)
122.174*** (29.955)
16.295*** (5.332)
.$&* Access to handwashing facilities
1.193 (1.981)
-0.273 (0.302)
0.026 (0.051)
Access to handwashing facilities
0.186 (0.615)
0.177* (0.094)
-0.001 (0.016)
.$&
-52.184 (88.247)
12.526 (13.461)
-1.115 (2.262)
Education and gender equality
.$&* Primary education completion (F)
-0.309*** (0.092)
-0.052*** (0.018)
-0.003 (0.004)
Primary education completion (F)
-3.721*** (0.219)
-0.740*** (0.041)
-0.135*** (0.008)
.$&
23.502*** (5.250)
3.807*** (0.999)
0.306 (0.202)
.$&* Gender parity index (education)
-0.777*** (0.268)
-0.128** (0.051)
-0.006 (0.010)
Gender parity index (education)
-4.771*** (0.582)
-0.941*** (0.107)
-0.222*** (0.021)
.$&
75.214*** (24.037)
12.224*** (4.547)
0.584 (0.892)
.$&* Women, business, & the law index
-0.536 (0.476)
-0.065 (0.091)
0.006 (0.018)
Women, business, & the law index
-0.105 (0.377)
-0.008 (0.071)
0.044*** (0.014)
.$&
42.966 (31.640)
5.443 (6.020)
-0.271 (1.204)
Each row represents a separate regression. Standard errors are reported in parentheses. *, **, and *** denote statistical
significance at the 10%, 5%, and 1% levels, respectively. GDP per capita, physician density, primary education
completion (F), and country and year effects are included in all estimations. F: female.
28