Human fertility after a disaster: a royalsocietypublishing.org/journal/rspb systematic literature review D. Susie Lee1, Ewa Batyra1,2, Andres Castro1,2 and Joshua Wilde1,3 1 Fertility and Well-being, Max-Planck-Institute for Demographic Research, 18057 Rostock, Mecklenburg- Vorpommern, Germany Evidence 2 Centre for Demographic Studies (CED), Barcelona, 08193, Spain synthesis 3 Institute of Labor Economics (IZA), 53113 Bonn, Germany Cite this article: Lee DS, Batyra E, Castro A, DSL, 0000-0002-0761-0069; EB, 0000-0002-2967-1508; AC, 0000-0003-1032-3869 Wilde J. 2023 Human fertility after a disaster: a Fertility is a key demographic parameter influenced by disaster. With the systematic literature review. Proc. R. Soc. B growing risk of disasters, interest in the fertility response to a disaster is 290: 20230211. increasing among the public, policy makers and researchers alike. As yet, https://doi.org/10.1098/rspb.2023.0211 a synthesis of the current evidence on how fertility changes after disaster Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 does not exist. We reviewed 50 studies retrieved from a systematic search based on a pre-registered protocol. We found an overall negative impact of disasters on fertility. If any, increases in fertility were mostly linked Received: 26 January 2023 with weather-related physical disasters. We also identified 13 distinct mech- Accepted: 6 April 2023 anisms which researchers have considered as underlying the fertility effects of disaster. By contrast to the common belief that disasters are more likely to increase fertility in contexts with already high fertility, we found little evi- dence to suggest that the total fertility rate of the studied populations was an important predictor of the direction, timing or size of fertility impacts. Subject Category: While this may be because no relationship exists, it may also be due to Behaviour biases we observed in the literature towards studying high-income countries or high-cost disasters. We summarize the methodological limitations ident- Subject Areas: ified from the reviewed studies into six practical recommendations for future environmental science, biological applications, research. Our findings inform both the theories behind the fertility effects of disasters and the methods for studying them. behaviour Keywords: fertility, disaster, humans, systematic literature 1. Introduction review, birth counts Disasters can significantly affect the lives of those who experience them. Beyond the proximate perils of displacement, death, disease and injury, disasters are often accompanied by broader challenges such as economic crisis, reduced access to health facilities and other services, and food insecurity. Since child- Author for correspondence: bearing is one of the most consequential choices people make, it is not D. Susie Lee surprising that disasters affect whether and when people have children, and e-mail: lee@demogr.mpg.de how many. With the growing frequency and severity of disasters [1,2] especially due to the global climate crisis [3], how disasters affect fertility has become a topic of interest increasingly shared among the public, policy makers and researchers alike [4,5]. Yet in spite of the importance of and broad interest in the topic, a synthesis of the current evidence on how disasters affect fertility (the number and timing of births) does not yet exist. Doing so is a complex task, for multiple reasons. First, ‘disaster’ is a blanket term for many types of catastrophic events, each of which may vary from one another in their impact on fertility. Second, even within disaster type, the direction and magnitude of the effects thereof differ depending on population characteristics [6]. For example, it was pre- dicted that the COVID-19 pandemic would lead to a reduction in births in high-income, low-fertility contexts, but to an increase in low-income, high- fertility contexts [7]. Third, disasters can have different effects in the short, Electronic supplementary material is available medium and long runs. Indeed, the very first scientific observation on online at https://doi.org/10.6084/m9.figshare. the topic, made in 1892 [8], noted that fertility dropped around 9–10 months c.6607452. © 2023 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. (i.e. the average human gestation length) after the spike of knowledge—no systematic reviews are available. The primary 2 mortality during the 1889–1890 influenza pandemic, and outcome of interest was live birth counts, a measure with a direct impact on population structure and size. We excluded royalsocietypublishing.org/journal/rspb then increased above and eventually converged to the expected average. Although this sequence of initial trough, other reproductive health measures beyond live birth counts, given already existing reviews on the sexual and reproduc- rebound and stabilization mirrors the temporal dynamics of tive health consequences of disasters [21] or adverse birth mortality crises [9], we do not yet know the degree of support outcomes [22–27]. for similar fertility dynamics within the existing evidence We selected studies that were: (a) either published or in press base. Fourth, the mechanisms underlying fertility change in peer-reviewed journals as of early February 2022, (b) on are numerous and complex, making it difficult to differentiate human populations regardless of publication year and specific pathways by which disasters affect fertility [10]. languages, as long as (c) information on the direction or magni- Lastly, methodological differences hinder the comparison of tude (or both) of the relationship between disaster and fertility fertility effects across studies. This is particularly important was reported. We restricted studies to those on (d) the relatively since methodological quality determines the degree to acute impact of a disaster, not extending to studies that had which an association between disaster and fertility affords assessed exposure to post-disaster environments created by a dis- Proc. R. Soc. B 290: 20230211 causal interpretation, given that experimental study on this aster itself, such as displacement camps or food insecurity, or the intergenerational impact of disaster on fertility. We focused on topic is nearly impossible. studies which examined (e) one of the disaster types—both natu- We conducted a systematic review [11] of studies that ral and technological—listed in the International Disasters quantified the change in live birth counts after disasters, to Database [28] and (f ) fertility in terms of live birth counts, Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 provide a comprehensive view of the state-of-the-art evidence excluding studies that used a combined measure of stillbirth in this field. Building on existing reviews or continent-level and live birth [29]. Lastly, we reviewed (g) studies that analysis on other catastrophic events such as wars and con- operationalized disaster as a relatively concentrated and/or flicts, humanitarian crises and economic recession [12–16], unanticipated exposure, as explained below. we reviewed fertility change after natural or technological disasters. They included disease outbreaks, physical disasters (e.g. earthquake, weather-related events) and technological (b) Operationalizing exposure to disaster Different identifications of exposure yield different findings and disasters (e.g. radiation events, poisoning, explosion). We interpretations of the impact of an exposure [17]. In the present use the term ‘disease outbreak’ to refer to community-level study, we focused on studies that operationalized disaster as a disease occurrence including epidemics and pandemics, simi- ‘concentrated’—meaning having a major impact on many lar to the approach taken by the Center for Disease Control communities—and/or ‘unanticipated’—meaning seemingly (https://www.cdc.gov/outbreaks/index.html). We focused random, non-cyclic and extreme—exposure to entire commu- on studies that operationalized disasters as relatively concen- nities. Doing so would allow interpreting the impact of a trated and unanticipated exposure [17] experienced at the disaster as a population-level fertility response to significantly and community-level, and thus excluded those that measured unexpectedly abrupt disruptions to routine. This operationalization exposure at individual-level or with a continuous scale (see of disaster had three implications for the scope of this review. Methods for our eligibility criteria). First, exposure to disaster is not reduced to individual-level Based on the literature identified through a predefined exposure. For example, we omitted studies that compared ferti- lity between infected and uninfected individuals, as the protocol, we analysed (1) how disaster–fertility relationships literature on the fertility effect of HIV commonly does [30,31]. have been studied in terms of spatio-temporal coverage and The current review thus approached fertility from a demographic methodology, and (2) the direction, magnitude and mechan- perspective, where fertility is a process by which populations isms by which disaster affects fertility. Given the diversity of replace themselves, as opposed to a strictly clinical perspective. methodological approaches and measures employed by the Second, exposure to disaster should not be intermittent (e.g. sea- reviewed studies, we deemed the synthesis of effect sizes sonal heat waves) or persistent (e.g. endemic diseases). In this via meta-analysis infeasible. We instead opted to present a regard, identifying a specific disaster onset should be reasonably narrative synthesis accompanied by statistical analysis as possible. Third, exposure to disaster is measured entirely at the needed, and offer methodological recommendations for population-level. We also excluded studies that measured future research. exposure on a continuous scale, such as degree of rainfall [32], seismic intensity scale [33], or the prevalence of a disease [34], which would not allow interpreting the impact of a disaster event as a whole. 2. Methods The protocol for this review was pre-registered at PROSPERO (ID: CRD42022320294) before data extraction and analysis. (c) Literature search and study selection Details on the database search procedure and results, and the In February 2022, we searched databases and conducted an data extraction template and data extracted can be found on additional search in non-English databases for which we had the project website (https://osf.io/dpgqf/). language expertise and access. We used Boolean operators, whenever they were supported by databases, to combine search terms for each disaster and fertility (e.g. ‘hurricane AND (a) Scope and eligibility criteria fertility’). We searched for the presence of any combinations The scope of our review was determined by our two core con- either in the title or abstract. A total of 12 661 studies were cepts—disaster [exposure] and fertility [outcome]. Disaster is a found, of which 4190 deduplicated unique studies were screened difficult concept to define [18] and ranges from everyday emer- (figure 1). We used Covidence, an online record management gencies, economic shocks, conflicts and wars, to various types system for systematic review [35], to conduct screening and of catastrophic events which may entail distinct patterns of data extraction. First, the title and abstract of each article disruptions [19,20]. The present review focused on natural and were read by two researchers, who independently rated technological disasters, for which—to the best of our whether or not a study met the eligibility criteria (yes, no or 3 identification of studies via database searches royalsocietypublishing.org/journal/rspb Records identified from*: identification Scopus (n = 3316) Pubmed (n = 2054) Records removed before Proquest (n = 1353) screening: EBSCO-Host (n = 1825) Duplicate records removed World of Science (n = 4113) (n = 8471) Supplementary1 (n = 91) Records screened for abstract 3851 records not meeting (n = 4190) eligibility criteria were excluded. Proc. R. Soc. B 290: 20230211 Reports sought for retrieval Access to full-text was not (n = 339) available for 4 records.2 Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 screening Reports assessed for full-text 285 records not meeting (n = 335) eligibility criteria were excluded. 1The following non-English databases were searched: Russian Science Citation Index, SCIELO, Korean Citation Index. 2Pérez (2021) ‘The Spanish flu in Bilbao. Demographic analysis of the pandemic and its consequences from different information Studies included in review sources’, Oberson (2010) ‘Environmental included (n = 50, including 2 articles pollutions and effects on reproduction: A added from backward citation glance at the past’, Scragg (1977) ‘Pollution search3) and sex ratio of births’, Lyster (1981) ‘Pollution and sex ratio of births’ 3We identified 44 new articles cited in the 48 articles initially selected, and found 2 of the 44 to meet the eligibility criteria, hence the total number of finally selected articles was 50. Figure 1. Flow chart of studies selection process. maybe). Only the studies unanimously voted either ‘yes’ or (e) Fertility effect of disaster ‘maybe’ entered the full-text screening. Second, the full In this review, a fertility effect refers to ‘changes in live birth text of each article was read by two researchers and rated in counts after a disaster’. If an effect is reported as a numeric reference to the eligibility criteria. In both screening procedures, value, then it conveys both information about the direction and any discrepancies in assessments were discussed and resolved the magnitude of effect (i.e. effect size). The direction of an by the whole team. Third, based on the 48 articles chosen for effect can be smaller [negative] or larger [ positive] than 0, or review, we extracted information on study characteristics, not different from the control group [null]. Where possible, we methods and findings. Data were extracted twice each by extracted standardized effect sizes (e.g. percentage difference or two researchers, and any discrepancies were resolved in change). If there were multiple effect sizes, we identified a sum- discussion among them. Lastly, during the data extraction, mary effect, which is either the effect size reported from an we identified 44 potentially relevant additional articles author’s preferred analysis or the effect size mentioned as the through backward citation search. Two of these studies met the main finding of a study. eligibility criteria and were included for review. We did not per- From the 50 studies, we extracted 108 data points on fertility form a forward citation search, since studies on the COVID-19 effects, of which 82 were summary effects. Out of the 108 effects, pandemic are rapidly increasing. We identified 50 studies only 76 had further information about magnitude because some (figure 1, table 1) which quantified how live birth counts studies did not report fertility effects in numeric forms for mul- change after disasters. tiple reasons. For example, a multi-country study reported effect values from the study’s preferred analysis for only select countries with large and statistically significant effect [51], or, (d) Data and analyses in some studies, fertility of case versus control groups was The unit of analysis was either studies (figure 2), countries reported side by side across multiple time points without a matched with disasters (figure 3), effect sizes (figures 4–6), or single summary effect [40,52–54]. For these studies, we could mechanisms suggested by studies (figure 7). While there were still extract information on effect direction based on the authors’ 50 studies reviewed, not all studies covered just one country or conclusion and/or visualization of effects. We could not extract reported only one effect size or mechanism, leading to different any information on effects from one study [55] that only com- samples depending on the unit of analysis. pared post-disaster fertility of the affected population with the 1.00 mechanisms that could be behavioural or physiological depend- 4 ing on contexts: conception given intercourse would be affected behaviourally via (no. 6) contraceptive use but also physiologi- royalsocietypublishing.org/journal/rspb reviewed studies by year of publication cumulative proportion studied disasters by year of disaster occurrence 0.75 cally via (no. 10) infertility, and successful delivery given conception would be affected behaviourally via (no. 7) abortion but also physiologically via (no. 11) miscarriage. 0.50 Although it is challenging to integrate the 13 mechanisms into one framework, we follow the approach taken by a previous review on fertility after wars, humanitarian crises and displace- 0.25 ment [13]. Here, different pathways by which fertility changes were seen as a combination of indirect (distal) and direct ( prox- imate) determinants of fertility, following the frameworks 0 proposed by Davis & Blake [57] and later by Bongaarts [58]. 1880 1900 1920 1940 1960 1980 2000 2020 Among the 13 mechanisms we identified, the proximate mechan- year isms which can affect fertility directly are (i) intercourse that is Proc. R. Soc. B 290: 20230211 Figure 2. Temporal distributions of the 50 studies reviewed and disasters driven by partnership dynamics (no. 5), (ii) conception given intercourse (no. 6,10) and (iii) successful delivery given con- covered by the studies. The x-axis refers to the year of publication (solid ception (no. 7,11). Partnership dynamics can influence fertility line) or disaster occurrence (dashed line), and starts from 1880 because either through marital or coital rate; conception given intercourse the earliest disaster covered by the reviewed studies was the 1882 measles can be affected by contraceptive use or women’s ability to con- Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 outbreak in Iceland. The year range of disasters covered by the 50 studies is ceive (i.e. fecundity). Mechanisms affecting fertility through wider than the year range of publications, reflecting the recent increase in the successful delivery given conception encompass different types body of works that quantified fertility changes after disaster. More than 75% of pregnancy termination (both related to the decision to termi- of studies were published post 2000. nate - through abortion- and involuntary termination, for example miscarriage). The remaining mechanisms are either indirect ones (no. 1–4, 9) or ones that cannot be easily classified unaffected population, making it impossible to separate the as direct or indirect (no. 8, 12–13). For example, economic con- impact of disaster from any pre-existing difference between the ditions are a clear example of an indirect mechanism through compared populations. which disasters can affect the proximate determinants mentioned It was not feasible to summarize effect sizes via meta-analy- above. Conversely, a broad group of mechanisms frequently sis, due to differences in research designs, effect size types, mentioned in the studies, which we call ‘postponement/family duration of effect measurement, as well as the unavailability of size reduction’, is not easily classifiable, as it could be an out- information necessary for meta-analysis. Instead, we conducted come of an interplay of both indirect determinants (e.g. various alternative analyses. First, we used a vote-counting economic conditions) and proximate determinants (e.g. partner- method [56], which compares the numbers of studies with posi- ship dynamics) included in mechanisms groups no. 2–7. tive versus negative effects using a binomial/sign test. The null Similarly, preterm births and reproductive age mortality (i.e. hypothesis states the equal chance of positive versus negative either women or men of reproductive age) cannot be easily effects based on binomial distribution. All available information classified according to the proximate determinants framework. on the direction of point estimates regardless of statistical signifi- cance entered vote-counting, because vote-counting can be misleading if a subjective decision (e.g. an arbitrary alpha level for statistical significance) is used to determine a vote. Second, we used Chi-square tests and log-linear models to examine 3. Results differences in the counts of positive versus negative effects by disaster types and countries’ total fertility rate (TFR) at the year (a) What has been studied, and how? of disaster. Third, we analysed effect direction by time lag (i) Temporal and geographical scope of reviewed studies from disaster to the beginning of outcome measurement, first The 50 studies were published between 1975 and 2022 and by pooling all effect directions reported in the studies and then covered 24 different disasters, which occurred between 1882 by confining to 13 studies that examined fertility effect across (the measles outbreak in Iceland) and 2020 (the COVID-19 multiple post-disaster time points. We calculated the time lag pandemic) across 52 countries. Our study sample is biased in months. Lastly, we examined patterns in effect sizes among toward studies that examined certain disaster types and those reported in comparable metrics, specifically % change in regions. Twenty-two out of the 50 studies (44%) were on fertility adjusted for seasonality and time trends in fertility. There were 18 such effect sizes. either the Zika epidemic, the 1918 influenza pandemic, or the COVID-19 pandemic, and the majority of these studies were based on samples from Brazil, Italy, Japan and the US. (f ) Mechanisms In this regard, 44% of the research on the disaster–fertility Out of the 50 studies, 36 postulated at least one mechanism nexus has thus far focused on disease outbreaks from underlying the fertility effects of disaster. middle- and high-income countries. Indeed, 10 out of the We classified mechanisms into 13 groups, which were also 50 studies covered the US, and more than half (52%) covered categorized as either behavioural or physiological mechanisms either Brazil, Italy, Japan or the US. (electronic supplementary material, tables S1 and S2). Of the The literature is relatively young, as most works were 13, the behavioural mechanisms were: child/community replace- published over the last two decades (solid line, figure 2). ment effect (no. 1), economic conditions/uncertainty (no. 2), migration (no. 3), psychological (no. 4), partnership dynamics More recent disasters were more likely to be studied, with (no. 5) and postponement/reduction of family size (no. 8). The more than half of the studies covering post-2000 disasters physiological mechanisms were: disruption of reproductive ser- (dashed line, figure 2). This trend resulted in shorter vices (no. 9), preterm births (no. 12), and mortality of men or time lags between disaster and publication for more recent women of reproductive age (no. 13). There were two additional disasters, such as the COVID-19 pandemic and the Zika disaster type / / IRQ 5 7 2.0 disease outbreak IRN technological disasters royalsocietypublishing.org/journal/rspb total fertility rate at disaster occurrence physical disasters 6 1.8 5 1.6 ISL 4 The TFR was 4.3 1.4 when the measles outbreak occurred in Iceland in year 3 1882 1.2 2 1.0 Proc. R. Soc. B 290: 20230211 KOR 1 27 articles (53%) 48 country–year (56%) 0.8 // 1880 1900 1920 1960 1980 2000 2020 2010 2015 2020 Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 year of disaster occurrence year of disaster occurrence Figure 3. For more recent disasters, the disaster–fertility relationship is studied in a low fertility context. Each marker represents a country matched with a disaster and the country-level total fertility rate (TFR; y-axis) at the time of disaster occurrence (x-axis). Three markers distinguish disaster types (disease outbreaks, technological disasters, and physical disasters). To show the demographic context of each studied disaster, grey lines represent the time series of TFR five years before and after the disaster. For some disaster–country combinations we found more than one study. These include the 2015 Zika epidemic in Brazil (six papers), the Covid-19 pandemic in Italy (four papers) and the US (two papers), the 2011 Japanese earthquake (three studies), the 1918 influenza epidemic in the US (three papers), and the 2010 Chilean earthquake, 2004 Indonesian Tsunami, and the Chernobyl accident in Ukraine and Belarus (two papers). This information on the number of papers is not reflected in the graph. TFR data before 1950 comes from Statista (interpolation was conducted to fill single years series). TFR data after 1950 comes from the World Population Prospects 2022. epidemic. Such a trend may reflect an increase in both the shows that the concentration is among low and relatively interest in the topic and data availability, in particular for stable fertility contexts represented by countries in the individual-level and panel data [89]. Global North (figure 3, right panel). Most of these countries However, the recent expansion of the literature does not are included in multi-country studies. One multi-country correspond to the actual distribution of recent disasters and study that contributed significantly to this pattern [51] exem- disaster risks. For example, our search did not identify plifies a trend that favours multi-country comparative studies on several natural disasters that occurred during analyses—especially in high-income countries—based on 2000–2019 with more than 100 000 casualties [2], notably the availability of harmonized and comparable data such as the 2008 Cyclone Nargis in Myanmar, the 2005 Pakistan the Human Fertility Database (https://www.humanferti- earthquake and the heatwaves in Europe (2003) and Russia lity.org) and the Short-Term Fertility Fluctuations Data (2010). The deadliest natural disaster of the last two decades, Series (www.humanfertility.org/Data/STFF). Studies on the 2004 Indian Ocean Tsunami, affected 12 countries in Asia more recent disasters (especially post-2010) also concern rela- and Africa. Still, only two studies in the reviewed literature tively low fertility contexts with TFR below two (figure 3, examined the fertility consequences of this disaster, and right panel), driven by research on the COVID-19 pandemic only in Indonesia [84,85]. in high-income economies and the Zika epidemic in Brazil To further understand the demographic context of the lit- and Colombia. Among the whole 50 studies, there are very erature, we examined the distribution of countries studied few studies conducted in settings with TFRs above four according to country-level total fertility rate (TFR) and year (figure 3, left panel); among them, most studies were on of disaster occurrence. Figure 3 displays which countries physical or technological disasters that occurred in the were covered by the 50 studies, according to the disaster Global South. year and the country-level total fertility rate (TFR) at the time of disaster. The lines trace the time series of a country’s TFR 5 years before and after the disaster. Figure 3 shows that (ii) Methods for measuring fertility effect of disaster our study sample covers a wide range of TFR levels (from 0.8 In nearly two-thirds of the studies, one population serves as to 7.0) and temporal trends. Fertility exhibits substantial its own comparison group (within-population comparison; declines before and after the disasters—particularly for table 1). Fertility was compared (i) between before (control) those before 2010—making it important to consider time and after (case) the disaster or (ii) between what is expected trends of fertility within study designs. This is also important under the assumption of no disaster (control) and actually for disasters occurring in low fertility contexts, where fertility observed (case). In the rest of the studies, at least two popu- continues to fluctuate yearly. lations, one not affected (control) and the other affected (case) Figure 3 speaks again to the strong concentration of by the disaster were compared before and after the disaster studies in recent years (also figure 2), and additionally (between-population comparison). In most of these studies, total fertility rate groups 6 [0.8,1.5) [1.5,2) [2,3) [3,7.1] royalsocietypublishing.org/journal/rspb (a) all disaster types (b) disease outbreak (c) physical disasters (d) technological disasters 59 effects 42 effects number of summary effects 18 effects Proc. R. Soc. B 290: 20230211 9 effects 10 effects 9 effects 7 effects 1 effects 0 effects 1 effects 0 effects 0 effects Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 negative null positive negative null positive negative null positive negative null positive effect direction Figure 4. Directions of effect by country-level total fertility rate (TFR), for all disasters (a), disease outbreaks (b), physical disasters (c), and technological disasters (d ). To assess the potential relationship between the effects’ direction and the fertility context, bars are divided according to country-level TFR groups. Due to small sample size and to warrant consistency of cross tabulations, the Chi-squared test ignored rows and columns with no counts. The figure is based on 78 effect sizes available with information on effects direction, regardless of their magnitude and statistical significance. From the 82 summary effect sizes we had, four were not used because they were either associated with multiple countries (i.e. no single TFR could be assigned) [36,37] or they did not have corresponding TFR information for the year of disaster occurrence [38,39]. the subnational area(s) hit hard by a disaster served as the search for the disaster), to inductively identify the month ‘affected’ group, then compared with other subnational at which the exposure (or the perceived exposure) peaked areas or the national-level average. [36,41,42,61]. Studies employed three types of effect sizes to quantify the difference in fertility between case and control. Counts or rates were the most common type. The fertility measure (b) How does fertility change after disaster? of choice was calculated for each of the case and control (i) Fertility effects groups over a certain period of time, and then compared The effect of disasters on fertility was generally negative, either in terms of raw differences or ratios. Only some of even when controlling for secular time trends. Out of the these studies reported information on size of the population 108 effect sizes for which information on direction was avail- from which the births were counted (i.e. person-years of able, 77 were negative and 26 were positive. According to the exposure) for both case and control—the information binomial test of the assumption that the counts of negative required for meta-analysis of effect sizes on counts or rates and positive effects are even, the proportion of negative data. Other less common ways to calculate effect sizes effects was 0.66 (95% confidence interval [CI] = 0.57 to 0.79, included correlating mortality at its peak with fertility 9–10 p = 0.006). Similarly, among the subset of 81 summary effect months after [38,39,61] or calculating differences between sizes (effect sizes upon which each study based its main con- average births of case versus control groups [51,59,78,82], clusion), the proportion of negative effects was 0.71 (95% either in the aggregate or in rates. CI = 0.59 to 0.82, p < 0.001). The tendency for negative effects The majority of the reviewed studies (84%) presented became weaker when we further restricted to 46 effect sizes information on the onset of a disaster at least in monthly that adjusted for time trends, but the majority of effect sizes units, a prerequisite for defining the post-disaster period. (59%) were still negative (95% CI = 0.43 to 0.73, p = 0.302). The onset of a disaster was defined in two ways. First, Effect directions by disaster type and fertility context. Figure 4 researchers defined the onset of disaster prior to observing displays the distribution of fertility effects for all disaster data, by referring to the time point when a disaster is types ( pooled) and for each disaster type separately. To known to have created an acute and major shock. Infor- assess the potential relationship between the effects’ direction mation on the timing was available a priori, if (i) a disaster and the fertility context, bars are divided according to the is unanimously said to have occurred on a specific day or country-level TFR at the time of disaster. Negative effect across a few days (e.g. earthquake) or (ii) the researcher sizes dominated the studies on disease outbreaks and techno- makes an explicit assumption about when the disaster logical disasters, but not the studies on physical disasters. likely had the most impact on fertility. The latter was a fre- Negative and positive effects were reported across all TFR quent choice if there was a lag between the emergence of levels at similar proportions (figure 4a), and the same was hazards and their development into a disaster, as in the true within disease outbreaks (figure 4b) and physical disas- case of the COVID-19 pandemic where the lockdowns are ters (figure 4c). This comparison was not possible within often regarded as its onset of impact in respective countries. technological disasters, for which no positive effects were Second, some studies analysed monthly data on proxies of reported (figure 4d ). We further examined the counts of exposure to disaster (e.g. disease incidence rate, Internet effects by direction ( positive versus negative), TFR and nine effect sizes reporting positive effect 9 months after a disaster onset 7 royalsocietypublishing.org/journal/rspb positive one effect size reporting null effect 24 months after a disaster onset effect direction null Proc. R. Soc. B 290: 20230211 negative Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 0 3 6 9 12 15 18 24 36 48 60 months since disaster Figure 5. Directions of fertility effects reported at different time lags from disaster onset. Each grey circle represents an effect size available with information on the effect’s direction, regardless of the effect’s magnitude and statistical significance. Effect sizes are classified into either positive [fertility increase], negative [fertility decline], or no change (y-axis), and aligned by time lag of post-disaster fertility follow-up (x-axis). Assuming that follow-up began randomly, we can expect that the temporal distribution of positive, negative, or zero effects will give ideas about the temporal dynamics of disaster impact. One model of such dynamics often invoked is ‘initial fertility decline followed by fertility rebound’. We used all 105 effect sizes for which information on time at effect measurement was either reported or could be estimated at monthly intervals. disaster type using log-linear models. Among the models 9.5% increase. From these studies, two interesting patterns with different combinations of variables, the model with emerge. First, for studies that examine precisely the same con- the interaction between TFR and disaster type yielded the text and disaster, effect sizes and directions agree very closely. highest goodness of fit as suggested by lowest model accu- One example is the 2011 Japanese earthquake, where two racy metrics (electronic supplementary material, table S3). studies found a 9–10% reduction in fertility. Second, for This finding is likely driven by the effect sizes from physical studies that analyse the same disaster and country context disasters, because the majority (78%) of positive effects but different sub-national regions, effect sizes for smaller reported from physical disasters pertained to contexts with areas tend to be larger in absolute terms. For example, four the TFR above 2.0 (figure 4c). These results again suggest studies examine the effect of Zika in Brazil: a study on the that the direction of effects may not differ by the TFR of the Northeast region alone found a 25% decline in births, whereas studied populations. a broader analysis of the 36 largest cities found only a 7.7% Effect directions by time points. Figure 5 shows the direction decline, and two yet broader analysis (one of all municipalities of the fertility effects ( positive, null and negative) by time with at least one case, and another at the national level) found since the disaster. Follow-ups began on average 9 months a very similar 1.7% and 1.9% decline. This is possibly due to from the disaster onset (interquartile range = 6–10 months). publication bias towards studies which find large effects— Negative effects were most commonly reported between 0 regional analyses with null effects are less likely to be pub- and 12 months after the disaster, especially at 0 (i.e. at the lished, leaving only those with large effects. Similar patterns onset) and 9 months after the disaster. There were less posi- are found among the studies on the 1918 influenza pandemic tive or null effects reported. Nonetheless, both negative and in the US and COVID-19 in Italy. positive effects were most commonly reported at 9–10 months from the disaster onset. Among the 13 studies that reported effect sizes across (c) Mechanisms multiple time points since the onset of a disaster (figure 6), Among the 50 reviewed studies, 36 (72%) referred to at least only one study on the 1918 influenza in India reported a pat- one possible mechanism to explain the observed effects of dis- tern of fertility decline, rebound above baseline, and asters on fertility. Figure 7 shows how frequently studies stabilization [43]. In five studies births declined then invoked (if they did any) behavioural or physiological mechan- rebounded; however, exact time points at which negative isms to explain findings. For example, for disease outbreaks, and positive effects were reported varied across the studies studies considered behavioural and physiological mechanisms [36,38,40,43,44]. at a similar frequency, whereas physiological mechanisms Magnitude of fertility effects. There were 18 effect sizes that were considered more often for technological disasters. It reported effects as percentage changes and adjusted for should be noted that, although studies on the two disaster seasonality and time trends in fertility (electronic supplemen- types are shown to concern with mechanisms underlying tary material, table S4). Most of them were negative, ranging negative effects, this does not mean that these studies did not from −1.7 to −43%. Only one effect was positive, showing a consider any mechanisms regarding positive effects if they Japanese earthquake (Kurita 2019) 8 - royalsocietypublishing.org/journal/rspb Japanese earthquake (Körblein 2021) studies reporting multiple effect sizes Japanese earthquake (Hamamatsu 2014) Guatemalan earthquake (Houdaille 1982) - Chernobyl accident (Bertollini 1990) Zika (Triaca 2021) Zika (Gamboa 2019) Icelandic measles outbreak (Gunnarsdottir 2014) 1918 Influenza (Wilson 2019) 1918 Influenza (Mills 1986) - Proc. R. Soc. B 290: 20230211 after ten months from the disaster, 1918 Influenza (Chandra 2018) - fertility declined, increased, and stabilized over the seven months observation period 1918 Influenza (Chandra 2015) 1918 Influenza (Bloom-Feshbach 2011) Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 0 3 6 9 12 15 18 24 months since disaster Figure 6. Studies reporting effect sizes across multiple time points since the onset of a disaster. For each unique study (y-axis), effect directions (decrease [↓], no change [−], increase [↑]) are aligned by time lag of post-disaster fertility follow-up (x-axis). The grey horizontal bars represent the time window of the observation period, which starts from the first post-disaster month at which an effect direction is reported and ends at the last post-disaster month at which an effect direction is reported. We used 41 effect sizes extracted from 13 studies that reported fertility effects across multiple post-disaster time points. physiological 7% –psychological –child/community replacement effect al –partnership dynamics our ehavi b 93% –postponement and family size reduction physiological 28% ility fert % –behaviours limiting successful delivery beh 51% 49 given conception (e.g. abortion) avi –psychological our –partnership dynamics l a –behaviours limiting conception given ral –economic conditions/uncertainty physical ity intercourse (e.g. contraceptive use) behaviou –psychological disasters 51% 72% fertil 36 studies –migration –postponement and family size reduction –economic conditions/uncertainty –partnership dynamics considered disease –migration mechanisms outbreaks 100% y lit technological ferti disasters fer 0% phy 67% tili 10 –physiology limiting delivery sio ty log given conception (e.g. miscarriages) ic –physiology limiting conception al given intercourse (e.g. infertility) –reproductive age mortality cal –physiology limiting delivery given conception ogi hysiol (e.g. miscarriages) behavioural 33% p 49% –preterm births –psychological –disruption of reproductive service –physiology limiting conception given intercourse (e.g. infertility) Figure 7. Mechanisms driving the impact of disaster on fertility based on 36 studies that mentioned at least one possible mechanism underlying the disaster– fertility relationship. A total of 123 mechanisms mentioned by the 36 studies were classified by disaster types, effects direction, and whether a mechanism can be considered as behavioural or physiological. Mechanisms most frequently mentioned are shown on the top of each group, and mechanisms mentioned only once are not shown. For a full list of mechanisms by disaster types and fertility effect directions, see electronic supplementary material, tables S1 and S2. mentioned any. Rather, this means that the studies concluded reductions in completed family size, or abortion, as well an overall negative effect based on their analyses. as psychological factors (e.g. fear of health risks). Some of Among the 20 studies on disease outbreaks that mentioned the studies also considered physiological mechanisms, such at least one mechanism and reported a negative impact on as those preventing successful delivery (e.g. miscarriage), fertility, over half considered behavioural mechanisms. These reproductive age mortality and preterm births. Most of these included contraceptive use, postponement of childbearing, 20 studies on disease outbreaks were on the Zika epidemic Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 Table 1. Summary of 50 reviewed studies arranged by disaster types, specific disasters within types, and alphabetical order of countries studied. Follow-up lag is marked with an asterisk if it could be approximately estimated due to limited information available from a study. Direction of fertility effects is based on summary effects, which are information from the authors’ preferred analysis or mentioned as the main finding of their study. The dash sign (−) means that a study did not provide relevant information. follow-up fertility effects (months since (based on study (last name of research design: disaster summary the first author, disaster country comparison sample size data onset) effects) heterogeneity in effects publication year) disease outbreaks (48% of the 50 studies) measles outbreak Iceland within-population ∼70 000–80 000 yes 7 ↓ Gunnarsdottir [42] people 1918 influenza Canada within-population ∼10 baptisms per year no 24* ↓ Herring [59] 1918 influenza India within-population 50 000 births yes 6* ↓ Mills [43] 1918 influenza Japan within-population — yes 9–10 ↓ Chandra—Japan [38] 1918 influenza New Zealand within-population — yes 12 ↓ by race: larger decline in Maori Wilson [50] 1918 influenza Taiwan within-population — yes 9–10 ↓ Chandra—Taiwan [39] 1918 influenza US within-population — yes 0–3, 9–10 ↓ Chandra [41] 1918 influenza US within-population 21 334 births no 9–10 ↓ Dahal [60] 1918 influenza Denmark, Norway, within-population — unclear 6 ↓ by region (urban/rural): Copenhagen versus Bloom-Feshbach [36] Sweden, US rural Denmark were similar Measles outbreak Japan within-population — yes 9–10 ↓ Mizumoto [61] Zika Brazil within-population 3 820 304 births yes 5* ↓ within-region heterogeneities Castro [62] Zika Brazil within-population 77 977 births yes 9–10 ↓ Diaz-Quijano [63] Zika Brazil within-population 373 136 births yes 9–10 ↓ by regions: larger decline in cities with higher Diaz-Quijano [64] prevalence of microcephaly, but reduction also occurred with no or few cases Zika Brazil within-population 2 857 800 births yes 3* ↓ by age, education, regions: larger decline in Marteleto [65] younger women, highly educated, and in northeast Zika Brazil between-population all registered births yes 9–10 ↓ by age and education: larger decline in older Rangel [66] women and educated Zika Brazil between-population 33 366 births yes 12* ↓ by age: larger decline in younger women Triaca [45] Zika Colombia between-population 15 680 births yes 7–10 ↓ by age or education: no evidence for Gamboa [46] heterogeneity (Continued.) 9 Proc. R. Soc. B 290: 20230211 royalsocietypublishing.org/journal/rspb Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 Table 1. (Continued.) follow-up fertility effects (months since (based on study (last name of research design: disaster summary the first author, disaster country comparison sample size data onset) effects) heterogeneity in effects publication year) Zika Singapore between-population 657 women no 0 ↓ some interactions by age, desired fertility, Tan & Pang [67] education, and regions COVID-19 22 high-income within-population — yes 9–10 mixed across Aassve [51] countries countries COVID-19 Colombia within-population — yes −3* ↓ by age: larger decline among 10–14 years Mendoza [68] COVID-19 Italy within-population 1579 (Turin), 2656 unclear 9–10 ↓ DeRose [69] (Milan) COVID-19 Italy within-population 4611 births yes 9–10 ↓ Somigliana [70] COVID-19 Italy within-population 555 live deliveries no 6 ↓ Trombetta [71] COVID-19 US within-population 12 099 births no 9–10 ↓ by regions: larger decline in New York City McLaren [72] technological disaster (18% of the 50 studies) Kyshtym disaster USSR between-population 1063 people unclear — — by age: lower percentage of people having Buldakov [55] children among population born at the moment of accident Iraq methyl mercury Iraq within-population 66 549 births no 27 ↓ Greenwood [73] poisoning Minamata poisoning Japan between-population 1 904 534 births yes 60 ↓ Yorifuji [54] Chernobyl accident 15 former Soviet Within-population — unclear −4 ↓ Grech [37] republics Chernobyl accident Belarus between-population — unclear 44 ↓ by mode of conception (natural versus artificial Shakhot’ko [74] reproductive technology) Chernobyl accident Italy Within-population 43 905 births yes 9–10 ↓ by regions Bertollini [44] Chernobyl accident Ukraine between-population — no 8 ↓ Donets’ [75] Washington water US between-population — yes 0 ↓ Edwards [52] crisis Flint water crisis US between-population 15 425 births yes 3 ↓ by race: slight reduction in Blacks and increase Wang [76] in Whites (Continued.) 10 Proc. R. Soc. B 290: 20230211 royalsocietypublishing.org/journal/rspb Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 Table 1. (Continued.) follow-up fertility effects (months since (based on study (last name of research design: disaster summary the first author, disaster country comparison sample size data onset) effects) heterogeneity in effects publication year) physical disasters (34% of the 50 studies) Avalanche UK between-population 445 births yes 15 ↑ by bereavement status: bereaved parents Williams [77] contributing to increased fertility Guatemalan Guatemala within-population — yes 0 ↑ by regions: fertility changes were stronger in Houdaille [40] earthquake the capital Flood Poland between-population 362 280 women unclear 6* ↑ Neuberg [78] Tajikistan drought Tajikistan Within-population — no 7 ↓ Clifford [79] Red river flood US between-population 57 007 births yes 0 ↓ Tong [80] Hurricane Hugo US between-population — yes 4 ↑ by regions: no differential impact by severity Cohan [81] of hurricane Hurricane Katrina US between-population all regi stered births yes 6 mixed across by race: decline in Blacks but increase in Seltzer [82] racial groups Whites Indian earthquake India between-population 266 290 women no 12 ↑ by age, education, minority caste group, Nandi [83] regions: fertility increases larger in rural areas, among scheduled tribe women, 20– 30 years, and uneducated women; fertility reduction in highly educated women Indonesian tsunami Indonesia within-population 252 women no 1 ↓ Kinoshita [84] Indonesian tsunami Indonesia between-population 6363 women no 24 ↑ by age: fertility increases larger among 20–24 Nobles [85] years Chilean earthquake Chile within-population 3279 deliveries no 1 ↓ Oyarzo [86] Chilean earthquake Chile between-population all registered births yes 9–10* ↑ Scapini [87] Japanese earthquake Japan between-population 32 078 births yes 9–10 ↓ Hamamatsu [47] Japanese earthquake Japan within-population ∼28 000 births yes 7* ↑ by regions: fertility decline is larger in regions Körblein [48] exposed more to radiation Japanese earthquake Japan within-population — yes 0 ↑ Kurita [49] Iranian earthquake Iran within-population 44 265 women no 5* ↓ Bahmanjanbeh [88] Philippine typhoon Philippines within-population — yes — — Mangada [53] 11 Proc. R. Soc. B 290: 20230211 royalsocietypublishing.org/journal/rspb and the 1918 influenza pandemic, but notably, the former was clearest for disasters with prolonged periods of uncertainty, a 12 mostly concerned with behavioural mechanisms (87% of the finding that is in line with our above interpretation. Fertility royalsocietypublishing.org/journal/rspb time) and the latter with physiological mechanisms (80% of declined more in areas already vulnerable to the risk of the time). This diverging pattern reflects how different charac- droughts, suggesting that continuing uncertainty from disas- teristics of risks associated with each disaster can result in the ters could be a contributing factor to the observed fertility consideration of different mechanisms by researchers. decline in response to disasters. On the other hand, the average This pattern also manifests itself in technological and change in fertility was either unclear or non-substantial after physical disasters. For these disaster types, the mechanisms short-duration disasters, such as floods, earthquakes, tropical considered in each study were dominated by either behav- cyclones, other storms and epidemics. Of note, the definition ioural or physiological mechanisms, but not both (figure 7). of epidemics in that study only included rapid outbreaks, and Researchers interpreted fertility change from technological excluded endemic diseases such as HIV/AIDS. In other disasters mainly through the lens of physiological mechan- words, the study operationalized epidemics conceptually isms relating to conception and successful delivery, perhaps similar to physical disasters in terms of the time scale of risk because all the disasters studied involved the spread of development. It can thus be summarized that disasters invol- Proc. R. Soc. B 290: 20230211 hazardous chemical substances. By contrast, physical disas- ving higher uncertainty about risks may contribute to fertility ters were postulated to drive fertility change mainly reduction, whereas the evidence of fertility decline is weaker through behavioural mechanisms. Here, mechanisms invol- for relatively short-duration disasters. As such, the study ving positive fertility effects have been most explicitly suggests that the general findings from our review may hold Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 considered, including psychological factors (e.g. attachment), even in relatively high fertility contexts such as in Africa. child/community replacement effects, and partnership While we find clear uniformity in the direction of summary dynamics (see electronic supplementary material, table S1 effects for some disaster types, overall the results were hetero- for specific mechanism(s) each study mentioned). geneous. Among physical disasters, the direction of effects was variable, but also even within the same disaster, and even within the same population groups. As an example of between-population heterogeneity, one study on the 2004 4. Discussion Indonesian tsunami conducted in the most severely affected This systematic literature review demonstrates that disasters regions found a strong negative effect [84], while another reduce fertility, albeit heterogeneously by disaster type. cross-regional study found a fertility increase [85]. Some Specifically, while physical disasters resulted in either posi- studies also examine within-population heterogeneities in the tive or negative fertility responses, for disease outbreaks fertility effects of disasters, as summarized in table 1. This het- and technological disasters the summary effects were uni- erogeneity urges more future research to consider contextual formly negative. When positive effects were found for factors driving the divergence of fertility effects, especially physical disasters, they were more likely to occur in higher given that disaster vulnerability is highly unequal both fertility contexts (TFR > 2.0). However, we find this tendency across and within countries and socioeconomic groups [91]. was not strong enough to support the oft-postulated notion We also note that the current literature is biased toward that positive fertility effects are in general more likely in research on certain contexts and disasters, in spite of a sub- higher TFR countries [6,7]. In addition, we found little stantial expansion of research on the disaster–fertility nexus evidence that the direction of fertility effects differs by time during this period. Particularly noticeable were gaps for since disaster: negative effects were reported anywhere low-income countries, and for certain high-casualty disasters from the onset to five years after the disaster, suggesting which occurred during the last two decades. This is consist- that positive effects are likely to be short-term, if any, even ent with a preoccupation with disasters causing high for physical disasters [32]. economic losses rather than high casualties [92]. Similarly, The concentration of positive effects from physical disas- our analysis of the magnitude of fertility effects suggests a ters may reflect differences in how risks develop and are potential publication bias for studies finding large impacts, perceived during these events. Physical disasters are charac- since sub-national effects were on average larger than terized by a relatively acute, recognizable shock. As such, national ones. We could not formally test publication bias one can postulate that the waning of risks would also be per- due to the heterogeneity in disaster types, timeframes, ceived relatively easily, leading to a higher probability of methods, and types of effects reported. However, it is unli- fertility increase during the recovery period. By contrast, kely that negative effects were published at a higher rate disease outbreaks and technological disasters bring risks than positive effects, since both positive and negative effects that are often unclear, prolonged or difficult to track (e.g. were hypothesized a priori. the spread of viruses, toxic materials such as radiation, In the remainder of this paper, we identify common limit- etc.). Such uncertainty, together with the direct and often ations from the reviewed literature and summarize them into long-term physiological effects of virus infection or toxic six practical recommendations for future research. While not materials on fecundity and health in general, may underlie prescriptive, these recommendations can help researchers the predominant pathways by which disease outbreaks and produce and evaluate evidence on causal effects—essential technological disasters are thought to reduce fertility. for a field involving humans where experimental evidence Our findings are corroborated by studies that were not is problematic. Whenever possible, we highlight examples included in the present review because they did not meet our of methods from reviewed studies that can be usefully eligibility criteria [32–34]. A recent study [90] published six applied for future research. months after our literature search also concluded fertility First, data infrastructures for collecting data in low- reduction during 5 years after various disasters in African resource settings should be supported, particularly during countries. In this study, the evidence for fertility decline was disasters (Recommendation 1). Conducting regular yet well- spaced population censuses is essential for monitoring the [65]). A few studies used synthetic control methods to 13 effect of disaster on fertility and mortality; however, doing create a comparable control [46,76,94]. royalsocietypublishing.org/journal/rspb so remains difficult for many low-income countries [93]. Disasters do not happen in a vacuum, and fertility within Researchers can take an active role in improving data collec- countries is rarely stable. For example, fertility can rapidly fall tion by openly appraising the quality of data and data in countries undergoing the demographic transition, and infrastructure used in their research. Doing so would also oscillate in countries that have already achieved low fertility. help assess the risk of bias in their findings. For example, Many of the studies reviewed here did consider and account errors surrounding the under-counting of births due to for these context-specific fertility trends, and such efforts delayed reporting, house births in case of historical data, would continue to benefit the literature in future (Recommen- and outmigration due to disaster were frequently reported dation 5). One common technique is to analytically adjust for in the reviewed studies. Another common source of bias seasonal fertility patterns and long-term fertility time trends. was the omission of pre-disaster fertility data from women In our reviewed studies, the majority (80%) addressed sea- who died from the disaster, particularly in cases where pre- sonality, mostly by comparing fertility at least one year disaster fertility was determined by retrospective self-reports apart matched in months or by including month dummies Proc. R. Soc. B 290: 20230211 from survivors [84]. This case underscores a need for contin- in regression analyses. However, only half (52%) reported ued and regular monitoring of demographic data. Other effect sizes that took time trends into account. This is proble- possible efforts to overcome the bias present in the current lit- matic, given that a still large portion of the studied disasters erature include mapping data availability (e.g. demographic occurred in contexts of declining fertility. Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 and health surveys, and multiple-indicator surveys) and dis- Lastly, one of the most basic requirements for causal infer- aster occurrence jointly. This initial mapping would identify ence in this area is to establish that the disaster actually existing gaps that can be addressed readily by future studies. preceded the observed fertility change [95]. However, defin- Improved data availability, in turn, will lay the ground- ing the onset of a disaster can be difficult. For example, for work for more evidence building on how disaster affects disease outbreaks, the timing of the first case may be very fertility in a global perspective. Findings that do not meet different to the timing of the maximum effect on the popu- the conventional threshold of ‘statistical significance’ should lation. In eight out of the 50 reviewed studies, there was no be published (Recommendation 2), as long as the data, ana- clear indication of the timing of disaster onset, the timing at lytic approach, and estimated effects are clear (accompanied which fertility measurement began, or both. As an alterna- by the necessary information for meta-analysis and standard- tive, some studies identified the disaster onset post hoc, by ization, e.g. person-year exposure for birth counts [65]). subtracting 9–10 months from the period associated with Doing so will not only facilitate open science practices in gen- the largest or statistically significant fertility effect. This eral, but also give a much-needed opportunity to compare approach might be useful as an exploratory technique, but the fertility effects of disasters across different contexts. is subjective and vulnerable to questionable research practices Future research should present more detailed information such as post hoc hypothesizing. on the population characteristics of comparison groups, Instead, some authors made explicit their beliefs about or at least indicate if such information is not available the timing and magnitude of fertility impacts, a practice (Recommendation 3), and wherever possible, analytically that could aid in evaluating the evidence at hand and build- adjust for differences between compared groups that may ing theory (Recommendation 6). Some studies stated how confound the disaster–fertility relationship (Recommendation large an effect is expected to be, in the form of either qualitat- 4). Researchers, either implicitly or explicitly, assume that ive statements (‘little to no effect’ [65]) or even setting a groups that are being compared are generally similar aside threshold of what the researchers considered to be a large from the exposure to disaster, otherwise known as the enough effect size given the fertility trends of the study popu- ‘homogeneity assumption’. Satisfying this assumption helps lation [71]. Regarding timing, one study on the 2011 Japanese rule out alternative explanations that attribute observed earthquake [47] defined a ‘disaster impact period’ based on fertility changes to factors unrelated to disaster. When authors’ reasoning about when the disaster’s impact is comparing fertility within populations before and after a likely to manifest, and another study on the 1997 Red River disaster, researchers should recognize that pre-disaster ferti- flood [80] specified ‘pre-disaster’ and ‘post-disaster’ periods lity differences across subpopulations may contribute to for precise months and years based on known information post-disaster gaps. Likewise, researchers should acknowledge about the progression of the disaster. In doing so researchers that the effects of disasters on fertility may be direct and clarify not only the expected timing of shock, but also after indirect. The frequency of sexual intercourse in a population how long the impact of a disaster is expected to last. The may be lowered in the aftermath of a disaster due to physical latter is particularly important since different mechanisms damages or poor health (direct effect). At the same time, birth will be operative across different timeframes. For example, if rates may decline due to changes in population structure a specific disaster would significantly increase the mortality (indirect effect), for example, if individuals of reproductive of pregnant women, one might hypothesize an immediate age were more likely to die due to the disaster than the over- reduction in births. Very long lags between exposure and fer- all population. Less than 10 out of the 50 studies tested tility impacts could indicate a longer-run change in fertility specific threats to population homogeneity before and after intentions via postponement effects. Unfortunately, very long the disaster, or even presented background characteristics time lags also increase the likelihood that other post-disaster of each compared group. These threats included both events will confound the estimation, making it difficult to attri- spatial (spillover of fertility effect to control cohorts [45]) bute observed fertility differences to the disaster of interest. In and temporal (conflation with concomitant events such as nearly half of the reviewed studies, outcome measurement economic crises, which was indirectly tested in one study began after a lag longer than 9–10 months, the average by examining socioeconomic gradients in fertility effect length of gestation. However, among such studies, only one provided a rationale for why a longer lag was chosen [85], assumption; (5) adjust for seasonality and time trends in fer- 14 demonstrating the need for more explicit hypotheses tility; (6) be explicit about when and how the expected royalsocietypublishing.org/journal/rspb surrounding the timing of fertility outcome measurement. fertility effects of disasters are likely to occur. Given the pre- dicted increase in disasters—particularly in the context of global climate change—even subtle changes in fertility in 5. Conclusion response to disasters may have far-reaching consequences in population structure. Thus, the importance of studying Across the 50 studies examining the changes in live birth how disasters affect fertility is higher than ever, and should counts after disasters, we found that disasters generally remain a major priority in the research community across dis- have a negative fertility impact, depending on disaster type. ciplines, from demography to population studies, economics, If any, studies showing an increase in fertility were more anthropology, sociology, public health and epidemiology. likely to come from studies on physical disasters, which included avalanche, drought, earthquake and tsunami, Data accessibility. The original data extracted from the 50 studies, and flood, hurricane and typhoon in the reviewed studies. We detailed information about the screening and selection procedure also identified 13 distinct mechanism groups which research- Proc. R. Soc. B 290: 20230211 are available at the project website (https://osf.io/dpgqf/) [96]. ers have considered as underlying the fertility effects of Supplementary material is available online [97]. disaster. By contrast to the common belief that disasters are Authors’ contributions. D.S.L.: conceptualization, data curation, formal more likely to increase fertility in contexts with already analysis, project administration, resources, visualization, writing— original draft, writing—review and editing; E.B.: conceptualization, Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 high fertility, we found little evidence to suggest that the data curation, formal analysis, resources, writing—original draft, underlying fertility level was an important predictor of the writing—review and editing; A.C.: conceptualization, data curation, direction, timing or size of fertility impacts. While this may formal analysis, resources, visualization, writing—original draft, be because no relationship exists, it may also be due to writing—review and editing; J.W.: conceptualization, data curation, biases in the literature towards studying high-income resources, supervision, writing—review and editing. All authors gave final approval for publication and agreed to be countries or high-cost disasters. This bias may also arise held accountable for the work performed therein. from a lack of standardization by methods and timeframes Conflict of interest declaration. The authors declare no competing interests. in this literature, leading to a small number of studies Funding. E.B. acknowledges funding from the Marie Sklodowska-Curie across which to compare. We thus need more studies from Action Individual Fellowship from the European Commission (grant high fertility contexts, and also more efforts to measure ferti- no. 101060237-DISPARFAM). lity effects comparably and across multiple time points. Acknowledgements. We thank colleagues who helped in the process of Doing so will also help disentangle key mechanisms of ferti- protocol development (Saskia Morwinsky) and study selection and lity effects which are numerous and complex. To encourage data extraction by providing support in non-English language articles (Aliakbar Akbaritabar; Magnus Gottfreðsson; Marina Kolobova; moves toward a more robust literature on this critical topic, Aiva Jasilioniene; Henrik Alexander Schubert; Silvia Loi; Jiaxin Shi; and eventually toward evidence-informed theory and Hanna Semenchenko and Nataliia Levchuk). We are grateful for policy, we offer six recommendations for future research: (1) the assistance from the library and the IT team at the Max Planck increase efforts to build and maintain data infrastructure for Institute for Demographic Research (MPIDR). We thank the audience the collection of quality data in resource-limited settings; (2) who attended and provided helpful feedback at the MPIDR labora- nurture a publication culture that incentivizes reporting of tory talks series, European Human Behaviour and Evolution Association 2022 conference, CATA ‘Community in the Anthropo- results (including null results) to be compared across con- cene’ seminar series, and the Population Association of America texts; (3) collect and report characteristics of compared 2023. We are grateful for the constructive feedback provided by the groups; (4) analytically address threats to the homogeneity reviewers and the editors. References 1. Bettencourt UMD et al. 2013 Building resilience: 6. Skirbekk V. 2022 Decline and prosper!, 1st edn. 11. Higgins JP, Thomas J, Chandler J, Cumpston M, Li T, integrating climate and disaster risk into Berlin, Germany: Springer. See https://link.springer. Page MJ, Welch VA. 2019 Cochrane handbook for development. Washington, DC: The World Bank. com/book/10.1007/978-3-030-91611-4. systematic reviews of interventions. Chichester, UK: 2. United Nations Office for Disaster Risk Reduction 7. Aassve A, Cavalli N, Mencarini L, Plach S, Bacci ML. John Wiley & Sons. 2020 Human cost of disasters 2000–2019. Geneva, 2020 The COVID-19 pandemic and human fertility. 12. Bolouki A, Zal F. 2020 Impact of war on fertility and Switzerland: United Nations Office for Disaster Risk Science 369, 370–371. (doi:10.1126/science. infertility. Arch. Iran. Med. 23, S16–S22. (doi:10. Reduction. abc9520) 34172/aim.2020.s4) 3. WMO. 2021 Weather-related disasters increase over 8. Bertillon J. 1892 La grippe à Paris et dans quelques 13. Hill K. 2004 War, humanitarian crises, population past 50 years, causing more damage but fewer deaths. autres villes de France et de l’étranger en 1889– displacement, and fertility: A review of evidence. See https://public.wmo.int/en/media/press-release/ 1890: extrait de l’Annuaire statistique de la ville de Washington, DC: National Academies Press. weather-related-disasters-increase-over-past-50-years- Paris pour l’année 1890. Paris, France: Imprimerie 14. Neal S, Stone N, Ingham R. 2016 The impact of causing-more-damage-fewer. Municipale. armed conflict on adolescent transitions: a 4. Grace K. 2017 Considering climate in studies of 9. Cornia GA, Paniccià R. 2000 The mortality crisis in systematic review of quantitative research on age of fertility and reproductive health in poor countries. transitional economies. Oxford, UK: Oxford University sexual debut, first marriage and first birth in young Nat. Clim. Change 7, 479–485. (doi:10.1038/ Press. women under the age of 20 years. BMC Public nclimate3318) 10. Sellers S. 2022 Environment and fertility. Health 16, 1–11. (doi:10.1186/s12889-016-2868-5) 5. Muttarak R. 2021 Demographic perspectives in In International handbook of population 15. Thiede BC, Hancock M, Kodouda A, Piazza J. 2020 research on global environmental change. Popul. and environment (eds LM Hunter, C Gray, Exposure to Armed Conflict and Fertility in Sub- Stud. 75, 77–104. (doi:10.1080/00324728.2021. J Véron), pp. 441–461. Berlin, Germany: Saharan Africa. Demography 57, 2113–2141. 1988684) Springer. (doi:10.1007/s13524-020-00923-2) 16. Sobotka T, Skirbekk V, Philipov D. 2011 Economic 30. Lewis JJC, Ronsmans C, Ezeh A, Gregson S. 2004 The epidemic in Brazil. Pap. Reg. Sci. 100, 795–813. 15 recession and fertility in the developed world. Pop. population impact of HIV on fertility in sub-Saharan (doi:10.1111/pirs.12591) royalsocietypublishing.org/journal/rspb Dev. Rev. 37, 267–306. (doi:10.1111/j.1728-4457. Africa. AIDS 18(Suppl 2), S35–S43. (doi:10.1097/ 46. Gamboa LF, Rodriguez Lesmes P. 2019 The fertility- 2011.00411.x) 00002030-200406002-00005) inhibiting effect of mosquitoes: socio-economic 17. Grace K, Billingsley S, Van Riper D. 2020 Building an 31. Zaba B, Gregson S. 1998 Measuring the impact of differences in response to the Zika crisis in interdisciplinary framework to advance conceptual HIV on fertility in Africa. AIDS (London, England) 12, Colombia. Econ. Hum. Biol. 35, 63–72. (doi:10. and technical aspects of population-environment S41–S50. (doi:10.1097/00002030-199807000- 1016/j.ehb.2019.05.001) research focused on women’s and children’s health. 00001) 47. Hamamatsu Y, Inoue Y, Watanabe C, Umezaki M. Soc. Sci. Med. 250, 112857. (doi:10.1016/j. 32. Davis J. 2017 Fertility after natural disaster: 2014 Impact of the 2011 earthquake on marriages, socscimed.2020.112857) Hurricane Mitch in Nicaragua. Popul. Environ. 38, births and the secondary sex ratio in Japan. J. Biosoc. 18. Perry RW. 2018 Defining disaster: An evolving 448–464. (doi:10.1007/s11111-017-0271-5) Sci. 46, 830. (doi:10.1017/S0021932014000017) concept. In Handbook of disaster research (eds H 33. Behrman JA, Weitzman A. 2016 Effects of the 2010 48. Körblein A. 2021 Reduction in live births in Japan Rodríguez, W Donner, JE Trainor), pp. 3–22. Berlin, Haiti Earthquake on Women’s Reproductive Health. nine months after the Fukushima nuclear accident: Germany: Springer. Stud. Fam. Plann. 47, 3–17. (doi:10.1111/j.1728- an observational study. PLoS ONE 16, e0242938. Proc. R. Soc. B 290: 20230211 19. Quarantelli EL, Perry RW. 2005 A social science 4465.2016.00045.x) (doi:10.1371/journal.pone.0242938) research agenda for the disasters of the 21st 34. Boberg-Fazlic N, Ivets M, Karlsson M, Nilsson T. 49. Kurita N. 2019 Association of the Great East Japan century: theoretical, methodological and empirical 2021 Disease and fertility: evidence from the 1918– Earthquake and the Daiichi nuclear disaster in issues and their professional implementation. What 19 influenza pandemic in Sweden. Econ. Hum. Biol. Fukushima City, Japan, with birth rates. JAMA Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 is a Disaster 325, 396. 43, 101020. (doi:10.1016/j.ehb.2021.101020) Network Open 2, e187455. (doi:10.1001/ 20. Quarantelli EL. 1993 Community crises: an 35. Veritas Health Innovation. 2017 Covidence systematic jamanetworkopen.2018.7455) exploratory comparison of the characteristics and review software. Melbourne, Australia: Covidence. 50. Wilson N, Turner N, Baker MG. 2019 One hundred consequences of disasters and riots. J. Contingencies See www.covidence.org. years ago in 1919: new Zealand’s birth reduction and Crisis Manage. 1, 67–78. (doi:10.1111/j.1468- 36. Bloom-Feshbach K, Simonsen L, Viboud C, Mølbak shock associated with an influenza pandemic. N Z 5973.1993.tb00009.x) K, Miller MA, Gottfredsson M, Andreasen V. 2011 Med. J. 132, 57–62. 21. Loewen S, Pinchoff J, Ngo TD, Hindin MJ. 2022 The Natality decline and miscarriages associated with 51. Aassve A, Cavalli N, Mencarini L, Plach S, Sanders S. impact of natural disasters and epidemics on sexual the 1918 influenza pandemic: the Scandinavian and 2021 Early assessment of the relationship between and reproductive health in low- and middle-income United States experiences. J. Infect. Dis. 204, the COVID-19 pandemic and births in high-income countries: a narrative synthesis. Int. J. Gynecol. 1157–1164. (doi:10.1093/infdis/jir510) countries. Proc. Natl Acad. Sci. USA 118, Obstet. 157, 11–18. (doi:10.1002/ijgo.13768) 37. Grech V. 2014 The Chernobyl accident, the male to e2105709118. (doi:10.1073/pnas.2105709118) 22. Jeffers NK, Glass N. 2020 Integrative review of female ratio at birth and birth rates. Acta Medica 52. Edwards M. 2014 Fetal death and reduced birth pregnancy and birth outcomes after exposure to a (Hradec Kralove) 57, 62–67. (doi:10.14712/ rates associated with exposure to lead- hurricane. J. Obstet. Gynecol. Neonatal Nurs. 49, 18059694.2014.41) contaminated drinking water. Environ. Sci. Technol. 348–360. (doi:10.1016/j.jogn.2020.04.006) 38. Chandra S, Yu Y-L. 2015 The 1918 influenza pandemic 48, 739–746. (doi:10.1021/es4034952) 23. Harville EW, Xiong X, Buekens P. 2010 Disasters and and subsequent birth deficit in Japan. Demogr. Res. 53. Mangada LL, Permejo Cuaton G. 2020 Fertility, sex, perinatal health: a systematic review. Obstet. 33, 313–326. (doi:10.4054/DemRes.2015.33.11) and reproductive health dynamics after Typhoon Gynecol. Surv. 65, 713. (doi:10.1097/OGX. 39. Chandra S, Yu Y-L. 2015 Fertility decline and the Yolanda in Tacloban North, Philippines. Soc. Sci. 0b013e31820eddbe) 1918 Influenza pandemic in Taiwan. Biodemogr. Diliman 16, 50–72. 24. Stieb DM, Chen L, Eshoul M, Judek S. 2012 Ambient Soc. Biol. 61, 266–272. (doi:10.1080/19485565. 54. Yorifuji T, Kashima S, Suryadhi MAH, Abudureyimu air pollution, birth weight and preterm birth: a 2015.1062718) K. 2017 Temporal trends of infant and birth systematic review and meta-analysis. Environ. Res. 40. Houdaille J. 1982 The Guatemala earthquake of outcomes in Minamata after severe methylmercury 117, 100–111. (doi:10.1016/j.envres.2012.05.007) February, 1976 - its effects on birth and marriage exposure. Environ. Pollut 231, 1586–1592. (doi:10. 25. Shah PS, Balkhair T, births KSG on D of P. 2011 Air rates. Population 37, 429–434. (doi:10.2307/ 1016/j.envpol.2017.09.060) pollution and birth outcomes: a systematic review. 1532209) 55. Buldakov LA et al. 1990 The medical sequelae of Environ. Int. 37, 498–516. (doi:10.1016/j.envint. 41. Chandra S, Christensen J, Mamelund S-E, Paneth N. the radiation accident in the Southern Urals in 1957. 2010.10.009) 2018 Short-term birth sequelae of the 1918–1920 Medicinskaa Radiologia 35, 11–15. 26. Evans J, Bansal A, Schoenaker DA, Cherbuin N, Peek influenza pandemic in the United States: state-level 56. Bushman BJ, Wang MC. 2009 Vote-counting MJ, Davis DL. 2022 Birth outcomes, health, and analysis. Am. J. Epidemiol. 187, 2585–2595. (doi:10. procedures in meta-analysis. In The handbook of health care needs of childbearing women following 1093/aje/kwy153) research synthesis and meta-analysis, pp. 207–220, wildfire disasters: an integrative, state-of-the- 42. Gunnarsdottir S, Briem H, Gottfredsson M. 2nd edn. New York, NY: Russell Sage Foundation. science review. Environ. Health Perspect. 130, 2014 Extent and impact of the measles 57. Davis K, Blake J. 1956 Social structure and fertility: 086001. (doi:10.1289/EHP10544) epidemics of 1846 and 1882 in Iceland. an analytic framework. Econ. Dev. Cult. Change 4, 27. Zotti ME, Williams AM, Robertson M, Horney J, Hsia Laeknabladid 100, 211–216. (doi:10.17992/lbl. 211–235. (doi:10.1086/449714) J. 2013 Post-disaster reproductive health outcomes. 2014.04.539) 58. Bongaarts J. 1978 Framework for analysing the Matern. Child Health J. 17, 783–796. (doi:10.1007/ 43. Mills ID. 1986 The 1918–1919 influenza proximate determinants of fertility. Popul. Dev. Rev. s10995-012-1068-x) pandemic—the Indian experience. Indian Econ. Soc. 4, 105–132. (doi:10.2307/1972149) 28. Guha-Sapir D, Below R, Hoyois P. 2016 EM-DAT: the Hist. Rev. 23, 1–40. (doi:10.1177/ 59. Herring DA. 1993 ‘There Were Young People and CRED/OFDA international disaster database. Brussels, 001946468602300102) Old People and Babies Dying Every Week’: The Belgium: Université Catholique de Louvain. See 44. Bertollini R, Di Lallo D, Mastroiacovo P, Perucci CA. 1918–1919 Influenza Pandemic at Norway House. www.emdat.be. 1990 Reduction of births in Italy after the Chernobyl Ethnohistory 41, 73–105. (doi:10.2307/3536979) 29. Mamelund S-E. 2004 Can the Spanish influenza accident. Scand. J. Work Environ. Health 16, 60. Dahal S, Mizumoto K, Bolin B, Viboud C, Chowell G. pandemic of 1918 explain the baby boom of 1920 96–101. (doi:10.5271/sjweh.1803) 2018 Natality decline and spatial variation in excess in neutral Norway? Population 59, 229–260. 45. Triaca L, Ribeiro F, Tejada C. 2021 Mosquitoes, birth death rates during the 1918–1920 Influenza (doi:10.2307/3654904) rates and regional spillovers: evidence from the Zika Pandemic in Arizona, United States. Am. J. Epidemiol. 187, 2577–2584. (doi:10.1093/ 72. McLaren RA et al. 2021 COVID-related ‘lockdowns’ 85. Nobles J, Frankenberg E, Thomas D. 2015 The 16 aje/kwy146) and birth rates in New York. Am. J. Obstet. Gynecol. effects of mortality on fertility: population dynamics royalsocietypublishing.org/journal/rspb 61. Mizumoto K, Chowell G. 2020 Temporary fertility MFM 3, 100476. (doi:10.1016/j.ajogmf.2021.100476) after a natural disaster. Demography 52, 15–38. decline after large rubella outbreak, Japan. Emerg. 73. Greenwood MR. 1985 Methylmercury poisoning in (doi:10.1007/s13524-014-0362-1) Infect. Dis. 26, 1122–1129. (doi:10.3201/EID2606. Iraq. An epidemiological study of the 1971–1972 86. Oyarzo C, Bertoglia P, Avendaño R, Bacigalupo F, 181718) outbreak. J. Appl. Toxicol. 5, 148–159. (doi:10.1002/ Escudero A, Acurio J, Escudero C. 2012 Adverse 62. Castro MC, Han QC, Carvalho LR, Victora CG, França jat.2550050305) perinatal outcomes after the February 27th 2010 GVA. 2018 Implications of Zika virus and congenital 74. Shakhot’ko L. 1996 Demographic impact of the Chilean earthquake. J. Matern. Fetal. Neonatal. Med. Zika syndrome for the number of live births in Chernobyl nuclear power station disaster. Voprosy 25, 1868–1873. (doi:10.3109/14767058.2012. Brazil. Proc. Natl Acad. Sci. USA 115, 6177–6182. statistiki (Moscow, Russia : 1994), 66–72. 678437) (doi:10.1073/pnas.1718476115) 75. Donets’ MP. 2004 Demographic situation in 87. Scapini V, Vergara C. 2021 Natural disasters and 63. Diaz-Quijano FA, Chiavegatto Filho ADP. 2017 Chernigiv area and its connection with Chernobyl birth rate: evidence from the 2010 Chilean Reduction of the birth rate in São Paulo: a probable catastrophe. Likars’ka sprava/Ministerstvo okhorony Earthquake. J. Popul. Soc. Stud. 29, 274–285. effect of the panic caused by the Zika-associated zdorov’ia Ukraïny, 17–20. (doi:10.25133/JPSSV292021.017) Proc. R. Soc. B 290: 20230211 microcephaly epidemic. Ann. Epidemiol. 27, 76. Wang R, Chen X, Li X. 2021 Something in the pipe: 88. Bahmanjanbeh F, Kohan S, Yarmohammadian MH, 616–617. (doi:10.1016/j.annepidem.2017.08.009) the Flint water crisis and health at birth. J. Popul. Haghshenas A. 2016 Evaluation of reproductive 64. Diaz-Quijano FA, Pelissari DM, Chiavegatto Filho Econ. 35, 1723–1749. (doi:10.1007/s00148-021- health indicators in women affected by East ADP. 2018 Zika-associated microcephaly epidemic 00876-9) Azarbaijan earthquake on August 2012. Downloaded from https://royalsocietypublishing.org/ on 09 June 2023 and birth rate reduction in Brazilian Cities. 77. Williams RM, Parkes CM. 1975 Psychosocial effects Iran. J. Nurs. Midwifery Res. 21, 504–509. (doi:10. Am. J. Public Health 108, 514–516. (doi:10.2105/ of disaster: birth rate in Aberfan. Br. Med. J. 2, 4103/1735-9066.193414) AJPH.2017.304260) 303–304. (doi:10.1136/bmj.2.5966.303) 89. Ruggles S. 2013 Big microdata for population 65. Marteleto LJ, Guedes G, Coutinho RZ, Weitzman A. 78. Neuberg M, Jakubowska-Szwed B, Neuberg J. 2001 research. Demography 51, 287–297. (doi:10.1007/ 2020 Live births and fertility Amid the Zika Reproductive behavior after the flood disaster in s13524-013-0240-2) Epidemic in Brazil. Demography 57, 843–872. Klodzko region–July 1997. Ginekol. Pol. 72, 90. Norling J. 2022 Fertility following natural disasters (doi:10.1007/s13524-020-00871-x) 1037–1041. and epidemics in Africa. World Bank Econ. Rev. 36, 66. Rangel MA, Nobles J, Hamoudi A. 2020 Brazil’s 79. Clifford D, Falkingham J, Hinde A. 2010 Through 955–971. (doi:10.1093/wber/lhac011) missing infants: Zika risk changes reproductive civil war, food crisis and drought: trends in fertility 91. Rodríguez H, Donner W, Trainor JE. 2018 Handbook behavior. Demography 57, 1647–1680. (doi:10. and nuptiality in post-Soviet Tajikistan. of disaster research. 2nd edn. Cham, Switzerland: 1007/s13524-020-00900-9) Eur. J. Popul. 26, 325–350. (doi:10.1007/s10680- Springer International Publishing. 67. Tan PL, Pang T. 2022 Residence in infected 010-9206-x) 92. Kharb A, Bhandari S, Moitinho de Almeida M, Castro neighborhoods and fertility decline during the Zika 80. Tong VT, Zotti ME, Hsia J. 2011 Impact of the Delgado R, Arcos González P, Tubeuf S. 2022 epidemic in Singapore. Popul. Environ. 43, Red River catastrophic flood on women giving Valuing human impact of natural disasters: a review 393–422. (doi:10.1007/s11111-021-00389-3) birth in North Dakota, 1994–2000. Matern. Child of methods. Int. J. Environ. Res. Public Health 19, 68. Mendoza V, Velilla P, Hussein S, Maya W. 2021 Health J. 15, 281–288. (doi:10.1007/s10995-010- 11486. (doi:10.3390/ijerph191811486) Impact of the Covid-19 Pandemic on Birth Rates in 0576-9) 93. Chandy L, Zhang C. 2015 Stuffing data gaps with 2020: the case of Colombia. Revista Brasileira De 81. Cohan CL, Cole SW. 2002 Life course transitions and dollars: What will it cost to close the data deficit in Ginecologia E Obstetricia 43, 492–494. (doi:10. natural disaster: marriage, birth, and divorce poor countries? See https://www.brookings.edu/ 1055/s-0041-1731380) following Hurricane Hugo. J. Fam. Psychol. 16, opinions/stuffing-data-gaps-with-dollars-what-will- 69. De Rose AF, Ambrosini F, Mantica G, Terrone C. 2021 14–25. (doi:10.1037/0893-3200.16.1.14) it-cost-to-close-the-data-deficit-in-poor-countries/. Impact of COVID-19 on birth rate trends in the 82. Seltzer N, Nobles J. 2017 Post-disaster fertility: 94. Nobles J, Hamoudi A. 2019 Detecting the effects of Italian metropolitan cities of Milan, Genoa and Hurricane Katrina and the changing racial early-life exposures: why fecundity matters. Popul. Turin. Public Health 198, 35–36. (doi:10.1016/j. composition of New Orleans. Popul. Environ. 38, Res. Policy Rev. 38, 783–809. (doi:10.1007/s11113- puhe.2021.06.026) 465–490. (doi:10.1007/s11111-017-0273-3) 019-09562-x) 70. Somigliana E, Esposito G, Viganò P, Franchi M, 83. Nandi A, Mazumdar S, Behrman JR. 2018 The effect 95. Bhrolcháin MN, Dyson T. 2007 On causation in Corrao G, Parazzini F. 2021 Effects of the early phase of natural disaster on fertility, birth spacing, and demography: issues and illustrations. Popul. Dev. of the COVID-19 pandemic on natural and ART- child sex ratio: evidence from a major earthquake in Rev. 33, 1–36. (doi:10.1111/j.1728-4457.2007. mediated birth rates in Lombardy Region, Northern India. J. Popul. Econ. 31, 267–293. (doi:10.1007/ 00157.x) Italy. Reprod. Biomed. Online 43, 765–767. (doi:10. s00148-017-0659-7) 96. Lee DS, Batyra E, Castro A, Wilde J. 2023 Protocol, 1016/j.rbmo.2021.07.017) 84. Kinoshita M, Suhardan S, Danila Danila D, Chiang C, inclusion criteria, and search results. OSF. (https:// 71. Trombetta A, Travan L, Elefante P, Canton M, Rispoli Aoyama A. 2016 Estimating post-emergency fertility osf.io/dpgqf/) F, Maso G, Barbi E, Risso FM. 2021 The first Italian among disaster-affected adolescents: findings from 97. Lee DS, Batyra E, Castro A, Wilde J. 2023 Human COVID-19 lockdown reduced births and voluntary a case-control study in Aceh Province, Indonesia. fertility after a disaster: a systematic literature terminations by just under a fifth. Acta Paediatr. Disaster Med. Public Health Prep. 10, 80–86. review. Figshare. (doi:10.6084/m9.figshare.c. 110, 2387–2388. (doi:10.1111/apa.15862) (doi:10.1017/dmp.2015.104) 6607452)