Research

INTRODUCTION

Forest peoples rely on wild resources, such as forest products, not only for their livelihoods, but also to cope with adverse shocks, i.e., as natural insurance (Wunder et al. 2014, 2018, Angelsen and Dokken 2018, Razafindratsima et al. 2021). With limited coping options against covariate shocks such as natural disasters and epidemics, natural insurance can serve as a critical safety net for the rural poor. Whether this is the case under the COVID-19 pandemic remains unknown (Fiorella et al. 2020, McNamara et al. 2020), and understanding the impacts of pandemics on local ecosystems, wildlife, and conservation efforts is desperately needed (Lindsey et al. 2020, Stokes et al. 2020, Cooke et al. 2021, Lawler et al. 2021, Diffenbaugh 2022, Gibbons et al. 2022) as evidenced by HIV/AIDS and the Ebola epidemics in Africa (Talman et al. 2013, Olivero et al. 2017). More generally, understanding how poor populations cope with COVID-19 is critical in order to design policies that respond effectively to pandemics (Walker et al. 2020, Egger et al. 2021, Josephson et al. 2021, Miguel and Mobarak 2022, Mobarak et al. 2022). In countries where people have limited access to social assistance, health services, communication, and transportation, available data are scant; this is especially the case in remote rural regions of the developing world where conservation is a major concern.

Known for their rich biological diversity, the potential of natural insurance in tropical forests is particularly high. During the COVID-19 pandemic, van Vliet et al. (2022) found that fishing and hunting for food increased among forest peoples in South America. Our study explores whether nature provides insurance against the most devastating shock people may experience—the death of a close family member—in the Peruvian Amazon. Peru ranks among the countries most severely affected by COVID-19 (Abizaid et al. 2020, Taylor 2021). Following an in-person survey of almost 4000 households in 235 randomly selected communities before the pandemic, we conducted telephone surveys with over 400 communities during the early phase of the pandemic (Takasaki et al. 2022). These surveys were part of the Peruvian Amazon Rural Livelihoods and Poverty (PARLAP) project (Coomes et al. 2016, 2020), which covers over 900 riverine communities without road access in four major river basins: the Amazon, Napo, Pastaza, and Ucayali (near 120,000 km², counting for approximately 27% of the Peruvian Amazon, Fig. 1), the largest as yet undertaken in tropical forests.

Our study is novel in three important ways. First, to the best of our knowledge, this is the first empirical study on natural insurance specifically against health shocks during the COVID-19 pandemic. By combining unique survey data before and during the early phase of the pandemic, we show how the role of nature as insurance changed during the COVID-19 pandemic. Extant works examine the consequences of mortality caused by other viruses, including HIV/AIDS and Ebola in various biomes, but as of yet, none have focused on tropical forest environments (Yamano and Jayne 2004, Beegle 2005, de la Fuente et al. 2020). Previous studies in tropical forests examine natural insurance against adverse shocks, such as natural calamities, income loss, and illness, but not mortality (Pattanayak and Sills 2001, Takasaki et al. 2004, Coomes et al. 2010, Wunder et al. 2014); this is primarily because mortality is too uncommon for statistical analysis of small survey samples.

Second, distinct from other telephone surveys on COVID-19 conducted at the household level, our large-scale community telephone surveys capture wild resource use at a basin-wide spatial scale (i.e., landscape scale), thus addressing an important lacuna in forest livelihoods research: limited external validity of studies covering a small number of communities (Oldekop et al. 2020, Hajjar et al. 2021). In particular, our surveys considered differences in social and cultural identity among forest peoples, covering both Indigenous and mestizo communities. Mestizos (folk peoples; locally known as ribereños) are descendants over many generations of Iberian and Indigenous peoples living in the region (Chibnik 1991). Concerns about the fate of Indigenous peoples, including in Amazonia, have been prominent in media reports and research on COVID-19, which are important, but tend to overlook other marginalized populations (Flores-Ramírez et al. 2021, Mallard et al. 2021, Soto-Cabezas et al. 2022).

Finally, our study focuses on the often-overlooked nexus among risk sharing, wild resource use, and gender. We assess two forms of natural insurance: (1) self-insurance, whereby households experiencing an adverse shock, in our case the death of a household member, rely on wild resources to supplement household cash income and consumption, which has been the focus of previous studies (Takasaki et al. 2004, Wunder et al. 2014); and, (2) mutual insurance, whereby households increase wild resource use to help other households that experienced an adverse shock (i.e., risk sharing; Takasaki 2011, 2012). For poor forest peoples with limited coping options, natural mutual insurance can be significant. Because wild resource harvesting is often gender dependent (Mai et al. 2011, Sunderland et al. 2014, Colfer et al. 2016), the effectiveness of natural self-insurance against mortality may vary based on the gender of the deceased person. In our study area, the harvest of wild resources is done predominantly by males (Espinosa 2010). As such, the death of working-age male adults reduces available household labor for resource use, even if households may employ natural self-insurance against the death of others. Similarly, gender may drive natural mutual insurance if risk sharing arrangements also depend on the gender of the deceased person. Our study is the first to address gendered risk sharing against mortality.

METHODS

Study area and COVID-19

The Departments of Loreto and Ucayali, where our study area is located, cover about 85% of the area of the Peruvian Amazon (Fig. 1), consisting of humid tropical forest and extensive wetlands at < 200 m of elevation. The estimated population of Loreto and Ucayali in 2017 was about 1,380,000 with 73% living in urban areas (INEI 2018). The two primary cities of Iquitos and Pucallpa serve as major markets and administrative centers. Iquitos can be reached only by riverboat or by air; Pucallpa has also been connected with Lima, the capital city of Peru, by road since the 1940s. Small towns and many smaller communities line the main rivers and tributaries. Flood waters rise and fall over a range of 8–10 m, demarcating the seasons. Rural communities are situated on the upland (terra firme) above river flood levels, or in the lowland (várzea) on the river floodplain. Forest peoples (both Indigenous and mestizo) practice agriculture, fishing, hunting, timber and non-timber forest product (NTFP) gathering, and small livestock raising for subsistence and cash earnings, sending produce to markets by boat (Chibnik 1994, Kvist et al. 2001, Takasaki et al. 2001). The river flood pulse shapes people’s livelihoods (Castello et al. 2015, Endo et al. 2016, Tregidgo et al. 2020): hunting (fishing) is most productive during a high-water (low-water) season when the area of habitat available for game (fish) species is reduced.

In the Peruvian Amazon, COVID-19 spread in two waves during 2020 (April–June; August) and mortality was highest early during the first wave, especially in Iquitos and Pucallpa (Fig. S1; Álvarez-Antonio et al. 2021). A national lockdown, which was declared in mid-March 2020, lasted until early May 2020, when restrictions were gradually relaxed (Calderon-Anyosa et al. 2021). By the end of June 2020, the lockdown was lifted, though various regional restrictions such as curfews and mobility restrictions were maintained.

Surveys

Community and household surveys before COVID-19

Table 1 summarizes our surveys. In the PARLAP project (https://parlap.geog.mcgill.ca), we selected four major river basins, the Amazon, Napo, Pastaza, and Ucayali, to capture the diversity of ecological conditions, economic activities, history, and indigeneity of peoples in the region (Fig. 1). We sought to cover all communities in each river basin in the study area. Two field teams were guided by data from the 2007 Census from the Peruvian Instituto Nacional de Estadística e Informática (INEI 2015), maps from the Instituto del Bien Común (IBC) for their census of Indigenous communities (Smith et al. 2003, Benavides 2010), and Google Earth imagery, supplemented by local enquiries by the teams to identify unmapped settlements. The community survey conducted from September 2012 through March 2014 reached a total of 919 communities (436 Indigenous, 470 mestizo, and 13 colonist), which we estimate represents 92% of all communities in the study area (i.e., a near census). Each community was geo-referenced using a handheld Garmin GPS unit. The survey collected information through a focus group-based in-person interview among community leaders and elders following a structured questionnaire.

For the household survey, we randomly sampled a total of 235 communities using sub-basins, indigeneity, and resource endowments at community establishment at the current site as strata (Fig. 1). From August 2014 through July 2016, the household survey was conducted among all available households if the community consisted of up to 20 households, or 20 randomly sampled households in communities with more than 20 households (mean: 16.7 households per community). The household survey, which covered a total of 3929 households, collected information through in-person interviews with a head of household following a structured questionnaire. The analysis sample consisted of 3923 households in 235 communities with no missing information for the key variables analyzed. This sampling design ensures that our community and household samples for the household survey are representative for the PARLAP study area.

COVID-19 surveys

During the early phase of the pandemic in 2020, it was infeasible to conduct an in-person survey and we adopted a telephone survey. Excluding district capitals and communities with a health center from the 919 communities covered in the community survey, the remaining 893 communities were eligible for the COVID-19 survey. Our baseline telephone survey, which was conducted in July 2020, covered 469 communities (53% of the 893 eligible communities; 369 in Loreto, 100 in Ucayali). We subsequently conducted a follow-up telephone survey in August and early September 2020, reaching 435 of the 469 communities in the baseline sample (7% attrition). The timing of both surveys coincided with the low-water season.

Our telephone surveys sought information from community leaders following a structured questionnaire. With the suspension of public telephone service since November 2019 and an unreliable radiophone system, we relied mostly on cell phone contact. As the lockdown was gradually relaxed beginning in May 2020 and the economy was reactivated, people became more available and mobile. Our field teams visited ports and markets in Iquitos and Pucallpa to find people from the target communities. Some telephone interviews were arranged through an intermediary when people from the target communities visited a town where the intermediary lived. In these ways the surveys contacted people in communities without telephone access.

The analysis sample consisted of 421 out of the 435 communities in the follow-up sample without missing information for the key variables analyzed (3% missing data). Out of these 421 communities, 123 were also covered in the household survey (Fig. 1).

Comparability of community samples

Distinct from the communities in the household survey, the non-randomly sampled communities in the COVID-19 surveys are not representative for the PARLAP study area: larger Indigenous communities located closer to cities (Iquitos or Pucallpa) and with greater non-main channel open water are somewhat overrepresented (Table S1, column 1). Sampling was not correlated with the availability of public river transportation, telephone access, and the availability of a health facility. The communities in the COVID-19 and household surveys are similar (Table S1, column 2). These results suggest that while the two community samples are comparable, the findings from the COVID-19 sample may not be generalizable to the whole PARLAP study area. The limitation of external validity is a common problem of telephone surveys during the pandemic.

Mortality and resource use outcomes

Before COVID-19

The household survey in 2014–2016 collected information about all deceased household members, including (1) the age at which the individual died and (2) his/her age at the time of the survey if the person had been alive. We consider death with the difference of these two measures of ages (henceforth age difference) being 0 or 1 as our base measure of recent death. This includes all deaths in the 12 months prior to the survey and some deaths 12–24 months prior, depending on birthdays. Table S2 reports the number of deaths by gender-age cohort.

The household survey also collected information about the quantity harvested and quantity sold in the 12 months prior to the interview for each of the following wild resources: fish during a low/high-water season; game meat during a low/high-water season; NTFPs (moriche palm fruit, Mauritia flexuosa; heart of palm, Euterpe precatoria; leaf products); timber; and aquarium fish. Using the prevailing market price for each product in 2015 based on data collected in all 16 major markets in the region, we calculated income (cash earned and subsistence combined). Fishing, hunting, gathering of bulky palm products, timber harvesting, and aquarium fish collection are typically carried out by males; gathering of leaf products is done by both males and females.

During COVID-19

The COVID-19 baseline survey collected information about the cumulative number of deaths from mid-March 2020 until the time of the baseline survey. This period roughly covers the first four months of the pandemic, including the first wave of the spread of COVID-19 (April–June 2020; Fig. S1).

The COVID-19 follow-up survey conducted during the second wave of the spread of COVID-19 asked whether “people rely more on wild resources” in reference to COVID-19 and lockdown measures (this question was not included in the baseline survey). This measure captures increased reliance on wild resources among people in the community as perceived by community leaders. After asking about wild resources in general, the survey solicited information about wild resources using the following categories: fish, game, timber, NTFPs, and other resources. Increased reliance on timber and other was much less common than fish, game, and NTFPs.

Empirical design

Before COVID-19

We estimate impacts of mortality using ordinary least squares (OLS) regression specifications of the form

(1)

where Y_icb is an outcome such as private transfers received and income from wild resource of household i at community c in basin b; Death in household_ic is an indicator variable for death of at least one household member with age difference being 0 or 1 in household i at community c; Death in community_c is a vector of three indicator variables for death of at least one working-age male adult (ages 18–64), working-age female adult, and child (ages 0–17) or elder (ages 65+) with age difference being 0 or 1 at community c; X_ic is a vector of household-level covariates, including interviewer fixed effects (9 interviewers) and interview year-month fixed effects (which capture seasonality and aggregate time trend); X_c is a vector of community-level covariates; ϕ_b is a vector of basin fixed effects (6 basins); and ε_icb is an error term. Inference is based on robust standard errors clustered by community. Death in household captures risk sharing and natural self-insurance against mortality experienced by own household, and, conditional on this measure, Death in community captures potential natural mutual insurance against mortality experienced by other households. Our key estimates are β₁ and β₂, which capture the combination of ex ante and ex post coping responses to death; an ex ante response corresponds to an anticipated death (Beegle 2005). Note that the reference periods of the outcome (past 12 months) and death (all deaths in past 12 months and some deaths 12–24 months ago) overlap. Because only one working-age male adult in the sample died from a serious accident in the previous 12 months, it is safe to assume that deaths were not caused by wild resource use; that is, reverse causality is ruled out. Table S2A shows the definition and descriptive statistics of these shock measures and covariates.

The identifying assumption in equation (1) is that controlling for basin fixed effects, interviewer fixed effects, interview year-month fixed effects, and covariates, mortality shocks are exogenous determinants of the outcome variables. Death in household is mostly uncorrelated with household- and community-level covariates except for female headship and age of the household head (the joint significance test is not significant, Table S3). The correlations of female headship and age of household head with a recent death in the household are obvious: after the passing of her male partner, a widow became a household head; an elderly household head tended to have an elderly spouse or a spouse who passed away. The three measures for Death in community are not correlated with community-level covariates (the joint significance tests are not significant, Table S3).

During COVID-19

We estimate impacts of COVID-19 case and mortality using OLS regression specifications of the form

(2)

where Y_cbt is an indicator variable for increased reliance on: wild resources, fish, game, NTFPs of community c in basin b at follow-up survey t; COVID-19_c,t-1 is an indicator variable for at least one COVID-19 case including suspected one by the time of baseline COVID survey t-1 at community c; Any death_c,t-1 is an indicator variable for at least one death due to any cause from mid-March through the time of baseline survey t-1 at community c; X_c is a vector of covariates, including interviewer fixed effects (8 interviewers); ϕ_b is a vector of basin fixed effects that capture basin heterogeneity (6 basins); and ε_cb is an error term. Inference is based on robust standard errors. With very limited health facilities and testing for COVID-19 in the study area (19% of communities had health facilities), information about confirmed cases is incomplete. Although suspected cases may be inaccurate, any COVID-19 case captures people’s perceived risk of infection in the community that would have underlain local behavior. Death incidence, which is a memorable event, should be reliable, even if respondents’ perceptions about whether deaths were caused by COVID-19 may not necessarily be accurate. We consider death incidence, not the number of deaths, because multiple deaths were relatively uncommon across communities. Our key estimates are β₁ and β₂, which capture ex post coping responses to case and death incidence, respectively. Table S2B shows the definition and descriptive statistics of these shock measures, as well as disaggregated shock measures by the cause of death, the age (non-elderly vs. elderly) and gender of the deceased, and covariates.

The identifying assumption in equation (2) is that controlling for basin fixed effects, interviewer fixed effects, and covariates, the health shocks at the baseline are exogenous determinants of reliance on wild resources at the follow-up. COVID-19 case and Any death are mostly uncorrelated with the covariates (the joint significance tests are not significant; Table S4). This is also the case for mortality due to COVID-19 (COVID-19 death) and mortality due to other causes (Non-COVID-19 death), although the latter mortality incidence was more likely in larger communities.

Heterogeneity analysis

When we conduct heterogeneity analysis by one of the covariates, we add an interaction term of one of the shock variables and the covariate to equation (1) or (2), estimating the marginal effects of the shock variable according to the value of the covariate.

RESULTS

Natural insurance before COVID-19

Among the 3923 households in 235 randomly selected communities surveyed in 2014–2016, a total of 69 households (1.8%) in 61 communities (26%) had recently experienced the death of a household member (73 deaths in total, within approximately two years, Table S2A; S5A reports the number of deaths by gender-age cohort). In the 12 months prior to the household survey, 88% of households participated in wild resource extraction and the mean income (subsistence and cash earnings) was S/ 3522, or 36% of total household income (S/ 9845, or US$3077; S/ 3.2 = US$1); participation in fishing was much more common than in hunting and NTFP gathering (85% vs. about 30%), and income from fish, game, and NTFPs, respectively, constituted 82%, 7%, and 3% of the total income from wild resources (see SI text, S.1. Spatial distribution). Private transfers (remittances, in either cash or in-kind, from people living outside the community) were received by 15% of households and the mean amount was S/ 90.

Households experiencing a death received more substantial private transfers (35% increase), signaling risk sharing against mortality (Fig. 2; Table S6 reports the full regression results). The estimated impacts were greater among households with stronger social networks (i.e., both male and female head of household were born in the community; Fig. 3A) and larger kin groups (Fig. S2), suggesting stronger risk sharing within these networks and groups as found elsewhere in the literature (Fafchamps 2011). At the same time, households experiencing a death decreased income from wild resources (by 45%), with specific responses varying across resources: fishing and NTFP gathering decreased and hunting increased. Admittedly, these estimates for uncommon death events are imprecise because of weak statistical power except for hunting, which had a large impact (55%). The responses to the death of other community members depended on the gender of the deceased. On one hand, income from wild resources increased in response to the death of a working-age male adult in the community (by 36%). Fishing, hunting, and NTFP gathering all increased, although hunting and NTFP gathering increased more relative to fishing (Fig. 2, Table S6). On the other hand, the estimated impacts of a death of a working-age female adult and a child or an elder are small and not statistically significant. These estimation results are robust to a battery of robustness checks (see SI text, S.2. Robustness check).

The results for receipt of private transfers and participation in wild resource extraction are consistent (Fig. S3) as are the distributions of private transfers and income from wild resources (Fig. 4). The distribution of private transfers clearly indicates that gendered risk sharing occurred: greater risk sharing against the death of a working-age male adult in household vis-a-vis others’ death (Fig. 4A).

When we consider any death of a working-age adult (male and female combined; with uncommon death, it is infeasible to distinguish gender) and any death of a child or an elder in household separately, households that experienced the death of a working-age adult decreased wild resource use and fishing, but increased hunting in a statistically significant way (Fig. 5). Given that fishing and hunting are done primarily by males, the reduction in fishing (and thus wild resources) was driven by reduced labor endowments due to the loss of a working-age male adult and the increased reliance on hunting was driven by self-insurance against the death of a working-age female adult. The distributions of income from wild resources are consistent with gendered natural self-insurance: a greater decrease in wild resource use and fishing for the death of a working-age male adult than for other’s death and a greater increase in hunting for the death of a working-age female adult than for other’s (Figs. 4B-D).

Natural insurance during COVID-19

COVID-19 spread throughout the Peruvian Amazon in two waves in 2020 (April–June; August; Fig. S1). Mortality was highest early in the first wave (primarily in the cities of Iquitos and Pucallpa). The cumulative number of deaths from mid-March 2020 until the time of the COVID-19 baseline survey in July 2020 was 123 deaths in total regardless of cause, or 1.03 out of 1000 persons (Table S5B, C). Mortality due to COVID-19 (including suspected cases) was more common than that due to other causes (0.61 vs. 0.42 out of 1000); this was especially so for the elders and females.

According to official data from the Peruvian Ministry of Health, the mortality rate due to COVID-19 from mid-March until the last day of the baseline survey (2 August 2020) in Loreto and Ucayali (Fig. S1) was 2.70 out of 1000 persons (the comparable figure for the whole country was 1.86; Table S5C). We focus on this specific period to make a comparison between our survey data and the government data. These results suggest that mortality due to COVID-19 in our surveyed communities was lower than estimates for the region, which were driven primarily by mortality in urban areas.

According to death with age difference being 1, which effectively captures deaths during the one-year period before the pandemic, the mortality rate was 2.21 out 1000 persons (Table S5C). This figure was higher than the predicted annual mortality rate because of causes other than COVID-19 (0.41×3) and lower than the predicted annual overall mortality rate including deaths due to COVID-19 (1.03×3; Table S5C).

In our analysis sample (n = 421 communities), any COVID-19 case and any death were reported in 91% and 17% of communities, respectively (123 deaths in total, Table S2B; S5B reports the number of deaths by gender-age cohort). At the time of the follow-up telephone survey, people relied more on wild resources (86% of communities), especially fish (83%), game (71%), and NTFPs (58%) (see SI text, S.1. Spatial distribution). People in nearly all communities used traditional medicine to prevent and treat COVID-19, i.e., medicinal plants, such as ginger, matico, and lemon, and traditional healers among Indigenous peoples (Takasaki et al. 2021).

Our estimation results for the impacts of health shocks on people’s reliance on wild resources among communities show that (1) with any COVID-19 case, people decreased reliance on game (by 9%) without altering reliance on fish and NTFPs; and, (2) with any death, people increased reliance on wild resources, especially fish (by 13%), but reliance on game and NTFPs did not change significantly (Fig. 6; Table S7 reports the full regression results). These results are robust to a battery of robustness checks (see SI text, S.2. Robustness check). Results are similar when we consider any death due to COVID-19 and any death due to other causes separately (Fig. S4).

DISCUSSION

Natural insurance before COVID-19

Before the pandemic, did wild resources serve as mutual insurance against death? Hunting and gathering responses to the death of a working-age male adult in the community are significant only among households with strong social networks as evidenced by private transfers described above (Fig. S2). These matched responses point to gendered natural mutual insurance: people increased hunting and NTFP gathering to supplement risk sharing for households that experienced the death of a working-age male adult. Because hunting (NTFP gathering) is a high(low)-risk, high(low)-return activity (Takasaki et al. 2004), in combination, these two activities can be effective for mutual insurance against mortality.

In contrast to private transfers, the self-insurance role of hunting was significant only among households with weak social networks (Fig. 3B), suggesting that households that had limited means to share risk socially resorted to hunting for self-insurance as an alternative coping strategy. As such, gender drove natural self-insurance not only through changes in labor endowments, but also through gendered risk sharing.

The comparison of deaths with different timing shows that both natural self-insurance and mutual insurance were short-run phenomena corresponding to a recent death (see SI text, S.3. Timing of death before COVID-19). These findings on mortality are distinct from the limited role of hunting as self-insurance against other shocks (e.g., large flood, illness) found in previous works in the region (Takasaki et al. 2004, Coomes et al. 2010), although game meat sharing was found to be a common practice (Nunes et al. 2019) and game is a major source of traditional remedies (Ferreira et al. 2013, Lee et al. 2020).

Vulnerability and natural insurance during COVID-19

Wild resource use in response to mortality during the pandemic was different from the pre-pandemic period: people relied more on fishing, and interestingly, hunting decreased in response to a COVID-19 case.

Vulnerability to COVID-19 while hunting

Distinct from illness, which is an idiosyncratic shock, COVID-19 can be considered as a covariate shock for people in rural communities. Considering the risk of infection, people could alter their behavior. In contrast to fishing and NTFP gathering, which can be done near communities, hunting typically requires longer and more distant trips from communities, effectively leaving the hunter’s family alone for the duration of a hunting trip. A decrease in hunting in response to a COVID-19 case was significant only in communities without telephone access and without a health facility (Fig. 7A). Poor communication infrastructure is likely to have caused important delays and gaps in information, or misunderstandings about COVID-19, and compounded with the limited access to health services, people were likely seriously concerned about infection. These results suggest that people decreased hunting to avoid leaving their families alone while at risk of infection. In contrast, the role of hunting as self-insurance and mutual insurance (as well as NTFP gathering as mutual insurance) before the pandemic were significant only in communities without a health facility (Figs. 3B-D). These opposite patterns buttress the perceived risk of infection driving the distinct hunting behaviors during the pandemic.

Fishing and NTFP gathering as natural insurance

The shift from hunting to fishing is consistent with previous works showing the significant role of fishing as self-insurance against various adverse shocks in the region (Takasaki et al. 2004, Coomes et al. 2010), including a cholera outbreak in the early 1990s (McDaniel 1997). Although we cannot test risk sharing per se or distinguish self-insurance and mutual insurance during the pandemic, we can infer from our findings on hunting before the pandemic that fishing also served both purposes (self-insurance and mutual insurance).

How well fishing served as natural insurance must depend on market access, because although public river transportation was commonly available, COVID-19 restrictions on mobility reduced active transportation, weakening market access for communities (especially access to markets in towns and cities; Fig. 1). The fishing response to mortality was significant only in communities with active public transportation at the time of the follow-up survey (64% of communities; Fig. 7B), indicating that people sought to increase cash earnings by selling their catch. Natural insurance was thus constrained by mobility restrictions.

In contrast with hunting before the pandemic, fishing as a response to mortality was not differentiated by the age (non-elderly vs. elderly) or gender of the deceased person. People increased NTFP gathering in response only to a female death during the pandemic (Fig. S4), thus suggesting that people also adjusted NTFP gathering as natural insurance. Unfortunately, we are unable to discern whether gendered risk sharing was altered.

Although our results for hunting and fishing may have been driven by other factors such as seasonality, return migration from cities, the provision of a public safety net, social norms, and a shift away from agriculture, we found no evidence of such (see SI text, S.4. Other potential mechanisms during COVID-19). This buttresses our conjecture that vulnerability and natural insurance were key driving forces.

Indigeneity and heterogeneity

Prior to the pandemic, private transfers and hunting in response to a death in the household were not differentiated by indigeneity, yet an increase in hunting and NTFP gathering in response to the death of a working-age male adult in the community was significant in Indigenous communities only (Fig. 3). Thus, although gendered risk sharing and the self-insurance role of hunting were significant in both Indigenous and mestizo communities, natural mutual insurance through hunting and gathering was significant only in Indigenous communities. This may be because Indigenous peoples had fewer pooled resources to be shared with one another and/or followed distinct socio-cultural norms. In contrast, during the pandemic, both hunting and fishing results are similar for Indigenous and mestizo communities (Fig. 7), suggesting that Indigenous peoples and mestizos converged in how they adjusted natural insurance to deal with a death during the pandemic. Consistently, heterogeneity analyses based on population size, distance to city, and environmental conditions point to significant changes in fishing as insurance before and during the pandemic and suggest that natural insurance was not constrained by local resource endowments (see SI text, S.5. Heterogeneity).

CONCLUSION

Our study in the Peruvian Amazon revealed that nature provided insurance, both self-insurance and mutual insurance, against mortality before and during the early phase of the COVID-19 pandemic, but in different ways. Although caution is needed regarding the generalizability of our findings, they can have the following broad implications for research and policy.

Natural insurance can constitute a vital safety net for forest peoples with limited access to social assistance and health services, and as such, its scope is shaped by conservation and sustainable management of local wild resources. Local resource use including natural insurance shapes the dynamics of wild resources and this feedback effect merits further study.

The relative importance of different types of natural insurance is contextually contingent, varying based on their logistics (fishing near communities vs. hunting in distant places) in specific locales. In our study setting with rich aquatic habitat, fishing can have an advantage in reducing vulnerability under pandemics as people are able to have access to wild resources without leaving their families alone while at risk of infection. Depending on locales, balanced conservation across resources (fish, game, NTFPs) is needed to maintain multiple options of natural insurance and the resilience of forest peoples.

The effectiveness of natural insurance for forest peoples depends on market access, which can be constrained during pandemics because of mobility restrictions imposed to reduce contagion. Although relaxing such restrictions can contribute to the reactivation of the economy and the enhancement of the potential for natural insurance, such measures could further the spread of the virus (Okyere et al. 2020, Takasaki et al. 2022). In this way, natural insurance adds to the complexity of this central trade-off in policy making under pandemics. The scope of natural insurance can be shaped by broad government responses to pandemics in complex ways.

People in different socio-cultural groups can react in similar ways under pandemics as indicated by the observed convergence of natural insurance among Indigenous peoples and mestizos. Policy design and implementation need to consider such social dynamics depending on social contexts.

Gendered wild resource harvesting, which is largely fixed by socio-cultural norms, may underlie broad informal institutions as seen in gendered risk sharing. Further gender analysis may shed new light on forest peoples’ welfare, equity, norms, and local resource management (Razafindratsima et al. 2021).

Landscape-scale studies (Coomes et al. 2016, 2020) are promising to further understanding of the link between human factors and behavior at a micro level, such as gender and risk sharing, and large-scale environmental features including land cover and biodiversity.

LITERATURE CITED

Abizaid, C., L. Á. Collado Panduro, and S. Gonzales Egusquiza. 2020. Pobreza y medios de subsistencia en la Amazonía Peruana en tiempos del COVID-19. Journal of Latin American Geography 19(3):202-214. https://doi.org/10.1353/lag.0.0141

Álvarez-Antonio, C., G. Meza-Sánchez, C. Calampa, W. Casanova, C. Carey, F. Alava, H. Rodríguez-Ferrucci, and A. M. Quispe. 2021. Seroprevalence of anti-SARS-CoV-2 antibodies in Iquitos, Peru in July and August, 2020: a population-based study. Lancet Global Health 9(7):e925-e931. https://doi.org/10.1016/S2214-109X(21)00173-X

Angelsen, A., and T. Dokken. 2018. Climate exposure, vulnerability and environmental reliance: a cross-section analysis of structural and stochastic poverty. Environment and Development Economics 23(3):257-278. https://doi.org/10.1017/S1355770X18000013

Beegle, K. 2005. Labor effects of adult mortality in Tanzanian households. Economic Development and Cultural Change 53(3):655-683. https://doi.org/10.1596/1813-9450-3062

Benavides, M. 2010. Atlas de comunidades nativas y áreas naturales protegidas del nordeste de la Amazonía peruana. Instituto del Bien Común, Lima, Peru.

Calderon-Anyosa, R. J., U. Bilal, and J. S. Kaufman. 2021. Variation in non-external and external causes of death in Peru in relation to the COVID-19 lockdown. Yale Journal of Biology and Medicine 94(1):23-40.

Castello, L., V. J. Isaac, and R. Thapa. 2015. Flood pulse effects on multispecies fishery yields in the Lower Amazon. Royal Society Open Science 2(11):150299. https://doi.org/10.1098/rsos.150299

Chibnik, M. 1991. Quasi-ethnic groups in Amazonia. Ethnology 30(2):167-182. https://doi.org/10.2307/3773408

Chibnik, M. 1994. Risky rivers: the economics and politics of floodplain farming in Amazonia. The University of Arizona Press, Tucson, Arizona, USA. https://doi.org/10.2307/j.ctv1qwwjqj

Colfer, C. J. P., B. S. Basnett, and M. Elias, editors. 2016. Gender and forests: climate change, tenure, value chains and emerging issues. Routledge, London, UK. https://doi.org/10.4324/978131566624

Cooke, S. J., W. M. Twardek, A. J. Lynch, I. G. Cowx, J. D. Olden, S. Funge-Smith, K. Lorenzen, R. Arlinghaus, Y. Chen, O. L. Weyl, et al. 2021. A global perspective on the influence of the COVID-19 pandemic on freshwater fish biodiversity. Biological Conservation 253:108932. https://doi.org/10.1016/j.biocon.2020.108932

Coomes, O. T., Y. Takasaki, and C. Abizaid. 2020. Impoverishment of local wild resources in western Amazonia: a large-scale community survey of local ecological knowledge. Environmental Research Letters 15(7):074016. https://doi.org/10.1088/1748-9326/ab83ad

Coomes, O. T., Y. Takasaki, C. Abizaid, and J. P. Arroyo-Mora. 2016. Environmental and market determinants of economic orientation among rain forest communities: evidence from a large-scale survey in western Amazonia. Ecological Economics 129:260-271. https://doi.org/10.1016/j.ecolecon.2016.06.001

Coomes, O. T., Y. Takasaki, C. Abizaid, and B. L. Barham. 2010. Floodplain fisheries as natural insurance for the rural poor in tropical forest environments: evidence from Amazonia. Fisheries Management and Ecology 17(6):513-521. https://doi.org/10.1111/j.1365-2400.2010.00750.x

de la Fuente, A., H. G. Jacoby, and K. G. Lawin. 2020. Impact of the West African Ebola epidemic on agricultural production and rural welfare: evidence from Liberia. Journal of African Economies 29(5):454-474. https://doi.org/10.1093/jae/ejaa002

Diffenbaugh, N. S. 2022. COVID-19 and the environment: short-run and potential long-run impacts. Annual Review of Environment and Resources 47:65-90. https://doi.org/10.1146/annurev-environ-120920-125207

Egger, D., E. Miguel, S. S. Warren, A. Shenoy, E. Collins, D. Karlan, D. Parkerson, A. M. Mobarak, G. Fink, C. Udry, M. Walker, J. Haushofer, M. Larreboure, S. Athey, P. Lopez-Pena, S. Benhachmi, M. Humphreys, L. Lowe, N. F. Meriggi, A. Wabwire, C. A. Davis, U. J. Pape, T. Graff, M. Voors, C. Nekesa, and C. Vernot. 2021. Falling living standards during the COVID-19 crisis: quantitative evidence from nine developing countries. Science Advances 7:eabe0997. https://doi.org/10.1126/sciadv.abe0997

Endo, W., C. A. Peres, and T. Haugaasen. 2016. Flood pulse dynamics affects exploitation of both aquatic and terrestrial prey by Amazonian floodplain settlements. Biological Conservation 201:129-136. https://doi.org/10.1016/j.biocon.2016.07.006

Espinosa, M. C. 2010. Why gender in wildlife conservation? Notes from the Peruvian Amazon. Open Anthropology Journal 3(1):230-241. https://doi.org/10.2174/1874912701003010230

Fafchamps, M. 2011. Risk sharing between households. Pages 1255-1279 in J. Benhabib, A. Bisin, and M. O. Jackson, editors. Handbook of social economics. Volume 1. Elsevier, Amsterdam, The Netherlands. https://doi.org/10.1016/B978-0-444-53707-2.00007-4

Ferreira, F. S., H. Fernandes-Ferreira, N. A. L. Neto, S. V. Brito, and R. R. Alves. 2013. The trade of medicinal animals in Brazil: current status and perspectives. Biodiversity and Conservation 22(4):839-870. https://doi.org/10.1007/s10531-013-0475-7

Fiorella, K. J., J. Coffin-Schmitt, K. M. Gaynor, G. H. Gregory, R. Rasolofoson, and K. L. Seto. 2020. Feedbacks from human health to household reliance on natural resources during the COVID-19 pandemic. Lancet Planetary Health 4(10):e441-e442. https://doi.org/10.1016/S2542-5196(20)30199-6

Flores-Ramírez, R., A. A. Berumen-Rodríguez, M. A. Martínez-Castillo, L. E. Alcántara-Quintana, F. Díaz-Barriga, and L. Díaz de León-Martínez. 2021. A review of environmental risks and vulnerability factors of indigenous populations from Latin America and the Caribbean in the face of the COVID-19. Global Public Health 16(7):975-999. https://doi.org/10.1080/17441692.2021.1923777

Gibbons, D. W., C. Sandbrook, W. J. Sutherland, R. Akter, R. Bradbury, S. Broad, A. Clements, H. Q. Crick, J. Elliott, N. Gyeltshen, et al. 2022. The relative importance of COVID-19 pandemic impacts on biodiversity conservation globally. Conservation Biology 36(1):e13781. https://doi.org/10.1111/cobi.13781

Hajjar, R., J. Z. Cheek, P. Jagger, J. Kamoto, P. Newton, J. Oldekop, and O. H. Razafindratsima. 2021. Research frontiers on forests, trees, and poverty dynamics. Forest Policy and Economics 131:102554. https://doi.org/10.1016/j.forpol.2021.102554

Instituto Nacional de Estadística e Informática (INEI). 2015. Perú: Estimaciones y Proyecciones de Población Urbana y Rural por Sexo y Edades Quinquenales, según Departamento, 2000-2015. Boletín Especial 19, INEI, Lima, Peru.

Instituto Nacional de Estadística e Informática (INEI). 2018. Resultados Definitivos de los Censos Nacionales 2017 XII de Población, VII de Vivienda y III de Comunidades Indígenas. Tomo 1. INEI, Lima, Peru.

Josephson, A., T. Kilic, and J. D. Michler. 2021. Socioeconomic impacts of COVID-19 in low-income countries. Nature Human Behaviour 5(5):557-565. https://doi.org/10.1038/s41562-021-01096-7

Kvist, L. P., S. Gram, A. C. Cácares, and I. Ore B. 2001. Socio-economy of flood plain households in the Peruvian Amazon. Forest Ecology and Management 150(1-2):175-186. https://doi.org/10.1016/S0378-1127(00)00689-7

Lawler, O. K., H. L. Allan, P. W. Baxter, R. Castagnino, M. C. Tor, L. E. Dann, J. Hungerford, D. Karmacharya, T. J. Lloyd, M. J. López-Jara, et al. 2021. The COVID-19 pandemic is intricately linked to biodiversity loss and ecosystem health. Lancet Planetary Health 5(11):e840-e850. https://doi.org/10.1016/S2542-5196(21)00258-8

Lee, T. M., A. Sigouin, M. Pinedo-Vasquez, and R. Nasi. 2020. The harvest of tropical wildlife for bushmeat and traditional medicine. Annual Review of Environment and Resources 45:145-170. https://doi.org/10.1146/annurev-environ-102016-060827

Lindsey, P., J. Allan, P. Brehony, A. Dickman, A. Robson, C. Begg, H. Bhammar, L. Blanken, T. Breuer, K. Fitzgerald, et al. 2020. Conserving Africa’s wildlife and wildlands through the COVID-19 crisis and beyond. Nature Ecology & Evolution 4(10):1300-1310. https://doi.org/10.1038/s41559-020-1275-6

Mai, Y. H., E. Mwangi, and M. Wan. 2011. Gender analysis in forestry research: looking back and thinking ahead. International Forestry Review 13(2):245-258. https://doi.org/10.1505/146554811797406589

Mallard, A., M. A. Pesantes, C. Zavaleta-Cortijo, and J. Ward. 2021. An urgent call to collect data related to COVID-19 and indigenous populations globally. BMJ Global Health 6(3):e004655. https://doi.org/10.1136/bmjgh-2020-004655

McDaniel, J. 1997. Communal fisheries management in the Peruvian Amazon. Human Organization 56(2):147-152. https://doi.org/10.17730/humo.56.2.377747752t432gu2

McNamara, J., E. J. Robinson, K. Abernethy, D. M. Iponga, H. N. Sackey, J. H. Wright, and E. Milner-Gulland. 2020. COVID-19, systemic crisis, and possible implications for the wild meat trade in Sub-Saharan Africa. Environmental and Resource Economics 76(4):1045-1066. https://doi.org/10.1007/s10640-020-00474-5

Miguel, E., and A. M. Mobarak. 2022. The economics of the COVID-19 pandemic in poor countries. Annual Review of Economics 14:253-285. https://doi.org/10.1146/annurev-economics-051520-025412

Mobarak, A. M., E. Miguel, J. Abaluck, A. Ahuja, M. Alsan, A. Banerjee, E. Breza, A. G. Chandrasekhar, E. Duflo, J. Dzansi, D. Garrett, P. Goldsmith-Pinkham, G. S. Gonsalves, M. M. Hossain, A. Jakubowski, G. Kang, A. Kharel, M. Kremer, N. Meriggi, C. Nekesa, B. A. Olken, S. B. Omer, F. Qadri, H. Rees, B. Salako, M. Voors, S. Warren, and W. Więcek. 2022. End COVID-19 in low- and middle-income countries. Science 375(6585):1105-1110. https://doi.org/10.1126/science.abo4089

Nunes, A. V., R. D. Guariento, B. A. Santos, and E. Fischer. 2019. Wild meat sharing among non-indigenous people in the southwestern Amazon. Behavioral Ecology and Sociobiology 73(2):26. https://doi.org/10.1007/s00265-018-2628-x

Okyere, I., E. O. Chuku, B. Ekumah, D. B. Angnuureng, J. K. Boakye-Appiah, D. J. Mills, R. Babanawo, N. K. Asare, D. W. Aheto, and B. Crawford. 2020. Physical distancing and risk of COVID-19 in small-scale fisheries: a remote sensing assessment in coastal Ghana. Scientific Reports 10(1):22407. https://doi.org/10.1038/s41598-020-79898-4

Oldekop, J. A., L. V. Rasmussen, A. Agrawal, A. J. Bebbington, P. Meyfroidt, D. N. Bengston, A. Blackman, S. Brooks, I. Davidson-Hunt, P. Davies, et al. 2020. Forest-linked livelihoods in a globalized world. Nature Plants 6(12):1400-1407. https://doi.org/10.1038/s41477-020-00814-9

Olivero, J., J. E. Fa, R. Real, A. L. Márquez, M. A. Farfán, J. M. Vargas, D. Gaveau, M. A. Salim, D. Park, J. Suter, et al. 2017. Recent loss of closed forests is associated with Ebola virus disease outbreaks. Scientific Reports 7:14291. https://doi.org/10.1038/s41598-017-14727-9

Pattanayak, S. K., and E. O. Sills. 2001. Do tropical forests provide natural insurance? The microeconomics of non-timber forest product collection in the Brazilian Amazon. Land Economics 77(4):595-612. https://doi.org/10.2307/3146943

Razafindratsima, O. H., J. F. M. Kamoto, E. O. Sills, D. N. Mutta, C. Song, G. Kabwe, S. E. Castle, P. M. Kristjanson, C. M. Ryan, M. Brockhaus, and T. Sunderland. 2021. Reviewing the evidence on the roles of forests and tree-based systems in poverty dynamics. Forest Policy and Economics 131:102576. https://doi.org/10.1016/j.forpol.2021.102576

Smith, R. C., M. Benavides, M. Pariona, and E. Tuesta. 2003. Mapping the past and the future: geomatics and indigenous territories in the Peruvian Amazon. Human Organization 62(4):357-368. https://doi.org/10.17730/humo.62.4.3d82x31cywm8wl25

Soto-Cabezas, M. G., M. F. Reyes, A. N. Soriano, J. P. V. Rodríguez, L. O. Ibargüen, K. S. Martel, N. F. Jaime, and C. V. Munayco. 2022. COVID-19 among Amazonian indigenous in Peru: mortality, incidence, and clinical characteristics. Journal of Public Health 44(3):e359-e365. https://doi.org/10.1093/pubmed/fdac058

Stokes, G. L., A. J. Lynch, B. S. Lowe, S. Funge-Smith, J. Valbo‐Jørgensen, and S. J. Smidt. 2020. COVID-19 pandemic impacts on global inland fisheries. Proceedings of the National Academy of Sciences in the United States of America 117(47):29419-29421. https://doi.org/10.1073/pnas.2014016117

Sunderland, T., R. Achdiawan, A. Angelsen, R. Babigumira, A. Ickowitz, F. Paumgarten, V. Reyes-García, and G. Shively. 2014. Challenging perceptions about men, women, and forest product use: a global comparative study. World Development 64:S56-S66. https://doi.org/10.1016/j.worlddev.2014.03.003

Takasaki, Y. 2011. Do the commons help augment mutual insurance among the poor? World Development 39(3):429-438. https://doi.org/10.1016/j.worlddev.2010.08.019

Takasaki, Y. 2012. Natural disasters, gender and handicrafts. Journal of Development Studies 48(1):115-132. https://doi.org/10.1080/00220388.2011.598509

Takasaki, Y., C. Abizaid, and O. T. Coomes. 2022. COVID-19 contagion across remote communities in tropical forests. Scientific Reports 12:20727. https://doi.org/10.1038/s41598-022-25238-7

Takasaki, Y., B. L. Barham, and O. T. Coomes. 2001. Amazonian peasants, rain forest use, and income generation: the role of wealth and geographical factors. Society and Natural Resources 14(4):291-308. https://doi.org/10.1080/08941920151080237

Takasaki, Y., B. L. Barham, and O. T. Coomes. 2004. Risk coping strategies in tropical forests: floods, illness, and resource extraction. Environmental and Development Economics 9(2):203-224. https://doi.org/10.1017/S1355770X03001232

Takasaki, Y., O. T. Coomes, and C. Abizaid. 2021. COVID-19 among rural peoples in the Peruvian Amazon: Policy brief. Peruvian Amazon Rural Livelihoods and Poverty (PARLAP) Project, Policy brief, University of Tokyo, Tokyo. https://doi.org/10.5281/zenodo.4683343

Talman, A., S. Bolton, and J. L. Walson. 2013. Interactions between HIV/AIDS and the environment: toward a syndemic framework. American Journal of Public Health 103(2):253-261. https://doi.org/10.2105/AJPH.2012.300924

Taylor, L. 2021. Covid-19: Why Peru suffers from one of the highest excess death rates in the world. BMJ 372:n611. https://doi.org/10.1136/bmj.n611

Tregidgo, D., J. Barlow, P. S. Pompeu, and L. Parry. 2020. Tough fishing and severe seasonal food insecurity in Amazonian flooded forests. People and Nature 2(2):468-482. https://doi.org/10.1002/pan3.10086

van Vliet, N., A. Sirén, E. A. D. Paemelaere, E. Y. Cuesta Rios, M. S. Santiago Antúnez Correa, T. B. Williams, and S. Quintero. 2022. Immediate effects of the COVID-19 pandemic on the use of wildlife as food among indigenous people and local communities in South America. Ecology and Society 27(3):42. https://doi.org/10.5751/ES-13570-270342

Walker, P. G., C. Whittaker, O. J. Watson, M. Baguelin, P. Winskill, A. Hamlet, B. A. Djafaara, Z. Cucunubá, D. O. Mesa, W. Green, et al. 2020. The impact of COVID-19 and strategies for mitigation and suppression in low- and middle-income countries. Science 369(6502):413-422. https://doi.org/10.1126/science.abc0035

Wunder, S., J. Börner, G. Shively, and M. Wyman. 2014. Safety nets, gap filling and forests: a global-comparative perspective. World Development 64:S29-S42. https://doi.org/10.1016/j.worlddev.2014.03.005

Wunder, S., F. Noack, and A. Angelsen. 2018. Climate, crops, and forests: a pan-tropical analysis of household income generation. Environment and Development Economics 23(3):279-297. https://doi.org/10.1017/S1355770X18000116

Yamano, T., and T. S. Jayne. 2004. Measuring the impacts of working-age adult mortality on small-scale farm households in Kenya. World Development 32(1):91-119. https://doi.org/10.1016/j.worlddev.2003.07.004