Figure 1 provides a concise summary of our conceptual framework for how the suppression of malaria over a sustained period could have a long-term impact on the income of an agricultural household. It illustrates a continual process in which an increase in household income both depends on the health of household members and affects their health.

The figure begins by showing how anti-malaria programs could improve the health of an agricultural household in two ways. First, they may have a direct impact on reducing malaria morbidity and mortality. Second, the suppression of malaria may increase the income of the household, which in turn may affect its nutrition security.

Next, the figure illustrates that the decisions of an agricultural household affecting its income depend on the health of its members, its income from the previous growing season and its expectations regarding productivity in the next growing season. The productivity of adults in the household during the next growing season will depend on the number of malaria cases experienced by members of the household.

The impact of suppressing malaria on the harvest value of the household will depend on the decisions the household makes and the productivity of its members. Therefore, the suppression of malaria can impact the income of an agricultural household by affecting harvest values and household expenditures related to malaria morbidity and mortality. Finally, to close the loop, the impact of suppressing malaria on nutrition security depends on changes in household income from the suppression of malaria.

In the previous section, we described a conceptual framework for understanding the potential long-term impact of suppressing malaria on the incomes of agricultural households. In this section, we describe a research framework for examining each component of that conceptual framework.

We focus on four types of decisions made by agricultural households. The first two are which crops to plant and how much land to cultivate. The third is how much investment should be made in agricultural inputs, defined as fertilizer and any technology that could increase crop yield. The fourth is the decision made regarding the level of resources to devote to purchasing anti-malaria interventions.

Our framework assumes that the choices made regarding these four key decisions depend on the resources available to the household and its expectations regarding the productivity of household members. The resources available depend on the value of the harvest during the previous growing season and on expenditures during that growing season.

Few studies have examined the impact of malaria on these four decisions within agricultural households. A study conducted in Paraguay in the early 1970s did examine the relationship between malaria and the decisions made by farmers regarding which crops to plant. It found that an expectation of loss of labor due to malaria led farmers to plant less labor-intensive crops, even though those crops would provide lower revenues than more labor-intensive crops ( Conly, 1975). A study in southern Africa in the 1950s suggested that, after malaria was suppressed, farmers began farming lands that had previously been considered too malarious ( Annecke, 1950).

Suppressing malaria may increase the income of an agricultural household by increasing productivity. Our research framework focuses on two ways that suppressing malaria could increase productivity. First, the suppression of malaria would reduce the number of work days that are lost by adults due to adult malaria cases or to the need for adults to provide care for other household members with malaria. Second, suppressing malaria may increase gender equality between men and women in the household by reducing the number of caregiving days provided by women to household members.

Work days lost during the time labor is most needed for cultivation activities would have a higher value than work days lost during other times. Very few studies have attempted to quantify the value of work days lost based on the time of year the cases occurred, although an extensive number of studies have examined the number of work days lost due directly to malaria morbidity among adults or for caregiving ( Asenso-Okyere & Dzator, 1997; Alaba & Alaba, 2009; Audibert, 1986; Asante & Asenso-Okyere, 2003; Akazili et al., 2007; Akazili, 2000; Badiane & Ulimwengu, 2013; Cropper et al., 2000; Ettling et al., 1994; Nur, 1993; Ettling & Shepard, 1991; Guiguemdé et al., 1997; Girardin et al., 2004; Hailu et al., 2017; Leighton & Foster, 1993; Miller, 1958; Mohamed, 2012; Nur & Mahran, 1988; Oluyole et al., 2011; Omotayo & Oyekale, 2013; Sauerborn et al., 1991; Sauerborn et al., 1996a; Sauerborn et al., 1996b).

A wide range of methodologies have been used to estimate the number of work days lost and caregiving days provided for each case of malaria. As a result, it is difficult to determine the extent to which differences in estimates are due to methodological differences or other factors. Most estimates of work days lost by adults due to a malaria case range from 1 to 9 days, while work days thought to be lost due to caregiving range from 1 to 7 days. Table 1 summarizes the figures arrived at by 12 studies.

One of the first studies to examine the impact of malaria on the productivity of farmers analyzed the natural history of malaria among men aged 20 to 30 in Liberia ( Miller, 1958). Using data from 1957, the study found that the clinical duration of a malaria case was 6.5 days and that the average number of work days lost per case was 3 ( Miller, 1958). An analysis of the economic impact of malaria in Kenya arrived at estimates that two to four work days were lost by adults due to adult malaria cases and caregiving on the basis of focus group interviews conducted in 1992 ( Leighton & Foster, 1993). In Nigeria, focus groups in rural areas estimated that one to three work days were lost ( Leighton & Foster, 1993).

Three studies examined the impact of malaria on the productivity of agricultural households in Ghana ( Akazili, 2000; Asante & Asenso-Okyere, 2003; Tawiah et al., 2016). A study in the Kassena-Nankana District of northern Ghana used a cross-sectional survey of 423 households to estimate 5.3 work days lost for a case of severe malaria and 2.6 work days lost for a mild case ( Akazili, 2000). Another study in Ghana, published in 2003, provides one of the few detailed analyses of differences in the number of lost work and caregiving days between men and women ( Asante & Asenso-Okyere, 2003):

These estimates were based on a cross-sectional survey conducted in three districts from March to May of 2003 among households in which at least one member had experienced a case of malaria during the preceding month ( Asante & Asenso-Okyere, 2003). A more recent cross-sectional study of 947 households in Kintampo, Ghana estimated that a malaria case resulted in 6.4 days of work being lost on the basis of data collected from October 2009 to July 2011 ( Tawiah et al., 2016).

The intensity of malaria transmission in Ethiopia is generally lower than in the equatorial regions of Africa. As a result, we would expect that adults in Ethiopia would have lower levels of acquired immunity and longer durations of symptomatic infections. These expectations are consistent with two studies of the impact of malaria on work days and caregiving days ( Cropper et al., 2000; Deressa et al., 2007). Cropper et al. (2000) found that a malaria case resulted in 18 lost work days for an adult and the provision of 11 days of caregiving by a household member. These findings are based on a cross-sectional survey of 569 individuals in subsistence agricultural households conducted in the Tigray region of northern Ethiopia in January of 1997 ( Cropper et al., 2000). Another study that used cross-sectional data collected from 2,195 households in the Adami Tulu district from October to November of 2003 found that a malaria case resulted in 6.8 lost work days and 6.7 caregiving days ( Deressa et al., 2007).

A study of agricultural households in a district in Burkina Faso assumed that a malaria case in an adult led to a loss of 1–5 work days and that cases among children resulted in one-third of a day to one and two-thirds days being lost to caregiving activities ( Sauerborn et al., 1991). Those assumptions are consistent with a study using cross-sectional data collected from 3,065 patients from three villages in the Bobo-Dioulasso region of Burkina Faso, which found that a malaria case resulted in an average of 4 lost work days ( Guiguemdé et al., 1997). A third study in Burkina Faso collected longitudinal data from 566 households at six time points in 1992, carrying out two surveys during the dry season and four in the rainy season ( Sauerborn et al., 1996b). By collecting data from the same households at different time points, the study could examine seasonal differences in the impact of malaria. Household revenue was lower during the rainy season than during the dry season, but the opportunity cost of time during the rainy season was higher, as that this was the peak production period for the household. The lower revenues and higher opportunity costs during the rainy season led to a reduction in work days missed and caregiving days provided per case ( Sauerborn et al., 1996b).

While studies of the impact of malaria on work days provide a range for days missed per case, the coping mechanisms used by households when work days are lost are less well understood. Two studies of farmers in Gezira, Sudan provide the most detailed evidence in this regard ( Nur, 1993; Nur & Mahran, 1988). The first study concluded that the malaria burden was highest during the months in which labor was most needed for agricultural activities ( Nur & Mahran, 1988). When men were not able to work due to malaria, women and children compensated for the loss of labor, which affected school attendance ( Nur & Mahran, 1988). The same Sudanese community was studied again later ( Nur, 1993), and it was confirmed that although the malaria burden was primarily on men, other members of the household felt an indirect impact:

A growing body of research has shown how gender inequality, also described as the gender gap, affects the productivity of agricultural households in sub-Saharan Africa. A report developed by the United Nations and the World Bank found that reducing the gender gap in agricultural productivity could enable as many as 119,000 people to escape poverty in Uganda ( Women & Others, 2015). A separate study by the World Bank found that women farmers produce significantly less per hectare than men ( O’Sullivan et al., 2014). For example, in Uganda the production of farm plots managed by women was 13% less per acre than the plots that men managed ( O’Sullivan et al., 2014). This study suggests that the provision of community-based childcare centers could reduce the gender gap ( O’Sullivan et al., 2014), positing that childcare duties are, in part, responsible for the gender gap. A study in Ghana found that 83% of caregiving activities for malaria cases in agricultural households were provided by women ( Asante & Asenso-Okyere, 2003). Although the importance of addressing gender inequality to increase agricultural productivity has been recognized, only two studies have rigorously examined differences in malaria-related caregiving provided by women versus men within a household ( Asante & Asenso-Okyere, 2003; Leighton & Foster, 1993).

Agricultural households face three primary types of expenditures related to malaria morbidity and mortality: for interventions to prevent malaria, for interventions to treat malaria cases, and related to a death due to malaria. These expenditures depend on the number of malaria cases a household expects to experience and the number of malaria cases and deaths the household actually experiences. Our framework assumes that the impact of suppressing malaria on the harvest value of a household depends on the impact of malaria suppression on its decisions and productivity.

Many of the studies listed above that examined the number of work days lost due to malaria also quantified the expenditures by households on interventions to prevent or treat malaria infections ( Asenso-Okyere & Dzator, 1997; Asante & Asenso-Okyere, 2003; Akazili, 2000; Asante et al., 2005; Akazili et al., 2007; Alaba & Alaba, 2009; Ettling & Shepard, 1991; Ettling et al., 1994; Guiguemdé et al., 1997; Hailu et al., 2017; McElroy et al., 2009; Omotayo & Oyekale, 2013; Sauerborn et al., 1991; Shepard et al., 1991; Sauerborn et al., 1996a; Tawiah et al., 2016). An additional set of studies focused exclusively on such expenditures ( Guiguemde et al., 1994; Sauerborn et al., 1996b). For example, a study conducted in agricultural households in Burkina Faso found that expenditures on malaria prevention and treatment interventions were, on average, about 4.5% of family income over a 6-month period ( Guiguemde et al., 1994). While a significant number of studies have examined these expenditures over a short period of time, often less than 1 year, we are not aware of any studies that have analyzed how suppressing malaria transmission leads to a long-term change in household expenditures on anti-malaria interventions.

Studies of the cost of treatment for a malaria case often include expenditures for antimalarial drugs, transport and care for the patient, and the time spent seeking care. Expenditures by households to treat malaria cases often fluctuate throughout the year. Sauerborn et al. (1996b) found that expenditures during the rainy season were significantly less than during the dry season, due, in part, to differences in the opportunity costs of time. In some settings, the opportunity cost of the time for a patient to seek care may be higher than the cost of the treatment ( Asenso-Okyere & Dzator, 1997).

Households are often unable to pay for treatment in cash and must use other coping mechanisms ( Sauerborn et al., 1996a) to generate sufficient resources to cover treatment costs. A study of agricultural households in Burkina Faso identified multiple strategies:

Only two studies have examined the relationship between malaria and harvest values. The first examined the impact of malaria on the revenues of farmers from vegetable cultivation ( Girardin et al., 2004). The revenues from cabbage were 53% lower for farmers who missed more than two days of work due to malaria than the revenues of farmers who missed less than two days of work ( Girardin et al., 2004). The second study found that providing agricultural households with access to vector control interventions increased harvest values by 14.7%, likely due to the additional number of days worked ( Fink & Masiye, 2015).

Table 2 summarizes our research framework for examining the long-term impact of suppressing malaria on the income of agricultural households.

The two types of data that could be used to examine the research topics described in the previous section involve the collection of either cross-sectional data or longitudinal data. Cross-sectional data are collected from different individuals in a study population over one or more surveys in order to examine some aspect of that population. Longitudinal data are data collected from the same study population over time in order to understand the effect of changing conditions in the community. Collecting cross-sectional data from a community is generally easier and requires fewer resources than collecting longitudinal data.

With the exception of two studies ( Nur, 1993; Sauerborn et al., 1996b), which used longitudinal data collected within 1 year, each study that examined the impact of malaria on the decisions, productivity, expenditures and harvest values of agricultural households used cross-sectional data. Almost all of these studies, 25 of the 28, used less than 2 years of cross-sectional data.

The objective of our research framework is to provide a means of enriching our understanding of the cumulative impact of suppressing malaria over time on the income of agricultural households with regards to all of these topics. Given this objective, long-term longitudinal data are needed. For example, with data collected from the same agricultural household over time, a researcher could examine how the suppression of malaria transmission affected its decisions, productivity, and income.

In Table 3 we display the results of our analysis of the number of studies that have examined each research topic and the type of data that each study used. None of the studies we examined used longitudinal data. Furthermore, only two studies examined the impact of malaria on the decisions of agricultural households, and both of these studies were published before 1980.

Most of the studies that we identified that examined the productivity of individuals in agricultural households addressed the number of work days lost due to malaria cases directly or from caregiving. No studies examined the impact of malaria on gender inequality in agricultural households.

For topics related directly to the income of an agricultural household, most studies examined household expenditures on interventions to prevent or treat malaria infections. Only one studied the impact of malaria on harvest values.

In Figure 2, we illustrate the number of years of cross-sectional data that were used in each of the 27 studies that examined one or more of the topics in our research framework. Each line in Figure 2 represents the number of years of data that were used in each study. For example, the line in the lower left section of the figure indicates that Annecke’s findings that the suppression of malaria led to farming of areas that were previously considered too malarious were based on his observations from 1932 through 1935 ( Annecke, 1950). The number on the vertical axis represents the number we assigned to each of the 28 studies. Most of the studies, 23 of 27, used cross-sectional data collected over a period of less than 1 year. There were two studies that used cross-sectional data from a period of more than 2 years ( Annecke, 1950; Audibert, 1986; Girardin et al., 2004) and two studies used longitudinal data ( Nur, 1993; Sauerborn et al., 1996b). Table 4 provides additional information about each study included in Figure 2.

Our research framework for examining the long-term impact of malaria on the incomes of agricultural households uses longitudinal data to examine the impact on household decisions, productivity, expenditures and harvest values. By using longitudinal data, researchers will be able to examine how the decisions, productivity, expenditures and harvest values of households change over time and their relationships with changes in the intensity of malaria transmission. Our framework focuses on the impact of malaria on four decisions within an agricultural household: which crops to plant, how much land to cultivate, how much to invest in agricultural inputs, and how much to invest in interventions to reduce the risk of malaria infection. The available research evidence suggests that these decisions will depend on the household’s resources and expectations regarding the productivity of its members during the next growing season. Second, our research framework includes three means by which malaria can have a long-term effect on a household through its impact on the productivity of individuals: work days lost due to adult malaria cases, work days lost due to caregiving, and the impact on gender inequality. Finally, the framework defines the impact of malaria on the income of an agricultural household as being dependent on its expenditures related to malaria morbidity and mortality, and the value of the harvest. The harvest value will depend on the decisions made and productivity of individuals in the household. Existing systems for collecting longitudinal data from households in malarious communities should begin to collect the data necessary to address the topics described in this research framework.

In many malarious regions, a significant number of households are engaged in agricultural activities. We focus on agricultural households because of the unique relationship between these households and malaria as compared to non-agricultural households. The malaria burden creates a unique challenge for these households given the labor-intensive nature of agriculture. Labor is particularly critical for these households during certain periods of the year, namely the planting and cultivation seasons. Given that rural agricultural households are often poorer than non-agricultural rural households, the additional burden that malaria imposes on agricultural households could be a significant challenge to their efforts to escape poverty.

Historically, improvements in agricultural productivity have played an important role in enabling developing countries to increase their economic growth and escape poverty ( Awokuse & Xie, 2014; Datta, 2013; Gallup & Sachs, 2001; Mccarthy et al., 2000; Thirtle et al., 2003; Valdés & Foster, 2010). Four times the level of poverty reduction can be achieved by increasing growth in the agricultural sector than by increasing growth in other sectors of an economy (as outlined by the Gates Foundation).

One of the challenges in developing the research framework presented in this study is that malaria elimination has never been achieved and sustained in an agricultural region in sub-Saharan Africa. As a consequence, the identification of the research topics in this study has been informed by studies conducted in communities in which elimination was never achieved. Therefore, there may be other channels through which suppressing malaria may have a significant long-term impact on the income of an agricultural household that were not identified in this study. If malaria elimination were to be achieved in southern Africa over the next decade, evidence generated from that region may result in modifications to this framework as applied to equatorial regions of Africa where there are higher levels of malaria transmission.

We recommend that researchers examine how the suppression of malaria transmission over the next two decades may affect the duration of symptoms from a case of malaria. The suppression of malaria may lead to a reduction in acquired immunity levels among individuals in agricultural households, which may affect the probability of a malaria inoculation leading to a symptomatic case of malaria as well as the duration of the symptoms. Researchers should also examine how suppressing malaria transmission affects the seasonality of the impact of malaria on agricultural households and how the coping mechanisms used by households change over time.

We also recommend an analysis of the factors that contribute to heterogeneities across communities and over time in the long-term impact of suppressing malaria on the incomes of agricultural households. Given that the cost of illness methodology has been used more than any other methodology to examine the economic impact of malaria on agricultural households, we recommend a critique of this methodology to determine if it should continue to be used. Future research should also develop models that can utilize the longitudinal data that would be collected with this framework in order to develop prospective estimates of the impact of suppressing malaria on the welfare of agricultural households