Cultivating Diversity: Field Scattering as Agricultural Risk Management in Cuyo Cuyo, Department of Puno, Peru

By Carol Goland, 1993.

Chapter 2 - Theoretical Foundations of a Risk Perspective

Food Storage

(page 18)

Storage is a means of evening out the temporal availability of food (Binford 1980; Cashdan 1989). Storage appears rather infrequently among ethnographically documented hunter-gatherers. In contrast, it is an integral component of agricultural economies. This is a logical outcome of both the need to store seed from harvest to subsequent planting, and the temporary bounty created by this form of production. In both agropastoral and pure pastoral communities, livestock are frequently considered "storage on the hoof," the analog of the granary (Dyson-Hudson and Dyson-Hudson 1970). Dugdale and Payne (1988) use simulation techniques to demonstrate that storage is a relatively low cost, effective means of buffering agricultural producers from year-to-year variability in crop yields.

The ability to store may be limited by several factors. First, the capacity to amass a reserve depends on a surplus beyond what is needed until the next round of production. Storage is unlikely to be a routine practice in environments where periodic abundances are not predictable and frequent (Goland 1991). Second, appropriate technology for preservation must exist, and the environment must be favorable for long-term preservation (Keeley 1988; Testart 1982; Suttles 1968). Even under these conditions, the costs of processing, or losses in storage due to rotting and pests, may be unacceptably high (Bourne 1977; Dugdale and Payne 1987). Third, dependence on storage limits opportunities for mobility (Binford 1980; Rowley-Conwy and Zvelebil 1989). Storage logically co-occurs with greater sedentariness, although the use of scattered caches is possible for more mobile groups.

Mobility

Mobility is a way of buffering when the availability of resources varies spatially. Among hunter-gatherers, mobility is a frequent method for coping with resource scarcity (Lee and DeVore 1968), and among pastoralists constant movement in search of ephemeral pockets of good pasture is a definitive feature of the economy (Legge 1989; Little et al. 1988). For more settled agriculturalists, the possibilities for mobility are constrained by the need to remain near fields in order to tend them. But regardless of the particular subsistence base, there are several potential limitations to mobility. First and most obvious, there must be alternative areas nearby where resources above the needs of the resident population are available: if the factor(s) causing resource scarcity are geographically extensive, mobility may not be an effective countermeasure. Second, widespread environmental monitoring must be undertaken in order to acquire information about the resource availability in alternative areas. Third, land tenure rules and/or social ties to these other regions must be constructed and maintained to assure access to these areas (Smith and Boyd 1990). Scudder (1962) describes how Valley Gwembe Tonga--residents of an area subject to frequent drought and food shortage--maintain ties to distant kin in upland areas in order to have a refuge during crises.

Exchange

While mobility moves people to resources, exchange moves resources to people.⁴ Exchange alleviates problems of resource scarcity by distributing resources across social units.

(page 19)

Forms of exchange include sharing, trade, and markets. Ford (1972) describes how food is exchanged and distributed as an integral part of the ceremonial cycle of Southwestern Pueblos. Localized sharing or pooling of resources is a common feature of hunter-gatherer societies (Kaplan and Hill 1985 Smith and Boyd 1990). When their foraging successes are not highly correlated, relatively few cooperators who forage independently and pool resources can dramatically reduce their individual variance in food consumption (Winterhalder 1990). For this system to operate effectively, it may be critical to have some manner of limiting the scope of the network of social obligation. Simulation of resource pooling in an agricultural economy suggests that unrestricted sharing may make it impossible for individuals to survive the effects of widespread failure, even if they themselves have not been directly affected (Hegmon 1988, 1989).

Exchanges between parties may be an immediate trade of food for an item valued at equal worth (labor, materials goods, etc.), or exchange of non-food items may establish obligatory social relations which may be called upon during times of emergency. In this sense, exchange of valued but non-essential items during non-stress periods establishes and maintains a network for accessing foods during crises (Halstead and O'Shea 1982; O'Shea 1981; Wiessner 1977, 1982). When exchange partners inhabit distinct regions, resources may be spread across both social units and resource areas.

It is the case that for both exchange and mobility that as the spatial scale of environmental disruption widens, distance to unaffected areas increases (Harpending and Davis 1977). Thus, the cost of exchange and mobility increases at the same time that their effectiveness is reduced.

Diversification

Colson (1979) considers diversification to be the most common buffering strategy among subsistence producers, be they hunter-gatherers, agriculturalists, or pastoralists. Diversification is an effective device to counter both the spatial and temporal dimensions of environmental variability, by spreading impact over alternative food resources, labor opportunities, spaces, and time periods. Diversification of production systems is analogous to the diversity characteristic of mature ecosystems, and is presumed to provide the same advantages of stability and resilience.⁵

In the context of agricultural production, one way of achieving diversity, and thus presumably greater stability, is through intercropping: planting more than one crop in the same field simultaneously (Vandermeer 1990). Intercropping is viewed as a way of imitating the

(page 20)

diversity and assumed stability and resilience of mature stages of ecological succession (Rhoades and Bebbington 1990). Multiple cropping takes advantage of the distinct agronomic and climatic needs and the unique susceptibilities of different crops: if the production of one crop is diminished, another crop may still thrive (Forbes 1976, 1989; Lynman et al. 1986; Scudder 1962). Ecologically, intercropping may enhance stability due to several factors: fuller capture of solar energy, protection from wind and water erosion, and higher disease and insect resistance (Mountjoy and Gliessman 1988; Wilken 1987).

Agricultural producers also buffer risk by planting diverse crops and many varieties of each crop, in varying environmental settings (Norgaard 1989). Planting in several fields buffers against micro-climatic variation and highly localized infestations of disease and pests (Forbes 1976, 1989; Ford 1972; Scudder 1962). Forde (1931) describes field dispersion in Hopi agriculture as a method for spreading the risks associated with localized and unpredictable rainfall. Each clan owns land in several topographically distinct zones, and families of each clan cultivate several fields within each of these holdings. Planting an array of crops and varieties takes advantage of their varying requirements and tolerances to different climatic, disease, and edaphic conditions. Spreading planting dates of crops buffers against the possibility of abnormally early or late rains or frosts (Bradfield 1971).

The same logic holds for hunter-gatherer subsistence pursuits: if factors affecting the availability of important staple food resources are independent, then disruption to one food source will be mitigated by the continued supply of others (Gould 1975). In many extant foraging societies, wild resources are supplemented with cultivated foods. In some cases, the same group cultivates and forages (Hawkes et al. 1987). In others, garden foods are obtained through trade with agriculturalists (Hart 1978, 1979; Spielmann 1983) or gained as payment for labor or by cash from wage-earning pursuits (Bahuchet 1988; Bailey and Peacock 1988; Hitchcock 1982). Likewise, in many agricultural economies, subsistence production is supplemented with wild resources (Harris 1979) or with wage labor (Cancian 1989; Schumann 1985). In some years, the cash may be used to purchase non-essential items (or set aside as a precaution against future calamity), while in other years, it may be used to purchase foodstuffs in case of agricultural failure.

The effectiveness of each of the coping strategies listed above lies in the particular dimensions of variability which it addresses (spatial and temporal scales, predictability, etc.). The transmission of knowledge is particularly important when long periods separate environmental disruptions. Food storage is effective under more limited temporal scales, and it is probably not capable of buffering the impact of more than one or two years of scarcity. Mobility adjusts populations to available resources over regional spatial scales. Exchange evens the availability of resources over social units, whether these are separated by small or large distances. Diversification, perhaps the most fundamental response to variability, may be so pervasive precisely because it functions effectively along both temporal and spatial scales.

Variables other than environmental ones may also be significant in determining buffering behaviors. Cashdan (1985) describes how Basarwa (Bushmen) and Bantu-speaking agriculturalists in the same locale (and thus subject to the same environmental conditions) nonetheless rely on different coping mechanisms. The Bantu-speaking groups, by virtue of their more permanent residence, have larger agricultural fields, and so are able to produce sufficient surplus to put into storage. In contrast, Basarwa households shift residence more often. As a result, fields are smaller, and there is insufficient production for storage. The primary buffering mechanism they use is sharing.

Discussion of each of these strategies individually may suggest that they are more independent than is actually the case. Certain combinations naturally co-occur. For example, mobility as a strategy to cope with environmental variability implies a store of knowledge

(page 21)

about alternative resource areas. The viability of exchange may be conditioned by the existence of a stored surplus within a feasible distance of the population suffering the shortage. In contrast, other strategies do appear to be mutually exclusive: mobility and storage are an unlikely combination, for instance.

Many of these strategies are based on a foundation of social relations. Kin and social networks underlie the viability of mobility and exchange. Migrants rarely have the opportunity to venture into an uninhabited landscape, but move instead to areas where ties of mutual obligation insure their access. Likewise, in non-market economies, exchange relations are frequently built on similarly binding social ties. Even opportunities for wage labor may be based upon long-term social relations (e.g., patron-client relations).

In this study, I assess the importance and effectiveness of each of these strategies for coping with environmental variability in the Andes. Although each of these strategies is employed by Andean people to varying degrees, I argue that diversification is key, the defining feature of Andean production. Household production is diversified across economic spheres (e.g., combining wage labor with agricultural subsistence production), and within the subsistence sphere, it is diversified in space, through time, and across crops (Chapters 3 and 4). The analysis relies on formal models derived from evolutionary ecology and micro-economics.

ENVIRONMENTAL VARIABILITY AND OPTIMIZATION MODELS

Evolutionary ecology theory provides models that consider adaptive responses to temporal fluctuations and spatial heterogeneity (see Wiens 1976 and Thomas et al. 1979 for summaries). This method usually draws on optimization models that examine which trade-off between costs and benefits will give maximum advantage to the individual. In biology, advantage is usually expressed in units of fitness, while in economics, the concept of utility is invoked.⁶ The models have a general form: they specify a strategy or decision set (a range of options available to the subject) that is the object of analysis (e.g., choice of prey type); they select a currency used to assess costs and benefits (frequently energy, serving as a proxy for fitness); and they make a series of explicit assumptions about constraints that link currency and decisions (e.g., in the conventional diet choice model, searching and handling prey may not occur simultaneously). The model is then solved for the optimum, that is, the choice which maximizes advantage, given the specified constraints (Maynard Smith 1978; Pyke et al. 1977). Anthropologists also have adopted these models, and beginning in the 1980's, investigation of hunter-gatherer foraging behaviors using optimization models became prominent within ecological anthropology (Hames and Vickers 1982; Hawkes et al. 1982; Smith 1981; Smith and Winterhalder 1992; Winterhalder 1981a).

Early ecological and economic optimization models generally focused on maximizing harvest. While these models have proven heuristically valuable, both within anthropology and elsewhere (reviews in Smith 1983; Winterhalder 1981b), more recently researchers in diverse subject areas--evolutionary ecology (Caraco 1981; Stephens and Charnov 1982), agricultural economics (Roumasset 1976), and economic history (McCloskey 1975, 1976)--have proposed similar models with risk minimization criteria. These models examine the mean and standard

(page 22)

deviation associated with each production decision, and select the choice with lowest probability of falling below some minimum fixed requirement.

Risk minimization models extend the earlier generation of optimization models in several ways. First, they replace an energy maximization criterion with risk avoidance, where risk is taken as probability of loss (of life, or some less catastrophic diminishment of fitness, income, prestige, etc.). Second, rather than incorporating deterministic environmental variables, some variables in the model are given probability distributions, with both a mean and variance. Empirical results demonstrate that varying environmental circumstances and condition of the subject may have important consequences for decisions about behaviors.

RISK REDUCTION MODELS

In evolutionary ecology, attention to risk reduction begins with the observation that foraging does not yield certain returns. Rather, short-term capture rates are probabilistic. Each foraging choice can be characterized by a mean reward, and variance around it. The models begin by positing a forager who must capture a minimum number of calories in order to survive until the next foraging opportunity (Figure 2.1). Imagine a forager who needs at least 100 calories (R = 100) in order to avoid starvation overnight, and who has two prey options: (a) one yields on average 35 calories, but one in ten times yields as many as 50 or as few as 20 (prey choice 1 in Figure 2.1); (b) the other choice on average yields 60 calories, but one in 10 times yields as many as 120 or as few as 0 (prey choice 2 in Figure 2.1). If the forager must capture 100 calories to survive, it has no viable option other than pursuing the more variant prey (prey choice 2). This choice at least provides a one in ten chance of survival; the other provides only certain death. On the other hand, if the forager must fulfill a minimum requirement of only 25 calories (R = 25), it would do better to pursue the less variant prey (choice 1). Though less rewarding on average, the less variant prey leads to a lower probability of failure. In the literature, these different behavioral options are generally referred to as risk preference and risk avoidance. However, it may be more correct to say that all individuals will be risk averse (where risk is probability of loss), but under varying circumstances may do so by preferring or avoiding variance.⁷

Some of the strongest empirical support for the importance of risk reduction has been provided by studies of birds and other species with high metabolic demands (for a comprehensive review of these studies, see Real and Caraco 1986; Stephens and Krebs 1986). In experiments by Caraco and colleagues (Caraco et al. 1980; Caraco 1981; Caraco and Lima 1985) birds were presented with two feeding options, both with equal means: one yielded constant rewards, while the other presented a variable reward. The variable choice contained two alternate rewards, each with a probability of 0.5. The bird's consumption was altered prior to the onset of the experiment. On some occasions the birds had been given enough food to meet their energy requirement (positive energy balance); on others they faced a potential shortfall (negative energy balance). The result of these experiments indicates that the choice of either certain or variant rewards depended on the bird's energy balance. Birds with positive energy balances showed preference for certain rewards, while birds with negative energy

(page 24)

balances demonstrated preference for variable rewards. This result, termed the "expected energy budget rule," is generalized as follows: avoid variance if your energy budget is positive; prefer variance if your energy budget is negative. Like the model presented above, probability of survival is determined by requirement (energy budget) and the probability distribution associated with each dietary choice. When budget is negative (or requirement is greater than average rewards) the probability of survival is increased by choosing larger variance.

Balanced energy budgets have also been simulated with the result that the birds exhibit indifference to variance. Indifference has also been found when variance is very small, or mean reward very large. Thus, preferences for variance appear to be governed by both energy balance and statistical properties of the reward. This can be critical: when a forager has a fixed requirement that must be met, then appropriate response to variance may very well determine its chance of survival.

The generalization provided by the expected energy budget rule has been formalized by the Z-score model (Stephens and Charnov 1982; Stephens 1990), which is concerned with maximizing the probability of survival by avoiding energetic shortfalls. The model solves for the optimal choice within a feasibility set that includes rewards where both mean and variance change.⁸ The Z-score model shows that a forager's preference for variance depends on its energy requirement (R), expected food reward (æ), and variance (å), where Z = (R - æ)/å. Here Z is the standard normal deviate.⁹ Minimizing this value is equivalent to maximizing the odds of obtaining more than a needed requirement. Increased variance always increases the absolute value of Z. If the numerator is negative (requirement less than expected reward), decreased variance decreases the value of Z (enhances the probability of survival). If the numerator is positive (the requirement is greater than expected reward), then increased variance decreases the value of Z, the presumed goal of the forager. More colloquially, the results generalize to a rule that suggests that if you can expect to get more than you need, opt for small variance, but if your requirement is greater than you can expect, choose large variance. Important results of this model demonstrate that (1) risk avoidance is not constant, that is, preference or avoidance of variance depends on mean (Stephens and Paton 1986), and (2) risk can be avoided by manipulating both mean and variance (in contrast to the expected energy budget rule). This means that under some circumstances, a forager may be willing to accept a lower average reward (with higher variance) in order to enhance probability of survival. Although most researchers have interpreted R as some starvation threshold, there is no logical reason why this must be so. R could represent the food requirement necessary for successful breeding or dominant status (Stephens 1990) or, in the social realm, that needed to maintain prestige (Hegmon 1989).

The treatment of risk in economics converges with the evolutionary ecology literature, although there are several notable differences. In economics, attention to risk dates back at least to Knight's (1921) study, in which risk and uncertainty were defined. Knight is credited with identifying risk as objective probability (i.e., measurable, known probabilities), and

(page 25)

uncertainty as subjective probability (i.e., absence of prior information).¹⁰ Some economists follow this definition. Others, objecting to it on the basis that all information must ultimately be considered subjective, adopt alternative definitions for risk and uncertainty (e.g., Ellis 1988). As a result, there is no universal agreement about how risk should be defined (Rothschild and Stiglitz 1970). Fleisher (1990:16) defines risk as "a situation in which resolution of uncertainty will affect the well-being of a firm or decision maker." Uncertainty corresponds to its colloquial usage: it describes situations where a decision maker cannot know the outcome of at least one of the alternatives, because of multiple possible outcomes.

The economic literature on risk is vast, ranging from applications in portfolio theory to agricultural economics. Equally diverse are proposals for measuring risk (expected value, variance, and probability of disaster), and an array of models for analyzing subjects' behavior under conditions of risk (see Roumasset 1979a for a review). One of these, the safety first model (of which there may be several variants [Roumasset 1976, citing Day et al. 1971]) most closely matches the models of risk avoidance developed in evolutionary ecology, since it centers on the idea of avoiding disaster. In safety first models, risk is the probability that returns will fall below some critical "disaster" level. Risk aversion, then, is the willingness to accept a lowered average return in order to avoid a worst-case shortfall (Roumasset 1979b).¹¹

In agricultural economics, empirical studies often rely on data derived from agricultural experiment stations, where conditions are superior to those of the average farm. As a result, inferred yield response may be highly overestimated. Use of historical records to derive objective probabilities also may be unreliable due to the high level of aggregation of the data (often national or provincial level statistics). Such data will likely underestimate variability, through time and across space. They may say little about what is actually experienced in the typical farmer's field (Binswanger 1979; Dillon and Anderson 1990:158). These studies have focussed on several areas of inquiry, such as the effect of credit on production decisions, land tenure arrangements (e.g., sharecropping) as risk reduction strategies, farm diversification, and perhaps most prominently, adoption of new technologies