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

(page 26)

and products (fertilizers, improved seed varieties).¹² In this study I rely completely on localized empirical data on agricultural practices. Although time depth is limited to two years, extremely detailed information for individual households--and individual fields held by each household--allows a more complete estimate of the variability experienced by agricultural producers than has usually been the case. In this analysis I focus on field dispersion as a risk reduction behavior

FIELD DISPERSION AND RISK REDUCTION

Field dispersion is one of the risk buffering behaviors which falls under the more general category of diversification. It is a common feature of non-industrial agricultural systems world-wide. Because it is usually considered to be an inefficient system of holding land, it has been the source of much scholarly attention and development effort aimed at consolidation. However, under certain ecological conditions, field dispersion is an important risk reduction behavior.

The Meaning of Field Dispersion: Definitions and Examples

I define field dispersion as the division of a farm unit into two or more non-contiguous parcels. This spatial organization of agriculture has also been termed "scattering" (Farmer 1960) and, more frequently, "fragmentation," (Bonner 1983; Heston and Kumar 1983; Igbozurike 1970; Johnson 1970; King and Burton 1982; Smith 1959). Given inconsistencies in use of the term fragmentation, ¹³ I use it only in the restricted sense of field size (i.e, referring to the condition of diminutive parcels). Non-contiguity of fields will be denoted either by dispersion or scattering.

Bentley (1987), and King and Burton (1982) review measures which have been proposed for field fragmentation, highlighting the several meanings it imbeds. Parameters included in these measures are size of the total holding, the number of plots, size of plots, size distribution of plots, spatial distribution of plots, and the characteristic shapes of plots. For example, both Simmons (1964) and Januszewski (1968) present indices based on total number of plots, and their size. Distance between plots is not integrated into the measures. Igbozurike (1974) does include distance and average size of plots, but his methodology for calculating distance has been criticized (Bentley 1987; King and Burton 1982). Dovring (1965) incorporates only distance, developing a measure based on the distance traveled by a farmer to each plot, and returning to home between visits to different plots. Schmook's (1976) index is

(page 27)

based on the ratio of the area circumscribed by all farm plots divided by the total area of all plots.

Numerous suggestions for the origins of field dispersion have been advanced. Most authors recognize more than one probable cause. Some argue that scattering is the remnant of open field systems of the past (Binns 1950; Chisolm 1962; Clout 1971). Others recognize that natural or human-made physiographic and topographic barriers may produce dispersion (Binns 1950; Igbozurike 1970; King and Burton 1982). In some cases, scattering is explained with reference to patterns of land distribution which stressed equitability of holdings. Given heterogeneous qualities of soil, drainage, slope, etc., scattering fields attempted to insure that each individual or family would receive some "good" and some "bad" fields (Binns 1950). Smith (1959) considers this to be part of the cause of the extreme degree of scattering in northwest Spain, and Heston and Kumar (1983) make a similar argument for Pathan tribes of Pakistan.

By far, the most common explanations offered base their arguments on some combination of inheritance patterns and population growth (de Vries 1974; Igbozurike 1970; Johnson 1970; King 1977; King and Burton 1982; Smith 1959). According to this view, partible inheritance laws and population growth cause reduction in the land base available per capita, and thus the size of individual parcels. De Vries (1974) describes this process of morcellement in Holland, arguing that it led to increased intensification, reduced productivity of labor, and greater exposure to the risk of famine for the poorest of peasant farmers. Another view leaves aside inheritance customs, and focuses instead only on population growth and land shortages. This argument holds that under conditions of scarce land, the farmer who wishes to expand holdings may have little choice but to buy or let land wherever parcels can be found (Ilbery 1984; Johnson 1970; King and Burton 1982). Downing's work (1973, 1977) presents important counterpoints to these arguments. He demonstrates that among Zapotecan peasants, population growth cannot account for field fragmentation.

Field dispersion, as a world-wide feature of agriculture, has received a great deal of attention from economists and others concerned with development issues. Although the potential benefits of this pattern of landholding are sometimes acknowledged, on the whole, the associated costs and inefficiencies are thought to outweigh any possible advantage. Three categories of adverse effects are most commonly cited. The first is that scattered holdings are thought to be incompatible with the modernization of agricultural technologies and economies of scale (Clout 1971; Smith 1959). Related are concerns about the difficulty of making infrastructural improvements, such as irrigation works (Bonner 1983). So viewed, field dispersion is an obstacle to progress.

The second category draws attention to the potential losses of yield thought to result from sharing borders with less meticulous neighbors. Bonner (1983) suggests that the overall quality of weed and pest control falls to the standard of the worst-tended field in the vicinity. Several authors claim that the neighbor effect produces greater dispute and litigation (King and Burton 1982; Smith 1959). In addition, the need to maintain borders as unplowed strips or "balks" is presumed to result in a loss of arable land (Bonner 1983). Downing (1973) estimates that due to fragmentation in Oaxaca approximately 2% of arable land is devoted to establishing boundaries between fields. (Given average field size in the study community of Diaz Ordaz, this would amount to 350,000 square meters of land among 7000 fields).

The third category concerns loss of time, principally arising from the increased movement and transport between fields (Smith 1959; Bonner 1983). This is the inefficiency most commonly cited in criticisms of field dispersion. Related to time loss are suggestions that scattering makes supervision (of labor, crop growth, and bird and animal intrusions) more difficult (Bonner 1983), and that the additional borders and corners create the necessity for

(page 28)

extra hand work (King and Burton 1982). Chisolm (1962) reviews studies which indicate that as distance between home and fields increases, intensity of inputs (labor and fertilizers) are reduced significantly and in turn, gross and net productivity per unit of land decreases dramatically.

Two qualifications should be made to such judgments. First, the degree to which time is "lost" must be assessed against some measure of its opportunity costs. This can be quite difficult. In economics, opportunity costs are often measured in units of income, but in an economy with few off-farm employment opportunities to establish relative prices, this may be inappropriate currency. On the other hand, time may be used to assess a number of vital--if not wage-gaining--pursuits; but incorporating measures of these activities has been notoriously elusive. Opportunity costs may vary dramatically over time. Time spent in movement between fields may represent a true loss if labor shortages are acute during some phase of the agricultural cycle (e.g., harvesting), but may be a negligible "cost" during other periods.¹⁴

The second qualifier concerns how mode of transport may alter the cost of field dispersion. Some authors note that mechanized means of transport will reduce the disadvantages of dispersion (Johnson 1970). But this may not be true under all circumstances. Mechanical means of transport will likely require the infrastructure (i.e., roads) in which to operate, and thus potentially limit choice of route and access (and increase the waste of arable land). This may actually add time and cost to movement between fields.

There has been little empirical verification of these suggested costs of scattering (King and Burton 1982), nor has there been much empirical attention devoted to demonstrating that its benefits really occur. Almost all of the authors cited above recognize that under some circumstances, there can be significant benefits to field dispersion. Specifically, field dispersion provides advantage when agroecological conditions are locally heterogeneous. The scattering of fields does this by distributing a household's exposure to risk across the agricultural landscape (Binns 1950; Bonner 1983; Ilbery 1984; King 1977; King and Burton 1982; Smith 1959).

Anthropologists have done much to further the argument for field dispersion as a risk buffering mechanism. The benefits of field dispersion have been stressed especially in alpine environments (Friedl 1974; Netting 1972, 1981; Rhoades and Thompson 1975; Weinberg 1972). In mountainous areas, steep gradients, and the micro-climatic variations they create, permit diverse productive activities, reduce conflicts in the scheduling of labor, and most importantly, provide security against disaster.

Netting (1972, 1981) described the vertical economy of the Swiss village of Torbel. The areas used by the community include altitudinally differentiated hay meadows, gardens, orchards, forests, and summer pasture. Each household held land in each zone, and within each zone had access to several non-adjacent parcels. Netting stressed the ecological rationale for this practice. At the micro-level, within the same zone differences in slope, aspect, soil quality, and moisture moderated and spread the impact of climatic conditions on yield. Holding more than one parcel within each zone reduced the probability of disaster. On a larger scale, the mix of production zones (1) enhanced the diversity of economic activities

(page 29)

(agriculture and herding, different crops), further insuring against complete loss, and (2) spread the scheduling of labor.

Galt (1979) explores the ecological advantage of field scattering on the Italian island of Pantelleria. In this Mediterranean setting, rainfall is concentrated in winter, but exhibits extreme year-to-year variability. The highly dissected topography of the island creates numerous ecozones. Differences in topography and micro-climate produce variations in crop maturation rate, resistance to suboptimal rainfall, yield, and sweetness of grapes (the main cash crop). Households maintain fields in an average of four different zones of the island. The maximum distance from home to field is about 5 kilometers. Galt details how the dispersed landholding pattern saved one household from complete ruin in a year of low rainfall. While there was no harvest in one zone, in another zone where soils maintained moisture, yield was sufficient to provide an adequate income for the year. In all years, regardless of climatic conditions, the sequential ripening of grapes in different zones allows a typical household to harvest its own crop with only limited additional labor. In very wet or particularly productive years, timely harvesting becomes an urgent matter. Again, scattered holdings facilitate labor exchanges between cooperating households, since ecozone and micro-climatic differences spread the range of maturation dates.

Forbes (1976, 1982, 1989) describes a similar pattern on the Greek peninsula of Methana. Here, variability in rainfall is the prime environmental concern. The impact of insufficient rainfall is mediated by differences in altitude and in soils, especially their water retention capacities. An average holding is typically divided among about 18 different plots. Forbes argues that this diversity promotes stability in year-to-year total yield. He tests this by comparing the coefficient of variation on single plots over several years to the coefficient of variation for the aggregate set of plots over the same period (for the wheat crop). The variance for the aggregate is significantly lower. For the particular household examined, the variability of the total wheat harvest was about half of the variability of the average plot. Forbes' study does not incorporate estimates of reduction in net yields due to the time and energetic costs of movement between dispersed plots, however.

It is unlikely that we will ever understand in detail the origins of field dispersion; it is equally unlikely that there is a single cause responsible for all cases. A more interesting question is that of persistence. Field dispersion is documented in agricultural systems around the world. Most investigators, on the basis of limited empirical evidence, have stressed the negative aspects of it. Almost all agree, however, that under certain ecological circumstances, there are benefits: specifically, field dispersion is a response to locally heterogeneous agroecological factors. These include climate conditions (temperature, rainfall, wind, hail), soil qualities, and disease and pest invasions. Spreading fields over the landscape diminishes the proportional loss due to any of these factors. Although net production may be lowered (due to movement and transport among other costs), the probability of complete disaster is reduced. Although this argument is logically compelling, it must be tested with direct empirical data. This is required in order to understand why this landholding system--so inefficient on first appearance--persists.

Field Dispersion as a Problem in Risk Reduction

The economic historian Donald McCloskey (1972, 1975, 1976, 1986) has employed a safety first model to explain field scattering in the common field agriculture of Medieval England. As this work guides the present study, I deal with it in some detail. In the common field system, each peasant's holdings were divided among three great fields. Following community regulation, these were devoted to wheat, barley, or left to fallow in any given year.

(page 30)

Within each of the great fields, the typical peasant held up to a dozen scattered small plots. This system was widespread over much of Europe for centuries, but by the mid-nineteenth century, acts of enclosure effectively had transformed European and English agriculture into a system based on consolidated and private holdings.

Historical opinion holds that the common field system was simply inefficient, and that consolidation righted a wasteful practice. Among the presumed inefficiencies ("costs") of holding land in scattered plots were several of those noted above: (1) time lost to movement and transport between plots; (2) land lost to unplowed strips and balks between plots; (3) neighboring effects; (4) the potential damage from overgrazing when the animals were let in following harvest or during fallow of the great field; and (5) the time and effort required for community control to coordinate cropping and grazing usage (McCloskey 1975:78-87; Dahlman 1980).

Given these presumed costs, the fact that common field agriculture persisted as long as it did is a puzzle to the historians. Explanations offered for its origin and persistence are numerous. Again, several were noted above: (1) that labor was available in such large quantities that the inefficiency did not matter; (2) that differences between plots made the scheduling of agricultural tasks more even throughout the growing season, and thus avoided labor bottlenecks; (3) that scattering made it easier for peasants to engage in common plowing arrangements, with the timing of plowing through the season equitably spread among all participants and their fields; (4) that scattering was the result of a sociopolitical insistence on egalitarianism, ensuring that each would have a mix of better and poorer quality fields; (5) that the arrangement arose piecemeal with the clearing of new lands and allocation of fields among families; and (6) that scattering was due to the never-ending fragmentation which results from partible inheritance. Although each of the explanations is effectively countered in detail by McCloskey (1975:88-113), he challenges them collectively with the argument that if scattering had been undesirable, the extant market in land would have informally affected consolidation. That consolidation did not occur suggests that farmers must have perceived some benefit to the dispersion of their fields.

McCloskey (1972, 1975, 1976, 1986) sets forth an argument that scattering was a response to risk, the desire to avoid some disastrous level of income shortfall. It served as insurance to reduce the year-to-year variability of income, an important factor for families living near the subsistence margin who had no alternative means of buffering. (In turn, new employment opportunities in later centuries provided strategies to reduce income variance that made the costs of scattering unacceptable.) The safety first model used by McCloskey is a direct analog of the Z-score model reviewed above. It relates average income (m ), the variability of income (s , measured in standard deviations), and disaster level (D, equivalent to Stephens and Charnov's R), to the probability of a falling below the critical threshold (Figure 2.2). Probability of disaster is reduced as (m - D)/m increases. McCloskey compares the frequency of disaster for scattered and consolidated fields, and finds that scattering reduced the frequency of disaster from once every 9 years to once every 13 years. Stated another way, scattering doubled the probability of surviving 20 years and tripled the probability of surviving 30 years without disaster.

The formula used by McCloskey to calculate the effects of field scattering solves for the coefficient of variation in income after yields from all fields are combined (s ), using the coefficient of variation of income on an individual field prior to pooling (s), the number of individual plots (N), and the average correlation in yields between plots (R):

1 + (N - 1) R ^1/2

s = s[-------------------]

N

(page 32)

Several points make the logic of this equation obvious (see Figure 2.2). In the extreme case that R = 1 (correlation between plots is perfectly positively correlated), then s = s, and no reduction in variance can be achieved by pooling. Conversely if R = -1 (correlation between the plots is perfectly inverse), then s will equal 0, and all variance will be eliminated by pooling yields from only two fields.¹⁵ For positive values of 0 < R < 1, reduction in variance rises rapidly at first for small values of N, eventually reaching an asymptote. When R = 0, reduction in variance is continuous, though the rate decreases as N grows (see also Winterhalder 1990; Goland 1991).

Using historical records to identify variable parameters (average income, disaster level, correlation between plots, etc.), McCloskey calculates the degree of scattering predicted by the model. The model also includes the reduction in net yield due to the "cost" of scattering, based on historical records on changes in net yield following the enclosure movement of the eighteenth century. The optimal number of plots predicted by the safety first model (8.3) and actual number of plots held by a typical farmer (8) correspond closely.

McCloskey confirms that field dispersion in the Medieval Midlands of England is an effective risk buffering strategy for coping with year-to-year variability in income. Insurance in this case is provided by field-to-field production variability.¹⁶ Cultivation of multiple, scattered fields reduces the probability of falling below a critical threshold of income. The important parameters in this argument are the year-to-year variation in production from an individual field, the number of fields, and the average correlation among them. Together, these specify the variation in income after harvests from all the fields have been combined. McCloskey includes calculation of loss of net income due to field scattering (especially, the cost of neighbor effects) at 10%.¹⁷

The formula used by McCloskey is necessary because of the highly aggregated data available to him. He must assume a normal distribution and extract from partial historical records relevant quantitative information on mean and variance. Despite differences in the specific calculation, subject, and data quality, the model provides the same general result as does the Z-score model, which was developed and empirically tested within the context of

(page 33)

foraging species. Here, I apply the same risk minimization logic. In contrast to the sparse data available to McCloskey, the present study benefits from direct empirical data on production in individual fields, providing precise information on distributions, means, and variance. The number of plots and the production on each is precisely known. The investigation is thus empirically derived and the risk analyses are specific to individual household landholdings and fields. Calculation of costs due to scattering is more exact than that permitted McCloskey, because the number of visits to each field, number of workers, distance from home to plot, etc. have been measured. This allows me to quantitatively refine and extend McCloskey's analysis of field scattering as risk minimization for the case of contemporary Andean agricultural production.

SUMMARY

Environmental variations pose problems in the provisioning of a secure food base. Although environment has often been described in normative terms, this is insufficient for understanding processes of adaptation. Individuals and populations are not adapted to an abstract statistical average, but to the very real and important variations in their biotic, climatic, and social environments. For complete analysis of human biocultural adaptation, attention to environmental fluctuation is critical.

Environmental fluctuations can be described by parameters of magnitude, temporal scale, and spatial distribution. Although the precise dimensions of fluctuation can never be predicted with complete accuracy, they are a regular feature of all environments. Because problems are recurrent, environmental variation is fundamental in shaping human response. The specific characteristics of dimensions of fluctuation determine which of several coping strategies will be most effective in buffering the population from the risk of food shortage. Buffering strategies to cope with predictable and minor stresses are highly integrated into production decisions and activities. The ability to respond to stress without compromising future capacities to cope with change is a critical measure of adaptability.

The effectiveness of routine buffering mechanisms can be examined using formal models of risk reduction. When risk is defined as the probability of loss, it is assumed that all organisms will be risk averse. In these models, reduction of risk is shown to depend on decisions between outcomes with different distributions of mean and variance.

Diversification of economic activities is perhaps the most widespread strategy for avoiding risk. In agricultural economies, diversification is achieved in several ways; among the most common is dispersing agricultural fields across the landscape. This strategy is effective when the environment is heterogeneous with respect to climatic, edaphic, and pest and disease conditions.

Scattering fields is presumed to reduce net productivity, given additional costs of transport and movement between individual plots. Models and empirical studies in economics and ecology account for scattering by using a safety first principle: higher productivity is traded for greater security. The argument is that, as a first priority, production decisions are aimed at insuring that a minimum requirement will be met.

In this study I test the hypothesis that field scattering reduces the probability of failure for Andean agriculturalist households. I use a safety first (Z-score) model for this examination. To do this, I estimate both the risk reduction benefits of field scattering, as well as the costs of transport and movement between plots. I use information on the production of the individual fields of Andean peasant households for this evaluation.