The effect of ambiguous risk, and coordination on farmers' adaptation to climate change — A framed field experiment

Abstract

The risk of losses of income and productive means due to adverse weather can differ significantly among farmers sharing a productive landscape, and is of course hard to estimate, or even “guesstimate” empirically. Moreover, the costs associated with investments in reduced vulnerability to climatic events are likely to exhibit economies of scope. We explore the implications of these characteristics on farmer's decisions to adapt to climate change using a framed field experiment applied to coffee farmers in Costa Rica. As expected, we find high levels of risk aversion, but even using that as a baseline, we further find that farmers behave even more cautiously when the setting is characterized by unknown or ambiguous risk (i.e. poor or non-reliable risk information). Secondly, we find that farmers, to a large extent, coordinated their decisions to secure a lower adaptation cost, and that communication among farmers strongly facilitated coordination.

Introduction

There is an extensive literature on the effects of climate change on agriculture (see e.g., Adams, 1989, Mendelsohn et al., 1994, Schlenker et al., 2005). Technological innovations, changing patterns of land use and changes in ecosystem dynamics could either lessen or exacerbate the effect of climate change on agriculture (see Antle, 1995, Reilly, 1995 for early reviews). From a farmer's perspective, climate change, whether it comes as slow changes in temperature or precipitation, changes in seasonal weather patterns or more frequent extreme events (drought or flood spells), can be seen as a technology shock affecting possibly both the deterministic and the stochastic components of the production function (see for example Kelly et al., 2005). Accordingly, farmers can adapt³ tactically (e.g. with changes in input use and timing of planting and harvesting) and/or strategically (e.g. changing the selection of crop varieties, increased diversification of crops and/or crop insurance) to deal with both effects of climate change (Bradshaw et al., 2004).

Global climate analysis signals Central America as one of the most affected tropical regions in the world, in terms of higher temperatures (Aguilar et al., 2005), a clear “drying signal” (Rauscher et al., 2008) and increased variability (Giorgi, 2006). Most important, global studies predict more intense rain and more widespread drought spells (Aguilar et al., 2005, Magrin et al., 2007, Sheffield and Wood, 2008). In a recent report for Central America (Curry et al., 2009), a predicted 0.6 °C increase in sea surface temperature is associated with between 0 and 1 additional tropical cyclones per year and 10–26% damage increase.

This article starts by constructing a standard baseline of farmer's adaptation behavior and risk aversion under known risk. We then move to work on our two research questions. First, we test whether farmers are ambiguity averse, and if this explains adaptation behavior. Second, we explore if, and to what extent, farmers are able to coordinate their adaptation efforts in pursuit of economies of scope in adaptation costs.

We use a framed field experiment (Harrison and List 2004), conducted with small-scale coffee farmers in Costa Rica, in which farmers were asked to act as if the decisions reflected their actual behavior. The experiment also involved monetary payoffs. The baseline for the experiment was a standard risk experiment, such as that of Holt and Laury (2002), to which a realistic frame, and treatments to explore ambiguity and cooperation were introduced. Previous risk experiments with farmers in developing countries include, for example, Binswanger, 1980, Binswanger and Sillers, 1983, Wik et al., 2004, Cárdenas and Carpenter, 2010. In our case, the values were chosen to give the farmers a familiar decision on whether to adapt to climate change, associated in this case to increased risk of extreme events and income losses.

There are several reasons for using a framed field experiment instead of using actual production data (see e.g., Antle, 1987, Antle, 1989, Pope and Just, 1991, Chavas and Holt, 1996). First, with actual production data, it is difficult to disentangle adaptation due to changes in risk and risk perception from other reasons, such as changes in soil fertility or new market opportunities. Second, it is not clear whether farmers actually are aware of changes in climate over time, such as global warming, as opposed to the usual climatic variability. Third, climate change might bring about production conditions, particularly for extreme events, that had no historical parallel in that coffee-producing region.

In addition to risk aversion, some authors have argued that ambiguity aversion is a key factor hindering the adoption of new technologies. In economics, the interest in unmeasurable uncertainty⁴ or ambiguity was spurred by the Ellsberg paradox (Ellsberg 1961). A number of experimental studies have shown that people are ambiguity averse; see, e.g., Fox and Tversky, 1995, Moore and Eckel, 2006, Slovic and Tversky, 1974.⁵ By ambiguity aversion, we mean that there is a preference for known over unknown risks.⁶ In the case of technology adoption, the status quo is perceived to have a known level of uncertainty, given the agent's experience with the old technology. On the other hand, the benefits of the new technology in good or bad scenarios are ambiguous, leading agents to reject it in favor of the old. In simple terms, the status quo is perceived as a safe, known bet. A recent paper that uses this setting is Engle-Warnick et al. (2007), who conducted a field experiment with coffee farmers in Peru. More recently, Cárdenas and Carpenter (2010) conducted a multicity experimental study exploring risk and ambiguity of urban populations. This study was an unframed field experiment that included the capital of Costa Rica and hence provides an interesting reference point to our results.

In the context of climate change and technology adoption, both the status quo (no adaptation) and the new state (adaptation) can be characterized by both risk and ambiguity. Climate change is a complex phenomenon, and the estimates of future increases in temperature or the likelihood of extreme events, for example, are very uncertain. The risks associated with not adapting to climate change could therefore be described as unknown or unmeasurable (IPCC 2007). If farmers are ambiguity averse, they will more likely adapt to climate change when the risk of a disaster is unknown to them, compared to a similar situation with known risk. In our experiment, we explored the relevance of reliable risk information for adaptation decisions.⁷

There is one other aspect that we investigated: the capacity and willingness of farmers to coordinate in pursuit of lower adaptation costs. The cost of technology adoption is potentially a function of the behavior of others, either due to learning from others (Bandeira and Rasul, 2006, Besley and Case, 1993) or due to economies of scope. This opens the door for government intervention, as Dybvig and Spatt (1983) also suggested, aimed at insuring early adopters against the possibility that others do not follow, for example.

In our experiment, we designed a situation where the cost of technology adoption is lower if everybody (the farmers, in our study) in the group adopts. However, farmers face different risks and have different risk preferences. This means that the decision can be viewed as a coordination game, where there could be multiple equilibria; see, for example, Ochs, 1995, Cooper et al., 1990.

Depending on a number of factors, including the physical and social distance between farmers and the quality of the institutions, farmers are more or less able to communicate with each other in pursuit of reduced costs, as described above. It is important, then, to differentiate between situations where coordination is possible with and without communication. Evidence from other experiments points consistently to the fact that communication leads to increased cooperation in public good settings even with heterogeneous subjects (Cardenas et al., 2004, Hackett et al., 1994, Ledyard, 1995, Sally, 1995). Moreover, studies also show that the link is not unequivocal because players might react negatively, if they identify noncooperating behavior in the course of group discussions.

Some explanations for the effect of communication on group decisions include persuasion, verbal promises, creation of a group identity that favors cooperation, and improved understanding of the game; see, for example, Buchan et al., 2006, Bochet et al., 2006, Ostrom et al., 1994, Bochet and Putterman, 2008. In our experiment, we investigated the extent to which farmers reduce their vulnerability to extreme events when there are economies of scope in the adaptation cost. This is done with and without communication between the farmers. We also conducted treatments with and without communication, when there were no strategic reasons for communication, in order to isolate any learning effect.

The rest of the article is organized as follows. Section 1 provides background information on our sample and the study area where the experiment was conducted. Section 2 introduces the experiment design and procedure. Section 3 presents the results, and Section 4 concludes the paper.

We conducted our experiment with coffee producers of the Tarrazu region of Costa Rica. All coffee producers are organized in a cooperative, a common type of organization in Costa Rica, which provided our sampling frame. Still, individual farmers are completely free to make decisions for their land. This region is well known for its premium quality coffee, which results from the mix of high altitude, cold weather, and lots of sun. According to a census of coffee producers (ICAFE-INEC 2007), there are 672 coffee farmers in Tarrazu. Coffee plantations make 76% of the all the farms in the region. Average coffee farm size is 10.3 ha, but 64% of the farms are smaller than 5 ha. Almost all of the farmers own their land and, in 2006, only 10% had outstanding loans on their land. This gives a picture of a prosperous region that has an equal distribution of income — but at the same time farmers are highly vulnerable to changes in the profits from their land. Because the farms are small, profits generally are just enough to cover the household's day-to-day expenses. The possibility of finding work outside the farm is limited, given that 84% of the farmers have only basic or no education.

Changes in climate relate to reduced coffee production via the plants' sensitivity to extreme temperatures (warm and cold) and humidity (both excessive moisture and drought). Wind and extreme rain can also bring severe losses during the flowering period and if physical damage to the plant results (Tucker et al., 2010).

In early 2008, tropical storm Alma hit the region with full force. The occurrence of such extreme weather events in this region is rare because it faces the Pacific Ocean. Based on historical records from 1949, only five extreme weather events have come near the Pacific coast of Costa Rica; Alma came nearest to the country and furthest south (IMN 2008). Only on two occasions have extreme climatic events originating in the Pacific Ocean seriously affected the Central American region, one in 2005 and Alma in 2008. The Tarrazu region was one of the most heavily affected, with approximately 12% of all coffee plants destroyed by landslides and overflowing rivers.

In total, 211 farmers participated in our experiment. Table 1 provides descriptive statistics of our sample in the Tarrazu region, based on the 2007 coffee census, which is highly representative of the coffee farmers in this cooperative.