Essays On The Effects of Risk and Ambiguity Attitudes On Production Choices Of Smallholder Fish Farmers In Southern Ghana

This thesis contains four empirical chapters which together contribute to behavioural economics in the area of fish production in a developing country context. The key thread connecting all the empirical studies is the behavioural characteristic of farmers (risk and ambiguity attitudes) elicited through incentivised field experiments and general survey questions.

The first empirical chapter seeks to answer the questions: What is the risk attitude of a typical smallholder fish farmer in a developing country? Do risk attitudes of fish farmers remain stable across different elicitation methods and contexts of validation? Risk attitude measures are known to be sensitive to the method of elicitation and context (Bauermeister and Mushoff, 2016). The purpose of this chapter is three-fold.

1. It elicits and compares the risk attitudes of within-subject sample of smallholder fish farmers in southern Ghana using three of the frontier methods used to elicit risk attitudes in the literature. The risk attitudes elicited from these methods are employed in the subsequent chapters of this thesis to investigate how risk preferences affect production efficiency and technology adoption.

2. It investigates how the risk attitude measures correlate with each other, and how they vary with farmer characteristics.

3. It assesses whether the risk attitude measures can predict farmer responses to questions on hypothetical economic choices.

The results show that a typical smallholder fish farmer is risk preferring in the gains-only lottery experiment, risk averse in the gains-and-losses lottery experiment but is risk neutral from the self-reported risk attitude scale. However, the risk attitude measures from the two lottery experiments are positively correlated, consistent with the assumption that the two experiments capture similar traits of the same farmer. This confirms that risk attitude measures are influenced by the method of elicitation and the context being examined. Some personal characteristics of the farmers influence their risk attitudes. Finally, while risk preferences from the lottery experiments failed to explain hypothetical economic choices, the stated risk preferences were significantly correlated with some hypothetical economic choices, perhaps due to hypothetical bias. These results indicate that care should be taken to tailor the elicitation of risk attitudes to contexts and domains farmers are familiar with.

The second empirical analysis attempts to answer the question: to what extent does a fish farmer's risk attitude affect his/her level of economic efficiency? This is predicated on the assumption that the types, levels and frequency of application of inputs could be influenced by the risk attitudes of farmers. Data on the units of inputs, outputs and prices are collated from the farmers in an earlier survey, and their risk attitudes obtained from the previous chapter are then juxtaposed on their production data. The economic efficiency of the farmers is assessed with both the Stochastic Frontier Analysis (SFA) and the Corrected Ordinary Least Squares (COLS) techniques. While the former assumes that all deviations from the cost frontier are due to farmer-specific factors (including risk attitudes) and stochastic factors, the latter, a deterministic procedure, attributes all deviations from the frontier to farmer-specific factors. The evidence from this chapter suggests that over 80% of the total deviation from the cost frontier results from stochastic factors beyond the control of the farmers. It is also found that risk attitudes play no significant role in the economic efficiency of fish production in the study area. Based on the findings, it is concluded that stochastic factors, such as government policies, may have a greater impact on economic efficiency rather than risk attitudes of farmers.

The third empirical study assesses how risk attitudes of fish farmers affect the speed of technology adoption; adoption decisions are modelled with duration models. This study focuses on the adoption of Floating Cages, Extruded Feed and Akosombo Strain of Tilapia (AST) technologies in the fish farming sector in southern Ghana. Contrary to most existing literature on speed of adoption of technologies (e.g. Liu, 2013), the results from this chapter show that risk averse farmers have a higher proclivity to adopt the AST, Extruded Feed and Floating Cage technologies at a point in time. This novel outcome is due to the nature of the technologies in question, as perceived by the farmers. Liu's (2013) study, for instance, focuses on the adoption of cotton seeds modified genetically with Bacillus thuringiensis (Bt) bacteria, which enables cotton plants to produce phytotoxins to kill pests. The subjective risks posed by these phytotoxins to the farmers themselves may be an additional source of uncertainty and a likely reason for the delayed adoption by risk averse farmers. However, in this chapter, even though the AST is also genetically modified, it produces no toxins and yet it is more disease-resistant than the local breeds, therefore it may be perceived by the farmers as risk-reducing and hence it may not be surprising that risk averse farmers adopt this technology earlier.

In the final empirical study, attention is on how ambiguity attitudes affect the farming decisions of smallholder fish farmers, using the speed of adopting the AST technology as an example of such decisions. The speed of technology adoption is analysed with the hazard/survival model. Additionally, this chapter introduces and interacts the number of previous adopters in the same village with ambiguity attitude as a better test of the effect of ambiguity aversion on farmers' decisions. Where a farmer cannot predict with certainty the yield to be obtained from the new technology, an ambiguity averse farmer is expected to adopt the technology late. Ambiguity attitudes are elicited with Ellsberg's (1961) two-colour urn experiment. The results from this chapter show that the average fish farmer is ambiguity averse. However, risk aversion, but not ambiguity aversion, has a significant effect on the speed of adopting the AST technology in the study area, confirming the robustness of the finding in the previous chapter. I also find that the speed of adopting this technology increases with the number of prior adopters in the same village. The lack of any significant impact of ambiguity attitudes in determining the speed of adopting this technology suggests that there are other important determinants of adopting this technology, rather than lack of information about it, that affect other technology adoption decisions.

Overall, this thesis demonstrates and presents the elicitation of risk and ambiguity preferences outside the usual laboratory setting by engaging fish farmers in a field experiment involving real cash incentives, as well as field surveys. The experiments and methods employed are at the frontier of research in the field of development economics. The results of the analysis presented in this thesis indicate that that risk preferences are sensitive to the method of elicitation, as well as the context or domain in which it is elicited. While contrary to findings from other studies, risk averse farmers are more prone to adopt improved fish farming technologies earlier than farmers who are not risk averse. This conclusion is plausible because the technologies may be perceived as risk-reducing by the farmers. This outcome remains robust when ambiguity aversion is introduced into the analysis of the technology adoption decision. Therefore, research on farmer production choices should take their risk attitudes into account, and such risk attitude measures should be elicited in a manner that is compatible with the context of operation of the farmers.