Learned helplessness bias in the agricultural investment
decisions of smallholder farmers in eastern Uganda
By Floris Burgers
Contact: [email protected] Student number: 11131969
Supervisor: Prof. Arjan Verschoor (University of East Anglia) Second reader: Dr. Nicky Pouw (University of Amsterdam)
Research Master in International Development Studies University of Amsterdam
Acknowledgements
Because this thesis is part of my education, and my research skills will be assessed and graded on the basis of it, I decided to write it in singular to emphasise the full responsibility I take for all the decisions made throughout the research. However, before I decided this, I had already written extensive drafts of the thesis in plural and that says everything about the extent to which it has been teamwork rather than individual work which has led to the present document. In fact, this thesis would be non-existent without the support, engagement and patience of a large group of people, many of whom I did not know before the start of this project but I now consider friends. Because of these people, I always felt, and still feel, uncomfortable to speak about ‘my argument’, ‘my decision’ or even worse, ‘my research’, because clearly in most cases ‘our’ instead of ‘my’ would be much more appropriate. I therefore hope, that the words ‘I’, ‘me’ and ‘my’ will be interpreted, in this thesis, in a manner taking into account the fact that there would have been no ‘I’ without them:
First and foremost, my supervisor, professor Arjan Verschoor, who has made fantastic contributions to the quality of this thesis, was incredibly supportive to my personal development over the past two years and made this project practically feasible in the first place. Second, Joshua Balungira and Zam Namutosi, the heads of the Field Lab, who have been essential during the process of data collection and who made me feel at home in Uganda. Third, all research assistants of the Field Lab team, who were the kind of supporters a researcher can only wish for: Appollo Wamumbi, Charles Gidoi, Isaac Musedde, Isaac Namonyo, Jackline Naloko, Pamela Wamayi, Peter Budeyo and Kate Namisano. Last, dr. Michaela Hordijk, whose flexibility, enthusiasm and solution-oriented thinking gave me the opportunity to work with the mentioned people.
Summary
The farm income of smallholder farmers in developing countries, who often lack the resources or access to institutions to protect themselves against all kinds of income shocks, is highly sensitive to external forces such as weather and crop price fluctuations. Consequently, the relationship between the farm investments of farmers and the income they obtain through farming often seems rather imperfect, and is hard to assess for them, as many different factors co-determine the farm income. This could lead to a perceived lack of control over the farm among some farmers who would then be increasingly susceptible to learned helplessness bias when it comes to new investment opportunities. They may underestimate the extent to which their investment decisions have an impact on their farm income as a consequence of repeated experiences of a lack of control over the farm. Such a bias may lead to reduced agricultural investment in lucrative opportunities, and, in a worst case scenario, frustrate agriculture’s potential to contribute to the reduction of food insecurity and poverty, but has never been empirically examined as a potential explanatory factor for agricultural underinvestment.
In this context, this thesis investigates the psychological phenomenon of learned helplessness as a potential mechanism of counteracting agricultural investment among smallholder farmers in a rural area in eastern Uganda. Data was collected through a sequential explanatory mixed methods research design in which a quantitative research stage (field-experiment and survey) was succeeded by a qualitative research stage (semi-structured interviews). The outcomes of the quantitative analysis show that experiencing a lack of control over income losses can lead to reduced willingness among smallholder farmers to invest in an opportunity to reduce the risk of farm income shocks, presented to them in an experimental session. The qualitative analysis shows, however, that these results may lack external validity to the extent that farmers, in real life, attribute uncontrollable factors to the will of God – thereby creating an illusion of control – and highly value the social consequences of their investment decisions. Especially the latter, makes them persistent in their farm investments and makes them acknowledge, rather pro-actively, the importance of the value of perseverance in the context of farm investments. After an integration of both research stages, the thesis concludes that learned helplessness may indeed lead to a motivational deficit to invest in the farm, but this potential is not universal, nor inevitable, as it depends on the attributional style of farmers, their (illusionary) perceptions of control over the farm income and social considerations they take into account in the decision making process proceeding agricultural investment.
Content
Acknowledgements ... 3
Summary ... 4
Figures ... 6
Tables ... 6
Boxes ... 6
1.
Introduction ... 7
1.1. Problem Statement ... 71.2. Learned Helplessness: What is it? ... 8
1.3. Learned Helplessness: What is it not? ... 9
1.4. The current research: Location and methods ... 10
1.5. Structure of the thesis ... 10
2.
Theoretical Framework ... 12
2.1. Early learned helplessness studies: From ‘contiguity’ towards ‘contingency’ ... 12
2.2. Learned helplessness behaviour: Motivation, learning and emotion ... 14
2.3. The generality of learned helplessness in humans: Attributional style... 15
2.4. Potential imperfect learning: Factual vs. perceived uncontrollability ... 17
2.5. Smallholders in developing countries: An uncontrollable environment... 19
2.6. Learned helplessness bias and agricultural investment: Conceptual framework ... 20
3.
Methodological Design ... 24
3.1. Introducing critical realism... 24
3.2. Applying critical realism ... 26
3.3. The research questions ... 27
3.4. The research location ... 29
3.5. Methodological design: Sequential explanatory mixed methods ... 30
4.
The context: Social life of the Gisu ... 43
4.1. On Gisu manhood: Managing social expectations and pressure ... 44
4.2. On gender: Allocation of resources and division of labour ... 46
4.3. On supernatural beliefs: Ancestral spirits, religion and witchcraft ... 47
5.
Quantitative analysis: Learned helplessness in an experiment ... 49
5.1. The pre-treatment effect ... 49
5.2. The effect of attributional style on agricultural investment ... 54
5.3. The effect of income shocks on agricultural investment ... 55
5.4. Overview of the quantitative results ... 56
6.
Qualitative analysis: Learned helplessness in real life ... 58
6.1. The causes of income shocks: “The income from my farm drops once in a while” ... 59
6.2. The importance of the weather: “If there is rain, there will be enough food” ... 61
6.3. Perceptions of control: “It is all in God’s hands”... 62
6.4. Persistent farmers: “They encouraged me not to give up” ... 64
7.
Discussion ... 68
7.1. Integration of the two research stages ... 68
7.2. Theoretical implications ... 70 7.3. Practical implications ... 71 7.4. Methodological reflections ... 73
8.
Conclusion ... 77
References ... 80
Appendix ... 85
Figures
Figure 2.1: The traditional learned helplessness experiment ... 13
Figure 2.2: Conceptual Framework ... 21
Figure 3.1: Map of the research area ... 29
Figure 3.2: Sequential explanatory mixed methods research design ... 31
Figure 5.1: Distribution plot of pre-treatment loss by treatment ... 50
Figure 5.2: Bar chart investment by treatment and pre-treatment loss category ... 51
Figure 6.1: The context of the relationship between farm investments and farm income ... 61
Tables
Table 3.1: Features of the three domains of reality ... 25Table 5.1: Pairwise treatment comparisons ... 50
Table 5.2: Regression analysis interaction effect treatment and pre-treatment loss ... 53
Table 5.3: Regression analysis attributional style composites ... 54
Table 5.4: Regression analysis shock composites ... 56
Table 6.1: The causes of income shocks ... 59
Boxes
Box 3.1: Research questions ... 281. Introduction
In this thesis, I investigate the psychological phenomenon of learned helplessness as a potential mechanism of counteracting agricultural investment among smallholder farmers in a developing country context. Learned helplessness refers to a psychological state in which someone believes, because of previous experiences, that their actions have no effect on the outcomes of a situation. This belief is the consequence of repeated experiences of having no control and leads to a motivational deficit to act, a disrupted ability to learn about the effects of actions and, in some cases, to anxiety and depression (Maier and Seligman 1976; Peterson et al. 1993; and Seligman 1975). The motivational deficit and disrupted learning capacity can last even if control over the situation has been regained, giving rise to cognitive bias. A person who has learned to be helpless could be reluctant to invest, even when the investment would be profitable, because they underestimate the extent to which their actions can make a difference. Smallholder farmers in rural areas of developing countries do not typically have the resources or access to institutions to protect their farm income against different kinds of income shocks (Banerjee and Duflo 2011; and Ray 1998). This makes the incomes of these farmers highly sensitive to various external factors such as the weather and crop price fluctuations (Collier and Gunning 1999; Dercon 2008; World Bank 2000). To a large extent, these external factors co-determine the income they acquire through farming, rather than the effort they put into their farms or the investment strategy they apply. Consequently, it is hard for individual farmers to accurately assess the relationship between their farm investments and their farm income, and, in the context of farm income fluctuations, this relationship may seem highly imperfect to them. Farmers may, therefore, perceive little control over their farm income and, conforming to the learned helplessness theory, may then become reluctant to invest in their farm, at some point, as they no longer expect themselves to be capable of making a difference.
1.1. Problem Statement
Underinvestment in lucrative agricultural investment opportunities among smallholder farmers could frustrate the potential for agriculture to reduce food insecurity and poverty in developing countries (Dethier and Effenberger 2012; World Bank 2007). Studies have shown, for instance, that investment in fertiliser could seriously increase the average yield of small farms of about an acre in size in certain regions of developing countries (Duflo et al. 2008; and Verschoor et al. 2016, p. 147). If profitable technologies, such as fertiliser, are only limitedly adopted by smallholder farmers a considerable amount of potential profit for these relatively poor farmers remains unrealised.
An improved understanding of the investment decisions of smallholder farmers, taking into account learned helplessness as a potential explanatory factor, may improve the effectiveness of agricultural development policies concerned with the realisation of this potential. These policies often rely on the (investment) decisions of their ‘beneficiaries’ in order to be successful and thus benefit from a more complete understanding of how small farmers make their decisions (Oppong 2014; World Bank 2015).
Yet, while the complexity of decision making and the importance of psychological, social and cultural aspects in the decision making process is increasingly acknowledged by policy makers and consultants in the field of international development (see World Bank 2015), learned helplessness has never been empirically examined in this light. This thesis thus adds a new perspective to the understanding of investment decisions of smallholder farmers, which may help explain both underinvestment and, to a certain extent, why lucrative (new) technologies, when appropriate implementation is not self-evident and requires serious exploration, may not be adopted (cf. Duflo et al. 2008).
Note that the aim of this thesis is to explain a ‘piece’ of the puzzle rather than the puzzle at large. There is ample literature available in which the importance of other determinants of farmers’ technology adoption is illustrated, or that farmers can behave perfectly sensible when not investing in certain technologies (see for instance: Adrian et al. 2005; Nyamwanza et al. 2014; Yang and Fang 2015). Learned helplessness may be only a piece of the agricultural investment puzzle and does not make alternative perspectives any less relevant or appropriate. However, as learned helplessness has never been examined in this context, while many smallholder farmers in developing countries lack control over their farm income, it is an intriguing new piece and potentially an important one too.
1.2. Learned Helplessness: What is it?
The term learned helplessness has been used to capture a phenomenon encountered for the first time in the 1960s by Martin Seligman and his colleagues at the University of Pennsylvania: ‘experiences of a lack of control in the past can lead to expectations of little or no control in the future, resulting in passive behaviour’ (Overmier and Seligman 1967; and Seligman and Maier 1967). Although learned helplessness is generally thought of as a negative phenomenon, possibly because of its link to depression (see for instance Seligman 1975), essentially it is not inherently negative. People learn to be helpless all the time and this saves us a lot of time and energy that would otherwise be wasted. A woman who tries to open a locked door will learn, after having failed to open the door a couple of times, that she has no control over that door. Consequently, she will stop trying to open the door at some point because she no longer expects she can; an expectation she developed through her previously experienced uncontrollability. Imagine what would have happened if the woman had not learned to be helpless in the situation with the door: she would end up pulling the door for the rest of the day expecting it to open every time she tries!
However, most scientists concerned with learned helplessness, including Seligman, focus on its potential to result in disproportional passivity, hence the negative connotation attached to the term (see for instance Fosco and Geer 1971; Thornton and Jacobs 1971; and Maier and Seligman 1976). The term disproportional passivity refers to passive behaviour in situations where activity actually leads to a better outcome. Learned helplessness occurs in this undesirable form if its basic mechanism is accompanied by one of the following two conditions: (I) expectations are based on a misidentification of the cause of
uncontrollability, or (II) the situation changes while expectations do not change accordingly. The following two examples illustrate this using the example of the woman and the door. Example condition
I: the woman attributes her lack of control over the door to the perceived fact that it is locked while, in
reality, the door is just jammed. Consequently, the woman gives up on trying to open the door even though pulling a little harder would actually lead to the desired outcome. Example condition II: the door was initially locked, but unbeknownst to the women it was unlocked by someone after she had already stopped trying to open the door. The woman will remain passive even though she could now easily open the door. The woman’s behaviour is disproportionally passive in both cases because at some point she could have easily opened the door but she failed to do so.
These examples illustrate that learned helplessness can lead to a cognitive bias under specific circumstances. In this thesis, I am concerned with such a learned helplessness bias in the context of agricultural investment decisions of small farmers in a developing country; i.e. the idea that farmers may not invest in a lucrative opportunity because previous experience has taught them their investments do not make any difference. Given this focus, learned helplessness is here defined as a psychological state in which a person underestimates the extent to which their efforts have an impact on the outcomes in a particular situation and, therefore, behaves disproportionally passive in the situation, as a consequence of perceived uncontrollability experienced earlier. This definition captures the three essential components of the learned helplessness theory which will be elaborated later in the thesis: experience, expectations and behaviour (Peterson et al. 1993).
1.3. Learned Helplessness: What is it not?
The learned helplessness theory, as it is applied in this thesis, should not be confused or inappropriately associated with attitudinal theories of poverty (Kane 1987). Attitudinal theories of poverty argue that the poor have collectively adapted to their situation through the creation of a culture of poverty, including a sense of helplessness when it comes to their situation, which is perpetuated through intergenerational transmission (Kane 1987; Lewis 1965). During some of the presentations I gave of preliminary versions of this thesis, I noticed that some of the people in the audience associated my application of the learned helplessness theory with the idea of a culture of poverty (c.f. Rabow et al. 1983). In their critiques, they typically referred to the literature on resilience to argue that the poor are rather active and creative when it comes to protecting their livelihoods against income shocks and restoring them in response to an income shock (c.f. Akter and Mallick 2013). These critiques, however, are based on a misunderstanding which I would like to clarify before I continue to elaborate further on this research.
The way in which the learned helplessness theory is applied in this thesis does not compete with the idea of resilience. Nor does it have much in common with the concept of the culture of poverty as described by Oscar Lewis back in 1965. The learned helplessness theory, as opposed to the notion of a culture of
poverty, by no means refers to a general mechanism which shapes the majority of the decisions people make. Instead, the way learned helplessness is understood here (as a cognitive bias) suggests that it exclusively affects people’s behaviour in rather specific situations. Although learned helplessness bias may be facilitated by a culture of poverty, as will be extensively discussed later in the thesis (see section 2.5), the two concepts are by no means referring to the same mechanism. Moreover, when looking at the overall behaviour of farmers, ‘resilience’ may be a very accurate term to capture their intentions and endeavours. Yet, that does not mean a resilient farmer is insusceptible to a learned helplessness bias when it comes to some of their investment decisions. Whereas resilience refers to a lifestyle, learned helplessness can be limited to a rather specific set of decisions or situations and typically emerges under specific circumstances. Hence, rather than replacing resilience as a new label to capture the lifestyle of smallholder farmers, the concept of learned helplessness is used here to explain just a fraction of the investment decisions farmers make, which means that both concepts are not necessarily mutually exclusive.
1.4. The current research: Location and methods
Data for this research was collected in Bwikhonge, a rural area in eastern Uganda located on the slopes of Mount Elgon, near the Kenyan border. The majority of the people living in this area are Bagisu (Gisu), a Bantu-speaking people whose primary means of subsistence is farming. The agricultural economy that shapes everyday life in Bwikhonge exemplifies the situation in most rural areas in Uganda, as well as rural Africa at large: irrigation is rare, reliance on sophisticated agricultural inputs such as fertiliser and improved seeds is not widespread, good agronomic advice is lacking, land pressure is extremely high and, most importantly, income shocks are a continuous threat to smallholder farmers.1
Data was collected by means of a mixed methods research design including a quantitative research stage using a field-experiment and a survey, followed by a qualitative research stage in the form of semi-structured interviews. A representative group of 198 farmers, both male and female, living in ten different villages in Bwikhonge were selected, using a cluster sampling strategy, to participate in the quantitative research stage. A subsample of these participants was then later selected to also participate in the qualitative research stage. Data was collected between September and December 2016.
1.5. Structure of the thesis
The thesis is structured as follows: Chapter 2 provides an extensive discussion of the learned helplessness theory. The chapter starts with a discussion of the fundamental learned helplessness experiments, which were conducted with dogs in the 1960s, before moving on to outlining how the theory was reformulated in the 1970s and discussing some of the aspects of the theory that have been
1 The information provided in this section was obtained from Uganda Bureau of Statistics (2014), the Bulambuli
District Local Government Five Year District Development Plan (2010/11-2014/2015), and Verschoor et al. (2016).
somewhat underemphasised. The chapter ends with an application of the theory in the context of the agricultural investment decisions of smallholder farmers, providing the conceptual scheme upon which the remainder of the thesis is based. Chapter 3 presents the methodological design of the thesis as it introduces the study’s ontological and epistemological foundation, its research questions and the research methods used. Chapter 4 sets the context of the research on the basis of Heald’s (1998) ethnography of the Gisu, the ethnic group to which most people living in the research area belong.
Chapter 5 presents the results of the quantitative analysis, while Chapter 6 does so for the qualitative
analysis. These chapters are then integrated and reflected upon in Chapter 7, the discussion chapter and final conclusions are presented in Chapter 8.
2. Theoretical Framework
When a group of psychologists at the University of Pennsylvania were running experiments about learning processes with dogs in the early 1960s, coincidently they encountered something interesting. After having been exposed to high-pitched tones followed by brief electric shocks for a couple of days, in a subsequent session, the dogs failed to jump over a small barrier to avoid the electric shocks. Opposed to what the psychologists expected, the dogs remained passive and did not attempt to avoid the shocks (see Overmier and Leaf 1965). Whereas others were merely annoyed by the irresponsive dogs, Martin Seligman, a graduate student at the time, saw something significant going on. He thought that the dogs, after they had been exposed to inescapable shocks, had learned to be helpless. This is why, according to Seligman, the dogs gave up attempting to escape the shocks and did not respond to the option to avoid the shocks in the second part of the experiment. This explanation opposed the dominant ‘behavioural theory’ of learning which assumed that all of a subject’s behaviour is determined by their history of rewards and punishment and not by cognitive processes such as perceptions, beliefs and expectations. The irresponsive dogs, and Seligman’s interpretation of them, mark the start of the development of the theory of learned helplessness, which has since been applied in all kinds of contexts to explain disproportionally passive behaviour observed in both humans and animals (see Peterson et al. 1993). The remainder of this chapter is concerned with the details of this theory and it continues by discussing early learned helplessness studies (section 2.1), learned helplessness behaviour (section 2.2), the generality of learned helplessness (section 2.3) and the implications of the concept of imperfect learning to the learned helplessness theory (section 2.4.). Eventually, the theory will be applied to the context of the agricultural investment decisions of smallholder farmers in a developing country context and, in doing so, the theory is used to form a new line of enquiry for understanding such investments. Section 2.5 discusses the uncontrollable environments that smallholder farmers typically have to deal with, which makes them susceptible to a learned helplessness bias. Section 2.6 sets out the conceptual framework underlying this research, summarising the main theoretical links involved in the application of the learned helplessness theory in this thesis.
2.1. Early learned helplessness studies: From ‘contiguity’ towards ‘contingency’
The first experiment in which learned helplessness was formally tested and observed was published by Maier and Seligman in 1967, a couple of years after Seligman’s initial revelation. To test Seligman’s alternative explanation of the irresponsive dogs, Seligman and Maier designed a new experiment in which 24 dogs were divided over three treatments and exposed to shocks in two consecutive sessions. In the first session of the experiment, dogs were either exposed to controllable shocks in a rubberised hammock (T1), uncontrollable shocks in a rubberised hammock (T2) or not exposed to shocks at all (T3). The dogs in T1 could end the shocks by pressing a bar placed in front of the hammock. The dogs in T2 could not end the shocks by themselves, but they were yoked to the dogs in T1 so that they receivedshocks of the same duration.2 In the second session of the experiment (taking place 24 hours after the
first session), all dogs were exposed to the same escape/avoidance task in a two-way shuttle-box. Again, they were exposed to shocks but these could now be avoided by all dogs if they jumped over a small barrier placed in the middle of their box. This two-stage triadic design has become the traditional design to study learned helplessness and has been repeated since with various subjects and tasks to further explore different facets of the phenomenon. As I will refer back to this design a couple of times in the remainder of the thesis, a schematic overview of the design is presented in figure 2.1.
Figure 2.1: The traditional learned helplessness experiment
The results of the experiment confirmed Seligman’s hypothesis. The dogs in T1 and T3 managed to avoid the majority of the shocks in the escape/avoidance task while the dogs in T2 failed to escape the majority of the shocks. According to the authors, the latter group had learned to be helpless when it comes to escaping the shocks in session one, and had, therefore, failed to escape shocks in session two (see also Seligman 1972). Whereas the behavioural perspective on learning hypothesised that behaviour follows exclusively from contiguity, i.e. the extent to which a certain reinforcement followed a response in the past, they could now argue that animals can also learn about contingency, i.e. the causal relationship between their response and reinforcement based on both the extent to which reinforcement followed response and the extent to which it occurred in the absence of it (see also Overmier and Seligman 1967).
However, some scholars in the ‘contiguity camp’ were not yet convinced and offered several possible alternative explanations for these results (see Maier 1970; or Peterson et al. 1993, 29 for an overview). Most of their critiques revolved around the idea that the dogs in T2 learned something in session one that prevented them from performing the escape response in session 2. For example, the dogs could have
2 The shocks of a particular dog in T2 ended when its yoked counterpart in T1 pressed the bar. The yoking
procedure is an essential aspect of the traditional learned helplessness experiment, as the duration of shocks may otherwise confound with the effect of the extent to which participants have control over the shocks.
learned to sit still (as opposed to jumping over a barrier) as a way to cope with the shocks, assuming that shocks are less painful when received inactively (cf. Bracewell and Black 1974). However, this critique was proved incorrect a couple of years later by an experiment of Maier (1970). Maier’s design resembled the traditional learned helplessness experiment of Seligman and Maier (1967) (see figure 2.1), only this time the dogs in T1 could escape the electric shocks if they did not press the bar in front of them; i.e. they were taught to sit still – a response that is incompatible with the avoidance task to be performed in session 2. Maier’s results again confirmed the learned helplessness hypothesis: even though both dogs in T1 and T2 had now learned to sit still, dogs in T2 still escaped significantly fewer shocks in session two than dogs in T1 and T3. The contingency camp was winning the scientific battle against the contiguity camp.
In a next step, the learned helplessness effect was tested in humans in a range of experiments conducted in the 1970s. Again, the traditional learned helplessness experimental design was used (see figure 2.1). The earliest studies on human subjects exposed participants to uncontrollable aversive events (electric shocks or loud noise, for example) to produce learned helplessness behaviour (Fosco and Geer 1971; Hiroto 1974; and Thornton and Jacobs 1971). Later studies used insoluble problems (Benson and Kennely 1976; Hiroto and Seligman 1975; and Klein et al. 1976). The outcomes of these experiments were unambiguous: human participants in T2-like treatment groups (i.e. exposed to uncontrollability in session one) performed significantly worse in a subsequent task in which control was regained, compared to participants in T1 and T3-like treatment groups. It was now clear that both dogs and humans, as well as several other animals, learn about the extent to which they can control their environment (i.e. contingency) which in turn influences their behaviour in the future. This future behaviour, then, is a consequence of developed expectations of uncontrollability (Maier and Seligman 1976). This interpretation is the basis of the learned helplessness theory, which is further elaborated on in the following sections.
2.2. Learned helplessness behaviour: Motivation, learning and emotion
On the basis of these early learned helplessness experiments, Seligman (1972) started to distinguish between three potential behavioural consequences of repeated experiences of uncontrollability: a motivational deficit, a disrupted ability to learn and emotional disruption. First and foremost, the expected lack of control developed in the wake of experienced uncontrollability leads to a motivational
deficit to act. The expectation that control is impossible in a situation makes one less incentivised to
actively respond to that situation. Second, Seligman argues that experienced uncontrollability interferes with one’s capacity to learn about contingency. If, after having been exposed to uncontrollable shocks, a person has regained control and managed to avoid a shock for the first time, they often have difficulty understanding that the reason they avoided the shock was due to their own actions (see Miller and
Seligman 1975).3 This, in turn, makes them less incentivised to repeat those actions, avoid the shock
again and notice the contingency. Third, uncontrollability leads to emotional changes. When exposed to an uncontrollable aversive event, subjects develop more stress than when they are faced with a controllable equivalent (Seligman 1972). Seligman has later argued that uncontrollability leads to fear, which may disappear if control is regained, or can lead to depression if uncontrollability persists (see Seligman 1975).
Learned helplessness has been associated with depression because the behaviours of helpless and depressive individuals have a number of characteristics in common: a motivational deficit to initiate response, a negative cognitive capacity to learn about contingency, reduced aggression, loss of appetite and reduced susceptibility to overestimation of personal control (Seligman 1972, 1975; and Alloy and Abramson 1979). Moreover, a feeling of helplessness has been identified as one of the essential characteristics of depression (Seligman 1972). Conforming to this line of enquiry, several studies show that depressive individuals perform similarly in session two of the traditional learned helplessness experiment without any pre-treatment experience, as do non-depressive individuals in a T2-like treatment group who first learned to be helpless in a pre-treatment session (Klein et al. 1976; Miller and Seligman 1975). If learned helplessness and depression are similar phenomena, a cure may also follow from the same treatment. This is the basic idea behind the learned helplessness model of depression as introduced by Seligman (1975) and further developed in the past couple of decades (see LoLordo 2001).
2.3. The generality of learned helplessness in humans: Attributional style
As the learned helplessness effect had been observed in various animals including humans, and its basic mechanism had been conceptualised and even applied in the context of depression, the next aspect of the learned helplessness theory that deserved attention was the generality of the learned helplessness effect (c.f. Altenor et al. 1977; Cohen et al. 1976; and Miller and Norman 1979). In a revolutionary experiment, Hiroto and Seligman (1975) discovered evidence of a learned helplessness effect when human subjects were first exposed to a series of insoluble problems (session 1) and thereafter to escapable noise (session 2), as well as the other way around (i.e. firstly inescapable noise and, thereafter, soluble problems) (see also Gatchel et al. 1975; Gatchel and Proctor 1976; and Miller and Seligman 1975). Their results show that learned helplessness can be generalised across motivations and tasks, which led them to conclude that “learned helplessness may involve a trait-like system of expectancies that responding is futile” (p. 327). As the effect of uncontrollability can be generalised across different types of activities, learned helplessness may have greater consequences in real life than one would initially expect, based on the earliest learned helplessness studies (Seligman 1975, 32).
3 Note that making a conceptual distinction between the motivational deficit and the disrupted learning capacity in
Paradoxically though, Hiroto and Seligman’s results also revealed a major caveat of the learned helplessness theory when applied to humans. Their conclusion that learned helplessness can be generalised across different situations led to the realisation that the learned helplessness theory itself failed to address the boundaries of this generalisation. In other words, if a person learns to be helpless through a certain experience of uncontrollability, in which future situations will their behaviour be interfered and for how long? These questions could not be answered by the learned helplessness theory, as defined at the time, which led to a necessary reformulation of the theory by Abramson et al. in 1978. The reformulated learned helplessness theory of Abramson et al. solves two major problems of the original learned helplessness theory: (I) its failure to explain generality and (II) its failure to distinguish between universal helplessness (outcomes are uncontrollable for many people) and personal helplessness (outcomes are uncontrollable only for some people). The latter is particularly important with regards to the effect of learned helplessness on emotion. When learned helplessness is universal, someone’s self-esteem typically remains unaffected as the uncontrollability experienced is obviously not due to an individual’s lack of skill (indeed everybody lacks control). Yet, the opposite holds when uncontrollability is only experienced by one or a few people. As hardly anyone else is lacking control in such a situation, those experiencing uncontrollability, can easily blame themselves for their situation and, in doing so, damage their self-esteem.
Abramson et al. argue that a person who experiences uncontrollability automatically attributes their perceived lack of control to a specific cause, which in turn determines in which situations that person will behave helplessly, for how long and whether their self-esteem is harmed or not. A cause has three dimensions: internal vs. external (internality), global vs. specific (globality) and stable vs. unstable (stability). The last two dimensions influence the generality of learned helplessness; the first dimension its universality. A stable cause is hard to take away and will therefore be present in the future for a long time. If someone attributes uncontrollability to a stable cause, learned helplessness will be relatively chronic because expected uncontrollability is likely to persists as the cause upon which that person’s expectation is based is unlikely to disappear. A global cause affects relatively many situations. If uncontrollability is attributed to a global cause, learned helplessness will be generalised across relatively many situations as one expects to be lacking control in all situations in which control is determined by that cause. An internal cause is due to something about the self as opposed to something about other people or circumstances (in which case the cause is external). If uncontrollability is attributed to an internal cause, a decline in self-esteem is more likely to occur as people then believe it is something about themselves that is causing them to experience helplessness.
According to the reformulated learned helplessness theory, people who tend to attribute uncontrollability to internal, stable and global causes are more susceptible to learned helplessness than people with opposite attributional tendencies. One of the determinants of the causal attributions people make is their
attributional style (i.e. the way they tend to explain positive and negative events in life). Abramson et al. argue that people with a more internal and general attributional style, i.e. people who tend to attribute life events to internal and stable/global causes, are more likely to assign uncontrollability to causes that facilitate a learned helplessness effect. In line with this hypothesis, studies have shown that people who have an internal/general attributional style for negative life events and an external/specific attributional style for positive events (i.e. a pessimistic attributional style), are more susceptible to depression (see Joiner and Wagner 1995 for an overview).4
In sum, the reformulated learned helplessness theory adds a new component to the original learned helplessness theory and thus captures the following cognitive and behavioural steps: (I) experience of uncontrollability; (II) causal attribution to explain uncontrollability; (III) expectations towards future situations; (IV) motivational deficit to act, disrupted ability to learn about contingency and emotional changes. And, it argues that causal attribution (step II) depends on people’s attributional style: i.e. the extent to which they tend to attribute positive and negative life events to internal, global and stable causes. The theory has been most thoroughly applied in the context of depression, as it is argued that people with a more internal/general attributional style are more susceptible to learned helplessness, which in turn has been associated with clinical depression (cf. Ball et al. 2008). However, it has also been used to explain all kinds of other social problems related to disproportionally passive behaviour such as burn-out and academic underachievement (see Peterson et al. 1993 chapter 7 for an overview of various applications of the theory and Evans et al 2005; Fridley 2016; Houston 2016 for more recent examples).
2.4. Potential imperfect learning: Factual vs. perceived uncontrollability
Before applying the reformulated learned helplessness theory (henceforth just ‘learned helplessness theory’) to understand agricultural investment decisions of smallholder farmers, this section first sheds some light on an aspect of the theory which has been somewhat underemphasised: potential imperfect
learning. The cornerstone of the learned helplessness theory is the idea that people can learn about the
extent to which they control a certain situation; i.e. they do not only assess the extent to which a certain outcome follows their action (contiguity) but also the extent to which a certain outcomes occurs in the absence of their action, and they assess their level of control (contingency) by comparing both figures. A situation is uncontrollable if a certain outcome occurs equally often in the presence as in the absence of action because then action does not make any difference.
However, the fact that one can learn about contingency does not necessarily mean their learning is always accurate. The learned helplessness theory suggests that uncontrollability can lead to
4 An internal/general attributional style of negative events and an external/specific attributional style for positive
events are often referred to in terms of a pessimistic attributional style. The opposite is then referred to as an optimistic attributional style.
disproportional passivity, but in fact it is perceived uncontrollability that may do so. Imperfect learning may lead to a mismatch between factual uncontrollability and perceived uncontrollability, which means that people may perceive more (situation 1) or less (situation 2) control than they factually have. When applying the learned helplessness theory it is thus important to look at both factual uncontrollability and perceived uncontrollability. I distinguish between three factors that may affect people’s perceptions of control either resulting in situation 1 or situation 2 type biases:
Subsample assessment. It is not easy to remember exactly how often a reward followed a certain action,
or how often that reward occurred in the absence of action for every situation in life. Previous studies have shown that people therefore rely on small subsamples of their total number of past experiences if they have to make a decision (Hertwig et al. 2004). However, the typical subsample of past experiences, used for decisions making, is often not perfectly randomly obtained because people tend to exclude, unconsciously, rare cases. This may result in imperfect learning in situations where a certain action only incidentally generates a high reward. Imagine a certain action which leads to a loss in most cases, but incidentally to a reward that is high enough to leave you with a small profit on average as long as you continue to perform the action. This situation is factually controllable because action leads to a more desirable outcome than no action. However, due to a biased subsample assessment, one may perceive this situation as uncontrollable because the rare but profitable reward is more likely to be underrepresented in the typical subsample. Consequently, one may stop performing the action at some point even though that would be an undesirable passive act. Reward prevalence can thus lead to imperfect learning causing a situation 1 type bias (c.f. Teodorescu and Erev 2014).
(Healthy) illusion of control. Several studies have shown that people tend to overestimate the extent to
which they can control a situation, especially if factual controllability is low, resulting in a situation 2 type bias (Alloy and Abramson 1979; and Thompson et al. 2007; see Presson and Benassi 1996 for an overview). This tendency has been labelled human’s ‘illusion of control’ and is often considered a healthy phenomenon because it is believed to benefit psychological well-being. An illusion of control is seen as a way to cope with potential aversive emotions, such as anxiety and fear, which may occur in the wake of perceived randomness and/or unmanageability (Kay et al. 2010b). For example, Alloy and Abramson (1979) show that depressive individuals make more realistic estimations of control as opposed to non-depressive individuals who tend to overestimate their personal control.
Kay et al. (2010a) argue that a religious belief in a controlling God is a convenient way to create an illusion of control as events are no longer seen as entirely random, but “controlled or willed, even if not by the self” (p. 38). They summarise a large body of experimental research which shows that people are more sensitised to a greater confidence in God after having been exposed to uncontrollability (see Kay et al. 2008 for an example). They conclude that these studies “suggest that the prevalence of beliefs in God may reflect, at least in part, a psychological process set in place to help relieve the anxious
uncertainty associated with the threat lowered personal control poses to beliefs in an orderly world” (Kay et al. 2010a, 40)
Social influence. Helplessness can be learned through experience, but it can also be taught by word of
mouth; for instance, through intergenerational transmission. A person who grows up in an environment in which many people were taught to be helpless in a certain situation may become biased regarding that situation as others influenced their expectations. Consequently, they may not sufficiently explore that situation to be able to accurately estimate their factual control. If a culture of poverty, as described by Lewis (1965), exists, and small farmers would be part of it, they may stop exploring new agricultural technologies rather easily because others have already told them that their attempts are futile. Accurate assessments of factual controllability can thus be disrupted by the social environment a person lives in, potentially giving rise to either a situation 1 type bias (in the case of the culture of poverty example), or a situation 2 type bias (if a social environment highly believes in the possibility of control).
2.5. Smallholders in developing countries: An uncontrollable environment
Smallholder farmers in developing countries often live in highly uncertain environments in which their agricultural income is prone to all kinds of shocks (Fafchamps 1999). The relatively low income of most small farmers makes it difficult for them to protect their farms against various external factors. For example, whereas the typical farmer in a more affluent country makes use of irrigation systems to protect their farm against long periods of droughts, such facilities are typically unavailable to small farmers in developing countries, which makes their farm income highly dependent on the weather. According to the literature, some of the other factors causing income fluctuations are: loss of harvest due to pests; theft or violence; crop price fluctuations; eviction or resettlement; death of family members; and illness (Dercon et al. 2005, 563; Sinha and Lipton 1999, 8; World Bank 2000, 136).
Formal insurance structures, designed to cover unexpected income losses, are usually unavailable to most people in rural areas of developing countries. Successful insurance requires a certain degree of verifiability so that the insurance company is able to audit damage claims and avoid moral hazard on the side of the claimants. The possibility of verification, however, is often limited in these areas, which makes it hard for financial institutions to employ formal insurance in such locations (Alderman and Paxson 1994; Banerjee and Duflo 2011; and Ray 1998). Informal risk-reducing strategies have taken over the function of formal insurance structures (Alderman and Paxson 1994), but these informal strategies are often subject to numerous constrains (see Fafchamps 1999). Consequently, income shocks are a persistent threat to the stability of people’s livelihoods. For example, in Ethiopia, experiencing a drought or illness at least once in the previous five years lowers a household’s per capita consumption by 20% (in the case of drought) and 9% (in the case of illness) on average when compared to a household without such an experience (Dercon et al. 2005)
Thus, the farm incomes of many smallholder farmers in developing countries are not only the outcomes of their personal efforts and investment decisions, but are, to a large extent, determined by external factors which are hard to control by individual farmers. As these factors are important co-determinants of the farm income, it becomes hard for individual farmers to assess the effect of specific investment decisions. For instance, if a farmer decided to use a new type of seed but failed to get a good harvest after trying the seed for the first time, it may not be entirely clear why things did not work out as expected. Especially in the absence of good advice about agronomic practices, the farmer might wonder whether failure was due to the investment decision of trying the new seeds or one of the many other factors that affect the yield, such as the weather, soil fertility, or pests. Moreover, a sizeable group of farmers may have frequently had an experience whereby they tried certain investments and got a low yield in one season, while, on the other hand, similar investments gave them a good yield in another season. The relationship between what a farmer does with their farm and what they get out of it may then seem highly imperfect. In this context, learned helplessness may take hold among smallholder farmers and lead to a cognitive bias when it comes to their future investment decisions as they have been ‘taught’ that their investments are not crucial to their farm income.
2.6. Learned helplessness bias and agricultural investment: Conceptual framework
This section is concerned with the potential for this learned helplessness bias to take hold. As mentioned in the introduction of the thesis, learned helplessness only leads to a cognitive bias if its basic mechanism is accompanied by either of the following two conditions: (I) uncontrollability is attributed to an inappropriate cause and therefore inappropriately generalised across situations, or (II) unbeknown to the helpless individual, control is regained, meaning that their helpless expectations are maintained inappropriately. (Note that the second condition is often the consequence of the first. When uncontrollability is attributed to a stable cause while, in fact, the cause of uncontrollability is rather unstable, control will most likely be regained earlier than expected, possibly without the helpless individual knowing.) With this in mind, this section aims to answer the question of to what extent these conditions are likely to be present in the context of the potential lack of control over the farm income experienced by smallholder farmers. The argument to be made here is summarised in a conceptual framework presented in figure 2.2. This framework functions as a summary of the discussion below and is therefore frequently referred to in the remainder of this section. In the framework, the basic steps of the learned helplessness theory as employed in the thesis are captured by the arrows A-B-C-D, while context-specific conditions are captured by arrows I to IV.Repeated experience of income shocks to the farm income, obscuring the effect of specific investment decisions, may lead to a perceived lack of control over the farm income (arrow A). However, as mentioned in section 2.4, the extent to which factual controllability and perceived controllability coincide partly depends on subsample assessment, the illusion of control and social influence (arrow I).
The latter, social influence, is of particular importance for the sake of this thesis because several scholars have argued that a life in poverty gives rise to a collective sense of helplessness and a ‘culture of poverty’ (Dixon and Frolova 2011; Kane 1987; Rabow 1983). As a life in poverty is, according to some, characterised by limited control over future possibilities, poor communities would be rather susceptible to feel collectively helpless and be pessimistic about their capability to improve their situation (Dixon and Frolova 2011; Kane 1987). Such a collective sense of helplessness observed in poor communities – typically considered an aspect of a culture of poverty – may in turn cause bias in people’s individual perceptions of control over their farm income towards the extreme of a sense of uncontrollability. A person who is surrounded by people who do not believe in their personal capability to effectively change their situation may be discouraged to explore the extent to which they can actually affect their farm income and may thus end up more easily in a state in which they believe they have no control over that farm income (which makes arrow A in figure 2.2 more realistic).
Note that this is not to say that a culture of poverty, and the sense of helplessness that is part of it, is the same concept as the learned helplessness bias this thesis is concerned with. The concept of a culture of poverty does not capture ‘individual learning by experience’ and ‘causal attribution’ – as the concept of learned helplessness does – but, instead, focuses entirely on an individual context overriding sense of helplessness existing at the community level which fashions the majority of expectations and motivations of the people living in that community and is perpetuated by word of mouth over time. Though a sense of collective helplessness may lead to learned helplessness at the individual level as it may affect people’s learning process and causal attribution, this thesis argues that learned helplessness itself is a more sophisticated phenomenon, because it is contextual, dependent on specific conditions and may occur irrespective of whether or not a culture of poverty exists in society. Hence, the idea of a culture of poverty is important to this research to the extent it may stimulate learned helplessness bias, but unimportant to the extent this bias may as well occur in the absence of it.
Once a situation is perceived as uncontrollable people naturally attribute their lack of control to a certain cause, which in turn fashions their expectations of control towards new situations in the future. In figure 2.2 it is suggested that farmers’ may attribute perceived uncontrollability in the context of their farm income to an inappropriate cause, which in turn, is more likely to be a pessimistic cause (i.e. an internal/general cause). People in rural areas of developing countries have often relatively low levels of education and/or reduced access to knowledge systems.5 Consequently, they may be less capable of
identifying the appropriate cause of uncontrollability in the context of their farm income, or be less certain about their thoughts, which makes them more susceptible to an interpretation bias (arrow II). Attributional style has been identified as an important explanatory factor of causal interpretation and is increasingly important as straightforward interpretations are unavailable. Yet, although the roots of attributional style are not entirely clear, it has been argued that people with a history of negative life events are more prone to adopt an internal/general attributional style for negative life events over time (cf. Peterson et al. 1981, 258). As life in rural areas of developing countries is often associated with a high incidence of negative life events such as disease, death, theft and violence, people living in these areas are more likely to have a pessimistic attributional style for negative life events (see arrow III). A more general attributional style leads to more general expectations of uncontrollability regarding future situations. Hence the more general a farmer’s attributional style, the more likely they are to expect little control over their farm income in future seasons (see arrow C). An expected lack of control may lead to a reluctance to invest in the farm which in turn may lead to a learned helplessness bias if lucrative investment opportunities later become available (arrows I and D). Indeed, these opportunities may not
5 For example in Bulambuli district, where data for this research was collected, only 33.2% of the males and 36.4%
of the females who were in the age of primary school were enrolled in 2002, and still more than one-third of the total population is illiterate (see Bulambuli District Local Government Five Year District Development Plan (2010/2011-2014/2015).
be perceived to be lucrative by the farmers, as they have already given up on the idea that their investment can make a difference. Reframed in terms of the two processes that lead to learned helplessness, the conceptual scheme shows that the situation of farmers can change, i.e. control over the farm income can be regained through lucrative investment opportunities; however, due to an overly pessimistic attributional style causal interpretation may be inappropriately general and, as a consequence, farmer’s expectations may not change accordingly.
3. Methodological Design
The theoretical links this thesis is concerned with, as summarised in the conceptual scheme presented in the previous chapter, are captured in a set of research questions and operationalised into research methods. However, these questions and methods – to be presented in this methodological chapter – are not exclusively the outcome of the theoretical framework, but also of the ontological and epistemological foundation of this research: critical realism. Hence, the first two sections of this chapter introduce critical realism as a philosophical theory of ‘reality’ (section 3.1) and explain how the mixed methods design of this research is related to this ontological position (section 3.2). The remainder of the chapter presents the research questions (section 3.3), the research location (section 3.4) and the methodological design (section 3.5). The latter also explains how sampling was carried out and reflects on the main ethical dilemmas related to the chosen methods.
3.1. Introducing critical realism
Critical realism emerged as an ontological theory in response to the so called ‘epistemic fallacy’ of positivism, the dominant ontological perspective which has shaped our perceptions of science throughout its existence (Bhaskar 1975, 12). The term ‘epistemic fallacy’ was introduced by Roy Bhaskar, one of the earliest writers about critical realism. In his 1975 work A Realist Theory of Science he used the term to point to the misleading implications of positivism’s focus on empirical measurement, i.e. a focus on ‘what we can know about the truth’ rather than ‘what is truth’. This focus implicitly reduces ‘what it is’ (ontology) to ‘what we can know about it’ (epistemology). From a critical realist perspective, Bhaskar argued, this perspective is philosophically wrong, as it fails to address the relationship between the three fundamental domains of reality: what we experience (the empirical), what actually happens (the actual) and the underlying mechanisms that determine what happens (the real) (Bhaskar 1975; and Danermark et al. 2002, 21).
The importance of distinguishing between these three domains of reality becomes clear when analysing the fundamental nature of experiments (the main tool to study reality within the positivist tradition). In experiments, the investigator tries to make ‘visible’ some aspect of reality (causal law) that is usually invisible by manipulating the sequence in which certain events occur. This inevitably makes the researcher a causal agent of the sequence of events in an experiment, but not of the causal law they try to make visible through their interference (Bhaskar 1975, 33). Hence, the key reality the investigator tries to make visible is not the sequence of events (though the focus of positivism), but the mechanisms that determine these events. Hence, ‘the actual’ (i.e. the sequence of events) is ontologically distinct from ‘the real’ (i.e. the underlying mechanism causing events) (ibid).6 Or, in critical realist terms, the
events that happen in the actual domain, which we could (but not necessarily do) experience, measure
6 Explained differently: the researcher produced a certain chain of events but not the causal law that leads to the
or observe in the empirical domain, are caused by the underlying mechanism of reality (the real domain) and may, therefore, be a reflection of reality but not mere reality itself (see table 3.1 for a summary of the three domains of reality). The focus of research should, therefore, be on unravelling the relationships between the domains of reality, rather than simply searching for regularities in the events we observe. Table 3.1: Features of the three domains of reality
The real domain The actual domain The empirical domain
Mechanism X
Events X X
Experiences X X X
Source: Bashkar (1975, 13)
Furthermore, critical realism argues that the underlying mechanisms of reality exist largely independent of our knowledge of them (Sayer 2006). Even if ‘science’ had not existed there would still be gravity and photosynthesis, two mechanisms we have come to know about through science. Our knowledge of gravity and various other mechanism has changed through science, but the very nature of these mechanisms has not. This position opposes non-foundational or interpretivist ontologies which argue that all knowledge is infinitely relative as it is inevitably shaped by our language, concepts and ideas (Danermark et al. 2002, 17). Although critical realists agree with such ontologies, that our knowledge is theory and language dependent because facts are shaped by our theories7, it rejects their assumption
that facts are theory-determined (Danermark et al. 2002, 15). Epistemologically, this means that all knowledge is fallible, although not equally fallible, and that science has the function to produce knowledge that is the least fallible.
This leads to questions about which methodologies, from a critical realist perspective, are most appropriate in capturing the relationships between the real, actual and empirical domains of reality so that they can adhere to the principle of least fallibility. In their answers to such questions, critical realists have made the distinction between the study of natural phenomena (embedded in closed systems) and the study of social phenomena (embedded in open systems). Whereas natural objects are socially defined and naturally produced, social objects are both socially defined and socially produced. Though equally real, this implies that social mechanisms differ across contexts as they are not produced in a universal way throughout time and space (i.e. they are part of an open system), while natural mechanisms do not (i.e. they are part of a closed system). Conforming to this line of inquiry, some critical realists have argued that experiments are often less suitable for the study of social objects, as opposed to natural objects, because (I) the observed social phenomenon may not coincide with its equivalent in the real world as it is context dependent and (II) the original object of study may be lost after the experiment as
7 For example, theories inform the way we manipulate events in experiments and thus what may come around as
the experiment itself interferes with social phenomena (Danermark et all. 2002, 35). How social phenomena should be studied alternatively, according to critical realists, depends on the necessary and constituent properties of the object that is being studied. It is considered to be essential that the chosen methodology captures those properties that determine the object’s causal powers and susceptibilities, i.e. determine its generative mechanisms (ibid, 70; and Sayer 2006).
3.2. Applying critical realism
With this critical realist perspective in mind, this section aims to determine which methods are most suitable to investigate the impact of learned helplessness on agricultural investment. In critical realist terms, that means addressing the following question: how should learned helplessness, as a generative mechanism in the real domain, potentially leading to reduced investment in the actual domain, be studied in the empirical domain? Given the discussion of the previous section, the typical critical realist answer to this question is often a mixed methods design so that the generative mechanism under study can be explored from different perspectives in order to capture as fully as possible the different processes that determine its causal potentials and susceptibilities. In their respective contributions to this purpose, quantitative methods help to identify patterns and correlations in the actual domain, while qualitative methods help to get to the underlying mechanisms structuring these patterns and correlations (Harrits 2011, 162).
Hence the research design employed in this thesis is a mixed methods design, which includes a survey, field-experiment and semi-structured interviews. The quantitative methods employed focus on the actual domain of reality as they provide the necessary data to examine some of the expected patterns of events in the wake of the current application of the learned helplessness theory (as presented in the conceptual scheme). The semi-structured interviews aim to contextualise these quantitative findings into a real-life setting by focusing on the ‘real’ domain of reality; i.e. the underlying perceptions, beliefs and expectations that explain the real-life equivalents of the learned helplessness-related events as observed in the quantitative research stage.
Note that the current design does not entirely fit the critical realist tradition as critical realists have argued against the use of experiments to study social phenomena. The reason for including a field-experiment in the mixed methods design – despite critical realist critiques – is practical and revolves around the observability of the ‘event’ this thesis is concerned with, that is reduced agricultural
investment. The conceptual scheme presented earlier implies that smallholders in developing countries
may learn to be helpless when it comes to their farm income and therefore decide to invest less in their farm. The word ‘less’ in the previous sentence is crucial for the argument to be made here as it illustrates the intention to observe something relative to something else; i.e. a helpless farmer invests less relative to what they would have invested if they had not learned to be helpless. To experience, measure or observe in the empirical domain a ‘relative’ event of this kind taking place in the actual domain (the
main focus of the quantitative research stage) is impossible without the creation of a counterfactual, which in turn can only be created through a random assignment of participants to a control group in an experiment.
A field-experiment is thus necessary to objectively observe the kind of event this thesis is concerned with. That does not mean, however, that critical realist critiques on the usage of field-experiments to study social phenomena are no longer valid. Social phenomena are indeed context dependent and may, therefore, play a different role outside the field-lab; i.e. experiments may lack external validity. As the field-experiment is only one partial and artificial way of covering the empirical and cannot be fully equated to the empirical at a 100% rate, the results of the field-experiment should not be seen as a direct reflection of reality but rather as an indication of what reality could look like if its main circumstances resemble those created in the lab. The field-experiment explores whether learned helplessness can take hold under certain circumstances and, if so, what these circumstances look like, However, these results are of little value if they are not considered in combination with qualitative data which focuses on the real domain of reality, in order to reflect upon the generative mechanisms underlying the events observed in the field-experiment as they appear in real life (see also Jackson 2009, 2011).
Note that the chosen order in which both types of data are collected and analysed is not motivated/discussed here, while in fact the sequence in which both types of data occur matters to the overall value of the study. In the discussion chapter (subsection 7.4.1) it is argued that the (necessary) triangulation of experiments with qualitative research may be more effective if the qualitative research stage proceeds the experiment.
3.3. The research questions
With this epistemological approach in mind, the theoretical focus of the thesis has been captured in the main research question, which has been further divided into six questions. Half of these sub-questions are quantitative and half are qualitative sub-sub-questions (see box 3.1).
The quantitative sub-questions were formulated in line with the experimental design (set out in the next section) in which farmers are given an investment opportunity to reduce the risk of a farm income shock in real life. Each quantitative sub-question focuses on a separate aspect of the theoretical application of the learned helplessness theory as a potential explanatory factor for the investment decision. Sub-question 1 focuses on the extent to which a lack of control over income losses in an experimental pre-treatment task affects an agricultural investment decision with real-life implications. Sub-question 2 is concerned with the extent to which real-life income shocks experienced in the past two years affect this agricultural investment decision. The underlying idea of this question is that people who experienced more different income shocks with a larger impact are more likely to perceive a lack of control over their farm income, which in turn would make them more susceptible to a learned helplessness bias. Sub-question 3 tests the impact of attributional style on investment.