What do we know about Nigerian farmers attitudes to uncertainty and risk? A systematic review of the evidence

In developing countries, specifically Nigeria, the vulnerability of smallholder farmers to agricultural uncertainties and risks is a major concern. Several studies that examine Nigerian farmers have highlighted how risk attitudes influence farming decisions. However, there is a gap in the synthesis of evidence necessary to enhance our understanding of farmers' attitudes to uncertainty and risk. Thus, we conducted a systematic review of the literature to understand farmers' attitudes to uncertainty and risk and the implication on farm-decision making. After the final screening, 39 papers met the inclusion criteria for our review. We find that the literature on the topic in Nigeria is small but expanding. We also find that risk attitude is measured through both econometrics and experiment methods, with the former being the most popular method. Twenty-six (26) papers (67%) reported that most farmers are risk-averse, although at different levels. The most consistent predictors for risk attitudes are age, household size, education, income, poverty status, farm size and access to credit. Other significant determinants of farmers’ attitudes towards risk are household size, years of farming experience, health status, leadership position, and organisation membership. Attitudes to risk were also used to explain the differences in several farm decisions and livelihood strategies. Although the reference to risks took a relatively different perspective and constructs, the overarching finding was that a risk-averse farmer is more likely to make production and management decisions that will result in a lower average return if the associated uncertainty is lower. One benefit of our findings is that policymakers will be informed to accurately predict the choices and behaviour of farmers under different policy interventions. We conclude that there is a need for more empirical research, and accordingly, we suggest a research agenda that could address the gaps identified.


Introduction
Farmers, especially those in low-and middle-income economies, face several sources of uncertainty and risk (such as weather, climate, diseases, price, market, and institutional, among others). As a result, these farmers operate in precarious conditions (Assouto et al., 2020;Mwangi et al., 2021). Previous global studies have found that attitudes to uncertainty and risk play an important role in farm decisions (Greiner et al., 2009;Arinaitwe et al., 2017;Sarwosri & Mußhoff, 2020;Kiggundu et al., 2021). Similarly, several studies in Nigeria have highlighted the influence of risk attitudes on farmers' decision-making. However, as these studies have examined the effect of risk attitudes on specific decisions (e.g., adoption, participation in intervention schemes or specific farm management decisions), there is no holistic information necessary to generate sufficient attention to the role risk attitudes play in farm decision making. Besides, independently, the available studies cannot provide strong support for or dispute the assertion that farmers are generally risk-averse. Thus, we systematically reviewed the literature to provide a higher-level synthesis of the evidence. The objective of this paper was to (a) evaluate the extent to which risk attitude is considered in farm decision-making, (b) identify the approaches used, (c) provide empirical insight into farmers' risk attitudes, (d) evaluate the quality of the studies and (e) suggest a research agenda to address the gaps identified.
Our goal is for a holistic understanding of farmers' attitudes to uncertainty and risk, which would guide the design of risk management instruments and strategies that could help reduce the potentially disruptive effects. Therefore, this review could effectively increase accuracy in predicting Nigerian farmers' future choices and behaviour, at least in the aggregate, and guide policy design.

The importance of measuring risk attitudes of Nigerian farmers
In Nigeria, over 80% of farmers are categorised as smallholders (Anderson et al., 2017).
These farmers constitute 50% of working Nigerians. Notably, half of these smallholder farmers fall within the poorest 40% of the population (World Bank, 2016). Considering that many farmers are already in a precarious situation, the uncertainty and risk associated with farming threaten farmers' resilience and sustainability. Besides, the absence of insurance markets exacerbates the problem (Akpan et al., 2015;Olaoye et al., 2017). Recently, there have been interventions aimed at improving the livelihood of smallholder farmers. To achieve this goal, it is important that there is empirical evidence of farmers' behaviour in the context of risk and uncertainty and discernment of its sensitivity to acceptance of the intervention.
Globally, there are generalisations about the risk attitudes of farmers. Specifically, smallholder farmers are stereotyped as being highly risk-averse. However, these generalised assumptions have been disputed with evidence (Henrich & McElreath, 2002;Vieider et al., 2019). Also, findings from studies beyond Nigeria may not reflect the behaviour of Nigerian farmers. This is because evidence shows that risk attitudes vary across regions, impacted by the differences in cultural and institutional factors (Henrich & McElreath, 2002;Ferreira, 2018). Hence, it is important to collate the relevant empirical evidence on Nigerian farmers to provide a holistic interpretation of the findings in past studies and produce reliable and accurate conclusions.
The rest of the paper is organised as follows. In section 2, we briefly present the relevant concepts and context. Section 3 presents the steps in selecting the eligible studies for the systematic review. In Section 4, we report our findings, then discuss the finding in Section 5 and propose a research agenda. Finally, in section 6, we conclude the paper.

Concepts and context
In the agricultural economics literature, econometric estimation based on empirically observed data has been used to determine risk attitude. The variance of return is employed to measure risk as with the portfolio optimization mean-variance approach of Markowitz (1959). Other methods are the safety-first criterion (Roy, 1952) which is premised on a decision-maker considering outcomes below a particular level as a disaster and the threshold varying per individual. Thus, in a risky situation, each individual tends to make decisions that minimize the probability of attaining a disaster. Besides econometric methods and selfassessment of risk attitudes, a predominant elicitation instrument is the lottery-choice tasks.
This task takes numerous forms, e.g., the easier to comprehend Eckel and Grossman task where participants of the experiment are given a set of 50-50 gambles, a sure outcome and a number of risky outcomes with linearly increasing expected payoffs and associated risk that is estimated as the standard deviation of expected payoffs). On the other hand, in the popular Holt and Laury task, participants are presented with ten decision tasks from which they have to choose between a pair that is riskier than the other. The probabilities are systematically varied, and the expected value changes with a new lottery pair.
As opposed to estimations based on field data, experimental methods permits that the elicitation process and conditions are controlled and identical for all participants. Considering that the attitudes of every participant can be observed individually, this method is advantageous over the field data methods. For a detailed discussion of the different elicitation instruments, readers can refer to Andersen et al. (2010), Lönnqvist et al. (2015), and Crosetto & Filippin (2016).

Concept and sources of uncertainties and risks in agriculture
Farmers make decisions in the presence of a multitude of uncertainties and risks. Across studies in the literature, risk has been used interchangeably with uncertainty. We distinguish between risk (known probabilities) and uncertainty (unknown probabilities). For simplicity, we adopt De Groot & Thurik (2018) definition for risk as those circumstances in which the outcome is unknown, but the probability distribution governing that outcome is known, while uncertainty refers to those circumstances in which both the outcome and probability distribution are unknown. Unlike the case of risk, under circumstances of uncertainty, there is no valid basis for classifying instances. As such, there are no probabilities to not know, and probability logic does not exist for such cases. In the literature, an individual is generally categorised as risk-averse, risk-neutral or risk-seeking (Kahneman & Tversky, 1982;Iyer et al., 2020;Hannus & Sauer, 2020), with a similar classification for uncertainty. For simplicity, we refer to risk aversion (respectively risk-seeking) in a standard albeit weak context in which an individual prefers a random variable less (more) than the degenerate random variable with the same mean (Chateauneuf & Cohen, 1994). There is also evidence that risk attitude is domain-specific (Weller & Tikir, 2011). Hence, most decision-makers exhibit risk aversion for gains and risk seeking over losses (see Tversky & Kahneman, 1992).
A categorisation of the sources of uncertainties and risks in agriculture suggests that there are five main groups, namely: (i) production uncertainties (which are attributed to natural growth processes of crops and livestock, including weather related factors), (ii) price and market uncertainties (input and output price fluctuations that characterises agriculture in many lowand middle-income economies), (iii) financial uncertainties (v) institutional uncertainties (brought about by government actions) and (v) uncertainties from human or personal factors (Ullah et al., 2016;Komarek et al., 2020).

Information sources and search strategy
We conducted a comprehensive systematic literature search using multiple electronic databases, i.e., Web of Science, Google Scholar and SCOPUS. The combination of multiple databases is crucial to conducting efficient searches in systematic reviews as it eliminates or reduces the chances that eligible studies on a particular topic are left out (Bramer et al., 2017). In developing the strategy to search the electronic databases, we combined the key terms (risk, uncertainty, attitudes, behaviour, preference, farmers, crop, livestock, animal, and Nigeria) using Boolean operators. We scanned abstracts of papers returned by keyword searches for relevance. We also scanned the references of papers returned from the keyword searches (backward reference searching) to confirm no eligible papers were missed.

Eligibility criteria
Studies eligible for inclusion met the following criteria. The papers (i) measured or elicited risk or uncertainty attitudes, (ii) were published between 1991 and 2021, and (iii) focused on farmers in Nigeria (iv) the paper was published in a peer-reviewed journal or a conference proceeding. We excluded other grey literature, review and conceptual papers as these fell outside the scope of the review.

Data extraction/screening of articles
For data extraction and quality assessment, each paper was independently reviewed by two authors in line with the inclusion and exclusion criteria of the screening tool. The third author resolved any inconsistency.

Assessment for methodological quality
We assess each study for methodological quality following the UK Department for International Development (DFID) guidance for assessing the strength of evidence based on the conceptual framing, methodology, analysis, relevance, and contribution. We assigned each paper a rating ranging from low to high for each domain.

Data analysis
We extracted information on the sample size, farm types, methodological approaches, and main findings. We used descriptive statistics to summarize the relevant information from the eligible studies. However, several factors did not permit a meta-analysis of the findings.
Specifically, the limited number of papers, the heterogeneity of the studies due to the different elicitation methods and scoring, e.g., among studies that used Likert type self-report vs multiple price list vs hybrid experiment to categorize farmers' risk attitudes. In addition, some studies are missing important information that would otherwise have been used to compute the meta-analysis. These issues implied that a meta-analysis might be meaningless.

Description of included studies
The search yielded 496 articles comprising journal papers, conference papers and grey literature ( Figure 1). The 244 papers consisted of articles returned in more than one database.
Thus, we dropped 273 duplicates. Further screening of the title, abstract and full text resulted in a further 214 being dropped. After the final screening, 39 papers met the inclusion criteria for our review. The assessment of the strength of evidence was low-to-moderate, with only 12 papers (31%) satisfying the criteria for strong strength of evidence.
[ Figure 1 here] The studies that constitute this review measured risk attitudes from three farm types: namely crop farmers (30), poultry farmers (6) and fish farmers (3). Over the years, there has been an increase in the number of studies suggesting that the literature is growing. As shown in Figure 2, 76% of the papers that met the inclusion criteria in this review were in the last ten years, with peaks in 2019 (6 publications). Notably, there were only two empirical papers on uncertainty/ambiguity. The remaining 37 studies examined risk attitudes. A common feature across the studies that measured uncertainty/ambiguity was that farmers' behaviour under uncertainty/ambiguity was similar to that of risk, albeit not a replica. Hence, our discussion onwards will refer mainly to risk attitudes.

Measurement methods and elicitation techniques
Risk attitude estimation by econometric methods is predominant in the literature examined (Table 1). Of the different econometric methods, the Safety-first approach was the most popular (29 papers). However, only 5 studies were based on an experiment. Of these experiment-based studies, 2 employed the Multiple Price List, 2 used hybrid tasks, while the Becker-DeGroot-Marschak method was used in 1 study. For the (11) studies that employed alternative measures of risk attitude using attitudinal scales, farmers were asked to rank the level they are willing to take risks on a scale. Otherwise, farmers were asked questions that required them to state their risk attitude, e.g., "What is your general attitude toward risk?".

Empirical insight into farmers risk attitude
On the one hand, compared to 4 papers in which most of the farmers were risk seeking, 26 papers (67%) reported that the majority of farmers are risk averse, albeit at different levels (i.e., low to high risk aversion). On the other hand, the findings in 4 of the papers were mixed as there was no majority (Table 2). That is, farmers, risk attitudes in those papers were more heterogeneous. Also, contrary to previous assumptions, we find (2) studies that observed risk neutral attitudes among farmers. This review highlights that although most Nigerian farmers are risk averse, all farmers do not have homogenous attitudes toward risk. In this light, we categorise the studies into regions to examine whether risk attitudes differ by region given the heterogeneous settings of Nigeria in terms of factors such as agroclimatic conditions and culture, which are known to influence risk attitudes. The summary in Table 3 indicates that farmers in the Southwest region of Nigeria appear more risk-seeking compared to other regions. As such, a 'one size fits all' risk management approach is likely to be ineffective.
[ Table 2 here] [ contradictory to many past studies in other countries as we find mixed results. One set of studies (N = 11) found that the relationship between age and risk attitude was mainly negative, suggesting that older farmers tend to be more risk averse. While another group of studies (N = 6) reported that risk aversion decreases with age, considering that older farmers have experience in making economic decisions under uncertainty. Most studies find that larger farm owners were more likely to take a risk. Since most smallholder farmers are in the lowest income category in Nigeria, the postulation is that farmers become less willing to commit their meagre earnings to chance.
Similarly, poverty was predominantly associated with higher levels of risk aversion. That is, farmers whose incomes fall below the poverty line were less willing to take risks.
Availability of social networks was also reported to drive risk seeking attitude. Further, neighbourhood-effect was a significant determinant of farmers' attitude toward risk, i.e., residing in a cluster of farmers with similar risk attitudes tends to impact individual farmers' risk attitude.

The role of risk attitude in farm decisions and livelihood strategies
Attitudes to risk were used to explain the difference in the adoption of technologies and agronomic practices (5 studies), farm management strategies (9), poverty (2), participation in intervention schemes (3), willingness to pay (1) and off-farm participation decision (3). The estimation was mainly through regression analysis. The findings on the impact of risk attitude on adoption are mixed. For some risk management tools or strategies (e.g., high-yielding variety crops and the use of modern technologies), risk-averse farmers were less likely to adopt. On the other hand, risk averse farmers mainly adopted other risk management strategies such as crop diversification, polling resources via cooperatives and informal precautionary savings). Also, risk averse farmers mainly participated in national, or state intervention programs targeted at food security. Although authors refer to risks from a relatively different perspective and construct, the overarching finding is that risk averse farmers are more likely to make production and management decisions that will result in a lower average return if the associated uncertainty is lower. Considering that many farmers in Nigeria are trapped in the cycle of poverty, these studies contend that risk aversion contributes to many farmers remaining poor.

Implications of the findings
In agriculture, as compared to several occupations, there is a peculiarity in the decisionmaking environment since it is dominated by risks. Also, there is the fact that in many cases, the decisions are made by a single farmer. Thus, the farmers' risk attitude plays a significant role in the decision process. Further, risk aversion is reported to affect a single input use such as fertilizer. Also, the combination of inputs chosen by a risk-averse farmer will differ from those chosen by a risk-neutral or risk-seeking farmer. Besides, the farmer's risk attitude influences the decision to adopt modern technology. These, in turn, have significant implications for yield and farm income. Therefore, to ensure that risk attitudes do not pose a limitation to making better farm decisions, the government could implement programs targeted at stabilizing farm incomes by decreasing the risks associated with farming, perhaps, through insurance.
Regarding direct policy implications, the impact could differ depending on farmers' risk attitudes. For example, the government promotion and dissemination of a high-yielding variety crop may not meet its goal if it is targeted at farmers that are risk averse. This is because most high yielding varieties require increased inputs to realize their higher responses.
Thus, carrying additional risk. Thus, despite the yields being considerably higher, risk averse farmers may be unwilling to adopt it. We discuss further the implications for policy in section 6.

To what extent does this review align with the global literature?
We compare the current review to previous global reviews on the topic. This systematic review uncovered several key findings. First, the literature on Nigerian framers' risk attitudes is nascent, unlike the global literature on the topic in general and those focusing on farmers in particular. However, it somewhat follows the trend in the wider literature wherein the number of studies increases over time.
In the global literature, the approaches for examining risk attitude range from first, an assumption of risk-indifference, and on that account, a goal of maximising expected gains or minimising expected losses and second, specifying a utility function. Third, employing a stochastic efficiency criterion that aligns with the assumptions of the expected utility theory.
Fourth is a direct measurement method of risk attitudes (see Marshall et al., 1996). This Compared to the wider literature (Sauter et al., 2015;Attanasi et al., 2018;Iyer et al., 2020), Overall, the evidence supports the assertion that Nigerian farmers are mostly risk averse.
However, we make this generalization with caution as risk attitudes are domain and context specific. Again, a similar (risk averse) attitude has been reported among European farmers (Iyer et al., 2020) and farmers in several low-and middle-income countries (Hurley, 2010).

Challenges to conducting experiments to elicit farmers risk attitude in Nigeria
The direction of the wider literature indicates that experimental elicitation of risk attitude is the way forward. However, there are challenges to implementing experiments among farmers in Nigeria. This is mainly due to the level of cognitive burden the experiment poses. Most farmers would find it difficult to vary probabilities and payoffs in the manner presented in standard lottery experiments. Nmadu et al. (2012) argue that many of the standard experiments used to elicit risk attitudes are unknown to or difficult to understand by farmers in Nigeria, thereby justifying the alternative methods adopted to measure farmers' risk attitudes. Other studies (Cardenas & Carpenter, 2008;Charness & Viceisza, 2016;Ihli et al. 2016) have reported that the time involved, the level of general trust of the participants, the economic conditions and the barriers the use of electronic devices other than mobile phone poses makes conducting such experiments challenging to conduct among low literacy participants.
Researchers measuring farmers' risk attitudes in Nigeria via experiments have used hybrid methods to address some of these concerns. This approach highlights that authors have to design the experiment to reduce the cognitive burden and ensure that it captures and retains participants' interest. This reduces or eliminates the perception of undergoing a 'test' that some standard lottery methods induce. Also, experimental and non-experimental methods could be combined to provide more reliable measures of risk attitudes. Further, acknowledging that cultural aspects (e.g., beliefs nothing happens at random) may interact with interpreting probabilistic events and making provision to disentangle these factors from the experiment is crucial to improving experimentally derived risk attitude in Nigeria.

Research agenda
We outline a research agenda with the aim of stimulating discussion about future research directions that emerged from the review of the existing literature. Four fundamental research needs emerge, which we discuss. First, in terms of the methodological and elicitation methods. Risk attitude estimated by econometric estimation has not been compared to experimental results in any study in Nigeria. Therefore, testing and comparing the outcome of elicitation methods for consistency of risk attitudes will help uncover the 'better-suited' method. Besides, a move towards more experimental-based studies will also provide the platform to test several theories, including the popular prospect theory, which provides one of the leading descriptions of how a decision-maker assesses risk in an experimental setting.
Also, future studies could estimate panel data to measure whether Nigerian farmers risk attitude is stable over time. Besides, the impact of new policies on farmers' risks attitude could also be examined from this dimension.
Second, it is necessary to improve the evidence base on uncertainty and ambiguity. Previous studies show that although the behaviour under uncertainty and risk are similar, one is not necessarily an accurate reflection of the other. The need to expand the literature on uncertainty is particularly important as many of the decisions that farmers take are under conditions of uncertainty. Third, the consideration in previous studies has been individual farmers or farm households as a single entity. There is a gap in our knowledge of whether and how an individual farmer's risk attitude is influenced by interactions with other farmers nearby. Thus, we observe a need to investigate the presence of spatial clusters of farmers with similar risk attitudes. The advantage to policymakers will be substantial. Fourth, to improve the quality and credibility of the data and future studies, more attention needs to be paid to the conceptual framing, methodology, analysis, relevance, and contribution of each study.
This will result in more robust studies that satisfy a high criterion in terms of the strength of evidence.

Conclusion
Previous empirical studies examined the relationship between risk attitude and farm decisionmaking. We identified gaps in the synthesis of evidence needed to have a holistic understanding of farmers' attitudes to uncertainty and risk and how these attitudes shape agricultural decisions. Therefore, we conducted a systematic review of the literature, which included 39 papers after applying exclusion and inclusion criteria. We found that the measurement of risk attitude is mainly through econometrics and experiment methods. Also, the results from our review suggest that most Nigerian farmers are risk averse but at different levels. Taken  This review consolidates previous findings in Nigeria and similar low-and middle-income countries that risk averse farmers tend to make production and management decisions that result in a lower average return if the associated uncertainty is lower. This suggests that researchers should focus on expanding the methodological and elicitation methods and carry out more studies on uncertainty and ambiguity to move the literature on attitudes and behaviour forward. The benefit of policy design is that policymakers will be informed to accurately predict the choices and behaviour of farmers under different policy interventions.
It will also help policymakers to judge the effects of policy on farmers' risk attitudes. For example, policies that reduce poverty or increase wealth may change farmers' risk attitudes in the case where risk is influenced by wealth. However, we caution that any assumption of homogeneous behaviour integrated into policy design without consideration for context and domains will be fundamentally flawed.

Declaration of interests
☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.    Backward and forward citation analysis (n = 9)