Maize farmers’ adaptation to drought: Do risk attitudes and perceived risk probability matter?

Abstract

Drought effect is the most common consequence among all aspects of climate variability phenomenon. Accordingly, understanding farm-level decisions to adopt strategies to mitigate the negative impact of extreme climate events such as drought is a vital ingredient for making policy suggestions necessary for effective adaptation to climate change. The study aims at investigating the effect of risk attitudes and perceived probability of drought occurrence on farmers’ decisions to adopt new drought-tolerant maize varieties to adapt drought associated with climate change. Data was collected from farmers who were sampled using multistage sampling techniques. Estimation of standard Bayesian Probit, completed by using Bayesian Reversal Jump Probit analysis was done to arrive at study results. Farmers were found to adapt to climate change using Integrated Pest Management (IPM), adjustment in planting and/or harvesting times, and growing special new varieties. Majority of farmers were willing to adopt new drought-tolerant maize variety rather than the traditional ones they grew in the past. Further, the study revealed that farmers with risk seeking attitudes who are less sensitive to losses, and also perceived the occurrence of extreme drought as highly probable, are more likely to adopt and cultivate drought tolerant maize variety, when recommended to them. As part of a planned course of action necessary to adapt to climate change, planning of appropriate extension communication approach that would facilitate farmers’ adoption of the new tolerant-variety must understand farmers’ decision-behaviour under risk, and their perceptions of the riskiness of the strategies.

Keywords:

• maize farmers
• extreme drought
• adaptation
• risk attitudes
• perceived drought probability
• Ghana

Public Interest Statement

Climate change has been manifesting in the central region of Ghana through increasing temperature and decreasing rainfall amidst high variability. Maize farmers have experienced incidents of droughts that have highly impacted negatively on their production and livelihoods in the Central region of Ghana. Most of the maize farmers have taken decision to adopt drought tolerant maize variety to adapt to extreme drought and related weather events associated with climate change in the central region of Ghana. Higher probability of drought (PPD), as perceived by farmers, increases the likelihood of farmers’ decision to adopt a drought tolerant maize variety in order to adapt to climate change. Farmers who have risk seeking attitudes and are less sensitive to losses are more likely to adopt and cultivate even “risky” new drought tolerant maize variety, when recommended to them.

1. Introduction

In Ghana, evidence abounds that temperatures in all the ecological zones are rising and at same time, rainfall levels and patterns are becoming erratic (Agyemang-Bonsu et al., 2008). Analysis of long-term climate data shows a general increase in temperature in the country with a steady annual rise of 0.06°C and an overall increase by about 1°C over the past 40 years (World Bank, 2020). Further, projections of future climate show that there will be an increase in temperature for all agro-ecological zones but changes in precipitation will vary considerably both spatially and temporally (World Bank, 2020; Nutsukpo et al., 2012). Minia (2004) reported projections by the Ghana Meteorological Agency (GMet) that suggested indicated that rainfall will decline by 8.8 percent by 2050 and 14.6 percent by 2080. In the Central Region where the current study was conducted, records show increased temperature, declining rainfall totals coupled with increased variability and changes in land use and land cover (Dadzie, 2019; Mahmood et al., 2014). Climate change provokes extreme and unpredictable events including prolonged droughts, heat waves, and erratic beginnings and ends to the rainy seasons resulting in shift in cropping season. The national economy stands to suffer from the impacts of climate change because it is dependent on climate sensitive sectors, most especially agriculture. Climate change and variability effects threaten food production systems and land use management, water and other natural resources. Climate change lowers agricultural productivity, with substantial yield gaps, livestock offtake and decrease opportunities for future prosperity. Postharvest losses are not left out due to erratic and unpredictable weather conditions. Climate change, therefore, creates challenges for food security and livelihoods of millions of Ghanaians who depend on agriculture.

In most of the sub-Saharan Africa’s semi-arid and sub-humid areas, of which Ghana is no exception, drought effect is the most common consequence among all aspects of climate variability phenomenon (Wilhite and Glantz, 1985). For instance, a range of physical consequences of drought have been documented in sub-Saharan Africa to include: damage to natural habitats, plant scorching, increased fire hazards, crop withering and drying, and lack of feeding and drinking water due to drying-out of water bodies (Hyland & Russ, 2019; IPCC, 2007b). In the case of drought destruction of natural habitats, the extreme effects can be seen as some plants and animals going into extinction (Hyland & Russ, 2019; IPCC, 2007a; Maxwell et al., 2019). Drought is a climatic condition characterized by temporal deficiency of rainfall significantly below the normal or expected amount in a specific time period, say a year, season, or month (Van Schaik and Reitsma 1992). Drought can further be described in the context of the farm as a condition of lack of sufficient water for full plant and animal growth and living which may cause losses in production (Van de Geet, 2004). The fact that different plants on farms for example require different amounts of moisture for full growth, makes realization of drought on the farm relative to specific plants and thus a difficult climate problem to investigate in its totality. In Ghana, just as the case in most of the sub-Saharan African countries, most farmers continue to engage in rain-fed agriculture with little or no irrigation; accordingly, drought appears to be the most severe climatic phenomenon impacting negatively on food production. Here, livelihoods are immediately impacted through drought’s effect on crop production and farm yields thereby undermining national harvests; a threatening situation that results in farm households and national food insecurity (Lolemtum et al., 2017; Shiferaw et al., 2014). In the Central Region where the study was conducted, there have been some incidents of extreme climatic conditions such as increased temperatures and variability in rainfall patterns that bring about drought effects at farm settlements (Dadzie, 2019; Quartey, 2010; Kemausuor et al., 2011; Limantol et al., 2016; Adu-Boahen et al., 2019). These episodes also occur in other parts of Ghana where maize farms are suffering from extreme events of erratic rainfall, long period of intense heat, and droughts associated with climate change (Limantol et al., 2016; Adu-Boahen et al., 2019). The adverse events and stresses associated with climate change pose a devastating risk of yield loss in maize production systems and, by extension, food security. Maize is one of the most important food staples produced by farmers; hence any decline in their production has serious repercussions for the country’s food security. In their study, Dadzie et al. (2011) found out that most farmers had consistently experienced a decreasing trend in the yield levels of their food produce, and consequently had experienced a decrease in their ability to afford right dietary and nutritious food to feed their households. Unexpected and high variability in rainfall brings about droughts and other extreme weather events leading to soil degradation and unpredictable growing seasons; resulting in reduced crop yields (Altieri et al., 2015; Lemi & Hailu, 2019).

This makes it imperative for farmers to adapt to climate change with measures/strategies that effectively can mitigate the negative impact of extreme events (drought). To tackle the challenge of climate change and simultaneously eradicating hunger, and achieving the SDGs more broadly, “will require a profound transformation of food and agriculture systems” (FAO, 2016). Adaptation technologies will inevitably play a vital role in this transformation. In the agriculture sector, common adaptation technologies include seasonal forecasts, which support planning decisions and early warning; water efficient irrigation, which helps farmers cope with drought and variable rainfall patterns; resilient crop varieties, which boost food security by enhancing resistance to water- and heat-stress, salinity, pests and other factors; and farmer-led sustainable agriculture practices, which help ensure farmer ownership, the sustainability of agricultural techniques, and their suitability to their context (Clarkson et al., 2022). In Africa, extensive research into farmers’ adoption of technologies and or measures to adapt to climate change effects have been done. These researches aim at findings that are expected to inform policy on increasing capacity to adapt thereby reducing vulnerability impact on the farm and subsequently increase food security in Africa. However, research into farm adaptation decisions have gone on over the years with significant gaps that still leave room for the need for further studies. This paper helps to fill some of the existing gaps with a focus on maize farmers’ adaptation to extreme drought in Ghana. The subsequent paragraph discusses some gaps in the farm adaptation literature that have motivated the paper.

The literature on the determinants of adaptation to climate change suggests that most previous studies analysed pooled data collected from a number of different countries in the sub-Saharan Africa. Many of these studies findings were drawn based on the aggregated data.

Thus, findings and conclusions might be too general to represent respective countries’ specific scenarios. The data aggregation approach used by most of the studies neglected some vital characteristics that play a vital role in affecting farm level decisions to adapt to climate change. Risk preference attitudes of farmers, for instance, have been widely neglected as determinants of farmers’ climate change adaptation decisions; even though there is a sufficient age-on theoretical basis to investigate the possible effect of risk preference attitudes on farmers’ climate change adaptation decisions (see Adejoro, 2000; Baffoe-Asare et al., 2013; Comoe & Siegrist, 2013; Kenyon et al., 2008; Liu, 2008; Mistiaen & Strand, 2000; Pennings & Wansink, 2004; Picazo‐Tadeo & Wall, 2011; Senapati, 2020, 2020; Shahabuddn et al., 1986; van Winsen et al., 2016). Even in classical agricultural technology adoption decision literature, there is still not much evidence of the effect of farmers’ attitudes toward risk as a factor that influences technology adoption decisions, especially in developing countries of which Ghana is no exception. Thus far in the literature, one can find only a few studies that accounted for the effect of risk attitudes on adoption (see Engle-Warnick et al., 2006; Liu, 2008; van Winsen et al., 2016); but none that empirically investigated causal relationships between farmers’ loss aversion behaviour, for example, and their adaptation technology adoption decisions. The available findings from the literature suggest that farmers who are more risk averse or more loss averse adopt technology late while those who over value small probabilities of gains adopt technology earlier.

This current study seeks to fill the identified gap in the literature on determinants of climate change adaptation decisions of farmers by examining how risk preference attitudes under Cumulative Prospect Theory (risk aversion, sensitivity to losses/loss aversion, probability weighting behaviour etc.) may influence maize farmers’ decision to adopt drought-tolerant varieties to cope with drought in upcoming seasons. This is motivated by the established gap that suggests that previous studies have not done much to investigate into empirical findings regarding farmers’ decision behaviour under risk (beyond risk aversion), and how their risk attitudes might influence decisions to adopt climate change adaptation measures or technologies. However, Kenyon et al. (2008) had asserted that one possible reason for low uptake of climate change–related adaptation measures is the economic agents’ risk behaviour. It has also been reported that interaction between risk attitudes and perception plays a role in decision-making (Arbis et al., 2016; Comoe & Siegrist, 2013; Pennings & Wansink, 2004; Tombu & Mandel, 2015) and that extreme weather events might affect adoption decisions through attitudes towards risks and uncertainty (Liu, 2008). This paper posits that threats posed by extreme weather events associated with climate change may significantly affect farmers’ decision behaviour under risk, and subsequently the decision to adopt climate change mitigating technologies.

Accordingly, the formulated research hypotheses that follow have been empirically investigated: (a) Attitudes towards risks have significant effect on maize farmers’ decisions to adopt new drought tolerant variety; specifically, risk and/or loss aversion (high sensitivity to losses) negatively influences farmers’ decisions to adopt new drought tolerant varieties to cope with drought associated with climate change. (b) Farmers’ perceptions of the higher probability of the occurrence of drought enhance their decision to adopt innovative technologies to adapt to drought associated with climate change. The paper makes a contribution by augmenting empirical factors affecting adaptation to extreme weather events by farmers in Ghana. This will help provide the basic information for further climate change adaptation studies to assist farmers increase their coping capacities. The paper specifically contributes to the empirical driving factors and constraining factors in adapting to drought so that it can assist in finding appropriate interventions for improving adaptation capacity of the farmers in future.

2. Materials and methods

2.1. The survey

In the framework of the cross-sectional survey research design, the study made use of cross-sectional data collected from 384 farmers to describe and explain phenomena as they exist in the study area at the time the survey was conducted. The population of the study was entirely maize farmers in the Central Region of Ghana who have been in the farming business for not less than twenty years. It was assumed that these farmers have had enough experience to be able to observe possible changes in climatic conditions, if any. The maize farmers referred to here were mostly smallholders, with landholdings of less than 2.5 hectares. They mostly lived in rural areas in the study area. Figure 1 shows map of Central Region of Ghana indicating the study area where farmers were sampled from.

Figure 1. Map of Central Region, Ghana showing the study area.

The 384 farmers were sampled using a multistage sampling technique. In the study, sampling was done across different selection stages (i.e., districts →towns/villages →farmers). Series of simple random sampling were done at each stage of the process to arrive at the sample size. In the first stage of the sampling process, there was random sampling of six out of seventeen districts in the Central Region where maize production was predominant. In the second stage, eight towns/villages where maize farmers were mostly located were randomly selected from the each of selected districts with the assistance of the District Departments of Agriculture. Lastly, eight farmers were randomly selected from each of the selected farming communities to be part of the study. In accordance with Williams (2007) and Taherdoost (2016), selecting the sample in several stages reduced the work and cost involved in the preparation and maintenance of the sample frame; it also enhanced administrative convenience of the survey implementation. Note that the study sample chosen through this method was representative of the targeted population consistent with some previous studies (see, for example, Leedy & Ormrod, 2001; Neuman, 2007; Taherdoost, 2016; Williams, 2007); thus, was helpful to ensure valid statistical conclusions.

Out of over 80,000 maize farmers, the use of the 384-sample size, was consistent with Fox, Hunn and Mathers (2009) as well as Krejcie and Morgan (1970); coupled with the rule of thumb from the principle that “more is better” proposed by Darlington (1990). Fox, Hunn, and Mathers (2009) suggested a required sample size of 350 for a quantitative research study that sought to allow a Margin for Random Error (MRE) of ± 5%. It is implied from their determination of sample size that a larger sample size will allow a smaller MRE. Also, in their earlier work, Krejcie and Morgan (1970) developed a table indicating random sample sizes required for population representation and suggested that for a population of 100,000 people, a randomly selected sample of 384 would be adequate. The sample size in the case of this study was determined following Krejcie and Morgan (1970) buttressed by Fox, Hunn and Mathers (2009); therefore is considered as large and adequate, and its use may help to improve Margin for Random Error (MRE).

Structured interview schedule (see Czaja & Blair 2005; Newcomer et al., 2015; Russano et al., 2014; Roberts, 2007 for details) was used as an instrument for data collection. The interview schedule was planned to take place on a face-to-face basis at the selected farmers’ premises where data was collected. This took a period of about eight weeks. The collected data was analysed using descriptive statistics (means and standard deviations, frequencies and percentages), principal component analysis, and Probit regression model which estimation was completed using Bayesian reversal jump probit estimation technique. The subsequent sub-sections present details of the theoretical and empirical econometric specification of the estimated probit model as well as the outline of the Bayesian estimation process.

2.2. Theoretical specification of the estimated probit model

As a discrete choice model, the Probit regression model is based on random utility theory (Greene 2003; Cascetta, 2009). The theory posits that if an individual is faced with a choice problem involving alternatives, the decision maker, j, will choose the option, i, among the alternatives such that the utility that the i-th option gives the decision maker is the greatest compared to its alternatives. Hence in a binary choice case, such as the study context where farmers were confronted with the decision to choose between a new drought tolerant variety to adapt to climate change and their old and existing variety, the concept of the utility maximization is presented as follows: $U_{ij}=Max\left(U_{1j},U_{2j}\right)$. Hence, if the decision maker chooses option 1, then it implies that $U_{1j}>U_{2j}$ where option $1\ne 2.$

In the Probit model, the study defines the dichotomous dependent variable, Y, that denotes a decision maker’s choice as a dummy;

$y=\left\{\begin{array}{cc}1& ifanoptionischosen\left(success\right)\\ 0& ifotherwise\left(failure\right)\end{array}\right.$

Following this, if an individual, j, chooses an option, y = 1, then it implies that the utility associated with that option (i.e., $U_{y=1}$) is higher than the utility associated with its alternative (i.e., $U_{y=0}$). Here, the difference in utilities is described as an unobserved (latent) random variable, y*, given as: $y\ast =U_{y=1}-U_{y=0}$. This defines a criterion function, which must be above a threshold value, 0, for an option y = 1 to be chosen (Fadare et al., 2014; Greene, 2008; Hill et al., 2011). It is assumed that the latent index, y*, is a function dependent on certain factors associated with the decision maker; the econometric specification of this relationship is presented below:

(1) $y\ast =x_j^{\prime }{\beta }_j+e_j$(1)

Since y* is not observed in the process, “y” is used as an indicator to determine whether y* is positive and/or above threshold value,0 as in following: $y=1ify\ast \ge 0$ and $y=0ify\ast <0$. where: y* - is unobserved random variable that reflects binary choice, x – is a set of explanatory variables associated with the decision maker, $\beta$ – unknown coefficients associated with the x variables, and e – is a random error term. e is a standard normally distributed error term.

Assuming that “e” in the latent specification follows standard normal distribution, the probability of y = 1 can be given as $Prob(y\ast >0)=\mathrm{\Phi }\left(x_j^{\prime }{\beta }_j\right)$. If the probability of y = 1 is $\mathrm{\Phi }\left(x_j^{\prime }{\beta }_j\right),$, then the probability of y = 0 is given as: $Prob(y\ast <0)=1-\mathrm{\Phi }\left(x_j^{\prime }{\beta }_j\right)$. Now conditional on estimates of ${\beta }_j$, the probability that option “$y_i$” is chosen defines the likelihood function. This likelihood function is determined as follows:

(2) $Prob\left(y_i{x}_j,{\beta }_j\right)=\mathrm{\Phi }{\left(x_j^{\prime }{\beta }_j\right)}^y(1-\mathrm{\Phi }{\left(x_j^{\prime }{\beta }_j\right)}^{1-y}$(2)

For y = 1, the likelihood function reduces to $Prob\left(y=1|x_j,{\beta }_j\right)=\Phi {\left(x_j^{\prime }{\beta }_j\right)}^y$ or

$L\left(y=1x_j,{\beta }_j\right)=\mathrm{\Phi }{\left(x_j^{\prime }{\beta }_j\right)}^y$

In the probit model, the probabilities are estimated using a normal distribution function (Greene 2003; Hill et al., 2011) as specified below:

(3) $Prob\left(y=1\right)=\mathrm{\Phi }\left(x_j^{\prime }{\beta }_j\right)=\underset{-\mathrm{\infty }}{\overset{x_j^{\prime }{\beta }_j}{\int }}1/\left(\surd 2\pi \right)exp\left(\frac{-z^2}{z}\right)dz$(3)

where:

$Prob\left(y=1\right)$ - is the probability of choosing option y = 1 given x

$\Phi$ - is the cumulative distribution function (CDF)

Z—is a standard normal variable

$exp$ - base of natural logarithm (approximately 2.718)

${\beta }_j$ - vector of unknown parameter to be estimated

$x_j$ - explanatory or predicting variables

2.3. Empirical specification of the estimated probit model

In the study context, the probit model was employed to ascertain the probability that a farmer in the sample would adopt new drought tolerant maize variety as an adaptation to climate change (extreme drought). The objective here was to examine how individual farmers’ risk behaviour (risk and loss attitudes have been estimated from the CPT model—see Dadzie, 2016) and perceptions of the probability of the occurrence of extreme drought associated with climate change could influence or predict farmers’ decisions to adopt drought tolerant variety in the midst of other potential predicting variables. Threshold decision-making theory suggests that when farmers are confronted with similar decisions to adopt a technology or otherwise, as in the context of the study, there is a reaction threshold triggered by a certain number of factors (see Hill & Kau, 1973; Pindyck & Rubinfeld, 1998). The implication of this theory is that a reaction to adopt new technology takes place at a critical threshold level below which adoption is not observed. The assumption underpinning this theory, coupled with the random utility theory, influenced the decision to employ the binary choice probit model in this study. Though not as very popular as the case of the logit model, the application of the probit regression model to investigate farmers’ climate change adaptation decisions is consistent with adoption literature (see, for example, Alabi et al., 2014; Etoundi & Dia, 2008; Fadare et al., 2014; Fosu-Mensah et al., 2012; Gbetibouo, 2009; Hasan & Nhemachena, 2008). The forecast probability in the probit model estimation lies between 0,1 interval. The model is preferred over the logit model because “the logistic law tends to attribute a higher probability than the normal distribution to extreme events” (Etoundi & Dia, 2008: p402). Also, in accordance with Etoundi and Dia (2008), it was envisaged that convergence to an optimal solution might be a problem in the logit regression model estimation but not in the probit model.

In the study, farmers were asked whether they would adopt new drought tolerant maize variety or use their traditional/conventional maize variety in the subsequent year growing season; when the expectation was that drought would occur. Farmers’ responses were coded as y = 1 if the farmer decided to adopt the new drought tolerant variety and y = 0 if otherwise. Following the theory modelling of the probit regression as explained earlier, the probability of a farmer’s decision to adopt the new drought tolerant variety (i.e., $Prob\left(y\ast >0\right)$) was regressed on some hypothesized predicting/explanatory variables as specified below.

Recalling $y\ast =x_j^{\prime }{\beta }_j+e_j$ and $Prob(y\ast >0)=\mathrm{\Phi }\left(x_j^{\prime }{\beta }_j\right)$; the probability that a farmer will adopt the drought tolerant variety was specified and empirically estimated as follows:

(4) $Prob\left(y\ast >0\right)=Prob\left(y=1\right)=\Phi \left[{\beta }_{0j}+{\beta }_{1j}S{e}_j+{\beta }_{2j}A{g}_j+{\beta }_{3j}E{d}_j+{\beta }_{4j}H{s}_j+{\beta }_{5j}F{e}_j+{\beta }_{6j}I{n}_j+{\beta }_{7j}C{r}_j+{\beta }_{8j}E{x}_j+{\beta }_{9j}AT{R}_j+{\beta }_{10j}PP{D}_j\right]$(4)

From the empirical equation, $x_{j\hspace{0.17em}}\left(S{e}_{j\hspace{0.17em}},A{g}_{j\hspace{0.17em}},E{d}_{j\hspace{0.17em}},H{s}_{j\hspace{0.17em}},F{e}_{j\hspace{0.17em}},I{n}_{j\hspace{0.17em}},C{r}_{j\hspace{0.17em}},E{x}_{j\hspace{0.17em}},AT{R}_{j\hspace{0.17em}},PP{D}_{j\hspace{0.17em}}\right)$ – represent the independent predicting variables. $S{e}_{j\hspace{0.17em}}$is the Sex of the farmer, $A{g}_{j\hspace{0.17em}}$is age, $E{d}_{j\hspace{0.17em}}$is education, $H{s}_{j\hspace{0.17em}}$is household size, $F{e}_{j\hspace{0.17em}}$is farm experience, $I{n}_{j\hspace{0.17em}}$is income, $C{r}_{j\hspace{0.17em}}$is credit, $E{x}_{j\hspace{0.17em}}$is agricultural extension, $AT{R}_{j\hspace{0.17em}}$is attitudes towards risk, $P{d}_{j\hspace{0.17em}}$is the perception of the probability of drought occurrence. The coefficients of these variables (${\beta }_{0j},{\beta }_{1j},{\beta }_{2j},{\beta }_{3j},{\beta }_{4j},{\beta }_{5j},{\beta }_{6j},{\beta }_{7j},{\beta }_{8j},{\beta }_{9j},{\beta }_{10j}$), are unknown parameters that were estimated in the probit model following the Bayesian estimation procedure below.

2.4. Bayesian estimation of the empirical probit regression model

Bayesian estimation of the posterior estimates of the parameters in the regression model (results presented in Table 6) was achieved with the procedure below.

Table 1. Explaining independent variables in the regression model

Variable ($x_j)$	Definition	Description of how measured	A priori sign
$S{e}_j$	Sex	Dummy: male=1, female=0	-
$A{g}_j$	Age	Farmer’s age in years at last birthday	+,-
$E{d}_j$	Education	Number of years of formal education	+
$H{s}_j$	Household size	Number of people living in household	+
$F{e}_j$	Farm experience	Number of years of farming	+
$I{n}_j$	Income	Average income from farm	+
$C{r}_j$	Credit	Dummy: Inaccessible=1, 0=otherwise	-
$E{x}_j$	Extension	Dummy: access=1 (i.e. at least 4–7 extension contacts in a year), 0=otherwise	+
$AT{R}_j$	Attitudes Towards Risk (Risk C1…Risk 7)	Factor scores for estimated CPT parameter estimates ($\alpha ,\beta ,$λ,$\gamma ,\delta ,\text{isin},\eta$)	+,-
$PP{D}_j$	Farmers’ perceptions of the probability of drought	Computed probabilities from the farmers’ expectations of drought within the next 10 years	+

Table 2. Results of the principal component analysis

FACTOR LOADING
Risk parameter	Risk C1	Risk C2
$\alpha$	−0.74	−0.62
$\beta$	−0.61	0.77
$\gamma$	−0.97	0.19
$\delta$	0.85	−0.51
$\text{isin}$	−0.15	0.96
$\eta$	0.96	0.10
$\lambda$	−0.20	−0.97
VARIANCE EXPLAINED BY COMPONENTS
Component	Variance	Cumulative variance
1 (Risk C1)	0.57	0.57
2 (Risk C2)	0.39	0.96
3–7 (Risk C3–7)	0.04	1
COMPONENT INTERPRETATION
	Gain domain	Loss domain
	Risk Aversion increases ($RA\uparrow$)	Risk Seeking increases ($RS\uparrow$)
Risk C1	‘Inverted-S’ probability weighting increases ($IS\uparrow$)
	Decreasing weighting of gains (i.e. Pessimism increases - $PESS\uparrow$)	Decreasing weighting of losses (i.e. Optimism increases - $OPT\uparrow$)
Loss aversion decreases ($LA\downarrow$)
	Gain domain	Loss domain
	Risk Aversion increases ($RA\uparrow$)	Risk Aversion increases ($RA\uparrow$)
Risk C2	Increasing weighting of gains (i.e. Optimism increases - $OPT\uparrow$)	‘S’ probability weighting increases ($S\uparrow$)
Loss aversion decreases ($LA\downarrow$)

Table 3. Distribution of farmers’ subjective views of probabilities of occurrence of extreme drought and related weather events in the study area

Adverse climatic events	Subjective probability range					Mean probability (%)	Probability definition
	<5%	5–10%	11–25%	26–50%	>50%
Drought	4.7%	7.3%	6.5%	47.1%	34.3%	41.74 (19.70)	High
Unusually erratic rainfall	2.3%	3.9%	10.9%	57.3%	25.6%	40.06 (18.60)	High
Long period of intense heat	2.9%	2.1%	7.8%	54.2%	33.0%	43.44 (20.62)	High

Table 4. Perceived impacts of EWEACCs on the production/livelihoods of farmers

Adverse climatic conditions	Perceived impact scale					Mean	Overall perceived impact
	1	2	3	4	5
Drought	5.7%	4.2%	16.9%	27.6%	45.6%	4.03 (1.14)	High
Unusually erratic rainfall	7.0%	3.1%	38.0%	27.6%	24.2%	3.59 (1.10)	High
Long period of intense heat	3.9%	1.6%	17.2%	25.5%	51.8%	4.20 (1.03)	High

Interpreting key: 1.0–1.4 = very low impact, 1.5–2.4 = low impact, 2.5–3.4 = moderate impact, 3.5–4.4 = high impact 4.5–5.0 = very high impact

Prompt: impact implies the extent of damage to production

Table 5. Farmers’ knowledge, usage and perceptions of the riskiness of adaptation strategies

Adaptation strategies	Knowledge		Use before		Perceived riskiness		known as CCAS
	Freq	Percent (%)	Freq	Percent (%)	Safe (%)*	Risky (%)*	Freq	Percent (%)
Special new varieties	380	99.0	329	86.5	21.3	68.7	291	76.6
Irrigation and drainage systems designs	361	94.0	196	54.3	87.2	12.8	284	78.6
Mixed cropping	381	99.2	354	92.9	96.8	3.2	245	64.3
Mixed farming	351	91.4	187	53.3	87.1	12.9	173	49.3
Crop rotation	345	89.8	205	59.5	91.8	8.2	192	55.5
Shifting cultivation	348	90.6	182	52.3	87.9	12.1	195	56.0
Crop insurance	219	57.0	0	0.0	92.1	7.9	55	25.0
Adjustment in planting and/or harvest times	344	89.6	205	59.6	80.2	19.8	200	58.1
Soil conservation technologies	320	83.3	174	54.4	91.3	8.7	197	61.5
Integrated pest management	291	75.8	128	44.0	22.3	77.7	106	36.2

*Percentage response from farmers.

Table 6. Bayes probit results for the determinants of farmers’ decision to adopt new drought tolerant variety

Predicting variables	Probit Standard			Reversal Jump Probit
	$\theta$ (Mean)	Std	Pseudo p-value	Mean	Median	Prob. Inc
Int.	2.2*	0.25	<0.001	2.06	2.05	1
Risk C1	1.70*	0.99	0.086	1.41	1.46	0.74
Risk C2	0.96	1.04	0.35	0.28	0.00	0.32
Risk C3	1.21	1.08	0.26	0.47	0.00	0.41
Risk C4	−0.047	1.02	0.96	−0.009	0.00	0.24
Risk C5	0.12	1.13	0.91	−0.026	0.00	0.25
Risk C6	0.14	1.09	0.89	0.022	0.00	0.27
Risk C7	1.34	1.45	0.35	0.52	0.00	0.40
PPD	11.70*	1.53	<0.001	11.27	11.23	>0.99
Sex	−0.016	0.35	0.96	0.003	0.00	0.11
Age	0.133	1.03	0.89	−0.015	0.00	0.23
Education	0.971	0.75`	0.19	0.546	0.00	0.47
Family size	0.41	1.03	0.68	0.208	0.00	0.29
Farming experience	−0.82	0.80	0.30	−0.336	0.00	0.36
Income	1.98*	1.28	0.12	1.34	1.13	0.64
Accessibility of credit	0.95*	0.40	0.017	0.54	0.51	0.62
Agric. extension contact	2.02*	0.49	<0.001	1.72	1.70	>0.99

*significant at the 10% level; at least 90% of the posterior mass is either negative or positive for the affected variables’ parameter distributions (i.e. Confidence Level: 90%). Risk C1, C2-C7: factor scores for the PT parameters.

2.4.1. Prior specification

From the regression model specified, the set of unknown parameters that must be estimated was ${\beta }_j=({\beta }_{0j},{\beta }_{1j},{\beta }_{2j},{\beta }_{3j},{\beta }_{4j},{\beta }_{5j},{\beta }_{6j},{\beta }_{7j},{\beta }_{8j},{\beta }_{9j},{\beta }_{10j})$. These parameters, unlike in the case of random parameter models such as the mixed logit, were fixed (i.e., have zero population variance); as such, they were assumed to be same across the population. Normal prior distributions given as ${\beta }_j\text{~}N\left({\mu }_j,{\delta }_j^2\right)$ (i.e., the most common priors for probit regression parameters (Train, 2003)) on ${\theta }_j$were assumed. Here, the mean, ${\mu }_j$, was allowed to take zero (0) value and the standard deviation, $\delta$, was allowed to take on a value of 3. All the explanatory variables were scaled to lie between 0 and 1 so that the impact of the priors is even across the variables.

2.4.2. Posterior distribution

Given $\left(y_j,x_j\right)\dots ,\left(y_n,x_n\right)$ as a random sample from the conditional probit distribution, the posterior distribution $K\left(\cdot \right)$ which is the full conditional distribution of $\beta ,\mu ,\delta$ conditional on the data observations was specified in the estimation as below:

(5) $K(\beta ,\mu ,\delta |y_j,x_j)\propto L(y_j|x_j,{\beta }_j)F_n(\beta |\mu ,\delta )$(5)

$K\left(\cdot \right)$- is the posterior density

$F_n\left(\cdot \right)$- is normal density with mean, $\mu$ and covariance, $\delta$.

Hence, effectively, the posterior distribution is proportional to the product of the likelihood function of the probit regression and priors on the $\beta$ parameters.

2.4.3. MCMC Simulation

Simulations from the posterior density for $\beta$ parameter estimates were performed to obtain mean statistics as posterior estimators; the results of which are presented in Table 6. The means were computed from the posterior mass by simulation using Markov Chain Monte Carlo (MCMC) methods. Here, draws were taken from $K\left({\beta }_j,\mu ,\mathrm{\Omega }|y_j\right)$ and then averaged to allow the computation of the means for all the parameter estimates. The MCMC procedure was completed using the Metropolis-Hastings algorithm and Gibbs sampling as outlined in Train (2003). The MCMC procedure (Ragier et al., 2009; Train, 2003) followed is summarized in the subsequent paragraph.

The MCMC simulation process starts with Gibbs sampling where draws of $\mu$ conditional on the initial values of $\mathrm{\Omega }$ and ${\beta }_j$are taken. Afterwards, draws of $\mathrm{\Omega }$ are taken from the Wishart distribution conditional on the drawn value of $\mu$ and the initial value of ${\beta }_j$. Afterwards, the Metropolis Hastings algorithm is used to take draws from the joint posterior of ${\beta }_j$ (i.e., ${\beta }_j{\mathrm{\forall }}_j$) conditional on the sequence of draws of the $\mu$ and $\mathrm{\Omega }$. The process is repeated until convergence is attained; enabling draws to be taken from the posterior density. The procedure involved a “burn-in” stage to ignore iterations that occur before convergence. Once convergence is reached, a number of draws from the posterior mass are used to compute the means such that only every kth iteration is retained. This is done to ensure that correlation across the Markov Chain is reduced (Kass et al., 1998; Ragier et al., 2009; Train, 2003).

2.5. Description of Variables in the Estimated Model

2.5.1. Dependent Variable

The dependent or response variable in the probit model is the farmers’ decision to adopt new drought tolerant maize variety. The variable is dichotomous and thus coded in the model as y = 1 if the farmer decides to adopt the new drought tolerant variety and y = 0 if otherwise. y = 1 was represented by the notation “$y\ast >0$” in the empirical latent specification of the probit regression model as used in Equations 4(4) $Prob\left(y\ast >0\right)=Prob\left(y=1\right)=\Phi \left[{\beta }_{0j}+{\beta }_{1j}S{e}_j+{\beta }_{2j}A{g}_j+{\beta }_{3j}E{d}_j+{\beta }_{4j}H{s}_j+{\beta }_{5j}F{e}_j+{\beta }_{6j}I{n}_j+{\beta }_{7j}C{r}_j+{\beta }_{8j}E{x}_j+{\beta }_{9j}AT{R}_j+{\beta }_{10j}PP{D}_j\right]$(4) . “$y\ast >0$” means adopt the new drought tolerant variety.

2.5.2. Explanatory variables

These are predicting variables that are hypothesized to have a potential effect in explaining the probability of farmers’ decisions to adopt drought tolerant maize variety to adapt to climate change. These independent variables as specified in the model included: Sex, Age, Education, Household size, Farm experience, Income, Credit, Agricultural Extension, Attitudes Towards Risk, and Perception of the probability of drought occurrence. These variables are assumed to either positively or negatively influence farmers’ climate change adaptation decisions, and their choice is consistent with agricultural technology adoption literature in general (see, for example, Akudugu et al., 2012; Asfaw et al., 2016; Engle-Warnick et al., 2006; Liu, 2013; Rogers, 1983) and literature on the determinants of farmers’ adaptation to climate change in specific cases (see, for example, Apata et al., 2009; Deressa et al., 2009; Fosu-Mensah et al., 2012; Gbetibouo, 2009; Hassan & Nhemachena, 2008; Kurukulasuriya & Mendelsohn, 2006; Maddison, 2006). The study aimed at testing the hypotheses of the effects of farmers’ attitudes towards risk and their perceptions of the probabilities of drought occurrence on the adaptation to drought. It is assumed that these variables have significant effect in explaining the farmers’ adoption decision to adapt to extreme drought associated with climate change. Having been revealed as other potential variables that also can explain adoption decisions as the literature suggests, the rest of the variables in the probit model were used purposely to serve as control variables. This was done to investigate if in the midst of these other potential explanatory variables’ effects, the farmers’ risk attitudes and perceptions of the probability of drought would have a significant effect in predicting the farmers’ decisions to adopt new drought tolerant maize variety to adapt to extreme drought associated with climate change. In Table 1, the details of how these variables were captured in the regression model and their a priori expectations are presented.

2.6. Factor Scores of Risk Attitudes, and PPD used in the Estimated Probit Model

The study used the individual farmers’ risk attitudes parameter estimates under cumulative prospect theory (CPT) which were estimated with a data collected through choice experiment from the sampled farmers. Here, lottery choice experiment was designed and implemented to elicit risk preferences of the sampled farmers using ball drawing randomization technique (BDRT) to obtain data (see Balkom et al 2019). Further, the choice experiment data was analysed in the framework of cumulative prospect theory (CPT) using Mixed Logit Model estimated with Bayesian analytical approach to simulate the risk attitude parameters ($\alpha ,\beta ,$λ,$\gamma ,\delta ,\text{isin},\eta$) estimates for individual sampled farmers (see Balcombe et al., 2019 for detailed description of the choice experiment data; Dadzie, 2016 for details of the risk attitudes parameter estimation procedure with mixed logit model). The seven individual risk parameter estimates simulated were then introduced/captured in the probit model estimation as latent endogenous explanatory variables which were regressed on the farmers’ decision to adopt new drought tolerant maize variety after applying principal component analysis (PCA) on them. High degree of correlation among the risk parameter estimates was noticed and therefore a decision was taken to use principal component analysis to generate factor scores whereby the seven CPT parameter estimates could characterize two components as presented in Table 2. This was done in order to reduce the possible problem of multicollinearity between the independent-explanatory variables in the probit model. Therefore, the component scores were used rather than the estimates of CPT parameters directly in the probit regression model. Even so, collinearity between CPT parameters and the other variables may be an issue. Hence, in order to produce a more robust analysis, two approaches were employed in the estimation process: first the standard Bayesian probit and completed by using the Bayesian Reversal Jump Probit. The estimates of CPT parameters that were used for the principal component analysis are $\alpha$,, $\gamma$ $\delta$ $\text{isin}$ $\eta$ and $\lambda$. The components generated were described based on the results as Risk C1, Risk C2, Risk C3, … Risk C7; the factor scores of each of the risk parameters entered the probit model as regressors under their respective components (Risk C1 – Risk C7).

The results of the principal component analysis in Table 2 suggest that 96% of the variation in the seven risk parameters can be explained by two principal components (i.e., Risk C1 and Risk C2 which are the first and second components respectively). The table gives factor loadings that each of the CPT parameters has on the two principal components, and the behaviour that the factor scores represent is summarised under the “component interpretation” section in the table. The $\alpha$ and $\beta$ parameters loadings in the first component, for instance, can be interpreted as: “an individual tends to be more risk averse in the gain domain and more risk seeking in the loss domain. It can also be said that an individual that is pessimistic in the gain domain is optimistic in the losses domain while having strong IS behaviour in the gain domain but without a particularly strong tendency for the same behaviour in the loss domain. The negative loading of $\lambda$ in both the first and second principal components suggests seemingly decreasing sensitivity to losses; accordingly, individuals with such scores will tend to overweight gains relative to losses when faced with risky choice problems involving gains and losses”. The results help to throw more light on the fact that the risk parameters that were estimated in the CPT interact so as to explain choices made by individuals. For instance, an individual who has high second factor score has less risk seeking attitude in the loss domain and would act to weight losses higher than gains. Here the decreasing value of $\lambda$ tends to counteract this effect.

Clearly, another key determinant of attitudes to adopting a drought resistant crop will be farmers’ perceived risk of drought. Accordingly, farmers were asked to make counts of the number of years they would expect to experience drought over the next 10 years based on their experience in the last decade. The farmers’ responses were then treated as outcomes from a binomial experiment from which an approximate “Perceived Probability of Drought” (PPD) was deduced through computations. The PPD was calculated as the average probability that would lead to the number of outcomes expected by farmers. The average PPD was high at around 45% indicating that many farmers believe they face potential drought year on year. Farmers’ PPD was used as an explanatory variable along with the factor loadings from the principal component analysis of the CPT parameters and other socioeconomic characteristics as covariates in the regression model estimation.

3. Results

3.1. Trends in Rainfall and Temperature from Objective Data Compared with Farmers’ Perceptions in the Study Area

Figure 2 entails summary results that compare farmers’ perceptions of rainfall and temperature with observations from the meteorological recorded data from 1980–2020. The results show that most farmers in the study perceived that there had been a decrease in rainfall over the past years of their farming experience. However, the analysis of the trend in the rainfall series indicates rather high variation; suggesting alternating periods of sharp decline and increase in rainfall amounts relative to the computed yearly average that reflect a slightly decreasing overall trend over the period under observation. The results in Figure 2 further indicate that most farmers perceived a rise in increase change in temperature over the years of their farming experience. This appears to be in line with the objective temperature data, which, while showing some fluctuations, also provides clear evidence of a rising trend over the years.

3.1.1. Probability of Occurrence of Extreme Drought and Related Weather Events in Study Area

In Table 3, results of the distribution of farmers based on their perceptions about the probability of occurrence of extreme drought and related weather events investigated in the study are presented. Farmers were asked to reflect on their experience in the past as they look into the coming decade, and indicate how many years would pass within it before drought or other related extreme event would occur; their responses were computed into probabilities (i.e., subjective or perceived probabilities of the occurrence of the drought events associated with climate change). As indicated in Table 3, the subjective probability ranges used to categorize farmers’ responses and the probability definitions presented were adopted from Hillson (2005).¹

The results in the table show that over 80% of the farmers’ perceptions of the probabilities of the occurrence of all the adverse climatic events presented to them are in the category of 26–50% and > 50% (i.e., high and very high probabilities respectively). The mean probabilities computed from the sample data for the occurrence of the extreme drought events are in the range of 41.74% to 43.44%, all of which are consistent with the defined probability category of 26–50%. This means that on average, farmers in the study attributed “high probability” to the occurrence of extreme drought and related weather events associated with climate change in the study area.

Figure 2. Farmers’ perceptions of climate change compared with trend in objective data.

3.1.2. Impacts of Extreme drought and related weather events in the study area

With the understanding that climate change extreme events may negatively affect agricultural production processes and consequently yields and livelihoods, farmers in the study were asked to indicate their perceptions on the impacts of EWEACCs on their production and livelihoods using a 5-point perception scale. The “impact” referred to here implies the extent of damage on production processes and farmers’ livelihoods. The 5-point perception scale ranges from “1 – very low impact” to “5 – very high impact”. The farmers’ perceptions of each of the adverse climatic events were sought exclusively on the principle of “all other things being equal”. The results in Table 4 depict that over two-thirds (i.e. 73.2% and 77.3% respectively) of the farmers perceived that drought and long periods of intense heat have had “high or very high impact” on their production and livelihoods. The results also indicate more than half of the farmers perceived that unusually erratic rainfall has at least a “high impact” on their production.

In Table 4, the last but one column to the right comprises mean impact values computed over the sample; these interpretations and definitions represent the average of farmers in the study. Below the means are the standard deviations in parentheses that determine the dispersions in the farmers’ perceptions of the impacts of the extreme drought events around the mean: the higher the standard deviations, the wider the variations in the farmers’ perceptions of the impacts of the extreme drought and related weather events associated with climate change. The results of the mean values indicate that long periods of intense heat, drought, and unusually erratic rainfall all have computed means (i.e., 4.20, 4.03, and 3.59 respectively) which, on approximation, are consistent with 3.5–4.4 range interpretation of elicited farmers’ perceptions of the impacts of extreme events associated with climate change.

3.2. Empirical investigation into Climate change adaptation strategies (CCAS) in Study area

In response to the devastating effects of extreme weather events associated with climate change, farmers in Ghana, like those in other vulnerable countries, are expected to implement measures to adapt to climate change. Depending on the adaptation goal (i.e., short or long term), scope and scale of adaptation to climate change may vary based on the decision maker. This may range from use of coping measures—short term responses to deal with projected impacts and return to status quo, more substantial adjustments—changes to some aspects of the system without complete transformation, and system transformation driven adaptation decisions (Dovers, 2009; Ekstrom et al., 2010; Moser & Ekstrom, 2010). Further, the decision maker builds capacity to detect and redefine the climate problem, develop options and select what is desired to implement leading to adaptation responses decisions being purposeful or reactive (Ishaya & Abaje, 2008; Moser & Ekstrom, 2010). Literature has revealed an extensive number of adaptation strategies (see, for example, Hazell et al., 2010; IPCC, 2007b; Lybbert & Sumner, 2010; Masters et al., 2010) whose implementation (whether planned or reactive) is considered to have the potential to mitigate climate change impacts on African agriculture. In the study, a list of some of these adaptation strategies was presented in the interviews with the farmers to investigate how popular such strategies were with the farmers in terms of awareness, usage and perceived riskiness. Table 4 presents results from the interviews on farmers’ knowledge, farmers’ usage, and their perceptions of the riskiness of the selected adaptation strategies recommended in the literature. These adaptation strategies are special new varieties that have resistant qualities (such as drought tolerant maize variety as modeled in the study context), irrigation and drainage systems, mixed cropping, mixed farming, crop rotation, shifting cultivation, crop insurance, adjustment in planting and/or harvesting times, soil conservation techniques, and integrated pest management (IPM).

The results in Table 5 indicate that, with the exception of crop insurance, over three-quarters (i.e., between about 77% and 99%) of the farmers reported they knew about all the adaptation strategies presented to them. Even in the case of crop insurance, the result shows over half (i.e., about 57%) of the farmers indicated awareness of it. However, when the farmers who had reported knowledge of the strategies were asked if they knew the strategies could be used as climate change adaptation strategies, the responses obtained dropped in numbers to between 25% and 77%. This shows that a significant number of the farmers did not know how they could use most of the strategies presented to them to adapt to climate change, even though they were aware of such agricultural techniques. Unsurprisingly, where usage is concerned, results also show a drop from the percentages of the farmers who had reported having general awareness of the strategies to claim that they had used them before as CCAS (for instance, 94% dropped to 54.3% in the case of irrigation and drainage, 91.4% dropped to 53.3% in the case of mixed farming, and 83% dropped to 54.4% in the case of adjustment in planting and/or harvest times). In the case of special new varieties and mixed cropping, the drop in responses was relatively quite low (i.e., 99% dropped to 86.5%, and 99.2% dropped to 92.9% respectively); whereas none of the farmers reported having used crop insurance before as a strategy to adapt to climate change. Another interesting observation worth making from the results presented in Table 5 is that most of the farmers (over 80%) who reported awareness of the listed strategies also perceived most of the strategies presented to them as “safe” to implement. However, about 69% perceived usage of special new varieties as “risky”; Another exception can also be made of the use of Integrated Pest Management (IPM), which was also perceived as “risky” by most farmers (i.e., about 78%).

3.3. Farmers’ Decision to Adopt Drought Tolerant Varieties to Adapt to Extreme Drought

The Ghana national action programme to combat drought and desertification which is ongoing (EPA, 2003) motivates the current study interest in the adaptation with a drought tolerant variety in Ghanaian context of climate change adaptation to drought effects in the study area. The adaptation action programme aims at sustaining high agricultural production and ensure food security and enhanced livelihoods in the changing climate. Core to the programme’s activities include: (i) strengthening research institutions in the development of drought tolerant crops varieties, (ii) promoting the dissemination of drought tolerant crops, and (iii) strengthening the extension services to effectively promote drought tolerant crop varieties. Accordingly, to deal with problem of low productivity due to frequent drought stress adverse effects in largely rain-fed farming systems in Ghana, issues of drought management using drought tolerant crop varieties have become paramount in the adaptation to climate change in Ghana. The paper supports the national agenda to promote use of drought tolerant crops; therefore, the need to design and implement an adaptation to drought experiment to elicit the sampled farmers’ decision to adopt drought tolerant maize variety. Here, the farmers interviewed were presented with an experimental question modelled about drought tolerant variety (a stated preference experiment). The objective was to investigate farmers’ decisions to adopt new drought tolerant maize crop and subsequently find out whether their risk preferences and perceived probability of drought (PPD) explain their adoption decision. To address this issue, the study employed the following experimental question within the survey instrument:

“The weather will be changing in the next coming years by becoming warmer and with more variable rainfall. The Ministry of Food and Agriculture (MoFA) through its extension directorate is promoting the use of new variety of maize that you have been growing over the years as a farmer. The ministry’s primary reason for this promotion is that the variety has been developed through research to be resistant to drought in a likely event that it may occur in the upcoming year’s growing season. This was because of the unexpected changes in climatic conditions resulting in drought in the past. As a farmer in the area, you are at liberty to choose to adopt the new variety or stick to the traditional one you are used to. This adaptation technology policy is not compulsory for farmers to adopt. Please indicate your choice between the following options: A (new drought tolerant variety) or B (traditional/old variety)”.

The farmers’ responses (from the stated preference experiment) to adopt the drought tolerant variety as a measure to adapt to climate change extreme events leading to drought have been presented in Figure 3 below.

Figure 3. Maize farmers’ decision to adopt drought tolerant variety or otherwise.

The results show that about 82% of the farmers reported they would adopt the drought tolerant variety whereas the remaining few (about 18%) reported they would retain their old maize variety.

3.4. Analysis of Farmers’ Drought Tolerant Crop Adoption Scenario: A Conceptual Decision Tree Model

Farmers in the study were faced with the decision to adopt a new drought tolerant variety or otherwise in the upcoming year with chances that extreme drought might occur. Accordingly, the farmers were to choose between new drought tolerant variety and their old crop variety. It can therefore be deduced that: the new variety has come to mitigate the extreme drought effect on production. Its performance is, however, based on the chances of success (P) or failure (1-P) of the embedded tolerant property to withstand drought. On the other hand, farmers are certain about the outcome of the performance of their old variety under the state of extreme drought occurrence or otherwise. Farmers were asked to act on the assumption that the new variety would cost the same as to invest in the farmers’ old variety; “all other things being equal”. Figure 4 4 is a framework of a decision tree that analyses the drought-tolerant crop adoption problem with which the farmers were faced with. As the figure suggests, readers must note that, the associated payoffs or outcomes in terms of yield value, more or less depend on the success or failure of the new variety even in the bad state of the weather. Regarding the performance of the old variety, it is assumed that farmers might be certain about the possible yield values they stand to gain or lose at the good or bad state of weather respectively. Accordingly, the study considered non-adoption of the new drought tolerant variety as safe option to the farmers; on the other hand, the adoption of the new variety was considered a risky option since to gain or lose would depend on the success or failure of the variety’s performance coupled with the state of weather.

Assuming the varieties are mixed prospect pairs, the following analysis can be made from the figure:

1. If a farmer decides to adopt the drought tolerant variety and there is success at probability (P), there will be gain of yield value (X) if drought occurs or otherwise.

2. If a farmer decides to adopt the drought tolerant variety and there is failure at probability (1-P), there will be loss of yield value (−X) if drought occurs or otherwise.

3. If a farmer decides not to adopt the new variety and drought occurs at probability (q), there will be loss of yield value (−X) that might be certain.

4. If a farmer decides not to adopt the new variety and drought did not occur at probability (1-q), there will be gain of yield value (X) that might be certain.

Figure 4. Framework of decision tree to analyse drought tolerant crop adoption problem (author’s construct).

Where: X defines gain (loss) of yield value, P defines probability of success of the new variety, and q defines probability of drought

The yield values (or utilities) with respect to the payoffs associated with farmers’ decision scenarios outlined above can therefore be interpreted as follows: Both gains recorded under adoption (success) as shown in the figure are higher in value than the sure gain recorded under non-adoption in the good state of nature: i.e. $X_{ng}<X_{ab}<X_{ag}$. Where $X_{ng}$ defines gain of yield under non-adoption in good weather (drought not occurred), $X_{ab}$ defines gain of yield under adoption in bad state of the weather (drought occurred), and $X_{ag}$ defines gain of yield under adoption in good weather (drought not occurred). Furthermore, both losses recorded under adoption (failure) as in Figure 4 have higher disvalue than the loss recorded under non-adoption if the bad weather occurred: i.e., $-X_{nb}>-X_{ag}>-X_{ab}$. Where $-X_{nb}$ defines loss of yield under non-adoption in bad weather (drought occurred), $-X_{ab}$ defines loss of yield under adoption in bad weather (drought occurred), and $-X_{ag}$ defines loss of yield under adoption in good weather (drought not occurred).

3.5. Regression of Farmers’ Decision to Adopt a Drought Tolerant Variety on Their Risk Attitudes and other Explanatory Variables

It is assumed that farmers who took decisions in same direction (e.g., farmers who decided to adopt the new drought tolerant variety) might have some similar characteristics that could help provide better explanations for their decisions. These common characteristics could be socio-economic (income, farm experience, credit), demographic (age, sex, family size), behavioural and perceptional (attitudes towards risk, perceived probability of drought occurrence) or institutional (agricultural extension services, education). The study, therefore, investigated the empirical underpinning factors that explained farmers’ reported decision to adopt or not adopt the new drought tolerant variety presented to them in the stated preference experiment. Here, the relationship between farmers’ decisions to adopt or not adopt (i.e., the dependent variable) and the specified explanatory variables (mainly demographic, socio-economic, behavioural or institutional) were modelled using the two-step Bayesian probit regression estimation process (i.e., probit-standard regression and completed with the reversal jump probit); the results of which are presented in Table 6. The study employed the two-step Bayesian probit estimation approach in order to produce a more robust analysis that could help addressed issues of collinearity between CPT parameters and other variables in the regression model. The reversal jump probit delivers a posterior probability that a given regressor has a non-zero coefficient. Tiffin and Balcombe (2011) revealed that unlike a classical p-value, the reversal jump probit delivers a measure of the probability that each variable explains the dependent variable (which, in the current study, is the farmers’ decisions to adopt the drought tolerant variety). However, for readers of a classical disposition, pseudo p-values from the probit-standard which are very similar to what would be produced by maximum likelihood estimation are also provided. For the purpose here, all explanatory variables are scaled to lie between zero and 1, so that the impact of the priors is even across the variables. Both sets of results below use normal priors for the coefficients with a mean of zero and a standard deviation of 3, but similar results are obtained if the standard deviation is halved or doubled.

The asterisk (*) indicated against the mean values of some variables in Table 6 implies that at least 90% of the posterior mass in their parameter distribution is negative or positive in explaining adoption. Effectively, one can discuss their effect based on the result with some 90% confidence of being accurate (while acknowledging that in 1 out of 10 cases, their predicting effect might not be accurate). Accordingly, these variables are considered to have significant effects on the adoption decision of farmers; however, in the study context, their effects are described as “strong”. The concept of statistical significance is mostly verified using p-values, especially in classical data analysis. In the current study, the strengths of the explanatory variables’ effects are judged by a system based on the pseudo p-values and Prob. Inc. The effect of those variables with pseudo p-values of ≤ 0.1 and Prob. Inc > 50% are considered “significant”; these variables have an asterisk (*) in the tables of results. The discussion of the probit regression results in terms of variables that best explain or predict the probability that a farmer would adopt a drought tolerant variety, given the data, therefore focused on these variables with some 90% confidence. It is worthy of note that inference in this circumstance may be difficult also due to the nature of collinearity between variables.

The study results in Table 5 indicate that Risk C1 (first component of risk scores), PPD (perceived probability of drought), income, accessibility of credit and agricultural extension contact have strong influence on the farmers’ decision to adopt a drought tolerant variety as a climate response measure or technology. The results also indicate that the likelihood that a farmer will adopt the new drought tolerant variety is positively influenced by Risk C1 (i.e., $\theta =1.70$), PPD (i.e., $\theta =11.70$), income ($\theta =1.98$), credit access (i.e., $\theta =0.95$) and agricultural extension contacts (i.e., $\theta =2.\mathrm{o}2$). This suggests that, given the data, a farmer in the study area is more likely to adopt the new drought tolerant variety, provided that close attention is paid to addressing issues pertaining to the aforementioned variables based on the way these variables affect farmers’ adoption decisions.

The results reported in Table 6 suggest evidence that the first component from the risk scores strongly influences farmers’ drought tolerant adoption decision. Risk C1 is the only component that has more than 50% probability (i.e., about 75%) that it has a non-zero coefficient suggesting its posterior median value is non-zero; it also shows significance at the 10% level which can only be said to be a moderate evident. All the other risk components do not have statistical evidence of influence on the farmers’ potential decision to adopt the drought tolerant variety. Unlike prior expectation that the risk components 3–7 might be insignificant, the fact that the second component also seems to have little explanatory effect perhaps give indication that the risk attitudes might be playing a secondary role in this stated preference adoption decision. In the table of results, the positive coefficient of the Risk C1 can be explained as follows: those farmers who are more risk seeking in the loss domain, and more risk averse in the gain domain, display relatively high IS behaviour, and are pessimistic about probable gains but optimistic about probable losses are more likely to say “yes” to the adoption of the new drought tolerant variety. With regard to the loss sensitivity’s (λ) effect, this implies that farmers who have less sensitivity to losses are more likely to adopt the drought tolerant variety. However, the role of the λ was not found to be very pivotal. This is because it is principally loaded in the second component that does not show evidence of a statistically significant effect; its loading in the first component indicating a significant effect is also quite low.

4. Discussion

4.1. Trends in Rainfall and Temperature in the study area

In line with the findings of Yaro (2013), who also studied in a Ghanaian context, and Akponikpe et al (2010) who worked in countries including Ghana, the results in the study imply that perception of a decrease in rainfall over the years, as mentioned by many farmers, is somewhat supported by the evidence in the meteorology data. Similarly, in studies elsewhere in Africa, farmers’ perceptions of decreasing change in rainfall were supported by evidence found in objective data. Gbetibouo (2009), for example, observed high variability in rainfall, with a decline only in the last few years that agreed with farmers’ perceptions in the Limpopo area of South Africa. The increasing temperature change over the years as perceived by the farmers which is consistent with the trend in objective data, buttresses the findings of most previous studies in an African context (see, for example, Akponikpe et al., 2010; Fatuase & Ajibefun, 2013; Gbetibouo, 2009; Osbahr et al., 2011; Tambo & Abdoulaye, 2013; Yaro, 2013). For instance, the study agrees with the argument of Osbahr et al. (2011), suggesting that the effects of rising temperature might influence decrease in rainfall. They argued that “temperature increases will result in higher evapotranspiration and greater demand on available water, faster development of water stress during dry spells, increase severity of pests and disease” (Osbahr et al., 2011: p 309).

4.1.1. Drought occurrence and impact in the study area

It can be deduced from these results that most farmers expect extreme drought events to be more frequent, basing their notions on the observations and experiences from the past years they have been farming; their opinion appears to be supported by some climate projections. Dyoulgerove et al. (2011), for example, discussed future climate trends for Ghana and noted that the mean annual temperature was projected to increase by 1–3°C by the 2060s. Accordingly, they reported that there would be increases in the frequency of intense heat periods, as suggested by all projections. They also reported a projected rainfall decrease of about 1.1%, and suggested increased incidences of droughts. The farmers’ perceptions of the higher likelihood of the incidences of drought events also conform to the findings of Enete et al. (2011), who studied the pattern of climate change effects on agriculture in Nigeria, and revealed that the majority of the farmers interviewed were of the opinion that uncertainties in the onset of the farming season and extreme weather events had been increasing. As also the case in the current study, the adverse climatic events identified in their study included erratic rainfall, long dry seasons and intense heat, whose effects were perceived to increase the incidences of pests, diseases, weeds, and signals of land degradation such as declining soil fertility and drying up of streams and rivers. The results of the impact of drought imply that, on average, farmers in the study perceive that long period of intense heat, drought, and unusually erratic rainfall respectively have highly impacted on their production and livelihoods. It is interesting to note that drought, unusually erratic rainfall and long periods of intense heat, which farmers perceive have high impacts on production, are all adverse climatic events that can be linked to periods of dry spells, resulting in a reduction of the water supply for crop growth.

4.1.2. Climate change adaptation strategies (CCAS) in Study area

The results about climate change adaptation strategies (CCAS) by the farmers in study area suggest that, to some of the farmers, the use of some of these strategies might be for intent other than to mitigate effects of extreme weather events associated with climate change. Of course, every farmer’s primary aim in their farming business would be to maximize output/yield most of the times (Juma et al., 2009); the way to achieve that might involve a combination of different intents. It could also be noted that, according to the farmers’ report, every strategy had been used before, with the exception of crop insurance. The percentages of farmers who reported that they had used some of the strategies before appeared to be consistent with the proportions who also claimed knowledge of the strategies’ potential as climate change adaptation strategies. This means that an effort is required to sensitize farmers more about the potential of the strategies to reduce vulnerability to impact of climate change. The farmers’ high general awareness of the strategies suggests that most strategies are common and may be available to farmers in Ghana; their strategic use as CCAS, however, might be an issue requiring a policy response. It has been argued that impacts of climate change are the greatest threat to African agriculture and thus implementation of effective adaptation strategies at farm level would be critical to reduce vulnerability and impacts (Reilly & Schimmelpfennig, 1999; Smit & Skinner, 2002; Smit et al., 2000; Tambang, 2009). This underscores the importance of farmers’ tactical implementation of climate change adaptation strategies presented in the study, in response to the potential negative effects of the extreme climatic events.

4.1.3. Discussion of farmers’ drought tolerant maize variety adoption

The results of drought tolerant maize variety adoption revealed that majority of the farmers would choose the new maize variety. The explanations obtained from few farmers who would reject the new drought tolerant variety suggest that, notwithstanding the touted benefit of the new variety in case of drought, the farmers also had some fear that the new variety might fail and consequently they would risk losing their investment in such an unfamiliar new variety. They might think this would end up making them worse off than not adopting the new variety. According to those farmers, they at least were used to their old and traditional varieties and would rather be comfortable with whatever outcome they got from the old variety, which they regarded with some amount of trust and/or certainty. Some were also of the view that they did not have a strong belief that extreme drought would occur and so did not see the danger for them in refusing to adopt the new drought tolerant variety, especially since the decision to adopt might require relatively more resource commitments.

From the Prospect Theory perspective, responses depend on the farmers’ reference point along with the perceived risk associated with adoption. It is worthy of note that many farmers would perceive the new crop technology as a mixed prospect (i.e., although the new variety might give improved yields if there is a drought, there is also a possibility to perceive a probability that it may either fail or give lower yields under other conditions). Obviously, there might be other devastating climatic events that confront farmers apart from drought. Farmers are, therefore, likely to doubt whether the new variety is necessarily as good as the existing one under non-drought conditions. This subjectively would project the adoption of the new variety as a “risky climate response measure”. It can therefore be asserted that the majority of the farmers who indicated they would choose to adopt the new variety might have risk attitudes that can be characterized to be consistent with “putting more weight on gains than losses”; including risk and/or loss seeking (less sensitivity to losses) and a high degree of “optimism”.

Further, from the framework analysing farmers’ drought tolerant crop adoption scenario, it can be observed that a risk or loss averse person might avoid taking chances associated with the new variety and thus would choose the safe option (old variety) since that gives him/her certain positive outcomes under occurrence of drought or otherwise. Also, a farmer who over weighs probabilities of large gains associated with success (of the new variety) and under weighs probability of large losses associated with failure (of the new variety) would take chances (gamble) to adopt the new variety. Furthermore, a loss averse person who perceives low probability of drought occurrence would choose a certain gain from the old variety over higher possible gain that the new variety would give. All other things being equal, it is the insight from this analysis that theoretically support the formulation of the hypothesis addressed in the adoption probit model.

4.1.4. Effect of risk attitudes and PPD on farmers’ decision to adopt drought tolerant maize variety

It is worthy of note from the results in Table 5 that some of the risk parameters’ effect are contrary to prior expectations. The fact that those that are more risk averse in the positive domain, and also that those that are more “pessimistic” in the gain domain are more likely to adopt, is contrary to the prior expectation. This is because while it is expected that a farmer who under weighs large losses (optimistic) associated with failure or impact of other extreme events will adopt the “risky climate response drought tolerant variety”, the results also suggest that a farmer who under weighs large gains (pessimistic) associated with the improved yield of the variety under drought will adopt; which is not expected. To be able to make a coherent meaning from such outcomes, given the fact that the majority of the farmers in the study sample responded “yes” to the adoption of the new drought tolerant variety, it can be assumed that most farmers might have reflected their risk tendency in their drought frequency responses. Thus gain-pessimists might have higher PPDs to accept the new variety. The very dominant role of PPD combined with the fact that the risk parameters are correlated with this variable, would suggest that common factors may jointly determine farmers’ adoption decisions.

While farmers see drought tolerant varieties as a mitigation of one of the risks, they may believe that other downside risks are potential risk sources. This can inform farmers’ perceptions of the adoption of the new variety as “more risky” in terms of loss. Accordingly, though not very pivotal, the compensating role λ (loss sensitivity attitude) plays in the adoption models makes the discussion of its effect along with risk seeking attitude worth considering. In the paper “Prospect theory, mitigation and adaptation to climate change”, it has been argued that when individuals are confronted with a decision problem in the climate change context, such as a choice between more and less risky climate response measures, risk/loss seeking individuals who perceived the climate impact as moderately probable are likely to prefer the riskier response measure that was considered more effective (Osberghaus, 2013). This theoretical argument sheds light on the negative effect of the risk and/or loss seeking (loss sensitivity) variables on the probability of farmers adopting the drought tolerant variety in anticipation of extreme drought, as the result indicates. In the study’s context, the new drought tolerant variety would be considered as a risky climate response measure, since farmers might be uncertain about how it would actually perform in their own changing climate. As a result, the decision to commit resources to adopt such a new response measure would be taken under risk. Thus, in agreement with Osberghaus (2013), it would be farmers whose risk behaviour was consistent with risk seeking and/or less sensitivity to losses who would be likely to adopt the drought tolerant variety when it was recommended to them. The policy implication of this is that much needs to be done to account for risk behaviour regarding risky climate response measures, as in the case of the new drought tolerant variety to be promoted to farmers in the study area as adaptations to climate change.

The most important variable by far is the PPD, which is expected, along with access to credit income, and agricultural extension contacts. Farmers’ adaptation to climate change can be influenced by their perceived vulnerability to the impact of extreme events associated with climate change. This logic is reflected in the argument that personal perceived risk (PPR) of extreme climatic events positively influences motivation to adapt to climate change; this suggests that “higher PPR leads to a higher motivation to adapt” (Osberghaus et al., 2010: p14). Accordingly, in this study, the perceived impact of climate change, and farmers’ perceived vulnerability to this impact, is linked to farmers’ perceptions of the probability of the occurrence of extreme climatic events like drought. By implication, higher perceived probability of the occurrence of drought leads to higher perceived impact of drought as well as higher perceived vulnerability (not an empirically investigated relationship in the study, though). This helps to explain why the probit regression result shows that a higher probability of drought (PPD), as perceived by farmers, increases the likelihood of farmers’ decision to adopt a drought tolerant variety in order to adapt to climate change. We find the result of the effect of perceived probability of drought (PPD) on the adoption of a drought tolerant variety to adapt to drought associated with climate change to corroborate the assertion that climate is a key driver behind the adaptation strategies farmers employ and therefore adaptation would deliver positive payoff (DiFalco & Veronesi, 2013); thereby increases food productivity (DiFalco et al., 2011).

From the results in Table 5, the positive effect of the accessibility of credit also indicates that the more farmers access financial support through credit, the more likely those farmers would be to adopt the new drought tolerant variety if it were recommended to them. This is in line with the findings in most previous studies conducted in similar contexts (see, for instance, Apata et al., 2009; Deressa et al., 2010; DiFalco et al., 2011; Fosu-Mensah et al., 2010; Gbetibouo, 2009; Hassan & Nhemachena, 2008; Juana et al., 2013; Mudzonga, 2011; Tizale, 2007). These studies were conducted in various African countries, including Ghana. They found empirical evidence that access to credit is vital in enhancing African farmers’ decisions to adopt various innovative strategies. Credit support gives farmers purchasing power and can enhance their capacity to afford resources necessary for the implementation of climate change adaptation measures. This accordingly has the potential to positively influence farmers’ decisions to adopt strategies for adapting to climate change. It has also been revealed in an empirical study in Ghana (Dadzie & Acquah, 2012) that access to microcredit inversely relates to risk aversion in farmers. Thus, all other things being equal, “depriving farmers of access to microfinancial services will make them prone to being more risk averse” (Dadzie & Acquah, 2012; p34); and can consequently reduce the probability that farmers would decide to adopt risky climate response measures, regardless of how effective they might be, compared with their non-risky alternatives as in the context of the study. Consistent with Liu (2008), the effect of income on adoption of technology can be explained by its association with greater access to resources. All other things being equal, increase in income of farm households would encourage savings, which eventually enhance the investment capacity of the farmers. This gives farmers greater access to resources that are required to implement measures to adapt to climate change. Thus, as the study result suggests, the greater the income of farm households, the higher the probability that farmers will adopt the climate change adaptation measure (i.e., drought tolerant variety) recommended to them. The positive effect of the agricultural extension services also corroborates extensive findings in the literature which suggest that farmers who have access to extension services are more likely to adopt new strategies for adapting to climate change (see for example, Gbetibouo, 2009; Maddison, 2006; Fosu-Mensah et al., 2010; Apata et al., 2009, Deressa et al 2010; Hassan & Nhemachena, 2008; Juana et al., 2013; DiFalco et al., 2011). As they are an important source of information about agricultural management practices and climate change for farmers in the study, it is not surprising that agricultural extension services positively influence farmers’ decisions to adopt the drought tolerant variety strategy presented to them. Information about new adaptation measures channelled through Agricultural Extension Agents (AEAs) can help to reduce uncertainty about the performance of the innovative strategies. This is because the AEAs are usually on the ground with the farmers to help them to solve their numerous agronomic challenges, and as a result, might build trust and credibility in the face of their client farmers. When uncertainty about adaptation measures’ performance is reduced, it helps to influence farmers’ intuitive evaluation from a merely subjective to an objective view and can, effectively, facilitate adoption.

5. Conclusion and Policy Implication

When some agricultural technologies selected from the literature were presented to farmers, it was noted that the majority of the farmers were aware of the strategies. However, a significant number of these farmers did not indicate they knew most of the technologies had potential to be tapped and strategically used as climate change adaptation strategies (CCAS). It was therefore not surprising that only smaller percentages of the farmers reported that they had used the technologies before as strategies to adapt to climate change. The agricultural practices (technologies) that were presented to farmers to assess their knowledge, usage and perceptions of riskiness were special new varieties that have resistant qualities, irrigation and drainage, mixed cropping, mixed farming, crop rotation, shifting cultivation, crop insurance, adjustment in planting and/or harvesting times, soil conservation techniques, and integrated pest management (IPM). It was further interesting to find out, in the case of crop insurance, that though the majority reported it as safe, little over half of the farmers claimed knowledge of crop insurance as CCAS, and none indicated previous usage. Majority of maize farmers were willing to adopt drought-tolerant variety instead of their traditional ones they grew in the past. It also be concluded that farmers who overweigh probabilities of gains associated with success of the new variety, and underweight probability of losses associated with failure of the new variety would adopt the new variety. The analysis of the determinants of the adoption decision concludes that farmers with risk and/or loss seeking (less sensitivity to loss) attitudes who perceive as highly probable the occurrence of extreme climatic events, will be more likely to adopt even “risky climate response strategies”, like cultivating a new drought tolerant variety, when recommended to them. It can further be concluded from the analysis that a farmer who has access to credit and or agricultural extension services is more likely to adopt innovative technologies to adapt to climate change extreme events.

Dispassionate assessment of farmers’ capacity to adopt farm-level measures to adapt to climate change reveals the need for a policy that will increase it still farther. It is suggested that work must be done to improve farmers’ knowledge about climate change adaptation strategies that they can implement at farm level. This can be done through increasing sensitization education about climate change adaptation strategies to highlight relevance, effectiveness and benefits to farmers. We emphasize that enhancing farmers’ awareness of the potential of using most of the existing and recommended measures to adapt to climate change can be vitally important in facilitating farm-level adaptation to climate change. In the special case of the need for crop insurance to help farmers to adapt to weather shocks, there is the need to strengthen policies to encourage the development of crop insurance products by the players in the insurance market as well as increasing awareness of the availability of insurance to farmers, so that they can use it as a shock absorbing mechanism against the impacts of extreme weather events associated with climate change; it also serves as protection, and as a means to boost farmers’ confidence in investing resources in their food crop production. For instance, offering insurance as a cover against the risk of failure if new drought tolerant varieties are used can help to enhance technology dissemination and adoption. This also applies to other strategies. As part of the policy agenda to build farmers’ adaptation capacities, farmers can also be encouraged to strengthen networking and information sharing in their farm neighbourhoods among themselves by forming effective working groups and Farmer-Based Organisations (FBOs) through which they could easily seek help for themselves and also be contacted by others in need of assistance. Furthermore, agricultural extension services to farmers must be increased. It is recommended that agricultural extension services must incorporate more climate change information as well as information about agricultural strategies to adapt to climate change. This can be made possible by establishing and strengthening linkages between the agricultural extension directorate of MoFA, GMS and research institutions.

The potential influence of farmers’ perceptions of the riskiness of new adaptation strategies which might be the function of their decision behaviour under risk (risk and loss attitudes) is a key predicting factor to be considered and accounted for, in order to pave the way for the successful introduction of new strategies to farmers for adoption against climate change extreme events. The study recommends the need to facilitate the development and diffusion of new effective strategies (such as new drought tolerant varieties in study context), so that farmers can adopt them to help reduce their vulnerability to the negative impact of climate change. It is further recommended that agricultural policies that seek to promote new strategies, such as the development of new drought tolerant varieties to help farmers to adapt to extreme events associated with climate change, may need to consider incorporating feasibility studies on the personal circumstances of the farmer-target group (particularly, farmers’ decision behaviour under risk as well as their perceptions of the riskiness of the strategies). This will help to establish the risk and/or loss attitudes of the target farmers and also, whether they have “optimistic” or “pessimistic” attitudes towards the new technologies. It is based on this that the content of the promotional message can be packaged to well-informed farmers about the new technologies in a manner that might prompt the farmers and facilitate their adoption decisions. Also, having established the risk behaviour of the target farmers, it might help to plan the appropriate extension communication approach that would help to facilitate farmers’ adoption of new technology, and other strategies, as part of a planned course of action necessary to adapt to climate change extreme events. Another essential policy that might expedite the diffusion of new technology to help farmers adapt to climate change would be the improvements of financial services to farmers, especially the provision of agricultural loans. In Ghana, financial institutions such as Agricultural Development Banks (ADB) as well as rural banks can make it a policy to increase the provision of credit assistance to give farmers the necessary financial capability to invest in practices recommended as effective adaptations to climate change. It is recommended that agricultural credit facilities should be affordable and secured by synchronizing the credit facilities with insurance. For sustainability of the credit facilities, this could protect investment in food crop farming against the risk of default due to crop failure that might result from weather shocks.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the University of Cape Coast.

Notes on contributors

Samuel Kwesi Ndzebah Dadzie

Samuel K. N. Dadzie is an Associate Professor at the Department of Agricultural Economics and Extension in the School of Agriculture, University of Cape Coast, Ghana. He has PhD in Agricultural and Food Economics from the University of Reading in the United Kingdom. As an applied agricultural economist, he has special interest in climate change adaptation and resilience, agricultural risks, choice modelling, and value chain and agribusiness development. He has worked on number of projects involving smallholder farmers’ adaptation to climate change, climate risk analysis, and impacts on food system sustainability. The current paper is one of the papers published from the projects he had worked on in the area of food crop farmers’ adaptation to extreme weather events associated with climate change in the Central Region of Ghana.

Notes

1. Hillson (2005) worked on “Describing probability: the limitations of the natural language” and presented an exhibit of probability definitions as preferred solutions recommended by several risk standards.