Invited paper presented at the 6th African

Conference of Agricultural Economists, September 23-26, 2019, Abuja, Nigeria

Copyright 2019 by [authors]. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.

Production risk and risk preference among small-scale pig enterprises in Southwestern

By

Mbah Leslie Tembei Centre for Independent Development Research, P.O. Box 58 , SWR, Cameroon

Ernest L. Molua, PhD Department of Agricultural Economics and Agribusiness Faculty of Agriculture, University of Buea, Cameroon, P.O. Box 63 Buea, SWR, Cameroon.

Mr Ajapnwa Akamin Department of Agricultural Economics and Agribusiness Faculty of Agriculture & Veterinary Medicine University of Buea, SWR, Cameroon *Corresponding author: Email: [email protected]

1

Production risk and risk preference among small-scale pig enterprises in Southwestern Cameroon

Abstract

This article analyses average pig production and output variability in southwestern Cameroon. After testing for the presence of production risk, the feasible generalised least squares (FGLS) technique is used to estimate the mean and variance functions and identify sources of output variability in pig production. Both descriptive and econometric results converge on the fact that pig producers become more risk averse when exposed to production risk. Income diversification and variable input use are identified as the main factors influencing expected output. Meanwhile, of all factors posited to cause variability in output, activity diversification alone tends to increase output risk. Keywords: Pig production, risk, risk preferences, Cameroon

Introduction The agriculture sector is the key sector of Cameroon’s economy, employing more of the active population than any other sector and contributing very significantly to the country’s gross domestic product (76.38% in 2017) (MINADER 2018). Over the years, the country has witnessed an upward trend in agricultural output (albeit a less than proportionate increase in productivity) which can be attributed to an expansion of the size of cultivated area rather than to an increase in productivity and efficiency in the agricultural sector (Dewbre & Borot de Batisti 2008). Cameroon’s agribusiness sector plays a vital role in the economy, not only at the national level but also at household level, as it provides a source of food, employment and livelihood.

Meat production is a particularly lucrative business in Cameroon both in terms of its contribution to the country’s GDP, as well as serving as a source of food and livelihood for many. Meat production includes ruminants, birds, as well as other livestock. Pig production contributes about 15% of the country’s total meat production, significantly lower than cattle (54%), but just slightly above sheep and goat (13%). Meanwhile, poultry and rabbits contribute 17% and 1% respectively (GESP 2011). The pig agribusiness in Cameroon is dominated by small-scale subsistent producers.

.

Figure 1. Evolution of pig production in Cameroon Source: Authors’ computation using FAOSTAT data.

2

Agribusinesses by their nature attract a lot of risk. As a result, decision-making is a complex issue for most agribusinesses (both small- and large-scale), because production is often subject to uncertainty and risk (Moschini & Hennesy 2001). Many sources of risk exist which influence agribusiness decisions, some of which include; political risk (war, political unrest, administrative bottlenecks), economic risk (financial, price, and credit risk), environmental risk (disease outbreak, climate and weather risk), amongst others. The focus of this study is on production risk only.

Uncertainty and risk are inherent features of the agricultural production process - both crop and livestock. Two main forms of these are fluctuations in output (production risk) and prices (price risk), and their combined effect significantly influences farm income. Production risk often arises because agricultural production depends on natural biotic and abiotic processes which cannot be controlled entirely by man. Various sources of production risk in agriculture and agribusiness include weather vagaries, plant and animal diseases, natural disasters, amongst others. Risk associated with crop and animal production is thus more pronounced in less developed agrarian economies as the technology available to them to curb output risk is limited compared to more industrially advanced countries (Asche & Tveteras 1999). As such agribusiness decision-making (under uncertainty) is best analysed by taking into consideration the risk preference behaviour of the latter. Conceptually speaking, agribusinesses exhibit three possible attitudes towards risk; risk- averse, risk-neutral, and risk-friendly preferences.

Materials and methods Conceptualisation of risk in agriculture Throughout this study, risk is viewed as the chance of a bad event occurring relative to the producer’s expected outcome. Risk is likely to have a potentially negative impact on the profitability of investments in the agribusiness sector. The various sources of uncertainty and risk in agriculture and agribusiness can be classified under four main categories; economic/financial, social, environmental, and political risk. Institutional risk - unpredictable changes in the provision of services from institutions that support farming – is also known to affect the functioning of agribusinesses. Such institutions can be both formal and informal and include banks, cooperatives, marketing organizations, input dealers and government extension services. Price support, subsidies, food quality regulations for export crops, rules for animal waste disposal and the level of price or income support payments are examples of decisions taken by government that can have a major impact on the farm business. Meanwhile, marketing risk refers to variations in prices beyond the control of the individual farmer. The price of farm produce is affected by its supply and demand, as well as the cost of production (Kahan 2008).

Risk preference refers to the level of tolerance agribusinesses exhibit when faced with risk. Risk aversion can thus be defined as the willingness to accept lower expected returns in a bid to reduce risk involved. Meanwhile, risk-loving agribusinesses will exhibit willingness to accept higher expected returns even if it means incurring higher risk in the process. Risk preference analysis is therefore very important for understanding why and how agribusinesses make decisions when faced with uncertain outcomes. Usually, smallholders either do not adopt or only partially adopt new technologies, even when these technologies could generate higher returns than the existing technologies. One possible explanation for this reluctance among smallholders in developing countries could be the perceived risk profile associated with these technologies (Hardaker et al. 2015). For instance, Smale et al. (1994) show that production risks lead to slow adoption of new technologies in maize production. Production decisions are thus greatly influenced by the level of risk preference of the agribusiness (risk averse, risk friendly or risk neutral).

The effects of production risk, if not properly managed, could result in misallocation of resources, low productivity, inefficiency, low investments and consequently slow rate of business growth. 3

Production risk influences agribusiness decisions in many ways. Agribusiness managers decide what inputs to use, where, when and how, depending on their level of tolerance to the associated production risks. This is particularly true for small-scale agribusinesses whose management decisions are highly influenced by exogenous processes. Such high dependence leads these agribusinesses to exhibit high aversion towards risk. In addition, micro agribusinesses usually lack adequate technology to mitigate the effects of the risks they face. As such the low investments lead to low productivity and by extension lower expected profits for the agribusiness (Cole et al, 2017).

In order to cope with production risk, farmers in developing countries adopt a range of risk- management strategies, ranging from income diversification and production strategies to common risk-sharing mechanisms based on kinship and social networks. The latter approach is more appropriate for covariate shocks as opposed to idiosyncratic shocks. Evidence suggests that in the absence of formal risk management, less risky but less profitable farming practices are adopted, resulting in lower productivity (Antonaci et al. 2014). In this paper, we focus on production risk. This refers to output uncertainty caused by the vagaries of the weather and other production-related shocks in the course of the production cycle. Production risk arises because agricultural production depends on natural and environmental processes which cannot entirely be controlled by the producer or agribusiness - weather, animal disease outbreak, political instability, and natural hazards, amongst others. It also arises from uncertainty around the use of technologies. As such, farmers decide to produce amid uncertainty about ex post production (Kahan 2008). The identification of the sources of risk is important because it helps to choose the appropriate risk management strategy: these strategies can either be ex-ante or ex-post.

Although awareness of the existence of risk is clearly important, the latter needs to be clearly identified and analysed for effective risk management to follow. Risks and the associated impacts, are best assessed by quantifying three main variables: hazard, vulnerability, and exposure. Hazard can be measured in terms of frequency of occurrence, severity of the risk and the extent of the risk (World Bank 2011). Given the highly susceptible nature of crop and livestock production in developing countries to production risk (Roll et al. 2006), and the attendant effects on food security and livelihood, it is important to understand the causes and consequences of agricultural production risk in these countries. Modelling production decisions made under such circumstances also helps to unravel information about why farmers make certain decisions when faced with risk.

Analytical framework Many approaches have been used in the literature to model production risk. One of such is the coefficient of variation, which measures randomness relative to the mean yield value (Hardaker et al. 2015). Other researchers have exploited this output variance method to determine the extent to which production risks influence production. Many attempts to explain risky decisions are however couched on the expected utility hypothesis refined by Neumann & Morgenstein (1944). According to this theory, the agribusiness decision-maker may choose between risky or uncertain prospects by comparing their expected utility values - the weighted sums obtained by adding the utility values of outcomes multiplied by their respective probabilities. Thus, agribusinesses chose activities with the highest expected utility, where utility quantifiable as satisfaction with regard to a particular production goal. In agricultural production, it may be derived from output, business profits, or in some other form. Maximizing expected utility is thus assumed to be the main goal of the agribusiness.

A common starting point in agricultural output analysis is the assumption that increasing inputs not only leads to an increase in output, but it also increases variation in output. Just and Pope (1978, 1979) introduced the parametric stochastic function which has become the most popular framework used in analysing production risk. Many studies that use this approach, such as Kumbhakar (2002), focus only on risk in agricultural production. Meanwhile, some other studies 4

ignore production risk and analyse product price risk only, while some analyse both production and price risk separately (for instance Sandmo [1971]). Production and price risk are however weakly separable in agricultural production since quite often expected outcomes of production are usually either output or profit (the latter being a function of the output price).

The expected utility theory can be used to illustrate a pig producer’s preference towards risk using the income variance approach. Assume a pig producer with an expected utility (EU) function EU = P1.U (I1) + P2.U (I2) and EMV = P1.I1 + P2.I2, where EMV is the expected money value, I1 and I2 are income levels, and P1 and P2 are their respective probabilities. Suppose a certain income level IA. If IA< EMV but both yield the same level of utility (that is the producer is indifferent between IA and EMV) and if the producer chooses IA and forgoes an amount of income equivalent to EMV – IA, then he is said to be risk-averse. Generally, a producer is risk-averse if E(U) ≤ U{E(X)} and vice versa. In the presence of risk and uncertainty, agribusinesses do not always choose the option with the highest expected outcome.

The Just-Pope production function The Just-Pope framework allows for modelling the marginal effects of inputs and environmental factors on output and production uncertainty. The Just-Pope production function takes the form

= ( ; ) + ( ; ) (1)

Where f(x;α) is the mean𝑦𝑦 production𝑓𝑓 𝑥𝑥 𝛼𝛼 function,ℎ 𝑧𝑧 𝛽𝛽 𝜀𝜀h(z;β) is the output variance function or risk function), x and z are input vectors with parameters α and β respectively. The function f(.) comprises factors of production while g(.) contains factors that are variance-increasing or variance decreasing vis-à-vis expected output. Parameters of both average output and risk functions could be identical or distinct. Meanwhile, production shock is captured by a homoscedastic disturbance term ε. The specification in equation (1) above enables us to rewrite the variance function as an additive (heteroscedastic) error term such that the production model becomes

= ( ; ) + (2)

Where E(u) = 0 and Var(u)=[h(.)]².𝑦𝑦 The𝑓𝑓 𝑥𝑥 Just𝛼𝛼 -Pope𝑢𝑢 Framework has been used in many cases to analyse production risk and estimate risk preferences (for instance Kumbhakar & Tsionas 2010; Czekaj & Henningsen 2013). The derivation of production risk preferences involves incorporating the expected utility theory into the agribusiness’ production function based on the objective of the business, or the businesses’ expected outcome at the end of the production season. The main business objectives usually adopted in production risk analysis are profit and output maximization, although producers could seek to achieve other objectives such as cost minimisation (Kumbhakar and Tsionas 2010).

Risk preference among pig producers Pig farming in the South West region of Cameroon is relatively small-scale, mostly practised by farm households. To simplify our analysis, we assume that, for all the pig farmers, the sole objective is to maximise expected utility derived from profits obtained from pig farming.

Thus ( ), where is profit from the pig farming business.

However,𝑀𝑀𝑀𝑀𝑀𝑀 𝐸𝐸𝐸𝐸 = py𝜋𝜋 – rx, where𝜋𝜋 y is output, x is the input vector, and p and r are the vectors of output and input prices respectively. Substituting equation (1) in the profit function yields 𝜋𝜋

5

= [ ( ) + ( ) ] (3a) = ( ) + ( ) (3b) 𝜋𝜋 𝑝𝑝 𝑓𝑓 𝑥𝑥 𝑔𝑔 𝑥𝑥 𝜀𝜀 − 𝑟𝑟𝑟𝑟 The objective of the pig farmer therefore,𝜋𝜋 𝑝𝑝𝑝𝑝 𝑥𝑥will− be𝑟𝑟𝑟𝑟 to 𝑝𝑝𝑝𝑝 𝑥𝑥 𝜀𝜀

( ) = ( ) + ( )

For the first-order conditions for𝑀𝑀𝑀𝑀 a𝑀𝑀𝑀𝑀𝑀𝑀 maximum𝜋𝜋 𝑝𝑝 to𝑝𝑝 be𝑥𝑥 satisfied,− 𝑟𝑟𝑟𝑟 𝑝𝑝 𝑝𝑝 𝑥𝑥 𝜀𝜀

( )[ ( ) + ( ) ] = 0 ( ) ′ ( ) + ( ) ( ) = [ ( )] 𝐸𝐸𝐸𝐸 𝜋𝜋 𝑝𝑝𝑝𝑝 𝑥𝑥 − 𝑟𝑟 (𝑝𝑝)𝑝𝑝 𝑥𝑥 𝜀𝜀 ′ ( ) + ( ) = (4) ′ ( ) 𝐸𝐸𝐸𝐸 𝜋𝜋 𝑝𝑝𝑝𝑝 𝑥𝑥 𝑝𝑝𝑝𝑝 𝑥𝑥 𝐸𝐸𝐸𝐸𝐸𝐸𝑈𝑈 ′𝜋𝜋𝜋𝜋𝜀𝜀 𝜀𝜀 𝑟𝑟 𝐸𝐸𝐸𝐸′ 𝜋𝜋 ′ 𝑝𝑝𝑝𝑝 𝑥𝑥 𝑝𝑝𝑝𝑝 𝑥𝑥 � 𝐸𝐸𝑈𝑈 𝜋𝜋 � 𝑟𝑟

Therefore, ( ) ( ) + ( ) = / (5) ′ ( ) 𝐸𝐸𝑈𝑈 𝜋𝜋 𝜀𝜀 ′ ( ) + ( ) = / 𝑓𝑓 𝑥𝑥 𝑔𝑔 𝑥𝑥 � 𝐸𝐸𝑈𝑈 𝜋𝜋 � 𝑟𝑟 𝑝𝑝 (6) 𝑓𝑓 𝑥𝑥 𝑔𝑔 𝑥𝑥 Θ 𝑟𝑟 (6 𝑝𝑝 )

( ) Where = is the risk preference function for pig producers whose aim is to maximise ′ ( ) 𝐸𝐸𝑈𝑈 𝜋𝜋 𝜀𝜀 expected utility from′ obtained profits. Θ � 𝐸𝐸𝑈𝑈 𝜋𝜋 �

Generally, = 0 = > 1 < 0 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 − 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 Θ � 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 Production decision-making entails taking𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 into− consideration𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 the𝑝𝑝𝑝𝑝 risks𝑝𝑝𝑝𝑝𝑝𝑝 involved with the associated inputs. Although several factors influence the production decisions of farm enterprises, we however consider only the socio-economic and demographic characteristics of farmers, the expected returns from the farm enterprise and the risks involved in production. As such, a pig farmer’s production decision based on the expected utility framework, is a function of expected output (profits) and risks associated with production.

Description of variables Variables in the production function The production function of pig-farm enterprises depends on both output and inputs of the production process. However, most of the conventional production inputs are not used in pig production in the Southwest region. As such, for most of the pig producers, zero values would be recorded for these variables. This is not uncommon for small-scale agriculture. In line with this, we follow the approach of Di Falco et al. (2007) where for most of the factors identified as likely to influence pig production or variability in output, we use dummies to control for their use or effect. The variables included in the production function in this research include: Output (Q): This refers to the total quantity of production at the end of each season, which in this case refers to the total quantity of pigs produced for sale by the farmer at end of the season. It could also be understood to mean expected returns at the end of the season.

6

Variable inputs: This refers to material inputs used in pig production such as pig feed, nutrients, and treatment, amongst others. These inputs are jointly assessed because pig farmers use a mixture of them throughout the production cycle, and so find it difficult to quantify a particular type. In addition, application is sometimes spontaneous for some of these inputs. As such, controlling for these variable inputs was done using a dummy which takes a value of one if the pig producer used at least some of these inputs, and zero otherwise.

Technology plays a very important role in pig production. In our model, this comprised biological capital mainly in the form of new or improved pig breeds and improved feed. As was the case with variable inputs, technology was captured using a dummy. Given that most of the pig producers live not far from the main town of Buea, and with other big towns in proximity, we also considered the effect of the pig producer engaging in another income generating activity, in addition to pig farming. In reality, most pig producers are either full-time farmers or engage in off- farm employment. They therefore seldom rely solely on pig production. So, we introduced a dummy for pig producers with second employment aside pig production. Additional dummies were included to capture male-female production differentials and formal schooling. We assume that expected output and its variability are increasing in inputs and ambivalent in terms of the other variables.

Factors influencing pig production decision The decision whether or not to keep pigs is determined by an array of factors that include demographic, socioeconomic, household and market characteristics. All these jointly determine the producer’s propensity to take or avoid risk with regard to pig production.

Pig production is economical if the pigs are sold on maturity to avoid wasting feed for marginal increase in size and weight. However, demand for pork, though regular, is affected by seasonality. Pig demand fluctuates with season as some periods (for instance festive periods) experience high increase in demand and consequently price and profits. Thus, the risk of having mature pigs during off-peak periods also influences a farmer’s decision to produce pigs as well as the quantity of pigs to be produced. Biological risk, as well as outbreaks and epidemics, also significantly influence pig production. The occurrence of outbreaks such as swine flu, and other diseases affecting pigs constitute a major risk for these businesses which significantly influence the pig producer’s decision to produce or not.

Technology-related risk in the form of output uncertainty vis-à-vis the adoption of new technologies constitutes another major risk factor for the pig business. The adoption of new technologies may not always yield the desired or desirable outcomes. Other factors that could influence risk aversion or risk seeking among pig producers and consequently production decision include age, sex and education. Here, we distinguish farmers who have achieved at least some formal schooling, from those who have never been to school at all. Also, we arbitrarily sort pig producers into teenage and non-teenage producers. We expect that, in general, pig producers are risk-averse vis-à-vis the aforementioned factors.

Estimation strategy Variability in expected output will normally appear in the form of heteroscedasticity. We first check for heteroscedasticity. If the presence of heteroscedasticity is confirmed, then a suitable estimator that downplays its effect (that is heteroscedasticity-consistent) can be used to estimate average output, as the latter not only provides reliable estimates but also allows for valid inference. This holds as long as focus is on expected output (Asche & Tveterås 1999). However, since we are also interested in the sources of output variability, the output variance (risk) function is equally estimated. For this purpose, the feasible generalised least squares (FGLS) technique was used to estimate the mean production and risk functions. With regard to the assumption of risk aversion in 7

terms of production decision, we do not empirically verify it using econometric techniques, but resort to exploratory and descriptive tools.

Description of study area and data The study was conducted in the Buea municipality in division, Southwest region of Cameroon. Fako is located at latitude 4° 10' (4.1667°) north and longitude 9° 10' (9.1667°) east, with an average elevation of 2,833 meters. It covers a surface area of 2093 km² and its population as at 2005 was estimated at 466, 412. The following sub-divisions make up Fako division: Buea, Idenau, Limbe, Muyuka and . Because of its location at the foot of Mount Cameroon, the climate of Fako is humid, with neighbourhoods at higher elevations (for example Buea, Muea, and Idenau) experiencing cooler temperatures while low-lying towns like Muyuka, Tiko, Limbe, Mutengene, and Ekona experience a hotter climate. Extended periods of rainfall, characterised by incessant drizzles which can last for weeks, are common during the rainy season, as are damp fogs, rolling off the mountain into the town.

Figure 2: Map of Study Area in the Fako division. Source: Limbe Meteorological station data.

The data for this study was collected through a survey with the use of a structured questionnaire administered to pig farmers in Buea. The study population comprised pig farmers and enterprises in Fako. Field visits were organised and a random sample of 60 male and female pig farmers was selected. The questionnaire used for the collection of the data was structured into three sections: demographic information, output and input information, and production risk information.

3. Results and discussion Socio-economic and demographic characteristics and production decision Of the pig farmers sampled, 36 were male (60%) while 24 were female (40%). Based on responses elicited from the producers, it was found that 50% of the male producers sampled were risk-averse, while 58% of the female producers revealed a risk-averse attitude. Usually, more educated farmers are more open to information, better at input application and are generally more productive than uneducated ones. This also influences their attitudes towards risk and thus production decisions 8

they adopt when faced with risk and uncertainty. Everything being equal, farmers with no formal education are more risk averse than farmers with some formal schooling (Cárcamo & von Cramon- Taubadel 2016). Training feeds producers with more information about imminent risks related to agriculture (technology, price, environmental) as well as strategies to manage these risks. Risk perception is also considerably influenced by the socio-economic characteristics of the farmers and of their farm business.

It was observed that of all the total respondents, 20 had no other income-generating activity other than pig farming while 40 (66%) engaged in other activities. This indicates that everything being equal, most producers keep pigs only part-time and either have a full-time job with the public service or some other non-farm employment. The results revealed equal frequencies of both risk- averse and risk-friendly farmers below the age of 20 and above 40 years of age, while farmers aged 20-40 are generally more risk friendly. The relatively lower risk aversion among farmers aged 20-40 could be a result of their limited dependence on the pig business which puts them under relatively lesser pressure to produce when they are faced with risk. This contrasts with teenagers and producers above 40 whose income may highly depend on the outcome of the production (due to lack of alternative sources of income for the former and family responsibilities for the latter). High dependence on business outcomes usually leads to higher levels of risk aversion amongst farmers.

Interaction of socio-economic indicators and production decision Gender, occupation, education and risk preference First, we considered the interaction between gender and other occupation to see how it influences farmers’ decision making or risk preference. Figure 3 below shows that both male and female educated pig agribusiness owners are more risk friendly than the less educated ones. In addition, both male and female producers with a second economic activity have higher propensity to take risk. This supports the finding of Sulewski & Kłoczko-Gajewska (2014) who observed that farmers with an off-farm employment are less risk-averse. Next, we disaggregate by gender, the effect of education on production decision. Both education and gender are key demographic indicators of farmers. However, their interaction effect on production decision could be different from their individual effects. It is observed that uneducated women and educated men are more risk-loving, while uneducated male and educated female pig producers are risk neutral.

Figure 3. Gender and education relationship to risk preference. 9

Age-education and risk preference The relationship between age and educational level is a very important indicator of producer behaviour. Educated farmers of a certain age group might react to risk differently from those of other age groups or compared to their uneducated counterparts. Our results showed that both educated and uneducated farmers between the age of 20 and 40 are more risk friendly as they will choose to produce when faced with risk. In terms of the relationship between male farmers’ educational status, other occupation and risk preference, educated male farmers, both with and without other occupation, as well as less educated male farmers without other occupation, are more risk friendly. Meanwhile, female farmers are more risk-averse than their male counterparts. Irrespective of educational status, with or without a second employment, female pig farmers exhibited a risk-averse behaviour. This had earlier been observed Cárcamo & von Cramon- Taubadel (2016).

Input, policy, technology, biological and market risk also have a significant influence on the decision to rear pigs. Variable inputs have been shown to be risk-increasing and thus influence producers’ production decision (Guttormsen & Roll, 2013). It was observed that the pig farmers are generally sceptical vis-à-vis the adoption of new technologies, as most of the pig farmers prefer the common breed of pigs and feed quality. About 65% of the pig producers reported that their decision to produce is strongly determined by their expectations of future demand for pigs.

Determinants of expected output and production uncertainty We ran a couple of tests prior to estimating our econometric models. As shown in tables 1 and 2, first, the choice of variables for our analysis was based on Akaike’s information criterion (AIC) and the Bayesian information criterion (BIC) which were estimated. Variables and the model retained was that with the least AIC (1,501.981) and BIC (1,514.547). Next, the Breush-Pagan test was used to verify whether the Just-Pope framework is suitable for our analysis. The null hypothesis of homoscedasticity is strongly rejected at 1% level, with the main source of variability in expected output being participation in a second income-generating activity, in addition to pig production.

Table 1. Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Variable Chi-squared Degrees of freedom p-value (unadjusted) Variable input 0.97 1 0.3259 Technology 4.53 1 0.0334 Sex 5.67 1 0.0173 Education 0.26 1 0.6112 Other employment 9.38 1 0.0022 Simultaneous 15.17 5 0.0097 Ho: Constant variance

Table 2. Akaike's information criterion and Bayesian information criterion

Model Observations Log- Log- Degrees of AIC BIC 10

likelihood likelihood freedom (null) (model) Model1 60 -756.9728 -744.979 7 1503.958 1518.618 Model2 60 -756.9728 -744.605 8 1505.21 1521.965 Model3 60 -756.9728 -744.9904 6 1501.981 1514.547 Model4 60 -756.9728 -744.979 7 1503.958 1518.618 Note: N=Obs used in calculating BIC; see [R] BIC note.

The FGLS estimation results for the expected output function and the variance function are reported in Table 3. The mean production function was estimated by fitting a simple linear model while the dependent variable of the risk function was log-transformed. Most of the explanatory variables had the expected sign in line with production theory. The average production function is more or less monotonic in inputs. In fact, apart from education, all variables in the model have positive marginal effects on average output. The results show that engaging in another economic activity apart from pig production and investment in, or the use of, variable inputs, have the highest and most significant positive marginal effects on expected pig production. Both coefficients are positive and statistically significant at 1% level and 5% level respectively). On average, pig producers who engaged in at least one other economic activity earned about 76,000 FCFA (US$131) more than their counterparts. Meanwhile, the application of pig feed, treatment and other variable inputs increased pig production and consequently pig income by 33,000 FCFA (US$57) on average. The adoption of improved production technologies has a positive effect on pig production and income but this effect is statistically infinitesimal. Same applies to the difference in average output between male and female pig producers, which is statistically zero. Conversely, educational attainment has a negative but statistically insignificant effect on mean production.

Table 3. FGLS estimates of average pig production function and variance function.

Mean production function Risk functiona VARIABLES (Linear) (Log-lin) Variable inputs 2,789** 0.0964 (14,115) (0.466) Technology 874.3 0.2699 (13,664) (0.408) Sex 9,299 0.3703 (13,457) (0.438) Education -1,494 -0.330 (16,443) (0.513) Other employment 76,334*** 2.4522*** (13,871) (0.438)

Weight 142,444*** (18,580) Constant 20.06*** (0.628) R-squared 0.920 0.176 Notes: a Reported coefficients for the risk function are obtained by transformation of the original coefficients using the formula100 1 . Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1. 𝛽𝛽�𝑖𝑖 ��𝑒𝑒 � − �

11

With regard to estimated coefficients of the output variance function, all the variables posited to influence fluctuations in expected output have the same sign as in the mean function. Although almost all the variables are found to be risk-increasing, the effect is statistically significant only for the variable other employment. In other words, the purported influence of education, sex, technology, and variable inputs on fluctuations in expected pig production can be attributed to mere chance. It is interesting to note that involvement in a second activity has a very significantly large influence on variation in mean output. However, although technology adoption in the form of improved feed or breed appears to increase variability in output, this effect is not statistically significant.

The contribution of income diversification to average output lends support to theories of the development of peasant households and time (re)allocation. The literature highlights two main pathways via which off-farm1 participation increases household income & welfare (Kousar and Abdulai 2013; Hoang et al. 2014; Adjognon et al. 2017). First, off-farm employment opportunities that burgeon as a result of transformation of the economy enable households to increase their non- farm income and by extension, their food and non-food consumption expenditures. Second, off- farm income provides financial capital for poor households who are almost always cash- constrained and this enables then to purchase inputs and carry out investments that boost expected agricultural output (Barrett et al. 2001; Kilic et al. 2009; Oseni & Winters 2011). Our finding that pig producers who diversify their income suffer higher exposure to fluctuations in their average pig production, is an indication that the second activities carried out by pig producers are often spontaneous and transitory, generating irregular income. The results are in line with those of Pandey & Pandey (2004). Our finding reveals interesting information vis-à-vis previous research. The results show that although production inputs increase expected output, they are not accountable for any fluctuations is expected pig production as argued in the literature (neither do they decrease output variability).

4. Conclusion and recommendations Agribusiness is very lucrative in Cameroon but production risks remain a major constraint in this sub-sector. The aim of this study was to identify factors influencing average pig production and output variability, as well as to explore determinants of production decision against the backdrop of production uncertainty. Although both descriptive and econometric techniques revealed a number of results, they all converge on the fact that pig producers become more risk averse when exposed to production risk. Variable input use and income diversification were identified as the main factors influencing expected production. Meanwhile, all factors posited to cause variability in output were found to have no statistically significant effect, apart from activity diversification that tends to increase output risk.

Due to imperfections in the credit and input markets, pig producers are predisposed to diversify their income sources by engaging in a second economic activity and this generates positive spillovers as the non-farm income facilitates investments have positive effect on average pig production. As such, our paper corroborates claims of the importance of off-farm employment for poor households. However, diversification in itself is not a silver bullet as it leads to higher fluctuations in expected output. This is subject to the caveat that certain important environmental variables have not been considered.

Based on the finding of this study, the following recommendations can be made:

1In the literature, off-farm employment generally refers to economic activities other than crop and livestock production. This is also sometimes called rural non-farm employment (RNFE) or off-farm income. However, the term as used in our empirical analysis refers to “non-pig” production only. 12

• Better policies should be put in place to reduce and/or manage risk in agriculture and agribusiness. Crop and livestock insurance can significantly increase investments in the production of crops and livestock. Insured farmers are more likely to decide to keep animals and thus allocate a larger share of their agricultural inputs to pig production. • Policies that differentiate between various societal strata ought to be effective given the differences in the preferences of the different groups as seen in the results above. • Increasing access to credit and encouraging diversification through farm and non-farm entrepreneurship are amongst measures that can help to boost production, notably via providing credit-constrained producers with financial capital for investment in pig production. • Government input subsidies can be a useful way of boosting production should not only enable production but also ensure the effects of risks from purchasing such inputs are minimized. • Enhancing the competitiveness of the agribusiness sector should also be an effective means of making farmers/producers more risk friendly. • The uptake/adoption and use of new technologies in production (which also depends on agribusinesses’ risk preference) needs to be to strengthened as the effect of technology adoption on pig production or output variability is yet to be felt. • Proper education and information available to agribusinesses as to the sources of risks and their expected consequences on the businesses will also enable these farmers to develop risk-coping strategies to minimize these production risks.

Production risk leads to investment disincentive, timid technology adoption and low productivity. For risk-averse producers, these problems are likely to persist unless adequate and appropriate measures are taken to encourage investment and the use of new technologies in production. This entails concerted efforts from the concerned stakeholders.

References Antonaci L, M Demeke, and A Vezzani. 2014. The Challenges of Managing Agricultural Price and Production Risks in Sub-Saharan Africa. ESA Working Paper No. 14-09. Rome: Food and Agriculture Organization of the United Nations (FAO). Asche F, and R Tveterås .1999. Modeling Production Risk with a Two-Step Procedure. Journal of Agricultural and Resource Economics 24(2): 424-439. https://doi.org/10.2307/40987031. Barrett CB, T Reardon and P Webb. 2001. Nonfarm Income Diversification and Household Livelihood Strategies in Rural Africa: Concepts, Dynamics, and Policy Implications. Food Policy 26: 315–331. https://doi.org/10.1016/S0306-9192(01)00014-8. Cárcamo J and S von Cramon-Taubadel. 2016. Assessing Small Raspberry Producers’ Risk and Ambiguity Preferences: Evidence from Field-Experiment Data in Rural Chile. Discussion Paper No. 1610. Department für Agrarökonomie und Rurale Entwicklung, Georg-August- Universität Göttingen. Cole S, X Giné, and J Vickery. 2017. How Does Risk Management Influence Production Decisions? Evidence from a Field Experiment. Review of Financial Studies. https://doi.org/10.1093/rfs/hhw080. Czekaj TG and A Henningsen. 2013. Panel Data Nonparametric Estimation of Production Risk and Risk Preferences: An Application to Polish Diary farms. IFRO Working Paper 2013/6, Department of Food and Resource Economics, University of Copenhagen. Dewbre J, and A Borot de Battisti A. 2008. Agricultural Progress in Cameroon, Ghana and Mali: Why it Happened and How to Sustain it. Paris, France: OECD Publishing. http://doi.org/10.1787/9789264039483-en.

13

Di Falco S, JP Chavas and M. Smale. 2007. Farmer Management of Production Risk on Degraded Lands: the Role of Wheat Variety Diversity in the Tigray Region, Ethiopia. Agricultural Economics 36(2): 147-156 https://doi.org/10.1111/j.1574-0862.2007.00194.x Guttormsen AG and KH Roll. 2013. Production Risk in Subsistence Agriculture. Journal of Agricultural Education and Extension, 20: 1-13. http://dx.doi.org/10.1080/1389224X.2013.775953. Hardaker JB, RBM Huirne, JR Anderson and G Lien. 2015. Coping with Risk in Agriculture: Applied Decision Analysis. Wallingford, UK: CABI Publishing. http://doi.org/9780851998312. Just RE and RD Pope. 1978. Stochastic Specification of Production Functions and Economic Implications. Journal of Econometrics 7: 67–86. Just RE and RD Pope. 1979. Production Function Estimation and Related Risk Considerations. American journal of Agricultural Economics 61: 276-284. Kahan D 2008. Managing Risk in Farming. FAO Working paper. Rome: FAO. Kilic T, C Carletto, J Miluka and S Savastano. 2009. Rural Nonfarm Income and Its Impact on Agriculture: Evidence from Albania. Agricultural Economics 40: 139–160. http://doi.org/10.1111/j.1574-0862.2009.00366.x. Kousar R and A Abdulai. 2013. Impacts of Rural Non-Farm Employment on Household Welfare in Pakistan”. Paper Presented at the 2nd AIEAA (Associazione Italiana di Economia Agraria e Applicata) Conference “Between Crisis and Development: Which Role for the bio- Economy, Parma, June 6-7. Kumbhakar SC, 2002. Risk Preference and Technology: A Joint Analysis. Marine Resource Economics, 17: 77-89. Kumbhakar SC and EG Tsionas. 2010. Estimation of Production Risk and Risk Preference Function: A Non-Parametric Approach. Annals of Operations Research, 176: 369-378. MINEPAT. 2011. Growth and Employment Strategy Paper. Moschini G and DA Hennessy. 2001. Uncertainty, Risk Aversion, and Risk Management for Agricultural Producers. In Handbook of Agricultural Economics 1 Part A, edited by B. L. Gardner, and G. C. Rausser, 88-153. Amsterdam: Elsevier. Neumann JV and O. Morgenstein. 1944. Theory of Games and Economic Behaviour. Princeton, New Jersey: Princeton University Press. Oseni G and P Winters. 2009. Rural Nonfarm Activities and Agricultural Crop Production in Nigeria. Agricultural Economics 40(2): 189–201. https://doi.org/10.1111/j.1574- 0862.2009.00369.x. Pandey A and S Pandey. 2004. Risk Fertilizer Use in the Rainfed Rice Ecosystem of Tarlac, Philippines. Journal of Agricultural and Applied Economics 36(1): 241-250. Roll KH, AG Guttormsen and F. Asche. 2006. Modelling Production Risk in Small Scale Subsistence Agriculture. Conference paper, International Association of Agricultural Economists Conference, Gold Coast, Australia, August 12-18, 2006 Sandmo A, 1971. On the Theory of the Competitive Firm Under Price Uncertainty. American Economic Review, 61: 65-73. Smale M, RE Pope and HD Leathers. 1994. Land Allocation in HYV Allocation Models: An Investigation of Alternative Explanations. American Journal of Agricultural Economics, 76(3): 535-546. Sulewski P and A Kłoczko-Gajewska. 2014. Farmers’ Risk Perception, Risk Aversion and Strategies to Cope with Production Risk: An Empirical Study from Poland. Studies in Agricultural Economics, 116: 140-147. World Bank. 2011. Weather Index Insurance for Agriculture: Guidance for Development Practitioners. Agriculture and Rural Development Discussion Paper 50. Agriculture and Rural Development Discussion Paper.

14