Effects of climate variability and insurance adoption on crop production in select provinces of South Africa

Increasing climate variability increases the risks in production and prices of agricultural products. Inarguably, Africa’s susceptibility to climate change is high because it hosts the majority of the world’s poor who cannot afford the costs of coping mechanisms. Agricultural insurance is being largely put forward as a coping measure of adapting to climate change to sustain farm production and farmers’ livelihood. The study critically reviewed numerous publications on climate change impacts and the role of insurance in the adaptation process. It examined the effects of varying weather conditions and insurance on net crop revenue using the instrumental variable regression approach on a Ricardian model. The study further identified factors influencing the purchase of insurance among the farmers with a probit model. The study data were collected from a cross section of farmers in three selected provinces of South Africa. Results of data analysis indicated that owning insurance, number of labourers employed, size of irrigated farmland and rainfall have significant effects on net revenue. It was also revealed that experience, indicated by years of farming and revenue, influenced farmers’ adoption of insurance. Consequently, the paper advocates for the provision of efficient irrigation facilities and promotion of insurance among farmers. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wcc.2018.020 om https://iwaponline.com/jwcc/article-pdf/9/3/500/484725/jwc0090500.pdf 9 Zelda Anne Elum (corresponding author) Michael Akwasi Antwi Department of Agriculture and Animal Health, UNISA Science Campus, P.O. Box X6, Florida 1710, University of South Africa, Johannesburg, South Africa E-mail: zeldaforreal@yahoo.com; zelda.elum@uniport.edu.ng Zelda Anne Elum Department of Agricultural Economics and Extension, University of Port Harcourt, P.M.B. 5323, Choba, Port Harcourt, Rivers State, Nigeria Godwell Nhamo Institute for Corporate Citizenship, University of South Africa, P.O. Box 392, UNISA, 0003, Pretoria, South Africa


INTRODUCTION
Agricultural commodity producers across the globe are facing increasing production and price risks attributed to climate change (Raju & Chand ). The adverse effects associated with climate change and weather conditions include flooding, drought, fire, hail, extreme temperatures, incidence of pest and diseases, among others, and have been shown to affect agriculture and livelihood in many ways, such as increasing input costs and reducing profits due to total failure or reduced harvest, and severe incidence of livestock deaths (CARE ; Müller ). Thus, climate change risks can adversely affect the current and future decisions of farmers on timing of farm operations, seriously affecting farm production and income. The term 'risk' has been defined as 'that which represents the probability of occurrence of an event which may have adverse consequences at any stage in the pathway of a production chain' (Pasaribu ). On this note, it is imperative that farmers adopt some form of measures to mitigate their losses from climate-related disasters and one such measure is the uptake of insurance.
Global population is increasing and certainly future population increases will lead to increasing competition for available land for human food, animal feed and recreational needs. This would be made worse by the increasing challenges posed by climate change, however, the adoption of insurance coverage provides an opportunity to encourage farmers to continue in business and sustain food production. Agricultural insurance is a financial protection given to farm investments and is one method by which farmers can minimise financial implications of production variability and guard incomes against the catastrophic effect of losses arising from extreme weather events. Insurance spreads farm losses over space and time, and encourages farmers to sustain investments in agriculture despite the occurrence of natural hazards. Insurance can be used together with other climate risk management tools to strengthen societal resilience (Raju & Chand ). This refers to an increase in the ability of the general public or group of persons to cope with natural disaster and a reduction in their vulnerability to climate change impacts.

The economic impact of climate change on South
African field crops has been studied (Akpalu et al. ), but to the best of our knowledge there is limited literature on the implication of climate variability for agricultural insurance. Thus, the main aim of this study is to examine whether changing weather condition and insurance significantly impact the net returns of crops across the provinces of Gauteng, Limpopo and Mpumalanga in South Africa and to also generate information on agricultural insurance with a special focus on South Africa. To this end, the work is structured into four parts as follows. The first section introduces the study and its objectives, then gives numerous literature references and discusses climate change impacts on agriculture, and the role of the insurance industry in mitigating climate risks on a global and national level. The next sections lay out the methodology, presents and discuss the analysed results, before drawing conclusions in the final section. Food systems globally consume about 30% of the world's total energy (Elum et al. ). On the other hand, agriculture provides positive effects for climate change, as all plant cover acts as a carbon sink. Nevertheless, going by reports, between 75 and 250 million people in Africa would be exposed to water stress by 2020 as a result of climate change (UNFCCC ). The changes in rainfall patterns could cause drought or flooding while rising temperature will cause changes in planting seasons, wildfires and increased incidence of pest and diseases (Chambwera & Stage ). The impact of climate factors such as higher temperatures and decreasing rainfalls on agriculture could lead to diminished production and higher prices that could result in socio-political instability (Sullivan ). Without doubt, climate change presents challenges to agricultural production and infrastructures and affects farmers' livelihoods in Africa. More so, the continent has been identified as highly vulnerable to changing climate due to varying prevailing factors such as poverty, illiteracy, lack of manpower, weak institutions, poorly developed infrastructure, poor health care, armed conflicts, unbridled corruption, population increase and land degradation (UNFCCC ).
Evidently, climate change has an impact on agriculture (Seo & Mendelsohn ). Two broad approaches often employed in the assessment of the economic impact of climate change on agriculture are the agronomic-economic and the Ricardian approaches (Kumar ). The agronomic-economic method uses a crop model that has been standardized from controlled experiments in which the crops are grown under a field or laboratory settings with different simulated climates and levels of carbon dioxide.
However, this approach has the disadvantage of not including adaptation in its estimates since it does not allow changes in the farming methods across experimental fields

Overview of South African agriculture
The agricultural sector in South Africa contributes 2.6% (DAFF ) to the country's GDP and employs 5.1% of the country's labour force (DAFF ). South Africa has a dual agricultural economy comprising of a commercial sector that is well developed and a considerably less developed subsistence sector. The major crops grown include maize, wheat, oats, sorghum, rye, barley, tobacco, cotton, deciduous fruit trees and citrus fruit. Notably, maize one of the main staple crops of South Africa is produced almost throughout the country. Maize accounts for 70% of total produced grains and occupies 60% of cropped area (Akpalu et al. ). South Africa is considered a dry country with more than two-thirds of its area experiencing less than 500 mm average annual rainfall (Durand ) and as such, agriculture activities have been to a large extent adapted to semi-arid conditions. Irrigated farming involves 1.3 million hectares of land and this places South Africa as the largest 'irrigation country in the Southern African region' (Durand ). Most of the 22 major rivers in the country provide water for numerous activities including agriculture; however, it has been projected that water requirement in South Africa will exceed available water by 2025 (UNEP ). Role of insurance as an adaptive strategy to climate change impact A UNEP Finance Initiative Report indicated that climate change-related risks can be viewed in terms of the frequency and severity of extreme weather events (e.g. flood, drought and storms) and in terms of slow onset but long-term events (e.g. sea-level rise and desertification). In recent times, discussions on the suitability of insurance in managing slow onset events in vulnerable countries have emerged. There has been a consensus that a risk transfer mechanism as insurance can help to address losses associated with the impacts of climate change (Balogun ). Notably, the IPCC recognizes insurance as one method of developing resilience to climate change impacts. The insurance system works on one of two principles which are: spreading the cost of those suffering losses over those commonly exposed to the possibility of loss, and spreading each individual's costs from random of the global agricultural insurance market premiums compared to Africa's mere 1% (Iturrioz ). In comparison to other continents, Africa has the lowest insurance penetration and insurance density (UNEP Finance Initiative ). Insurance penetration indicates the level of development of a country's insurance sector and it is measured as the ratio of total premiums to the country's gross domestic product (GDP) expressed in percentage terms (KPMG ). Insurance density refers to insurance premium per capita, that is, the ratio of total premiums to the total population (KPMG ). Although insurance is mainly created to provide relief after the occurrence of losses, it could be designed to motivate proactive actions that reduce risks (Schipper et al. ) through new products such as weather-based index insurance (WBII).
Insurance offers protection to farmers against various risks. Apparently, farmers who are insured tend to be less risk-averse than those not insured. The most developed type of agricultural insurance has been crop insurance (Iturrioz ). Traditional forms of claims-based crop insurance cover multiple risks, including weather-related risks such as hail and drought, as well as non-weather catastrophes such as pest and disease outbreaks. However, the price of premiums are often high and unaffordable for poor smallholder farmers and worse still, the payment of claims is seen as too slow to quickly alleviate the adverse impacts of losses on farmer's livelihoods (Daron & Stainforth ). Agricultural insurance product also consists of the earlier mentioned index-based insurance made up of area-based yield index and WBII. In area-based yield insurance, indemnity to beneficiaries are based on the average yield of all producers in the region, whether or not they purchased insurance. In WBII, claims payment is based on the occurrence of a specific weather parameter over a pre-speci- To achieve the objective of the study and as highlighted earlier, the Ricardian model was adopted, regressing the crop net revenue per hectare on climate and other exogenous variables. The model is expressed as: where Y is the crop net revenue per hectare as used by Gbetibouo & Hassan (), P i is the market price of crop i, Q i is the output of crop i, X is a vector of purchased inputs other than land, F is the vector of non-climatic variables, Z is a vector of climatic variables, G is a set of socioeconomic variables and R is a set of input prices. It is supposed that the farmer will choose X based on the characteristics of the farm and market prices in order to maximize net revenues. The relationship between the climatic variables and crop net revenue are assumed to be non-linear (Seo & Mendelsohn ). As such, the Ricardian model used in this study was that of quadratic formula and in a semi-log form: The analysis model also included an interaction term between air temperature and rainfall variables. An instrumental variable (IV) regression approach such as the two stage least squares (2SLS) was used in the regression analysis as it was assumed that for a farmer, having an insurance coverage may influence farm revenue just as farm revenue may influence a farmer's decision to purchase insurance (Di Falco et al. ). In this regard, insurance is treated as an endogenous variable since the factors affecting farm net revenue could also influence a farmer's decision to purchase insurance. The 2SLS is an IV estimator that entails two consecutive ordinary least-squares (OLS) regressions. This regression was done with the help of STATA 12.0. The IV model is as shown: where y i is the dependent variable for the ith observation, Y i stands for the endogenous regressors, X 1i represents the included exogenous regressors and X 2i are the excluded exogenous regressors. It is presumed that there is a nonzero relationship between U i and V i . Thus the study model is represented as: Insurance ¼ β 2 irrigation area þ β 3 labour þ . . . . . . þ β n X n þ π 1 education þ . . . . . . þ π n X n (6) where Y is the binary dependent variable for the ith observation and the Xs are the regressors. One of the a priori expectations of this analysis was that the likelihood of a farmer buying insurance would increase if farmers depended more on irrigation and if they had access to institutional credit. The variables used in the analysis are defined in Table 1. However, a possible limitation of the study data was the assumption that farmers may not want to disclose the true quantity (or revenue) of their produce while exaggerating on their expenditure.

RESULTS AND DISCUSSION
In this section, the results of the Ricardian and probit regression which were used to determine the economic impacts of climate change on crops and the determinants of insurance adoption among farmers, respectively, are presented and discussed. First, the farmers' descriptive statistics are shown in Table 2. It is inferred from the standard deviations which express the level of disparity in the variables that there was great disparity in the input expenditures and revenue obtained among the farmers. This may be attributed to the different crops planted by the farmers as the crops require different amount of inputs and also command significantly different market prices. The average age  of the farmers reflected that the farmers were in their youthful and active years. Also, the average size of cultivated farms which was very small indicated that the farmers were mostly subsistence and small-scale commercial farmers.
The result of the 2SLS regression is presented in Table 3.
It could be observed that the dummy for the adoption of insurance is statistically significant and positive (STATA takes zero as the base and reports for 1), implying that farmers who insured their farm business had relatively higher net revenue than those who did not. However, it is also possible that farmers with higher incomes are in a better position to afford to buy insurance. One implication of this is that insurance can be used as a means of reducing agricultural loss brought about by climate variability (Di Falco et al. ).
The labour variable was also significant, but with a negative sign which implies that increasing number of laborers reduces the farmers net revenue. This may be so if more labour, which means higher cost of wages, is employed without increasing cultivated area as this would reduce profit. The result of the probit regression is presented in

Number of observations 235
Notes: *significant at 10%; **significant at 5%; ***significant at 1%.  periods (1985-1999 and 2000-2014) to examine any significant differences in the climatic parameters (Elum et al. ). Table 5 shows that Limpopo province had experienced a significant change in rainfall and minimum temperature.
Similarly, change in minimum temperature was significant for Mpumalanga province. The covariances of rainfall and air temperature imply that as it gets hotter, there is less rainfall in the provinces. The persistence of such trends would indicate that these provinces are likely to face increasing challenges from climate change.

CONCLUSION
The study examined the effects of climate variability and insurance adoption on the net returns of crops across selected provinces of South Africa. It is inferred from the study that farmers with insurance coverage had higher net revenue. In addition, rainfall positively affects net revenue until it peaks (excessive rainfall) and then assumes a diminishing effect. It was further revealed that the adoption of insurance is significantly influenced by rainfall, length of farming experience and the extent/level of irrigated farming.
It is implied from the study that people would more likely buy insurance with increasing occurrence of flooding caused by excessive rainfall. However, farmers may choose not to participate in the insurance market due to various factors, e.g. cost constraint or trusting in their knowledge of agronomic practices for climate adaption.
The recommendations synthesized from the study are as follows. There is need for the promotion of insurance among  (2015) and Elum et al. (2017).
emerging farmers, who should also be encouraged to invest in efficient irrigation facilities as this is important for optimum crop yield. In addition, farmers should be provided with timely information on changing planting times that will not coincide with periods of heavy rains as well as adequate provision of inputs such as fertilisers and seeds. Despite the limitations of the study, the results of the analysis are consistent with those in economic literature. However, there is scope for further research for more robust results.