Crop Yield Prediction Using Time Series Models

INTRODUCTION Farming is a major source of income for many people in developing countries. In Ghana farming represents 36 percent of the country's GDP and is the main source of income for 60 percent of the population (http://earthtrends.wri.org, 2003 p. 1). In addition, agricultural growth in Ghana has been more rapid than growth in the non-agricultural sectors in recent years, expanding by an average annual rate of 5.5 percent, compared to 5.2 percent for the economy as a whole (Bogetic et al., 2007). As with other parts of the developing world and the African continent, climate and other environmental changes in Ghana has become a major threat to their agricultural economy (Etwire et al., 2013). Direct losses to farming include destruction of their assets (such as, crop, livestock) which push poor farmers into poverty traps from which they have little means of recovery. Indirect impacts include sub-optimal management of this financial risk exposure, for example by selecting low-risk, low-return asset and activity portfolios that reduce the risk of greater suffering, but limit growth potential and investment incentives, selling assets (at inopportune times), reducing nutrient intake, and withdrawing kids from school and hiring them out to work. The problem is exacerbated by the reaction of financial institutions, which may restrict lending to farmers to minimize exposure to agricultural risk. These indirect consequences hinder economic growth (Barnett et al., 2008). Traditional insurance is impractical in developing countries because of high transaction costs, adverse selection, information asymmetry, poor distribution, and other challenges which hinder the availability of protection (Skees, 2008). Furthermore, post-event response in the form of emergency aid, debt forgiveness, and grants are at risk following recent economic crises, and such public capital does not usually help create independent private solutions and can be inequitable and untimely. In recent years, index based insurance instruments have been piloted as a way for smallholder farmers to hedge their losses. Unlike traditional indemnity insurance, the payout on index insurance products is not based on actual farm level yield and/or revenue losses. It is rather based on realizations of an index which assumes correlations with actual farm yield (or revenue) losses. Since the indexes are based on objective and transparent sources of data, it is unlikely that informational asymmetries exist that can be exploited by index insurance contract purchasers. Thus, the inherent insurance problems of adverse selection and moral hazard, additionally the high transaction costs of implementation can be largely avoided (Deng et al. 2006). Index insurance may also have the benefit of crowding-in capital, and allow farmers to get loans for needed inputs, as the risk for agricultural losses and thus financial risk becomes more manageable (Carter et al. 2007). The two types of index products are parametric and sample-based. Examples of parametric indices in insurance include weather (with triggers based on variables such as rainfall, temperature, humidity, wind speed, etc.), flooding (water levels and durations triggers), wind speed (velocity and duration triggers) and seismic activity (Richter scale triggers). Sample based indices include area based yield insurance and sample based livestock index insurance. Area yield insurance is essentially a put option on the average yield for a production in a region/area. Payouts are triggered by shortfalls in that area average yield rather than farm level yield. For this reason, area yield insurance requires no farm-level risk underwriting or loss assessment. If the area is sufficiently large, area yield insurance is not susceptible to moral hazard problems, since the actions of an individual farmer will have no noticeable impact on the area average yield. Area yield insurance also has relatively low transaction costs since there is no need to establish and verify specific farm yields for each insured unit nor is there any need to conduct on-farm loss adjustment. …