Pork price forecast based on breeding sow stocks and hog-grain price ratio

Abstract: Hog market is full competitive in China. There are obvious and regular cyclical feature in the fluctuation of pork price. Since 2003, pork prices have experienced over two full cycles, the time intervals between the adjacent price valleys are 36 months and 41 months, and the lengths between the adjacent price peak are 41 months and 43 months. Taking both price peak and valley intervals into account, the average length of the cycles is approximately 40 months. Hog production has its own specific regularity that producers enlarge their production scales when profit rate increases, reduce scales in contrast. Breeding sow stocks are the keys of the hog production scale changes. Increase or decrease of breeding sow stock will cause pork supply changes in 10 months because of the rules of hog production. Granger causality test results show that current pork price and current hog-grain price ratio are unidirectional Granger causality to current breeding sow stocks, breeding sow stocks and hog-grain price ratios 10 months ago are unidirectional Granger causality to current pig prices. The contribution of this paper is the multiplicative pork price forecast models which synthesize seasonal factor, trend factor and cyclical factor based on breeding sow stock deviations and hog-grain price ratio deviation 10 months ago. Long term trend equation of pork price is a linear function of time. By using least squares method and pork price time series data, the trend equation parameters can be obtained. The calculations of each month pork price devided by the Feb pork price in the same year separately are the seasonal factor values in one year. Averaging the same month seasonal factors of different years, we get seasonal pork price feature curve, fit the curve with a sine function in which the length of cycle is 12 months. Deviding real pork prices by long term trend prices and seasonal feature fitting values of the same month, cyclical factor curve is obtained. The cyclical curve is fitted by a sine function also, but in which the length of the cycle is 40 months, the amplitude is the functions of breeding sow stock deviation and hog-grain price ratio deviation 10 months ago. Using monthly hog-grain price ratio data from 2000 to 2011, the parameters of the forecast model can be estimated. We forecast 2012 pork prices with the model based on hog-grain price ratio deviation 10 months ago, the average forecasting accuracy rate is 95.6 percent. Using monthly breeding sow stocks data from 2009 to 2011, we obtain the forecast model parameters by least square method and forecast 2012 pork prices based on breeding sow stock deviation 10 months ago, the average accuracy rate is 97.2%. There will be a cyclical pork price valley in April 2013 based on hog-grain price ratio deviation model forecast, and the price value is 19.89 Yuan per kilogram. A cyclical pork price valley will occur in May 2013 based on breeding sow stocks deviation model forecast, and the price value will be 20.88 Yuan per kilogram.