2. REVIEW OF LITERATURE
3. MATERIALS AND METHODS
3.1 Haldwani Mandi and its Operations
3.2 Data Collection
3.3 Data Analysis
3.4 Programming for computation of trend
3.5 Programming for computation of seasonal variations
4. RESULTS AND DISCUSSION
4.1 Trend Analysis
4.2 Seasonal variations in arrivals and prices
4.3 Future Forecasts:
5. SUMMARY AND CONCLUSIONS
The fruit trade in India is conducted through a network of wholesale markets called Mandies. These Mandies represent a complex interaction amongst seller (grower), buyer (trader) and the regulatory agency (government). Therefore, for a properly managed post-production system fruit Mandies are the nodal points through which all the primary production of the concerned production catchment must pass. Thus a suitably designed modern Mandi could provide a unique opportunity for processing of fruits, which may involve receiving, cleaning and grading, drying, storage and disposal. Such a system is also expected to afford substantial reduction in primary losses of the total fruit production. However, for a rational and logical design of the Mandi system, including the various processing facilities, the characteristics of inflow i.e. fruit arrival pattern must be known, or be capable of being forecasted.
These arrival patterns are strongly time variant. One of the widely applied techniques to describe the time dependent stochastic processes is time series analysis. Since, the Mandi arrivals and prices form an ordered sequence of observations; they also represent a time series. Therefore, their time dependence could also be established using time series analysis techniques. Once the mathematical model to describe the time dependent structure of the series is developed it could be extended to generate future forecasts and help in Mandi design. The present study is a step in this direction. Specific objectives of the study are:
1. To analyse fruit arrival and price patterns using the historical arrival and price records of Haldwani Mandi,
2. To develop a program in MATLAB 7.0 for time series modelling in order to describe the time dependent structure of the fruit arrival and price series; and
3. To extend the time series models for prediction of monthly arrival and average monthly price value of the series.
2. REVIEW OF LITERATURE
Generally, there are two approaches to model the forecasts. One is the so-called classical econometric approach which involves an analysis of the factors on which the variable to be predicted depends. It concentrates on the specification of a system of behavioural equations which characterize the underlying variable to be predicted by means of multiple regression or simultaneous equation techniques. Another category of forecasting models is the time series approach, which is in contrast to the econometric approach in that no independent variables are used. This approach bases its predictions solely on current and past behaviour of the variable, by assuming that the actual observed series may be considered as a realization of some stochastic process with a structure that can be characterized and described. Examples of this approach include autoregressive (AR) models, moving average (MA) models, exponential smoothing (ES) models, a combination of AR and MA models and the Box-Jenkins (BJ) techniques.
Baudentistl (1971) examined marketing trends for the total milk supply and individual dairy products in Austria for 1960-70 period with a view to predict likely future trends in the market from developments in the past. Total milk supply during the period increased steadily until 1967. After a stagnation period, supply increased again from 1969, although to a lesser extent.
Singh and Sidhu (1972) investigated the trends of market arrivals and prices of wheat, paddy, maize and gram in different size markets of the Punjab state, for the period from 1964-65 to 1969-70. The arrivals of all the four commodities were much higher in the bigger markets because of high prices prevailing there. Arrivals of wheat, maize and gram in large markets were between 60 and 70% and those of paddy were about 85% of the total arrivals. The major portions of the arrivals were during the post-harvest period. The arrivals varied inversely with the prices.
Blyn (1973) investigated into the price series correlations as a convenient measure of market integration. Eight year collection of monthly wheat price data in eight Punjab markets and Delhi was analysed to eliminate trend and seasonal influences and detrained series were then correlated. He concluded that correlation among prices ruling in a set of markets was imperfect measure the degree of market integration and that correlation among the prices adjusted for trend and seasonality was a better measure.
Singh (1975) examined arrivals and prices for 1960-61 to 1970-71 periods in the Khanna market of Punjab for the possible correlation. The results showed high correlation. The results showed high correlation between arrivals and prices i.e. when the arrivals were high, prices were low and vice-versa. This is in agreement with the findings of Singh and Sidhu (1972) and Venkataramanan and Murlidharan (1972). He also observed that the lagged prices of groundnut affected its production.
Hanssens (1977) illustrated the use of invariable and multiple time series analysis in the development of descriptive marketing models. He used Box-Jenkins ARIMA model for modelling the individual time series. His approach is different earlier investigators, in that none of them employed time series analysis for examining market arrival patterns and its correlation with the market price. He also explored the relationship between econometric models and time series models and argued that a combined use of both techniques is most favourable for marketing model building.
Singh and Sindhu (1979) examined seasonal variations in production, marketed surplus and process of milk in the urban and rural areas of Punjab in 1977. They found that the production in urban areas varied from 85,111 litres in May to 98,666 litres in November and from 28,750 litres in June to 61,250 litres in January in the rural areas. Marketed surplus of milk accounted for 91.5 % and 71.5% of total milk production in the urban and rural areas respectively. The marketed surplus varied with the level of production, consumption and price over the different months.
Sharma and Roy (1979) analysed the trends in Indian food grain consumption and factors influencing it. They observed that per capita consumption and factors influencing it. They observed that per capita consumption declined due to rise in food prices. The other factors which might have caused this decline are identified to be the shift in taste away from food grains and deterioration and distribution of income.
Jain and Kaul (1980) analyzed the seasonal fluctuations in the potato market and the existence of cycles and their periodicities using series data relating to arrivals and prices from four major potato markets in Punjab. A multiplicative time series model was fitted to the data to analyzed trend, seasonal and cyclic variations in potato prices and arrivals. Prices showed an upward trend but large fluctuations in a year confirmed by cyclic variations with three year cycle period. No attempt is however, made for forecasting future arrivals and prices.
Boyle (1980) made a time series analysis of monthly pig deliveries at Irish bacon factories using Box-Jenkins technique. The forecasts were made for 24 months ahead with an average monthly error of 4.4 %.
Luandahl and Peterson (1982) calculated monthly correlation coefficients, using 1969-74 price data for five food grain products from Haiti markets, to show seasonal pattern and what they indicate about the structure of the marketing network. The numbers of markets considered for each product were: 10 for rice, 8 for millet, 20 for maize, 11 for ground maize and 15 for red beans. The results showed a lower correlation during the harvest months when most of the deliveries were made. The patterns show by the correlation coefficients for all the products were contrary to the expectation, probably because of the uni-directional trend.
Ngenge (1982) used time series models to investigate the dynamic relationship between the weekly cash prices of corn, sorghum and soybeans in three regions of the USA. In general, multivariate models performed better than univariate models, all markets were inefficient and most markets adjusted to changes in supply-demand conditions within a week, although residual adjustments took several weeks.
Singh and Gupta (1982) analysed grain arrival patterns and marketing practices of five food grains (paddy, wheat, maize, rapeseed and gram) for Rudrapur grain market of Uttar Pradesh. The maximum weekly arrivals varied from 3845 T of wheat to 180 T of gram. Annual volume of arrivals for wheat and paddy over the period 1971-73 were comparable with an average of 16470 T followed by maize (2400T), rapeseed (1940 T) and gram (660T). The combined monthly volume of all grains varied considerably (10 - 12280 T) over the three year period. No effort was made to analyse the trends and seasonality and the forecasting models which is an important aspect of effective planning of Mandi operations and modernization programmes.
Narain (1984) analysed the grain arrival and price patterns in a wholesale grain market (Mandi) for trend, seasonality and periodicity in order to develop forecasting models for the grain arrival process so as to rationalize an important input to grain Mandi system design. Historical time series data on monthly arrivals and average monthly prices was collected from the Rudrapur Mandi records for the period 1968-69 to 1983-84. Arrivals and prices of five grains (wheat, common paddy, Basmati, rapeseed and masoor) were considered. These taken together constituted more than 90 % of the total Mandi arrivals and represented the three major classes of food grains i.e. cereals, oilseeds and pulses. Forecasting models were developed on the basis of first 168 months (July 1968 to June 1982) data using time series analysis technique. Forecasts were generated for the next 36 months (July 1982 to June 1985). These forecasts were compared with the actual arrivals for July 1982 to June 1984 period. Model adequacy was tested by the independence of the generated residuals.
Yuan (1985) developed a model of fruit and vegetable wholesale markets in Taiwan, based on size, location, transport costs, flows and quantity and destination. The results of the model indicated that 27 fruit and vegetable markets were needed in southern Taiwan by 1981, and a further six by 2001, i.e. one per region. Size should vary according to volume of trade. Transport costs should be considerably reduced by increasing the number of markets. The township is recognized as the base unit for market planning.
Kallolo et at. (1988) studied the relationship between grade and quality, and between price and quality of a product using groundnuts as an example. Data were collected from three regulated markets (Gokak, Hubli and Dharwad) in Karnataka state, India. A grade response model was fitted for each of the selected markets with grade value of the sample as dependent variable. To test these models multiple linear regression analysis and step-wise linear regression analysis were used. The study revealed that the grade value and price are closely related. The main variables affecting price are, dryness of the pods, maturity, soil content of the pods, shelling percentage, oil percentage and moisture percentage. The existing grading systems in the regulated markets were far from satisfactory.
Wegner (1989) in his theoretical and empirical study analysed production structure and developed, demand, international trade and agreements and prices for selected fruits. These include three with wide spread productions (apples, pears, kiwifruits), three tropical fruits (bananas, pineapples, mangoes) and three subtropical fruits (sweet citrus fruits, grapefruit, avocados). Conclusion are that although demand for all type of fruit is increasing with rising income (detailed demand analyses relate to the USA and GFR), there is a tendency for production of the more popular varieties to expand more rapidly than demand, with consequent choking of the market. Changing consumer tastes as well as rising incomes are important determinants of demand and are increasing the market for different varieties of fruits and particularly for exotic fruits. This helped by the increasing liberalization of trade products which greatly benefiters’ developing countries.
Pouch (1997) studied that in year 1994 world fruit production reached 388 Mt, of which 10 % were exports of fresh fruit. In the same year production of temperature climate fruit (apples, pears, peaches, plums, apricots, grapes and kiwis) reached 125 Mt, of which 8 Mt were exports, a 30 % increase since 1985. There has been a considerable increase in demand for fresh fruit has attracted many producers in the southern hemisphere, particularly Chile, Argentina, Mexico and Brazil. The general decrease in prices for tropical product exports has resulted in a diversification of production towards temperate climate fruit, which are in demand in the northern of hemisphere during the off season. A more detailed analysis of world apple, table grape and kiwi markets is presented and the position of France is examined briefly.
Yadav et al. (1999) studied the probability analysis of some of the fruits and vegetables arrival in Haldwani Mandi for eight years and ten years data, using the same five distributions and concluded that all the five distributions were good at five percent level of significance with probability factor highly significant. It was seen that for one month arrival at 80 percent probability level percentage deviation was minimum for Normal distribution and at 90 percent probability level it was minimum for Log Normal distribution and at 50 percent probability level percentage deviation was minimum for Log Pearson type HI distribution. The estimated values of these distributions and the actual arrival in next two years were compared. For two months consecutive maximum annual arrival, percentage deviation was minimum for normal distribution at 80 percent probability level and for Log Normal at 90 percent probability level and at 50 percent probability level percentage deviation was minimum for Log Pearson type III distribution. Picchi (2002) studied the factors affecting the interval between picking various fruits and their appearance on supermarket shelves is discussed. The categories considered are; most perishable, moderately perishable (apples, pears, citrus, nuts, etc.) and frozen or dried fruits. Times taken in transport, loading and unloading, sorting, storing, packaging and labelling clearly vary with the type of produce. Management of the stock on arrival in the supermarket is also considered. The topic is illustrated with the help of flow charts and diagrams.
Tijskens et al. (2003) analysed that a considerable proportion of the fresh produce (such as fruit, vegetables and flowers) sold around the world is exported. While most of the export is successful, a disturbingly high proportion of shipments encounter quality problems on arrival at destination. New technologies have been developed for monitoring refrigerated shipping containers that makes it possible for actual storage conditions during transport to be monitored in real time. This enables speedy rectification of any problems and alerting to any quality problems on arrival. Besides the considerable benefits that this technology achieves for all involved in the shipping and fresh produce industry, it is possible to interface storage prediction software and real time monitoring of storage conditions to provide a premium service to the industry that can fine tune the quality delivered to market requirements. Advantages of such a model are that decisions can be made before shipment as to potential storage life remaining for a particular crop (e.g. apples stored in controlled atmosphere prior to export). Then, if there is a problem for a particular voyage length, the shipment can be cancelled or storage conditions adjusted to assure a longer storage life. The development of a fully robust model to cover a wide range of products, a wide range of pre-shipment conditions and a range of problems during shipping will obviously take some time but the benefits will be considerable. Fitting a wide range of models to data for over 30 crops suggests that the best choices are the exponential and a modified quadratic model.