Seasonal availability, price behaviour and statistical forecasting of soybean: Evidence from the Nagpur agricultural market

Introduction

Soybean (Glycine max L.) is a major oilseed crop of the world, contributing significantly to food, feed, and edible oil supplies and plays an important role in agricultural economy and industrial growth. India is one of the major producers of soybean and Maharashtra contributes a large share of the total cultivated area and production in India besides being a leading producer. Soybean is also very crucial for improving farmers' income, industrial development, and trade. Market arrivals and prices in soybean are prone to high variations over a period of time due to the seasonality of production, production variability, post-harvest management, consumer preferences, and domestic and global demand conditions. Such fluctuations make it necessary to understand the seasonal market behavior in terms of arrivals and prices in order to facilitate efficient planning of production, management of marketing operations and stabilization of prices for all stakeholders involved in the agricultural system like producers, intermediaries, processing units, and policy makers (Rozi et al., 2025; Martignone et al., 2024; Margarido & Turolla, 2024) [1,12,23] . In agricultural commodity markets, production of crop commodities takes place only during a specific time of the year whereas demand for the commodities exists through the year. Thus, during the peak harvesting season, supply in the market usually exceeds demand, resulting in low prices while market arrivals and supply in the lean season are low, causing prices to rise. Past research has indicated that there is a negative relationship between market arrivals and prices, which shows the significance of seasonal behavior of markets for proper marketing of the produce and reduction in postharvest losses in agriculture (Pardhi et al., 2024; Jadeja et al., 2025; Santhosh et al., 2024) [4,5,7] . Additionally, better transport, storage facilities, market integration and the development of international trade have brought changes to the conventional seasonal nature of the prices which requires continuous monitoring of market behavior and assessment of the impact of the changing economic conditions of the sector (Macarringue et al., 2023; Martey et al., 2025; Wilson & Bullock, 2025) [2,3,18] . Accurate price forecasting is an essential element of agricultural market planning as it helps producers, traders, processors and policy makers in taking correct decisions regarding procurement, storage and marketing of commodities. Statistical models such as exponential smoothing and Autoregressive Integrated Moving Average www.extensionjournal.com (ARIMA) are commonly used techniques in the analysis of the price forecasting of the agricultural commodities (Cortazar et al., 2025; Cortazar et al., 2023) [9,10] . Similarly, decomposition analysis and calculation of seasonal index can give an insight into the seasonal pattern of prices and the magnitude of variation in the prices which can help improve the price forecasting (Jadeja et al., 2025; Pardhi et al., 2024) [5, 4] . Therefore, the study was conducted to analyze the seasonal behavior of soybean arrivals and prices in the Nagpur market and forecast the prices using decomposition analysis, seasonal indices, and different statistical models. This study aims to facilitate better market intelligence for informed storage and marketing decisions and also assist the stakeholders in developing effective policy interventions to improve the price stability and market efficiency of soybean in soybean growing areas.

Literature Review

Due to the important role it plays in providing food security, edible oil, feed to animals and other industries, soybean has become one of the strategic crops in recent years. Soybean production systems have been the subject of research for enhancing sustainable agriculture and achieving global food security. As noted by Rozi et al. (2025) [1] , enhancing the economic efficiency of soybean production and marketing can improve food security and help address food crises globally. Similarly, Martignone et al. (2024) [12] noted that the growing significance of soybean in commodity markets can be traced to an increasing demand for soybeans in food processing and animal feed industries that has driven an upsurge in international soybean trade. Margarido and Turolla (2024) [23] discussed prospects for the development of Brazil's soybean industry and underscored the role of transportation, seasonal production and price transmission for market competitiveness. Richards and Arima (2026) [17] observed that increases in soybean profitability have contributed to land values and land use changes in important production areas. The cited studies establish the importance of soybean as an economically important crop whose market success is dependent on production and marketing systems. In recent years, the seasonal behaviour and price dynamics of soybean markets have been widely researched as a means of enhancing marketing efficiency and reducing market uncertainty. For example, Pardhi et al. (2024) [4] reported that market arrivals and prices had an inverse relationship in soybean markets in Maharashtra such that high post-harvest arrivals translated into low market prices. Jadeja et al. (2025) [5] noted a similar seasonal pattern of market arrivals and soybean prices in domestic and international soybean markets. Santhosh et al. (2024) [7] found that high market arrivals in the harvest season can contribute to price instability while Santos (2023) [13] noted that an annual cyclical fluctuation in soybean prices in Brazil follows patterns of production and marketing seasons. Kang and Lee (2025) [15] noted that soybean price seasonality has been influenced by structural changes in energy markets and rising demand for renewable diesel. Studies have also emphasized the importance of storage, transportation, logistics and market integration. Martey et al. (2025) [3] reported that the quality of storage facilities significantly influences postharvest losses and trader profitability. In addition, Macarringue et al. (2023) [2] reported that soybean freight rates are influenced by the harvest seasonality and market prices and Marsola et al. (2025) [8] emphasized the influence of climatic conditions, congestion and transportation infrastructure. The influence of tariff regulations, seasonal production and international trade conditions on the dynamics of soybean markets was also noted by Wilson and Bullock (2025) [18] and Song et al. (2025) [20] . The risk spillover effect of soybean markets on domestic and international trade in Brazil was confirmed by Zhang and Shi (2024) [24] . The focus of soybean research in recent years has been on how to enhance forecasting accuracy and agricultural decision-making. While statistical techniques still have a prominent role in predicting soybean market behaviour, other studies have incorporated climate, remote sensing, optimization, risk assessment and other variables to better understand soybean markets. For instance, Cortazar et al. (2025) [9] estimated risk premiums on agricultural commodities through soybean futures price and analyst forecasts, while Cortazar et al. (2023) [10] quantified risk premiums in soybean, corn and wheat markets using futures market data. Basco et al. (2025) [21] highlighted that early season climate indicators can be used to predict soybean yields and improve crop production plans in Argentina. Santana et al. (2024) [19] developed a satellite-based approach for estimating sowing and harvesting dates of soybean with MODIS-NDVI data. In addition, Zhang et al. (2025) [22] observed that soybean classification using a combination of optical and SAR time-series improved early season soybean identification. Byer et al. (2025) [16] employed mathematical programming techniques to maximize soybean production profits with disease management constraints, while Ilves (2025) [6] applied Monte Carlo simulation to examine soybean marketing decisions, emphasizing the need to understand seasonal price behaviour to enhance risk-adjusted returns. Taken together, the reviewed studies indicate that soybean markets are influenced by multiple factors such as seasonal production behaviour, logistics, trade policies, climatic factors and statistical methods, suggesting the need for further market-oriented analysis to help farmers make informed marketing decisions and policy-makers to formulate appropriate policies.

Research Methodology

This research analyzed the soybean market in Nagpur (Maharashtra) from April 2020 through March 2026 to assess price variability, availability and seasonality, and to forecast the prices in the upcoming season of soybean. The wholesale prices (Rs./ quintal) and arrival data of soybean (M.T.) were gathered from the Agricultural Marketing Information Network (AGMARKNET), Agricultural Produce Market Committee (APMC), Nagpur and Rasoya Proteins Ltd., Nagpur. A total 72 data points for both prices and availability were utilized which was considered enough to identify the long-run trend, seasonal changes and fluctuations in prices for soybean. The collected data were organized and subjected to MS Excel, MINITAB and R. For this study, the multiplicative time series decomposition and analysis of seasonal indices, seasonal price variability and the comparison of different forecasting models were carried out.

Multiplicative Time Series Decomposition

Seasonal movements in soybean arrivals and commodity prices tend to grow and diminish in proportion to the movement of the time series itself. For these reasons, we adopted the multiplicative decomposition model to extract soybean arrivals and price series into its trend, seasonal and irregular components. The multiplicative decomposition model is represented as

The multiplicative model assumes that the observed value is the combined product of trend, seasonal, and random variations.

Trend Estimation

The long-term trend was estimated using a twelve-month moving average. The moving average was computed as

Since twelve observations constitute an even period, centered moving averages were estimated as

where 𝑀𝐴 𝑡 = Twelve-month moving average 𝐶𝑀𝐴 𝑡 = Centered moving average The centered moving average represents the trend-cycle component after removing seasonal effects.

Estimation of Seasonal Indices

Seasonal-irregular ratios were estimated as

Monthly seasonal indices were obtained by averaging the ratios corresponding to each month. The adjusted seasonal index was then calculated as

These adjusted seasonal indices were used to identify peak arrival and lean marketing periods.

Seasonal Arrival Index

Monthly seasonal availability was measured using

Holt's Double Exponential Smoothing

Holt-Winters Exponential Smoothing

The multiplicative Holt-Winters model is expressed as

The forecasting equation is

Autoregressive Integrated Moving Average (ARIMA)

The ARIMA model is represented as

Forecast Accuracy Measures

The forecasting performance of all models was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The results of the decomposition analysis showed that both trend and seasonal effects exist in soybean prices. They rose steeply in 2021 and peaked at the end of the year. Then they declined and leveled off for the remainder of the period. Seasonal indices fluctuated to a moderate degree across the months. This suggests that seasonal effects were not the only drivers of soybean prices, as soybeans are also subject to the forces of demand, processing, and other external market effects. The irregular component contains the effects of other unmodeled factors such as weather and government policy. As illustrated in Figure 5, the level of arrivals of soybeans is more seasonal than the prices. The level of arrivals increases in October and November when price is on decline owing to excess supply and conversely prices increase on lower levels of arrivals in lean months.

Root Mean Square Error

Multiplicative Decomposition of Soybean Prices

Seasonal Behaviour of Soybean Arrivals and Prices

This clearly indicates that storage and marketing strategies play an important role in enhancing farmers' income. The average maximum price of soybean was recorded in 2021-22 (Rs. 6617.50 per quintal), whereas, the maximum arrivals in soybean were observed during 2022-23 (20287 MT). Similarly, seasonal arrival index registered its maximum value during 2022-23 (137.83), while seasonal price index found its highest value during 2021-22 (134.05). These suggest that soybean prices are affected by supply and demand-side situations, but they may not be always directly associated with the arrival of the crop. The graph shows great fluctuations in arrivals and prices from year to year. Years in which arrivals had reached higher volumes show a lower price index and years in which supplies were limited show better price development.

Inter-Annual Availability and Price Behaviour

Moreover, the prices and the indices show opposite trends, which confirms the seasonal market mechanism for the soybean market.

Seasonal Price Variation

IPR (Intra-Year Price Rise) was estimated as 42.25 per cent, showing significant differences between higher prices and lower prices. The ASPV (Average Seasonal Price Variation) was estimated at 35.51 per cent which revealed the presence of moderate to high seasonal variation in soybean prices. CV (Coefficient of Variation) was also found to be 10.65 per cent which implied low variation of prices across seasons. The relative evaluation of ASPV, IPR and CV showed that price fluctuation was more affected by irregular disturbance than only by seasonality. Even though a large share of the fluctuation of soybean price could be explained by seasonality, the remaining part may come from market disturbance, policy change, weather conditions and other international market events. The ANN model has performed better than all of the other forecasting methods in predicting the soybean price due to its lowest values in the RMSE, MAE, and MAPE error measurement, at $213.13, $191.39, and 4.34%, respectively. The second-best performance in predicting the soybean price is given to Holt's model, which is followed by SES and ARIMA methods with similar forecasting ability. However, the Holt-Winters model has produced the highest error values, thus this method is not recommended for soybean price forecasting. The ANN was selected as the best performing model due to its lowest MAPE value; this smaller percentage error illustrates its better performance in modelling the nonlinearity of the market prices and changing market conditions. The ANN model closely traced the observed soybean price path and accurately predicted the upward movement toward the end of the data period. On the other hand, ARIMA and SES gave relatively horizontal predictions and did not follow the movement of price in this period. During the testing phase, the ANN forecasts proved to be significantly nearer to the actual prices compared with the SES and ARIMA forecasts.

Forecasting Framework

Comparative Evaluation of Forecasting Models

Forecast Validation

While the linear models missed a portion of the 2026 recovery, the ANN was able to identify the price recovery that took place during the start of 2026. Visual inspection of the results indicated that the ANN predictions followed the soybean price trend relatively well, exhibiting small deviations from actual values. Therefore, the results suggested that it was possible for the ANN to successfully replicate both the trend and nonlinear behaviour of the soybean price series, thus demonstrating that ANN is a suitable model to be used for forecasting agricultural prices. The ANN predicted an increase in prices of soybean from INR 5299.95/quintal (Rs/quintal) in the month of April 2026 and the prices are likely to continue the rising trend during the entire forecasting period, reaching up to Rs 6272.25 per quintal in March 2027. Thus, the model predicted an upturn in soybean prices which is good for the soybean farmers during the future period. We find large discrepancies between these two types of prediction results. ANN forecasts a continuously increasing soybean price while the prediction of ARIMA is relatively stable. The better performance of ANN also shows the promise of using machine learning method for agricultural market forecasting and decision support system.

Future Price Forecasts

Comparative Analysis

Several studies have been carried out recently in the field of soybean markets focusing on analyzing the behaviour of the commodity in terms of seasonality, price behaviour and forecasting performance by using individual approaches. Pardhi et al. (2024) [4] studied seasonality in arrivals and price of soybean at markets of Maharashtra by computing seasonal indices and using moving averages whereas Jadeja et al.

(2025) [5] investigated the fluctuations in domestic and international price of soybean by trend and seasonality analysis. Recently, Oliva et al. (2025) [11] predicted price of soybean in Brazil by integrating climatic variables with LSTM models. Even though the findings of these studies provided substantial insights on soybean market research, the studies were either based on traditional seasonality analysis or used purely machine learning approach to analyse the commodities without considering market arrival, seasonal decomposition, and forecasting validation simultaneously. The current study intends to fill this gap by providing an analytical framework of multiplicative decomposition, seasonality, price behavior variation and ANN based forecasting model.

Methodologies Compared

• Pardhi et al. (2024) [4]: Seasonal Index Approach • Jadeja et al. (2025) [5]: Trend and Seasonality Analysis • Oliva et al. (2025) [11]: LSTM-Based Forecasting [4] Moderate High Moderate Jadeja et al. (2025) [5] Good Good Good Oliva et al. (2025) [11] Very Good Moderate Very Good Present Study Excellent Excellent Excellent

The suggested method achieved better results than the benchmark in all indicators. By applying multiplicative decomposition, it allowed to distinguish accurately the seasonal fluctuations, and by adding indices of price fluctuations, it permitted to identify and comprehend better the market trend. Moreover, in terms of predictive performance, the ANN reached the best results (MAPE = 4.34), reducing more than 3 times its error in relation to the traditional statistical approaches and machine learning approaches. Figure 1 is a bar chart that compares the four approaches in terms of three performance metrics: accuracy of the forecasts, capture of the seasonality, and applicability in the market. It is clear from Figure 1 that the proposed approach surpasses the other techniques in each of these three performance metrics, confirming the benefits of using decomposition and time-series forecasting with market indicators as part of the forecasting model. The results of this study suggest that the combination of classical techniques and nonlinear forecasting can form a more accurate and relevant forecasting system in the soybean market and support decision-makers in this market.

Comparative Performance Graph

Conclusion

This research explored seasonal availability, price dynamics and forecasting efficacy of soybean in Nagpur market using monthly arrival and wholesale price data for years 2020-21 to 2025-26. Decomposition method and seasonal indices revealed the presence of seasonality in soybean arrival; the peak arrivals occurred in the months of October and November after harvest and the market was in a lean state in August and September. An inverse relation between arrival and prices was noticed which signifies a downward pressure in price during harvest due to increased availability and high price in off-season months due to reduced availability. Estimated seasonal variation indices suggest that soybean prices are subjected to seasonal variability but the magnitude of variation is low and the prices are relatively constant throughout the year.

Comparative forecasting performance analysis revealed the Artificial Neural Network (ANN) approach as an improved technique over other conventional time series models like SES, Holt, Holt-Winters, and ARIMA for forecasting of soybean prices. ANN model recorded least forecasting errors and it was able to predict soybean prices by incorporating market trends and nonlinearities. The methodological approach used in the current study which included decomposition analysis, seasonal variability indices and ANN is found more efficient and improved over earlier reports.

The findings are expected to benefit the farmers, traders, processors and policy makers regarding storage decisions, procurement timing, marketing strategies and price risks associated with soybean market. The soybean market forecasting methods may be further improved by incorporating rainfall, temperature, international soybean prices, exchange rates, oil prices, export demand and government policy decisions in forecasting models and by using more sophisticated approaches such as Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU), Transformer, ARIMA-ANN hybrid model, etc. and the scope of the framework may be extended to include different markets with use of additional information on real time market intelligence, remote sensing data and supply chain indicators for developing dynamic decision support systems for the management of agricultural commodities.

References

1. Rozi F, Subagio H, Elisabeth DAA, Mufidah L, et al. Indonesian foodstuffs in facing global food crisis: economic aspects of soybean farming. Journal of Agriculture and Food Research. 2025.
2. Macarringue AMJS, Oliveira ALR, Dias CTS, et al. Multidimensionality of agricultural grain road freight price: a multiple linear regression model approach by variable selection. Ciência Rural. 2023.
3. Martey E, Goldsmith P, Etwire PM. Economics of storage and grain loss in the soybean trade in Northern Ghana. NJAS: Impact in Agricultural and Life Sciences. 2025.
4. Pardhi PS, Toor MS, Jahagirdar SW. Seasonal arrival and price behaviour of soybean in Amravati district of Maharashtra. Economic Affairs. 2024.
5. Jadeja RM, Thakar KP, Marviya PB, et al. Dynamics and seasonal fluctuations of soybean prices in domestic and international markets. Agricultural Science Digest. 2025.
6. Ilves J. Evaluating risk and return of corn and soybean marketing strategies using Monte Carlo simulation methods. 2025.
7. Santhosh KS, Kumareswaran T, et al. Analyzing price dynamics and instability in the soybean market of Gautampura: a decadal analysis. Asian Journal of Agricultural Extension, Economics and Sociology. 2024.
8. Marsola KB, Oliveira ALR, Utino MYR, Mann P, et al. The impact of exogenous variables on soybean freight: a machine learning analysis. Sustainability. 2025.
9. Cortazar G, Ortega H, Pérez JA. Using futures prices and analysts' forecasts to estimate agricultural commodity risk premiums. Risks. 2025.
10. Cortazar G, Ortega H, Pérez JA. Measuring risk premiums in corn, soybeans, and wheat using futures prices and forecasts. 2023.
11. Oliva JT, Martelli OP, Barbosa MAC, et al. The use of long short-term memory models to estimate soybean pricing: a regional climate data evaluation from Brazil. IEEE Access. 2025.
12. Martignone GMB, Ghosh B, Papadas D, Behrendt K. The rise of soybean in international commodity markets: a quantile investigation. Heliyon. 2024.
13. Santos BAG. Sazonalidade do preço da soja grão no Estado da Bahia entre os anos de 2010 e 2020. Bahia: Universidade Federal do Recôncavo da Bahia; 2023.
14. Wu B, Yu S, Lv SX. Explainable soybean futures price forecasting based on multi-source feature fusion. Journal of Forecasting. 2025.
15. Kang M, Lee S. Navigating the soybean market shift: river transport as a lens on the renewable diesel boom. 2025.
16. Byer T, Hatfield T, Kadelka C. A mathematical model informs optimal fungicide use against Sclerotinia stem rot to maximize profits in soybean production. Bulletin of Mathematical Biology. 2025.
17. Richards P, Arima E. High profits from soybean-corn agriculture are associated with increased land prices and deforestation rates in Mato Grosso's Amazon forests. Communications Earth & Environment. 2026.
18. Wilson W, Bullock DW. Non-tariff and tariff impediments affecting spatial competition between the United States and Brazil for soybean shipments to China. China Agricultural Economic Review. 2025.