Keywords
Rice supply chain, Rice import, Increasing rate of rice price, Hybrid optimization-regression model.
This article is included in the Agriculture, Food and Nutrition gateway.
This study addresses the challenge of stabilizing rice retail prices in Indonesia, where rice is a critical staple food, as in many Asian countries. The government prioritizes price stability to prevent sharp increases that could trigger social unrest during shortages. Price control approaches are categorized as direct or indirect. Direct controls involve immediate interventions, such as boosting rice stocks through imports to quickly influence market prices. Indirect controls consist of longer-term measures, like enhancing domestic production capacity for gradual stabilization. This study proposes an optimization model to determine the optimal rice import volume to minimize bi-monthly retail price fluctuations.
A linear programming model is formulated to minimize bimonthly price changes, subject to constraints including local production capacity, import limits, rice flow balance, and demand fulfillment. The monthly retail price is modeled using a compound linear regression approach with seven explanatory variables: the rupiah exchange rate against the US dollar, GDP per capita, the price of ground dry rice (GKG) per kilogram, domestic rice procurement, rice imports, rice distribution, and government-managed rice stock aimed at ensuring domestic availability and price stability. The explanatory variable is forecasted using methods best suited to its historical pattern.
The model was tested using data from 2020 to 2023. The results indicate that bimonthly rice prices increases can be effectively controlled, with maximum price change rates maintained between 0.75% and 1.14% and a standard deviation ranging from 0.20% to 0.34%. These values are significantly lower than the anticipated inflation rate of 2–3%.
The optimization model effectively determines the required volume of rice imports to minimize bimonthly retail price fluctuations. By regulating import volumes, excessive price increases can be prevented. Enhanced data-driven forecasting with granular historical data may further improve the accuracy of retail rice price predictions and strengthen price stabilization initiatives.
Rice supply chain, Rice import, Increasing rate of rice price, Hybrid optimization-regression model.
The revised version of the article incorporates an enhanced compound linear regression model that includes seven independent variables, offering improved accuracy in estimating optimal rice import volumes aimed at minimizing fluctuations in the monthly rice price change rate. To validate the model, a behavioral evaluation was conducted to assess the extent to which the model accurately replicates the real-world dynamics of the system. This evaluation followed a standard procedure involving the comparison of simulated outputs with historical data over time, along with calculating the average deviation between simulated and observed values. The primary performance metric used in this assessment was the Mean Squared Error (MSE), supplemented by its decomposition into bias (UM), variance inequality (US), and covariance inequality (UC) components. Ultimately, the model's application in determining monthly rice import quantities is presented, with simulations conducted over 12 months for four consecutive years: January–December of 2020, 2021, 2022, and 2023.
Additionally, a sensitivity analysis was performed by varying the lower limit of the strategic inventory to examine the effect of parameter changes on the model’s outcomes. The sensitivity analysis results indicate that, from 2020 to 2023, changes in the strategic inventory threshold had minimal impact on the monthly rice price change rate between consecutive months, with maximum monthly fluctuations ranging from 0.42% to 1.14%. In 2022 and 2023, the model produced more stable price change rates compared to the actual historical data. Conversely, in 2020 and 2021, the model-generated price change rates exhibited greater volatility than observed in reality. This discrepancy is likely attributable to the COVID-19 pandemic, which led to a decline in rice demand and subsequently reduced rice prices.
See the authors' detailed response to the review by Katsuhiko Takahashi
See the authors' detailed response to the review by Amelia Santoso
Rice is a staple food with a high level of consumption in Indonesia. In 2017, the average national per capita rice consumption reached 97.6 kg per year, significantly exceeding the average consumption of meat (2.5 kg per capita per year) and poultry (7.5 kg per capita per year). Corn, often considered an alternative carbohydrate source, was consumed at an average of only 2 kg per capita per year (Arifin, et al., 2018). Given Indonesia's growing population, rice demand is projected to rise in the coming years. From a population of approximately 275 million in 2022, Indonesia is expected to see an increase to 294 million in 2030 and 319 million in 2045 (BPS-Statistics Indonesia, 2018). This population growth will likely lead to greater demand for staple foods, particularly rice. Arifin et al. (2018) forecasted that per capita rice consumption in Indonesia will increase by 14% to 101.5 kg per year by 2025 and by an additional 1% to 102 kg per year by 2045.
Such trends highlight rice's critical role in national economic stability, food security, and political stability. Ensuring a stable rice supply, characterized by consistent availability and affordable prices, is essential. However, recent trends indicate a decline in rice production. According to estimates from the Area Sample Framework (Kerangka Sampel Area, KSA) conducted by the National Statistics Bureau (BPS-Statistics Indonesia), national rice production decreased from 59.2 million tonnes in 2018 to 54.6 million tonnes in 2019. This figure remained relatively unchanged in 2020 at 54.6 million tonnes but further declined to 54.4 million tonnes in 2021 (BPS-Statistics Indonesia, 2022b).
The combination of increasing demand and declining production has disrupted the balance of rice supply and demand, leading to scarcity and rising prices. Typically, rice prices are lower during the harvest period (February to September) but increase toward the year's end due to reduced supply. However, between September 2022 and September 2023, rice prices in traditional markets displayed a consistent increasing trend.
In response to these challenges, the government has implemented policies to stabilize rice prices, including the minimum purchase price for farmers and the highest retail price (Harga Eceran Tertinggi, HET) for consumers (Hermanto, 2017). With rising inflation, rice prices have continued to climb, prompting the National Food Agency (Badan Pangan Nasional) to issue updated regulations on HET based on rice quality and trade location, as outlined in the National Food Agency Regulation No. 7 Year 2023.
To manage national rice stocks, ensure food security, and stabilize prices (Bappenas, 2019), the Indonesian government has appointed a state-owned enterprise responsible for these objectives. When retail rice prices exceed the HET, this entity intervenes by importing rice. This study aims to develop a model to determine the required quantity of rice imports to minimize the rate of increase in retail rice prices between consecutive months.
Over the years, research on commodity distribution, price prediction, and price stabilization has evolved significantly. Early studies primarily focused on supply chain network design and optimization. Melo et al. (2006) explored dynamic supply chain network design, focusing on relocating, adding, or reducing distribution center capacity. This study sought to optimize facility operations, capacity movement, and the quantity of goods produced, stored, or delivered. Min et al. (2007) and Li & Bing (2007) studied supply chain equilibrium under varying demand conditions, examining how product distribution among factories, retailers, and customers affects overall network performance. Wu & Zhang (2014) further investigated multi-commodity supply chains, focusing on cost reduction through optimized distribution center locations and retailer allocations. Teng et al. (2007) focus on developing a multi-commodity flow supply chain network equilibrium model that accounts for random demand. Afterwards, Xu and Zhu (2007), and Dellaert et al. (2021) extend the multi-commodity flow supply chain network equilibrium model by incorporating stochastic choice.
Athanasiou et al. (2008) focused on price stabilization in commodity markets through a nonlinear Cobweb model. The study showed that government interventions, such as adjusting stock levels, could stabilize prices. Paksoy et al. (2012) presented a fuzzy multi-objective linear programming model to minimize transportation costs for commodity distribution across multiple stages in the supply chain. Gouel (2013) and Dawe & Timmer (2012) further explored government policies aimed at stabilizing food prices, highlighting the role of subsidies, stock management, and trade regulations in minimizing price volatility.
Other studies concentrated on supply chain optimization under specific conditions. Serra & Gil (2012) found that blending biodiesel with diesel could mitigate fuel price fluctuations in Spain. Dorosh & Rashid (2013) examined rice price stabilization in Bangladesh, finding that stock management and trade interventions were key to price stabilization during crises. Similarly, Possamai et al. (2015) used a spatial-temporal network framework to model price stabilization for agricultural commodities, emphasizing the importance of interseasonal storage.
Several studies, such as Mogale et al. (2017), Gholamian and Taghanzadeh (2017), and Cheraghalipour et al. (2019), focused on minimizing logistics costs, but did not explore the interaction between product distribution quantities and market prices, an important factor for controlling price inflation. In contrast, Fitrawaty et al. (2023) investigated how domestic production levels, exchange rates, international prices, and GDP affect rice prices in Indonesia, providing policy insights but not actionable strategies for price control.
Research also examined various computational models for optimizing supply chain performance. Zhou et al. (2018) and Boccia et al. (2018) explored vehicle route optimization and facility location for multi-commodity distribution. Gu et al. (2021) and Guimaraes et al. (2019) combined inventory policies with route optimization to minimize distribution costs. Studies by Mohamed et al. (2023) and Zhang et al. (2022) addressed the complexities of stochastic demand in e-commerce, using algorithms to optimize facility locations and vehicle routes.
In terms of price prediction and stabilization, several studies have developed models to forecast commodity prices and mitigate fluctuations (Lin and Xu, 2019; Kwas et al., 2022; Elyasi & Teimoury, 2022). Ohyver & Pudjihastuti (2018) employed an ARIMA model to forecast rice prices in Indonesia, while Mele et al. (2020) explored suboptimal price stabilization as central banks influence market expectations. Anggraeni et al. (2019) combined hybrid neural network and ARIMA models to predict rice prices, Fitrawaty et al. (2023) developed a regression model that takes into account various factors influencing rice market prices. Menawhile, Maulana et al. (2023) created a model to optimize the equilibrium within the rice supply chain network using the Method of Successive Average (MSA).
While existing studies have made significant contributions to supply chain optimization and the analysis of price fluctuations, there remains a notable shortcoming in research that integrates rice distribution quantities with price control mechanisms capable of generating immediate effects. To address this shortcoming, the present study aims to develop a quantitative model for determining the optimal monthly volume of rice imports necessary to balance supply and demand, thereby minimizing increases in retail rice market prices. By optimizing import volumes, the GE authorized for rice imports can ensure a stable rice supply while maintaining its affordability as a staple food, particularly in developing economies where rice is a primary dietary component.
The primary model to be developed is an optimization model aimed at minimizing monthly changes in rice retail prices over two consecutive months. This is achieved by determining the quantity of rice imports necessary to control fluctuations in monthly rice retail prices, subject to 10 constraints governing the rice supply chain mechanism from sources to markets. To support the development of this optimization model, several auxiliary models are required.
The rice supply chain in Indonesia commences with farmers who cultivate rice during the harvest season, producing Harvested Dry Rice and Ground Dry Rice. The harvested rice is subsequently sold to intermediaries, who then distribute it to domestic rice mills. These mills process the rice using GKG as the primary raw material. The processed rice is acquired by private distributors and government enterprises for domestic rice procurement. Private distributors then sell the rice to retail traders, including supermarkets and traditional markets, which serve as the primary purchase points for rice consumers in Indonesia.
In procuring rice, government enterprises have two sources: domestic rice mills and rice imports. The rice supplied by government enterprises is stored and subsequently distributed to consumers to meet their demand. When the retail price of rice in the market is high, government enterprises conduct market operations by increasing the distribution of rice to the market or importing rice from source countries to enhance rice availability. Consumers purchase rice from government enterprises through traditional markets. This study models the rice supply chain network, which includes domestic rice mills, import source countries, national rice warehouses owned by government enterprises, and traditional markets as points of purchase for consumers. Figure 1 illustrates the rice supply chain network in Indonesia and the scope of the modelled rice supply chain network in this study.
Fitrawaty et al. (2023) stated that rice prices are influenced by several factors, including the followings.
In addition, according to Rusono (2019), rice prices in Indonesia are also influenced by the availability of rice, which includes rice stocks as well as the mechanism for its procurement and distribution of rice. Some of the activities carried out by government enterprise to maintain rice price stabilization in Indonesia include (Rusono, 2019):
The increase in rice supplies in Indonesia, encompassing domestic procurement, rice imports, distribution to consumers, and stocks managed by government enterprises, can lead to a reduction in rice prices. The availability of rice in Indonesia is also significantly influenced by the extent of agricultural land; larger areas of rice farming correspond to higher production levels, thereby increasing rice availability for consumption. According to Makbul & Ratnaningtyas (2017), the prices of Harvested Dry Rice (GKP) and Ground Dry Rice (GKG) per kilogram also play a critical role in determining the retail market price of rice. An increase in GKP and GKG prices per kilogram directly leads to a rise in retail rice prices.
Based on the aforementioned factors, a compound linear regression model with 10 independent variables, representing these determinants, can be used to predict rice prices for a given month. The model parameters, specifically the intercept and slope coefficients, were estimated using the Ordinary Least Squares (OLS) method. To develop this predictive model, historical data on rice prices and the factors influencing them are required, as outlined in Appendix A.
The variables of GDP per capita per month, domestic rice procurement, rice imports, and rice distribution are readily available, but require estimation. The monthly GDP per capita for Indonesia is estimated by dividing the annual GDP per capita (at current prices) by 12 (BPS-Statistics Indonesia; 2022a). Monthly domestic rice procurement by government enterprises for the period January 2018–December 2022 is approximated using historical data on the monthly proportion of domestic rice procurement to national rice production from January to December 2017 (Rusono, 2019) and historical data on monthly rice production for the same period from the National Statistics Bureau (BPS-Statistics Indonesia; 2022b).
Table 1 summarizes the national rice production, rice procurement, and the proportion of rice procurement to national production for January–December 2017. Historical data on monthly rice production for January 2018–December 2022 is provided in Table 2.
The monthly proportion of rice procurement to national rice production from 2017 is applied to estimate domestic rice procurement for January 2018–December 2022 using Equation (1). Similarly, monthly rice import data (BPS-Statistics Indonesia, 2022c) for the same period is estimated using Equation (2)
To estimate the monthly rice distribution by government enterprises for the period January 2018–December 2022, the 2018 rice distribution data (BPS-Statistics Indonesia, 2022d), totaling 1,449,000 tonnes, was used as a baseline. Additionally, national rice consumption data for the 2018–2022 period, obtained from the National Statistics Bureau (BPS-Statistics Indonesia, 2022e), is presented in Table 3.
Year | National rice consumption (tonnes) |
---|---|
2018 | 25,792,716 |
2019 | 25,214,232 |
2020 | 25,348,135 |
2021 | 25,596,828 |
2022 | 25,941,627 |
Based on rice distribution and national rice consumption data for 2018, the proportion of rice distribution to consumers relative to national rice consumption was calculated using Equation (3), yielding a value of 0.056. After determining this proportion, monthly national rice consumption data for the period January 2018–December 2022 was required. However, actual monthly data for this period was unavailable.
To estimate the required data, the average monthly rice consumption for the 2018–2022 period was first calculated using Equation (4). The results, including the average monthly rice consumption and the annual standard deviation for the 2018–2022 period, are presented in Table 4. To derive the standard deviation of monthly rice consumption, Equation (5) was applied under the assumption that the standard deviation remains constant throughout the 2018–2022 period.
Year | Average National Rice Consumption per Month (tonnes) |
---|---|
2018 | 2,149,393 |
2019 | 2,101,186 |
2020 | 2,112,345 |
2021 | 2,133,069 |
2022 | 2,161,802 |
Standard Deviation | 25,133.1 |
With the estimated average monthly rice consumption and its standard deviation, the monthly national rice consumption data for January 2018–December 2022 was generated, assuming the data follows a normal distribution. The estimated monthly national rice consumption values for this period are presented in Table 5.
Once the monthly national rice consumption data for the period January 2018–December 2022 is obtained, the monthly rice distribution data for the same period can be estimated using Equation (6). For months without rice imports, the rice distribution data is estimated using Equation (7)
The compound linear regression model was built using 60 training data, namely time series data in the period January 2018 - December 2022. The dataset used in developing the model is available from the authors upon request. The rice price prediction model formed is shown by Equation (8)
The coefficient value, t-statistic, and p-value of each variable can be seen in Table 6.
The calculated F-statistic of this compound linear regression model is 17.66 which is greater than the critical F-value = 1.99 for 60 training data and 10 independent variables. The p-value of the resulting calculation of F-statistic is <0.001 which is less than α = 0.05, indicating that the 10 independent variables together significantly affect the price of rice in a month. The R-squared value obtained from this model is 0.783. This indicates that the ten independent variables collectively explain 78.3% of the variance in the dependent variable. As three independent variables has Variance Inflation Factor (VIF) higher than 10, the model was revised by eliminating the variables and re-calculate the statistics of the revised model.
The new compound linear regression model is as follow
The coefficient value, t count, p-value, and VIF value of each independent variable of the revised model can be seen in Table 7.
By using the significance value of α = 0.05, the critical t value is obtained = 1.671. If the calculated t value of each variable is compared with the critical t value, it can be seen that the independent variables that significantly affect the price of rice are:
1. Gross Domestic Product per capita in month t (PDBt)
2. Ground Dry Rice price per kg in month t (GKGt)
3. Rice stock at the end of month t.
In alignment with the research objective of stabilizing rice price fluctuations through import mechanisms, this study emphasizes the overall validity of the model rather than the marginal contributions of individual independent variables. The model's validity is demonstrated as follows: The computed F-statistic for this compound linear regression model is 22.07, exceeding the critical F-value of 2.17 for a sample size of 60 training observations and 7 independent variables. The p-value associated with the F-statistic is less than 0.001, which is below the significance threshold of α = 0.05, indicating that the combined effect of the 7 independent variables significantly influences monthly rice prices. Furthermore, the model exhibits an R-squared value of 0.748, suggesting a strong explanatory power. Additionally, the Variance Inflation Factor (VIF) values for all independent variables are below 5.0, confirming the absence of multicollinearity in the developed rice price prediction model.
The objective function aims to minimize the rate of increase in rice prices between consecutive months, specifically between the current month (t) and the previous month (t−1), as expressed in Equation (9). The rice price for month is predicted using a compound linear regression model with 10 independent variables, as represented in Equation (8)
The set of constraints that define the flow of rice in the rice supply chain network of government enterprise can be described as follows
Constraint (11) ensures that the quantity of rice procured by government enterprise from domestic rice mills in month t does not exceed the production capacity of domestic mills in month t. Constraint (12) guarantees that the quantity of rice imported from all import source countries in month t does not exceed the total rice export capacity of all import source countries to Indonesia.
Constraint (13) ensures that the rice distributed by government enterprise in month t comes from the procurement of domestic rice in month of t, the import of rice in month t, as well as the difference between the rice stock at the end of the previous month (t-1) and the lower limit of the strategic inventory. Constraint (14) represents the balance of rice flow in the rice warehouse of a government enterprise. Rice stocks at the end of month t are rice stocks at the end of the previous month (t-1) plus domestic rice procurement in month t plus rice imports in month t minus rice distribution by government enterprise in month t.
Constraint (15) states that the quantity of rice stored in the national rice warehouse at the end of month t is not less than the lower limit of the strategic inventory of rice in a month and does not exceed its storage capacity. Constraint (16) ensures that the amount of rice distributed by government enterprise in month t is equal to consumer demand for rice in month t. Finally, Constraints (17) to (20) ensure non-negativity for the variables representing domestic rice procurement, rice imports, rice distribution, and rice stock at the end of month t.
The value of consumer rice demand parameters is estimated using the triple exponential smoothing method, based on monthly rice distribution data for the period January 2018 - December 2022. The procedure is outlined as follows:
1. Setting the value of initials Ft, Trt, and Bt.
2. Calculating the value of Ft, Trt, and Bt for each actual data.
3. Estimating variable value using actual data.
4. Calculating the value of variable in the future.
The value of consumer rice demand parameters in January–December 2023 can be seen in Table 8.
The lower limit parameter of the strategic inventory of government enterprise is estimated based on the minimum monthly rice stock value at the end of each month during the period January 2018 - December 2022, which is 326,763 tonnes. This value is rounded to 300,000 tonnes. Similarly, the rice storage capacity parameter is estimated using the maximum monthly rice stock value (Satudata Indonesia, 2022) at the end of each month during the same period, which is 2,418,511 tonnes, rounded to 2,400,000 tonnes.
The rice production capacity of domestic rice mills for each month (t) in the period January–December 2023 is estimated using the maximum domestic rice production for each corresponding month (t) during the period 2018–2022, calculated using Equation (21). The estimated rice production capacity of domestic mills for January–December 2023 is presented in Table 9.
The parameter for the rice export capacity of the source country of import to Indonesia is estimated using historical rice export data from Indonesia's seven largest rice import source countries. Table 10 shows rice export data from these seven countries to Indonesian during the period 2018 – 2022, measured in units of tonnes, as obtained from the National Statistics Buerau (BPS-Statistics Indonesia, 2022b; UN Comtrade, 2024).
The total rice exports from all source countries during this period amounted to 109,238,309.19 tonnes. The average annual rice import between 2018 and 2022 are calculated to be 21,847,661.94 tonnes. To estimate the monthly rice export capacity of the seven source countries to Indonesia, the average annual rice import volume during the 2018–2022 period is divided by 12 months and rounded to the nearest value, resulting in a monthly export capacity of 1,820,638.49 tonnes, rounded to 1,820,000 tonnes.
For the rice production variables, including the rupiah exchange rate against the dollar, the area of agricultural land (BPS Statistics Indonesia, 2022f), the price of GKP per kg, the price of GKG per kg (BPS Statistics Indonesia, 2022g), and domestic rice procurement, the values are estimated using the triple exponential smoothing method, based on actual data from January 2018 to December 2022. For the rupiah exchange rate against the dollar, the value is determined using actual data from January 2019 to December 2022. The methodology for estimating these values follows a similar approach to that used for estimating consumer demand parameters.
The value of the GDP per capita variable in month t, PDBt, is estimated using the double exponential smoothing method, based on annual GDP per capita data at current prices from 2014 to 2022. Table 11 presents the GDP per capita value for Indonesia at prevailing prices during the period 2014–2022.
The procedure is outlined as follows:
1. Setting initial value of Ft and Trt.
2. Calculating the value of Ft and Trt for each actual data.
3. Estimating variable value using actual data.
4. Calculating the value of variable in the future.
5. The GDP per capita per month variable is obtained by dividing the estimated GDP per capita by 12 months.
After taking steps 1–5, the estimated monthly GDP per capita for the period January–December 2023 is IDR 6,399,433.59. The rice import variable for month t, Yt, is determined based on two conditions. If the rice stock at the end of the month It is less than the lower limit of the strategic inventory ss, imports are required to bring the rice stock to the strategic inventory threshold. However, if the rice stock at the end of the month It is greater than or equal to the lower limit of the strategic inventory ss, no imports are necessary. The change in the value of rice stocks at the end of the month, prior to any imports, is due to an increase in stocks from domestic rice procurement and a reduction in stocks for distribution to consumers. Mathematically, Yt can be calculated using the Equation (22).
The rice stock at the end of month t, It, is determined based on the inflow of rice into and outflow of rice from rice stock in month t-1 or one month before. Rice stocks at the end of month t are rice stocks in month t-1 or one month before plus rice inflows that come from domestic rice procurement and rice imports, then minus rice distribution to meet consumer demand. Mathematically, It can be calculated with the Equation (23). The variable of rice distribution in month t, Ot, is set to be the same as consumer demand for rice in month t, dt. Mathematically, Ot can be calculated with the Equation (24)
Evaluation of model behavior is conducted to assess the extent to which the model accurately replicates the actual behavior of the system. A standard approach involves analyzing simulation outputs, comparing them against historical data over time, and calculating the average deviation between simulated and observed values. The primary performance metric employed in this evaluation is the Mean Squared Error (MSE), along with the proportion of MSE attributed to bias (UM), variance in equality (US), and covariance inequality (UC). MSE, UM, US, and UC can be calculated using the equation (25)–(28) successively.
where,
X(a) = actual data
X(m) = model output
n = number of data
The smaller the value of MSE, the better the model is at representing the real system.
MSE can be decomposed into three separate components which are called ad Theil inaquality statistics (Morecroft, 2015). Theil statistics measure the portions of MSE explained by the three components, namely:
1. Bias UM represents the difference between mean of actual data and mean of model outputs. It is calculated using equation (26).
2. Variance inequality (US), measures the difference between standar deviation of actual data and that of model outputs. Equation (27) is used to calculate US.
3. Covariance inequality (UC), represents unexplained variation between actual data and model outputs.
Each of the following three equations is used to calculate UM, US, and UC respectively. Sum of the three components is 1
where,
= mean of actual dat
where,
= standard deviation of actual data
= standard deviation of model ouputs
where r is correlation coefficient between actual data and model output, and calculated using the following equation
where,
= actual data i
= model ouput i
In this section, the results related to the problem of determining the quantity of rice imports each month are presented. The simulation period spans 12 months over 4 years: January–December 2023, 2022, 2021, and 2020.
The model recommends no rice import in 2020, the GE interventions were in forms of domestic rice absorption, rice distribution, and stock storage. With this scheme, the developed price prediction model during the 2020 period resulted in rice a relatively fluctuating rate of rice price. The standard deviation of the predicted price change was 0.40%, which was higher than the standard deviation of the actual price variation of 0.24%. In 2020, rice prices tended to decline due to the impact of the COVID-19 pandemic that emerged after March 2020, with its effects carrying over into 2021. Figure 2 illustrates the behavior of the model compared to actual monthly rice price data and the monthly price increase over two consecutive months.
Based on table in Figure 2, the highest actual rice price change rate observed in 2020 was 0.54%, with a standard deviation of 0.24%. The developed model predicts a highest price variation rate of 0.57%, close to the actual one. However, the price change rate predicted by the model exhibits greater fluctuation, as evidenced by a standard deviation of 0.40%. This fluctuation can be attributed to the effects of the COVID-19 pandemic, which began impacting Indonesia after March 2020, with lingering effects into 2021. These effects led to a decrease in rice demand and a subsequent reduction in rice prices. Despite this, the predicted rice price increase rate remains lower than the expected range of 2–3%.
The behavior of monthly rice prices can be further analyzed using the U M, U S, and U C components of the Mean Squared Error (MSE). As shown in Table 12, the differences between actual and predicted monthly rice prices in 2020 are primarily attributed to discrepancies in the mean values of the predicted and actual prices, rather than differences in standard deviations.
The model recommends a total rice import volume of 332,311 tons for 2021, whereas actual imports from the seven main rice-exporting countries amounted to 406,981.3 tons. The actual import volume includes special varieties of rice not produced domestically, which are permitted to be imported by private companies. The model suggests imports of 44,473.95 tons in March, 111,703.72 tons in November, and 176,133.02 tons in December, respectively. Incorporating rice imports along with other forms of government intervention, the model forecasts monthly rice prices for 2021. Figure 3 presents the actual and predicted monthly rice prices, as well as the rate of price change over two consecutive months.
In 2021, the COVID-19 pandemic continued to impact Indonesia, leading to a decrease in rice demand and a reduction in rice prices per kilogram. Based on the table in Figure 3, the highest actual rice price change rate observed in 2021 was 0.29%, with a standard deviation of 0.15%. The developed model predicts a highest price change rate of 0.80%, and the price change rate predicted by the model exhibits greater fluctuation, as evidenced by a standard deviation of 0.41%. This phenomenon occurred due to the effects of the COVID-19 pandemic which resulted in a decrease in rice demand and a decrease in rice prices. Despite this, the predicted rice price change rate remains lower than the expected range of 2–3%. The variation Paramaters of actual and predicted monthly rice price for the period of 2021 is presented in Table 13. Similar to the results for 2020 period, the differences between actual and predicted monthly rice prices in 2021 are primarily attributed to discrepancies in the mean values of the predicted and actual prices, rather than differences in standard deviations.
Application of the model for the year 2022 yielded a total recommended rice import volume of 254,905.08 tons, allocated across two shipments in March: 124,259.5 tons and 130,645.58 tons, respectively. This volume is substantially lower than the actual rice imports recorded in 2022, which totaled 428,843.2 tons. Nevertheless, despite the lower import recommendation, the model’s predicted monthly rice prices closely matched the actual observed prices throughout 2022. A detailed comparison of the predicted and actual monthly rice prices, along with their respective rates of change, is illustrated in Figure 4.
Compared to the outputs of model for 2020 and 2021, the MSE value resulting from the comparison between predicted and actual rice prices is much lower IDR 20,509.04, indicating that the developed model represents the actual price in 2022 quite accurately. This MSE is mostly due to difference in standard deviation between the predicted and actual prices, while their mean values are relative the same. The paramaters of variation between predicted and actual monthly rice price for 2022 period is presented in Table 14.
Parameter | Value |
---|---|
MSE | Rp 20,509.04 |
<0.01 | |
0.55 | |
0.45 | |
1.00 |
In 2022, the COVID-19 pandemic began to subside in August, which caused rice demand to increase and rice prices per kg to increase. Based on table in Figure 4, the highest actual rice price increase rate observed in 2022 was 2.22%, with a standard deviation of 0.82%. In comparison, the developed model results in a more controlled rice price increase, with the highest rate reaching 1.13% and a standard deviation of 0.49%. These values are notably lower than the expected increase rate of 2–3%.
The estimated rice import requirement for 2023 increased sharply, reaching a total of 1,033,392.11 tonnes. The imports were distributed across several months, including 279,313.43 tonnes in January, 234,887.87 tonnes in February, 194,608.20 tonnes in March, 140,088.17 tonnes in November, and 184,494.44 tonnes in December. Dibandingkan dengan volume import beras aktual pada 2023, which amounted to 3,049,924.4 tonnes (BPS-Statistics Indonesia, 2024), volume impor yang direkomendasikan oleh model jauh lebih rendah, tetapi jumlah yang direkomendasikan oleh model telah memenuhi konstrain including local production capacity, import limits, rice flow balance, and demand fulfillment.
According to the National Food Agency of Indonesia (BPN-National Food Agency, 2024), the sharp increase in rice imports in 2023 was an unavoidable measure necessitated by a decline in national rice production, primarily due to climate change and the effects of El Niño. This reduction in domestic rice production has raised concerns about a potential monthly rice balance deficit in early 2024. Consequently, the Government of Indonesia, through the National Food Agency (NFA), commissioned GA to import 2 million tons of rice, with an additional 1.5 million tons in 2023.
Beyond the impact of El Niño, other factors contributing to the increased demand for rice and the surge in retail rice prices include: (1) large-scale social assistance distributed to 10 million households in the context of the general election, and (2) inefficiencies in the Food Information System, which hindered accurate forecasting and disrupted supply chains (Department of Agricultural Socio-Economics, UGM, 2024). The diminishing impact of COVID-19 also contributed to rising rice demand in 2023, further driving price increases.
Compared to the predicted monthly rice prices generated by the model, despite the government's policy to drastically increase rice imports, the actual monthly rice prices in 2023 are significantly higher from the predicted values. Figure 5 illustrates the behavior of actual and predicted monthly rice prices in 2023, while Table 15 presents the variability parameters.
Parameter | Value |
---|---|
MSE | Rp 3,248,225.36 |
0.83 | |
0.14 | |
0.03 | |
1.00 |
Based on table in Figure 5, the highest actual rice price increase rate observed in 2023 was 4.35%, with a standard deviation of 1.47%. In contrast, the model developed yields a more controlled rice price increase, with the highest increase rate reaching 1.14% and a standard deviation of 0.45%. These values are yet lower than the expected price increase rate of 2–3%. Further empirical investigation may necessary to determine why the substantial rice imports—more than three times the recommended volume—failed to mitigate price fluctuations as predicted by the model.
Sensitivity analysis was conducted by varying the lower limit of the strategic inventory (ss) to examine the influence of changes in model parameters on the model's solution. The lower limit values of the strategic inventory considered were 300,000 tonnes (set for the base model), 500,000 tonnes, 1,000,000 tonnes, and 1,500,000 tonnes. The results of sensitivity analyses are summarized in Table 16.
From 2020 to 2023, changes in the lower limit of the strategic inventory (ss) have minimal impact on the change rate of monthly rice price between two consecutive months, with the maximum price change rate ranging from 0.42% to 1.14% and the standard deviation of the price change rate ranging from 0.40% to 0.50%. These values remain significantly lower than the expected price change rate of 2–3%.
In 2022 and 2023, the price increase rates generated by the model solutions were more stable than the actual price increase rates. This stability is demonstrated by the smaller standard deviation of the price increase rates produced by the model compared to the actual rates, indicating that the developed model effectively controls the rate of rice price change between consecutive months.
However, in 2020 and 2021, the price change rates generated by the model solutions fluctuated more than the actual price change rates. This is reflected in the larger standard deviation of the price change rates produced by the model compared to the actual rates. This phenomenon can be attributed to the COVID-19 pandemic, which caused a decrease in rice demand and a subsequent decline in rice prices.
This study develops a model to determine the optimal quantity of rice imports required to minimize the rate of increase in rice prices between consecutive months. The model is formulated as a linear programming optimization problem, incorporating flow constraints within the rice supply chain network. These constraints include the capacity of domestic mills, the export capacity of rice from source countries, components of rice distribution, balance of rice flow, storage capacity, fulfillment of consumer demand, and non-negativity of variables.
The retail market price of rice is predicted monthly using a compound linear regression model with seven independent variables: the rupiah exchange rate against the US dollar, GDP per capita, the price of ground dry rice (GKG) per kilogram, domestic rice procurement, rice imports, rice distribution, and government-managed rice stock aimed at ensuring domestic availability and price stability. The model is trained using a dataset of 60 observations, covering the period from January 2018 to December 2022. The F-statistic of the model is 22.07, exceeding the critical F-value of 2.17 for 60 observations and seven independent variables, indicating that the independent variables collectively have a significant impact on rice prices. The model achieves an R-squared value of 0.748, demonstrating a strong level of explanatory power.
The values of each independent variable are estimated based on historical data patterns using appropriate forecasting methods. The rice production rate, rupiah exchange rate against the US dollar, agricultural land area, GKG price per kilogram, and domestic rice procurement from mills are estimated using the triple exponential smoothing approach. GDP per capita is estimated using the double exponential smoothing method, while rice consumer demand is projected using triple exponential smoothing. The quantity of rice distributed to consumers each month is determined to meet demand. End-of-month rice stocks are calculated as the sum of beginning-of-month stocks, domestic procurement, and imports, minus the distributed rice. Imports are triggered when end-of-month stocks fall below the strategic inventory lower limit, ensuring that stocks meet the minimum required threshold.
The model is applied to determine rice imports aimed at mitigating monthly price fluctuations for the years 2020, 2021, 2022, and 2023. The results indicate that the recommended import volumes closely replicate actual price behaviors for the years 2020, 2021, and 2022, with Mean Square Error (MSE) values of IDR 48,821.12, IDR 206,539.7, and IDR 20,509.04, respectively. In 2023, respecting the model’s constraints, the total volume of rice imports increased fourfold compared to 2022. The recommended import volume was able to maintain a maximum monthly price change rate of 0.41% and a standard deviation of 0.45%.
However, the recommended rice import volume for 2023 was significantly lower than the actual import volume, which amounted to 3,049,924.4 tonnes. Despite the higher actual import volume, the retail price of rice in 2023 was paradoxically higher than the predicted price for the same period, resulting in a large MSE value of IDR 3,363,769.44 between actual and predicted monthly rice prices. Several specific factors are identified as contributors to the unique conditions of 2023, distinguishing it from previous years (2020–2022). These factors include: (1) extreme El Niño weather conditions, (2) large-scale social assistance targeting 10 million households in the context of the 2024 general election, and (3) inefficiencies in the Food Information System, which impaired accurate forecasting and disrupted supply chains (BPN-National Food Agency of Indonesia, 2024; Department of Agricultural Socio-Economics, Universitas Gadjah Mada, 2024). Future models can be developed to incorporate these factors for improved predictive accuracy.
In conclusion, the model effectively controls the rate of change in rice prices between consecutive months through planned rice imports, achieving a maximum price increase rate of 0.42% to 1.14% and a standard deviation of 0.40% to 0.50%, which are lower than the expected rate of 2–3%. These findings demonstrate the model's effectiveness in determining the necessary rice import volumes required to stabilize prices in Indonesia.
Future research could expand the rice supply chain network to include upstream stakeholders, such as farmers and intermediaries. However, this would require accurate historical data on rice production and distribution from farmers to mills. Additionally, if reliable data on domestic rice procurement, rice imports, and rice distribution become available, the accuracy of the rice price prediction model can be further enhanced. Furthermore, social, environmental, and political factors could be incorporated to refine projections of independent variables influencing rice prices.
Ethical approval and consent were not required.
Third party ethics
The data used in this research are publicly available and were obtained from BPS-Statistics Indonesia: https://www.bps.go.id/en
The data used to estimate the parameters of the rice price prediction model is published under a CC0 Public Domain Dedication, which does not retain any rights to the data.
The authors would like to acknowledge Mrs. Endah Ayu Ningsih and Mrs. Nurhayati from the Indonesian Ministry of Trade, who contributed to the data tabulation required for conducting this research.
Views | Downloads | |
---|---|---|
F1000Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: production systems engineering, supply chain management, industrial engineering
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Production & Inventory Systems, Supply Chain Engineering, Humanitarian Supply Chain
Is the work clearly and accurately presented and does it cite the current literature?
Yes
Is the study design appropriate and is the work technically sound?
Yes
Are sufficient details of methods and analysis provided to allow replication by others?
Partly
If applicable, is the statistical analysis and its interpretation appropriate?
Partly
Are all the source data underlying the results available to ensure full reproducibility?
Yes
Are the conclusions drawn adequately supported by the results?
Yes
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Production & Inventory Systems, Supply Chain Engineering, Humanitarian Supply Chain
Is the work clearly and accurately presented and does it cite the current literature?
Yes
Is the study design appropriate and is the work technically sound?
Yes
Are sufficient details of methods and analysis provided to allow replication by others?
Partly
If applicable, is the statistical analysis and its interpretation appropriate?
Partly
Are all the source data underlying the results available to ensure full reproducibility?
Yes
Are the conclusions drawn adequately supported by the results?
Partly
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: production systems engineering, supply chain management, industrial engineering
Alongside their report, reviewers assign a status to the article:
Invited Reviewers | ||
---|---|---|
1 | 2 | |
Version 2 (revision) 05 Jun 25 |
read | read |
Version 1 20 Mar 25 |
read | read |
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Already registered? Sign in
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)