This chapter explains how the functioning of public stockholding programmes was modelled using the OECD-FAO Aglink-Cosimo model. It describes the data, the data sources, and the assumptions made to construct the baseline projections. This baseline scenario reflects a business-as-usual scenario that assumes current stockholding policies (as described in Chapter 1) are maintained in the selected Asian countries for the entire projection period (2018 to 2030).
The Economic Effects of Public Stockholding Policies for Rice in Asia
Chapter 2. Modelling public stockholding programmes for rice
Abstract
2.1. Modelling public stocks in Aglink-Cosimo
The main objective of this report is to examine the potential impacts of stockholding policies on domestic and international markets over the medium term. The analysis is conducted using the OECD-FAO Aglink-Cosimo model. This model is a comprehensive partial equilibrium model for global agriculture, which can simulate developments of annual market balances and prices for the main agricultural commodities.
The standard Aglink-Cosimo model does not include important policy variables related to public stockholding policies such as public stock norms or guaranteed procurement prices, nor does it distinguish between private and public stockholding. Furthermore, the stock equations are standardized for most countries.1
Using the information collected in Chapter 1, which describes in detail the specific functioning of public stockholding programmes for rice in eight Asian countries up to the end of 2016, the equations for these countries in Aglink-Cosimo have been adjusted to achieve the following objectives:
Incorporate the three possible procurement and distribution channels of public stocks.
Separate private from public stocks.
Include stock norms, procurement prices, and subsidised prices.
Annex 2.A gives an overview of Aglink-Cosimo, describes how stocks have been modelled in other dynamic models, and explains in detail the adjusted and new equations in Aglink-Cosimo. Since the descriptions in Chapter 1 reflect the functioning of public stockholding programmes up to and including 2016, any changes thereafter are not reflected in the model.
2.2. Assumptions for baseline projections
New variables were introduced in Aglink-Cosimo in order to model the functioning of public stockholding programmes and data needed to be collected for each of them. For certain variables, time series data were available for relatively long periods of time, while there were gaps in the historic series for other variables. This section describes the assumptions made to fill these gaps, as well as the assumptions made to extend the series over the Aglink-Cosimo baseline projection period (2018-2030).
Table 2.1 lists the new variables and indicates by country which information needed to be collected. It also lists other variables that are crucial for the modelling but are already available in Aglink-Cosimo (indicated by *). For these latter variables – namely total stock volumes, producer prices, consumer prices, export prices, import prices, and trade policies – historic time series as well as data for the projection period are available in the Aglink-Cosimo database.2
For the new variables related to the acquisition and release of rice from public stocks, the data requirements varied by country. For example, for a country that exclusively buys domestically-produced rice at prevailing market prices, it was not necessary to collect information on procurement prices. In addition, for each country data needed to be collected on the volumes of public and private stocks, and public stock norms.
Table 2.1. Data requirements for modelling
Note: * Indicates variables for which data are already available in the Aglink-Cosimo database.
The projection period covers 2018-2030 and the base year was set at 2017. Several rules were implemented when collecting data for the new variables and constructing the baseline projections:
The same source was used for historic data, i.e. the data were not combined from different sources. When multiple sources were available, the source with the most complete recent time series was selected.
Data were converted to milled equivalent.
For the projections of several variables a five-year average was used to capture the most recent developments in the country. For example, public stocks can be expressed as a share of total stocks and Aglink-Cosimo includes projections for total stocks. To extend public stock levels over the projection period, the average share of public stocks in total stocks over the last five years was calculated. This average share was then used to calculate public stock levels over the projection period. The five-year period ideally comprised the years 2013-2017. However, depending on data availability, an earlier or shorter period was used. The five-year average methodology was applied when recent data did not reveal any clear trend.
As is common practice when constructing baseline projections, it is assumed there are no policy changes over the projection period. This means that for the baseline projections, it is assumed that public stockholding programmes will function as described in Chapter 1.3
Chapter 3 examines how markets are affected by changes in these programmes. Other specific assumptions are described in detail below.
Stock volumes
There are four variables related to stock included in the model: total stock volume, public stock volume, private stock volume, and public stock norm. The Aglink-Cosimo database contains historic and projection data for total stock volumes. Historic public stock data were collected from the AMIS Market Database,4 official governmental sources, and USDA'S Global Agriculture Information Network (GAIN) reports. To extend public stock data over the projection period, the share of public stocks in total stocks over the last five years was calculated and the average of this share was then maintained over the projection period. Private stock data were calculated as the difference between total stocks and public stocks. Information on the public stock norm was converted into days of national food consumption. The most recent norm figure was used for the projections. Since projections of national food consumption are included in Aglink-Cosimo, the projections of the norm follow the same trend.
Figure 2.1. Total and public stocks of rice, averages for 2018-2030
Note: * The level of public stocks in China is not based on an actual number; instead it is artificially set to the equivalent of 52 days of domestic food consumption. The choice for this level of public stocks in China is explained in the next section. Public stock data for Japan and Korea refer to public stocks composed of domestic rice. To extend public stock data over the projection period, the share of public stocks in total stocks over the last five years was calculated and the average of this share was then used over the projection period.
Source: Historic data and projections on total stock volumes are obtained from the Aglink-Cosimo database. Historic data on public stock volumes come from different sources: USDA GAIN reports for Bangladesh and Korea; AMIS Market Database for Indonesia, India, and Thailand; the Ministry of Agriculture, Forestry and Fisheries (MAFF) for Japan; and the National Food Authority (NFA) for the Philippines.
Figure 2.1 illustrates the average levels of public and total stocks as well as the average share of public stocks in total stocks over the projection period. Since the official information on the level of public stocks in China is kept confidential, the numbers are not based on historic data. For the modelling purpose, it was assumed that China's public stock level is equivalent to 52 days of national consumption. This number was artificially selected for reasons explained in the next section.
There are no private stocks in Korea (GAIN-KS1804, 2018), which implies that total stocks equal government stocks. In India, almost all rice stocks are held by the government. The private sector has been marginalised from stockholding activities as a result of the government's huge public stock programme which has discouraged the private sector from participating. In addition, private stockholding is limited under the Essential Commodities Act (1955).
The public stock data for Japan and Korea in Figure 2.1 refer to public stocks composed of domestic rice. As described in Chapter 1, there are two types of public stocks in these countries. The first type is composed of domestically produced rice, while the second one is built with imported rice. The amount of rice in the second type of public stocks depends on the countries' respective Tariff Rate Quota (TRQ) levels. In recent years, the TRQs for both countries have been filled and it is assumed this will continue over the projection period. Since the amount of rice in the second type of public stock is assumed to remain constant, this stock is excluded from the modelling of public stock scenarios. Furthermore, in Japan public stocks of imported rice are rarely used for domestic food consumption and the amounts are relatively small. Any change in public stock levels under the scenario analysis in these two countries hence occurs strictly via the first type of public stocks (composed of domestically-produced rice). Consequently, all graphs and modelling results related to public stocks in Chapters 2 and 3 refer to the first type of stock. The second type of stocks is incorporated in the total stock volumes for both countries.
Public stock levels in Thailand are projected to be almost zero during the projection period. This reflects the government's stockholding policy in 2016, whereby it planned to liquidate remaining public stocks during 2017 and started to encourage the private sector to hold stocks (FAO, 2017a; GAIN-THA8020, 2018; GAIN-THA7011, 2018).
Procurement volumes
Rice for public stocks can be procured domestically or internationally. Historic data on domestic procurement is obtained from official government sources, OECD, and USDA GAIN reports. For most countries, projections for domestic procurement volumes are obtained by calculating the average share of procurement in total production during the five most recent years. This share is then applied to the rice production data (already in Aglink-Cosimo) to obtain procurement volumes for the projection period. Table 2.2 lists these shares and the sources for historic data on procurement.
In Japan, projections for domestic procurement were obtained differently. Specifically, domestic procurement is set by the government at 200 000 tonnes of brown rice per year. This is equivalent to 181 200 tonnes of milled rice, using the conversion factor of 0.906 (MAFF, 2015). There is no information on procurement volumes for either Thailand (no public stocks during baseline projection) or China.
Rice for public stocks can also be imported. For those countries that regularly use imported rice to build public stocks, the baseline projections were obtained by calculating the average share of imports in total procurement for the last five years and then applying this share over the projection period. The other countries in the study have occasionally imported rice to add to their public stocks. In the model, these imports are modelled to occur when the domestic to international price ratio exceeds a certain threshold.
Table 2.2. Sources for procurement data and average share of domestic procurement in total production (2018-2030)
Distribution volumes
Rice from public stocks can be released on the international market (exports) or on the domestic market at either market prices or subsidised prices. The total amount of rice distributed from public stocks each year was calculated residually. The formula used is:
The baseline projections used the historic information on the share of public stocks released through the three different channels. For example, if during the last five years a country released on average 10% of public stocks on the domestic market at the prevailing market prices and 90% on the domestic market at subsidised prices, then these same percentages were applied for the projections. Figure 2.2 shows these shares for the projection period. China and Thailand are not included in these graphs because China has no data on procurement and distribution volumes, and Thailand does not build public stocks of rice under the baseline scenario. India has historically exported rice from its public stocks, but these shares have been minimal in the most recent years. This explains why the share of exports is practically zero for India over the projection period.
Figure 2.2. Average relative shares of public stocks released on the domestic and international markets, 2018-2030
* Public stock data in Japan and Korea refer to public stocks composed of domestic rice. Relative shares of the different distribution channels during the projection period are based on historic shares. Release on the international market is minimal under the baseline for the countries in the figure.
Source: Historic relative shares from Food Corporation of India (2018) for India and GAIN Reports for Indonesia.
Prices
Six sets of prices are crucial for the modelling. Four are already included in Aglink-Cosimo, namely: producer, consumer, import, and export prices.5 The producer prices in Aglink-Cosimo are derived from wholesale prices, while the consumer prices are derived from retail prices. The producer and consumer prices in Aglink-Cosimo were cross-checked with wholesale and retail price series in GIEWS (2018) and, where necessary, adjustments were made to the Aglink-Cosimo data.
Historic information on procurement prices and subsidised consumer prices was collected from several sources. When multiple price series were available, one series was selected or an average was calculated. Table 2.3 explains how the procurement and subsidised prices were obtained and lists the sources.
For the baseline projections, the procurement price in each country is assumed to follow a trend reflecting its recent relationship with the domestic producer price. The procurement price is assumed to remain below the producer price. Subsidised consumer prices are assumed to remain constant in nominal terms, which is consistent with the overall trend in those countries. Figure 2.3 illustrates the relationship between the producer, consumer, procurement, and subsidised prices for each country in the study.
Figure 2.3. Consumer, producer, procurement and subsidised prices of rice, 2000-2030
Note: All prices are in milled equivalent. Shaded area indicates the projection period 2018-2030.
Source: Aglink-Cosimo for producer and consumer prices. Table 2.3 lists the sources for procurement and subsidised prices.
Table 2.3. Selected procurement and subsidised consumer prices
Note: Paddy prices were converted to milled equivalent.
2.3. Limitations of the model and data
When examining the baseline and other scenarios, it is important to keep in mind the limitations of the model and data. First, the modelling of stockholding programmes is based on the description given in Chapter 1, and thus this report does not take into account how stockholding policies changed in 2017 or thereafter. Second, there are many varieties of rice which differ considerably. Aglink-Cosimo does not separate between these varieties and uses a single aggregate figure. Third, the parameters in the model are not based on estimates, but are selected based on plausibility considerations and the literature. Fourth, even though stocks and other variables can vary significantly within a year, it is not possible to do any intra-annual analysis since the data and model are annual.
References
AMIS (2016), AMIS Policy Database, http://statistics.amis-outlook.org/policy/ (accessed 7 February 2018).
FAO (2017a), Rice Market Monitor July 2017, Trade and Markets Division Food and Agriculture Organization of the United Nations, vol. 20(2), pp. 1–35, http://www.fao.org/fileadmin/templates/est/COMM_MARKETS_MONITORING/Rice/Images/RMM/RMM-Jul17.pdf.
FAO (2017b), Commodity Policy Developments Tool, http://www.fao.org/economic/est/est-commodities/commodity-policy-archive/en/ (last accessed December 2017).
Food Corporation of India (2018), “Homepage of the Food Corporation of India” (FCI), http://fci.gov.in/ (last accessed February 2018).
Food Planning Monitoring Unit (FPMU), Ministry of Food, People's Republic of Bangladesh (2017) (Online consultation November 2017), Available at: http://fsnis.fpmu.gov.bd/dms/advanced.html#graphcontainer.
GAIN-BG6005 (2016), “Bangladesh: Grain and Feed Annual”, USDA FAS, 31 March.
GAIN-BG5003 (2015), “Bangladesh: Grain and Feed Annual”, USDA FAS, 5 May.
GAIN-BG4002 (2014), “Bangladesh: Grain and Feed Annual”, USDA FAS, 31 March.
GAIN-BG3004 (2013), “Bangladesh: Grain and Feed Annual”, USDA FAS, 28 March.
GAIN-BG2001 (2012), “Bangladesh: Grain and Feed Annual”, USDA FAS, 22 February.
GAIN-BG1101 (2011), “Bangladesh: Grain and Feed Annual”, USDA FAS, 4 March.
GAIN-CH17017 (2017), “China – Peoples Republic of: Grain and Feed Annual”, USDA FAS, 4 April.
GAIN-CH16027 (2016), “China – Peoples Republic of: Grain and Feed Annual”, USDA FAS, 8 April.
GAIN-CH15014 (2015), “China – Peoples Republic of: Grain and Feed Annual”, USDA FAS, 8 May.
GAIN-CH13015 (2013), “China – Peoples Republic of: Grain and Feed Annual”, USDA FAS, 29 March.
GAIN-CH8012 (2008), “China – Peoples Republic of: Grain and Feed Annual”, USDA FAS, 1 March.
GAIN-ID1707 (2017), “Indonesia: Grain and Feed Annual”, USDA FAS, 30 March.
GAIN-ID1610 (2016), “Indonesia: Grain and Feed Annual”, USDA FAS, 30 March.
GAIN-ID1412 (2014), “Indonesia: GOI New regulation on rice exports and imports”, USDA FAS, 6 May.
GAIN-ID1318 (2013), “Indonesia: Grain and Feed Annual”, USDA FAS, 11 April.
GAIN-KS1804 (2018), “Korea – Republic of: Grain and Feed Update”, USDA FAS, 29 January.
GAIN-KS1613 (2016), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 30 March.
GAIN-KS1511 (2015), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 1 April.
GAIN-KS1417 (2014), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 2 April.
GAIN-KS1323 (2013), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 2 April.
GAIN-KS1230 (2012), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 17 April.
GAIN-KS1121 (2011), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 6 May.
GAIN-KS1014 (2010), “Korea – Republic of: Grain and Feed Annual”, USDA FAS, 3 May.
GAIN-TH8020 (2018), “Thailand – Grain and Feed Update”, USDA FAS, 26 January.
GAIN-TH7011 (2018), “Thailand – Rice Market and Policy changes over the Past Decade”, USDA FAS, 18 January.
GAIN-TH6029 (2016), “Thailand – Grain and Feed Annual”, USDA FAS, 18 March.
GAIN-TH5030 (2015), “Thailand – Grain and Feed Annual”, USDA FAS, 19 March.
GAIN-TH4120 (2014), “Thailand – Grain and Feed Update”, USDA FAS, 4 December.
GAIN-TH4021 (2014), “Thailand – Grain and Feed Annual”, USDA FAS, 7 March.
GAIN-TH1035 (2011), “Thailand – Grain and Feed Annual”, USDA FAS, 22 March.
GAIN-TH9161 (2009), “Thailand – Price Insurance Starts to Replace Mortgage Scheme”, USDA FAS, 29 October.
GIEWS (2018) Food Price Monitoring and Analysis Tool, http://www.fao.org/giews/food-prices/tool/public/ (accessed February 2018).
Hertel T., J. Reimer and E. Valenzuela (2003), “Incorporating Commodity Stockholding Behavior into a Short-run General Equilibrium Model of the Global Economy”, Prepared for presentation at the joint meetings of the American Agricultural Economics Association, Canadian Agricultural Economics Society, and Rural Sociological Society, 27-30 July 2003, Montreal, Quebec.
Kozicka M., M. Kalkuhl and J. Brockhaus (2017), “Food Grain Policies and their Implications for Stocks and Fiscal Costs: A Dynamic Partial Equilibrium Analysis”, Journal of Agricultural Economics, Vol. 68(1): 98-122(25), https://doi.org/10.1111/1477-9552.12176.
MAFF (2016a), “Demand and Supply of Government-Owed Rice, Wheat and Barley” (in Japanese), Tokyo, http://www.maff.go.jp/e/tokei/kikaku/nenji_e/89nenji/index.html.
MAFF (2016b), “Current Situation of Rice” (in Japanese), Tokyo, Online available at: http://www.maff.go.jp/j/seisan/kikaku/pdf/ref_data3.pdf.
MAFF (2016c), Monthly Report of Rice (in Japanese), Tokyo, Online available at: http://www.maff.go.jp/j/seisan/keikaku/soukatu/pdf/mr160705.pdf.
MAFF (2015), Monthly Report of Rice, March 2015 (in Japanese), Tokyo, Online available at: http://www.maff.go.jp/j/seisan/keikaku/soukatu/attach/pdf/mr-20.pdf.
NFA (2018), Homepage of the National Food Authority, http://www.nfa.gov.ph (last accessed February 2018).
NFA (2017) NFA Palay Procurement 2011-2016, available at: http://nfa.gov.ph/images/files/statistics/palayprocurement.pdf.
NFA (2016a) NFA Palay Procurement 2006-2014, available at: http://nfa.gov.ph/files/statistics/palayprocurement06-14.pdf.
NFA (2016b) Summary of NFA Rice Import Arrivals 2000-2012, available at: http://nfa.gov.ph/files/statistics/summary.pdf.
NFA (2016d) NFA Total Rice Distribution 2006-2014, available at: http://nfa.gov.ph/files/statistics/ricedistribution06-14.pdf.
NFA (2015) Annual Accomplishment Report 2015, available at: http://www.nfa.gov.ph/files/Transparency/AAR2015.pdf.
NFA (2014) Annual Accomplishment Report 2014, available at: http://www.nfa.gov.ph/files/Transparency/AAR2014.pdf.
OECD/ICRIER (2018), Agricultural Policies in India, OECD Food and Agricultural Reviews, OECD Publishing, Paris, https://doi.org/10.1787/9789264302334-en.
OECD (2015), Managing Food Insecurity Risk: Analytical Framework and Application to Indonesia, OECD Publishing, Paris. http://dx.doi.org/10.1787/9789264233874-en.
OECD (2012), OECD Review of Agricultural Policies: Indonesia 2012, OECD Publishing. http://dx.doi.org/10.1787/9789264179011-en.
Reserve Bank of India (2017), Handbook of Statistics on Indian Economy 2016-17, available at: https://www.rbi.org.in/Scripts/PublicationsView.aspx?id=17803.
Annex 2.A. Adjustments in Aglink-Cosimo
Overview of the Aglink-Cosimo model
The analysis undertaken in this study uses the 2017 version of the Aglink-Cosimo model.6 Aglink-Cosimo is a recursive-dynamic, partial equilibrium model used to simulate developments of annual market balances and prices for the main agricultural commodities produced, consumed, and traded worldwide. It is managed by the Secretariats of the OECD and the Food and Agriculture Organization of the United Nations (FAO), and used to generate the OECD-FAO Agricultural Outlook and for policy scenario analysis. Aglink-Cosimo covers the whole world and projections are developed and maintained by the OECD and FAO Secretariats in conjunction with country experts and national administrations. Several key factors and assumptions are listed below.
World markets for agricultural commodities are competitive, with buyers and sellers acting as price takers. Market prices are determined through a global or regional equilibrium in supply and demand.
Domestically-produced and traded commodities are assumed to be homogeneous and thus perfect substitutes by buyers and sellers. In particular, importers do not distinguish between commodities by country of origin as Aglink-Cosimo is not a spatial model. Imports and exports are nevertheless determined separately. The homogeneity assumption extends to the rice market, but some heterogeneity is accounted for by assuming a lower substitutability of domestic production and imports in countries with high shares of Japonica rice consumption.
Aglink-Cosimo is a partial equilibrium model for the main agricultural commodities. Non-agricultural markets are not modelled (except for the biofuel sector) and are treated exogenously to the model. Therefore, hypotheses concerning the paths of key macroeconomic variables are predetermined and there is no feedback from developments in agricultural markets to the economy as a whole.
Aglink-Cosimo is recursive-dynamic. Consequently, each year is modelled over the projection period and depends on the outcome of previous years. Aglink-Cosimo simulates generally ten years into the future with a version available until 2030. This latter version is used in the current study.
The upper part of Figure 2.A1 provides a general overview of one domestic commodity market in Aglink-Cosimo with the main linkages between endogenous and exogenous variables.
Supply and demand are balanced by clearing for a domestic price. This price is an important explanatory variable in each of the supply and demand components. The supply side consists of three major elements: production, imports, and beginning stocks (which are equal to the ending stocks of the previous year). Production equations are not only influenced by the endogenous domestic price, but also by exogenous variables, such as energy prices, weather conditions, and domestic policies. Imports of a commodity are a function of the ratio between domestic and international prices as well as trade policies.
Demand is composed of food use, feed use, other use, exports, and ending stocks. Food use equations depend on demographics, real income developments, and the consumer price ratios between the commodities of the food basket. Feed use quantities depend on the price ratios between the different feed bulk commodities, but also on the production quantities of the animal products using this feed. Depending on the commodity, the category “other use” might be broken down into sub-categories,7 but in general includes other industrial uses and losses occurring along the supply chains. Exports are – like imports – a function of the ratio between international and domestic prices and of trade policies.
Figure 2.A1. Aglink-Cosimo model with stock equation extensions
Note: The rectangular shapes contain the endogenous variables while the oval shapes indicate the exogenous variables.
Modelling of stocks in dynamic models
Only recursive dynamic models can be used to trace the behaviour of annual stock variables. Comparative static models usually simulate a stable equilibrium that is not influenced by annual fluctuations and thus are unable to address short-term variations. In addition to Aglink-Cosimo, commonly used dynamic models are the global models developed by the Food and Agriculture Policy Research Institute (FAPRI) and by the Economic Research Service (ERS) of the United States Department of Agriculture, as well as some Computable General Equilibrium (CGE) models.
Most stock equations in Aglink-Cosimo have the same structure. Ending stocks are modelled according to the principle that stocks of a specific commodity are built when the price of this commodity falls below a certain threshold, when supply is high or demand is low. The price considered is the producer price and the threshold is the average producer price of the previous three years. Conversely, stocks are released when the price is higher than the threshold, supply is low or demand is high. Specifically, the stocks of a specific commodity in a specific country at a specific time are modelled as follows in Aglink-Cosimo (subscripts are omitted to improve the legibility):
Where:
ST: ending stocks
QP: quantity produced
QC: quantity consumed
PP: producer price (domestic clearing price)
TRD: trend
α, β: equation-specific coefficients
The above equation reflects private stockholding behaviour although the data in the model relates to total stocks, which includes both private and public stocks. For India and China, additional equations related to public rice stocks are specified. In India, stock procurement also occurs when stock levels are below a specified norm or when the producer price is below the minimum support price. In both cases, rice is assumed to be procured from domestic producers. China is modelled to import rice when its stock level falls below a specified norm.
In the FAPRI stochastic model, there is no single specification of the stock equations.8 For most crops, stocks are a function of the commodity price and of one or multiple supply variables. These supply variables differ across commodities and countries. In some cases only production is considered, while in others beginning stocks or imports are included. Production in the next year (period t+1) is used as an additional explanatory variable since it is assumed that in the months before the harvest the market actors already have a good estimate of the size of the crop. As a result, prices and stocks adjust accordingly. For example, in the case of a drought, the production in period t+1 is expected to be lower, which will motivate market actors to hold onto stocks, which in turn will drive up prices in period t. In addition, public stockholding policies are incorporated in certain cases. One example is the US nine-month loan programme which provides a storage subsidy. Separate equations for stocks under that programme exist and imperfect substitution with commercial stocks is assumed.
The USDA outlook model used at ERS represents stocks in different ways, depending on the country and the commodity. There are cases where stocks are just an identity and other cases where demand for stockholding is treated as a function of prices and trends. In the case of big importers, stocks are a function of consumption and domestic prices. Policy variables also occur in some cases (e.g. India). In general, stocks are drawn down when weather-induced supply shortfalls result in rising prices. The incidence of weather-induced movement in stocks is more prevalent in countries with highly variable precipitation, or no or limited irrigation capacity (e.g. North Africa). When such countries are modelled, yield is considered the best predictor of stocks. Unfortunately no publicly available documentation of these equations exists.
Most dynamic CGE models focus on long run growth effects and not on short run issues like stockholding behaviour. Hertel et al. (2003) introduced a simple stock behaviour equation as a function of expected revenues and costs from stockholding as well as max and min conditions on stock levels. The authors tried to improve the validation of CGE models, considered as one of their main limitations, by generating stochastic experiments and comparing the price and stock variations to those historically observed.
Even though some of the global models listed above distinguish to a certain extent between private and public stockholding behaviour, none contain the envisaged detail of stock modelling required for the analysis in this study. These requirements are: i) separate private from public stocks; ii) incorporate the three possible procurement and distribution channels of public stocks; and iii) include stock norms, procurement prices and subsidised prices. In the current literature only one model was identified which includes some of these requirements. Kozicka et al. (2016) developed a partial equilibrium model for rice and wheat to analyse the stockholding system in India. They introduced a distinction between private and public stocks, and model private stocks as a function of commodity supply and the level of public stocks. The way in which the authors incorporated how the level of public stock affects private stocks, i.e. the crowding out effect, has been incorporated in the revised stock equations in Aglink-Cosimo (see below). Their model also distinguishes between the procurement and distributional aspects of stockholding behaviour and incorporates policy variables such as minimum support prices, subsidised prices, and stock norms.
Adjustments to rice stockholding equations in Aglink-Cosimo
As explained above, the standard Aglink-Cosimo model does not include important policy variables such as stock norms or procurement prices, does not separate public from private stocks, nor does it distinguish between the different channels of procurement and distribution of public stocks. In addition, the stock equations are standardised for most countries. This section explains how the equations in Aglink-Cosimo have been adjusted to incorporate this information for the countries in this study.
The lower part of Figure 2.A1 summarises the extensions made for rice in the eight countries of this study. The ending stock equation is replaced by two separate equations: one for private stocks and another for public stocks. Private stocks are mainly market-driven as indicated by their link with the domestic clearing price, and are also influenced by public stocks through the crowding out effect. Public stocks have additional policy variables that determine their behaviour and are influenced by procurement activities, distribution activities, and loss. Procurement can occur on the domestic market (production) or the international market (imports), and distribution can also take place in the domestic market (food use) or the international market (exports).9 Losses from public stockholding activities are linked to the component “other use”, and, more specifically, to the sub-category of this variable that covers losses along the value chains.
In the equations below, the commodity and country subscripts are omitted to improve their legibility. All these equations apply to rice and are activated for the eight countries in the study. Furthermore, all the behavioural functions include a calibration factor which is not explicitly specified in the equations below. This factor is used to calibrate the model equations to the baseline projections for all stock-related variables. As explained in Chapter 2, these baseline projections were obtained using specific rules and are based on the 2017 version of the Aglink-Cosimo model. Hence, the values for the new stock-related variables were created according to the rules specified in Chapter 2 and the model equations (specified below) were calibrated to those values using the equation-specific calibration factors. More specifically, this calibration is done by plugging the projected values into the left hand sides of the equations, then solving the equations for the equation-specific calibration factors, and finally calculating their values.
Total stocks
Total stocks (ST) are the sum of private (PRST) and public stocks (PUST).
Private stocks
Private stocks follow the “buy low – sell high” principle. If the current market price (PP) is above the average of the previous three years, then stocks decrease and vice versa. In addition, it is assumed that public stocks can crowd out private stockholding to a certain extent, which means that private stocks decrease as public stocks increase. As mentioned above, this behaviour was based on Kozicka et al. (2016). However, whereas Kozicka et al. (2016) use available supply as an explanatory variable, this study uses the price ratio instead. Since supply and prices are linked – with higher supply triggering lower prices and vice versa – the price ratio is considered a better proxy for the potential benefits of buying and selling rice from private stocks.
Price ratio elasticities () average around -0.8 and crowding out elasticities ( range between -0.05 and -0.2 for the eight countries.
The crowding out behaviour of private stocks by public stocks has been linked to the stock norm, rather than to the total public stock level. The reason for this is that in a situation with a one year shock (as with the simulated yield shock in the analysis) it is most likely that the actors holding private storage realise that the reduced public stock levels are only a temporary deviation from the equilibrium and thus they do not start to increase procurement. Therefore, only price ratios explain private stock change in this situation.10
Public stock levels
Public stock levels depend on four components:
1. a behavioural equation (PUST..EQ) reflecting similar price behaviour as in the private stock case, but with lower elasticities,
2. a component that activates additional procurement (PRCU..ADD) as soon as the market price falls 10% below the procurement price (MSP),11
3. the 90% of the stock norm (PUSTN) as a lower bound,12 and
4. a component that allows an additional release during an emergency situation (PUSTD..EM), in which case stocks are allowed to go below the norm but not below 10% of the previous year's level of stocks.
In the central baseline projections it is assumed that none of the MAX conditions in the PUST equation are triggered so that PUST..EQ = PUST. MSP and PUSTN are part of the stockholding policy variables mentioned in Figure 2.A1.
The PUST..EQ equation assumes that public stock levels are dependent on the stock norm in a country. Therefore the PUST..EQ equation is defined relative to that norm. However, the relationship changes with the level of the norm following the assumption that at higher norm levels the relation is likely to weaken.
The price ratio elasticities () were generally taken from the former total stock equations in Aglink-Cosimo and average around -0.4. The elasticities range between -0.3 and 0 for the eight countries.
The additional procurement component reflects the assumption that the public sector buys any quantity at the guaranteed procurement prices (MSP):
The PRCU..ADD equation guarantees that additional procurement becomes completely price elastic when the market price drops below 90% of the MSP. That is, as soon as the market price is lower than 90% of the MSP, then any quantity will be procured in order to keep the price above that value. These types of equations are regularly used in Aglink-Cosimo to switch between a market that clears for prices to one that clears for a certain quantity – in this case additional procurement and implicitly total stocks. The choice of 90% of the MSP is motivated by the fact that market price support is unlikely to work perfectly and as a result it takes some time for government procurement to react to price drops.
Public stock releases
Total public stock releases (PUSTD) are a function of subsidised consumer prices (CP..SUB) relative to open market consumer prices (CP), the market price (PP) relative to its average over the past three years, the stock norm (higher norms increase distribution – but only slightly) as well as the population of a country (POP). A further restriction guarantees that the stock release is at least as big as the decrease in public stocks and that distributions do not occur if public stocks are zero. The latter is guaranteed through the last term (1/1+e^(…)) at the end of the PUSTD equation. This term evaluates to values close to but below 1 as long as public stocks are greater than 20 kt and evaluates to values close to but above 0 if public stocks approach 0 kt. For any volumes in between 20 kt and 0 kt, it smoothly reduces from values slightly below 1 to values slightly above 0. These types of sigmoid functions are widely used in equilibrium models to allow a variable to perform a ‘step’ around certain values. A classic example of this are applied tariffs which depend on the fill rate of tariff rate quotas (see equation 70 in OECD, 2015).
The elasticity is only relevant for countries that distribute at subsidised prices and averages around -0.1. The price ratio elasticity ( averages around 0.3 and the elasticity of the stock norm ( around 0.05. The population elasticity () was set at about 0.15.
Release from public stocks can occur through three channels: in the domestic market at market prices, in the domestic market at subsidised prices or in the international market at export prices. The countries in this study only use one or two of these three channels. For the countries that only have one channel of distribution, total distribution is set equal to the distribution through this channel. This means that total distribution is set equal to distribution at subsidised prices for Bangladesh and the Philippines, while total distribution is equal to distribution at domestic prices in Japan and Korea, and equal to exports in Thailand. Indonesia and India have a combination of subsidised and domestic distribution at market prices. For those two countries, the equation for subsidised distribution is exactly the same as the one for PUSTD..EQ, but elasticities are generally less reactive to price changes to ensure that subsidised release is more stable than release at market prices.
Emergency public stock releases can occur when rice availability is at risk. In the model, emergency stock releases are triggered when market prices increase above a certain threshold. This threshold is set at 5% above the reference price PP..REF, which corresponds to the price in the baseline scenario.13 In this case, a certain amount of public stocks will be released. The parameter γ identifies how much of the beginning stocks should be available for emergency releases. It can also – when set to zero – be used to deactivate the mechanism. In the yield shock scenario γ is chosen to be quite high at 95% making almost the entire beginning stocks available for emergency releases.
Public stock procurement
Public stock procurement (PRCU) is modelled residually: it is a function of public stocks’ change (which incorporates a loss factor) and distribution.
The additional procurement variable (PRCU..ADD) and total procurement (PRCU) are linked via the public stock equation. The public stock equation explicitly contains PRCU..ADD and since total procurement is calculated as shown above, PRCU..ADD enters into PRCU as well.
As is the case with distribution, countries use not more than two of the three existing procurement channels. For those that use only one procurement channel (Japan and Korea exclusively procure in the domestic market at prevailing market prices, while India exclusively procures in the domestic market at the MSP), total procurement is set equal to procurement through that one channel. The other countries use a combination of procurement at guaranteed prices and procurement from imports. Procurements from imports (PRCU..IM) are defined as a share of total procurement. This share is defined as:
The exponent δ defines the substitutability between the two procurement channels and μ is used to calibrate to observed shares. For some countries, the general instrument to procure rice from international markets does exist, but is rarely used (Bangladesh, India, Indonesia, the Philippines). In those countries, PRCU..IM was assumed to be zero when building the baseline framework. In order to make it possible to calibrate the PRCU..IM equation to those specific situations with zero observations, the μ parameter is chosen so that the fraction evaluates to a value smaller than 0.01 which causes the MAX operator to evaluate to zero. This specification hence allows scenarios in which procurement from imports will kick in when the relation between the domestic price and the sum of domestic and import price becomes larger than 0.01. The choice of 0.01 is arbitrary and could benefit from further sensitivity analysis in future.
Subsidised procurement is then the residual of total procurement and those from imports.
Notes
← 1. These equations are described in the documentation of the Aglink-Cosimo model, http://www.agri-outlook.org/abouttheoutlook/Aglink-Cosimo-model-documentation-2015.pdf
← 2. Aglink-Cosimo data were obtained from the OECD-FAO Agricultural Outlook 2017-2026 and extended to 2030.
← 3. Chapter 1 describes the functioning of public stockholding programmes up to the end of 2016.
← 5. Detailed information on these prices can be found in the documentation of the Aglink-Cosimo model, http://www.agri-outlook.org/abouttheoutlook/Aglink-Cosimo-model-documentation-2015.pdf
← 6. The documentation of the Aglink-Cosimo model can be consulted at http://www.agri-outlook.org/abouttheoutlook/Aglink-Cosimo-model-documentation-2015.pdf.
← 7. Examples are biofuel feedstock and the crushing of oilseeds.
← 8. The FAPRI model documentation is available at https://www.fapri.missouri.edu/wp-content/uploads/2015/02/FAPRI-MU-Report-09-11.pdf.
← 9. Procurement on the domestic market can occur at prevailing market prices or at procurement prices, while distribution on the domestic market can occur at prevailing market prices or at subsidised prices.
← 10. In general, private stocks should help mitigate price variability. However, the private stock levels are much lower than the public ones in several of the countries in this study, and therefore their variation cannot compensate that of public stocks.
← 11. The procurement price is modelled as a Minimum Support Price, which explains the choice for the acronym MSP.
← 12. Even though it initially seemed appropriate to use the norm itself as a lower bound for public stocks, historical data show that public stock levels fall below the norm from time to time. To reflect this, the lower bound in the model was set at 90% of the norm.
← 13. The threshold of 5% above the reference price was chosen after several trial and error simulations based on the 10% yield shock scenario. The stockholding policies of the eight countries can contain rules for emergency stock release, but usually there are no clearly defined triggers that explain when exactly emergency distribution will occur as the trigger is likely to be different depending on the situation.