Rice price stabilization in Bangladesh Assessing the impact of public farm-gate and consumer price stabilization policy instruments on the overall grain market and developing policy orientations with a greater role for the private sector Nicholas Minot, Shahadat Hossain, Razin Kabir, Paul Dorosh, and Shahidur Rashid INTEGRATED FOOD POLICY RESEARCH PROGRAM | WORKING PAPER 011 AUGUST 2021 CONTENTS Acronyms and abbreviations ......................................................................................................... 1 Chapter 1. Introduction .............................................................................................................. 2 Chapter 2. Background on food price stabilization ................................................................. 4 2.1 Measuring price instability ............................................................................................... 4 2.2 Sources of food price instability ....................................................................................... 4 Changes in demand .............................................................................................................. 4 Changes in supply ................................................................................................................. 5 International price volatility .................................................................................................... 5 2.3 International patterns in grain price instability .................................................................. 6 2.4 Patterns of grain price instability in Bangladesh ............................................................... 8 2.5 Economics of public grain reserves ............................................................................... 11 Economic arguments for public grain reserves .................................................................... 12 Research on storage and public grain reserves ................................................................... 13 International experience with public grain reserves ............................................................. 14 Public grain reserves in Bangladesh ................................................................................... 16 2.6 Effect of food price instability on welfare ........................................................................ 19 Chapter 3. Analysis of rice prices ........................................................................................... 23 3.1 Introduction.................................................................................................................... 23 3.2 Methods and data .......................................................................................................... 23 3.3 Results .......................................................................................................................... 25 Econometric characteristics of the data ............................................................................... 25 Econometric estimation ....................................................................................................... 33 3.4 Summary ....................................................................................................................... 38 Chapter 4. Analysis of relationship between rice prices and PFDS operations .................. 40 4.1 Introduction.................................................................................................................... 40 4.2 Methods and data .......................................................................................................... 40 Sources of Data .................................................................................................................. 40 Estimation method............................................................................................................... 41 4.3 Results .......................................................................................................................... 41 Descriptive analysis ............................................................................................................. 42 Effect of rice prices on public food distribution ..................................................................... 47 Effect of public food distribution on rice prices ..................................................................... 50 4.4 Summary ....................................................................................................................... 51 Chapter 5. Impact of policy options on rice price stability ................................................... 52 5.1 Introduction.................................................................................................................... 52 5.2 Methods - Description of the model ............................................................................... 53 Demand for and supply of rice ............................................................................................. 54 Commodity balance and prices ........................................................................................... 54 Private and public storage ................................................................................................... 55 International trade ............................................................................................................... 55 5.3 Data sources for the model ............................................................................................ 56 5.4 Results of the simulations .............................................................................................. 63 Baseline scenario ................................................................................................................ 63 Simulations of changes in scale of PFDS ............................................................................ 68 Simulations of changes in rice import tariff .......................................................................... 69 Simulation of adaptive buffer stock operation ...................................................................... 72 5.5 Summary ....................................................................................................................... 74 Chapter 6. Summary and conclusions ................................................................................... 76 6.1 Relationship among rice prices in Bangladesh .............................................................. 76 6.2 Relationship between rice prices and government purchases and distribution ............... 77 Descriptive analysis ............................................................................................................. 77 Impact of prices on PFDS procurement and distribution ...................................................... 78 Impact of PFDS on rice prices ............................................................................................. 78 6.3 Simulation of alternative policies and price stabilization ................................................. 78 Bangladesh Rice Market Model ........................................................................................... 78 Base scenario ..................................................................................................................... 79 Impact of changing the scale of rice procurement and distribution ....................................... 80 Impact of alternative rice import tariffs ................................................................................. 80 Impact of a buffer stock policy ............................................................................................. 81 Impact of an adaptive buffer stock policy ............................................................................. 81 Reducing price instability ..................................................................................................... 82 Appendix - A. Equations of the model ................................................................................... 83 About the authors ......................................................................................................................... 86 Acknowledgments ........................................................................................................................ 86 References .................................................................................................................................... 87 TABLES Table 2.1: Price instability in wholesale rice markets .................................................................. 7 Table 2.2: Price instability in wholesale wheat markets............................................................... 8 Table 3.1: Test for unit roots in rice prices and first differences .............................................. 29 Table 3.2: Engle-Granger test for cointegration ......................................................................... 34 Table 3.3: Johansen cointegration test ....................................................................................... 34 Table 3.4: Cointegrating vector.................................................................................................... 35 Table 3.5: Vector error correction model .................................................................................... 35 Table 3.6: Test for normality, serial correlation and heteroscedasticity ................................... 36 Table 4.1: Regression analysis of net distribution of rice by the PFDS .................................... 47 Table 4.2: Regression analysis of the gross distribution of rice by the PFDS ......................... 48 Table 4.3: Regression analysis of rice procurement by the PFDS ............................................ 49 Table 4.4: Regression analysis of real rice prices ...................................................................... 50 Table 5.1. Per capita consumption of rice by Division .............................................................. 56 Table 5.2. Population by Division ............................................................................................... 56 Table 5.3. Rice production by season and by Division in 2016 ................................................ 57 Table 5.4. Coefficient of variation of production (%) .................................................................. 58 Table 5.5. Cross-Division correlation of rice production by season ......................................... 59 Table 5.6. Monthly distribution of harvest ................................................................................. 60 Table 5.7. Description of storage rule function ......................................................................... 61 Table 5.8. Main results of base scenario .................................................................................... 67 Table 5.9. Simulation of the impact of changing the scale of rice PFDS ................................. 69 Table 5.10. Simulation of the impact of changing the rice import tariff ................................... 70 Table 5.11 Floor and ceiling prices for the buffer stock simulation ......................................... 71 Table 5.12. Simulation of the impact of a buffer stock with different price bands .................. 72 Table 5.13. Simulated impact of an adaptive buffer stock with different price bands ............ 74 FIGURES Figure 2.1: Bangladesh and international (Thai) rice prices ........................................................ 9 Figure 2.2: Bangladesh and international wheat prices ............................................................... 9 Figure 2.3: Moving coefficient of variation (CV) of Dhaka wholesale coarse rice price ........... 10 Figure 2.4: Moving coefficient of variation (CV) of Dhaka wholesale wheat price ................... 11 Figure 2.5: Diagram of effect of food price instability on household welfare ........................... 19 Figure 3.1: Coarse rice wholesale price for each Division ......................................................... 26 Figure 3.2: Coarse rice wholesale prices in levels and first-differences for Barishal, Chattogram, Dhaka, and Khulna .................................................................................................. 27 Figure 3.3: Coarse rice wholesale prices in levels and first-differences for Mymensingh, Rajshahi, Rangpur, and Sylhet .................................................................................................... 28 Figure 3.4: Autocorrelation by month lags for Barishal, Chattogram, Dhaka, and Khulna ..... 30 Figure 3.5: Autocorrelation by month lags for Mymensingh, Rajshahi, Rangpur, and Sylhet 31 Figure 3.6: Scatter plot of prices and first difference of prices of Dhaka with Barishal, Chattogram, Khulna, and Mymensingh ....................................................................................... 32 Figure 3.7: Scatter plot of prices and first difference of prices of Dhaka with Rajshahi, Rangpur, and Sylhet ..................................................................................................................... 33 Figure 3.8 Impulse response function ......................................................................................... 37 Figure 4.1: National coarse rice wholesale nominal and real prices ......................................... 42 Figure 4.2: Monthly variation in coarse rice wholesale real price ............................................. 43 Figure 4.3: National public rice procurement ............................................................................. 44 Figure 4.4: Monthly variation in public rice procurement .......................................................... 44 Figure 4.5: National public rice distribution ................................................................................ 45 Figure 4.6: Monthly variation in public rice distribution ............................................................ 45 Figure 4.7: National public rice net-distribution ......................................................................... 46 Figure 4.8: Monthly variation in public rice net-distribution ...................................................... 46 Figure 5.1. Private-sector storage decision rule ........................................................................ 61 Figure 5.2. Wholesale price of rice in New Delhi, historical and simulated ............................. 63 Figure 5.3. Sample of 10 years of results for rice production, consumption, and price .......... 65 Figure 5.4. Sample of 10 years of results for public and private stocks .................................. 65 Figure 5.5. Estimated size of private stocks by month ............................................................. 66 1 ACRONYMS AND ABBREVIATIONS BIDS Bangladesh Institute of Development Studies BIHS Bangladesh Integrated Household Survey BULOG Indonesian Bureau of Logistics CFW Cash-for-Work CV Coefficient of variation DG Food Directorate General of Food EPP Export Parity Price FAO Food and Agriculture Organization of the United Nations FFW Food-for-Work FOB Freight on Board FPMC Food Planning and Monitoring Committee FPMU Food Planning and Monitoring Unit GAMS General Algebraic Modeling System GIEWS Global Information and Early Warning System HIES Household Income and Expenditure Survey IFPRI International Food Policy Research Institute IFPRP Integrated Food Policy Research Program IPP Import Parity Price JV IFPRI-BIDS-UIUC Joint Venture MFSP Modern Food Storage Facilities Project MoF Ministry of Food NFA National Food Authority PFDS Public Foodgrain Distribution System UIUC University of Illinois at Urbana-Champaign USD United States Dollars 2 CHAPTER 1. INTRODUCTION Price instability is a fact of life. In a market economy, domestic prices change in response to changes in supply, consumer preferences, policy, world prices, and other factors. Crop prices tend to be particularly volatile because harvests occur only once or a few times per year and because the size of the harvest varies due to weather, prices, and other factors. For internationally-traded commodities, volatility in world prices can be another source of instability in domestic prices. At the same time, most people are risk averse, meaning they prefer income and consumption to be steady. In economic terms, they are willing to pay to reduce income volatility. That is, they are willing to accept a lower average income in order to have a more stable income. Evidence of this can be found in the demand for various types of insurance, the fact that risky investments must offer higher rates of return, the fact that employers must pay higher wages for dangerous or short-term jobs relative to safer and more stable positions with similar skill requirements, and the fact that households use savings to maintain stable consumption in the face of fluctuating income. Low-income food-deficit households are particularly vulnerable to changes in food prices in general and staple grain prices, in particular. This is because staple foods represent a significant share of the expenditure of low-income households, so price changes have a larger effect on the purchasing power and standard of living of these households. Similarly, households that depend on the sale of agricultural commodities are sensitive to instability in agricultural prices. Thus, it is not surprising that various governments pursue policies to reduce food price instability. The most common strategy is for the government to purchase grain during the harvest and sell it during the off season, maintaining public reserves of grain for emergency use and other periods of scarcity. For example, the Food Corporation of India purchases and distributes roughly 60 million tons of rice and wheat each year (Rashid et al., 2007; FCI, 2017). In Indonesia, the Bureau of Logistics (BULOG) plays a similar role, managing rice imports and distributing more than 3 million tons of subsidized rice through a program called RASKIN (Fernandez, 2015, Rashid et al., 2007). The Ministry of Food in Bangladesh has responsibility for the procurement, storage, and distribution of rice and wheat throughout the country. And in sub-Saharan Africa, grain marketing boards in Kenya, Malawi, and Zambia buy and sell maize and other staples to manage food prices. At the same time, the costs of managing public procurement and distribution of food grains can be high. The Food Corporation of India receives annual budget support amounting to more than US$ 10 billion per year (FCI, 2017). In Indonesia, the RASKIN rice distribution programme costs US$ 1.7 billion per year (Fernandez, 2015). Given the scale of these programs, it is important to evaluate the benefits they generate. This paper examines rice price stabilization in Bangladesh and uses a model to simulate the impact of alternative policies on rice price stability. Although a comprehensive assessment of the costs and benefits of the public grain distribution system in Bangladesh is beyond the scope of this paper, this paper has four objectives: • To review previous research on price instability and policy options to manage it; • To examine the relationship among rice prices in Bangladesh; • To estimate the determinants and impact of public procurement and distribution of rice in Bangladesh; • To use a model to simulate the impact of alternative policies on rice price stability in Bangladesh. 3 The paper is divided into six chapters. After this introduction, Chapter 2 provides some background on food price stabilization, including measurement, sources of price instability, international patterns, instability in Bangladesh, and the economics of public grain reserves. Chapter 3 gives an econometric analysis of rice prices in Bangladesh, assessing the extent to which rice prices in different Divisions are co-integrated (move together). Chapter 4 presents an econometric analysis of the effect of rice prices on public rice distribution and procurement, as well as the effect of government intervention on rice prices. Chapter 5 describes the IFPRI Bangladesh Rice Market Model, a monthly spatial equilibrium model of rice markets in Bangladesh, which is used to simulate the consequences of various policy options on rice price stability and fiscal costs. Finally, Chapter 6 summarizes the results and discusses some policy implications. 4 CHAPTER 2. BACKGROUND ON FOOD PRICE STABILIZATION This chapter provides a review of food price stabilization. The topics covered include the reasons for price instability, international patterns in food price stability, the arguments for and against public grain reserves, the international experience with public grain reserves, and the link between price instability and welfare. 2.1 Measuring price instability Food price instability is typically measured using the coefficient of variation (CV), which is defined as the standard deviation of prices (s) as a percentage of the average price (µ): CV = 100 ( 𝑠 𝜇 ) (2.1) If the variable has a normal distribution, then the CV indicates the range over which it will vary 68% of the time. For example, if the average is 100 and the CV is 15%, this indicates that the price stabil- ity is such that 68% of the time it will be between 85 and 115 and 32% of the time it will be outside that range. A CV of zero indicates that there is no variation in prices. In theory, the coefficient of var- iation has no upper limit, but the CV of monthly grain prices is generally in the range of 10% to 40%. Any upward or downward trend in the data leads to an artificial increase in CV. Therefore, when there is an upward or downward trend, it is better to use the adjusted CV, also called the Cuddy- Della Valle Index (Cuddy and Della Valle, 1978). The adjusted CV can be calculated as the CV of prices after removing the time trend1 or calculated more simply as follows: Adjusted CV = ( 𝑠 𝜇 ) (1 − 𝑅2)0.5 (2.2) where R2 is the coefficient of determination, which measures the correlation between price and a time trend variable. The adjusted CV is equivalent to the calculation of the coefficient of variation on prices after the time trend has been removed from the data series. 2.2 Sources of food price instability In simple terms, domestic food price changes can be the result of changes in demand or changes in supply. A shift in domestic supply may be caused by changes in the cost of production, changes in the price of competing crops, changes in policy (e.g. fertilizer subsidies), or weather-related varia- tion in yields. A shift in food demand may be caused by changes in income, changes in the price of competing foods, or changes in policy (e.g. food price subsidies). Changes in international prices can influence both the supply of imported food and the international demand for domestic food, thus affecting domestic prices. Changes in demand The demand for certain foods may vary by season if there are holidays or festivals. For example, in Vietnam, the Tet (New Year) holiday is marked by an increase in consumption of sticky rice and 1 The time trend can be removed statistically by regressing prices on time and then using the residuals, that is, the difference between the original price and the estimated price based on the regression equation. The residuals show the variation in prices around the trend line. 5 other traditional foods. However, these changes are relatively small and predictable, so they are not a major contributor to food price instability. Over the years, rising income shifts food demand from starchy staples like rice and wheat toward fruits, vegetables, animal products, and processed goods. However, the effect of rising income on food demand is slow and steady, with effects that are visible only over a decade or more. As a result, changes in demand are generally not a major contributor to food price volatility. Changes in supply Changes in food supply are a more important cause of food price instability, particularly for non-trad- able commodities. First, food prices follow a seasonal pattern due to the agricultural cycle, with prices lowest at harvest and rising throughout the off-season. The economics of storage suggests that the rise in price during the off-season reflects the cost of storage since the harvest. This ex- plains several features of price seasonality: • Prices continue to rise in the off-season, reaching a maximum just before the next harvest. • Seasonality in prices is negligible for non-perishable goods like processed foods, modest for grains and other somewhat-perishable goods, and greatest for fruit and other highly-perisha- ble products. • Grain prices show the greatest seasonality in regions with one harvest per year, such as southern Africa, and the least seasonality in places with multiple harvests, like Bangladesh. Second, in addition to seasonality, the size of harvests varies from year to year as a function of weather, pests, disease, and policy changes, adding to food price instability. Minot (2014) examined patterns of grain price volatility in sub-Saharan Africa and found that price volatility was greatest among commodities with little international trade (including cowpeas, maize, and sorghum) and low- est among internationally traded goods such as wheat and rice. This suggests that domestic supply shocks are a more important source of price volatility than international markets, at least in Africa. Similarly, Dorosh and Shahabuddin (2002) argue that the main source of price instability in Bangla- desh is domestic supply shocks associated with weather. International price volatility How do international prices affect domestic prices? In a market economy, international prices set an upper and lower limit on domestic prices. Domestic prices are generally kept below the import parity price (IPP), defined as the world price plus the cost of shipping the commodity to the country includ- ing any applicable taxes. As soon as domestic prices rise to the level of the IPP, it becomes profita- ble to import, which prevents further price increases. Likewise, international trade keeps domestic prices above the export parity price (EPP), calculated as the world price minus the cost of exporting rice to international markets including any taxes. As soon as domestic prices fall to the EPP, it be- comes profitable to export, and the reduction in domestic supply prevents further decline in the price. Domestic prices will closely follow international prices in the following circumstances: • If a commodity is regularly traded (imported or exported) in the world market, then spatial ar- bitrage will link local and international prices. For example, as noted above, rice and wheat prices in Africa are more likely to be linked to world markets than maize prices, presumably because almost all African countries are regular importers of rice and wheat, while they are generally self-sufficient in maize (Minot, 2011). • If the cost of transportation of goods to and from the country is low, domestic prices will be more closely linked to international prices. Low transportation costs reduce the import parity 6 price and raise the export parity price, thus narrowing the band within which domestic prices may fluctuate. • When trade policy with respect to the commodity is relatively open, local prices are more closely linked to international prices. Low taxes and minimal non-tariff barriers (like low transportation costs) reduce the gap between IPP (the upper limit) and EPP (the lower limit), thus limiting domestic price volatility. • The locally-produced commodity is a close substitute for the internationally-traded commod- ity, meaning that consumers consider them to be equivalent and will not pay a premium for either. Conversely, domestic prices may not follow international prices under other conditions: • If the country is self-sufficient in most years, then the domestic price fluctuates between IPP and EPP, being uninfluenced by either. • If the country is landlocked or the cost of transporting commodities to and from international markets is high for other reasons, this will increase the IPP (the upper limit) and reduce the EPP (the lower limit) creating a wider gap between them. This allows more price variation in response to supply shocks. • If there are high taxes, quotas, or administrative restrictions on imports or exports, this will also create a wider gap between the IPP and the EPP within which domestic prices can move. • If there are important quality differences between the domestic commodity and the interna- tional commodity, then their prices may diverge even after taking transport costs into ac- count. By limiting extreme movement of domestic prices upward or downward, international trade can re- duce domestic price instability. At the same time, instability in international prices can be transmitted to domestic markets, as was the case during the food crisis of 2007-08. The net effect of interna- tional trade on domestic price stability is an empirical issue, depending on the volatility in interna- tional prices, the width of the band between the EPP and the IPP, and the degree of instability in do- mestic supply. 2.3 International patterns in grain price instability This section describes patterns of price instability in rice and wheat markets around the world. Table 2.1 and Error! Reference source not found. show the CV and adjusted CV for wholesale rice and wheat prices in various countries, based on data from the FAO Global Information and Early Warn- ing System (GIEWS) database. Because the original prices are expressed in US dollars, the infla- tionary time trend is small, and there is little difference between CV and adjusted CV. The adjusted CV for medium rice prices in Dhaka wholesale markets is 13.6%. This indicates that rice price instability in Bangladesh is relatively low, below the international average of 16.8%. The adjusted CVs of other prices range from 7.9% in Ecuador to 30.6% in Thailand. There are no obvi- ous patterns by region or by importer/exporter status. 7 Table 2.1: Price instability in wholesale rice markets Source: Authors’ analysis of FAO GIEWS database Country Market Rice type Coefficient of varia- tion Adjusted coefficient of variation Ecuador Quito Long grain 11.6 7.9 Dominican Republic Santo Domingo First quality 9.8 9.2 Myanmar Yangon Emata, ehyv-fq 12.6 10.7 Bolivia La Paz First quality 11.5 11.0 Viet Nam An Giang 20% broken 15.3 13.0 Panama Panama City First quality 15.0 13.1 Honduras San Pedro Sula Second quality 14.1 13.2 Nigeria Lagos Imported 13.5 13.4 Bangladesh Dhaka Medium 13.9 13.6 Mali Bamako Local 13.7 13.7 India New Delhi Unspecified 26.5 14.2 Peru Lima Milled, superior 18.5 14.7 Nicaragua Managua (oriental) First quality 25.5 14.8 Philippines Metro Manila Regular milled 33.3 15.0 Brazil National Average Paddy 15.5 15.1 Guatemala Guatemala City First quality 23.8 15.1 Rwanda Kigali Unspecified 16.8 15.5 Uruguay National Average Grade 1 27.4 16.4 Niger Niamey Imported 17.0 16.9 Cambodia Phnom Penh Mix 19.9 17.8 El Salvador San Salvador Unspecified 19.1 18.4 Burkina Faso Ouagadougou Imported 18.8 18.6 Uganda Kampala Unspecified 22.7 19.0 Tanzania Dar es Salaam Unspecified 24.7 20.5 Ghana Accra Local 27.3 22.0 Mexico Mexico City Morelos 27.9 22.4 Colombia Bogotá First quality 37.4 23.8 Djibouti Djibouti Belem 24.9 24.5 Italy National Average Paddy, Arborio Volano 29.1 28.8 Thailand Bangkok 25% broken 39.6 30.6 Mean 20.9 16.8 Median 19.0 15.1 8 Table 2.2: Price instability in wholesale wheat markets Source: Authors’ analysis of FAO GIEWS database 2.4 Patterns of grain price instability in Bangladesh In Bangladesh, rice and wheat prices are strongly influenced by international prices. Figure 2.1 com- pares the wholesale price of rice in Dhaka with two benchmark Thai rice prices, 5% broken and 25% broken, all prices expressed in US dollars per kilogram. From 2000 to 2007, the Bangladesh rice price tracked the Thai prices quite closely. During the 2007-08 food crisis, both prices spiked, though the increase in Bangladesh was much smaller. Between 2008 and 2013, the Bangladesh and Thai prices diverged, at times moving in the opposite direction. Since 2013, the prices have re- mained relatively close to each other. Figure 2.2 compares wholesale wheat prices in Bangladesh with international benchmark prices in Argentina and the United States. The Bangladesh wheat flour price is also included, starting in 2008. The link between domestic and international prices is clear, with domestic wheat prices rising Country Market Rice type Coefficient of varia- tion Adjusted coefficient of variation Ecuador Quito Wheat flour 6.6 2.3 India New Delhi Wheat 26.4 12.7 Uzbekistan National Average Wheat flour 14.0 13.9 Peru Lima Wheat flour 15.1 15.0 Sudan Khartoum Wheat 35.2 16.6 Bolivia La Paz Wheat flour 19.8 19.2 El Salvador San Salvador Wheat flour 19.3 19.3 Kazakhstan National Average Wheat flour 19.6 19.6 Djibouti Djibouti Wheat flour 21.1 20.0 Chile National Average Wheat 23.4 21.1 Bangladesh Dhaka Wheat 30.5 22.7 Uruguay National Average Wheat flour 26.3 23.2 Ukraine National Average Wheat flour 26.8 24.9 Ethiopia Addis Ababa Wheat 37.3 25.0 Israel National Average Wheat 26.0 26.0 Colombia Bogotá Wheat flour 31.0 27.3 Brazil National Average Wheat 30.8 28.6 South Africa Randfontein Wheat 41.6 29.0 Russian Federation National Average Wheat 32.3 31.8 Italy National Average Wheat 37.8 37.8 Argentina Buenos Aires Wheat 41.9 39.9 Mean 26.8 22.7 Median 26.4 22.7 9 in tandem with international prices in 2007-08, falling in 2009, and then rising again in 2010. In re- cent years, domestic prices of wheat seem to be trending above international prices. Figure 2.1: Bangladesh and international (Thai) rice prices Source: Analysis by Paul Dorosh. Figure 2.2: Bangladesh and international wheat prices Source: Analysis by Paul Dorosh. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 J u l- 9 8 F e b -9 9 S e p -9 9 A p r- 0 0 N o v -0 0 J u n -0 1 J a n -0 2 A u g -0 2 M a r- 0 3 O c t- 0 3 M a y -0 4 D e c -0 4 J u l- 0 5 F e b -0 6 S e p -0 6 A p r- 0 7 N o v -0 7 J u n -0 8 J a n -0 9 A u g -0 9 M a r- 1 0 O c t- 1 0 M a y -1 1 D e c -1 1 J u l- 1 2 F e b -1 3 S e p -1 3 A p r- 1 4 N o v -1 4 J u n -1 5 J a n -1 6 A u g -1 6 P ri c e ( U S $ /k g ) Bangladesh, Dhaka, Rice (coarse- BR-8/ 11/ Guti/ Sharna), Wholesale, (USD/Kg) INTERNATIONAL PRICES, Thailand (Bangkok), Rice (5% broken), Export, (USD/Kg) INTERNATIONAL PRICES, Thailand (Bangkok), Rice (25% broken), Export, (USD/Kg) 0 0.1 0.2 0.3 0.4 0.5 0.6 J a n -0 0 J u l- 0 0 J a n -0 1 J u l- 0 1 J a n -0 2 J u l- 0 2 J a n -0 3 J u l- 0 3 J a n -0 4 J u l- 0 4 J a n -0 5 J u l- 0 5 J a n -0 6 J u l- 0 6 J a n -0 7 J u l- 0 7 J a n -0 8 J u l- 0 8 J a n -0 9 J u l- 0 9 J a n -1 0 J u l- 1 0 J a n -1 1 J u l- 1 1 J a n -1 2 J u l- 1 2 J a n -1 3 J u l- 1 3 J a n -1 4 J u l- 1 4 J a n -1 5 J u l- 1 5 J a n -1 6 J u l- 1 6 P ri c e ( U S $ /k g ) Bangladesh, Dhaka, Wheat, Wholesale, (USD/Kg) Bangladesh, Dhaka, Wheat (flour), Wholesale, (USD/Kg) INTERNATIONAL PRICES, Argentina, Wheat, Up River, FOB (USD/Kg) INTERNATIONAL PRICES, US (Gulf), Wheat (US No. 2, Hard Red Winter), Export, (USD/Kg) 10 We can measure changes in grain price instability. Figure 2.3 shows changes in rice price instability, measured by the 12-month lagged CV of the real wholesale price of coarse rice in Dhaka in US dol- lars. In other words, each point measures the standard deviation of the previous 12 months of prices as a percentage of the average price over the previous 12 months. The CV of rice prices ranged from just 2% to over 20%. The period of greatest rice price stability over this period occurred be- tween 2003 and 2006, when the CV remained below 5% for most months. The highest points of price instability occurred in April 2007 and June 2008 during the global food crisis of 2007-08. Dur- ing most of this period, however, the CV of wholesale rice prices was less than 10%, which is low by international standards. As of October 2017, the CV was about 10%. Figure 2.4 shows the 12-month lagged moving coefficient of variation for the wholesale wheat price in Dhaka expressed in dollars. The most stable period for wholesale wheat prices was 2000-03, when the CV was in the range of 2-3%. The most instable period was between 2007 and 2011, then the CV was frequently above 15% and three times passed 20%. Since 2011, wheat prices have been more stable, with the CV falling from around 10% to about 5%. Wheat prices in Bangladesh are more stable than in most countries; the mean and median CV for the countries in Table 2 were above 20%. Figure 2.3: Moving coefficient of variation (CV) of Dhaka wholesale coarse rice price Source: Analysis by authors using data from FAO GIEWS. 0 5 10 15 20 25 J a n -9 9 J a n -0 0 J a n -0 1 J a n -0 2 J a n -0 3 J a n -0 4 J a n -0 5 J a n -0 6 J a n -0 7 J a n -0 8 J a n -0 9 J a n -1 0 J a n -1 1 J a n -1 2 J a n -1 3 J a n -1 4 J a n -1 5 J a n -1 6 J a n -1 7 J a n -1 8 1 2 -m o n th l a g e d m o v in g C V 11 Figure 2.4: Moving coefficient of variation (CV) of Dhaka wholesale wheat price Source: Analysis by authors using data from FAO GIEWS. 2.5 Economics of public grain reserves Government-managed grain stocks go by several names, depending on the objectives. A strategic grain reserve generally refers to a public stock of grain used to meet emergency food requirements and to relieve temporary shortages while commercial imports or food aid are being arranged (FAO, 1998). A buffer stock is a public storage system specifically for the purpose of price stabilization. Buffer stocks and strategic grain reserves are generally operated by a semi-autonomous state- owned enterprise. The enterprise can impose a floor price by offering to buy unlimited quantities at that price. Eventually, however, the stocks need to be disposed of, either through sales, exports, or donation as food aid. Similarly, it can set a ceiling price by offering to sell unlimited quantities at that price. Eventually, the public stock will need to be refilled, either by domestic purchases or imports. The buffer stock can also adopt a price-band policy, keeping the market price within a specified band by offering to sell at the ceiling price and buy at the floor price. The main objective of public grain reserves is to stabilize price. Even emergency reserves can be thought of as strategy to avoid price spikes following poor harvests or natural disasters. But it is im- portant to examine the arguments for public grain reserves, as opposed to allowing grain storage to be maintained by private agents, such as farmers, traders, and millers, guided by market forces. In a competitive market, spatial arbitrage ensures that price gaps between markets in two locations cannot exceed the cost of transporting the product from one market to another, where costs are de- fined broadly to include returns to capital and management and a risk premium. If the price gap ex- ceeds cost, traders can make a profit by transporting goods, but in doing so they raise the price in the source market and reduce the price in the destination market. Temporal arbitrage works in the same way. If the expected price increase between (say) January and June exceeds the cost of storage over this period, then traders, millers, and farmers will have an incentive to store in January and sell in June. In doing so, however, they reduce the supply and increase the price in January, while increasing supply and reducing the price in June. When the ex- pected price increase falls below the full cost of storage, the incentive to store additional quantities will be eliminated. In this way, profit-oriented storage decisions result in a socially beneficial out- come: reducing price fluctuation by raising the prices during periods of surplus, such as during the harvest, and reducing the price during periods of scarcity, such as during the off-season. 0 5 10 15 20 25 30 J a n -9 9 J a n -0 0 J a n -0 1 J a n -0 2 J a n -0 3 J a n -0 4 J a n -0 5 J a n -0 6 J a n -0 7 J a n -0 8 J a n -0 9 J a n -1 0 J a n -1 1 J a n -1 2 J a n -1 3 J a n -1 4 J a n -1 5 J a n -1 6 J a n -1 7 J a n -1 8 J a n -1 9 1 2 -m o n th l a g g e d m o v in g C V 12 An important difference between spatial arbitrage and temporal arbitrage is that goods can be trans- ported in either direction between two markets, but storage is asymmetric in that it can only “move” goods forward in time, not backward. It should be noted that competitive storage does not eliminate price instability; it only reduces ex- pected price increase between two periods to the cost of storage between the two periods. One im- plication is that, in a competitive market, reducing the cost of storage or the risks associated with storage will reduce price instability. Economic arguments for public grain reserves Economic arguments for public grain reserves focus on ways in which the government may be able to improve on the performance of private agents in stabilizing prices. There are at least four types of economic justification for public grain reserves, most of which are based on some type of market failure. The first possible economic justification is market power. If a group of grain traders were able to col- lude, they could make monopoly profits at the expense of society. To do this, they would need to control a large share of the storage capacity, to agree to act together, and enforce this agreement on each other. The agreement would involve storing less than what they would in a competitive mar- ket, thus lowering the harvest price and raising the off-season price, which increases the returns from storage. This is not an easy task, given the large number of traders and millers, the risk of “cheating” (agents who store more than agreed to), and the possibility that farmers or other agents will store grain in response to the increased profitability. The second possible justification is imperfect information. If the government has better information about future market conditions, they could do a better job of knowing how much needs to be stored for future needs. This argument was probably more persuasive in the 1960s and 1970s when pri- vate markets and communication infrastructure were less developed, than it is today (Rashid et al, 2008). The third rationale is economies of scale in storage. If large-scale public grain storage is less costly than smaller-scale private grain storage, this would be a possible justification for maintaining public grain reserves (Newberry, 1989). This argument has been undercut by studies showing that public grain storage tends to have higher costs than storage in the private sector. Fourth, public reserves to stabilize grain prices may be justified on the grounds of equity. Poor con- sumers tend to spend a large share of their budgets on staple grains and are risk averse. The gov- ernment may have an interest in grain price stabilization to protect poor households from food inse- curity (Turnovsky et al, 1980). As mentioned above, competitive private sector grain storage does not eliminate price instability, so the government may take measures to further stabilize prices. Regardless of economic arguments, political pressure is often brought to bear during periods of ex- treme prices. Consumers lobby for intervention when grain prices spike, and farmers may protest during gluts when prices are low (Islam and Thomas, 1994; Poulton et al, 2006). Governments are pressured by power consumers to force traders who have accumulated grain to surrender stocks to the government…to limit “speculation” in grain markets…Antici- pation of such treatment discourages private storage in times of plenty, for distribution at a high price in time of need (Wright and Cafiero, 2011). Even if private traders are not accused of hoarding, the operation of public grain reserves to stabi- lize prices will reduce the incentive for private traders to store grain. In effect, public grain reserves tend to displace some private storage activity. 13 Arguments against public grain reserves often focus on the high cost of managing them, the poten- tial for corruption or favoritism in procurement and/or distribution of grain, and the fact that govern- ment agencies managing the reserves may be slow to respond to evolving market conditions (New- berry and Stiglitz, 1981; Rashid et al., 2008). Research on storage and public grain reserves Economic research to simulate the operations of public grain reserves focuses on the case of buffer stocks, which have clearly defined rules for buying and selling stocks. A buffer stock is usually rep- resented as defending a price floor by being willing to buy unlimited quantities at that price and de- fending a price ceiling by being willing to sell unlimited quantities at that price. These operations keep the market price in the band between the two prices. A key contribution to the study of grain storage under uncertainty was made by Gustafson (1958), who developed an approach to identify optimal storage rules given stochastic harvests. This ap- proach only worked, however, when supply was assumed to be perfectly inelastic. Wright and Wil- liams (1982) extended this analysis to take into account the case where supply responds to market prices. Williams and Wright (1991) provided a comprehensive analysis of the impact of buffer stock operations in the context of competitive price-sector trade with rational expectations. They found that the market price rarely stays in the center of the band; instead, it is often “challenging” the price ceiling or the price floor. Because private storage is less profitable in the presence of a buffer stock, there is actually more price volatility near the price ceiling, though of course the buffer stock elimi- nates the risk of the price exceeding the ceiling. Furthermore, there is no long-term tendency for sales and purchases to offset each other. Their simulations show that the costs will eventually grow unsustainably. Gouel and Jean (2013) expand this model to take into account international trade. They assume a market for a storable commodity in a small open economy where world price is given, per-unit transport is constant, consumers are risk averse, and domestic food price volatility is driven by both stochastic output shocks and a stochastic world price. The optimal trade policy to stabilize prices in the absence of storage consists of subsidizing imports when availability is low and the taxing of ex- ports when both availability is bountiful and the world prices are high. By combining trade policy and public grain storage, the price of grains is significantly more stable than in the absence of these poli- cies. However, the redistribution of benefits between producers and consumers is large compared to the efficiency gains. In addition, the effect of these policies on international trade, if implemented by many countries, would exacerbate grain price spikes in world markets. The choice of price band has major implications for the fiscal cost, the necessary storage capacity, and the impact on the food prices. One important dimension of the buffer stock rules is the width of the price band. • The wide price band would limit purchases or sales to cases of serious shortages or a large surplus. Thus, government intervention in the market would only occur every few years. The cost and storage requirements would be relatively small, but the degree of price stabilization would also be modest. It would reduce inter-annual price instability but leave seasonal cy- cles largely unaffected. • A narrow price band would result in annual purchases during the harvest season when prices are lowest and annual sales during the off-season when prices are highest. This ap- proach would reduce both inter-annual and seasonal fluctuation in food prices. If the width of the price band is less than the cost of seasonal storage, seasonal storage of grain would not be profitable, causing traders to withdraw completely from seasonal storage. Furthermore, a 14 band this narrow prevents the buffer stock from covering its costs on seasonal storage (Knudsen and Nash, 1990). • At the extreme, the buffer stock could attempt to eliminate all price instability by setting the buying and selling price arbitrarily close to each other. This would probably be infeasible from a cost point of view because the buffer stock could end up being forced to purchase or sell a large share of annual production. Furthermore, complete stabilization would probably be undesirable from an economic point of view because price variation helps farmers and consumers respond to surpluses and deficits, thus bringing the market to equilibrium. Another dimension of the buffer stock rules is the level of the price band. • If the price band is set too high relative to the market price, the buffer stock will be purchas- ing more often than it is selling, resulting in the accumulation of larger stocks each year. Storage capacity or budget constraints will eventually prevent further purchases, making it impossible to continue supporting the price. • Conversely, if the price band is set too low, the buffer stock will be selling more often than it is buying. Eventually, the stock will be exhausted and it will be impossible to continue to de- fend the ceiling price. Because of uncertainty regarding the “normal” market price, it is often recommended that buffer stocks use a moving average of the previous 3-5 years as the mid-point for the price band (Knudsen and Nash, 1990). However, simulations by Williams and Wright (1991) suggest that it is very difficult to defend a price band with a finite budget. Finally, there is the issue of whether the price band is the same across all buying stations or not. • If all stations are defending the same price band (pan-territorial pricing), it will reduce or elim- inate the incentive for private traders to move grain from one location to another. Depots in surplus zones will be paying above-market prices and will be forced to purchase the entire surplus. Meanwhile, in deficit zones, the depots will be selling at below-market prices, so they will be forced to supply large quantities of grain. The buffer stock becomes a grain mar- keting parastatal, responsible for all grain transport from surplus to deficit zones. Further- more, the grain transport will be done at a loss because the price difference will often be less than the cost of transport, particularly if the band is narrow. • Alternatively, if each station sets a different price band, it would be possible to maintain in- centives for private traders to move grain from deficit to surplus zones. To achieve this, the mid-points of the price bands would have to be set according to the normal market price in each location. This means that the price difference between locations would be large enough to motivate private traders to handle transport in normal years. The above is an idealized view, in which the buffer stock has one objective (price stabilization) and makes purchases and sales based on a clearly defined price band. The actual operation of public food reserves is more complicated, as discussed in the next section. International experience with public grain reserves In practice, most public food reserves are designed to serve multiple objectives, including public dis- tribution, emergency relief, and/or price stabilization. In addition, many reserves do not announce buying and selling prices ahead of time. Instead, procurement and distribution decisions are made on an ad hoc basis, depending on prices, the expected size of the harvest, and political pressure from stakeholders. These patterns are illustrated by three case studies described below. 15 In Indonesia, BULOG, the government food logistics agency, is responsible for public food reserves, which it uses mainly for price stabilization. From 1967 to the mid-1990s, BULOG defended a mini- mum farm price and maximum ceiling price for rice, which was adjusted roughly every year. It main- tained a legal monopoly on rice imports, restricting imports and supporting the price to promote self- sufficiency. Studies have shown that BULOG keeps prices higher but more stable than world prices, but there have been problems of smuggling and corruption. One study estimated that the costs to consumers of the high rice price were US$ 400 million per year (Arafin, 2008). In 1997, the Asian financial crisis and a poor harvest meant that the country was forced to turn to the International Monetary Fund for emergency credit. Over 1998-2000, rice trade was liberalized, leading to a surge in imports. In 2001, BULOG was restructured as a state-owned enterprise and was once again given a monopoly on rice imports. A tariff keeps the domestic price above interna- tional prices in an effort to promote rice self-sufficiency. A study showed that a majority of rural farmers in Indonesia are net rice buyers, relying on other crops, non-farm businesses, and wage labor to purchase rice. This implies that most rural farmers are hurt by the policy of supporting rice prices above the import parity level. There remains concern about the high cost and inefficiency of BULOG, but strong political interest in maintaining current policy has preserved its central role in Indonesian rice marketing (Arifin, 2008). In the Philippines, the National Food Authority (NFA) was formed in 1972 with the goal of stabilizing price and promoting rice self-sufficiency. Like BULOG, it carried out local procurement in surplus zones, as well as maintaining a monopoly on rice imports. It also maintains emergency and strategic reserves and carries out rice distribution which is supposed to be targeted at poor households. However, NFA has been accused of importing unneeded quantities of rice, of implementing emer- gency distribution systems which overlap with other government programs, and allowing 50% leak- age to non-poor households in its public distribution system. In addition, the above-market price of rice in the Philippines imposes high costs on urban consumers and many net buyers in rural areas. The government has debated various proposals to reform the NFA to focus on rice and announce a price band to improve predictability in rice markets. Others have suggested privatizing the NFA or allowing the private sector to compete with the NFA in importing rice (Clarete, 2008). In Vietnam, before 1986, most of the economy was under centralized government planning. Under the socialist regime, agricultural production was carried out on collective farms, where farmers were paid a salary based on the output of the farm. Government cooperatives and state enterprises man- aged agricultural marketing, including rice milling and distribution to consumers. Vietnam had rela- tively low rice yields and, as a result, was a chronic rice importer. In 1986, the government began to implement the Doi Moi policy, which allocated collective farmland to individual farm households. Initially, farmers were required to sell a quota to the government at fixed prices, but were allowed to sell any surplus to emerging private traders. Eventually, the quotas were phased out and milling and marketing was increasingly carried out by small private entrepre- neurs. With greater incentives to maintain irrigation works and use modern inputs, Vietnam quickly transformed itself to a major rice exporter. However, two large state-owned enterprises, Vinafood 1 in the north and Vinafood 2 in the south, were given a monopoly on rice exports, as well as managing large rice mills and storage facilities. Unlike in Indonesia and the Philippines, they were not involved in price stabilization, neither support- ing farmgate prices nor limiting consumer prices. In the late 1990s, the rice export quota was gradu- ally lifted and then eliminated, allowing Vietnam to become the second-largest exporter in the world. 16 In spite of partial export liberalization, Vinafood 1 and 2 retain 70% of Vietnamese rice exports, op- erating primarily as commercial firms. The state enterprises incur high costs and require annual sub- sidies to continue operating, causing some researchers and policy makers to question their contin- ued role in Vietnamese rice marketing (Son and Thang, 2008). Public grain reserves in Bangladesh Since independence in 1971, Bangladesh has moved progressively from a government-managed grain market to one in which markets play a much larger role. In the 1970s and 1980s, Bangladesh operated a grain rationing system in urban and rural areas, combined with numerous restrictions on private sector grain trade. A number of studies in the 1980s demonstrated both the high cost of the grain rationing system, and the fact that much of the grain was “leaked” to non-poor households. In the late 1980s and early 1990s, Bangladesh removed restrictions on internal movement of grain, phased out the rural rationing program, and legalized the private-sector imports of rice and wheat. The combination of social safety net programs and private-sector grain imports are credited with the successful response to the extensive 1998-9 floods (Ali et al, 2008). Bangladesh maintains a system of public procurement of rice and wheat, a network of public ware- houses, and open market sales of grain. However, the procurement price is not a floor price in that the government does not attempt to “defend” the price by purchasing all available grain at those prices. Similarly, the open market sales are limited in quantity, so they do not represent an effective ceiling price. Although some research has questioned the economic rationale for price stabilization (Goletti, 2000), there is political support for continued public grain procurement and open-market grain sales. Various studies have attempted to study the impact of price stabilization. Building on the work of Ah- med and Bernard (1989), Shahabuddin (1991) presented a model for rice and wheat price stabiliza- tion in Bangladesh. This model was used to estimate the volume of domestic procurement and open market sales of food grains that are needed to support prices during the harvest seasons and ac- commodate the prices during peak seasons with a fixed price ceiling for price stabilization. Using a fixed floor (procurement) price and a fixed ceiling (sales) price, the results indicate that most pro- curement would take place in the aman and boro harvests. The model indicated that there was hardly any seasonality displayed in the ration (monetized) distribution of food grains. The offtake in rice and wheat, which comprised for nearly 60% of the sales by the Public Food Distribution System (PFDS) and roughly 7% of the market supply, acted as a cushion for market rice prices in Bangla- desh. On the other hand, open market sales of rice showed notable seasonality in the model, specif- ically in the lean season where higher prices were predicted, versus the harvest season where low prices were predicted. One limitation of the model, however, is that the model assumes that the gov- ernment is fully responsible for price stabilization, with no role for the private sector. Dorosh and Haggblade (1997) use a partial-equilibrium model to simulate the impact of potential food-aid monetization which would shift poverty assistance away from food-for-work (FFW) pro- grams with in-kind deliveries of wheat to a cash-for-work (CFW) program. Food aid would be mone- tized (sold at market prices), and cash transfers would be provided to beneficiaries. The results indi- cate that the program would improve beneficiary welfare and reduce commodity handling costs which have the potential to lead to an increase in available funding for development and poverty- alleviation programs. From a supply standpoint, a small increase of 100,000 MT in wheat supply un- der a year-round CFW monetization program was found to lower lean season prices by 4% with only a 0.8% decrease in harvest season prices. Replacing the FFW program with a monetized CFW program would alter the seasonality of wheat prices, lowering lean season prices and raising har- vest season prices which would greatly benefit poorer households. Dropping the FFW program’s in- kind wheat delivery and switching to a wholesale CFW program would impact supply and demand 17 by providing program participants with slightly higher incomes because of lower wheat prices along- side about $10 million in savings for commodity handling which is currently funded through a combi- nation of government subsidies and in-country FFW rations. In an analysis on the value of price stabilization and the management of public food grain stocks in Bangladesh, Goletti (2000) says that demand for government price stabilization is in decline. He ar- gues that price stabilization in Bangladesh has provided very small microeconomic benefits, unde- terminable macroeconomic benefits, and negligible effects on poverty. Seasonal and inter-year price fluctuations have diminished because of significant growth of the boro rice harvest, which allows for rice to be available throughout the year. Built-in price stabilization has been made available through private traders who have been pioneers in importing grains during periods of domestic shortfalls. However, price stabilization and the scale of public intervention remain important politically, rather than economically. He argues that 700,000 to 800,000 tons would be sufficient for emergency needs. At these stock levels, there would be an estimated savings for the public food program of about $130 million per year. Dorosh and Shahabuddin (2002) examine the causes of grain price instability in Bangladesh, argu- ing that it is primarily the result of fluctuations in production, due to floods and droughts. According to their analysis of rice price data over 1970s-1990s, the annual fluctuations in nominal prices of rice in Bangladesh were more stable in the 1980s compared to the 1970s. However, in the 1990s fluctu- ations in rice prices increased. Agricultural seasonality remains an issue among the poorer popula- tions; vulnerable households can slip into poverty temporarily, and those who are poor experience more hardships. The seasonality of rice prices in Bangladesh from the 1970s to 1990s, calculated by taking the average price of rice for each month as a ratio of a twelve-month moving average, pro- duced three major changes. First, the ratio of peak price to trough price showed a gradual decline over time, with the most significant decline occurring between the 1980s and 1990s. Second, the peak price shifted from July in the late 1970s to April in the 1980s and 1990s. Third, the late 1970s showed prices rising after a slight decline from April to May, while the 1980s showed a drop in prices from April to June and stable prices from June to August. Within the world market, the fluctu- ation of nominal prices increased in the 1980s when compared to the 1970s, while during the 1990s, the range of price fluctuations declined to nearly the same range of the 1970s. Before liberalization of the private rice trade in 1994, government imports and stock policies were the two factors most responsible for determining rice prices. As a result, the Bangladesh market was partially protected from world market fluctuations. After liberalization, domestic rice prices of the 1990s were below import parity levels. Over 1970-2000, world prices of rice became more stable as the volume of world trade increased. Bangladesh domestic rice prices (in Taka) were relatively as stable in the 1990s. Although Indian rice prices were more stable than Thai and Bangladesh rice prices, Bangladesh rice prices were marginally more stable than international (Thai) prices. Through trade liberalization, the import parity price sets a ceiling on domestic prices. Del Nino et al (2001) show that private-sector imports were instrumental in stabilizing rice prices in Bangladesh for 1997/98 as well as 1998/99 after major rice production shortfalls. The reason that prices did not rise any further during this time was that competitive private-sector importers were able to import the grain necessary to meet domestic demand. While private sector imports have provided stable prices, one cannot rule out the need for rice stocks. The sales of subsidized government imports can avoid price spikes caused by domestic production shortfalls combined with high international prices. Export parity prices should form a price floor, but this has not worked because market links are not established, and there are no established uniform grades or standards in Bangladesh. In or- der to avoid large price declines, the establishment of grades and standards that comply with inter- national trade in addition to significant investments in mechanized graders should be considered. 18 Similarly, Dorosh (2001) analyzes trade liberalization and its effects on food security and price stabi- lization. Because of the late 1990’s rice production shortfalls, Bangladesh’s private sector importers relied heavily on Indian rice harvest imports in order to alleviate Bangladesh’s food security prob- lems. Dorosh (2001) raises the question of whether the private sector and international markets can be relied on as a source of food grain. The analysis shows that there is low correlation among In- dian and Bangladesh rice harvests. In particular, there is a lack of correlation between poor Indian rice harvests and poor Bangladesh rice harvests. Nonetheless, one cannot completely rule out the possibility of India and Bangladesh having poor harvests in the same year. He recommends that the Bangladesh government prepare itself for such a situation. In this case, rice imports would most likely come from Thailand which would come at a higher cost than if imported from India. For these reasons, continued government support of private import trade is essential. Because of trade liberal- ization in both India and Bangladesh, large-scale trade took place and private sector importers were able to help to prevent a food crisis in the late 1990s. Brennan (2003) developed a model of Bangladesh grain markets that government intervention takes place in the context of competitive storage by private-sector agents using rational expectations. The paper examines the effect of private storage, government subsidization of storage, and government imposition of a price ceiling on the distribution of market prices over time. The key findings from this analysis were that the more inelastic the demand, the greater the incentive to store which led to lower prices in times of oversupply and higher prices in times of shortage. In turn, the increase in storage under inelastic demand led to a higher consumption stability. The reverse was found to be true for more elastic demands, which lowered the incentive to store and increased the variability of consumption. She also found that government stockholding tends to displace private sector storage because the price ceiling reduced the incentive for private-sector storage during periods of lower production. In general, price ceiling methods were found to be unfavorable due to the large amount of stock that the government would need to hold, the government’s infrequent participation in the market, and overall variability in prices. The option of public subsidization of private sector storage would be a cheaper alternative for stabilizing rice prices, but this policy does not always prevent price spikes. If the goal is to protect poor urban consumers by preventing extreme price peaks, then buffer stock schemes with a price ceiling are a better choice than the subsidization of private stor- age. One limitation of this study is that the model assumes a closed economy, meaning that there are no grain imports or exports. As discussed above, grain imports have proven to be an effective way to prevent price spikes at no fiscal cost. Rashid et al (2005) reviews the benefits associated with grain market liberalization, particularly the legalization of private-sector grain imports. Until the early 1990s, the government of Bangladesh maintained a monopoly over international grain trade, and restrictions on the movement of grains by the private sector. Concessional credit to government operations and preferential access to trans- portation were also used to discourage private-sector trade. The government paid higher import prices for grain than the private sector, even when the government's imports were much greater than those of the private sector. Allowing private grain trade and stabilizing grain prices around the international parity price in Bang- ladesh had a large payoff. The savings associated with the addition of private-sector imports were an estimated 3.22 million tons of grain over the ten-year time span of 1992/93-2001/02 which is roughly $422 million USD. In addition, liberalization contributed to an overall decline in the annual public distribution of rice and wheat. Price stability was not threatened by reforms which were ac- complished through the PFDS. Because of these savings, Bangladesh has been able to allocate ad- ditional resources to development and anti-poverty projects. In general, investments in agricultural productivity, roads, education, and safety net programs are considered a better investment than price stabilization (Rashid et. al, 2005; Ali et al, 2008). 19 2.6 Effect of food price instability on welfare Food price instability has a negative effect on household welfare because it contributes to fluctua- tions in income and consumption. Studies of human behavior (and common sense) confirm that most people are risk averse, meaning that they prefer a steady level of income to a highly variable income that has the same average value. The relationship between price instability, income, con- sumption, and welfare is illustrated in Figure 2.5: Diagram of effect of food price instability on house- hold welfare . Figure 2.5: Diagram of effect of food price instability on household welfare Source: Authors. As shown in the diagram, the effect of price instability of a given commodity on household welfare depends primarily on four factors: • The degree of price instability. Generally, the greater the instability, the larger the effect on household welfare. • The effect of a given level of price instability on real income (or purchasing power) of the household. The proportional effect on income depends on the net sales (or net pur- chases) of the commodity as a proportion of income. Households that spend a large share of their budget on the commodity are most affected, as are households that derive a large share of their income from the sale of the commodity. • The degree to which the variability in income is translated into fluctuations in con- sumption. Households with high income or valuable assets are better able to smooth con- sumption during income shortfalls by drawing on savings, borrowing, and selling non-produc- tive assets. Poor households, and those with few assets, are not able to smooth consump- tion as easily, being forced to reduce non-food or even food consumption during hard times. • The degree to which variability in consumption affects household welfare. Again, high- income households can experience a reduction in consumption with less adverse effect on welfare. Households close to subsistence cannot reduce consumption without risking health and malnutrition. Turnovsky et al (1980) were one of the first to rigorously examine the effect of price instability on consumer welfare. They confirmed earlier results that a risk-neutral consumer actually gains from 20 instability in consumer prices, provided the instability does not affect income. Their analysis also showed that for consumers to be adversely affected from the price instability of a commodity, they must be risk averse and the income elasticity of the commodity must be low. These conditions are likely to hold in the case of food price stabilization and the poor in developing countries. Newbery and Stiglitz (1981) pioneered methods for analyzing agricultural price risk and the effects of price stabilization programs. We highlight four of their findings: • They emphasized the fact that the objective of policy should not be price stabilization per se. Rather, price stabilization is only useful to the extent that it reduces the instability of the in- come of farmers and consumers. • Second, price stabilization does not always stabilize farmer income. Supply shocks create a negative correlation between output and prices: during good years, output is high but prices are low, and during bad years, the reverse is true. Thus, during bad years, price stabilization would lower farm income by not allowing the shortage to result in higher prices. Conversely, during good years, price stabilization would raise farm income by not allowing the surplus to depress prices. In other words, a program that is successful in stabilizing agricultural prices could actually destabilize farm income and reduce farmer welfare. • Third, food price stabilization is likely to have a positive effect on food supply, motivating farmers to produce more at a given average price. However, if food is non-tradable, this will reduce the equilibrium price, transferring some of the benefits of price stabilization to con- sumers. • And finally, one of their most important contributions was a method for estimating the welfare gain for farmers associated with price stabilization and assumptions about the degree of risk aversion, as measured by the Arrow-Pratt measure of relative risk aversion (R). Newbery and Stiglitz (1981) applied their formula using a range of plausible parameters describing food markets in developing countries and the degree of risk aversion. Under a range of assump- tions, the results showed that the gains to farmers from complete price stabilization is relatively small, about 0-3% of household income. They conclude that the benefits of price stabilization have been exaggerated and that government resources would be better off allocated to other types of programs, such as investment in infrastructure, research, and other public goods. Later studies that used the Newbery-Stiglitz approached confirmed that the static gains to price sta- bilization are quite small. For example, Jha and Srinivasan (1999) develop a multimarket model of the Indian grain economy with random supply fluctuations and use it to simulate the effect of alter- native trade and marketing policies on price stabilization and welfare. They find that the policies with the most stable prices were not the ones that increased welfare the most. Islam and Thomas (1996) compare free-trade and a partial stabilization policy in terms of their effect on rice farmers for five Asian countries: Bangladesh, Indonesia, Pakistan, the Philippines, and Thai- land. Using historical data on rice price variability in each country and three alternative assumptions about risk aversion, they showed that the gains in risk reduction from price stabilization ranged from 0.5% to 5% of farm income. Adopting the middle assumption on risk aversion (R=1.5), the gains were 1.5% to 3.5% of farm income. More recently, Myers (2006) extended the Newbery-Stiglitz approach to incorporate the impact of food price stabilization on households that both produce and consume the commodity. He uses pa- rameters describing four representative households: poor consumers, affluent consumers, poor pro- ducers, and affluent producers. With low levels of price instability (CV=10%) and low risk aversion (R=1), the gains from price stabilization on all groups are less than 1% of income. With high price instability (CV=30%) and high risk aversion (R=3), all four groups gain from price stabilization but 21 the size of the gain varies. Affluent producers gain the most (9% of income) because they have large sales of the commodity. Poor producers gain less from stabilization (3%) because their sales are a smaller share of income. Affluent consumers are hardly affected by price stabilization because the commodity represents only a small share of their budget. And poor consumers gain barely 1% from price stabilization. These and other similar studies have led to the conclusion among most economists that the benefits to farmers and consumers of food price stabilization are modest. These findings have contributed to the general skepticism regarding food price stabilization programs at the World Bank and other do- nor organizations (Byerlee et al, 2006). There is, however, some dissent. One line of research explores possible dynamic effects of food price stabilization. In other words, food price stabilization may contribute to a higher rate of eco- nomic growth. Even a small contribution to economic growth would generate large benefits for food price stabilization, but several studies have failed to find such a connection (Myers, 2006). Another approach is to question the relationship between income variability and welfare. The ex- pected utility model (upon which the Newbery-Stiglitz results are based) is a simplification of peo- ple’s complex preferences regarding risk, and some alternative models of risk lead to larger esti- mates of the benefits from price stabilization (Aizenman, 1998). Finally, one study argued that the benefits of price stabilization may be greater than previously esti- mated if we examine price stabilization for multiple commodities. Bellemare et al (2011) uses panel survey data from rural Ethiopia to estimate the effect of price stabilization of seven staple foods. The benefits are estimated to be worth 6%, 15%, or 32% of household income, depending how risk averse farmers are assumed to be. The practical implications of this finding may be limited by the fact that stabilizing seven food prices is beyond the financial and logistical capacity of most coun- tries. Another interesting finding from the Bellemare et al (2011) study is that the benefits of price stabili- zation accrue mainly to the richest 40% of rural households who are surplus farmers; the other 60%, many of whom are net buyers, gain less or actually lose as a result of price stabilization. In this sense, it confirms the results of Myers (2006) that food price stabilization may have the largest ben- efits for medium and larger farmers who are surplus food producers. This has implications for the political economy of food price stabilization. In contrast, Timmer (1996) assessed the impact of price stabilization methods through the lens of the Indonesia Bureau of Logistics (BULOG) to emphasize that price stabilization is no longer crucial for economic development because the rice sector no longer acts as a gauge for the welfare of the economy. Rice price stabilization has been implemented largely in Asian and poorer countries that are heavily dependent on staples for caloric intake, jobs, and economic activity. Timmer’s analysis of BULOG’s role in the Indonesian rice economy provided convincing evidence that BULOG was able to successfully stabilize prices via the implementation of price bands, domestic procurement from rural markers, and through urban market operations which supply rice to markets, but that the model for stabilization was not suitable for planning purposes (Timmer, 1996). During periods of in- stability in global and local rice economies (such as the mid-1990s), unstable rice prices were found to have the capability to slow economic growth considerably. Over time, the benefits of price stabili- zation on economic growth and investments in Indonesia have declined in conjunction with the de- crease in the total share of rice in the economy and the subsequent decline in the significance of spillovers from rice to other sectors, as well as the increases in per capita income. There is little doubt that food price stabilization is politically popular in many developing countries. When food prices rise, consumers exert strong political pressure on governments to bring prices 22 down. The food riots that took place in a number of countries during the 2007-08 food crisis are an example of this pressure. Likewise, when food prices drop following a bumper harvest, farmer or- ganizations lobby for measures to support prices. Poulton et al (2006) argue that donors and re- searchers should work to improve the design price stabilization programs because “full liberalization is often not a credible strategy for political reasons.” A key question is whether this political pressure for food price stabilization represents the real interests of society, responding to gains from stabiliza- tion that are not captured by current economic models? Or does it represent pressure from special interests, such a small number of well-organized commercial farmers, who would gain from price stabilization at the expense of taxpayers and the economy as a whole? 23 CHAPTER 3. ANALYSIS OF RICE PRICES 3.1 Introduction In Chapter 2, we showed that rice price instability in Bangladesh (as in most countries) is related to the seasonality of rice production, inter-annual changes in rice production, and fluctuations in inter- national prices. In this chapter, we explore the degree to which rice prices in different Divisions are integrated, meaning that they move together. Market integration is important for two reasons. First, it is an indicator of the performance of rice markets. If prices are closely integrated, it is a sign that they are operating well in redistributing rice from areas with surpluses (where prices are low) to areas with deficits (where prices are high). Sec- ond, market integration may provide some information on the competitiveness of markets. There are numerous indicators of market competitiveness, including the degree of market concentration, the size of marketing margins relative to costs, and the ability of market agents to coordinate to set prices and punish those who do not follow the agreement. Market integration has long been consid- ered another indicator of competition, though it is an indirect measure. The idea is that if (for exam- ple) wholesalers colluded with each other, they would need to reduce the volume of rice moving from one Division to another in order to increase price differences and the profitability of inter-Divi- sional trade. By raising the inter-Division price differences, it would reduce the frequency of move- ment of rice between Divisions, creating periods when the two prices are not linked by spatial arbi- trage. In this chapter, we will use time-series econometrics to explore the statistical relationship among the monthly coarse rice prices between Divisions over the last eleven years. Section 3.2 describes the data and methods used in this study, and Section 3.3 presents the results of the analysis. 3.2 Methods and data This analysis uses monthly coarse rice prices data for each of the eight divisions in Bangladesh to explore the price movement across the divisions. The econometric analysis makes use of monthly wholesale coarse rice price data for each division over the 11 years from 2008 to 2018. The data were obtained from the Directorate of Agricultural Marketing (DAM). The data were available at the district level, so we selected the district with the largest city in the Division. We found one missing value for the Barishal division and replaced it using linear interpolation. In time series econometrics, existence of a valid long run relationship among the variables depends on whether the variables are stationary2 or not. If the variables are non-stationary, the ordinary least squares (OLS) regression estimates may lead to spurious regression problems with the t and F tests being non-standard. The problem with non-stationary time series may arise if the variables are generated either by trend stationary process (TSP) or by difference stationary process (DSP). The TSP variables are made stationary through de-trending them and the DSP variables need to be dif- ferenced until they are stationary. The Unit Root tests are the formal tests to determine whether the data generating process is TSP or DSP. This study uses the vector error correction model (VECM) to examine the relationship between wholesale coarse rice prices for each division with the wholesale coarse rice price in Dhaka division. The VECM is appropriate if (1) each variable is nonstationary and integrated to degree 1 and (2) the 2 Basic properties of stationarity are that mean, variance and covariance are constant over time. Stationary prices fluctuate around a mean, so they have a tendency to return toward the mean. In contrast, non-stationary prices tend to “wander” in a pattern often called a random walk, with no tendence to return toward a mean. 24 variables are cointegrated, meaning that there is a linear combination of the variables that is station- ary. We analyze prices in Dhaka division price with the prices in other divisions, so that the cointegrating equation would take the form of P1 = α + βP2 + ε or P1 – α – βP2 = ε, where ε is stationary. For each pair of division-wise prices, the analysis consists of three steps. First, we test the price var- iables individually to see if they are I(1)3. This is done with the Augmented Dickey-Fuller (ADF) test and the Phillips-Perron test. Second, we use the Engle-Granger and the Johansen test to determine whether the two series are cointegrated, meaning that each variable is I(1) and a linear combination of the two variables is I(0). In terms of our analysis, these tests whether there is a long-run relation- ship between the domestic price and the corresponding world price. Third, if the Johansen test indi- cates that there is a long-run relationship between the two variables, then we estimate the VECM. The model takes the following general form: Δpt = 𝛼 + Πpt−1 + ∑ Γk 𝑞 𝑘=1 Δpt−k + 𝜀𝑡 (3.1) where, pt is an n x 1 vector of n price variables; Δ is the difference operator, so Δpt = pt – pt-1; εt is an n x 1 vector of error terms; α is an n x 1 vector of estimated parameters that describe the trend component; Π is an n x n matrix of estimated parameters that describe the long-term relationship and the error correction adjustment; and Γk is a set of n x n matrices of estimated parameters that describe the short-run rela- tionship between prices, one for each of q lags included in the model. The VECM tests for the effect of each variable on each other variable. In the context of this study, the two-variable VECM tests the effect of Dhaka division price on other division prices as well as the effect of other division prices on the Dhaka division price. In addition, tests indicate that one lagged term is generally sufficient. For our purposes, then, we are interested in only one portion of the VECM. This portion can be simplified as follows: Δpt od = 𝛼 + 𝜃(𝑝𝑡−1 𝑜𝑑 − 𝛽1 − 𝛽2𝑝𝑡−1 𝑑ℎ ) + 𝛿Δpt−1 dh + 𝜌∆𝑝𝑡−1 𝑜𝑑 + 𝜀𝑡 (3.2) where pt od is the wholesale coarse rice price per kg in tk for any division other than Dhaka; 𝑝𝑡−1 𝑑ℎ is the wholesale coarse rice price per kg in tk for Dhaka division; Δ is the difference operator, so Δpt = pt – pt-1; α, θ, β1, β2, δ, and ρ are estimated parameters; and εt is the error term. As described above, if the original price series are I(1), then the first differences (Δp) will be station- ary, or I(0). Since the prices are expressed as tk/kg unit, the cointegration factor (β) is the long-run changes in tk/kg of the division-wise coarse rice price with respect to the change in coarse rice price in Dhaka. Thus, β is the long-run coefficient of price transmission. Considering the impact of coarse rice price in Dhaka the expected value for imported commodities is 1 > β > 0. 3 I(1) refers to integration of degree 1, meaning that the variable (xt) is not stationary, but the first difference (xt-xt-1) is stationary. I(0) means integrated of degree 0, meaning that the variable is stationary. 25 The error correction coefficient (θ) reflects the speed of adjustment. We expect it to fall in the range of -1 < θ < 0. The term in parentheses represents the deviation or “error” between the prices in the previous period and the long-run relationship between the two prices. If the error is positive (the price in other division is too high given the long-term relationship), then the negative value of θ helps “correct” the error by making it more likely that the pt od is negative. The coefficient on change in the price in Dhaka (δ) is the short-run coefficient of the price in other division relative to the price in Dhaka. In this case, it represents the absolute adjustment of rice price in other divisions one period after a unit (i.e. tk) shock in Dhaka price. The expected value is 0 < δ < β. The coefficient on the lagged price change in other divisions (ρ) is the autoregressive term, reflect- ing the effect of each change in the other division’s price on the change in price Dhaka division in the next period. The expected value is -1 < ρ < 1. 3.3 Results This section presents the results from the econometric analysis investigating the relationships among the price in the Dhaka division with the prices in other divisions. We started by checking whether the price variation in divisional markets follows any trend and seasonality over the last 11 years. Then, our analysis investigates the existence and nature of the relationship among the spatial markets. In doing so, we compared the Dhaka rice markets with the rice markets in the other seven divisions. For each pair of market combinations, we use the Engle-Granger and Johanson test for measuring the cointegration among the market pairs. Next, we examine the magnitude and direction of the short-run relationship among the market pairs by using the vector error correction model (VECM). Finally, we measured the impulse response function to see how one market responds to the shocks in other markets. Econometric characteristics of the data Before presenting the econometric results, this section provides the econometric characteristics of the data. Figure 3.2 shows the wholesale price movement for coarse rice over the last 11 years across the eight divisions. As we see from the figure, the price widely varies across the years. For instance, the price reached a record low in the year 2009, which was around 30% drop from the price of 2008. The price again jumped by 50% by mid-2010. However, the geographical markets seem to be well-integrated, as we do not observe much price differences across the markets over those years with some exceptions in 2015 and early 2016. The year 2015 and 2016 are marked as politically unstable years in the country, which constrained the normal grain flow from surplus zones to deficit zones resulting in some variation in the prices across divisions. 26 Figure 3.1: Coarse rice wholesale price for each Division Source: Authors’ calculations using data from the Department of Agricultural Marketing (DAM). To better understand the price year-to-year price variation across different geographical locations, the left panel of Figure 3.2 and Figure 3.3 separately present the price movement over 11 years for eight divisions. The right panel of Figure 3.2 and Figure 3.3 plot the first difference4 of the prices for each division. The first difference of each division-wise price shows a random fluctuation around zero relative to the prices in the previous months. 4 First difference means the difference in prices for the two consecutive years. It is represented as Δpt = pt – pt-1 27 Figure 3.2: Coarse rice wholesale prices in levels and first-differences for Barishal, Chattogram, Dhaka, and Khulna Source: Authors’ calculations using data from DAM. 28 Figure 3.3: Coarse rice wholesale prices in levels and first-differences for My- mensingh, Rajshahi, Rangpur, and Sylhet Source: Authors’ calculations using data from DAM. In the following, we will examine the stationarity of the wholesale coarse rice prices for eight divi- sions by applying the Augmented Dickey-Fuller (ADF) test and Phillips-Perron (1988) tests of unit roots. Table 3.1 reports the results of ADF and Phillips-Perron tests estimated using both the level form and the first difference form of the wholesale coarse rice price for each division. The ADF and Phil- 29 lips-Perron tests help us to determine the order of integration5 in the prices in divisional markets. Ta- ble 3.1 shows that for the price level, the ADF and Phillips-Parron test statistics are not statistically significant for any of the divisional markets at 5% level of significance, implying that the presence of unit roots in the prices for each division. On the other hand, both ADF and Phillips-Perron test statis- tics for all the markets on their first differences are statistically significant even at 1% level of signifi- cance, implying the absence of unit root and the stationarity in the first differenced form of prices. Therefore, it is reasonable to use the first difference forms of the prices for estimation to ensure sta- tionarity. Table 3.1: Test for unit roots in rice prices and first differences Source: Authors. Note: *, ** and *** indicate significance at 10%, 5% and 1% levels respectively. We further investigate the stationarity in the market prices for each division by plotting the autocor- relation function (ACF) against the number of lags. For non-stationary variables, the ACF usually declines slowly demonstrating either a gradual decline or a constant trend in the curve plotting the autocorrelation coefficients. On the other hand, for a stationary variable the ACF curve declines al- most instantly and then shows random movement. We plot the ACF function curve for both price level and the first difference in Figure 3.4 and Figure 3.5. As we see from the figures, the first difference form of the wholesale coarse rice prices for each division is stationary, implying integrated of order 1 (I(1)). 5 According to Engle and Granger (1987) a time series is said to be integrated of order d (denoted as ~I(d)) with d is the number of times the series needs to be differences in order to become stationary. Price variable ADF test Phillips-Perron test Barishal (level) -2.675 -2.413 Barishal (1st difference) -7.059*** -10.32*** Chattogram (level) -2.913 -2.381 Chattogram (1st difference) -4.778*** -8.899*** Dhaka (level) -2.985 -2.552 Dhaka (1st difference) -5.008*** -9.974*** Khulna (level) -3 -2.497 Khulna (1st difference) -4.792*** -9.615*** Mymensingh (level) -2.583 -2.63 Mymensingh (1st difference) -6.429*** -10.808*** Rajshahi (level) -2.955 -2.435 Rajshahi (1st difference) -4.791*** -9.347*** Rangpur (level) -3.132* -2.461 Rangpur (1st difference) -4.198*** -11.009*** Sylhet (level) -3.073 -2.562 Sylhet (1st difference) -7.692*** -8.755*** 30 Figure 3.4: Autocorrelation by month lags for Barishal, Chattogram, Dhaka, and Khulna Source: Authors’ calculations using data from DAM. 31 Figure 3.5: Autocorrelation by month lags for Mymensingh, Rajshahi, Rangpur, and Sylhet Source: Authors’ calculations using data from DAM. Finally, in Figure 3.6 and Figure 3.7, we present the scatter plots of price levels and the first difference in prices in Dhaka with the prices in the rest of the divisions in Bangladesh. A positive relationship between each of the market pairs is shown in both the level and the first differences, implying the potential existence of short and long-run influence. To further investigate, we use the Engle-Granger procedure and the 32 Johansen cointegration test to see if there is a long-run relationship between each division-wise price and the price in Dhaka and measure the magnitude of this relationship (if any) between the seven pairs of markets. Figure 3.6: Scatter plot of prices and first difference of prices of Dhaka with Barishal, Chattogram, Khulna, and Mymensingh Source: Authors’ calculations using data from DAM. 33 Figure 3.7: Scatter plot of prices and first difference of prices of Dhaka with Rajshahi, Rangpur, and Sylhet Source: Authors’ calculations using data from DAM. Econometric estimation Once we have identified variables as non-stationary, we apply some cointegration techniques to in- fer about the long-run relationship. In this section, we use both the Engle-Granger and Johansen procedures for estimating the magnitude of the long-run relationship betwee