: n K“ we}... r firm ... HR; 2, if... $4 «mu. .fi :1 its ..rx.. .3.) II! OI?!- SI on“ 3 iii: ‘- . .1 i 2... “MN: .. . ., . .a 31%.. k .. mmmme 3am ma )umflflfonff. . 41.08.. Ilia; nl. . 3%.... 3.: .u... i: . . . ,rxwrrz... .‘a." 3...??? £33.»...fi (inn. .. t. .- li‘va‘S 4:213. as: 31. 3.5:“? Ida-1.. .2 :1 1‘7"...“ .5. . I. .. . . .3293513511 . drift: .5... 11...... 5.53.. I... l... rialiitstlla i. aciiiit... :Cll. .. . a ‘3; .i. . i. I... a. 6% . n 3.. . 53:13..) I. 2L3?!) . ... ...u....):...lr. 5.. .12.}? 3’37. Viil’Y-bi’otl £50.”.I'vl‘ up 1...... 8.... .. r iii... 2'1! ‘0. L112 - 5.5.3.3..‘I Di \a In...” t . E. o I; 4.3.! X This is to certify that the thesis entitled AN ANALYSIS OF IMPORT TARIFF ESCALATION: A CASE OF MAIZE TRADE BETWEEN SOUTH AFRICA AND MOZAMBIQUE presented by Michelle N. Corzine has been accepted towards fulfillment of the requirements for the MS. degree in Agriculture, Food and Resource Economics Major Professor’s Signature I Z/ / 2 /08 ’ / Date MSU is an Affirmative Action/Equal Opponunity Employer AN ANALYSIS OF IMPORT TARIFF ESCALATION: A CASE OF MAIZE TRADE BETWEEN SOUTH AFRICA AND MOZAMBIQUE By Michelle N. Corzine A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Agriculture, Food and Resource Economics 2008 Copyright by MICHELLE N. CORZINE 2008 ABSTRACT AN ANALYSIS OF IMPORT TARIFF ESCALATION: A CASE OF MAIZE TRADE BETWEEN SOUTH AFRICA AND MOZAMBIQUE By Michelle N. Corzine Recent World Trade Organization negotiation rounds have focused on import tariff escalation, which occurs when import tariffs increase as the processing level increases. Although typically researched as a trade policy used by developed countries to protect their agro-processing sector, this research examines the effects of import tariff escalation when used by a developing country. Specifically, this thesis examines import tariff escalation (and the value-added tax) applied by Mozambique to imports of maize and maize flour from South Africa. Using econometrically estimated domestic supply and demand elasticities for each region, this thesis uses a spatial, partial equilibrium model to maximize social welfare subject to material balances and price constraints to model the changes in prices and export quantities due to a removal of import tariff escalation, the value—added tax, and a simulated ‘free’ trade environment. Although data availability was limited and thus decreased the statistical validity of the econometrically estimated elasticities, sensitivity tests indicated the models robustness to small changes in the elasticities. The model simulations indicated the removal of the VAT had very little effect of changes in prices or export quantities, while considerable changes were found in the simulated removal of import tariff and the ‘free' trade simulation. To My Parents iv ACKNOWLEDGMENTS I would like to thank all those who have provided support during this ongoing research project. First, I would like to thank my committee members. Dr. David Schweikhardt, my major professor, thank you for the continued guidance, support, and patience during this process. Your opinion and input has been invaluable throughout my time here at Michigan State. Thanks to Dr. Cynthia Donovan, for her vast knowledge of Mozambique’s maize market, which provided the building block for this thesis. Finally, Dr. Lisa Cook, thank you for agreeing to be a part of my thesis committee and for your constructive criticisms and suggestions. Special thanks to Dr. David Tschliey and Dr. Thom Jayne for their input and expertise on the maize markets and maize trade in South Africa and Mozambique. Thanks to Lulama Ndibongo-Traub and Danilo Abdula for their help and suggestions during the collection of data. Also special thanks to Elton R. Smith Endowment for partial funding for this research. I would also like to thank my friends and family for their support and encouragement. First, to all my friends and classmates at MSU for being a sounding board and providing advice, my experience and education at MSU would not have been the same without you. To my family and friends, your encouragement during this has been irreplaceable. Finally to my parents, your unconditional love, support and continued emphasis that I can achieve whatever I set out to do, is why I am where I am today. TABLE OF CONTENTS LIST OF TABLES ........................................................................................................... viii LIST OF FIGURES ............................................................................................................ xi KEY TO SYMBOLS AND ABBREVIATIONS .............................................................. xii 1. INTRODUCTION ....................................................................................................... '1 1.1 Import Tariff Escalation ............................................................................................ 1 1.1.1 Developed Countries .......................................................................................... 2 1.1.2 Developing Countries ......................................................................................... 4 1.2 Overview of the Problem .......................................................................................... 5 1.4 Research Objectives .................................................................................................. 7 1.5 Thesis Organization ................................................................................................... 8 1.6 Conclusion ................................................................................................................. 9 2. MAIZE MARKET .................................................................................................... 10 2.1 Mozambique ............................................................................................................ 10 2.1.1 Market Structure ............................................................................................... 12 2.1.2 Production ........................................................................................................ 14 2.1.3 Consumption .................................................................................................... 17 2.1.4 Milling Industry ................................................................................................ 20 2.2 South Africa ............................................................................................................ 22 2.2.] Production and Consumption ........................................................................... 23 2.2.2 Milling Industry ................................................................................................ 24 2.3 South Africa —- Mozambique Trade ......................................................................... 25 2.3.1 Import Tariffs ................................................................................................... 29 2.3.2 Value Added Tax (VAT) .................................................................................. 29 2.4 Conclusion ............................................................................................................... 31 3. LITERATURE REVIEW .......................................................................................... 32 3.1 Introduction ............................................................................................................. 32 3.2 North - South Import Tariff Escalation Analysis ................................................... 32 3.2.1 Nominal Tariff Wedge and Effective Rate of Protection ................................. 32 3.2.3 Partial Equilibrium Models .............................................................................. 37 3.2.4 Other Economic Based Methods ...................................................................... 42 3.3 South-South Import Tariff Escalation Analysis ...................................................... 45 3.4 Conclusion ............................................................................................................... 47 4. SOCIAL WELFARE MAXIMIZATION ................................................................. 48 4.1 Introduction ............................................................................................................. 48 4.2 Conceptual Model ................................................................................................... 49 4.3 Conclusion ............................................................................................................... 53 vi 5. ECONOMETRIC ESTIMATES OF DEMAND AND SUPPLY ............................. 55 5.] Conceptual Model ................................................................................................... 55 5.2 Empirical Model, Data and Results ......................................................................... 58 5.2.1 Southern Mozambique —Demand for Maize Flour .......................................... 58 5.2.2 Southern Mozambique - Supply of Maize Grain ............................................. 66 5.2.3 Central Mozambique — Demand for Maize Flour ............................................ 66 5.2.4 Central Mozambique — Supply of Maize Grain ............................................... 70 5.2.5 South Africa — Demand for Maize Flour ......................................................... 75 5.2.6 South Africa - Supply of Maize Grain ............................................................. 79 5.3 Conclusion ............................................................................................................... 83 6. SIMULATION .......................................................................................................... 85 6.1 Operational Model ................................................................................................... 85 6.1.1 Excess Supply and Demand ............................................................................. 88 6.1.2 Excess Supply and Demand Slopes and Intercepts .......................................... 89 6.1.3 Data .................................................................................................................. 90 6.2 Simulation ............................................................................................................... 99 6.2.1 Model Validity ............................................................................................... 100 6.2.1.1 Baseline ....................................................................................................... 100 6.2.2.2 Sensitivity Tests .......................................................................................... 101 6.2.2 Remove of the Value Added Tax (VAT) ....................................................... 106 6.2.3 Remove Import Tariff Escalation ................................................................... 108 6.2.4 Free Trade ...................................................................................................... 1 10 6.3 Conclusion ............................................................................................................. 1 12 7. CONCLUSION ....................................................................................................... 1 13 7.1 Summary ............................................................................................................... 113 7.2 Limitations ............................................................................................................ 115 7.3 Implications ........................................................................................................... 117 7.4 Additional Research .............................................................................................. 120 APPENDICIES ............................................................................................................... 1 22 APPENDIX A: VARIABLES OF SUPPLY/DEMAND ELASTICITY ESTIMATION .......................................................... 123 APPENDIX B: UNIT ROOTS DISCUSSION ........................................................... 128 APPENDIX C: VARIABLES/PARAMETERS OF SOCIAL WELFARE MAXIMIZATION MODEL ............................................................. 137 APPENDIX D: SENSITIVITY ANALYSIS OF ESTIMATED SUPPLY/DEMAND ELASTICTITIES AND ‘EXPORT’ QUANTITIES ......................... 143 BIBLIOGRAPHY ........................................................................................................... 158 vii LIST OF TABLES Table 2.1: Percent Maize Production and Sales, 2001/2002 ............................................. 17 Table 2.2: Food Expenditure Allocated to Cereals, National ........................................... 18 Table 2.3: 1996 Food Expenditure Allocated to Cereals, Urban and Rural ...................... 19 Table 2.4: Food Expenditure Allocated to Cereals, by Province and Urban/Rural .......... 20 Table 2.5: Historic Trade Patters between Mozambique and South Africa ...................... 26 Table 2.6: Imports of Maize Grain and Maize Flour to Mozambique, 2002-2006 ........... 27 Table 2.7: Exports of Maize Grain and Maize Flour from South Africa, 2002-2006 ....... 28 Table 3.1:: Results, Golub and Finger (1979) ................................................................... 39 Table 3.2: Results, Tangermann (1989) ............................................................................ 40 Table 3.3: Results, Wailes et a1 (2004) ............................................................................. 42 Table 4.1: Variables Description for Durand-Moral and Wailes (2003) Welfare Model ........................................................................................................... 52 Table 5.1 :Parameter Description for Maize Flour Consumption Calculation .................. 60 Table 5.2: Southern Mozambique Demand Estimation .................................................... 63 Table 5.3: Central Mozambique Demand Estimation ....................................................... 69 Table 5.4: Central Mozambique Supply Estimation ......................................................... 74 Table 5.5: South African Demand Estimation .................................................................. 77 Table 5.6: Top 5 Maize Producing Provinces in South Africa, 2002 ............................... 80 Table 5.7: South African Supply Estimation .................................................................... 82 Table 6.1: Variables, Parameters and Subscripts Description for Welfare Model ........... 86 Table 6.2: Historic Maize Prices and Production for Southern/Central Mozambique ...... 91 Table 6.3: Parameter Description for Maize Flour Consumption Calculation ................. 93 Table 6.4: Baseline Results ............................................................................................. 101 viii Table 6.5: Sensitivity Test - Change in Domestic Elasticities ........................................ 104 Table 6.6: Sensitivity Analysis - Quantity of Central Mozambique Maize Flour .......... 105 Table 6.7: VAT Removal Results ................................................................................... 107 Table 6.8: Import Tariff Escalation Removal Results ..................................................... 108 Table 6.9: Free Trade (Removal of VAT and Import Tariffs) Results ........................... 111 Table A. 1: Variable Description: Southern Mozambique — Demand for Maize Flour... 123 Table A2: Descriptive Statistics: Southern Mozambique - Demand For Maize Flour.. 123 Table A3: Variable Description: Central Mozambique — Demand for Maize Flour ..... 124 Table A.4: Descriptive Statistics: Central Mozambique — Demand for Maize Flour ..... 124 Table A5: Variable Description: Central Mozambique — Supply of Maize Grain ......... 125 Table A6: Descriptive Statistics: Central Mozambique -— Supply of Maize Grain ........ 125 Table A.7: Variable Description: South Africa — Demand for Maize Flour ................... 126 Table A8: Descriptive Statistics: South Africa — Demand for Maize Flour .................. 126 Table A9: Variable Description: South Africa — Supply of Maize Grain ...................... 127 Table A. 10: Descriptive Statistics: South Africa — Supply of Maize Grain ................... 127 Table B. 1: Dickey Fuller Test Results, (Continued) ....................................................... 130 Table 82: Southern Mozambique Demand Estimation, ARMA Model ........................ 132 Table B.3: Central Mozambique Demand Estimation, ARMA Model ........................... 133 Table B.4: Central Mozambique Supply Estimation, ARMA Model ............................. 134 Table 3.6: South African Supply Estimation, ARMA Model ........................................ 135 Table C.1: Variable/Parameter Description, Social Welfare Maximization Equation, (Cont’d) ................................................................................................................... 137 Table C.2: Parameter Description, Consumption Calculation for Mozambique ............ 142 Table D. 1:Sensitivity Analysis - Domestic Supply Variation, Values and Percent Difference from Observed Value, (Cont’d) ............................................................ 149 Table D2: Sensitivity Analysis - Domestic Demand Variation, Values and Percent Difference from Observed Value, (Cont'd) ............................................................. 151 Table D.3:Sensitivity Analysis - Domestic Supply/Demand Variation, (Cont'd) ........... 153 Table D4: Sensitivity Analysis - Central Mozambique Flour 'Expons' ......................... 156 LIST OF FIGURES Figure 1.1: Real Staple Food Prices in Maputo, 1993-2007 ............................................... 7 Figure 2.1: Mozambique Gross Domestic Product, per capita ......................................... 11 Figure 2.2: Map of Mozambique with Regions and Maize Flow Patterns ....................... 13 Figure 2.3: Total Maize Production vs. Total Area Harvested for Maize ......................... 15 Figure 2.4: Total Rainfall and Total Maize Production, Central/Southem Mozambique ...................................................................... 16 xi KEY TO SYMBOLS AND ABBREVIATIONS Abbreviation Definition 04 Met 2004 Mozambique new Metical’s APC African & Pacific Countries CIF Cost, Insurance & Freight CIM Compannia Industrial da Motala CPI Consumer Price Index EEC European Economic Community ERP Effective Rate of Protection EU European Union FAO Food and Agriculture Organization FEWSNET Famine Early Warning System FOB Free on Board GATT General Agreement on Tariffs and Trade GDP Gross Domestic Product IAF Inquierito aso Agregados Familiares lNE Instituto Nacional de Estatistica KG Kilogram MEREC Merec Industries MT Metric Ton OECD Organization for Economic Co-operation and Development SADC Southern African Development Community UNCTAD United Nations Conference on Trade & Development V&M Vonk Industries VAT Value-Added Tax WTO World Trade Organization xii 1. INTRODUCTION 1.1 Import Tariff Escalation As globalization continues to evolve and countries adjust to increased integration of markets, countries seek potential markets for their products while continuing to implement barriers that obstruct the expansion of those markets. These barriers used to distort trade incentives and protect domestic industries include export subsidies, export taxes, quotas and import tariffs. Though past multilateral negotiating rounds of both the General Agreement on Tariffs and Trade (GATT) and the World Trade Organization (WTO) have focused on reducing these market barriers, many are still present and continue to distort trade and hinder development in many countries. One distorting trade policy that received attention during the Uruguay Round of negotiations was import tariff escalation in agricultural processing chains. According to OECD (1997), import tariff escalation occurs when there is a zero or lower tariff on unprocessed or raw commodities, but a larger tariff on a processed form of the commodity. An example would be when a country has a low or zero import tariff on cocoa beans, a higher import tariff on cocoa powder and an even higher import tariffs on chocolate bars. Import tariff escalationI has become a major issue during trade negotiations because it adds a higher level of effective protection to a country’s processing and retail sector (OECD, 1997). As global trade expands, the protection of a country’s processing sector has two separate negative effects dependent on if the escalating tariff is applied by I Export tariff escalation, which is a less used trade barrier, occurs when the export tariff is greater on the lower the processing level product. In other words, if there is a higher export tariff on the raw commodity than the processed commodity then export tariff escalation exists. Nevertheless, the goal of the escalating export tariff is the same, to protect the local processing industry. a developed or developing countries. The most discussed and researched effect of import tariff escalation is the use by developed countries and its negative impacts on developing countries. More specifically, the use of import tariff escalation by developed countries inhibits developing countries’ ability to diversify their export portfolio, increase their trade in processed commodities, and reduce their reliance on exports of raw commodities (Elamin and Khaaira, 2004). Alternatively, the use of import tariff escalation by developing countries, specifically on imports of staple commodities, can create negative impacts, especially for its consumers, as the protected processing industry is allowed to produce less efficiently, while using market power to increase the retail price of the processed commodity. 1.1.1 Developed Countries Historically most developing countries have focused on exportation of raw commodity products. However, as noted by Elamin and Khaira (2004) many different factors have lead to an increased interest of developing countries to play a larger role in the processing of agricultural commodities. The first factor is the continued emphasis on diversification of a country’s export portfolio for sustainable economic growth. Developing countries acknowledge that focusing exports on only one or two raw commodities makes the country highly sensitive to instability caused by poor production years, natural disasters, or declining world market prices. Another factor leading to the change in developing countries is the growing urban labor force in developing countries due to migration from rural to urban areas. Processing industries would thus provide another employment option for the growing urban population. Third, there is growing awareness of environmental problems caused by over—exploitation of natural resources, causing developing countries to pursue greater control over their own raw commodities. The final factor is the growth in global demand of agriculture processed products as compared to raw commodities. Developing countries see the growth of demand for processed products as compared to demand for raw commodities and acknowledge that if a shift is not made, their economy will decline due to a loss in exports. This final factor brings the emphasis back to the most important point of the need for diversification in the export portfolio. Elamin and Khaira (2004) indicate that the agriculture processing sector as a share of world trade has increased since 1980. In a twenty year period, from 1980 to 2000, processed agriculture products increased its share in total value of world trade by 6 percent, while primary products trade value only increased by 3.3 percent of the total value traded. The decrease in importance of primary products has been caused by low income elasticity of demand, coupled with the decline in economic activities in which primary products are intensely used, along with the overall structure of the commodity markets (Elamin and Khaira, 2004). This shift towards trade of processed products, as compared to primary products, has large negative effects on developing countries economies as they tend to be the leader in exports of primary products. In addition to the decreasing importance of primary products trade, developing countries are losing their share in the trade of agriculture processed products. Elamin and Khaira (2004) note that overall the share of agriculture exports of processed products from developing countries decreased by 2 percent, while the share of agriculture processed products as a whole was increasing. Countries classified as least developed countries also had a decrease in the share of agriculture exports in processed products from 0.7 percent to 0.3 percent over the same time period. This trend can be seen in an example of cocoa. Elamin and Khaira (2004) compiled the top ten cocoa producing developing countries and found that as the stage of processing increased the share that developing countries had in that product decreased. The results indicate that the top ten developing countries shares of world exports from 1996 to 1999 were 83 percent for cocoa beans, 30 percent for cocoa butter, 28 percent for cocoa powder and 1 percent of chocolate. To illustrate that the growth of the export processing sector the authors note that in 1970 chocolate exports were 20 percent of total cocoa products world exports, however by 1996 chocolate exports were 56 percent to total coca products world exports. The authors also noted that the same trend could be found for the top ten developing countries producing coffee. Therefore, due to the importance that agriculture plays in developing country economies and the changing preferences in world trade of agriculture products, it is important that trade-restricting policies not be used to further hinder developing countries export potential. 1.1.2 Developing Countries As mentioned above, developed countries are not alone in applying escalating import tariffs on agriculture commodities. Developing countries use import tariff escalation as a means of protecting their own agricultural processing industries. Most developing countries throughout Africa, Asia and Central and South America use some type of import tariff to protect their main staples (WTO, 2008). However, literature that examines the impacts of import tariff escalation on domestic consumers and producers in developing countries is limited. Valenzuela et a1 (2004) are one of the few authors to address the impacts caused by the removal of import tariff escalation. The authors examine social welfare changes due to tariff removal on smallholder livestock producers in developing countries. The authors conclude that both in the short and long run, poverty for the majority of countries examined would be reduced through the elimination of domestic import tariffs 1.2 Overview of the Problem Mozambique is a prime example of a developing country using import tariff escalation as a means of protecting its maize processing industryz. Mozambique and South Africa are both members of the South African Development Community (SADC), which is a free trade block of fourteen southern African countries. Trade in maize, however, which is considered a sensitive commodity, is exempt from this free trade agreement. Mozambique currently applies a 20 percent import tariff on maize flour entering from South Africa, while only applying a 2.5 percent import tariff on maize grain imports from the same country3. Mozambique has an additional import policy to further protect its industrial maize processors. The country applies a 17 percent Value Added Tax (VAT) on all maize grain entering the country. The combination of the VAT and import tariff on grain equals the import tariff applied on maize flour, thus making it appear that Mozambique has high import barriers on all products rather than an escalating 2 Mozambique is currently discussing applying export tariff escalation to the cashew sector to protect the cashew processing sector. Given the current “food price crisis”, Mozambique is discussing changing their import tariff policy on basic staples, including maize. import tariff problem. However, further investigation into the VAT reveals that the 17 percent VAT is reimbursed as long as the grain is being imported to be processed into maize flour by an industrial processing company. On the other hand, if the grain being imported for use at smaller hammer mills or homes, then the VAT is not reimbursed. Therefore, the combination of a high escalating import tariff on maize flour, combined with a high VAT on maize grain that is exempt for industrial maize processors further illustrates the Mozambique govemment’s protection of the industrial maize millers in the country. The importance of this topic is heightened by the recent trends in local maize prices in Mozambique. Maize grain has traditionally been the number one staple for the urban poor and rural populations in Mozambique. However, maize grain and flour prices, which were traditionally below the cost of rice, have been increasing more rapidly than rice prices (Figure 1.1). Tschirley and Abdula (2007) argue that this increase in maize grain and maize flour prices in Mozambique is not caused by world price changes but the domestic maize market structure. The authors explain that beginning in 2002, maize flour prices in Maputo began increasing, eventually peaking in late 2004 and then stabilizing in 2006, with similar trends occurring in the center region of Mozambique, sometimes to a higher degree. During this same time period, Zambian maize flour prices were significantly lower than Mozambique prices with a surge in late 2005, due to a sharp appreciation of the Zambian Kwacha. Even with this surge, maize flour prices in Zambia were still almost half the cost per kilogram (kg) as compared to Mozambique, with similar results found in other surrounding countries. Affordability of maize grain and maize flour, especially for poor consumers throughout Mozambique, is heightened by the current increasing price of maize and maize flour on the world market. It therefore becomes imperative to examine the domestic impacts of the restricting import policies applied to maize grain and maize flour, specifically, given the importance of maize and maize flour as a food staple. Figure 1.1: Real Staple Food Prices in Maputo, 1993-2007 20 g ' Retail Rice Price 18 9 I\ / I’ ‘\ 16 g I \‘ Retail Maize Flour Price 14 o E L/ as ‘2 E 10: .v‘-- 4 8 1 ’ 6 4 2 0 -,_-. .L_.L.-- , , ,,,,_M,-,z “a h ‘9 Q ’\ Q Q Q N “t ’5 Dr ‘2 Q ’\ Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q x“\°’\°’\9s°’s°’\°’~°w“~°~“w°w°’w“~° Data Source: SlMA (2008) 1.4 Research Objectives The general objective of this research is to empirically measure the changes in prices and quantities traded that would occur in Mozambique with a change in import tariff policy. The specific objectives of this research are to: a) Describe the maize market structure and industrial maize milling industry in both Mozambique and South Africa, including the current trade situation between the two countries, and the import tariffs and VAT applied to maize grain and maize flour originating from South Africa. 7 b) Provide a review of prior research on import tariff escalation. c) Create a model of maximum net social monetary welfare under different tariff scenarios. (1) Estimate elasticities of supply and demand for maize grain and maize flour in three relevant geographic regions that will be used to make the net social monetary welfare model operational. e) Make policy recommendations on Mozambique’s import tariff escalation and VAT policies based on the results of the tariff simulations. 1.5 Thesis Organization This thesis is organized to accomplish the objectives defined above. Chapter 2 will examine the maize sector in Mozambique and South Africa and the trade relationship between these two countries, including an in-depth discussion of the import tariff and VAT policies applied to maize and maize flour traded between South Africa and Mozambique. Chapter 3 will review the literature of past research in import tariff escalation analysis to identify the knowledge gaps that need to be filled and the methods used to analyze import tariff escalation. Chapter 4 will describe the theoretical framework of the social welfare maximization model used in the analysis. Chapter 5 provides the theoretical framework for the supply and demand estimations, in addition to the empirical supply and demand models used, data description, and results from the estimations that will be used to make the social welfare model operational. Chapter 6 starts with a detailed description of the empirical model and parameters used for the net monetary social welfare model and finishes with results from the baseline and alternative tariff scenarios. Finally, Chapter 7 concludes with an overall summary of the research and findings, implications, and opportunities for further research on this topic. 1.6 Conclusion This chapter has provided a brief introduction to the economic consequences of import tariff escalation that occurs when applied by both developed and developing countries. Mozambique was introduced as the subject of this research due to their application of escalating tariffs import on maize grain and maize flour imported from South Africa. A short introduction to the current situation in Mozambique has been provided. Before analysis of the import tariffs can be conducted, the maize market system in Mozambique and, to a lesser degree, South Africa must be understood. The following chapter will provide detailed information related to the complex maize market in Mozambique, including details on both production and consumption of the commodities. In addition, the chapter will provide a brief description of the maize market system in South Africa and details on trade patterns between the two countries and a more in-depth look at the tariffs used by Mozambique. 2. MAIZE MARKET 2.1 Mozambique Mozambique is a large country on the south-eastern coastline of Africa stretching north-south along 2500 km of the Indian Ocean. The country is divided into ten provinces and three regions. In this study, Maputo, Inhambane and Gaza provinces form to make the southern region, while Manica, Sofala and Tete make the center region, leaving Zambezia“, Nampula, Cabo Delgado, and Niassa to form the northern region. According to the Instituto Nacional de Estatistr’ca (INE), almost 20 million people populated Mozambique in 2006, of which 25 percent were located in the southern region, 15 percent were residents of the central region and the remaining 60 percent were in the northern region of Mozambique (1997). While the majority of the country, 70 percent, is labeled as rural, over 40 percent of the southern region is classified as urban, and approximately 80 percent of the population in Maputo province is categorized as urban (INE, 1997). Mozambique’s economic development has suffered various setbacks. Economic growth began to decline with the end of Portugal’s colonial rule of Mozambique in 1975 and further decreased through the 1980’s, as civil war engulfed Mozambique. As one can see in Figure 2.1, with the end of the civil war in 1992, economic growth increased rapidly5 throughout the post-war era (1993 to 1998) due to good crop production years, increased political stability and large foreign investments (Jones, 2006). Due to importance of agriculture to the GDP, slight dips in economic growth occurred during 4 While official statistics often group Zambezia with the center region, for market analysis, Zambezia is more integrated into the northern region. The bridge currently being constructed over the Zambezi River may change the current market integration dynamics of Mozambique. 5 GDP growth averaged around 8 percent during the post — war era, with the largest growth occurring in 1993 with a 20 percent increase. 10 1999 and 2000 caused by poor rainfall and intensive flooding that led to poor production years. Economic growth continued after 2000, but at a slower, less intensive rate (approximately 3 percent) than experienced during the early post-war era. In recent years, economic growth has been rapidly increasing (around 7 percent) due to large investments in industry. Figure 2.1: Mozambique Gross Domestic Product, per capita 16000000 » . 3 5 i 5 i , ; : . , 3 . 3 [ 40000000 14000000 -~ ‘ w ' ‘ ‘ ‘ ‘ fl 35000000 12000000 ~ t- 30000000 A m ' l a 9-1 3 10000000 , 25000000 5 9 § l a C? 29. 30000000 -1 l 20000000 g g ; . i = :5 60000000 ; 1- 15000000 :5 '5 40000000 -. }- 10000000 V 20000000 1 ,,, I: t _ w 9 5000000 —1 N M '1' if: \G I\ oo GK 6 —4 N M ‘I' If: a 5‘ G's a ¢\ ON a a a G G G G G G a a ON a 6 ON OK OK ON 6 G G G c c III II! '-1 v-1 '-1 fl F1 1-1 "I N N N N N N +Total GDP - - GDP fromAgriculture + GDP from Manufactorlng Source: INE (2008) Agriculture is still the main contributor to the Mozambique economy, providing over 75 percent of employment for the population and contributing slightly less than 25 percent to the GDP. The majority of production occurs on small household farms“ with commercial farms focusing on exportable and/or cash crops. Arable land availability in Mozambique amounts to 36 million hectares. while only around 10 percent of the 36 6 Small farms are defined as less than 10 hectares. million hectares is currently being used. In addition, the country offers a wide variety of soil types and climate conditions making it suitable for production of a wide range of crops (Sebei, 2002). 2.1.1 Market Structure In Mozambique, maize grain7 is the most widely produced, sold and consumed staple among the country’s several food staples. Additionally maize is the only staple that is exported on a regular basis, primarily from the northern region, but also occasionally from the center providing income to rural maize-producing households. The majority of maize grain, around 90 percent, is produced in the northern and central regions of Mozambique, while the majority of the urban population and rural net purchasers of maize grain live in southern Mozambique. Transport is an important factor in the flow of maize grain from the production to the consumption zone. Mozambique has poor infrastructure in all main methods of transport, including roads, railroads and sea links, and rail options from the north to south are nonexistent. During colonial years, railroads were built east to west to move raw materials from inland to the coast, however these are not consistent with today’s need to move staples north to south. 7 Unless otherwise noted, maize grain refers to white maize grain only. Mozambique uses white maize grain for human consumption, while yellow maize grain is used for animal feed and for human consumption only during extreme food insecurity. Therefore, in this research, yellow maize will only be considered for use other than human consumption. 12 Figure 2.2: Map of Mozambique with Regions and Maize Flow Patterns North Malawi Zambia ‘ Nampula City . o. ‘ I . I I I v ‘ ' U n . \ Zimbabwe . 4 Center \ / \ ‘ \ Beira City Xai-Xai City South Maputo City Africa Source: Abdula (2005) Movement of grain from the north and center to the south by road is difficult and costly due to distance, combined with low road quality and increased fuel prices. Many portions of the road are likely be poor as World Road Statistics (2008) estimated that in 2002 only 19 percent of roads were surface paved, with an additional 27 percent paved in gravel. Additionally, high diesel fuel prices, which increased by 610 percent in real price 13 terms from February of 1995 to December 2003, make transport of maize grain by road extremely expensive (Abdula, 2005). The distance from Lichinga, the capital of Niassa in the northern region and an important production zone, to Maputo is over 1800 kilometers and truck transport is the only viable option. Although the transport cost from the north per kilometer is comparable to the cost per kilometer from the center, the added distance increases overall transport cost to a minimum of 1800 Meticals/ton (SIMA, 2008). An additional access problem for movement of maize grain from the north to the southern region of the country is created by the Zambezi River. Only one large commercial ferry, between Tete and Zambezia provinces provides access across the river. In addition, during the rainy season movement across the river typically becomes impossible due to flooding. A new bridge is being constructed but until then, the Zambezi River forms a natural border between the north and the rest of Mozambique creating two natural, separate markets within the country (Figure 2.2). Therefore, this research will focus on the market area of central and southern Mozambique. 2.1.2 Production As noted earlier, maize is the most widely produced staple throughout Mozambique. Total maize production in Mozambique has been increasing throughout the last fifteen years (Figure 2.3). Data suggest that this increase in production is due more to additional land being added to production than increased productivity, as yields have consistently fluctuated around 1 mt/ha. Figure 2.3 illustrates the increasing trend in production and land cultivated for maize, although the additional land added into maize production is minor compared to the amount of arable land still available for production. It is estimated that 88 percent of arable land has not been cultivated in Mozambique. Figure 2.3: Total Maize Production vs. Total Area Harvested for Maize 1800 1600 1400 l 200 1000 800 600 400 E 200 0 Source: +Total Production (1000 mt) - r. ~Area Harvested (1000 ha) x“x“@s“@s"’s°’@w“~“~“~“~°w“~°a9 USDA-FAS Production, Supply and Distribution Online Database Most maize is produced on small household farms for home consumption with few inputs. The majority of households use seed from the previous year with no fertilizer8 or irrigation. Therefore, total production is highly dependent on rainfall. Figure 2.4 illustrates the high dependence of production on rainfall, as yield does not vary throughout the years but total production in the central and southern regions moves with rainfall. Rainfall varies throughout the country, thus affecting regional production differently. According to Tschirley and Abdula (2007), the northern region typically has higher total production because the region generally has more reliable rainfall (along with 8 Only 3.5 percent of farms in Mozambique use chemical fertilizer (Cunguara, 2008) 15 better soil quality). Rainfall in the north tends to be independent and is not correlated with rainfall in the central and southern regions of the country. However, rainfall in the center and southern regions due tend to be correlated (Figure 2.4). This lack of rainfall correlation between the north and the rest of the country can be seen during years of drought in the center and southern regions, while the north remains relatively unaffected, such as 1992/1993. The center also tends to have better rainfall than the southern region, but is still more variable than the north (Tschirley and Abdula, 2007). Figure 2.4: Total Rainfall and Total Maize Production, Central/Southern Mozambique 14 p m... Total Rainfall (100mm)— ‘ Center . ,,,,,,, ’4. - -Total Rainfall (100mm)- 12 i f' / 't South ’ I’ + Production (1 00,000 mt) i ' .3 10 8 6 j 4 Q 1 i -. i / 2 95/96 99/00 02/03 Drought Flooding 0 ‘ Drought 1995 1996 199" 1998 1999 2000 2001 2002 2003 200.1 2005 2006 Source: USDA-FAS Production. Supply and Distribution Online Database (2008). FEWSNET(2008) The trend of higher production in the northern and central regions can be seen for the 2006/2007 crop season. According to the 2007/2008 Famine Early Warning System (FEWSNET) Projected Food Balance Sheets (2007), a little over 1.5 million metric tons l6 of maize was produced. Of that 1.5 million tons, 57 percent was produced in northern Mozambique, 37 percent was produced in the central region and the remaining amount, approximately 6 percent was produced in southern Mozambique. Data on total production can be deceiving since a majority of the maize produced is never sold on the market. As Table 2.1 indicates, in the 2001/2002 seasong, 50 percent of maize was produced in the north, almost 40 percent was produced in the center and almost 10 percent was produced in the south, which is comparable to production percentages for the 2006/2007 season discussed above. Percent of national sales is equivalent to percent of production. However, the percentage of households selling on the market is more evenly distributed in the northern and center regions. In 2002, approximately 24 percent in the north, 23 percent in the center and almost 4 percent of households in the south were selling their maize on the market (Abdula, 2005). Table 2.1: Percent Maize Production and Sales, 2001/2002 North Center South Percent of National Production 50.5 39.8 9.6 Percent of National Sales 59.1 38.5 2.4 Percent of Households Selling Maize 24 23.4 3.8 Data Source: 1AF(2002), Abdula (2005) 2.1.3 Consumption Maize grain and flour are the primary staples in both rural and urban areas throughout Mozambique, although consumption patterns have been shifting in the recent years, due to the relative changes in retail prices of the basic staples (Figure 1.1) and a general shift in, specifically urban, preferences for rice and wheat products. As can be 9 . . . . The 2001/2002 season rs used because 11 rs the most current household survey data available. 17 seen in Table 2.2, when aggregated across the entire country to include both rural and urban areas, maize and maize derivatives continue to dominate total expenditure of cereals with a little over 15 percent of expenditure in both 1996 and 2002 according to Inquierito aso Agregados Familiares (IAF), a Mozambique government household consumption and budget survey, while rice only accounts for around 4 percent of total expenditure allocated to cereals. Table 2.2: Food Expenditure Allocated to Cereals, National IAF 1996 IAF 2002 Percent Maize and Maize Derivatives 15.44 15.45 Rice 3.84 3.76 Wheat and Wheat Derivatives 1.66 1.31 Data Source: IAF (1996), IAF (2002), and Abdula (2005) It should be noted that expenditure patterns vary greatly by rural and urban areas. Table 2.3, shows the total expenditure percentages for cereals for 1996 allocated to the same categories separated into rural and urban categories. The percent of total cereal expenditure allocated to maize and maize derivatives decreases to around 10 percent for urban areas while over 16 percent of expenditure allocated to cereals is spent on maize and maize derivatives in the rural area. Table 2.3 also illustrates that urban households decrease their total cereal expenditures allocated to maize and maize derivatives and increase their total expenditure of cereals spent on rice and wheat to over 5 percent. 18 Table 2.3: 1996 Food Expenditure Allocated to Cereals, Urban and Rural Urban Rural Percent All Cereals 22.7 22.8 Maize and Maize Derivatives 10.56 16.68 Rice 6.77 3.09 Wheat and Wheat Derivatives 5.28 0.74 Data Source: IAF (1996) and Abdula (2005) Expenditure patterns also vary across provinces (Table 2.4). In the highly urban province of Maputo, the urban population slightly increased total expenditure for cereals spent on maize from 1996 to 2002, although it still remained under 3 percent with a majority of their cereal expenditure spent on maize substitutes such as wheat products and rice. Overall, rural areas of Maputo province allocated slightly more of their total cereal expenditure to maize and maize products, however, from 1996 to 2002, total cereal expenditure spent on maize decreased by 7 percent. During the same time period, the rural Maputo population increased their total expenditure of cereals spent on rice by 5 percent. However, maize remained an important staple, over 10 percent was allocated to maize expenditure in both the rural and urban areas of Gaza and Inhambane, while over 40 percent of total cereal expenditure was allocated to maize in the rural and urban areas of Manica and Tete. Therefore, even with increasing prices of maize products as compared to other staples as discussed in Chapter 1, it still remains an important staple throughout Mozambique. l9 Table 2.4: Food Expenditure Allocated to Cereals, by Province and Urban/Rural Maputo Gaza and Manica and Sofala Inhambane Tete Urban 1996 2002 I 1996 2002 [ 1996 2002 L1996 2002 Percent of Total Food Expenditures Maize 1.1 2.4 10.1 14.5 24.6 39.9 19.4 27.5 Rice 15.0 7.8 16.2 9.8 6.1 4.4 8.9 9.2 Wheat 21.7 15.5 16.3 6.0 5.9 2.9 8.5 4.2 Rural 1996 2002 | 1996 2002 l 1996 2002| 1996 2002 Percent of Total Food Expenditures Maize 16.4 9.1 22.4 12.0 23.2 48.0 18.8 26.7 Rice 6.2 11.4 5.7 9.5 1.1 2.5 7.8 6.5 Wheat 6.1 7.4 3.4 3.2 0.9 1.4 0.5 1.7 Data Source: IAF (1996). 1AF(2002), Abdula ( 2005) 2.1.4 Milling Industry As explained by Tschirley and Abdula (2006), there are three main types of maize millers in Mozambique. The first is home milling, where the household buys or produces the maize grain and hand pounds the maize to produce the maize flour. Secondly, there are small-scale hammer mills, where the customer provides the maize grain and the hammer mill provides the service of pounding, usually producing a straight run meal. Finally, there are industrial millers who purchase maize and then process and sell various qualities of maize flour at wholesale and retail levels. Different qualities of maize flour can be produced from these different milling options. Extraction rates can go from 100 percent, where all of the maize grain is used in the creation of the flour (typical of small- scale hammer mills), to 65 percent extraction rate where the maize germ and other components are removed. For this study, only the highest quality of maize flour will be examined. Within the last 15 years, large industrial millers in Mozambique have started to gain market share. Throughout the 1980’s, the maize milling industry declined due to poor management and the inability to obtain inputs due to the political instability. By the mid-1990’s, Compannia Industrial da Motala (CIM), a large industrial maize milling company in the southern region of Mozambique, was privatized and operating as the only maize miller in southern Mozambique. Today Merec Industries (Merec) is the only real competitor for CIM in the commercial maize milling industry in southern Mozambique. According to a survey conducted in 2005, these two companies held over 70 percent of the maize flour market in both central and southern Mozambique, and 100 percent of the market in Maputo City (Tschirley and Abdula, 2007). However, CIM with its highest quality maize flour, Top Score, continues to dominate the market in Maputo City, which is reinforced by results from a 2007 survey (Tschirley and Abdula, 2007) that indicated in open air markets in Maputo, CIM held 70 percent of the market while Merec products were almost completely absent. This is consistent with results found in the supermarket during the same time period. A survey of the shelf space revealed that CIM’s products occupied 70 percent of the shelf space, while Merec products had 13 percent and imported maize flour from South Africa had 5 percent of the shelf space (Tschirley and Abdula, 2007). The dominance of CIM in the market may be important considering the increasing prices of maize flour. During the first few years of production, CIM’s price was compatible to similar maize flour in surrounding countries. Beginning around 2002, Top Score maize flour prices increased and by 2005 Top Score was three times the price of maize grain in Maputo, four times the price of maize grain in any other city in the center and southern regions of Mozambique, and double the equivalent brand of maize flour in Zambia and Malawi (Tschirley, et al, 2006). Nevertheless, competition from other large and small millers remains limited. In the last 5 years, other large industrial millers have opened in and around Maputo, including SMC and Inacio de Sousa, but, neither company’s brand of maize flour is aggressively competing for market share in Maputo City (Abdula, 2005). Vonk (V&M), one of the major industrial millers in the center region of Mozambique, has begun shipping maize flour to the south and targeting the Maputo market to gain market share. V&M, which uses maize grain produced in the center region of Mozambique, determined that even with the cost of shipping the maize flour from the center to the south, the current high cost of maize flour in Maputo offered an opportunity to enter the Maputo market. However, V&M’s attempt seem to have been unsuccessful as it was not mentioned in the 2007 survey on maize flour in either the open air markets or retail stores (Tschirley and Abdula, 2007). 2.2 South Africa Traub and Jayne (2008) explain that starting in the 1930’s, South Africa’s maize production was a single-channel system controlled by the government. A Maize Board set prices at every stage of the processing chain, including the producer maize grain and retail maize flour prices. In the 1980’s the South African government recognized the inefficiencies caused by setting prices and controlling distribution and began the process 22 of privatizing the maize marketing sector. By 1991, deregulation was in full force and maize flour prices were no longer set by the government. At the end of the 1996/1997 marketing season, the Maize Board was dissolved (Traub and Jayne, 2008). 2.2.1 Production and Consumption South Africa is the leading producer of maize in the southern African region. In a typical year, South Africa produces more maize than Zimbabwe, Malawi, Mozambique, Zambia, Swaziland, Lesotho, Botswana and Namibia combined (Jayne, 1995). According the South African Department of Agriculture (2008), maize is produced in five provinces of South Africa, with most production occurring in the Free State province, followed by North West, Eastern Cape, Mpumalanga and finally KwaZulu- Natal. Planting typically occurs between October and December due to the rainfall variation, usually starting in the eastern part of the country and moving west. South Africa plants an average 3.8 to 4.8 million hectares of maize per year, which accounts for approximately 25 percent of total arable land in South Africa. Due to the lingering reminisces from the regulated production by the Maize Board, most maize production occurs on large commercial farms, as small-farms did not have access to contracts when the Maize Board regulated production. More specifically, approximately 400 large commercial farms supplied maize grain to feed 40 million people during the regulated era. Today, these commercial farms still produce the majority of maize grain used for commercial processing and exportation due to the strict phytosantary standards enforced by domestic and trade laws. On average, these large commercial farms produce 4.3 million tons of white maize per year (in addition to 3.9 million tons of yellow maize). 23 Additionally smaller farms produce around a half a million tons of white maize, mostly for household consumption (South African Department of Agriculture, 2008). Maize is also considered the most important staple for domestic consumption by the population of South Africa. On average South African’s consume around 4.4 million tons of white maize (and 3.1 million tons of yellow) each year that is supplied from domestic production, in addition to imports from the United States, Argentina and Kenya (Dept of Agriculture 2008). 2.2.2 Milling Industry Traub and Jayne (2008) explain that before the deregulation of the maize market, a single-channel, regulated, government system created an institution where only certified producers could sell to the marketing board and only registered maize millers could purchase maize from the marketing boards, thus limiting competition in the production and processing sector. The number of industrial millers slowly decreased as many were consolidated. By the 1980’s, only six industrial maize millers remained. The pre- determined producer prices and set marketing margins were intended to keep maize flour prices in line, however, the maize millers worked together to create a cost-plus pricing system, which led to significant increases in maize flour prices, while new entrants were blocked due to the nature of the maize marketing system(Traud and Jayne, 2008). The deregulation of the maize marketing system was intended to eliminate the previous government created barriers to entry and in turn lead to a decrease the maize flour prices (Traub in Jayne, 2008). Deregulation permitted competitors, including small- scale millers, to enter the market and provide additional sources of supply for consumers. 24 Small-scale millers offered a range of different quality maize flours. Together these effects were intended to decrease maize marketing margins and in turn decrease maize flour prices. However, the intended results have not been observed. Traub and Jayne (2008) indicate that marketing margins rose up to 22 percent in the first three years of deregulation adding that today marketing margins continue to increase monthly. Unfortunately for small-scale maize millers, industrial maize millers again have a comparative advantage, as the South African government requires maize flour to be fortified with vitamins if it is to be sold at a retail level. The additional inputs, including machinery, are expensive and access to the needed materials can be difficult to obtain, therefore creating a new barrier to entry. Such a policy favors large maize miller’s creating a situation in which large maize millers could return to their oligopoly marketing practices (Jayne 2008). 2.3 South Africa - Mozambique Trade Trade between South Africa and Mozambique dates back to the late 19'h century. Originally a relationship built of the movement of migrant labor and transportation, trade and investment eventually became the focus. Although a seemingly obvious choice for trading based on their proximity to one another, Castel-Branco (2004) concluded that four main factors have shaped trade between Mozambique and South Africa. These are the regional strength of South Africa as a trading partner, South Africa’s international economic standing in comparison to major trading powers, the weakness of Mozambique’s economy and public policy, and finally the importance of minerals and energy in social, political and economic dynamics in South Africa. Table 2.5: Historic Trade Patters between Mozambique and South Africa Exports from South Africa to Exports from Mozambique to Mozambique South Africa Real 04 Meticals 2003 1,677,448,897 82,984,649 2004 1,479,930,030 59,703,003 2005 1,804,614,718 56,172,132 2006 1,680,744,273 85,806,] 1 1 Data Source: SA Government Info (2008) South Africa is Mozambique’s main trading partner. South Africa accounts for over 40 percent of Mozambique’s total imports, while Mozambique accounts for 20 percent of South Africa’s total imports. Table 2.5 shows the past four years of total trade for Mozambique and South Africa in real Meticals. Mozambique’s imports from South Africa are highly concentrated on maize, cereals, meal and pellets, while Mozambique’s exports to South Africa are focused on nickel ores and concentrate and cotton (SA Government Info 2008). South Africa’s importance as a main importer into Mozambique extends into Mozambique’s imports of maize and maize flour. In recent years, South Africa has typically been the main exporter of maize flour into Mozambique (supplying 100 percent of Mozambique’s maize flour imports), and has become increasingly more important as main source for maize grain imports in Mozambique (Table 2.6). In 2005, South Africa’s importance as a maize grain exporter to Mozambique increased from 30 percent of total maize grain imports to over 90 percent of Mozambique’s total maize grain imports. In 2006, South Africa continued to be the main exporter of maize grain to Mozambique and accounted for over 75 percent of total imported maize grain into Mozambique. 26 Table 2.6: Imports of Maize Grain and Maize Flour to Mozambique, 2002-2006 Imports to Imports to Percent of Total Mozambique - Mozambique - Imports from World South Africa South Africa Maize Grain (MT) 2002 4,723 1,000 21.15 2003 21,135 7,686 36.37 2004 74,734 22,633 30.28 2005 39,91 1 37,303 93.47 2006 107,439 82,256 76.56 Maize Flour (MT) 2002 l ,073 1 ,07 3 100 2003 1,259 1,259 100 2004 1,067 1,067 100 2005 3,453 3,363 97.39 2006 2,628 2,628 100 Source: United Nations UNCOMTRADE Database (2008) Although, Mozambique increasingly relies on South Africa for a main source of maize grain and maize flour imports, Mozambique is not the primary destination of South Africa’s exports of maize grain and flour (Table 2.7). In recent years, Mozambique has only accounted for a maximum of 30 percent of South Africa’s maize flour exports (2006) and has accounted for as small as 2 percent of South Africa’s maize flour exports (2002). Although increasing in recent years, exports of maize grain to Mozambique has only accounted for a maximum of 14 percent of South Africa’s total exports of maize grain and again occurring in 2006, with the lowest percent of total exports, less than 1 percent, occurring in 2002. In 1980, Mozambique was one of nine founding members of SADC whose original goal was to coordinate economic integration between the members and to lessen their dependence on then apartheid—controlled South Africa (SADC 2008). Shortly after South Africa ended apartheid rule in 1992, South Africa joined an additional 13 countries 27 in SADC. Since its establishment in 1980, the goals of SADC have been amended and it currently serves as a free trade region, along with other development and economic growth activities. Although the block focuses on free trade agreements between member countries, some commodities are exempt from free-trade status and countries are allowed to apply tariffs based on the sensitivity of the commodity. Maize is one of the exempt commodities due to its importance in domestic food security. Mozambique uses this exemption to apply an escalating import tariff on maize and maize flour, in addition to a VAT to maize grain that together provide significant protection for Mozambique’s domestic industrial maize processing industry (SADC 2008). Table 2.7 : Exports of Maize Grain and Maize Flour from South Africa, 2002-2006 Exports to Exports to Percent of Total the World Mozambique Exports to Mozambique Maize Grain (MT) 2002 478,275 1,000 0.21 2003 624,435 7,686 1.23 2004 423,535 22,633 5.34 2005 2,100,926 37,303 1.78 2006 603,861 82,256 13 .62 Maize Flour (MT) 2002 60,807 1 ,073 1.76 2003 9,053 1,259 13.91 2004 5,053 1,067 21.11 2005 7,1047 3,363 4.73 2006 8,832 2,628 29.76 Source: United Nations UNCOMTRADE Database (2008) 28 2.3.1 Import Tariffs Mozambique applies tariffs on all imports of maize grain and maize flour. An import tariff of 2.5 percent is applied to all maize grain imports and, as of January 1, 2006, a 20 percent tariff is applied to all maize flour imports. Prior to January 1, 2006, Mozambique applied a 25 percent tariff to all maize flour imports. The high tariff levels effectively protect the domestic maize processing industry, which can be seen by the low levels of maize flour imports (2,628 tons for 2006, all of which originated from South Africa) and the fact that only 5 percent of maize flour self space is devoted to imported maize flour (Tschirley and Abdula 2007). Mozambique is to have all import tariffs for maize flour eliminated by 2012 for all SADC member countries excluding South Africa, which will have the import tariffs eliminated by 2015 (Tschirley et a1 2006). 2.3.2 Value Added Tax (VAT) In 1999, Mozambique added a 17 percent VAT to their tariff schedule that is applied to maize grain. The VAT has two conditions that cause a disproportionate affect on maize grain as compared to its substitutes or maize flour. First, the VAT is not applied to imports of wheat or rice. As a result, imports of wheat and rice have a 2.5 percent import tariff applied to the free on board (FOB) price, while maize grain has the 17 percent VAT in addition to the same 2.5 percent import tariff applied to its FOB price. This leads imported maize grain to have a cost disadvantage compared to imported rice and wheat. The second disadvantage occurs between imported maize grain for retail level sales and maize grain imported to be processed by large industrial millers into maize 29 flour. The VAT regulation stipulates that imported maize grain that is purchased by large industrial millers and processed into maize flour will be reimbursed for the entire VAT. However, the VAT is not reimbursed, if maize grain is imported and sold on the retail level without processing or imported to be processed at a small-scale processor (Tschirley et a1 2006). These two conditions restrict imported maize grain on the retail level to almost zero. It also adds an additional level of market protection for the industrial maize processors, specifically in the south whose market is currently saturated by two maize processors. Southern industrial millers benefit from this current policy as it provides them a continual supply of maize grain without having to deal with unreliable supplies and high transport costs of maize grain from the center or northern regions of Mozambique. Southern industrial millers also benefit from higher quality maize grain from South Africa who impose and enforce higher quality standards than maize produced domestically within Mozambique. The only disadvantage industrial maize millers face from the VAT is a loss in opportunity cost for the money used to pay the VAT. Although the government guarantees that the VAT will be reimbursed within three months (or the government will provide interest for the additional time taken to reimburse), industrial millers report varied results. Industrial millers do report that they receive the reimbursement, though the process is long and difficult (Abdula 2005). 30 2.4 Conclusion This chapter has provided an in-depth look at the maize sector in Mozambique including detailed information on production and consumption trends throughout the country, and the maize milling sector. South Africa’s maize market has been briefly summarized. The importance of trade between South Africa and Mozambique, specifically for Mozambique, has been outlined and a description of the import tariffs applied to maize and maize flour have been provided. Based on the information in this chapter, it should now be clear the importance of maize grain and maize flour throughout Mozambique. The next chapter will review previous research on the topic of import tariff escalation and begin to examine the appropriate method for measuring the effects of import tariff escalation on prices and quantities traded. 3. LITERATURE REVIEW 3.1 Introduction There is a broad range of literature on the analysis of import tariff escalation, although no research has examined import tariff escalation in Mozambique. Nevertheless, the studies and methods of analysis used to examine import tariff escalation vary considerably, as do the countries and regions where the escalation occurs and has been studied. In addition, while import tariff escalation has been considered and researched as a north-south problem, it has begun to be examined as south—south trade issue as well, although the amount of research in that area is limited. 3.2 North - South Import Tariff Escalation Analysis As noted in Chapter 1, analysis of import tariff escalation has historically examined the impacts on developing countries due to the use of escalating import tariffs by developed countries. The most popular methods for this type of analysis of import tariff escalation include (1) nominal tariff analysis and effective rate of protection (ERP) measurement, (2) partial equilibrium models, and (3) other economic based models. 3.2.1 Nominal Tariff Wedge and Effective Rate of Protection Nominal tariff wedge analysis and Effective Rate of Protection (ERP) were Previously the two most commonly used methods of import tariff escalation analysis, and bOth methods are still used today. Nominal tariff wedge analysis is one of the more elementary methods to analyze import tariff escalation. The nominal tariff wedge is calculated as the difference between the tariff applied to the processed and raw commodity. If the tariff wedge is greater than zero, import tariff escalation between the commodities is present (Tangermann, 1989). Alternatively, the ERP is calculated as the difference between value added to a processed good at the distorted trade prices and the value added to the processed good at the free trade price divided by the value added at the free trade price (Balassa, 1965; Corden, 1966). Economists criticize the nominal tariff wedge approach because (1) it does not fully represent the protection caused by the import tariff escalation policy, (2) it provides no information on the effects of the value added to the product through the processing level, and thus cannot allow for comparison across commodities, and (3) it cannot be applied to the production function of multiple input or output analysis’ (Lindland, 1997; Tangermann, 1989). Critiques of the ERP have also been made from both a theoretical and methodological standpoint (Antimiani et al, 2003; Greenaway and Milner, 2003; Anderson, 1998; Ethier, 1971, 1977; Tangermann, 1989). The authors collectively note that the model holds some very restrictive assumptions including: ( 1) domestic and foreign products are perfect substitutes, ( 2) no other trade restrictions, including quota and non-tariff trade barriers are present, (3) competition is perfect and (4) the importing country is a small, price taking country with no impact on the world price. However, the strongest critiques occur with the assumptions made of fixed input coefficient and the neglect of general equilibrium repercussions from a change in import tariff escalation. The fixed input coefficient assumption does not allow for an accurate measurement of the protection added because the input-output ratio is not allowed to vary under different tariff rates, which would normally happen if substitutes for the inputs were available. Substitution elasticities would be needed to model this movement 33 between inputs, but is not used and thus biases the level of protection calculation (Tangermann, 1989). The disregard for the effects throughout the economy due to tariff changes is a major critique of both the nominal tariff wedge analysis and the ERP method. Specifically, in the analysis of a large sector of the economy or multiple small sectors that are aggregated together, the lack of acknowledgement of the change throughout the economy due to a tariff change is serious. Domestic incomes and expenditures, prices of inputs and substitutes, in addition to exchange rates may be affected due to a change in tariff policy and is not accounted for in either the nominal tariff wedge analysis or the ERP (Tangermann, 1989). Even with these critiques, due to the seemingly straight forward method of calculation for measuring import tariff trade barriers, both methods have been used in a variety of studies to analyze import tariff escalation. The Organization for Economic Co- operation and Development (OECD) argues that the ERP measurement is almost impossible to use in agriculture import tariff escalation, due to the fixed input coefficient assumption in the model and the need for flexibility of the input coefficient based on agricultural processing. The OECD argues that the nominal wedge is just as useful as an ERP measurement because an overall decrease in the absolute value of the nominal tariff wedge implies a decrease in protection of the industry and that the direction of change is more important than the overall calculation of protection (OECD, 1997). However, most research on import tariff escalation typically uses the nominal tariff wedge approach as an introduction and then calculates the ERP to further examine the degree to which the processing chain is protected (Lindland, 1997; Chevassus-Lozza and Gallezot, 2003; Humphrey, 1969; Milner, 1990; Greenaway and Milner, 1990; Greenaway and Milner, 1987; Hassan et al, 1992). Lindland (1997) examined the impact on import tariff escalation due to policy changes that were negotiated in the Uruguay Round of the WTO using a nominal tariff wedge approach. Lindland analyzed over 200 different commodities in the three main agriculture markets (the United State, the European Union and Japan) between the years of 1992 and 1994. He found that over 80 percent of all tariff wedges in all three markets had converged to zero and commodities with the highest bound tariff wedges had the greatest reduction. Data difficulty including aggregation, processed products using multiple raw commodities for production, raw commodities used for multiple processed products, and matching the FAO and FAOSTAT data code for the nominal wedge analysis made the approach difficult to implement (Lindland, 1997). In an additional paper, Lindland attempted to calculate the ERP using the same data. However, the data were too aggregated and could not be used to complete a meaningful analysis of the level of protection added from import tariff escalation (Lindland, 1997a). Chevassus-Lozza and Gallezot (2003) contended that the level of nominal protection, combined with the import tariff escalation along the production chain, provided reliable information on the sign of ERP, but not the degree of protection from the import tariff escalation, as claimed by the OECD (1997). However the authors felt the degree of protection caused by the escalating tariffs was needed to completely analyze import tariff escalation occurring between multiple developing countries and the European Union over major commodity groups. They sought to determine how a decrease in import tariff escalation, required by the Uruguay Round, would change the DJ 'Jt trading relationships between countries, specifically those with preferential trading agreements and sensitive geographic regions. The authors noted the difficulties of using the ERP model, including difficulty in identifying the processing chain and standardization/classifications of the tariffs, but used previous research for clarification. Chevassus-Lozza and Gallezot (2003) first created a baseline using the current trading relationships and accessibility to European Union (EU) markets. The authors remark that imports into the EU had been increasingly originating from countries with preferential agreements with the EU. However, when differentiating between the types of preferential agreements, the authors found that countries with generalized system of preferences imported higher quantities to the EU than countries with bilateral trade agreements. The authors also noted that although there had been a slight increase in the percent of products being imported as processed commodities, the majority of agricultural imports into the EU were raw commodities, while exports from the EU were mainly processed commodities. The authors then simulated the effect that the Harbinson Proposal10 would have on import tariff escalation and trade of raw and processed goods, and found that in general, it would lead to a drop in tariff protection, of an average of 10 percent. Generalized system of preference countries would benefit the most from this proposal, While African and Pacific countries (APC) would be negatively affected as their preference margins decrease, specifically causing a decrease in exports of cereal, '0 The Harbinson Proposal, presented to the WTO in 2003, proposed that (1) all ad-valorem tariffs greater than 90 percent be reduced to 60 percent, (2) reduce ad-valorem tariffs that are less than 90 but greater than 15 percent to 50 percent, and (3) for all levels below 15 percent ad- valorem, decrease by 40 percent. 36 fruits/vegetables, meat, sugar, and coffee/tea/cocoa, which are the primary export commodities in APC countries. Hassan et a1 (1992) analyzed the Egyptian agriculture sector through the nominal tariff wedge approach and effective rate of protection. The authors examined 22 commodities, which accounted for 78 percent of the total Egyptian agriculture trade value between 1980 and 1987. The authors did not aggregate the commodities in an attempt to gain the best individual analysis of the commodities. After using the nominal tariff wedge to verify the presence of import tariff escalation, both Cordon and Balassa’s approach were used to calculate the ERP. The authors found that when the market exchange rate was used, as compared to the official exchange rate, Egyptian agriculture production faced an overall disincentive due to the protective tariff structure. 3.2.3 Partial Equilibrium Models Multiple researchers (Clark 1985, Golub and Finger 1979, Tangermann 1989, Wailes er al 2004) have measured the effects of trade policies, including tariffs, through the use of partial equilibrium models. First used by Coumot and Marshall, partial equilibrium models occur under assumed constant prices of substitutes/complements of the commodity and constant levels of income, therefore allowing only prices of the given commodity to adjust to create an equilibrium condition where supply equals demand. Golub and Finger (1979) used a partial equilibrium model with fixed constraints to measure the effects of developing country export taxes and developed country import tariffs on trade, specifically on primary and processed products. The authors analyzed eight different commodities, including cocoa, cotton and coffee and aggregated all 37 countries into two categories, developed or developing countries. Using data from 1973 on trade flows and production levels, the authors ran three different scenarios through the fixed coefficient, partial equilibrium model, which was based on six behavioral relationships of supply and demand of the different processed level products. Table 3.1 presents the results found by Golub and Finger. As hypothesized, developing countries experienced overall increases in production of both primary and processed commodities and increases in export revenue under import tariff removal and the free trade scenario. Specifically, under the tariff removal scenario, developing countries would experience an average of 23 percent increase in processing, with cocoa, coffee and wool showing the greatest increases. Consumption of both the primary and processed commodities decreased under all three scenarios, however, the average change was under 5 percent in all three simulations. Although developed countries would see decreased processing under the tariff removal and free trade scenario, the average percent change was found to be no greater than 3 percent, with the copra processing sector experiencing the largest decline of around 40 percent under both scenarios. Tangermann (1989) also used a partial equilibrium model, to analyze two different commodities, cocoa and soybeans. All major trading countries were grouped as either net importers or net exporters and developing or developed countries. Tangermann used consumption and production data from 1981 to 1983 and supply and demand elasticities provided by the Food & Agriculture Organization (FAQ). Using the most favored nation tariff structure in ad valorem form and ignoring transportation cost, Tangermann analyzed both cocoa and soybean trade by comparing different free trade tariff scenarios to the current baseline situation. A brief outline of the results are presented in Table 3.2. Table 3.1:: Results, Golub and Finger (1979) World Developing Countries Developed Countries Scenario 1: Removal of Import Tari 3 Consumption Increase Decrease Increase Primary --- Increase Increase Production ProcessirgL --- Increase Decrease Export Revenue --- Increase --— Scenario 2: Removal of Export Taxes Consumption No Change Decrease Increase Primary -—- Increase Decrease Production Processing --- Decrease Increase Export Revenue --— Decrease --— Scenario 3: Free Trade Consumption Increase Decrease Increase Primary --- Increase Decrease Production Processing --- Increase Decrease Export Revenue --- Increase -—- Source: Golub and Finger ( 1979). see for detailed results Table 3.2: Results, Tangermann (1989) Results Cocoa Scenario 1: Import Tariff Removal Overall increase in trade. Developing Countries — small increase in processing. Scenario 2: Overall - decrease in export revenue. EXPO" Tax Overall - decrease in processing of beans and paste. Removal Overall - decrease in exports of butter and powder. Scenario 3: Exporters — decrease in foreign receipts. Free Trade Developing Countries — increase in first stage processing. Soybeans Scenario 1: Import Tariff Removal Increase in world price of raw and processed commodities. Developing Countries —- increase in processing. Developing Countries — negative foreign exchange. Scenario 2: Export Tax Removal Developing countries — increase in export revenue Overall — increase production. Overall — decrease in world market price. Scenario 3: Free Trade Developing countries — increase in export revenues. Developed countries — increase in export revenues. Importing countries — increase import expenditure on soybean commodities. Overall — small increase in world price of both raw and processed commodities. Source: Tangermann (1989), see for detailed results In the cocoa analysis, Tangermann (1989) noted that the import tariff removal scenario did not indicate that the change in policy would transfer significant amounts of processing to the developing countries. He hypothesized that the less than dramatic increase in the growth of the processing sector in the developing countries was due partly to the low processing margins in the European Economic Community (EEC) and the United States. With the change in tariff policy, the original margins created an 40 opportunity for large increases in processing of cocoa beans in both the US and EEC. The elimination of export taxes scenario also produced counterintuitive results, specifically while all other countries showed negative export revenues due to the change in policy, Cameroon and Ghana showed increases in export revenue. This result suggests that the export tax was encouraging inefficient processing and discouraging domestic raw bean production. With the elimination of the tax, revenue increased due to an increase in exports of raw beans and movement away from the less efficient processing and exportation of processed cocoa products. Overall, the scenarios in both cocoa and soybeans provided conflicting results, indicating less than hypothesized adjustments in processing between developed and developing countries due to the removal of tariffs. Many authors have also used a partial spatial equilibrium model to examine changes in prices and flow of commodity due to a change in trade policy (Bates and Schmitz 1969, Zusman er al 1969). Wailes et.al (2004) also chose to use a partial, spatial equilibrium model to analyze the effects of import tariff escalation on the United States and global rice trade and prices, specifically focusing on trade between the United States and Latin American countries. The authors used the RICEFLOW© model to maximize the net monetary social welfare to make comparisons of the effects on tariff removal. Through the use of 2002 trade flow and tariff data, the authors simulated two trade scenarios and compared the results to a baseline scenario (Table 3.3). The results of scenario one indicate that harmonization at higher tariff level has a higher overall negative effect on exports and production in the exporting countries than the import tariff escalation policy. Analysis on a country level indicates that the US. rice millers would be the only beneficiaries of a change in policy, as less paddy rice is 41 exported and thus must be milled in-country for domestic consumption. Although the aggregated results indicate no change in consumption of importing countries, consumers in Nicaragua and Guatemala would be worse off from decreased imports of paddy rice which is not supplemented by increased import levels of milled rice. The aggregated results of the second scenario are compatible with hypothesized theory that the removal of escalating import tariffs lead to increases in production and exports in the exporting countries, while consumption increases in the importing countries due to lower trade barriers. Table 3.3: Results, Wailes et a1 (2004) Importers Exporters Scenario 1: Harmonization at the milled rice tarifl level Consumption No Change Increase Production No Change Decrease Total Trade Decrease Decrease Scenario 2: Harmonization at zero tarifif level — free trade Consumption Increase Decrease Production Decrease Increase Total Trade Increase Increase Source Wailes er a1 (2004). see for detailed results 3.2.4 Other Economic Based Methods Production Economics Authors have chosen different methods besides examining the effects of import tariff escalation on a country through the use of ERP, nominal tariff wedge analysis, or partial equilibrium analysis. Guha-Khasnobis (2004) conducted research based on the 42 continual complaints originating from developing countries that tariff reductions in developed countries suggested under the Uruguay Round Agreement were not uniform. More specifically, deve10ping countries argued that there were large variations in reductions throughout different production chains and among the different importing countries. The author analyzed the impact of import tariff escalation by measuring the changes in factor prices, specifically wages, through the use of a simple production and trade model. Guha-Khasnobis chose wage as the primary measurement to illustrate the different affects that occur on unskilled and skilled labor forces due to the removal or decrease of import tariff escalation in developing countries. The author used a two country, two good model that represented one developed and one developing country and one processed and one raw commodity. After simultaneously solving the production and trade equations, the results indicated that if the developed country would harmonize all tariff rates (no escalation), then wages of unskilled labor would rise relative to that of wages for skilled labor. However, the author notes that institutional constraints, specifically lobbying ability of each prospective labor group, could be an unexamined cause of import tariff escalation. The author argues that if the skilled labor group has stronger organized lobbying skills than unskilled labor, the country would be more prone to having a government system that protected the service sector more than the manufacturing sector, resulting in an increase in wage for skilled labor relative to unskilled labor. Therefore, the author concludes that because of the high degree of lobbying power by skilled labor groups, attention tends to be diverted away from the benefits that would occur through de- escalation, i.e. increased wages for unskilled labor in both developed and developing 43 countries and instead escalation of tariffs continues to protect the skilled labor in developed countries. Industrial Organization McCorriston and Sheldon (2004) approached import tariff escalation analysis in a different manner. The authors used a three-stage game theory approach to test their hypothesis that import tariff escalation is not the only factor excluding developing countries from access to developed country markets. Instead, the authors believed it was the structure of the market and industry that was restricting assessability to developed country markets. Using descriptive statistics, the authors found that at the processing level in developed countries, specifically the European Union and the United States, there were high levels of market concentration. At the retailing level, the authors found high concentration levels in the European Union, however, the United States were much less concentrated. Based on these conclusions, the authors conducted a three-stage game to measure the importance of market access and import tariff escalation. The authors found that as theory suggested, a change in tariff at one processing level had a corresponding affect on a different processing level. However, the authors concluded that the degree of impact from a tariff removal or reduction would have differing affects depending on the nature of competition, in addition to the availability of substitutes of the commodity. 44 3.3 South-South Import Tariff Escalation Analysis Little analysis has been done on the use of import tariff escalation in developing countries, as most research examines the effects developing countries encounter when import tariff escalation is used to block access to developed country markets. What little analysis is available examining the effects of import tariff escalation by developing countries focuses on the issue of south—south trade through nominal tariff analysis and partial equilibrium models. Even fewer studies have examined the impact on prices and quantities traded due to the application of escalating import tariffs. Safadi and Yeats (1993) illustrate, through the nominal tariff approach, that import tariff escalation is not only a north—south problem but also a south-south trade problem. The authors analyzed data from 1970 to 1990 on 48 commodities, which were aggregated into four categories of agriculture materials, food/feeds, ores/metals, and energy products. They examined countries in Asia because the region provided a mix of developed and developing countries. Fifteen countries were selected and aggregated into South Asia, non-OECD East Asia, OECD East Asia or OECD Asia categories. The authors then used the European Economic Community (EEC) as the baseline for the analysis. The authors concluded that overall the Asian countries are more biased against processed products when compared to the European Economic Community. More specifically, the authors concluded that Asian countries are more biased against processed commodities in intra-regional trade as compared to non—regional markets. They also concluded that between 1970 and 1990 the data showed no evidence of a narrowing of the bias when compared to the EEC data. Based on these results, the authors examined the average import tariffs for both primary and processed products for ten selected countries. They found that countries differ greatly with applied tariffs, but that import tariff escalation occurs to some degree in almost all of the processing chains, with Japan having the highest percentage (90 percent) of processing chains with import tariff escalation. The authors conclude that import tariff escalation in Asian markets has a restrictive effect on intra-regional trade of processed goods. As a result, Safadi and Yeats, concluded that import tariff escalation is not just a north-south problem, as suggested in the Tokyo and Uruguay trade negotiation rounds, but also a south-south problem. Laird and Yeats (1987) used the UNCTAD trade policy simulation model, which is a partial equilibrium model to estimate changes that would occur from the elimination of import tariff escalation in a variety of developing countries on some fifteen different commodities, which were aggregated into broad categories. Before conducting their analysis, the authors noted that a high degree of import tariff escalation was evident in developing countries. Where escalation was not present, the authors determined it was partly due to the lack of production capacity of the unprocessed commodity and partly due to the fact that non-tariff barriers were present in the absence of tariffs. Using the preferential trade agreement tariff structure, which allows commodities from developing countries to enter a country with no applied import tariff as the baseline, the authors ran the UNCTAD under different scenarios of supply elasticities. Under the first scenario, of perfect elasticity, the authors found that free trade led to an increase of 6 percent in values of exchange, with an average increase of 1.4 percent in imports. Under 46 the second scenario, unitary elasticity, the results were less than desired, however, the estimated prices still remained 10 percent above 1981 prices. Valenzuel et al (2004) used a standard computable general equilibrium model to measure the changes in welfare on smallholder livestock producers in eight developing countries, including Mozambique. Their main objectives were to determine the impacts on overall income, income distribution, and poverty from the removal of tariffs in both developing and developed countries. They simulated the changes in tariffs through the use of a global trade model created by the World Bank, GTAP. The authors considered both short and long term effects from the removal of the tariffs and found that the majority of the countries had a decrease in overall poverty levels, both in the short and long run, from the removal of the tariffs, with the exception of the Philippines in the short run and Zambia, with no change in the long run. 3.4 Conclusion This chapter has provided a detailed discussion of past research on the analysis of import tariff escalation, including detailed methods used to analyze import tariff escalation, limitations to the methods, and results obtained from the research. Using the information found in this prior research, the next chapter will provide an in-depth theoretical discussion of the chosen method for modeling import tariff escalation in trade of maize and maize flour between Mozambique and South Africa. 47 4. SOCIAL WELFARE MAXIMIZATION 4.1 Introduction There are a variety of methods used to conduct analysis of import tariff escalation. Nominal wedge and Effective Rate of Protection are two specific methods previous researchers have used to examine import tariff escalation. As noted in Chapter 3, however, the disadvantages of these types of measurements outweigh the advantages. Linear programming has been extensively used in analysis of policy, but, the design of the model does not allow the price mechanism to simultaneously change production and consumption. To overcome this restriction, many researchers have used nonlinear quadratic programming models that permit for simultaneous interaction between production and consumption, specifically in general and partial equilibrium models (Durand-Morat and Wailes 2003). General equilibrium models, such as computable general equilibrium models (CGE), permit the simultaneous interaction of all actors in the system, including producers, consumers, importers, exporters, government, etc. The data intensity of this approach and the difficulty of identifying individual effects in such a large and complex data set often lead researchers to consider partial equilibrium models (Durand—Morat and Wailes 2003). Partial equilibrium models permit researchers to focus on a specific commodity and examine changes that occur in that market due to changes in policy or other factors. In addition, partial equilibrium models are easier to manipulate and understand. Partial equilibrium models are not without their faults, including the lack of interaction between the commodity being examined and the substitute or complement the products (Durand- 48 Morat and Wailes 2003). Though this limit of only examining a specific commodity in a domestic or international relationship is a disadvantage, it is appropriate for the problem examined in this research. 4.2 Conceptual Model A spatial, price equilibrium model is the foundation for this research. Spatial equilibrium models permit separate changes in prices over geographic areas or time periods. Specifically, this type of model examines inter-regional trade and spatial efficiency of regional prices and can range from a two-country, one-commodity model to a complex multi-country, multi-commodity model. Spatial equilibrium models have its weaknesses, including the assumptions of homogeneity of all commodity types or that the production or consumption of all commodity types occur at a given location in the region. In addition, traditional relationships between buyers and sellers are ignored, and the assumption that the decision to export the commodity is based solely on a chosen optimization rule is made. Despite these limitations, it is still believed that the spatial equilibrium model is the best method to measure social welfare in this situation due to separation of commodities along the processing chain and distinctions between the producing and consuming regions, along with the required data set (Tomek and Robinson 1977). This research will be a single—commodity spatial partial equilibrium model”. Myers (2008a) outlines the general model, created by Takayama and Judge (1971), by H From this point forward, the spatial, partial equilibrium model used in this analysis will be referred to only as a spatial equilibrium model with the acknowledgement that this model is also a partial equilibrium model. 49 assuming that there are two or more countries, trading two or more homogenous goods where each country has its own separate market, with the markets being reachable through transportation. The exporting country, i, has a supply equation, qf , that is a factor of in ut rices and ademande uation, 4 , for the im ortin countr , ', is a factor P P Cl q j P 8 Y J of product prices, which are maximized less transportation, (7, of the product (equation 4.1). In addition, the following constraints are included: (1) that the quantity demanded is less than or equal to total imports, x in the importing region, (equation 4.2) and (2) ii . that the quantity exported to all other regions is less than or equal to the quantity supplied in the exporting region (equation 4.3). qd (4.1) max 2 qujo )dqjl— Z qu (4i )dq, —Z 12:161.ij F —I 0 F —10 (4.2) ixl-j 2g? i=1 (4-3) jixij r 0011?,“ -EXPrW (5-5) QDZ’f’f = f 64,. Table 5.1:Parameter Description for Maize Flour Consumption Calculation Variable/Parameter Descriytion EXP Expenditure POP Population 96EXP Percent Expenditure from 1996 PPOP Percent Population GDP Gross Domestic Product C P Retail Price CON Consumption Subscript Description w Weighted r Region p Province f Flour urban Urban rural Rural pc Per capita Note: See Appendix C for a more detailed description. 60 The limitations of this method of creating the quantity consumed of maize flour should be acknowledged. Specifically, the 1997 percentage of urban and rural population are likely not consistent with the percentage of population in rural and urban areas throughout the entire time series. In addition, using an expenditure that includes maize and maize derivatives, almost undoubtedly skews the consumption of maize flour upwards. However, both the urban and rural breakdown and the expenditure percentage of maize and maize derivatives were the best available information. Maize Flour Price ( CPf )M In southern Mozambique, price data for maize flour was reported on a weekly basis from specific markets throughout the region provided that the commodity was available at the market (SIMA, 2008). The maize flour prices from the Maputo City markets were used. Observed prices from multiple markets within Maputo City that occurred on the same date, were combined to create aggregate daily Maputo maize flour prices, which were then aggregated to create monthly prices. Finally, using the Mozambique maize marketing year of April to March, observed prices were combined to create a yearly time series data set of maize flour prices. Price of Rice ( C P r ) The retail price of milled rice is used as the main substitute for maize flour in the southern region. Retail milled rice prices were also reported on a weekly basis from selected markets when the commodity was available (SIMA, 2008). To create '4 Due to the nature of prices, all prices for all elasticity estimations were tested for unit roots. The test results and discussion can be found in Appendix B. 61 consistency of reporting across retail prices, Maputo City market prices were again used for the southern Mozambique region. Rice prices were aggregated in the same manner as the maize flour prices, first daily, then monthly and finally a time series of yearly rice prices for the maize marketing year of April to March was created. Time Trend (T) Economic variables have a tendency to change over time and this increase cannot always be accounted through measurable explanatory variables. Thus, a time trend is included to capture the effect of the increase. In this case, if the coefficient is positive, consumption of maize is increasing over time, if it is negative, consumption of maize flour is decreasing over time. The time trend in this particular demand estimation is hypothesized to account for changes in tastes and preferences and income. 5.2.1.3 Results The estimated demand of maize flour results for southern Mozambique are outlined in Table 5.2. The model indicates that the explanatory variables explain the variation in consumption well with an R2 of 90 percent. In addition, the regression has explanatory power with a probability value of the F-statistic of0 percent. All coefficients on the explanatory variables are also consistent with economic theory. 62 Table 5.2: Southern Mozambique Demand Estimation Dependent Variable = Consumption of Maize Flour, kgs/cgfita ExplanatorLVariable Coefficient T-Stat (P-Value) Significance Price of Maize Flour, Meticals/k gs -0.89 -7.18 (0.0) *** Price of Rice, Meticals/kgs 0.21 1.21 (0.26) Time Trend 0.03 2.12 (0.06) * Constant 5.08 18.47 (0.0) *** F-Stat 30.27 Prob >F 0.000 R-squared 0.9008 Source: Estimated from secondary data source Note: Number of observations 14. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level However, the high R2 (a classic sign of multicollinearity), the manner in which per capita consumption of maize flour was created, and the general nature of the data could be the indication of estimation problems. First, multicollinearity was assumed to be present in the model and a correlation matrix indicated a high correlation between the dependent variable of maize flour consumption and the explanatory variable of maize flour prices, but, correlation between the explanatory variables remained under 50 percent. The presence of multicollinearity results in large standard errors that cause coefficients to be sensitive to small changes within the sample. Multicollinearity can only be corrected through the removal of one of the correlated variables, which can result in estimations not consistent with economic theory and have other major implications for the econometric model. Therefore, acknowledgement of the sensitivity of the estimated parameters needs to be made. In addition, the estimated parameters are sensitive to change due to the small data sample, which are undesirable in econometric estimation. Goldberger (1991) states that micronumerosity, small sample size, creates the same problems as data sets with 63 multicollinearity, including high standard errors relative to the coefficients, therefore causing coefficients to be highly susceptible to small changes within the sample. Goldberger suggests that estimation should be reconsidered if more observations are not available. However, this research acknowledges the sensitivity of the parameters used from the estimation results by conducting sensitivity tests that are reported in Appendix D. Heteroscedasticity, which causes the variance of the error term to be inconsistent across all observations, was tested for through the use of the Breusch-Pagan-Godfrey Test. The null hypothesis that heteroscedasticity was not present in the model was unable to be rejected at a 5 percent level. Due to the presence of heteroscedasticity, the variables in the model are biased and inconsistent (Hamilton, 1994). Autocorrelation, which causes the covariance between different observations to not equal zero, was tested for with the Durbin-Watson Test. The Durbin—Watson test statistic was found to be. in the indeterminate zone between the upper and lower bound statistics, failing to reject the null hypothesis that there is no first order autocorrelation in the model. Due to the autocorrelation in the model, the standard error terms are inconsistent and biased (Hamilton, 1994). To correct for both heteroscedasticity and autocorrelation, White’s Robust Standard Error Estimation was used. Unfortunately, as Hamilton (1994) notes, this correction only adjusts the variances, not the parameter estimates. With a large sample this would not be a concern, as the loss of efficiency in the parameters would be minimal. However, with a small sample size the efficiency loss in the parameters is greater. Again 64 sensitivity analysis on the elasticities and comparison of previously reported elasticities will be used to validate the estimated elasticities. As expected, the own-price elasticity of demand for maize flour is inelastic. However, it is more inelastic than one might expect due to its importance as a consumption staple in Mozambique. Consumption of maize flour decreases by .89 percent when there is a 1 percent increase in maize flour price. However, the shift of relative prices of maize flour as compared to rice (Figure 1.1) and a general shift of preferences in the south, particularly the urban areas, for rice could be the source of the highly sensitive own—price elasticity of maize flour. The cross—price of rice, although not statistically different from zero, does indicate that rice is a substitute for maize flour but little adjustment of consumption of maize flour is made when the price of rice increases. More specifically, with a 1 percent increase in rice prices, consumption of maize flour only increases 0.2 percent. However, the coefficient on the time trend does not correspond with the hypothesized change in preferences, as consumption of maize flour is increasing by 0.03 percent with each passing year. These conflicting results could be due to the manner in which the dependent consumption of maize flour variable was created and the resulting multicollinearity, in addition to the sensitivity of the parameters due to the small sample size. Comparison to ARMA model estimates (Appendix B) and sensitivity tests were conducted on the own- price elasticity of maize flour (Appendix D), and it was determined that -0.87 would be used as the own — price elasticity in the net social monetary welfare maximization equafion. 65 5.2.2 Southern Mozambique - Supply of Maize Grain In the southern region of Mozambique, supply elasticity is assumed to be zero and was not estimated. Estimation was not done due to the southern region being classified a consumption zone, its minor contribution to national production (Table 2.1), and the lack of producer prices reported by SIMA (2008). In addition, poor overall soil quality and high variability in rainfall during the production months create a situation where it is hypothesized that southern producers plant available land regardless of the price. However, Cruz (2006) used an assumed supply elasticity for maize in the south of 0.3 percent. Therefore, sensitivity tests were conducted and it was determined that 0 percent would be used as the supply elasticity in the social welfare model (Appendix D). 5.2.3 Central Mozambique - Demand for Maize Flour 5.2.3.1 Model Equation 5.6 was used to estimate central Mozambique’s demand for maize flour. Demand was estimated as a function of retail maize flour price (CPU),r ), retail rice price (CPcr, ), and a time trend (T). As discussed in section 5.2.1.1, a time trend was used in place of GDP'5 per capita due to the nature in which maize flour consumption was calculated and the uncontrollable characteristics of GDP in Mozambique. (5.6) In Q0}; 2 ,60 + ,6. 1n CPCr, + p2 In Cng + fl3T + ,u, 5 . . . . . . I Again attempts to stabilize the variable through the use of dummy varrables at the pornts of structural change were attempted, however, the variable could not be stabilized due to the lack of ‘normal’ growth data points of the GDP data set. 66 5.2.3.2 Data The central Mozambique time series data set for the demand estimations was for the years of 1992 to 2006. Quantities have been converted in kilograms (kg). Using the same non—food CPI deflator (Donovan, 2008), the prices observed in the center region have been deflated to 2004 new Meticals. Quantity Demanded — Flour ( QDf ) Consumption of maize flour for central Mozambique was not available. Using the same method explained in the data section for the demand of maize flour for southern Mozambique (5.2.1.2), consumption of maize flour was calculated using a static weighted 1996 total expenditure allocated to maize and maize derivatives for the central region that was applied to yearly expenditure GDP, which was then divided by the price of maize flour in the central region (equation 5.4 and 5.5). The limitations to the method of calculation still remain, including the use of static expenditure percentages (1996) and static urban and rural breakdowns (1997), due to data availability this is the best method that could found. Maize Flour Price ( CPf ) As in the southern region, prices were reported on a weekly basis when the commodity was available in selected markets throughout central Mozambique (SIMA, 2008). Beira was chosen as the best representative market for retail maize flour prices in central Mozambique and thus prices from the Beira markets were aggregated first by daily observations from multiple markets with the Beira classification then by month, and 67 finally by year to create an average yearly price for the Mozambique maize marketing year of April to March. Price ofRice (CPr) As in southern Mozambique, rice is the main substitute for maize flour. Retail milled rice prices from selected markets throughout Beira were selected as the main market for price analysis to eliminate inconsistencies due to location (SIMA, 2008). Prices were aggregated in the same manner as maize flour to create a yearly retail rice price. Time Trend (T) To create consistency throughout the country of Mozambique, a time trend was used in central Mozambique, as in southern Mozambique, to capture the increase in consumption of maize flour over time. Specifically, the time trend was used to capture the changes in preferences and tastes in addition to the income effect that was not captured due to omitting GDP per capita as an explanatory variable 5.2.3.3 Results . Table 5.3 shows the results of the estimated demand for consumption of maize flour in central Mozambique. The model had an R2 of 66 percent and was overall significant in modeling the consumption of maize flour at a 1 percent significance level. 68 Table 5.3: Central Mozambique Demand Estimation Dependent Variable = Consumption of Maize Flour, kg/capita Explanatory Variable Coefficient T-Stat (P-Value) Significance Price of Maize Flour, Meticals/kg -0.33 —0.88 (0.40) Price of Rice, Meticals/ kg 0.21 0.36 (0.72) Time Trend -0.024 -0.67 (0.52) Constant 4.578 4.48 (0.0) *** F-Stat 7.11 Prob >F 0.0045 R-sauared 0.6596 Source: Estimated from secondary data source Note: Number of observations 15. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level As with the demand estimation in southern Mozambique, multicollinearity was suspected to be present within the explanatory variables due to a high R2 and insignificant t-statistics on the explanatory variables. A correlation matrix indicated high correlation (77 percent) between the time trend variable and the retail price of rice. The time trend and dependent consumption variable were also correlated at 75 percent. The presence of multicollinearity combined with a small sample size, which results in similar negative effects as multicollinearity, again equates to estimated coefficients that are highly sensitive to change within the sample. The equation was tested for heteroscedasticity using the Breusch-Pagan-Godfrey Test. Again, the results failed to reject the null hypothesis that the model was homoscedastic. The Durbin Watson Test was used to test for first order autocorrelation. It was found that at a 95 percent confidence level, the null hypothesis was accepted and there was not autocorrelation in the model. The model was corrected for heteroscedasticity through the Whites Robust Standard Error correction, however, again 69 only standard errors were corrected not parameter estimates, and thus the parameter estimates have a loss of efficiency due to the small sample size. Although most variables were consistent with economic theory, none of the explanatory variables were significantly different from zero. A time trend was used in place of GDP for reasons previously explained in section 5.2.1 . 1 , however, the model now indicates that consumption of maize flour is decreasing overtime, which is not indicative of current consumption trends in central Mozambique. Additionally, when modeled with a time trend, the own—price and cross price elasticities were not significantly different from zero. Although, own-price elasticity of maize flour was relatively inelastic, which one would hypothesize to be correct, a decrease in consumption of 0.33 percent due to a 1 percent change in price may be low when compared to other estimates of demand elasticities. Due to the sensitivity of the parameters from the multicollinearity, small sample size, and heteroscedasticity, sensitivity tests were conducted on the own-price elasticity of maize flour for central Mozambique (Appendix D) and comparison with ARMA elasticity estimates (Appendix B), which resulted in -0.33 own—price elasticity of demand for maize flour to be used in the net social monetary welfare maximization equation. 5.2.4 Central Mozambique — Supply of Maize Grain 5.2.4.1 Model Supply of maize grain in central Mozambique was estimated as a function of a one period lagged producer price (PP/El ), rainfall (TTRC, ), and a time trend (T) (equation 5.7). Futures prices were not available for Mozambique, so it was determined 70 that the price of grain lagged one year would be the best indicator for expected prices for farmers, which is consistent with the habits of Mozambique producers. Rainfall data were used as the main input variable since fertilizer and irrigation are rarely used. Additionally, a time trend was added to account for changes in production of maize grain in the region. (5.7) In st, = 00 + ,6. 1n P1351 + ,62 In TTRC, + an +11, 5.2.4.2 Data A time series data set from 1993 to 2006 was used to estimate the maize grain supply for central Mozambique. As with all other time series data, quantities have been converted in kilograms (kg) and prices have been deflated to 2004 Meticals using the non-food CPI deflator (Donovan, 2008). Quantity Supplied — Grain ( Q5 3 ) Regional total maize production was not available for the entire time series and thus had to be calculated for some years. Total production for the years to 1993 to 1999 was obtained from the FAO. INE provided total production from 2000 to 2002 and the FEWSNET Food Balance Sheets provided regional production for the remainder of the time series. To create regional production numbers for data from 1992 to 2002, the 2006/2007 regional production breakdown (37 percent) was applied to all yearly total production. Although actual regional production numbers would be ideal, the 2006/2007 production year appears to be a consistent with regional production patterns when compared to the 71 regional production patterns from 2003 to 2005 provided by the FEWNET Food Balance Sheets, in addition to production patterns found in the 2002 IAF household survey (2002). Logged Producer Price ( Ppril ) The producer prices reported for the province of Manica were used in place of Chimoio, the largest production region in the center, due to lack of data availability in Chimoio (SIMA, 2008). Only prices reported from May through October were used, since that is the typical period producers sell maize. Prices were first aggregated by all prices reported on a specific date, then within a month and finally into the maize marketing year of April to March and then lagged one year. Total Rainfall (TTR) Total rainfall was reported on a ten days basis throughout the production year (September to April) for different locations throughout Mozambique (FEWNET, 2008). The central region station used was Sussundenga in Manica Province. Rainfall was summed throughout the production year to create a total rainfall variable. Rainfall data were not available for 1992 through 1994 and were thus calculated as the average rainfall throughout the years available (1995—2007) minus the standard deviation of that time period. Time Trend (T) Economic time series have a tendency to change over time and this increase cannot always be accounted through measurable explanatory variables. In this particular model, production steadily increases over time. The increase is hypothesized to be due to the increased political stability, but, other factors are also likely to contribute to production growth. Therefore a time trend was included to capture all the effects causing a change in maize production in central Mozambique. 5.2.4.3 Results Table 5.4 shows the regression results for the supply of maize grain for central Mozambique. The R2 is 81 percent and the F-value shows that the model is significant at a 1 percent level, however, many of the explanatory variables do not have statistically significant t-statistics, indicating possible multicollinearity in the model. The correlation matrix reveals that the time trend and production are correlated around 80 percent and again the estimation is done with a small sample size, therefore causing the parameters to be highly sensitive to changes in the data sample. The estimated equation was tested for both heteroscedasticity and autocorrelation. Results from the Breusch-Pagan-Godfrey Test failed to reject the null hypothesis that the model was homoscedastic at a 5 percent significance level. The Durbin Watson Test results also failed to reject the null hypothesis that there was no autocorrelation in the model. Therefore White's Robust Standard Errors method was used to correct for both heteroscedasticity and autocorrelation in the model. As with the other models, this correction method only corrects the standard errors, however in large samples the 73 coefficients are asymptotically normal due to the small loss in efficiency. Since this model was used on a small sample, the loss of efficiency on the parameter coefficients cannot assume to be to be asymptotical and thus the sensitivity of the parameters must be considered when interpreting the results. Table 5.4: Central Mozambique Supply Estimation Dependent Variable = Production of Maize Grain, kgs Explanatory Variable Coefficient T-Stat (P-Value) Significance Lagged Producer Price, Meticals/kg 0.17 0.88 (0.40) Total Rainfall, mm 0.46 1.43 (0.18) Time Trend 0.08 5.46 (0.0) *** Constant 15.99 6.63 (0.0) *** F-Stat 25.68 Prob >F 0.001 R-sguared 0.8105 Source: Estimated from secondary data sources Note: Number of observations 15. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level The estimated coefficients are consistent with economic theory, however, most are not statistically different from zero, especially the desired price elasticity of supply. Although not significant, the own-price supply elasticity indicates that as the price of maize grain increases by 1 percent, quantity of maize supplied increases by 0.17 percent. It is important to note that due to production systems throughout Mozambique, with low inputs and rain fed production, a low elasticity is not unreasonable as actual hectares planted are not always harvested. Nevertheless, the supply elasticity of maize does seems low when compared to a study by Cruz (2006) who assumed a supply elasticity in central Mozambique of 0.45 percent. Therefore, comparison to ARMA elasticity estimates (Appendix B) and sensitivity tests were conducted on this elasticity (Appendix D) and it 74 was determined that 0.17 supply elasticity of maize in central Mozambique would be used to estimate the excess supply elasticity of central Mozambique for the net social monetary welfare maximization equation. 5.2.5 South Africa - Demand for Maize Flour 5.2.5.1 Model Equation 5.8 shows the econometric model used to estimate the demand of maize flour in South Africa. Based on economic theory, the demand of maize flour in South Africa is being estimated as a function of the own price of maize flour ( CPaf, ) and the per capita income for South Africa (GDP ’7" ). (II (5.8) In Q03, 2 so + p, In GDP ”" + p, In CPU], + ,u, (If 5.2.5.2 Data For the South African time series analysis of demand, all observed data are from the years 2000 to 2006. All quantities have been converted to kilograms and all prices have been deflated to 2004 and converted to Meticals. Prices in South African Rand were deflated using the urban CPI deflator (Statistics South Africa, 2008) and were converted using exchange rates at the Interbank rate (OANDA, 2008). Prices reported in US. Dollars were deflated using an US. urban CPI deflator (US. Department of Labor Statistics, 2008) and converted to Meticals through an US. dollar — Mozambique-Metical exchange rate (OANDA, 2008). 75 Quantity Demanded - Flour ( QDf ) Data on consumption of maize flour in South Africa were not available. However, by using the food balance equation (equation 5.9), where the sum of production, imports, and beginning storage, equals the sum of consumption, exports, and ending storage, the disappearance method was used to determine quantity of maize flour consumed. Total maize milled per year (The National Maize Millers of South Africa, 2008), imports of maize flour and exports of maize flour per year (United Nations, 2008), and the assumption that beginning and ending storage was zero, a time series data set on consumption of maize flour in South Africa was created. Total maize flour consumption in South Africa was then divided by total population to achieve maize flour per capita (World Bank, 2008). (5.9) PROD + IMPORTS + STOR, = CONSUMP + EXPORTS + STOR, Maize Flour Price (CPf ) The consumer price of maize flour for South Africa was evenly aggregated for the twelve month maize marketing year in South Africa, May to April (SAGIS, 2008). GDP per Capita (GDPP‘) GDP per capita, which was reported in current local currency was deflated and converted to 04 Meticals and then divided by total population to obtain a data set of GDP per capita for South Africa (World Bank, 2008). 76 5.2.5.3 Results The results of the econometric estimation of demand for maize flour in South Africa are shown in Table 5.5. The model had an R2 of 80 percent, indicating the variation in consumption was explained well by the variables in the model, and a high F- value, which implies that the regression had explanatory power. In addition, although not all significantly different from zero, the estimate coefficients for the explanatory variables were consistent with economic theory. Table 5.5: South African Demand Estimation Dependent Variable = Consumption of maize flour, kg/capita Explanatory Variable Coefficient T-Stat (P-Value) Significance Price of Maize Flour, Meticals /kg —0.49 -1.35 (0.25) GDP per capita, Meticals/capita -l .88 -3.73 (0.02) ** Constant 26.95 4.48 (0.01) *** F-Stat 7.19 Prob >F 0.0474 R-sguared 0.8047 Source: Estimated from secondary data sources Note: Number of observations 7. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level Originally wheat flour, the major substitute for maize flour was included in the model to capture the cross price effects. However, under further investigation, the variable was insignificant (probability of t-statistic 0.988) and highly correlated (86 percent) with GDP per capita. Therefore, it was decided to remove the variable to help decrease the multicollinearity effects, specifically parameter sensitivity, in a data set that was already more susceptible to parameter sensitivity due to the small sample size. Multicollinearity does still exist in the model as indicated in the correlation matrix between GDP and consumption of maize flour. 77 As with the other demand estimations, heteroscedasticity and autocorrelation were tested. The Breusch-Pagan-Godfrey Test failed to reject the null hypothesis that the model was homoscedastic at a 5 percent significance level. The null hypothesis was accepted at a 5 percent significance level in the Durbin Watson first-order autocorrelation test, indicating that the model was not autocorrelated. However, Whites Roust Standard Errors correction still had to be used to correct the standard errors due to heteroscedasticity. Due to the small sample size, the same loss of efficiency of the parameters discussed in previous estimation results still is relevant. Although not the focus of this study, the model indicates that consumption of maize flour is expected to decrease by 1.88 percent as income increases by 1 percent in South Africa. This result is consistent with Bennet’s Law, which states as a country’s income increases, the proportion of income spent on staples decreases, once they reach a minimum level of income needed to meet everyday calorie needs. The own price elasticity, which indicates that consumption of maize flour will decrease by 0.5 percent as the price of maize flour increases by 1 percent, is consistent with economic theory. However, the t-statistic on this parameter indicates that it is not statistically different from zero. It is believed that the own price elasticity is valid and would be statistically significant if additional observations were available due to its comparability to the own— price maize flour elasticity estimated by Mabiso and Weatherspoon (2008). The authors use the same data set, including time frame and retail maize flour prices, except their analysis of demand was on a monthly basis, thus increasing their sample size to over 80 observations. The authors estimate comparable own-price demand elasticities for maize 78 flour using a seemingly unrelated regression and a double log single equation model and find in both models own-price demand elasticity for maize flour of -0.42 that is statistically significant at a 5 percent level. 5.2.6 South Africa — Supply of Maize Grain 5.2.6.1 Model South Africa’s econometric model of supply of maize grain was based on the economic theory of supply estimation discussed in section 5.1, with consideration taken for data availability and the production system in South Africa. Supply is modeled as a function of the futures price of maize grain (FPPtg ), fertilizer prices ( F ERT, ), and a time trend (T) (equation 5.10). Futures prices were chosen to apply Ferris’ (1998) belief that farm supply is based on expected prices and South Africa has an active futures market. Fertilizer was determined to be the main input used in South Africa and thus was chosen as the input price. In addition, rainfall is known to be an important variable in production, as most agriculture production is rain fed. However, after repeated attempts, rainfall data for South Africa were unavailable. (5.10) In Q5 ’4 = no + ,6, ln FPPtg + ,6, ln FERT, + T + ,u, at A crop that is a competitor with maize grain for land and inputs was indeterminate and thus not included. Table 5.6 illustrates that the main crop produced in South Africa are sugarcane, however, sugarcane and maize grain are not similar commodities and thus are not competitors for land or inputs. Intuitively yellow maize would be a major competitor for inputs and land. However, white and yellow maize future prices are highly correlated (96.1 percent) thus if it was included it would decrease 79 the precision of the estimated parameters, increase the confidence intervals, and increase the standard errors in a small data set. Wheat was also hypothesized to be a major competitor for inputs with maize, however, wheat production was minor compared to maize production (Table 5.6). Ideally, all of these commodities would be included as competitors for inputs and land, but, due to the number of observations in the data set, this variable was omitted. Table 5.6: Top 5 Maize Producing Provinces in South Africa, 2002 Eastern Cape Free State North West Mpumalanga KwaZulu- Natal ( 1000 MT) Maize 1,460 5,227 5,057 1,31 1 402 White 536 1,918 1,856 481 148 Yellow 924 3,309 3,201 830 254 Wheat 2 1,540 34 0 15 Data Source: Lehohla (2002) 5.2.6.2 Data The time series observed data for the analysis of supply for South Africa range from 1997 to 2006. As with the data used in the demand estimation for South Africa, all quantities have been converted to kilograms (kg) and prices have been deflated using the urban CPI deflator (Statistics of South Africa, 2008). Prices have also been converted to Meticals using the Interbank conversation rate (OANDA, 2008). Quantity Supplied — Grain ( QS g ) In South Africa, only white maize production was desired since yellow maize is not used for human consumption in Mozambique. To achieve this, total production of 80 maize was multiplied by the percent of white maize in total maize production (South African Department of Agriculture, 2008). Two percentages were used and applied to the time series data set. The first, 0.453 percent, was applied to 1997 through 1999, which was created based on production breakdowns of white and yellow maize provided during that time period (FAO, 1997). Using the 2005 percentage of production data, in addition, to a statement that white maize has surpassed yellow maize production, total maize production quantity for the remainder of the time series was multiplied by 0.633 to obtain total white maize production in South Africa (South African Department of Transport, 2005). Futures Price of Maize Grain ( F PPg ) Future prices of maize grain were used as the expected price in the supply estimation equation for South Africa (South African Federal Exchange Board, 2008). Prices were based on contracts that started on November 15‘h with a delivery date of July for the following year. Fertilizer ( F ERT) A fertilizer price index for South Africa was used in place of yearly fertilizer prices, which were not available (South African Department of Agriculture, 2008). Time Trend (T) Economic time series have a tendency to change over time and this increase cannot always be accounted through measurable explanatory variables. Production of 81 maize grain in South Africa has been increasing over the past years, therefore a time trend is included to capture all the effects causing a change in maize production in South Africa that is not due to prices of maize grain or fertilizer. 5.2.6.3 Results Table 5.7 illustrates the econometric results for the estimation for the supply of maize grain in South Africa. The model has an R2 of 17 percent and the overall model is not significant at a 10 percent level, according to the F-value. These results indicate that the model used to estimate supply of maize does not fully explain the variation in supply and does not have significant explanatory power for estimating supply of maize grain in South Africa. Table 5.7: South African Supply Estimation Dependent Variable = Production of Maize Grain, kgs Explanatory Variable Coefficient T-Stat (P-Value) Significance Future Price, Meticals/kg 0.13 0.46 (0.66) Fertilizer Price Index 0.38 0.44 (0.67) Time Trend -0.01 -0.14 (0.89) Constant 20.80 7.49 (0.0) *** F-Stat 0.58 Prob >F 0.6457 R-sgared 0.1 668 Source: Estimated from secondary data sources Note: Number of observations 1 1. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level Multicollinearity between the explanatory variables was not found in this model of supply. However, the sample size is small in this estimation, like all others, causing the estimation results to be treated as if multicollinearity was in the model, including 82 acknowledgment of high standard errors relative to the coefficients therefore causing coefficients to be highly susceptible to small changes within the sample. As with all other estimation models in this study, the equation was tested for heteroscedasticity through the use of the Breusch-Pagan-Godfrey Test. The results failed to reject the null hypothesis and thus heteroscedasticity is likely present in the model. Autocorrelation was also tested for through the Durbin Watson test. The results accepted the null hypothesis and no autocorrelation was found in the model of supply of maize grain in South Africa. To correct for the heteroscedasticity, White’s Robust Standard Error Estimation was used. This correction only adjusts the variances not the parameter estimates, with a large sample this would not be a concern as the loss of efficiency in the parameters would be minimal. However, with a small sample size, the efficiency loss in the parameters is greater. Although not statistically different from zero, the coefficients do correspond with economic theory, including the price elasticity of supply, which indicates that the quantity supplied increases by 0.13 percent as the price of maize grain increase by 1 percent. However, that supply elasticity is low. Cruz (2006), in his analysis of storage in Mozambique, assumed a supply elasticity for maize grain in South Africa of 0.65. Based on the sensitivity test conducted (Appendix D), a supply elasticity of 0.13 will be used for South Africa. 5.3 Conclusion This chapter presented the theoretical background for econometric methods used in the supply and demand estimation. In addition, the chapter has provided an in-depth discussion on the conceptual model chosen, the data used to estimate the models, and the results of the econometric estimations. The estimated supply and demand elasticities will be used to make the net social monetary welfare maximization model operational by using the estimated domestic elasticities to calculate the excess supply and demand elasticities. The use of the domestic elasticities is explained in the following chapter that begins with a detailed discussion of the conceptual social welfare model and data used. 84 6. SIMULATION 6.1 Operational Model Chapter 4 outlined the theoretical basis for using a spatial equilibrium model to simulate welfare changes due to a change in tariff policy. The Durand-Morat and Wailes (2003) RICEFLOW model (equation 4.4) that simulates tariff policy changes in world rice trade was used as a template to create an empirical model (equation 6.6) that maximizes net social monetary welfare of Mozambique, by allowing all prices to adjust, subject to material balances and price constraints. The net social monetary welfare equation uses the intercepts and slopes of the excess supply and demand curves to maximize the social welfare as a sum of total social revenue for the consumption of imports in southern Mozambique less the sum of the total production exported in central Mozambique and South Africa less transport costs of the commodities from the exporting to the importing region. Thus, the problem is modeled as: Maximize (a! oClFSf —%fl,f "(€1st )2)+(a_§ ~CIFSg —% 6f -(c1F,-" )2] (6.1) —y~[’ F,oe(+;6,,f (Foejfj—[y§-FOB§+l6§.(F03§l)) —(y,f Foe,f+— 26f (TOBfl-J—[Jq Toe,?+— :26}? (magi) —(TC,, , FL L,,,)— (ch FL§,)— (TCZ, -FL,f.::j-(TC,§ ~FL§S) Subject to: Price Restrictions: (6. 2) CIFf — Foef < ch+ (I — (IA (6.3) CIFf- FOB,.f 3 Eamon "N4 2:3. Ea: 65%: £25 32852 3 Bo v. 50:22 9:922 <25 3:280:22 .3 x lb $33.23» beueefimam <25 :3: .mZH 3.982% k 165% . \mb 0 «QQ cosflso—au 2.5m. 8.er \ . 653 we New 633.53 aeeezemeQ :e_~a_=o_a0>:eEEeU 8.:8m .ED czar—5» use-rm 332 3.. 23:39 1 a: 35332 Eofiaem :8: Eamon «33...; :4 ~35. ZO~H<2~Hmm >H—U—Pmdfim QZ§WQ>AmmDm hO mums—magmas» n< Nun—2954‘ 123 .5 5.2 8w «35$ 02.9 2 k b 3.2 nm wmozmm $82: 2 :5 3&2832 we $33.23» bSuEdem b 3.? 02% Roche $08.8 2 Eu 3&8).wa mESLS» :83:in E:E_xa2 85552 55359 95235 532 23.9330 @332; 32m onus—z .Sa 152:3— I a: 55334 .5550 ”85:55 9»: E82— 64 «Sub 124 m5 MAB $33.25» bactsmxm ESE 26m <25 3&3232 we 332 26m <25 MVESEE we <25 4“: .mZ_ ”mouScm x g :5 gQU ROE . u cog—5030 2&5. 2E8 \ 553 My. mam 3%.:3» 332855 :cufiafinvhcofificu 83cm E5 @3525» :53 3?: .Su “Eng—on I a: 5:5ch .95on unwantomon czar—5’ "mé «Bah . b NE «2 m: $33 3%? 3 3.28%: «9 ii mag 5v mnwmamm «.30.wa 3 8:5 MEL. 9633.53 baucfimxm 9% n _ w :2 m2 SV 3 m a? 8:36 mEShS EmnzwmwQ Esp—:32 SEE—:2 55359 6.3—=55 E32 msczwiamnc «33...; £30 332 .3 _ am I a: 5:532 .9550 32335 «>3 €369 63.x oEaH 5 m 2352 @3530 608-33 008-32 session 23:5: I 008-32 2352 owfio>< ”32 .N02 30:3th «Baa—2V c236 wconcammsm kamBmE E82222 at 59900 9 $2 .803 8:32 <25 33,4352 3 ”moi 833.23» Ecgceaiam 385m 85:5 voom HMZmBmE ”88-83 8m m7: HNo.88 cocoscoi O? 33:02 we .i RQD a E? 9 :2"; m6._8_62 3 26 833.23» b23253fi 2232 8:5 I memonflxm 23 @0838 I MOHm 2232 BE: .. flags: cosfluommxx E252 3:23 Z I DOME ”woohsom 3 cat + Exomxm + «3538 u is: + Execs: + Q9: gossaoau 85.2 wx \ QQ 3‘35: Ewwzmamfi :czsafiuxucoE—cau 83cm E5 933:; , ESE 332 .8.“ 1:959 I 32.2 5.5m :5: 2369 933...; #6. via. .5 2.9: 5.8 853.8 03% 2 K L i N: 85090 23 2 EC Engage €33.53 Eeuuasgmw . So 2% wwm m3 9% E x a; 85m: 3%.:3» 2533me E55532 Esp—:52 553:5 ESE—3m :32 2.83:330 czar—5w EEG 3?: .3 _ mam I 82.3. 5.5m "wuamsflm 959.58: ”Eé «Earn. $83 5?: 82m SEEK 89m w< mo n< £8522 3 hum: b3. 95 .52 592.952 ”Bab :Sm 68:50 anzm 9:28:02 3 unmmk 833.23» bScagmam 388 233 E083 I Atommcfih <9 mmod * coa SEE :39 388 38m w< mo n< oooméoom BEE BER :5ch I €030 "a4 ~35. 127 APPENDIX B: UNIT ROOTS DISCUSSION Due to the descriptive properties of prices”, prices tend to have stochastic trend (unit roots) causing the covariance to not be stationary, which violates basic econometric assumptions. Therefore, the data must be tested for unit roots. All prices and GDP data sets for all regions were tested for unit roots using the Dickey Fuller Test (Hamilton, 1994). All data sets were tested under two different conditions, first with a constant, and second with a constant and time trend (Table B. l ). In southern Mozambique, the Dickey Fuller test results failed to reject the null hypothesis that unit roots were in the data set with two lags with a Z(t) test statistic of - 2.423, with an approximated p-value of 13 percent. However, additional lags did not result in a rejection of the null hypothesis either and actually moved the estimated p—value further from the significance levels needed to reject the null hypothesis. The data set were tested out to four lags, after which the number of observations in the data set did not allow for further testing. When the unit roots were tested for with the additional of a time trend, the results were consistent with the results found when only a constant was in the model. A two-lagged Dickey Fuller test provided the lowest estimated p-value of 29 percent with a Z(t) statistic of -2.570, while additional lags increased the p-value. The data were again lagged to the fourth degree, after which the number of observations did not allow for further testing. The price of rice was also tested for unit roots. The Dickey Fuller test results rejected the null hypothesis that unit roots were present in the data set at a 1 percent significance level with a Z(t) statistic of -6.809 with three lags when the 20 Descriptive propenies of prices include: high variability, co—movement, seasonality, time- varying volatility, unpredictability, and trends (Myers 2008b). 128 model was ran with a constant only. When the time trend was added and the test results indicated that the null hypothesis could not be rejected at a 10 percent level with up to four lags, after which the number of observations did not allow for further testing. As in southern Mozambique, the price of maize flour and price of rice, in addition to the price of maize grain in central Mozambique were tested for unit roots. The tests results for the maize fiour when only a constant was added, the null hypothesis was rejected that unit roots were in the model at a 5 percent significance level with a Z(t) statistic of -3.640 with three lags. When a time trend was added, the results were similar, with rejection of the null hypothesis at a 1 percent significance level with a Z(t) statistic of -3.936 with three lags. When the price of rice for central Mozambique was tested for unit roots using a constant only, the Dickey Fuller test found that the null hypothesis could be rejected at a 10 percent significance level with a Z(t) statistic of -3.22 when three lags were used. The addition of a time trend to the test found that the null hypothesis that unit roots were in the model could be rejected at a 1 percent significance level with a Z(t) statistic of -4. l 85 with one lag. The price of maize grain, used in the supply elasticity estimation for central Mozambique was also tested for unit roots. The Dickey Fuller test results for this data set indicated that with a constant only, the null hypothesis was rejected at a 1 percent significance level with a Z(t) statistic of -4.894 with one lag. When the test was ran with a time trend the results mirrored the previous results, in that the null hypothesis was rejected at a 5 percent level with a Z(t) statistic of - 4.26] with one lag. I29 Table B.l: Dickey Fuller Test Results, (Continued) Z(t) Stat Significance Level 1% 5% 10% Southern Mozambique Price of Maize Flour Constant: 4.423 _3.75 _3_() —2.63 2-lags TIme Trend: -2570 -438 -3.6 -324 2-lags Constant: 0.959 _3.75 _3_() -2,63 4-lags TIme Trend: 492 -4.38 -3.6 -324 4-lags Price of Rice Constant: -6.809 _3.75 -31) -2.63 3-lags TIme Trend: _ l .433 -438 -35 -324 4—lags Central Mozambique Price of Maize Flour Constant: -3.640 _3_75 -30 —2,63 3-lags TIme Trend: -3.936 -4.38 -3.6 -324 3-lags Price of Rice Constant: 3.243 _3.75 -3.() —2,63 3-lags TIme Trend: -4.185 -4.38 -3.6 -324 l-lag Price of Maize Grain Constant: -4.894 3.75 -30 -2,63 l-lag TIme Trill]; .426] -438 -3.6 -3.24 Source: Estimated by Author using secondary data sources. I30 Table B.l: Continued Z(t) Stat Significance Level 1% 5% 10% South Africa GDP Constant: l .413 3.75 _3.0 -263 l-lag TIme Trend: -O.369 -4.38 -3.6 -324 l-lag Price of Maize Flour Constant: _ 1.883 3.75 _3.() -263 l-lag TIme Trend: 3.155 -4.38 -35 -324 l-lag Price of Maize Grain Constant: 4.09] -375 _3.0 -263 l-lag Constant: _ l .399 _3.75 -30 -263 3-lags TIme Trend: -8.588 -4.38 -3.6 -324 3-lags Source: Estimated by Author using secondary data sources The South African GDP, maize flour price, and futures maize grain prices were all tested for unit roots as well. The Dickey Fuller test results for both a constant only and the addition of a time trend failed to reject the null hypothesis at a 10 percent level with one lag. Dickey Fuller tests with additional lags was not an option, as the number of observations did not allow for additional lags. When the price of maize flour was tested for unit roots with only a constant the test results indicated that with one lag the null hypothesis was rejected at a 1 percent level that the data contained unit roots. However, ’when a time trend was added, the results failed to reject the null hypothesis at a l0 percent significance level. As with the GDP data set, additional tests with added lags could not be done due to the number of observations in the data set. Finally, the South 13] African futures maize grain prices were tested for unit roots. The Dickey Fuller test indicated that with a constant only the results failed to reject the null hypothesis that unit roots were in the data with three lags. As with previous data sets, additional tests with added lags were not possible due to the number of observations in the data set. However, when a time trend was added, the test results indicated that the null hypothesis could be rejected at a 1 percent level with a Z(t) statistic of -8.588 when three lags were used. Based on the results of the Dickey Fuller Test, which indicated that generally unit roots were present in at least one lag of the data, ARMA models, which were lagged based on the Dickey Fuller Test results, were ran to estimate the elasticities. Table B.2 shows the estimation results for the demand for maize flour in southern Mozambique using a three-lag ARMA model. The maize flour price elasticity using this estimation method is -l.02, which is slightly more inelastic than the OLS estimated elasticity of - 0.89, however both are statistically significant at the 1 percent level. The remaining variables estimated in the model are all significant at the 5 percent level or less and are very similar (with in 0.1) to the OLS estimated values. Eble B.2: Southern Mozamflue Demand EstimjaLtion, ARMA NIodel Dependent Variable = Consumption of Maize Flour, kgs/capita Explanatory Variable Coefficient Z-Stat (P- Value) Significance Price of Maize Flour, Meticals/kgs -l.02 —9.27 (0.0) *** Price of Rice, Meticals/kgs 0.32 2.33 (0.20) ** Time Trend 0.03 4.29 (0.00) *** Constant 5.12 18.1] (0.0) *** Chiz-Stat 389. I 5 Prob > Chi2 0.000 Source: Estimated from secondary data source Note: Number of observations l4. All estimates in log-log form. * Significance at the It) percent level ** Significance at the 5 percent level *** Significance at the l percent level ._. Using a three-lag ARMA model, the demand for maize flour in central Mozambique was estimated (Table B3). The estimation results indicate that the estimated price elasticity of maize flour is -0.48, which is slightly more elastic (in absolute terms) than the OLS estimated price elasticity of maize flour, -0.33. The ARMA estimated elasticity of maize flour, like the OLS estimation, is not statistically different from zero. The additional variables estimated in the demand function remain similar to the OLS estimated values (within 0.2), however no additional parameters are statistically different from zero, except the constant which is consistent with the OLS estimation. Table B.3: Central Mozambique Demand Estimation, ARMA Model Dependent Variable = Consumption of Maize Flour, kg/capita Explanatory Variable Coefficient Z-Stat (P-Value) Significance Price of Maize Flour, Meticals/kg -0.48 -l .20 (0.23) Price of Rice, Meticals/ kg 0.46 0.66 (0.51) Time Trend -0.01 -0.39 (0.70) Constant 4.24 3.57 (0.0) *** ChiZ-Stat 172.30 Prob > Chi2 0.000 _- Source: Estimated from secondary data source Note: Number of observations 15. All estimates in log-log form. * Significance at the ID percent level ** Significance at the 5 percent level *** Significance at the l percent level The supply of maize grain in central Mozambique was estimated, based on the results from the Dickey Fuller test, with a one-lag ARMA model (Table B4). The ARMA estimation shows the price elasticity of maize flour to be 0.22, which is slightly more elastic than the OLS estimate of maize grain elasticity, however, neither are statistically different from zero. The ARMA estimations of the additional parameters are within 0.1 of the OLS estimates and are all statistically significant at a 5 percent level or lower. l33 Table BA: Central Mozambique Supply Estimationz ARMA Model Dependent Variable = Production of Maize Grain, kgs Explanatory Variable Coefficient Z-Stat (P- Significance Value) Lagged Producer Price, Meticals/kg 0.22 0.89 (0.37) Total Rainfall, mm 0.55 1.96 (0.05) ** Time Trend 0.09 4.94 (0.0) *** Constant 15.27 7.35 (0.0) *** Chiz-Stat 32.56 Prob > Chi2 0.00 Source: Estimated from secondary data sources Note: Number of observations 15. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level Even though there were not enough degrees of freedom to reject the null hypothesis that unit roots were not in the data, thus indicating the number of lags to use for the ARMA model the failure to reject the null at a 10 percent level with one lag indicates the presence of unit roots. Therefore the South African demand for maize flour was estimated using an ARMA model with the most lags possible for this data set, one. Table B.5 shows the estimation results for this ARMA model and indicates that the price elasticity of maize flour is -0.55, but is not statistically different from zero. The ARMA estimated elasticity of maize flour is, like all ARMA estimations thus far, slightly more elastic than the OLS estimated elasticity of -0.49. The GDP variable is slightly less price responsive with the ARMA model than the OLS estimation, but is no longer statistically different from zero. 134 Table B.5: Southern African Demand Estimation, ARMA Model Degndent Variable = Consumption of maize flour, kg/capita Explanatory Variable Coefficient Z-Stat (P-Value) Significance Price of Maize Flour, Meticals /kg -0.55 -0.04 (0.67) GDP per capita, Meticals/capita -2.08 -2.85 (0.00) *** Constant 29.45 2.59 (0.01) *** Chiz-Stat 27.36 Prob > Chi2 0.00 _ Source: Estimated from secondary data sources Note: Number of observations 7. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level Finally, the supply of maize grain in South Africa was estimated using a 3-lag ARMA model (Table 8.6). The estimated supply elasticity of maize flour using the ARMA model was 0.77, but not statistically different from zero. This estimation of supply elasticity of maize grain is considerably more elastic than the OLS estimation of 0.1 l, which was not statically different from zero. The additional parameters, specifically fertilizer are also considerably different, however, besides the constant, just like the OLS estimations, none of the additional ARMA estimates are statistically different from zero. Table B.6: South African Supply Estimation, ARMA Model Dependent Variable = Production of Maize Grain, kgs Explanatory Variable Coefficient Z—Stat (P-Value) Significance Future Price, Meticals/kg 0.78 0.97 (0.33) Fertilizer Price Index 0.99 0.86 (0.39) Time Trend -0.08 -0.62 (0.54) Constant 19.07 6.08 (0.00) *** Chiz-Stat 478.67 Prob > Chi2 0.000 Source: Estimated from secondary data sources Note: Number of observations 1 1. All estimates in log-log form. * Significance at the 10 percent level ** Significance at the 5 percent level *** Significance at the 1 percent level 135 Although the Dickey Fuller test results indicated that unit roots were in the all the price data, at least through the first lag, and the estimated ARMA model elasticities seem reasonable, the OLS estimated elasticities are used in this research. This decision was made due to the critiques of the Dickey Fuller test and the characteristics of this data set. First, researchers main critique of the unit root tests, specifically the Dickey Fuller test, is its low power. In other words, the test tends to accept the null hypothesis that unit roots are present in the data, even when that is not the case. Critiques say the low power of the test is due not only to the size of the data set, but the time span of the data set. If a data set is spanned over 30 years as compared to 10 years, the data offers more explanatory power (Gujarati, 2003). In addition to the critiques that the test lacks power, the characteristics of this data set, specifically the low number of observations leads to decreased explanatory power of the data when unit roots are corrected for through a lagged ARMA model. Therefore, due to the short time span of the data and the overall loss of degrees of freedom with the use of the ARMA model to correct for the unit roots, combined with the overall observation that the estimated elasticities of the ARMA model were consistent with the OLS estimated elasticities, the OLS estimated elasticities were used. Although, the ARMA elasticities due help validate the accuracy of the OLS estimated elasticities. In addition, the ARMA elasticities were used as an alternative elasticity Option in the sensitivity tests, which can be found in Appendix D. muoonm ooze—am coon HmZmBmE “ooomrmOON 8m m7: moéoom v. COUODUOHQ O 3.0 ~35. Han—O2 ZO—HZNE—NAGA WMAGAHB A<~UOm m0 mmmhm2 ”U N—QZmA—mé. 8332 9 208an .3 02585 >20 9:922 9 £22202 mo oocfifln 5m 3:380 <25 FEE—mose—Z we 6003 ~20 HEB—352 v0 v.9 one E E» so: B ~80 223:9; 5:26 :22; 28m ooh been I 26 Good 2222 28.62 mo .32 b mzo> $83 2222 Eoeoo ow E 30E 33265 new 892m 3503 new; E882 mdo comm 33— :88.“th .mE . 2.6.2:. 5.2%. $321362 . 233,37 2,: 8:285 2653., £62852 3 m _ .0: m .2259 825 $2900 .6 8 232 go oawewwm 665:5 5 30cm $2605 <25 93:28:02 we 3&me mmbk $88 :2? 9 :2"; .6222 223 :80 22m w< .8 62:? 93:28:62 3 8.48m women 3:095 3:526 .88. B2098 mo 33> ESP 838m 355% £me 368 Brag co $2.2: .3535 Awoomv 9 BEBE H<> 02E 9 EEOE mo $52.52 a??? can zoEcowH m 22 0:35.582 *0 Esogoucoo 2: “8 3:850on 283 E 8 EE< ESSA N :45 25:. 265E322 :50th ON .,.m4<>Q< ova; oscfiESoE Eueom om ku<>Q< mtg; v. E mQ<~ElEOU 23 .52 amino. u. 3 magi/Bu z: :2 .580 u. x 3 maépzou z: :2 mmww K. x BEE 823 EN 2a muons: Ew wEEsmm< MQ :EKQQE ooE>oE N mZH ESQSE E 3:32 svfinmbmm m7: $5.30.: £3 m7: 80.3? $3 2262030 SE>oE : .98 9 v8.7.3.6: I 236530 v08 .3 SwacooSE mm votESM ES. coo. :SUSE E 355 zilxmoo 226598 855E a :29 9 833.6: I 236530 308 («o SwSSSSE mm notomom m5 coo. :SuSE E SE5 sfiixmg gm qlfiiucxfig. BEE: 8:235 . .. Sm QO SE:< :SoSE mm C .x m E cozflzoBD . + AQUQKSQXMOQ . QUQkBQQmANVH v N III \SKQXM LOSHD< uCOUhOnm VN.w «Suanm :¢_3_=u_aD>:oEEcD «8.5m 3.5 3:530 SEE—5r osEAEnnoE 58 532:3.3 559:3ch 5232809 Sac—52mm "ND ~35. 142 APPENDIX D: SENSITIVITY ANALYSIS OF ESTIMATED SUPPLY/DEMAND ELASTICTITIES AND ‘EXPORT’ QUANTITIES As briefly discussed in the model validly section of Chapter 6 (section 6.2.1.2), sensitivity tests were conducted on the model by making slight changes in both the domestic demand and supply elasticities and the quantity of maize flour traded between central and southern Mozambique. The sensitivity tests were ran for a couple of reasons. First was to verify the accuracy of the model. By making slight changes in both the elasticities and quantities of maize flour traded, the results indicate the robustness of the model to small changes. By illustrating the models resilience to small changes in the elasticities, the sensitivity tests also helps to provide confidence in the estimated elasticities, which is important due to the data constraints that occurred during the estimation of the domestic supply and demand elasticity estimation. Therefore, the sensitivity tests help to provide confidence in the estimated elasticities, in addition to the overall model. Table D.l contains the results for the sensitivity analysis on the estimated supply elasticities of maize grain with constant domestic demand elasticities of maize flour. Five of the six sensitivity runs are a variation of the different regional supplies being increased by 0.25 from the baseline (indicated by a bold elasticity). The sixth run uses assumed supply elasticities of maize grain from Cruz (2006). The results of the sensitivity tests are reported in total value or quantity and percent difference from the observed value. The sensitivity test runs for the variation in domestic SUpply elasticity for maize grain indicate the models overall resilience to small changes of the domestic I43 supply elasticities of maize grain. Although the South African maize grain variablez' appears to be more sensitive to changes of the supply elasticities than other variables, with the percent difference between the simulated and observed value ranging from 3 percent to 47 percent (absolute value), most variables stay within a 10 percent range of variation from the observed value. Sensitivity run 4 and 6 are considered the most accurate combinations of domestic supply elasticities due to their overall small variation from the observed prices and quantities. In addition to the South African maize grain quantity, the central Mozambique ‘export’ maize flour quantity also shows considerable variability from the observed value and variability between elasticity changes. However, special note should be taken when interpreting this variable, as the observed value is actually an assumed value by the author. Originally the value was assumed to be zero, but the model required that the quantity be a positive integer, resulting in 1 MT of maize flour to be the “observed” quantity moving from central to southern Mozambique based on the current maize market structure in southern Mozambique (see section 2.1.4). Due to the assumption made and the considerable percent differences between the simulated and “observed” values, sensitivity tests were conducted on the quantity of maize flour flowing from the center to the southern region of Mozambique (Table D4). The model shows considerable differences between the central Mozambique FOB observed price of maize flour and the simulated price of maize flour. This is hypothesized to be due to the observed FOB maize flour price used for central Mozambique. Since maize flour rarely moves from central Mozambique to southern “l Due to the variation ofthIs vanable, It Is shaded gray In all tables, Including D.l I44 Mozambique the accuracy of the price used for this variable is weak. However, the implications of this are believed to be small considering that all simulations in the sensitivity tests estimate the value of the price of maize flour to be within the same range. Since the removal of the VAT and import tariffs are compared to the simulated baseline used, the changes in price due to the removal will not be the difference between the observed prices and simulated changes from a change in trade policy, but the comparison between the simulated baseline results and the simulated change in trade policy results. Table D2 illustrates the results of the sensitivity tests where domestic supply elasticity was held constant and domestic demand elasticity, varied. Again indicated in bold within the table, 0.25 was added and subtracted from the estimated (baseline) domestic elasticities of maize grain for each region. A range of combinations were ran, including individual regional changes in demand elasticities of maize flour, while the other regions were held constant at the OLS estimated elasticity of demand. Results from the sensitivity runs when 0.25 was added to the estimated domestic supply elasticity of maize flour, runs 7 through 10, exhibited more variation from the observed prices and quantities as compared to the baseline. When 0.25 was subtracted from the estimated demand elasticities of maize flour, runs 1 I through 14, the simulated results seemed to be more consistent with the results found in the original simulated baseline results. South African export maize grain quantity also varied throughout this section of sensitivity tests more than other variables, ranging from 8 percent of the observed quantity to 57 percent of the observed quantity. The central Mozambique quantity and price of maize flour also varied considerably from the observed value in I45 these runs, however, the simulated price of maize flour stayed within a 10 percent margin. Run 13, where all the domestic demand elasticities for maize grain were increase by -0.25, was determined to be the ‘best’ run due to the least variation from the observed prices and quantities, excluding the South African maize grain quantity and central Mozambique maize flour price and quantity variables. Run 12, with a domestic demand for maize flour for southern Mozambique which was increased by —O.25, while the other regions domestic demand elasticities were held constant at the estimated baseline level, simulated the second best run with the least variation from the observed prices and quantities, again with the exclusion of the problem variables discussed above. Table D3 shows the results of the sensitivity tests for the simultaneous changes in the domestic demand and supply elasticities for each region, using the sensitivity runs with the least variation from Table DI and Table D2. The table indicates each demand and supply elasticity used for each region. In addition, the heading of the column notes the table and run for each demand and supply elasticity grouping. Runs 19, 20, 22, and 23 are the only sensitivity runs where both the supply and demand elasticities are changing from the baseline estimation, besides run 24 which uses the elasticities from the ARIMA model. Runs l9 and 20 use the same domestic demand elasticities (D.2-12) and there is little difference in the supply elasticities used, except run 20 uses a high (relative) supply elasticity of maize grain for South Africa. Both results show low percentages of variation between the simulated and observed prices. The percent differences between the observed and simulated quantities are overall less than the estimated baseline as well. The same comparison can be done between runs 22 and 146 23. The percentage differences between the simulated and observed values are similar to the results from runs 19 and 20. However, the quantity of maize grain from South Africa estimates continue to show their sensitivity to a change in supply and/or demand elasticities due to considerable variation across elasticities. Overall the model appears to be resilient to small changes in the elasticities. Most variables show little variation when small changes in elasticities of either supply and/or demand are made. Considerable variation still occurs in the South African maize grain quantity traded variable, although little can be done since the observed value comes from a reliable source and the value was double checked against another source. The price of maize flour in central Mozambique remains problematic, but as discussed above, because all simulated prices are in the same area, this problem is believed to be with the observed value, not the model. Since the quantity of maize flour from central to southern Mozambique was assumed to be 1 MT and the considerable variation that occurred during the sensitivity tests of the elasticities, the quantity of maize flour from central Mozambique was tested for sensitivity. Flow quantities from the center to the southern region of maize flour were increased for four different sensitivity runs. The results indicate that the model is not sensitive to an increase in quantity of maize flour traded under the baseline estimation as there was little variation in the simulated prices or quantities across the different sensitivity runs. More specifically, the percent difference between the simulated quantity and the observed quantity of maize flour traded to the center region did not notably change. The quantity of maize flour flowing from the center to the southern region of Mozambique was only increased to 500 MT of maize flour, however, that is 20 percent of I47 the total maize flour imports for 2006 and it is therefore believed that 500 MT would be the maximum amount of maize flour flowing from the center to southern Mozambique, which is compatible when using Abdula (2008) belief that no more than 5 percent of maize flour comes from central Mozambique. I48 I49 EE _ i I :3. 9.3.2 :3- :32 3.: £32 :2- $22 $3.:- m..:...:_ M 3:62 E. 25.: .8: 52:33. I PE): 2039:3ch 3.: :3: m: 2:: mm: 35.: :3 :3: 2:. :::.~ :3: =5: Rfi- :33: 3%». :E.:: 3...: 83: 3.5.- 533. 3::- ::m.:m $3: 520 8.2.? £3: $.32 no: ::.:N- «.5: ::.::- 3.: RS.- fi: :3:- mh: _ 3......— ~:.:m- 332 up: - R _ .w _ N :3:- :-.E :3:- _ a. _ mm 3::- :m:.£~ 35:: 58c .82 3:56 PE): x5550 L595. :3- ~::.: R:- xmo: :3.- ~::.: 3:. :5: ::.m- 8:: Rd: 28.... :: .:- 3%.: E N- : _ 3 av:- 32 2 .:- :32 .3:- ~R.~ :8: £95 at? :28: R $.- _:N.: Ric _ a: $3.. 9... _ .: NE.- 2: _ .: 3::- :~_.: 2:: .53... :3 N - v.8... 2.: \- mafi mamm- xcf NE \ - :2: E .S. :33 33 5:5 ES ESSD :3- a: :5- ES :2- :t: ::.m- 2a.: :2- :2K 33. :8...— m:.:. 43.: am:- SQN :3:- :::.N 2:- :5: RN. :3 :2: 526 we: 5:535. «GOP—0A— o=_a> uCOOLOn— o=_a> 2.59—0am 0:_a> unuahom 03—5, u—SUhum 0:1: > CI—Zbo—Zv wuotm :n: 3.: mm: m _ .: t .: ma: m _ .: «a: : mm: 5 _ .: : 2.: 2.: : Em=om - _ cam ESEESm - 258.5 S>SEO SEES> :5:ch 65.5» S>SEO E9... “go—SSEG ESSA— EE mos—5r éozata> bmmzm 93:25.9 . 535: fizzmSmu—d :33. Table D.l: Continued Variables Observed Sensitivity Run 5 - Sensitivity Run 6 - Values Author Est. + 0.25 Cruz (2006) SM CM SA SM CM SA 0.25 0.42 0.13 0.3 0.45 0.65 Prices (Met/MT) Value Percent Value Percent Southern Moz. Grain 3,140 2.900 -7.64 3,023 -3. 72 Flour . 7,391 7.261 —l.76 7,335 .0. 76 Central Moz.. Grain I 2,343 l,850.37 21.01 1,973 —15. 76 Flour . 9,302 6.21 1 33.23 6,285 -3244 South Afrieci Grain E 3,035 2,776 -8.54 2,894 4.62 Flour 6,137 ' 6.028 —/.77 6.090 -0. 77 ‘Export’ Quantity (MT) I Central Moz. Grain . 257,743 ; 183.647 -28. 75 199,393 -22.64 F101" E 1 i 0.55 -45.46 0.53 .4659 South Africci ‘ ' *Gr'ain E 85,256 § 88.254 3.52' "63346 425.11 ' F101" 2,628 ‘ 2.666 1.43 2.645 0.64 Consumption (MT) Southern Moz. ! Flour 196,045 195,771 0.42 190,120 -2-4 7 Source: Estimated by Author. Note: This sensitivity test holds the estimated domestic demand elasticities constant, SM=-0.87, CM=-0.33, SM=-0.5. Bold elasticity indicates a change from the baseline run 150 8:- 232 2.: 2.82 :5: :2.::_ :3- 25.2. 8.: 23:2 M :36: m 5:: 293 5.335% 1 3.2: comEEagU 3.: 22 :3- 22 2.2 2:2 2.: 22 2:2 52-: 2.2- 2:: 5.2- 52.2 :3:- 53: 2.:- 225 :22 56.5 5.5.5: :55: 8::- 2:.: 2.2. 8: :92- 2.: 2.2. 2.: 5:2- 5.: _ . 5:5 2:2- 22:2 25- :::.:~ :32- 2::._:2 2:5. 95.22 2.2- 5:52 2.52 m :55 $.33 23:69 CL): 5:530 .toaxm. :3- 22.: :2:- 25: 2::- E: ::...-- 22.: :2..- 2:: 5.: m 5:: 2.:- :2.2 2.2- 2:2 2.:- :::.2 :2- 22 2.:- 3:2 2:.»- 52v 5.2? 55:: 2.»:- :~_.: 2::- :_:.: :5.- :2.: 2.2- :2: 2:2- 23- N::.: m .52: 2.2- 2:; 2::- 251 :92- 5:2 :2- :22 2.2- 2:; :32 m 550 me: 3:59 :52- :22 2:. :::.: ::.:- 85.: :2- :22 :3.- :_:2 an.» . 5:5 5.:- 52 :2:- ng :2:- :2.2 2:2. :22 :2:- 22 :2: m :55 .95.: 53:33. :5 55> 6: 55> 6: 55> :5 55> :5 55> @253: 35: ::.:- :m.:- 2:- 2:. 8:. 3:. :3. 2:- 3:- :m.:. 2:. ::.:. E:- ..:.:- 5::- :5:ch 635’ 63.330 :8: 3:93.55 Eco-3m 6:: 823/ .5532; 65:3: games—ca . max—:5: bEzmgm "md :35- 151 m _ .on 5: 55> 5: 55> 5,: 55> @232: .65: ::.:- mm:- :m.:. ::.:- :2. :2. :2. :m.:. 3.5 ::.:- 2:- :2- - m— Cflm hemigm—sw . V— cay— xumifiw—SW .m— :5”— b—mfiummcvw . N_ ::m Azizzcom Uo>homLC m2£§5> 2:558“: :d 55: 152 Table D.3:Sensitivity Analysis - Domestic Sunny/Demand Variation, (Cont'd) Variables E Observed Baseline - Author Est. Sensitivity Run 16 Sensitivity Run 17 g Values 2 S: D.1(4) D: Baseline S: Baseline D: D.2(13) SM CM SA SM CM SA SM CM SA 5"PP’." 0 0.17 0.13 0.25 0.42 0.38 0 0.17 0.13 Demand -0.89 -033 -050 -O.89 -033 -050 -1.14 -0.58 -075 Prices (Met/MT) Value % Value % Value % Southern M02, Grain 3,140 2,896 -7. 77 2,974 -528 3,054 -2. 75 ”our 7,391 7,170 -299 7,301 -1.21 7,433 0.57 Central M02; Grain 2,343 1,846 -2179 1,924 -l7.85 2,004 44.47 Flour 9,302 6,120 -3421 6,251 -32. 79 6,383 -3l.38 South A friea Grain 3.035 2,772 -8.67 2,847 -6. I8 2,924 -3.66 “0‘" 6,137 5,953 —3.00 6.062 -l.22 6,172 0.57 ‘Export’ Quantity (MT) Central M02, Grain ' 257,743 214,059 -16.95 194.793 -2442 216,242 -I6.lO F'Our 1 0.73 -2736 0.55 -44.87 0.65 -3492 South Africo Grain 85,256 56,539 -33. 68 71,810 -15. 77 45,759 -4633 F101" 2,628 2,686 2.21 2,654 0.99 2,614 -0.52 Consumption (MT) Southern M03 3 , 1 F'O‘" 196.045 194,945 -0.56 190,120 -247 189,284 -345 153 Table B.3: Continued Variables Observed Sensitivity Run 19 Sensitivity Run 16 Sensitivity Run 17 Values s: D.1(4) D: 0.2(13) s.- D.1(6) D: 0.2(13) 8: Baseline s.- 0.2(12) SM SM SM SM CM SA SM CM SA SW’P’Y 0.25 0.42 0.38 0.3 0.45 0.65 0 0.17 0.13 Demand -1.14 -0.58 —0.75 -1.14 -0.58 -075 -1.14 -033 -050 Prices (Met/MT) Value % Value % Value % Southern M02, Oral" 3,140 3,076 -204 3,095 -143 3,040 -319 “0‘" 7,391 7,472 1.10 7,463 0.97 7,451 0.81 Central Moz Grain 2.343 2,026 -1351 2,045 -1269 1,990 -1505 “our 9,302 6,422 -30.96 6.413 -3l.06 6.401 -3119 South Africa Grain ' 3,035 2,945 -2.94 2,964 -233 2,910 -4.09 “0‘" 6,137 6,204 1.10 6,196 0.97 6,187 0.81 ‘Export’ Quantity (MT) Central M02, Grain . 257,743 199,185 -22. 72 200,508 -2221 226,711 -1204 Flour 1 0.48 -5205 0.46 -54.34 0.35 -64.98 South Africa Grain 85.256 61,325 -28.07 58.328 -3I.58 36.441 -57.26 “0‘" 2,628 2,599 -109 2,602 -097 2.609 —0. 74 Consumption (MT) Southern Moz. Fl 188,301 0‘" g ”6,045 g -4.11 187,216 -450 190,026 -3.07 l54 Table D3: Continued Variables Observed Sensitivity Run 22 Sensitivity Run 23 Sensitivity Run 24 Values 8.- D.1(4) D: D.2(12) s: D.1(6) D: D.2(12) ARIMA Estimates SM SM SM SM CM SA SM CM SA S‘W’y 0.25 0.42 0.38 0.3 0.45 0.65 0 0.22 0.78 Demand -1.14 -033 -050 -1.14 -033 —0.50 -1.02 -0.48 -055 Prices (Met/MT) Value % Value % Value % Southern Moz. Grain 3,140 3,073.30 -213 3.097 -l.36 3,064 -24 “0‘" 7,391 7.495 1.41 7,478 1.17 7,358 0.0 Central Moz. Grain 2,343 2,023 -l3.63 2,047 -1260 2.014 -1403 F10” 9,302 6.445 -30. 71 6.428 -30. 90 6,308 8219 South Africa Grain 3,035 2,943 -303 2,966 -227 2,934 -330 F10" 6,137 6,224 1.42 6,209 1.18 6,109 0.0 ‘Export’ Quantity (MT) Central Moz Grain . 257,743 209,680 -l8.65 21 1,079 -l8.10 217.917 -1545 Flour 1 0.58 -4202 0.56 -4438 0.645 -3550 South Africa Grain 85.256 51.076 -40. 09 47.567 -4421 42,549 -500 “our 2,628 2,591 -1.41 2.597 —1. 18 2.638 -00 Consumption (MT) Southern Moz. ' “0‘" 196,045 3 188,452 -3.87 187,087 -457 188,311 -3.40 Source: Estimated by Author. as S32 :3 $0.82 86 «8.2.. 86 $32 $5. $32 _ $32 so: as: Emizew. 1 8.2V :EEEJECU $12. 32.. AR- 53 mi. £3 03 9.5 EN £3 ”SJ :5...— mfifi 33m. no.2. 9.3m no.2. 3.35 3.2. 3.2m. more anew. cmfimw £20 at? 53% 3%. Nod 82%. 2m 3%. 2: him. cm 3.2. m2. _ 52m 32. $3.: 3.2. 3.34: 32. «.885 3.2- $3.: 32. 3.3a 35mm :35 was 3.530 C13: 5:550 .aacaxm. 9%. come 32.. 53 NE: Rem 3.4.. 3.3 8.4.. $5.4. Ea so: new. 92 3.x. 2.2 New. 22 New. Ed 2.x. Ea mm? 52o 4.3.2? £43m 2.9.- toe $3.. mac 824.. so: 3:3,. 226 $3.. 22 Sea as; 2.x. ex: 2:. as: 2:. 53 2:. e3: 2:. 0%.. $3 5.26 We: 3.5an 034. E: 38.4.- we: $4.. 3.3 Em. so; 92. E3 33 .56; Ex. 33 RN. 033 EN. 82 Rx. .53 RN. e2: 83 W EEO at? 2.3555 88 33> 88 83> s 2.3» a 33> e 33> £2 26 .825 Snuwumd Smumund oflumumd Edmond 1203a mo:_m> ”N :3”— hugmummflow INN =5"— .hugmummflvm ON :5”— ~m=>mumm=om mN 5.1m hzfizmcow 4mm ._053< 1 vE—Dmsm UPCUwLC m03fiffl> .mtcmxm. ESE 2.538332 3.550 . mam—23. 53.65.,“ 6.9 935. 156 BIBLIOGRAPHY 157 BIBLIOGRAPHY Abdula, D. 2005. Improving Maize Marketing and Trade Policies to Promoted Household Food Security in Southern Mozambique. MS thesis, Michigan State University. Abdula, D., email message to author, May 6, 2008. Anderson, J. 1998. “Effective Protection Redux.” Journal of International Economics, 21-44. Antimiani, A., P. Conforti, and L. Salvatici. 2003. The Effective Rate of Protection of European Agrifood Sector. Paper presented at Agriculture Policy Reform and the WTO: Where are we heading? June 23-26, in Capri, Italy. Amdt, C., R. Schiller, and F. Tarp. 2001. “Grain Transport and Rural Credit in Mozambique Solving the Space — Time Problem.” Agriculture Economics, 59-70. Balassa, B. 1965. “Tariff Protection in Industrial Countries, an Evaluation.” Journal of Political Economy, 573-594. Bates, T., and A. Schmitz. 1969. A Spatial Equilibrium Analysis of the World Sugar Economy. University of California, Giannini Foundation Monograph 23. Castel-Branco, C.N. 2004. What is the Experience and Impact of South African Trade and Investment on the Growth and Development of Host Economies? A View from Mozambique. Paper prepared for Conference on Stability, Poverty Reduction and South African Trade and Investment in Southern Africa for SARPN/HSRC, March 29-30, Pretoria, South Africa. Chevassus — Lozza, E. and J. Gallezot. 2003. Preferential Agreements — Tariff Escalation: What are the Consequences of the Multilateral Negotiations for Access of Developing Countries to the European Market. Paper presented at Agricultural Policy Reform and the WTO: Where an .ding?, June 23-26, Capri, Italy. Clark, DP. 1985. “Protection and Developing Country Exports: The Case of Vegetable Oils.” Journal of Economic Studies, 3- l 8. Cunguara, B. 2008. Pathways out of Poverty in Rural Mozambique. MS thesis., Michigan State University. Cordon, W.M. 1966. “The Structure of a Tariff System and the Effective Protective Rate.” Journal of Political Economy, 22 l -237. 158 Cruz, A. 2006. Maize Trade in Southern Africa: Comparative Advantage on Storage Costs. National Directorate of Studies and Policy Analysis Discussion Paper No. 32E. Mozambique Donovan, C. I996. Effects of Monetized Food Aid on Local Maize Prices in Mozambique. PhD diss., Michigan State University. Donovan, C. 2008. Non-food CPI for Mozambique. Unpublished. Durand-Morat, A. and E. Wailes. 2003. RICEFLOW©: A Spatial Equilibrium Model of World Rice Trade. Unpublished. Elamin, N. and H. Khaira. 2004. “Tariff Escalation in Agricultural Commodity Markets.” In Commodity Market Review 2003-2004, FAO. http://www.fao.org/docrep/006/y5l l7e/ y5 l l7e00.htm. Ethier, W. 197]. “General Equilibrium Theory and the Concept of Effective Protection.” In Effective Tarifl Protection, eds. H. Grubel and H. Johnson, H ., 17-44. Geneva: GATT. Either, W. 1977. “The Theory of Effective Protection in General Equilibrium: Effective- Rate Analogues of Nominal Rate.” Canadian Journal of Economics. FAO. 1997. White Maize: A Traditional Food Grain in Developing Countries. Rome: FAO. Ferris, J. 2005. Agricultural Prices and Commodity Market Analysis. East Lansing: Michigan State University Press. FEWSNET. 2008. Drought Days and Total Rainfall Database. Greenaway, D. and CR. Milner. 1987. “True Protection Concepts and the Evaluation of Trade and Industrial Policies in Developing Countries.” Journal of Development Studies, ZOO-219. Greenaway, D. and CR. Milner. I990. “Policy Appraisal and the Structure of Protection in a Low-Income Developing Country: Problems of Measurement and Evaluation in Burundi.” Journal of Development Studies, 22-42. Goldberger, AS. 199]. A Course in Econometrics. Cambridge: Harvard University Press. Golub, F. and J .M. Finger. I979. “The Processing of Primary Commodities: Effects of Developed-Country Tariff Escalation and Developing — Country Export Taxes.” Joumal of Political Economy, 559-577. l59 Guha — Khasnobis, B. 2004. “Who Gains from Tariff Escalation?” Journal of Economic Integration, 416-425. Gujarati, D. 2003. Basic Econometrics. Boston: McGraw Hill. Hamilton, J. 1994. Time Series Analysis. Princeton: Princeton Press. Hassan, R., D. Greenaway, G.V. Reed. 1992. “Nominal and Effective Protection in the Egyptian Agriculture Sector: A multi-commodity Analysis.” Applied Economic, 483-492. Humphrey, D. 1969. “Measuring the Effective Rate of Protection: Direct and Indirect Effects.” Journal of Political Economy, 834-844. INE. I997. National Population Census. Maputo: National Institute of Statistics. INE. 2008. http://www.ine.gov.mz. Jayne, TS. 2008. Maize Marketing Margins in South Africa. Presented for 2008 Brown Bag Series for Agriculture Economics Graduate Students, East Lansing, U.S. Jayne, T.S., Hajek, M. and J. Zyl. 1995. An Analysis of Alternative Maize Marketing Policies in South Africa. MSU International Development Working Paper No. 50. Jones, S. 2006. Growth Accounting for Mozambique (I980 - 2004). National Directorate of Studies and Policy Analysis Discussion Paper No. 22E. Mozambique Koo, Won W., and PL. Kennedy (2005). Intemational Trade and Agriculture. Malden: Blackwell Publishing. Laird, S. and A.J. Yeats. 1987. “Empirical Evidence Concerning the Magnitude and Effects of Developing Country Tariff Escalation.” The Developing Economies, 99-123. Lehola, P. 2002. Report on the Survey of Large and Small Scale Agriculture. Pretoria: Statistics of South Africa. Lindland, J. I997. “The Impact of the Uruguay Round on Tariff Escalation in Agriculture Products.” Food Policy, 487-500. Lindland, l997a. The Impact of the Uruguay Round on Tariff Escalation in Agriculture Products. Working FAO paper ESCP/No 3 Mabiso, A. and D. Weatherspoon. 2008. Fuel and Food Trade-offs: A preliminary Analysis of South African Food Consumption Patterns. Paper presented at I60 Annual Meeting for American Agriculture Economics Association, July 27-29, Orlando, FL. McCorriston, S. and I. Sheldon. 2004. Tariff Escalation and the Developing Countries: How Can Market Access Be Improved in the Doha Round of Trade Negotiations? Paper presented at Annual Meeting for American Agriculture Economics Association, August 1-4, Denver, Colorado. Milner, CR. 1990. “Identifying and Quantifying Anti-Export Bias: The Case in Cameroon.” Review of World Economics, 142-155. Myers, R. 2008. Demand for Commodities. Notes for ABC 845: Commodity Markets. January 14, 2008 Myers, R. 2008a. Spatial Equilibrium Models. Notes for ABC 845: Commodity Markets. April 2], 2008. Myers R. 2008b. Stochastic Properties of Commodity Price Time Series. Notes for ABC 845: Commodity Markets. April 7, 2008. National Chamber of Milling of South Africa. 2008. Industry Statistics. http://www.grainmilling.org.za/. Ndibongo-Traub, L.N. 2002. The Effects of Market Liberalization on Maize Milling/Retail Margins in South Africa. MS thesis, Michigan State University. Ndibongo — Traub, L. and TS. Jayne. 2007. “The Effects of Price Deregulation on Maize Marketing Margins in South Africa.” Food Policy, 224-236. OANDA. 2008. FXHistory: Historical Currency Exchange Rates. http://www.oanda.com/convert/thistory. OECD. I997. The Uruguay Round Agreement on Agriculture and Processed Agricultural Products. Paris: OECD Publications. Safadi, R. and A. Yeats. I994. “The Escalation of Asian Trade Barriers.” Asian Economic Journal, 147- l 79. SAGIS. 2008. Stats SA: Consumer Prices. http://www.sagis.org.za/. Sebei, E. 2002. Trade Potential Between South Africa and Mozambique. Department of Agriculture. Pretoria, South Africa. SIMA. 2008. Data. l6] South African Department of Agriculture. 2008. Abstract of Agricultural Statistics: 2008. Pretoria: National Department of Agriculture. South Africa Department of Agriculture. 2008. Maize. httpzl/www.nda.agric.za/docs/maize/maize.htm. South African Department of Transport. 2005. Grain and Seed Production. http://www.kzntransport.gov.za/public_trans/freight_databank/kzn/industries/grai n_miIIing_distribution/index_xml.htmI. South African Futures Exchange. 2008. Market Data. http://www.safex.co.za/. South African Government Information. 2008. South African President Thabo Mbeki to attend South African-Mozambique Heads of State Bilateral Economic Forum in Mozambique. http://www.info.gov.za/. South African Grain Information System (SAGIS). 2008. Stats SA: Consumer Prices. http://www.sagis.org.za/. Statistics South Africa. 2008. Key Indicators. http://www.statssa.gov.zal. Takayama, T. and G. Judge. I97]. Spatial and Temporal Price and Allocation Models. London: North-Holland. Tangerman S. I989. Tariff Escalation in Agricultural Trade: A Research Survey and Illustrative Case Study for Cocoa and Soya. Stuttgart: Wissenchaftsverlan g. Tomek, W. and K. Robinson. I977. “Agriculture Price Analysis and Outlook.” In A Survey of Agricultural Economics Literature: Volume 1, ed. L. Martin, 329- 409. Minneapolis: University of Minnesota Press. Tschirley, D. and D. Abdula. 2007. Toward Improved Marketing and Trade Policies to Promote Household Food Security in Central and Southern Mozambique: 2007 Update. Research Paper Series No. 62E. Tschirley, D., D. Abdula, M.T. Weber. 2006. Toward Improvement Marketing and Trade Policies to Promote Household Food Security in Central and Southern Mozambique. Research Paper Series No. 60E. United Nations. 2008. UN Commerce of Trade Statistics. http://comtrade.un.org/. FAO. 2008. FAO STATS. http://faostat.fao.org/. United States Department of Labor Statistics. 2008. Consumer Price Index. http://www.b|s.gov/. I62 Valenzuela, E., T. Hertel, M. Ivanic, and A. Pratt. 2004. Evaluating Poverty Impacts and Trade Policy Changes on Agricultural Producers. Paper presented at Annual Meeting for American Agricultural Economics Association, August 1-4, Denver, Colorado. Wailes, E., A. Durand-Morat, L. Hoffman, and N. Childs. 2004. Tariff Escalation: Impacts on US. and Global Rice Trade. Paper presented at American Agriculture Economics Association Annual Meeting Denver, Colorado. August 2004. World Bank. 2008. World Development Indicators. The World Bank Group. WTO. 2008. Tariffs. http://www.wto.org/english/tratop_e/tariffs_e/tariffs_e.htm. Yeats, A. 1984. “On the Analysis of Tariff Escalation: Is there a Methodological Bias Against the Interest of Developing Countries?” Journal of Development Economics,77-88. Zusmand, P., A. Melamend, and I. Katzir. I969. Possible Trade and Welfare Efifect‘s of EEC Tarifl and ‘Reference Price’ Policy on the European — Mediterranean Market for Winter Oranges. University of California, Giannini Foundation Monograph 24. “'11111111111111111111111111111s