Essays on South Africa: Exchange Rates, Bilateral Trade and Inflation by Peter Kariuki Kinyua A dissertation submitted to the Graduate Faculty of Auburn University in partial fulfillment of the requirements for the Degree of Doctor of Philosophy Auburn, Alabama May 6, 2012 Keywords: economics, exchange rates, real foreign sector, metals, inflation, South Africa Copyright 2012 by Peter Kariuki Kinyua Approved by Henry L. Thompson, Chair, Professor of Economics John D. Jackson, Professor of Economics Richard O. Beil, Associate Professor of Economics Asheber Abebe, Associate Professor of Mathematics and Statistics Brenda M. Allen, University Reader, Assistant Professor of Forestry and Wildlife Sciences ii Abstract South Africa shows how natural resources can be harnessed to build a successful economy. This success gives rise to peculiar macroeconomic issues that warrant analysis. Chapter one of this dissertation investigates the effects of exchange rate volatility on South Africa?s export of metals, using monthly data for the period 1980:01 to 2011:07. The study uses squared residuals from the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) process to generate a measure of exchange rate volatility, which is then tested in a model of South Africa?s metal exports. Utilizing conventional cointergration techniques, the study estimates both the short-run and long-run impacts of exchange-rate volatility (and other macro-variables) on South Africa?s export volumes of 11 major metals. Results suggest that exchange rate volatility increases export demand for South Africa?s base metals both in the short-run and in the long-run. Chapter two proceeds with an examination of relationships between South Africa?s bilateral trade volumes with 42 of its major trading partners. Annual data for the period 1970 to 2010 is used in the context of a gravity equation. The select variables represent importer/exporter?s real income, population, export prices, unemployment, real effective exchange rate, exchange rate volatility, and a set of dummy variables representing involvement in a Preferential Trade Area (PTA) and a dummy variable signifying a change in South Africa?s international trade pattern. Results show that real foreign sector has a significant impact on South Africa?s bilateral trade. Exchange rate volatility yields mixed results for import demand and iii depresses trade for exports. PTAs are found to be building blocks to trade, while income inequality within a PTA results to trade diversion. Finally, inclusion of the gravity equation?s intangible attributes such as language, colonial ties, and culture is justified. Lastly, the causes of rising demand pull inflation in South Africa are examined with an eye on the international price of gold given the importance of gold mining in the country. Effects of the money supply, exchange rate, foreign income, and an index of political stability are included in the model. Results show that money supply, exchange rates, and the price of gold and world income to be the major determining factors of inflation levels. The evolving monetary regime and political stability are also found to positively influence inflation levels. iv Acknowledgments I would like to offer special thanks to my academic advisor, Prof. Henry Thompson, for his guidance and insightful comments that have brought this work to a successful completion. I would also like to thank my committee members, Prof. John Jackson, Prof. Richard Beil, Prof. Asheber Abebe, and Prof. Brenda Allen for their invaluable comments and criticisms. Special thanks to Prof. Curtis Jolly, Chair of Agricultural Economics and Rural Sociology, and to Prof. Overtoun Jenda, Associate Provost for Diversity & Multicultural Affairs at Auburn University for providing full funding for my doctoral studies. Finally, I owe my family a big debt of gratitude and appreciation for their patience and support during these four and a half years. I however bear full responsibility for any opinions and/or errors in this dissertation. v Table of Contents Abstract.....??. .?????????????.???????... .?????? .???. .ii Acknowledgments..?????????????????????????????? iv List of Tables?.. ????????...?????????.??...????? ? ?? .......vii List of Figures ????? ???????????.????????..???????.ix List of Abbreviations???????????? ????????????.??? .??x i Chapter 1. Exchange Rate Volatility and South Africa?s Metal Exports??????????1 1. Introduction????????????????????????????.1 1.1. Outline of the Study???????????????? ?... ???...?2 2. Literature Review ???????????...???????.? ..????? 2 2.1. An Overview of South Africa?s Economy? ?????????... ???.3 2.2. Export Markets??? ???????????????.. ??????4 2.3. Export Prices, Exchange Rates, and Volatility? ???? ???????.6 2.4. South Africa?s Economic Sanctions????? ??????. ????..12 3. Model Specification???? ??????????.? ?????...........?.12 3.1. Data Sources and Variables??? ? ???????????????16 3.2. Data Plots..??? .? ?...?? ????????????? ..???? 17 4. Empirical Results ?................................................................... ..............................18 4.1. Stationarity Analysis?.. ? ..?????????? ??????.??.18 4.2. Model Estimation..???????????????..... ........................19 5. Summary and Conclusion... ?.............. ...................................................................23 Chapter 2. Exchange Rate Volatility and South Africa?s Bilateral Trade?????????44 1. Introduction ???????????????...... .......................????...44 1.1. Outline of the Study?. ?????????.. ?.??????????45 2. Literature Review ???????????...?????????????45 2.1. The Gravity Model?. ?.?? ?. ???????...?????????45 vi 2.2. Unemployment?. ??? .?? .????????...?? ??????..48 2.3. Preferential Trade Agreements? .? ..? ??? ???????????51 2.4. Gross Domestic Product??. ? ..? .???????????.????.56 2.5. Export Prices?.?? .....????? .???????????..???..57 2.6. Information/Information Costs.....?... ? ....??????????..??..58 3. Model Specification..??????????????????????...?59 3.1. Model Selection Criteria?? ??? ???????????????.64 3.2. Data Sources and Variables?? ?? ???...???????????.65 3.3. Data Plots?? ????? ?? ?????????????????65 4. Empirical Results and Discussion.?........................................ ..............................66 4.1. Import Demand??.. ?????????????????????.68 4.2. Export Demand??.. ? .???????????????..????..70 4.3. Unemployment??. ?????????..????????????71 4.4. Intangible Effects??.. ????????????????????..72 5. Summary and Conclusion..????????????????????.?72 Chapter 3. World Gold Prices and Demand Pull Inflation in South Africa???????.....114 1. Introduction ???????????? ?????????????..?114 1.1. Outline of Study??? ?...???????? ?????? ..??...?115 2. Literature Review ?? ??...?????????...???????...??115 2.1. Gold Mining and South Africa?s Economy??. ? ? ????? ? ??.115 2.2. Inflation??. ???? ....????? .???????????.??..118 2.3. Monetary Regimes??. ??. ?? .???????????..???...120 2.4. Exchange Rates and the Rand??.. ?? .??????????.. ??..120 2.5. Gross Domestic Product??... ?? ????.??????.. ????..122 3. Model Framework....???? ??????????? ???????.?.122 3.1. Deriving the Inflation Equation.....?... ??? ????????? ? ?.123 3.2. Data Plots and Sources??? ???????? ???.. ??????128 4. Empirical Results and Discussion .........................................................................129 4.1. Stationarity Analysis??? ??? ????? ??????????.129 4.2. Model Estimation??? ...?????????????????...?129 5. Summary and Conclusions??... ........................................................................131 vii References???? ????????????????????????????? 140 viii List of Tables Table 1.1. Stationarity Analysis for Commodities???????? ?????... ? ???..24 Table 1.2. Levels Model for Commodities?????... ?????? .????? ????27 Table 1.3. ECM Model for Commodities?????.. ?????????????? ??29 Table 1.4. Derived Effects for Commodities????... ??????? ?????? .??.31 Table 1.5. Short-Run Estimates for Aggregate and Disaggregated Data? ?... ??... ????33 Table 1.6. Summary Chronology of Trade Sanctions against South Africa?? ??? ???..35 Table 2.1. Stationarity Analysis for South Africa?s Bilateral Trade?... ????? ? ??.?..75 Table 2.2. Stationarity Analysis for USA and South Africa Bilateral Trade??... ?? ?.??.81 Table 2.3. Stationarity Analysis for Australia and South Africa Bilateral Trade??.....???.83 Table 2.4. Stationarity Analysis for Taiwan and South Africa Bilateral Trade?... ..? ????8 4 Table 2.5. Stationarity Analysis for Japan and South Africa Bilateral Trade?... ?? ????.8 5 Table 2.6. Results for South Africa?s Imports?... ?????????????? ...???.8 7 Table 2.7. Results for South Africa?s Exports..?. ????????????? ?????97 Table 2.8. Levels Model for South Africa?s Bilateral Trade??... ? ..????? ??? ?..106 Table 2.9. ECM Model for South Africa?s Bilateral Trade?... ..???? ???? ?...??.107 Table 2.10. Derived Effects on South Africa?s Bilateral Trade?... ??????? ????108 Table 3.1. Stationarity Analysis??... ????????????? ??????...??.133 Table 3.2. Inflation Model in Levels??... ???????? ??? ..???????.?.134 Table 3.3. ECM for Inflation??... ??????? ???????? ????????135 ix Table 3.4. Derived Inflation Effects??... ????????????? ?????...?..136 Table 3.5. Macro Policy and Inflation?... ???? ..?????????? ??...???.137 x List of Figures Figure 1.1. Coal Exports and Corresponding Export Price??? ?? .? ??????..??37 Figure 1.2. Iron Ore Exports and Corresponding Export Price....?? ??. ???????....37 Figure 1.3. Chromium Exports and Export Price?????? .? ???. ??????...?.38 Figure 1.4. Copper Exports and Corresponding Export Price??? ?? ??. .?????.?38 Figure 1.5. Manganese Ore Exports and Corresponding Export Price?? ?? ? .?...???..39 Figure 1.6. PGM Exports and Corresponding Export Price? ??? ??? ??? .?.???39 Figure 1.7. Nickel Exports and Corresponding Export Price?... ? ?????? .? .?.??..40 Figure 1.8. Gold Exports and Corresponding Export Price??? ?????? ..???? .?40 Figure 1.9. Granite Exports and Corresponding Export Price???? ??? ..????...?..41 Figure 1.10. Limestone Exports and Corresponding Export Price? ...? ???? ? ??.. ?..41 Figure 1.11. All Metals Exports and South Africa?s Producer Price Index? ????.?.??.42 Figure 1.12. All Metals without Gold and South Africa?s Producer Price Index?? ? ??.....42 Figure 1.13. Nominal and Real Effective Exchange Rate??? ??????????.??43 Figure 1.14. Rand ? USD Exchange Rate Volatility? ????? ????. ???????43 Figure 2.1. SA and USA Bilateral Trade????? ?? ?????. ??????? ??109 Figure 2.2. S.A. and U.S.A. Unemployment Rates Comparisons?? ??.. ? ? ??.???109 Figure 2.3. S.A. and Australia Bilateral Trade??? ?????... ??? ????...??..110 Figure 2.4. S.A. and Australia Unemployment Rates?? ???... ..??? ????...??..110 Figure 2.5. S.A. and Taiwan Bilateral Trade Flows?? ? ..????.. ?????????111 xi Figure 2.6. Comparing Unemployment Rates between S.A. and Taiwan...?? ??.. ???..111 Figure 2.7. S.A. and Japan Bilateral Trade Flows..?????? ??? ???.. ???...?112 Figure 2.8. South Africa?s Exports to Regional Markets???.. ??? ??. ???? ..? ....112 Figure 2.9. South Africa?s Imports from Regional Markets..?? ???... ???? ..?? ?.113 Figure 3.1. Aggregate Demand-Aggregate Supply Curve ?? ? ???... ?????..??.138 Figure 3.2. Cost Push Inflation????????? ? ?????... ????? ...?..??138 Figure 3.3. IS-LM-BP Equilibrium???? ? ???... ??? ???? .?? ?... ?...??127 Figure 3.4. Variable Series ??????? ...? ??? ?? ?????????... ???139 xii List of Abbreviations AR (1) ? Autoregressive Model of Order One ARCH (1) - Autoregressive Conditional Heteroskedasticity of Order One DF ? Dickey Fuller DFc ? Dickey Fuller with a Constant DFt ? Dickey Fuller with a Time Trend ADF ? Augmented Dickey Fuller ADF (2) ? Augmented Dickey Fuller of the Second Order ARCH ? Autoregressive Conditional Heteroskedasticity ARIMA ? Autoregressive Integrated Moving Average BGS ? Balance of Goods and Services BOT ? Balance of Trade COMTRADE - Commodity Trade Statistics Database ECM ? Error Correction Model EG ? Engle-Granger EMU ? European Monetary Union ERV ? Exchange Rate Volatility FDI ? Foreign Direct Investment GARCH ? Generalized Autoregressive Conditional Heteroskedasticity xiii GDP- Gross Domestic Product IMF- International Monetary Fund LDC ? Least Developed Countries OPEC ? Organization of the Petroleum Exporting Countries PGM ? Platinum Group of Metals PTM- Price-to-Market RER ? Real Effective Exchange Rates RGDP ? Real Gross Domestic Product SAB ? South Africa Central Bank WTO ? World Trade Organization DC ? Developed Countries LDC ? Least Developed Countries PTA ? Preferential Trade Area FTA ? Free Trade Area ROW ? Rest of World NAFTA ? North American Free Trade Agreement EU ? European Union SADC ? Southern African Development Community ASEAN - Association of Southeast Asian Nations AGM ? Augmented Gravity Model H-O-S - Heckscher-Ohlin-Samuelson C-H-O - Chamberlin-Heckscher-Ohlin GVAR ? Global Vector Auto regression xiv CGE - Computable General Equilibrium CAFTA ? China-ASEAN Free Trade Area CARICOM - Caribbean Community and Common Market COMESA - Common Market for Eastern and Southern Africa GATT ? General Agreement on Trade and Tariffs CEES - Central and Eastern European Countries AD ? Aggregate Demand SBC ? Schwartz Bayesian Criterion AIC ? Akaike Information Criteria CFA - Communaut? fran?aise d'Afrique, French Community of Africa CLS ? Continuous Linked Settlement SARB ? South African Reserve Bank 1 CHAPTER 1 Exchange Rate Volatility and South Africa?s Metal Exports 1. Introduction This chapter seeks to determine the effects of exchange rate volatility on South Africa?s export of raw metals. South Africa is a primary producer of metals, historically in great demand in industrialized countries and emerging markets. The study determines whether exchange rate volatility has exerted any impact on export volumes of eleven major metal exports and if so, the magnitude and direction of the impacts. The metals studied are gold, diamond, copper, manganese, iron, platinum group of metals (PGMs1), nickel, chromium, coal, granite and limestone. Other macroeconomic variables such as export price and world GDP are included in the model. A structural break is also included in the model to test for the effects of trade sanctions. The metals studied are primarily exported in their raw form for use in building and construction, heavy and light industrial manufacturing, weapons and electronics. Gold, platinum, and diamonds are also held for financial speculation purposes. This diverse demand for South Africa?s exports, coupled with a dynamic international market, sets the stage for this application of international trade theory. Disaggregated macroeconomic data for all variables is analyzed in a time series framework. The period under consideration is 1980:01 to 2011:07. All variables are monthly series. The model specified in the chapter is a hybrid derived from other past studies on 1 PGM refers to ?Platinum Group of Metals?. This group includes Ruthenium, Rhodium, Palladium, Osmium, Iridium, and Platinum. They have similar physical and chemical properties, and tend to occur together in the same deposits. 2 international trade. All variables are pretested for cointergration using conventional time series methods. An error correction model is further developed and results in the form of elasticities derived. 1.1. Outline of Study This study is organized as follows: Section 2 provides detailed theory, an overview of past literature and justification of variables used, while section 3 describes the data and the methodology used in the study. Sections 4 and 5 provide the discussion of results, and concluding remarks respectively, while the index includes all the results tables, explanation for abbreviations and data plots. 2. Literature Review Most of the studies available on export markets are based on developed countries such as the United States, the European Union, and Japan. The majority of these have employed aggregate trade data i.e., the imports and exports of one country with the rest of the world. Most studies fall within this category. Arize, Osang and Slottje (2006), and Weliwita, Ekanayake and Hiroshi (1999), studied Japanese export performance for manufactures, while Fountas and Bredin (2006), Caporale and Doroodian (2002), and Kenen and Rodrick (1986), examine U.S. sectoral exports. Because of the aggregation bias problem, the second group of studies has concentrated on using trade data at the bilateral level, i.e., import and export data between two countries. In this category, Choudhry (2008), Mckenzie (1999), Sercu and Vanhulle (1992), and Broll and Eckwert (1999) use cointergration analysis on U.S. - Canada bilateral trade. 3 The last category includes only a few studies that have disaggregated the trade data between two countries and have looked at the impact of exchange rate volatility on sectoral data. In this category, Onafowora and Owoye (2011) studied Nigerian export demand for oil and agricultural sectors, while Bleany and Greenaway (2001), and Bah and Amusa (2003) use panel data on South Africa?s export market for different sectors. Ekanayake and Thaver (2011), and Edwards and Lawrence (2006) use cointergration analysis on South Africa?s sectoral exports. This chapter utilizes similar models but goes further by disaggregating data by commodities within a sector. This has an advantage in that it allows the analysis to pay special attention to commodity attributes. This is how the present study intends to contribute to existing literature. 2.1. An overview of South Africa?s economy South Africa?s economic performance over the past 40 years has been rather disappointing. GDP per capita rose to a historical peak in the early 1980s and declined thereafter until a moderate return to growth after political transition in the early 1990s. Even with this recent improvement, GDP per capita as of 2011 remained only 40 percent higher than it was in 1960. Growth in South Africa?s exports during this period has been even more dismal. Although exports have grown in absolute terms over the past 40 years, exports per capita in constant dollars in 2011 are no higher than they were 40 years earlier. Over that period the data are volatile because of price swings in gold, which was over one-third of total exports, and also because of misreporting of exports due to international trade sanctions, but overall export performance is clearly dismal. One may attribute this weak export performance to South Africa?s status as a natural resource exporter, notwithstanding recent evidence from OPEC countries such as Canada and Russia that such an endowment is often an economic blessing .This is not simply due to bad luck 4 in international prices of South Africa?s primary exports, as the country?s relative performance in export volumes is equally poor compared to other natural resource exporters such as China and Australia. Looking across sectors, it seems that there has been a lack of structural transformation in South Africa. The country remains highly concentrated in mining. Even today, the only sectors with large net exports in South Africa are gold, coal, other mining and basic iron and steel. Yet while the endowment of mines is relatively fixed, the country?s population has been rising significantly, from 17 million in 1960 to 50 million today, and as a result, mining output per capita today is less than half its value in 1960 (IMF). At the same time, other sectors of the economy have not picked up the slack. Manufacturing output per capita today is lower than its levels in the 1970s (IMF). 2.2. Export Markets In International trade, the economy of South Africa continues to be heavily dependent on the export of gold and other metals. In 2010, gold accounted for 28 per cent of total South Africa merchandise exports (Mineral Bureau. 2011:5). Given South Africa?s dependence on commodity-based manufactures such as iron, steel and non-ferrous metals, the conventional separation of primary and manufactured goods does not suffice when studying its export performance. Other minerals which provide substantial foreign exchange earnings include platinum, and other platinum group of metals, coal, uranium, diamonds, copper, iron ore, manganese, chrome, nickel, asbestos, fluorspar, aluminosilicates, such as andalusite and heavy minerals (titanium, vanadium, and zirconium). Metals other than gold have contributed to foreign exchange earnings of nearly 20 per cent of total commodity export receipts over the last decade. 5 The direct contribution of the mining sector to the South African GDP in 2010 is estimated to have been about 13 per cent (Chamber of Mines, 2011: No 5). Known gold reserves in South Africa are considered to be approximately equivalent to what has already been mined for more than a century (Chamber of Mines, 2011, No. 4: 2). The extent to which this growth potential can be effectively exploited depends on a number of factors such as the international price of gold, the exchange rate, the tax structure, institutional environment, and development of new technologies. Since 1980, export-led growth has been a key element of the South African Government?s Growth, Employment and Redistribution strategy. Exports have been promoted through various supply-side measures and incentives, a program of tariff reductions and reforms, and the relaxation of exchange controls (Coetzee & Naude, 2004). Recent research into South Africa?s export performance by Edwards and Alves (2006) has, however, shown that the success of these policies has been mixed. Although the growth in exports seems impressive at first glance, it has not been enough to generate an export-led boom. South African exports remain resource-based or concentrated in products with a declining share in world markets. The policies and economic environments that promote exports can be examined at two levels. The first is an economy-wide level. For example, exchange rate depreciations show a positive relationship with export performance as these depreciations raise the profitability of export supply (Todani and Munyama, 2005). Similarly, tariff liberalization leads to improved export performance by reducing intermediate input costs and lowering the incentive to produce for the domestic market (Harding and Rattso, 2005). Edwards and Alves (2006) also find that 6 the availability of skills and infrastructure appears to be an important determinant of export growth. Skills and infrastructure are however, not only relevant at the aggregate, economy-wide level but also inseparable from the location of the exporter. The physical location of exporter is significant to the extent that the region determines the natural endowments available to and the distance faced by the exporter. Policy interventions in the form of human capital formation and infrastructure investment are also place specific. Recent research by Matthee (2007), and Matthee and Naude (2008) found that the regions in South Africa that have experienced faster export growth are those with higher GDP per capita, faster population growth, higher levels of skills, greater export diversification and shorter distances to ports. 2.3. Export Prices, Exchange Rates, and Volatility Among the many troubles of developing countries such as South Africa in recent years have been fluctuations in world prices of the commodities that they produce, especially mineral, oil and agricultural commodities, as well as fluctuations in the foreign exchange values of major currencies, especially the dollar, yen and euro. Some countries see the currency to which they are linked moving in one direction, while their principal export commodities move in the opposite direction. Frankel (2003) suggests a proposal, called PEP (peg the export price). The idea is most relevant for a country such as South Africa that is specialized in the production and export of a particular mineral commodity. The proposal is to commit to a monetary policy that fixes the local-currency to the price of the export commodity. It is not an attempt to stabilize the dollar price of the commodity because that would be futile, given the fact that South Africa is too small to affect the metal prices on world markets. This is implemented by the exporter?s central bank 7 announcing daily, an exchange rate against the dollar that varies perfectly with the daily dollar price of the metal in question on world markets, and to intervene to defend that exchange rate. That technique would be equivalent to fixing the price of the commodity in terms of local currency. The international price of any commodity should be determined by market forces including the exchange rate, industrial structure etc. This introduces a problem into trade between two countries when left to the forces of demand and supply. Fluctuation in the exchange rate introduces uncertainty (volatility) which could have a detrimental effect on trade flows. Volatility represents the extent to which a variable changes over time. The larger the magnitude of a variable change, or the more quickly it changes over time, the more volatile it is. Numerous research has been done on exchange rate volatility, and its effect on international trade, specifically relative prices and trade volumes. Majority of the researchers have confidently agreed to an inverse relationship between increased volatility in exchange rates and trade volumes, with the magnitude of the effect being small. A good number of research mostly using data from industrialized economies have posted different results ranging from a positive ERV ? trade relationship, no relationship at all, to ambiguous findings. Several reasons may explain the lack of a concrete consensus on this issue. First, even for risk-averse firms, the availability of hedging techniques makes it possible for traders to avoid most of the risk at insignificant costs (Cote, 1994). Secondly, ERV may actually offset some other forms of business risk, while creating profitable trading and investment opportunities (Arize, 1997) and (de Grauwe, 1988). In addition; an increase in risk resulting from ERV, does not necessarily lead to a reduction in the risky activity (Maskus, 1986). Lastrapes and Koray (1990) find that volatility has only a small effect on bilateral international 8 trade flows suggesting that the choice of exchange rate system on trade flows may be insignificant. Other studies such as Onafowora and Owoye (2011), Bleany and Greenaway (2001) and Bah and Amusa (2003) do show a negative relationship between exchange rate volatility and foreign direct investment. De Grauwe (1988) on the other hand finds a positive relationship arguing that in risky environment traders may trade more in order to avoid substantial decline in their revenue. Clarida and Gali (1994) identify the sources of ERV to be monetary shocks to money supply, and demand for real money balances. They find that demand shocks explain most of the variance in real exchange rate fluctuations, whereas supply shocks explain very little. In a similar study to this, Ekanayake and Thaver (2011) find that the long-run ERV has a negative and significant effect on the US exports to South Africa. Lastrapes and Koray (1990) used the moving average representation for ERV in U.S. multilateral trade and found extremely small quantitative effects on trade. Tenreyro (2007) finds no significant impact on trade, caused by nominal exchange rate variability. This is in contradiction to Kim and Lee (2007)?s study on Korea?s data; they find statistically significant impact on real exports? volume and prices, caused by fluctuations in nominal exchange rates. They further explain that the magnitude of the effect is stronger for volumes than quantities since Korean exporters prefer pricing to maintain market share rather than adjust export prices to reflect exchange rate changes. Betts and Devereux (2000) did an empirical investigation between price and exchange rate flexibility. The results showed that Pricing to Market (PTM2) increases exchange rate 2 Pricing-to-market (PTM) behavior implies that exporters adjust their prices to the prevailing prices in their export markets. For the importing country, PTM effects can be interpreted as a measure of the stability of domestic prices against foreign price and exchange rate developments. PTM behavior can be attributed to the level of competitiveness and price stickiness in the importing country. 9 volatility relative to one set price. Weliwita, Ekanayake and Hiroshi (1999), Fountas and Bredin (2006), Caporale and Doroodian (2002), Kenen and Rodrick (1986) and Dell?Ariccia (1999) and Pozo (1992) found that short - term ERV depresses trade. In addition; their results conclude that volatility has not diminished even after markets have gained experience with floating exchange rates. Choudhry (2005) and Chou (1999) suggests that ERV (from Chinese data) has especially large negative effects on manufactured exports than raw exports, whereas Kroner and Lastrapes (2003) determine that the magnitude of the effect is stronger for export prices than quantities. Arize (1997) uses data from eight Latin America countries with results showing very significant negative impacts on export demand caused by ERV both in the short and long run, with effects resulting in reallocation of resources by markets. Onafowora and Owoye (2011), Bleany and Greenaway (2001), and Bah and Amusa (2003) finds that real exchange rate instability for primary product exporters such as sub-Saharan African countries depresses investment in those sectors rather than export growth. Mckenzie (1999) finds that ERV significantly differ between traded goods sectors, thus the need to disaggregate trade data, while Hau (2002) found strong evidence that economic openness reduces ERV thereby reducing the resulting depressing effects on trade. In addition; Baccetta and Wincoop (2000) found no relationship between trade levels and welfare across exchange rate regimes. In unusual findings, Klein (2002) uses U.S. sectoral export data to seven major industrial economies. Results provide evidence that risk ? neutral firms increase supply of elastically demanded exports in response to an increase in ERV, thereby posting a significant positive relationship between increased ERV and export volumes. These are similar findings to 10 McKenzie and Brooks (1997), which used Germany ? U.S. trade flows. Rey (2006) found similar results for Israel and Morocco?s export volumes to the E.U., whereas Algeria, Egypt, Tunisia, and Turkey posted negative impacts on export volumes to the E.U. viz a viz ERV. Schnabl (2008) studied growth in the same countries (in addition to others in the EMU periphery) and found a strong negative relationship between their export growth and ERV viz the euro. In a subsequent study, Schnabl (2009) argues that real exchange rate stability reduces transaction costs for international trade, causes less uncertainty for international capital flows, and enhances macroeconomic stability. In addition, Schnabl (2009) isolates macroeconomic instability as the adverse cause of negative growth in emerging economies. Sercu and Vanhulle (1992), and Broll and Eckwert (1999) similarly conclude that increased ERV positively affects the value of exporting firms, which makes an exporting strategy more attractive relative to direct investment. Sauer and Bohara (2001) identified LDC exports from Africa and Latin America as being more sensitive to exchange rate uncertainties than those from Asian LDCs industrialized countries, while Vergil (2004) employs the less common measure of ERV: the standard deviation of the percentage change in the real exchange rate, and still found the expected results of depressed trade on Turkey?s export demand to U.S. and the E.U. In addition; Devereux and Lane (2003) explain that developing countries? bilateral exchange rate volatility (relative to creditor countries) is strongly negatively affected external debt. Exchange rate volatility also negatively affects U.S.?s foreign direct investment (FDI) (Campa, 1993), especially for firms bearing high sunk investments, for example, Japanese auto makers. Serven (2006) finds similar results on LDCs, especially in small open economies with less developed financial systems. 11 Sercu and Vanhulle (1992) and Daly (1998) found ERV to have ambiguous effects on trade volumes, using Japan?s bilateral trade flows, while Belke and Setzer (2003) find that ERV lowers employment growth in European markets and suggests that elimination of ERV could be a viable substitute for a removal of employment protection legislation. Belke and Kaas (2005) find this effect to be stronger in the E.U. than U.S. Hayakawa and Kimura (2009) using East Asian data determine that intermediate goods trade in international production is more sensitive to ERV compared to other types of trade. They additionally find ERV to have greater impacts than tariffs and smaller impact than distance- related costs. Kulatilaka and Kogut (1994) suggest that hysteresis (lagging effect) in the export prices caused by ERV is responsible for the persistence in U.S. current account deficit. In contrast, Guti?rrez (1992) finds ERV to have no significant effect on trade volumes and prices on U.S. - Canada bilateral trade. DeVita and Abbott (2004) uses U.K. ? E.U. bilateral trade data to support Guti?rrez (1992) findings, while Aristotelous (2001) uses U.K. ? U.S. bilateral trade data. They argue U.K. exports are unaffected by short-term ERV and relative price, but are largely income elastic. Long-term measure of ERV however yields negative and significant impact on trade volumes. Choudhry (2008) finds strongly positive impacts of exchange rate volatility on real imports using U.K. import data from Canada, Japan, and New Zealand, while Jung (2008) finds an interesting negative relationship between ERV and unemployment in Germany?s post- unification era. In addition; Canzoneri and Diba (2002) determine that higher currency substitution actually reduces ERV within the European Monetary Union (EMU). 12 2.4. South Africa?s Economic Sanctions From 1948 to 1994, the Nationalist party governed South Africa and enacted the apartheid system of laws. The system faced growing international criticism prompting some countries to restrict trade with South Africa. Restrictions on overseas investments in South Africa were first enacted by Sweden in 1979 (Table 1.6). This was followed by a sequence of measures against the importation of specific South African goods and services. In October 1985, the United States imposed a ban on the importation of some minerals. In the same month six Nordic countries imposed a ban in trade of almost all goods. Denmark specifically imposed a total ban of any form of trade with South Africa. In 1986, trade sanctions against South Africa reached a peak when the Commonwealth Group of Nations, the European Community, and the United States imposed measures that would reduce their imports from South Africa. As a result of these measures, and in response to adverse public opinion, many leading multinational firms reduced or completely sold off their investments in South Africa. Multinationals without their former South African subsidiaries ceased or reduced sourcing parts, components and raw materials from South Arica. Table 1.6 provides a chronological summary of sanctions against South African exports and foreign investments into South Africa. These events had significant adverse effects in South Africa export volumes, warranting the testing of a structural break in the time series analysis. 3. Model Specification The objective of this chapter is to examine the effects of exchange rate volatility on disaggregated South Africa?s mineral exports for the period 1980:01 through 2011:07, in monthly data series. Drawing on the existing empirical literature in this area, the study specifies that a standard long-run export demand function for commodity i may take the following form, 13 as suggested in Weliwita, Ekanayake and Hiroshi (1999), Fountas and Bredin (2006), Caporale and Doroodian (2002), Kenen and Rodrick (1986), Dell?Ariccia (1999) and Pozo(1992): ln Xit = ?0 + ?1lnYt + ?2lnPit + ?3lnRERt + ?4lnERVt + ?t (1.1), where Xit is real export volume in tons of commodity i in period t. The commodities studied in this paper are coal, iron ore, chromium, copper, manganese ore, PGMs, nickel, gold, granite, and limestone. Yt is the real world GDP in period t, Pit is the relative price of exports of commodity i in period t, RERt is the real exchange rate between the U.S. dollar and the South African rand, ERVt is a measure of exchange rate volatility, and ?t is a white-noise disturbance term. Economic theory suggests that the real income level of the domestic country?s trading partners would have a positive effect on the demand for its exports. Therefore, it is expected that ?1 > 0. If the relative price of exports rise (fall), domestic goods become less (more) competitive than foreign goods, causing the demand for exports to fall (rise). Therefore, one would expect that ?2, which measures the competitiveness of South Africa?s exports relative to world production, to be negative. Similarly, if a real depreciation of the rand, reflected by a decrease in the RER, is to increase export earnings of commodity i, a negative coefficient estimate for ?3 is expected. Consequently, a real depreciation of the rand, reflected by a decrease in the RER will at the same time imply that the import demand for commodity i is elastic. If, however, the import demand for commodity i were inelastic, it is expected that ?3 will be positive. The last explanatory variable is a measure of exchange rate volatility. Various measures of real ERV have been proposed in the literature. Some of these measures include (1) the averages of absolute changes, (2) the standard deviations of the series, (3) the deviations from the trend, (4) the squared residuals from the ARIMA or ARCH or GARCH processes, and (5) the moving sample standard deviation of the growth rate of the real exchange rate. Since the effects of ERV on 14 exports have been found to be empirically and theoretically ambiguous (Ekanayake and Thaver, 2011), ?4 could be either positive or negative. Following Ekanayake and Thaver (2011), the real effective exchange rate, RERt is constructed as: RERt = E PSA/PUS where RER is the real effective exchange rate, E is the bilateral nominal exchange rate of rand per U.S. dollar at time t, PSA is the consumer price index (2005=100) of South Africa at time t, and PUS is the consumer price index (2005=100) of the U.S. at time t. Exchange rate volatility (ERV) is obtained from the squared residuals from the GARCH process which takes the following form: ?lnRERt = ?0 + ?1lnREERt-1 + ?t where ?t ~ N (0, ?t2) (1.2) ?t2 = ?0 + ?t-12 + ?t (1.3) The estimated conditional variance (?t2) is used as the measure for ERV. Equations such as (1.1), where variables enter at their level and there is no lagged variable, are usually referred to as long-run relationships. Any estimate obtained for ?s are long- run estimates. In obtaining these long-run estimates, recent developments in time series analysis require incorporating the short-run adjustment process into the estimation procedure and making sure that when the adjustment takes place, the equilibrating error term (?t) decreases over time. The procedure to account for short-run dynamics is one of expressing (1.1) in an error-correction modeling format. The Engle-Granger (1987) error-correction representation theorem requires Equation (1.1) to be expressed as; 15 ?ln Xt = ?0 + ?i?ln Xt-i + ?i?lnYt-i + ?i?lnPt-i + ?i?lnRERt-i + ?i?lnERVt ?i + ?0lnXt-i + ?1lnPt-i + ?2lnYt-i + ?3lnRERt-i + ?4lnERVt-i + ?ECM (1.4), where ? is the difference operator and the other variables are as defined earlier. Ekanayake and Thaver (2011) use bounds testing approach to cointegration, which is based on two procedural steps. The first step involves using an F-test or Wald test to test for joint significance of the no cointegration hypothesis H0: ?0 = ?1 = ?2 = ?3 = ?4 = 0 against an alternative hypothesis of cointegration, H1: ?0 ? 0 or ?1 ? 0 or ?2 ? 0 or ?3 ? 0 or ?4 ? 0. This test is performed using Equation (1.4). The advantage of this approach is that there is no need to test for unit roots, as is commonly done in cointegration analysis (although the latter has still been done in this study to compare alternatives). Ekanayake and Thaver (2011) provide two sets of critical values for a given significance level with and without time trend. One assumes that the variables are stationary at the levels or I (0), and the other assumes that the variables are stationary at the first difference or I (1). If the computed F-values exceed the upper critical bounds value, then H0 is rejected signaling cointegration among the independent variables. If the computed F-value is below the critical bounds values, then fail to reject H0. Finally, if the computed F-statistic falls within the boundary, the result is inconclusive. After establishing cointegration, the second step involves estimation of the long-term elasticities and the error- correction model. Akaike Information Criteria (AIC) and the Schwartz Bayesian Criterion (SBC) for model selection are used to aid in model selection. Without a lagged error-correction term, Equation (1.1) is just a vector autoregressive (VAR) specification that is usually used to test Granger causality, a short-run concept. The addition of ?t-1 is designed to test whether, in the long run, the equilibrating error shrinks. If it does, the estimate of ? must be negative and significant. Note that a negative and significant ? 16 will also indicate that the dependent and independent variables in (1.1) are converging or, alternatively, they are cointegrated. The only requirement is that all variables must be non- stationary in levels or stationary when first differenced. The short-run effects of exchange rate volatility on exports is inferred by the sign, size and significance of estimated ?i , and its long- run effects by the estimate of ?4 that is normalized on ?0. Derived coefficients are calculated by multiplying the error correction coefficients in Equation (1.4) by each of the levels coefficients in Equation (1.1) respectively. The reported t- statistics are derived through error propagation procedure. This chapter estimates the short-run estimates of ERV effects on export volumes of all metals combined, combined volume without gold and diamonds, export volumes of gold, and those of diamonds. Only the long-run effects of the rest of metals are estimated. 3.1. Data Sources and Variables Equation 1 uses monthly time series data for the period 1980:01 to 2011:07. Data for mineral export volumes is obtained from the Central Bureau of Statistics of South Africa. These series are updated monthly and are available online at www.statssa.co.za and World Trade Organization (WTO)?s Commodity Trade Statistics Database (COMTRADE). Data for corresponding prices is available from IMF commodity index reports, also available online at http://www.imf.org/external/np/res/commod/index.aspx . Annual real GDP for South Africa and World are available at the World Bank?s portal and also at the Foreign Trade Division of the U.S. Census Bureau, while monthly series for South Africa and U.S. are available at South Africa?s Central Bank (SAB) and the Federal Reserve Bank of St. Louis (FRED II) respectively. Annual series for consumer and producer price indices are available at the IMF, while the monthly series for South Africa and U.S. are available at SAB and FRED II respectively. Data 17 for the nominal exchange rates viz a viz South Africa is available at SAB, Pacific Exchange Rate Service, Main Economic Indicators published by the OECD, and International Monetary Fund?s International Financial Statistics. 3.2. Data Plots Figures (1.1) through (1.12) show data plots for export volumes of the eleven metals and their respective export prices from 1980 through 2011. Export volume of coal (Figure 1.1) rises steadily through the years, registering only slight dips in 1983 through 1986, possibly because of economic sanctions. Corresponding coal prices show a sluggish trend but over the years, but start rising from 1983 through 2011. Export volume for iron ore and gold (Figure 1.2 and 1.8) shows a similar trend as coal, but the prices starts high in 1980, dropping to the lowest level in 1986 to 1989. The prices remain sluggish through 2002, then rises steadily through 2011. A good explanation for this would be increasing demand for iron ore as a raw material for the construction boom in India and China, and oil-induced construction boom in sub-Saharan Africa. Chromium, copper, manganese ore, Nickel and limestone (Figures 1.3, 1.4, 1.5, 1.7 and 1.10 respectively) show consistently rising export volume over the years, with sticky prices from over the years under consideration. The increase in PGM export volume (Figure 1.6) has not been as remarkable as other metals, largely due to their rarity. Their prices also show a falling trend from 1980 through early 2000, then rises slowly over the last decade to its highest in 2010. Granite (Figure 1.9) show export volume rising steadily, reaching a peak in 2003, and then falling to the lowest level in 2011. Corresponding prices show a similar trend. Falling export 18 volumes for granite maybe due to new building and construction technologies using recycled material, or due to diminishing granite resources within South Africa. 4. Empirical Results 4.1. Stationarity Analysis Variables in a time series regression should be stationary, converging to a dynamic equilibrium, or the standard errors would be understated (Enders, 1995). Therefore, prior to estimating the model, the study tests each series for a unit root using the Dickey-Fuller (Dickey and Fuller 1981) and the Phillips and Peron (1988) unit root tests. These tests verify that each series that enters the model is stationary. The analysis also checks for the presence of auto correlation and normality of the error terms. Durbin Watson statistics show that the models? error terms do not suffer from auto correlation. Inspection of the residual plots for several series concludes that the errors are normally distributed. Then, Shapiro-Wilk W-test, which is the ratio of the best estimator of the variance to the usual corrected sum of squares estimator of the variance (Shapiro and Wilk 1965) is applied. The test results confirm the visual inspection of the residual plots that the errors are normally distributed. Finally, to avoid problems that may arise from heteroscedasticity, the study uses ARCH (1) tests and reports robust standard errors. Table (1.1) provides a summary of stationarity analysis. All variables are difference stationary with white noise residuals (Table 1.1). All residuals are checked for white noise with zero means, low auto correlation by Durbin Watson statistics (DW>1.26 for lack of positive autocorrelation and DW>2.74 for lack of negative auto correlation), and homoskedasticity by ARCH (1) tests. 19 Chromium export volumes and export prices for coal, nickel, PGM, and gold are difference stationary by Dickey Fuller tests with no constant (DF), while chromium?s export price is difference stationary by Dickey Fuller tests with a constant (DFc). The series for real effective exchange rate (RER) is difference stationary by Dickey Fuller tests with a time trend (DFt). The rest of the series are difference stationary by Peron test. 4.2. Model Estimation Regressions in levels produce spurious results but variables are cointegrated by Engle-Granger EG tests. These results are reported in Table (1.2). ECM results are reported in Table (1.3). Relative prices for copper, nickel and gold bear positive and significant coefficients, while ERV yields negative and significant coefficient for coal, and a positive and significant coefficient for copper. The insignificant difference coefficients for the rest of the variables in Table (1.3) imply no transitory effects but the significant error correction terms imply adjustment relative to the dynamic equilibrium. Effects of exogenous variables on S.A.?s export of metals are reported in Table (1.4). Coefficients are derived by multiplying the error correction coefficients in Table (1.3) by each of the levels coefficients in Table (1.2). The reported standard errors are derived through error propagation calculation: ?? = ?((??/?) 2 + (??/?) 2).5, where, if ? = ? ? ? ? ?? = (??2 + ??2).5, and if ? = ?? or ? = ?/? ? ?? = ?((??/?) 2 + (??/?) 2).5. In Tables (1.2, 1.3, and 1.4), coefficient estimates are reported with standard errors in the parenthesis, and the corresponding t-statistics. For all metal exports combined, rising world income positively affects exports with an elasticity of 0.61, while rising export prices depress trade with an elasticity of -0.11. Appreciation of the rand relative to the U.S. dollar reduces export volumes with an elasticity of - 0.04, while exchange rate volatility depresses exports. The export volumes for all metals 20 combined without gold are adversely affected by rising world prices and rand appreciation, and exchange rate volatility, with elasticities of -0.1, -0.04, and -27.68. An increase in world income increases the export volumes for iron, copper, manganese, gold, and limestone, with elasticities of 0.61, 1.09, 0.88, 1.54, 1.75, and 0.5 respectively, consistent with Lastrapes and Koray (1990), and Ekanayake and Thaver (2011). The rest of the metals yield insignificant estimates, or estimates with wrong signs. An increase in export prices reduces export volumes for chromium, copper, manganese, nickel, and gold, with elasticities of - 0.46, -0.64, -0.64, and -0.45 respectively. The rest of the metals yield insignificant estimates, or estimates with wrong signs. Appreciation of the rand, caused by an increase in RER, reduces export volumes for coal, iron, copper, nickel, gold, and granite, with elasticities of -0.04, -0.04, - 0.01, -0.58, -0.43, -0.7, and -0.66 respectively. Iron, copper, manganese, PGMs, and limestone are not sensitive to currency appreciation since all yield insignificant estimates. All metals except chromium, PGMs, and limestone are sensitive to fluctuations in exchange rates. Coal, iron, PGMs, nickel, gold and granite yield significant negative estimates for ERV consistent with Ekanayake and Thaver (2011), Vergil (2004), Shnabl (2008), Weliwita, Ekanayake and Hiroshi (1999), Fountas and Bredin (2006), Caporale and Doroodian (2002), Kenen and Rodrick (1986), Dell?Ariccia (1999), Pozo (1992), Choudhry (2005), Chou (1999) and Arize, Osang, and Slottje (2006), meaning that their export volumes are depressed by fluctuations in exchange rates. ERV on the contrary improves the export volumes for Copper and manganese, since they yield positive estimates, consistent with Kim and Lee (2007), Klein (2002), Choudhry (2008), Sercu and Vanhulle (1992), and Broll and Eckwert (1999). Table (1.5) reports results for the short-run estimates (similar to Edwards and Lawrence (2006), and Ekanayake and Thaver (2011)), alongside the long-run estimates for all metal export 21 volumes combined, gold, and diamond. The effects of trade sanctions are also reported. Results in levels are spurious, and the ECM models produce insignificant estimates. For the derived effects, aggregate export volumes with and without gold decline with an increase in export prices, with elasticities of -0.39 and -0.06 respectively, while gold and diamond exports also decline with elasticities of -0.02, and -0.52 respectively. Rand appreciation in the short-run also depresses exports of all metals with an elasticity of -0.33, and reduces gold and diamond exports with an elasticity of -0.27 and -0.08. Exchange rate volatility in the short-run also depresses trade volumes in all metals combined, all without gold, diamond and gold individually, with elasticities of -0.33, -0.17, -0.27, and -0.08 respectively. Ekanayake and Thaver (2011) found similar results in mining sector, where data was an aggregate of the entire sector. Economic sanctions imposed on South Africa prior to 1994 by majority of its trading partners are found to depress export volumes of all metals combined, with and without gold, with elasticities of -0.04, and -0.02, consistent with Hufbauer, Elliott and Schott (2002) . Gold and diamond are not sensitive to those restrictions. Plausible explanation would be that most of the trade partners that imposed restrictions, allowed unrestricted trade in gold and diamonds due to their high value, but restricted trade in the rest of the metals. Although the above results are generally consistent with listed previous studies, direct comparison would not be appropriate, owing to various differences in the models. First, to generate a proxy of exchange rate volatility, one can pursue different methodologies. One of the most commonly employed methods to proxy for exchange rate volatility is the moving standard deviation of exchange rate changes. This methodology contains substantial correlation. Vergil (2004), Shnabl (2008), Caporale and Doroodian (2002), Kenen and Rodrick (1986) use this proxy in their models on studies predicting ERV?s effects on US exports. This chapter uses 22 squared residuals from the GARCH process as an alternative, similar to Ekanayake and Thaver (2011) and Edwards and Lawrence (2006), which use cointergration analysis in their studies. Alves and Edwards (2006) and Edwards and Golub (2004) use the GARCH process on South Africa?s non-gold merchandise exports using panel data for 28 manufacturing sectors. They obtain fixed effects and General Method of Moments (GMM) estimators, while Tsikata (1999) estimates ERV in both short and long-run in a reduced form export function OLS and 2SLS models. Bhorat (2008) estimates a similar model for South Africa?s paper and paper products export. Secondly, Dell?Ariccia (1999), Pozo (1992), Broll and Eckwert (1999). Estimate extended models that include tariffs, capacity utilization and infrastructure among other variables. Additional variables to a model lead to a loss in the degrees of freedom. Thirdly, difference in data frequency and period of analysis on similar studies will yield different results. Except for Kim and Lee (2007), and Klein (2002) that use monthly series on US sectoral exports, the rest of the studies use either annual or quarterly with differing period of analysis. Lastly, most of past studies in this area use aggregate data that leads to aggregation bias. This chapter adopted Ekanayake and Thaver (2011), Onafowora and Owoye (2011), Bleany and Greenaway (2001), and Bah and Amusa (2003). These studies use disaggregated trade data that looks at the impact of exchange rate volatility on disaggregated sectoral data for South Africa?s exports. This chapter goes further by considering individual commodities within a specific sector, which provides specific commodity attributes. 23 5. Summary and Conclusion This chapter examines the relationship between South Africa?s metal exports and the fluctuations in exchange rates among other macroeconomic variables, with monthly time series data for the period 1980:01 to 2011:07. Both short-run and long-run estimates are examined, as well as effects of trade sanctions on export volumes. Besides providing important commodity attributes lacking in previous studies, cointegration results clearly show that there exist both long and short-run equilibrium relationships between real exports and real foreign economic activity, relative prices, real exchange rate, and real exchange rate volatility in the eleven commodities. All the specifications yielded expected signs for the coefficients. Most of the coefficients in all the models considered are statistically significant. In the long-run, importer?s income is found to be important, having a positive impact on export volumes of five of the eleven metals. Export price and real effective exchange rates are also critical in determining export volumes, while ERV is perhaps the most important factor. All metals except chromium and limestone are found to be highly sensitive to exchange rate fluctuations. In the short-run, similar results are obtained for gold and diamond. Trade sanctions imposed on South Africa prior to 1994; appear to have depressed exports for all other metals except for gold and diamond, possibly due to gold and diamond?s high value and rarity. 24 Tables Table 1.1. Stationarity Analysis for Commodities Y-VARIABLE DF -1.95 1.26 for lack of positive autocorrelation and DW < 2.74 for lack of negative autocorrelation), and homoskedasticity by ARCH(1) tests. The rest of the variables are nonstationary. All variables are difference stationary by Dickey-Fuller DF tests (Table 3.1). The real exchange rate e is difference stationary with the Dickey Fuller test with a constant DFc. Inflation (?) and world gold prices (Gp) are stationary with the Dickey Fuller test without a constant DF. US disposable income Y is difference stationary with Augmented Dickey Fuller test ADF, and money supply Ms is stationary with the Dickey Fuller test with a constant DFc. 4.2. Model Estimation Regression in levels produces spurious results and variables are cointegrated by an Engle- Granger EG test. This regression is reported in Table (3.2), 130 ? = ?0 + ?1Ms + ?2e + ?3Y + ?4Gp + ?5STR + ?6MR1 + ?7MR2 + ?8MR3 + ?? (3.10) The residual ?? from the spurious model is stationary by the Engle-Granger EG test, satisfying the critical t-statistic -3.18. Analysis proceeds with an error correction model ECM. The residual ?? from the spurious model is included in the ECM ?? = ?0 + ?1?Ms + ?2?e + ?3?Y + ?4?Gp + ?5STR + ?6MR1 + ?7MR2 + ?8MR3 + ???? + ?ECM. (3.11) Regressions in levels produce spurious results but variables are cointegrated by Engle- Granger EG tests. These results are reported in Table (3.2). ECM results are reported in Table (3.3). Only gold prices and the error correction term yield significant estimates. The insignificant difference coefficients for the rest of the variables in Table (3.3) imply no transitory effects but the significant error correction terms imply adjustment relative to the dynamic equilibrium. Effects of exogenous variables on S.A.?s rate of inflation are reported in Table (3.4). Coefficients are derived by multiplying the error correction coefficients in Table (1.3) by each of the levels coefficients in Table (1.2). The reported standard errors are derived through error propagation calculation: ?? = ?((??/?) 2 + (??/?) 2).5, where, if ? = ? ? ? ? ?? = (??2 + ??2).5, and if ? = ?? or ? = ?/? ? ?? = ?((??/?) 2 + (??/?) 2).5. In Tables (1.2, 1.3, and 1.4), coefficient estimates are reported with their corresponding standard errors. An increase in the Money supply increases the rate of inflation, with an elasticity of 0.64 consistent with Gordon (1981). The -0.78 effective real exchange rate e elasticity is evidence that appreciation/depreciation of the Rand has a negative/positive effect on the domestic price level. Higher global gold prices have a positive effect on the rate of inflation with an elasticity of 0.54 131 reflecting the importance of gold as South Africa?s major export. Khamfula (2004) tested the effects of various metal prices on inflation, obtaining a similar result. Rising world income also raises inflation (with an elasticity of 0.54) level by raising the aggregate demand for gold and other exports. The lifting of trade sanctions by trade partners is also reported to increase the rate of inflation with an elasticity of 0.12. Export demand rose exponentially in 2004, surpassing aggregate supply as shown in Figure (3.1), thereby raising the general level. The first and second monetary regimes yield insignificant estimates, but the third one positively affects inflation levels with an elasticity of 0.19. This is the period from 1998 when SARB introduced a requirement that the repurchase interest rate be determined at an auction and earlier imposed direct controls were removed. This is inconsistent with Aaron (2004) since he found no change in interest rates across the three monetary regimes. 5. Summary and Conclusion The analysis concludes that money supply, the real effective exchange rate, gold prices, world income, and lifting of trade sanctions is important in determining the rate of inflation in South Africa. Elimination of direct controls by SARB in determining domestic interest rates is also found to be important. The data plots in Figure (3.4) show a general declining trend in South Africa?s inflation levels. This could have been as a result of the Growth, Employment and Redistribution (GEAR), a policy developed by South Africa?s government that included Inflation targeting. That is a monetary policy in which a central bank attempts to keep inflation in a declared target range, typically by adjusting interest rates. 132 According to the South Africa Reserve Bank (2000), adjusting interest rates will raise or lower inflation through the adjustment in money supply, because interest rates and money supply have an inverse relationship. SARB also publicly declares the forecasted interest rates such that if inflation appears to be above the target, SARB would raise interest rates and vice versa. 133 Tables Table 3.1. Stationarity Analysis AR(1) Coef+2(se)<1 DF -1.95