Exchange rate volatility and agricultural trade under policy reforms in South Africa

Abstract

Decades of government intervention have helped develop the South African agriculture sector to its present state. Policy reforms have included trade and exchange rate policies to increase the country's international competitiveness, reduce poverty and promote economic growth. These reforms are facilitating the growth in agricultural trade and South Africa's reintegration into the global economy. Annual agricultural exports and imports have increased. This paper uses annual data and a vector error-correction model to investigate the supply and demand relationships for agricultural trade flows in South Africa during the past four decades. The results show that prices, real exchange rates, domestic production capacity and real incomes have significant impacts on the country's agricultural trade. In particular, exchange rate volatility has negative impacts. This cannot be viewed solely as an exogenous source of macroeconomic instability in South Africa, as domestic policies play a crucial role in influencing the movement of exchange rates.

1. Introduction

Since Orcutt's (1950) pioneering work on the responses of trade flows to changes in exchange rates and prices, international trade studies have often focused on formulating and estimating the demand relationships for imports and exports in various countries (see, for example Houthakker&Magee, 1969; Bahmani-Oskooee, 1986; Tegene, 1991; Kargbo, 1995, 2005a). One aspect of international trade that has attracted significant research attention deals with the impacts of exchange rate volatility on trade. For instance, the International Monetary Fund investigated this issue in detail for the General Agreement on Tariffs and Trade and recently for the World Trade Organisation (IMF, 1984). The studies were motivated by an increase in protectionist pressures, large exchange rate fluctuations among the major currencies and significant slowdown in world trade recorded in the early 1980s and during the 2001–03 period (Clark et al., 2004). Developments in the global economy, including, for example, the liberalisation of capital flows and tremendous increase in the scale and types of international financial transactions, have exacerbated movements in exchange rates.

Very often it is hypothesised that exchange rate volatility has negative impacts on exports. Several studies, such as Asseery&Peel (1991), Koray&Lastrapes (1989) and Sauer&Bohara (2001), argue that exchange rate volatility has adverse effects on trade because it imposes costs on risk-averse entrepreneurs, who respond by favouring domestic rather than foreign trade. However, Clark et al. (2004) argue that there is no empirical evidence showing a systematic and definitive negative link between the volume of world trade and exchange rate volatility. Clark et al. noted that it is possible for huge exchange rate fluctuations to have impacts on the economy through other channels. Although the evidence so far is at best conflicting, policymakers and the general public are still concerned about the effects of huge exchange rate movements.

The government has implemented significant policy reforms in South Africa during the past decade (see for example Tsikata, 1999; NDA, 2001; Vink & Kirsten, 2003; and related papers in Niewoudt & Groenewald, 2003). The major objectives of the reform programmes include reduction of poverty and inequalities by increasing incomes and employment opportunities and building an efficient and internationally competitive agricultural sector to enhance economic growth. The liberalisation of trade and exchange rate policies is a key component of the reforms. The rand has experienced significant volatility since the implementation of exchange rate reforms. For instance, the rand depreciated on average 9–10 per cent per year against the British pound, US dollar and Japanese yen during the 1990–2002 period.

Agriculture is an important source of income for various segments of the South African population. For example, nearly 52 per cent of the 2.2 million employed people in the former homelands earned a living in agriculture, mainly in small scale and subsistence farming. Simbi&Aliber (2000) reported that as of 1998 employment in the agricultural sector accounted for 30 per cent of all employment for blacks living in rural South Africa. However, the contribution of the agriculture sector to South Africa's gross domestic product (GDP) has declined from 14–21 per cent during the 1920–60 period to about 4–5 per cent in the 1990s and early 2000s (see Jones & Müller, 1992; Stats-SA & NDA, 2000, NDA, 2004). The declining share of agriculture in the GDP is consistent with the stages of economic development countries go through as they move from an underdeveloped agrarian society to a highly industrialised economy. There are strong inter-sectoral links between agriculture and manufacturing/industrial sectors.

Significant research efforts have been devoted to understanding the effects of exchange rates on agricultural trade flows in various parts of the world (see for example, Henneberry & Kargbo, 1986; Kargbo, 1995, 2005a; Tsikata, 1999). A common characteristic of past trade studies is that supply relationships have generally been handled by assumption. Typically, the ‘small country case’ is assumed when a country's actions do not influence world prices. Thus, the import and export supply price elasticities faced by a particular country are taken to be infinite, or at least large. However, this assumption is inadequate when applied to the supply of exports of a country such as South Africa. As Goldstein&Khan (1978) argue, for the assumption of infinite supply price elasticity to hold there has to be idle capacity in the export/domestic sector; or, more generally, export production is subject to a constant or increasing returns to scale, otherwise an increase in the world demand for the country's exports will, at the very least, lead to an increase in the price of exports in the short term. Some researchers, including Bahmani-Oskooee (1986) and Tegene (1991) have investigated the effects of exchange rates on the aggregate trade flows of developing countries. Some of these studies do not include a variable that captures a country's openness or trade restrictions imposed by governments to collect revenues, conserve foreign exchange and protect domestic industries. Kargbo (2005a) and Lopez & Thomas (1990) argue that import demand models which do not include variables on relative prices, income and foreign exchange constraints are likely to yield biased estimates as a result of the omission of relevant variables or the simultaneity of import volumes and relative prices.

This paper investigates the supply and demand relationships for South Africa's agricultural exports and imports. We use a vector error correction model and annual data covering the 1960–2004 period. In particular, we examine the responses of export and import volumes to shocks on prices, real exchange rates, real incomes and domestic production capacity over specific time horizons. Thus specific variables that capture the effects of trade and exchange rate policies are included in our analysis. This method of modelling trade flows is useful from a policy perspective because it enables us to gauge the potential impacts of certain policies on variables in the system (see Sims, 1980; Tegene, 1991; Kargbo, 2005a). To further enhance our understanding of agricultural trade flows in South Africa, two models are specified in this paper: South Africa's import demand function for agricultural products, and an export supply function for South Africa's agricultural exports.

The rest of the analysis in this paper is divided as follows. Section 2 briefly discusses agricultural trade and policy reforms in South Africa and Section 3 presents the model used to analyse agricultural trade flows in South Africa. Section 4 presents the empirical results, and the conclusions are in the final section.

2. Agricultural trade and policy reforms in South Africa

This section discusses the South African agricultural sector within a changing policy environment that includes trade and exchange policies. There is a significant relationship between government policies (e.g. the broad-based Black Economic Empowerment Framework for Agriculture–AgriBEE, changes in domestic credit, etc.), exchange rates, real supply shocks, international monetary reserves and political instability and food and agricultural prices. Monetary variables (e.g. interest rates, exchange rates and money supply) are determined within domestic or international markets. Macroeconomic variables, such as trade policy instruments on imports and exports, are determined by domestic policymakers. These variables are predetermined in relation to the agriculture sector. The exchange rate determination process provides the direct link in models that seek to establish a relationship between the agricultural sector and levels of money supply in the economy. Chronic inequalities in the agricultural sector in particular, pertaining to land ownership and lack of access to land and other resources, hamper the country's efforts to realise its full agricultural potential and, if left unattended, could trigger political instability, with profound effects on the behaviour of the rand. See, for example, Kargbo (2000, 2003, 2005b) and Edwards (1989) for further evidence on the factors that determine the behaviour of exchange rates. This issue is discussed further in Section 3 of this paper.

The following section looks at some aspects of the agricultural sector that could have significant impacts on trade flows and the behaviour of exchange rates in South Africa.

2.1 State of agriculture in South Africa

Extensive government intervention has transformed the South African agriculture sector. Several studies, such as NDA (2001, 2004) and Simbi & Aliber (2000), report that the country's land laws have developed a highly skewed land ownership pattern wherein the white minority own approximately 87 per cent of the land while blacks and other racial groups own only 13 per cent. Lyne & Darroch (2003) report that 50,04 per cent of the farmland (including provincial parks) in KwaZulu-Natal was controlled by a small minority of white owners in 1994. Progress in implementing and realising the full benefits of land reform has been relatively slow during the past decade. Land reform will thus continue to be one of the most critical issues facing South African agriculture in the next decade. For further details on the land issue see for example Muendane (2005) and other papers presented at a recent summit in Johannesburg.

The impressive transformation of agriculture in South Africa has occurred almost exclusively on white-owned farms. The most glaring economic disparity between blacks and whites in South Africa is in the agriculture sector, where whites dominate the commercial farming sector with blacks being mainly in the subsistence sector. The average farming unit in white-owned farms ranged from 979 ha in 1970 to 1355 ha during the 1988–96 period. The average farm size in the subsistence sector is less than 2 ha. Furthermore, the StatsSA&NDA report (2000) shows that 39 per cent of whites engaged in agriculture are employers, compared to 95 per cent of blacks and 96 per cent of coloureds, who are employees in this sector. Overall, black employees were paid only 12 per cent of the wages received by white employees in the agriculture sector.

Increased rates of mechanisation and capital formation since the mid-1960s were fuelled with the aid of huge government subsidies (see Figure 1, panels A and B), tax concessions, and the forced resettlement of blacks in various parts of the country. There has been a significant decline in farm loans and government subsidies since the early 1990s. The adoption of labour-saving technologies has increased unemployment rates in rural South Africa. Employment rates in commercial farms decreased by 25,1 per cent over the 1988–1996 period (i.e. from 1,2 million people in 1988 down to 914 000 people in 1996). Simbi & Aliber (2000) noted that the dramatic reductions in farm employment during the past decade could be attributed to non-economic factors, such as white farmers fearing loss of their land to resident farm workers because of new legislations enacted by the post-apartheid government, and the perception that because of labour and human right laws farm workers are becoming more difficult to control than they were before the change of government in 1994.

Figure 1: Financial assistance, capital formation and farm debt (in million rands) in South Africa 1960–2003. Source: NDA data

Jones&Müller (1992: 236) argued that farmers in South Africa wielded more power over the national government than their counterparts in Europe or North America. Before the implementation of wide ranging policy reforms in the 1990s, the influence of the farm lobbies and farmers on the South African parliament was unprecedented in modern history. Jones&Müller reported, for instance, that by 1980 almost 87 per cent of all marketed agricultural products were sold under marketing arrangements with prices fixed by producer-dominated boards. The government has embarked on massive reforms of the economy with the aim of reducing poverty, enhancing economic growth and eliminating the inefficiencies that characterised large segments of the economy, including the agriculture sector. Since its publication in 2001, The Strategic Plan for South African Agriculture has been the guiding force for agricultural development in the country (NDA, 2001). The Marketing of Agricultural Products Act, which was passed in 1996, has led to a tremendous deregulation of agricultural markets in South Africa. All control boards and price controls on agricultural products were abolished by 1998.

The AgriBEE framework sets benchmarks for a broad-based black economic empowerment in agriculture that is consistent with the experiences gained from transformation efforts during the past decade (NDA, 2004). AgriBEE hopes to help the country achieve its goal of ensuring that 30 per cent of the agricultural land is owned by black South Africans by 2014. This could reduce or eliminate the possibility of political instability erupting, with disastrous effects on the exchange rate. Current events in Zimbabwe clearly demonstrate the strong links between exchange rate behaviour and political instability emanating from land ownership or lack of access to land and other resources in the economy. South Africa's political gains in recent years cannot be sustained in the long term with the current levels of inequalities in the country. Thus the adoption of the AgriBEE framework is a step in the right direction.

2.2 Trade and exchange rate policies

For decades, South Africa's agricultural trade regime was characterised by a maze of price controls, quantitative restrictions, specific duties, import and export permits and other regulations designed to stifle foreign competition. The trade and exchange rate reforms are improving the efficiency of the economy and increasing its productive capacity with rapid increases in exports. As Tsikata (1999) reported, trade reforms were initiated in the early 1980s, but were slowed by lobbying pressure from industrialists and politically powerful white farmers.The pace of reforms picked up after the change of government in 1994. The removal of international trade sanctions imposed on South Africa because of its racial policies of apartheid, coupled with policy reforms and strategic marketing efforts, have contributed to the boom in exports and imports since the mid-1990s (see Figure 2, panels A and B). South Africa's agricultural exports fluctuated between R0,3 billion and R25,46 billion per year over the 1960–2003 period, while annual agricultural imports ranged from R37,3 million to R14,94 billion during the same period. Annual agricultural exports accounted for 19,91 per cent of the country's total exports, while agricultural imports accounted for 5,03 per cent of total imports over the past four decades.

Figure 2: Agricultural trade flows in South Africa, 1960–2003 Source: FAO and NDA data

The main agricultural exports of South Africa and other Southern African Custom's Union (SACU) members have been corn (maize) and corn products, sugar, wine, grapes, citrus and deciduous fruits during the past few years (see Table 1). Maize and sugar are generally exported when there are surpluses. Henneberry&Kargbo (1986) reported that South Africa was among the world's top ten corn exporters during the 1970–83 period. According to the Abstracts of Agricultural Statistics published by the NDA (NDA, various years), maize was the most important crop in South Africa during the 1970–2003 period. For instance, it contributed 34–45 per cent to the gross value of field crops during this period. Field crops accounted for 31–51 per cent of the total value of agricultural products during this period. About 9 000 commercial farmers produce maize on nearly 3,4 million hectares and employ nearly 150 000 farm workers. The world market for sugar is highly distorted. The South African sugar market is an oligopoly with the two main players being Tongaat-Hulett Sugar Limited and Illovo Sugar Limited–accounting for 35 and 48 per cent, respectively, of the total sugar output in the country (see NAMC&DOA, 2003). Wine exports increased from R122 million in 1992 to R3,0 billion in 2002. These exports accounted for 3 and 12 per cent of total agricultural exports during this period. The Wine and Spirits Agreement between South Africa and the European Union signed in January 2002 is expected to further expand access for these products to the huge European market.

Table 1: Agricultural exports in Southern African Custom's Union, 1992–2002 (value in million rands)

Agricultural exports	1992		1995		1999		2002
	Value	%	Value	%	Value	%	Value	%
Sugar	0	0	545,687	6,70	1 470,256	9,95	2 280,789	8,81
Wine	122,112	2,51	332,895	4,09	1 222,731	8,28	3 000,144	11,59
Citrus fruit	540,382	11,12	722,183	8,86	1 866,80	12,64	2 273,005	8,78
Grapes	273,684	5,63	490,554	6,03	1 278,919	8,66	1 616,471	6,25
Preserved fruit and nuts	440,563	9,07	611,677	7,51	839,874	5,69	1 352,210	5,22
Apples, pears and quinces	595,763	12,26	791,142	9,72	887,980	6,01	1 248,177	4,82
Fruits and vegetable juices	214,516	4,41	247,414	3,04	489,917	3,32	964,741	3,73
Cigars, cheroots, cigarillos and cigarettes	28,663	0,59	107,413	1,31	391,846	2,65	522,440	2,01
Maize	264,607	5,49	566,202	6,95	498,469	3,37	1 417,841	5,48
Undenatured ethyl alcohol	70,710	1,46	282,970	3,48	371,800	2,52	957,690	3,70
Wool	311,765	6,42	298,514	3,67	382,733	2,59	676,273	2,96
Food preparations	34,181	0,70	237,393	2,91	151,152	1,02	363,131	1,40
Apricots, cherries, plums, peaches and sloes, fresh	87,843	1,81	115,169	1,41	276,004	1,87	304,760	1,18
Tobacco	52,213	1,07	51,791	0,63	224,516	1,52	334,769	1,29
Dates, pineapples, figs, mangoes, avocados and guavas	60,844	1,25	106,149	1,30	150,884	1,02	280,599	1,08
Raw skins of sheep and lambs	122,021	2,51	229,983	2,82	165,808	1,12	498,248	1,93
Sugar confectionary	50,743	1,04	57,509	0,70	128,792	0,87	182,027	0,70
Other meat and edible meat offal	5,897	0,12	21,907	0,26	101,398	0,69	341,762	1,32
Raw hides and skin of bovines	83,193	1,71	101,850	1,25	68,778	0,47	174,269	0,67
Other	1 500,483	30,87	2 224,181	27,31	3 804,886	25,76	7 005,962	27,06
Grand total	4 860,183	100,00	8 142,590	100,00	14 773,544	100,00	25 886,273	100,00
Source: Department of Agriculture (NDA), Abstract of Agricultural Statistics 2004, Pretoria, South Africa. Percentages are author's own calculations.

The major agricultural imports in recent years have been wheat, rice, whiskies, rum and other alcoholic beverages, soya bean oil cake, sunflower and cotton seed oil (see Table 2). The top ten destinations for SACU's agricultural exports since 1999 include the United Kingdom, the Netherlands, Belgium, Japan, Mozambique, the United States and Zimbabwe. The top six origins for SACU's agricultural imports are Argentina, the United States, the United Kingdom, Australia, Brazil and Spain. The liberalisation of trade and exchange rate controls is facilitating South Africa's reintegration into the global economy. This is consistent with a major objective of the post-apartheid government's strategy to transform South Africa into ‘a competitive, outward-oriented economy’.

Table 2: Agricultural imports in Southern African Custom's Union, 1992–2002 (Value in million rands)

Agricultural imports	1992		1995		1999		2002
	Value	%	Value	%	Value	%	Value	%
	Rice	302,450	6,75	496,411	7,26	864,113	9,63	1 253,927	8,27
Wheat	197,755	4,41	564,555	8,26	400,577	4,46	971,090	6,41
Whiskies, rum, gin, and other alcoholic beverages	232,817	5,20	345,267	5,05	582,932	6,49	745,536	4,91
Soya bean oil cake	47,439	1,06	61,717	0,90	332,663	3,71	944,147	6,22
Bread mixtures, jelly powders and other food preparations	58,000	1,29	173,428	2,54	294,975	3,28	475,480	3,13
Sunflower and cotton seed oil	114,984	2,57	426,268	6,23	443,809	4,94	758,428	5,00
Meat and edible offal of poultry	112,271	2,50	203,859	2,98	257,533	2,87	214,449	1,41
Tobacco	237,495	5,30	186,568	2,73	416,092	4,63	371,301	2,45
Palm oil	106,298	2,37	175,440	2,56	365,017	4,06	657,105	4,34
Cotton	140,984	3,14	290,440	4,25	239,935	2,67	633,177	4,18
Maize	1 177,144	26,29	412,661	6,04	245,035	2,73	1 177,243	7,77
Preparations used in animal feeding	20,792	0,46	79,030	1,15	216,491	2,41	227,757	1,50
Coffee	46,177	1,03	181,206	2,65	196,385	2,20	192,133	1,27
Hides and skins of bovine	3,511	0,07	55,115	0,80	119,998	1,34	144,102	0,95
Barley	40,087	0,89	0,500	0,01	65,145	0,72	317,380	2,09
Malt	102,221	2,28	140,195	2,05	164,429	1,83	251,584	1,66
Milk and cream	39,905	0,89	28,515	0,42	46,364	0,51	159,847	1,05
Meat of sheep or goats	62,720	1,40	35,803	0,52	103,019	1,15	64,980	0,42
Peptones, other protein substances and hide powder	35,812	0,80	54,548	0,79	133,851	1,49	173,817	1,15
Dried leguminous vegetables	144,232	3,22	165,150	2,42	142,257	1,59	239,913	1,58
Other	1 254,826	28,02	2 758,129	40,36	3 365,952	37,50	5 242,410	34,59
Grand total	4 477,918	100,00	6 834,325	100,00	8 976,550	100,00	15 155,805	100,00
Source: Department of Agriculture (NDA), Abstract of Agricultural Statistics 2004, Pretoria, South Africa. Percentages are author's own calculations.

It is widely accepted that developing countries are facing intense, and in some cases unfair, trade competition from the industrialised nations, in particular the United States and the European Union. For example, the OECD (Organisation for Economic Cooperation and Development) countries spent approximately US$360 billion on agricultural subsidies in 2001. Finance & Development (2004) reported that farm subsidies in industrialised countries were still high two years later, about US$235 billion in 2003. In addition to the domestic tariffs which limit access by developing countries to markets in the OECD countries, the subsidies depress commodity prices and farm incomes in developing countries, thereby increasing poverty. Krueger (2004) reports that production subsidies on cotton in the United States are approximately US$3,7 billion per year. This amount is about three times the annual aid given to Africa by the United States. The EU cotton subsidies depressed the world price by 12 cents per pound in 2001. The price declines in recent years mean West African economies and their cotton farmers lose nearly US$250 million per year. There is a growing demand for the agricultural subsidies to be reduced and eventually eliminated. The recent collapse of the World Trade Organisation talks in Cancun, Mexico, is attributed in part to the failure by delegates to satisfactorily address the thorny agriculture issues facing developing countries in the world market.

Despite the intense competition in the world market, it appears that South Africa has increased its competitiveness in recent years. For instance, an improvement in competitiveness is manifested through growth of export volume, assuming all other things are constant. An examination of the real exchange rate index suggests that South Africa appears to have made significant gains in competitiveness in recent years (see Figure 3). An increase in the index implies real depreciation of the domestic currency. Despite the high volatility of exchange rates, recent empirical evidence in South Africa and other countries shows that exchange rates display parity reverting behaviour after-shocks. Kargbo (2003, 2004, 2006) reports that 17–53 per cent of the deviation from purchasing power parity is corrected within two years in South Africa during the 1960–2002 period. Following the work of Edwards (1989), we use the formula below to construct the real exchange rate index reported in Figure 3:

where: MRER_t is the multilateral real exchange rate index in year t for South Africa; E_jt = index of the official nominal exchange rate between country j and South Africa in year t; α_j is the weight of partner j used in computing MRER;  = consumer price index (CPI) of partner j in year t; P_t is the CPI of South Africa; and j = 1, 2, …, n refers to partner countries included in constructing the MRER index. The author used the top 13 trading partners for South Africa, and 1994–96 total trade weights obtained from Fajgenbaum et al. (1998: 119–31) [the total trade weights (in parentheses) for the trading partners used in computing the real exchange rate index are: United States (11,53), United Kingdom (11,13), France (3,18), Japan (9,64), Belgium (3,28), Switzerland (2,98), Germany (12,52), Austria (2,19), Iran (2,09), Italy (5,17), South Korea (3,08), the Netherlands (2,78) and Zimbabwe (3,38)]. Figure 3 also presents an indicator for openness of the South African economy over the 1960–2003 period. Openness (calculated as the combined proportion of imports and exports of goods and services in South Africa's GDP) has averaged about 52 per cent per year over this period. Such levels of openness make the country highly vulnerable to external shocks. Smal (1996) argues that the adoption of an outward-oriented development strategy is the key to sustainable growth in South Africa. The implementation of appropriate exchange rate policies, coupled with complementary monetary and fiscal policies, can lead to stimulation of economic growth, creation of more jobs and increases in income levels. Without these policies, depreciations of the rand could lead to high domestic price increases and inflation that can readily develop into a vicious cycle of depreciations and price increases. Such a condition will eventually lower economic growth, employment and South Africa's international competitiveness.

Figure 3: Real exchange rates and openness of the South African economy, 1960–2003 Source: IMF data

3. Model specification

This section considers two models of agricultural trade flows for South Africa, an agricultural export supply model and an import demand function for agricultural products in South Africa.

The supply of agricultural exports is specified as: , CAPOR_t, WGNP_t, MRERV_t, GODUM, OPEN_t, Z_t); where:  = index of the quantity of agricultural exports supplied by South Africa (1989–91 = 100); PX_t = price index (unit value) of agricultural exports, 1989–91 = 100; WGNP_t = weighted average of real incomes of industrialised countries, including many that are South Africa's major trading partners; and PXW_t = weighted average of export prices of South Africa's trading partners. This last variable reflects competition among various suppliers to South Africa, rather than competition between South Africa's imports and its domestic output (Houthakker & Magee, 1969: 112). OPEN is an indicator for trade policy restrictions or openness of the South African economy. It has been discussed in the previous section (see also Kargbo, 2000). GODUM is a binary variable representing the democratic transition and change of government in 1994, with 0 representing the period 1960–93, and 1 the period 1994–2004. This variable captures a major (perhaps the most important) structural change that has occurred in South Africa during the past four decades. Dummy (binary) variables are widely used in the social sciences to capture the impacts of policy shifts and other structural breaks in the data. MRERV_t is a variable that neutralises the impact of temporary fluctuations in the real exchange rate. We use the Hodrick–Prescott filter as a smoothing technique to remove short-run volatility in the real exchange rate, thereby deriving a proxy for the long-run equilibrium values of this variable. This method is increasingly being used by macroeconomists to obtain the smooth estimates of the long-term trend components of a particular series (see QMS, 1998).

CAPOR_t = the level of domestic production capacity represented by the average capital-output ratio. In order to increase agricultural exports and growth of the economy, there must be an increase in domestic savings and investment. This is one major aspect of the Harrod–Domar growth model (see also Johnson, 1987). Z_t = other factors (e.g. changes in technology, weather, etc.) that affect supply, and t = 1, 2, …, n refers to annual time periods. The working hypothesis underlying the supply equation is that South African exporters will supply more as the price of exports rises relative to domestic prices, and as the domestic production of agricultural exports becomes more profitable. Furthermore, we expect agricultural exports to increase, ceteris paribus, as the country's capacity to produce increases. Since agricultural exports adjust to excess demand conditions in the rest of the world, the prices of agricultural exports are determined in South Africa. This is especially the case for white maize, as South Africa is the largest producer of this crop. The southern African region is the largest producer and consumer of white maize in the world.

The aggregate import demand for agricultural products in South Africa is specified as: MAP_t = f((PM/Pdc)_t, Y_dt, MRERV_t, DAP_t, GODUM, OPEN_t, … K_t); where: MAP_t = an index of the quantity of agricultural imports demanded, PM_t = price index (unit value) of agricultural imports; Pdc_t = domestic price level, represented by the consumer price index (CPI); DAP_t = domestic production of agricultural products. We expect agricultural imports to decrease as domestic agricultural production and real prices increase. Y_dt, = real per capita income, measured as real per capita GNP. Growth in incomes is expected to have positive impacts on imports. K_t = other factors (e.g. changes in tastes and preferences, population growth, etc.) that influence import demand. All other variables are as defined above.

Owing to the implementation of policy reforms, domestic prices are becoming fully flexible and are linked to international markets through the exchange rate. We know from our knowledge of the theory of monetary equilibrium and exchange rate determination that the domestic price level, for example: P_dc = MV(r, Y)/Y; where: M = nominal quantity of money; V is the velocity of money; r = the interest rate; and Y = real income (see Kargbo 2000, 2005b). According to the strict version of purchasing power parity, the domestic prices are equal to foreign prices (P_dc*) converted at the exchange rate. Thus, P_dc = e P_dc*. The above equations are combined to obtain: e = [(1/P_dc*)V(r,Y)M]/Y. The equilibrium exchange rate is a function of velocity, real output and nominal money. Furthermore, P_dc is made of both agricultural prices (PA) and non-agricultural prices (PN). Thus, PA_t = f (r_t, e_t, DAP_t, …, Q_t), where: e_t = nominal exchange rate (rands per foreign currency), and Q_t = other exogenous factors that determine agricultural prices. Chambers (1984) argues that a restrictive monetary policy depresses the agriculture sector by lowering agricultural prices and income. This occurs in two ways. First, the exchange rate effects of tight monetary policy worsen the competitive position of the domestic agriculture sector in the international markets. Prices are not very flexible during periods of extensive trade and exchange rate controls, as South Africa experienced in the 1960s through the early 1990s. Thus the transmission mechanism for price and other policies was dampened to some extent. Under normal circumstances, tight monetary policies will increase interest rates and lead to appreciation of the rand, thereby putting pressure on the exchange rate. The rise in interest rates depresses prices for agricultural products because a rise in interest rates could be associated with higher costs of storing agricultural commodities in South Africa. The high storage costs may induce exporters to bring products to the international markets earlier than expected, thus inducing excess supply and depressing prices. Secondly, the effects of high interest rates make it less likely that farmers will embark on additional investment and excess production in the future, eventually leading to a decline in agricultural exports.

There are time lags involved in the interaction of exporters and importers for agricultural products in South Africa and international markets. The use of the vector error correction (VEC) model is appropriate in this paper since the impulse response functions generated during the analysis allow us to trace the dynamic impacts of various policy shocks to variables in the system. From the above discussion and desire to maintain parsimonious models, we specify two VEC models with the following variables: for South Africa's aggregate agricultural export supply: X^s, PX, PXW, WGNP, GODUM, CAPOR, MRERV, and OPEN; and for South Africa's aggregate import demand: MAP, PM, Pdc, DAP, Y_d, OPEN and MRERV. Export supply functions were also investigated for agricultural products such as oranges, apples, maize, grapes and sugar.

3.1 Data sources

The data for official nominal exchange rates, gross domestic product, the GDP deflator, WPI, the exports and imports of goods and services, CPI and proxies for WGNP (we used the indices of GDP of industrial countries at constant prices) and PXW (we used indices of export unit values for industrial countries) were obtained from various issues of the International Financial Statistics Yearbook, published by the International Monetary Fund (IMF, various issues). Supplementary data were obtained from the World Bank Africa Database 2003 CD-ROM (World Bank, 2003). Per capita GNP at constant prices, and the capital–output ratio are from various issues of the Quarterly Bulletin, published by the South African Reserve Bank (SARB, various years). Data for the indices of quantities of agricultural exports and imports, with export unit values and import unit values, were obtained from the Food and Agriculture Organization of the United Nations (FAO) Agrostat Database (FAO, 2004). The data for maize, oranges and other agricultural products were obtained from NDA (various years).

4. Empirical results and discussion

Since economic time series are generally not stationary, differencing of the series is performed to achieve stationarity. We use the augmented Dickey–Fuller (ADF) and Phillips–Perron unit root tests to determine stationarity of the series. The tests were performed with the EViews 3.1 statistical package developed by QMS (1998). All variables (except GODUM) were expressed in logarithms. The results of the unit root tests show that all series are I (1), implying that stationarity was achieved after differencing the series once. Results of the unit root and cointegration tests are not reported here owing to space considerations. The analysis involved Johansen's method in performing the multivariate cointegration tests (Johansen, 2000). The author used various combinations of variables in an effort to understand the determinants of agricultural trade flows in South Africa.

The normalised VEC estimates for aggregate import demand and export supply for agricultural products are presented in Tables 3 and 4. The normalisations centre on the variables with the coefficient 1,000. Long-run parameters are provided for each cointegrating vector, along with the adjustment coefficients for each variable in the system. Real exchange rate volatility plays a significant role in South Africa's international trade decisions for agricultural products, in particular as shown by the import demand models. Moreover, real exchange rate volatility has negative impacts on agricultural imports in South Africa. Similarly, equation 2 of Model B in Table 4 shows that real exchange rate volatility has significant negative impacts on agricultural exports. The impacts of the other variables are mixed in the various trade models. Although not shown here, owing to space considerations, the long-run estimates for maize and other agricultural exports reveal that real exchange rate volatility (MRERV), change of government (GODUM), openness of the economy (OPEN), real prices for agricultural products (RPEX), changes in world income (WGNP) and domestic production capacity (CAPOR) play significant roles in influencing international trade decisions by South African producers. Our empirical findings are similar to those reported by Tsikata (1999) for South Africa's manufactured exports.

Table 3: VEC estimates of South African import demand for agricultural products, 1965–2001

Model and eqn. no.	MAP	RPIM	Y	MRERV	DAP	GODUM	OPEN	CONSTANT	TREND
A (1965–2000)^a	1,000	−0,448 (1,686)^b	−4,401 (2,733)	−1,633 (1,790)	10,049 (5,049)	–	–	5,037	−0,206
B (1965–2000)	1,000	−0,283 (1,265)	−6,624 (4,318)	−2,256 (2,390)	10,053 (5,549)	−0,162 (0,548)	–	27,464	−0,168
C Eqn no. 1 (1966–2001)	1,000	0,000	−6,782 (6,195)	−1,448 (2,125)	5,984 (8,905)	0,588 (3,519)	−0,800 (1,512)	41,635	−0,112
Eqn no. 2	0,000	1,000	−4,707 (4,088)	−0,394 (0,550)	−1,809 (2,558)	−0,035 (0,200)	−3,579 (6,434)	48,580	0,109
Adjustment coefficients
A	−0,083 (0,811)	0,090 (1,262)	−0,010 (1,579)	0,002 (1,229)	−0,154 (7,410)	–	–	–	–
B	−0,098 (0,899)	0,079 (1,048)	−0,009 (1,427)	0,001 (0,943)	−0,152 (6,387)	0,013 (0,229)	–	–	–
C Eqn no. 1	−0,306 (2,402)	0,266 (2,790)	−0,012 (1,273)	0,020 (4,865)	−0,145 (4,780)	0,081 (1,100)	0,005 (0,176)	–	–
Eqn no. 2	0,207 (1,329)	0,038 (0,324)	0,011 (1,018)	0,030 (5,850)	0,083 (2,242)	0,058 (0,643)	0,075 (2,389)	–	–
Notes: ^aEstimation period; ^bnumber in parenthesis is the absolute t-value.
– Not available.
The real price for imports (RPIM)=PM/Pdc.
Models A and B each had one cointegrating vector while Model C had two cointegrating vectors.

Table 4: VEC estimates of agricultural export supply model for South Africa, 1963–2000

Model and eqn no.	X	RPEX	WGNP	MRERV	CAPOR	GODUM	CONSTANT	TREND
A (1964–2000)	1,000	0,840 (0,936)	30,618 (1,321)	35,363 (1,296)	35,426 (1,274)	–	−220,944	−2,504
B Eqn no. 1 (1963–2000)	1,000	0,000	−0,676 (0,535)	0,587 (0,440)	−0,049 (0,224)	4,391 (1,655)	−4,186	–
Eqn no. 2	0,000	1,000	1,838 (6,508)	−0,518 (1,743)	−0,607 (1,235)	0,070 (0,119)	−5,851	–
Adjustment coefficients
A	−0,044 (1,554)	0,026 (1,223)	−0,002 (0,814)	−0,003 (8,985)	0,001 (0,762)	–	–	–
B no. 1	0,0002 (0,003)	−0,043 (1,175)	0,005 (0,543)	−0,004 (7,533)	0,003 (0,884)	−0,074 (2,840)	–	–
no. 2	1,413 (4,692)	−1,192 (5,292)	−0,031 (0,608)	−0,008 (2,650)	0,042 (1,677)	0,034 (0,220)	–	–
Notes: The real price for agricultural exports (RPEX)=PX/PXW. All other variables are as defined earlier.
– Not available.
Model A had one cointegrating vector, while Model B had two cointegrating vectors.

The adjustment to long-run equilibrium is borne by several variables in each system. The adjustment coefficients show the changes in agricultural trade flows required to remove past departures of actual values of variables from long-run equilibrium levels. The transition to error correction representation and interpretation of results becomes a little complex in the case where there is more than one cointegrating relation among variables in the system, because the adjustment process involves the combined result of movement in several variables. The short-run effects of changes in agricultural trade flows in South Africa are presented in Table 5. The coefficients for ECT show rapid adjustments to agricultural trade shocks. The speeds of adjustment suggest that deviations from long-run equilibrium trade levels are corrected within two years in South Africa. Again, our empirical results show that the change of government (as represented by the coefficients for GODUM in Table 5, and in the estimates for individual export products, such as maize, sugar, etc.) has mixed effects on agricultural trade flows in the country. The author's estimates were subjected to a battery of diagnostic tests, such as RESET, CUSUMSQ and ARCH effects, and they show no sign of model misspecification and parameter instability. Overall, the tests point to robust models.

Table 5: Short-run estimates of aggregate agricultural import demand and export supply models for South Africa, 1965–2001

Independent variables	Export supply	Import demand
Constant	0,171(1,373)^a	0,054 (0,266)
Δ X	0,242(-1)^b (0,686)	–
Δ RPEX	0,820(−1) (1,734)***	–
Δ OPEN	1,725(−1) (2,188)**	−0,447(−1) (0,578)
Δ WGNP	−0,881(−2) (0,818)	–
Δ MRERV	−7,537(−2) (1,965)***	2,266(−2) (0,367)
Δ CAPOR	3,888(−2) (3,241)*	–
GODUM	0,720 (2,190)**	−0,413 (0,979)
Δ RPIM	–	0,329(−2) (2,014)**
Δ MAP	–	−0,291(−1) (1,765)***
Δ Y	–	−0,673(−2) (2,044)**
Δ DAP	–	−0,376(−1) (1,080)
ECT	−0,495(−1) (1,793)***	−1,035(−1) (2,568)**
R^2	0,534	0,536
Adjusted R^2	0,379	0,381
SER	0,304	0,256
D-W statistic	2,214	1,910
F-statistic	3,442*	3,466*
LM: χ^2 (p)	(1)=2,261	(1)=0,292
LB Q-statistic	(16)=6,871	(16)=11,374
ARCH: χ^2 (1)	0,157	0,528
WHET: χ^2 (15)	21,495	5,202
JB statistic: χ^2 (2)	0,487	7,907
RESET	F(3,29)=0,474	F(1,28)=0,365
No. of observations	37	37
Notes: ^aThe number in parenthesis refers to the absolute t-value.
^bThe number of lags which were determined by minimising the Akaike information criterion are in parentheses.
* Significant at the 1,0% level.
** Significant at the 5,0% level.
*** Significant at the 10% level.
– Not available.
Regressions were corrected for autocorrelation and heteroscedasticity, with the autocorrelation coefficients for export supply and import demand equations being –0,465 and 0,520, respectively. LM is the Lagrange multiplier test of residual serial correlation. ARCH and WHET are tests for heteroscedasticity based on Engle and White, respectively. LB Q-statistic is a test on the correlograms of the squared residuals to check for ARCH in the residuals. JB statistics tests for normality in the residuals. RESET is a test for specification error. The degrees of freedom for the other tests are in parentheses adjacent to the distributions. ECT is the lagged residual from the cointegration regression, and Δ is the first difference operator. See QMS (1998) for further details on these tests.

4.1 Impulse response functions

The dynamic responses of each variable over a ten-year period to standard deviation (SD) shocks of aggregate agricultural exports and imports are presented in Figures 4 and 5. The graphs illustrate the responses of a single variable to all shocks in the system. The impulse response functions provide an easy way of tracing the impacts of shocks to each variable in the system because the innovations are orthogonalised by a Cholesky decomposition. However, as Sims (1980) noted, there is no unique way of ordering the variables, thus the impulse responses change with changes in the order of variables. The introduction of inflation targeting in South Africa means that monetary policy is now more forward-looking. In effect, monetary policy decisions are based on the current and expected developments in some key variables (e.g. developments in exchange rates, money supply, changes in wages and productivity, credit extension and import prices) that influence inflation (see SARB, 2002). Because of the links among these variables and frequent changes in the economy, we have tried several orderings of variables; for example, the variables for Model A of the export supply function are ordered as follows: X^s, RPEX, WGNP, CAPOR, MRERV.

Figure 4: Responses of agricultural exports to one standard deviation shock in other variables

Figure 5: Responses of agricultural imports to one standard deviation shock in other variables

The dynamic interactions among the variables in Figures 4 and 5 clearly show significant and persistent impacts of own-shocks and cross-variable effects throughout the ten-year period. In particular, real exchange rate volatility and relative prices have significant impacts on South Africa's agricultural trade flows. Agricultural exports increase immediately in response to shocks in export prices (RPEX), and remain positive during the first five years. On the other hand, real exchange rate volatility has negative impacts on agricultural exports throughout the forecasting horizon. A favourable global economic environment, such as growth in real incomes of South Africa's major trading partners (WGNP), will have a modest increase in the world demand for its agricultural exports after a three-year delay. South Africa's capacity to produce (CAPOR) responds positively to real exchange rate shocks by increasing export supply after a two-year delay, probably to allow producers to shift resources among production activities. Agricultural imports will remain basically unchanged during the first four years in response to shocks in real exchange rates; thereafter, agricultural imports are negative for the rest of the forecasting horizon (see Figure 5).

4.2 Variance decompositions

The decomposition of forecast error variance for a ten-year forecasting horizon is shown in Tables 6 and 7 for the aggregate import demand and export supply functions. The variance decompositions reveal the relative importance of each shock to variables in the system. Own-shocks generated by agricultural imports explain 19–100 per cent of the forecast error variance in the import demand model (see Table 6), compared to 48–100 per cent of the forecast error variance explained by own-shocks in the agricultural export supply model (Table 7). The impacts of shocks to agricultural exports (X) on all other variables is also fairly large. Shocks to the real incomes of South Africa's trading partners (WGNP) have significant impacts on all other variables in the aggregate export supply model. Shocks in international markets are easily transmitted to the South African economy through the exchange rates. The strong interactions among variables in the tables suggest the potentially high impacts of policy changes on the South African economy. The impacts of shocks to the capital–output ratio (CAPOR) are transmitted strongly to all other variables in the system. Macro variables, such as real domestic income (Y), real import prices (RPIM) and domestic agricultural production capacity (DAP), explain a fairly significant proportion of the error variance of real exchange rates (see Table 6); and see also Table 7 for linkages between CAPOR, GODUM and real exchange rates (MRERV). This suggests that to some extent, exchange rate volatility is not a purely exogenous source of macroeconomic instability in South Africa. Domestic policies play a crucial role in influencing exchange rate movements.

Table 6: Decomposition of forecast error variance k-years ahead for agricultural import demand model

Forecast error in	k-years	SE	Shocks to
			MAP	Y	RPIM	DAP	GODUM	MRERV
MAP	1	0,229	100,00	0,000	0,000	0,000	0,000	0,000
	3	0,415	92,771	0,100	5,606	1,000	0,520	0,001
	7	0,616	76,329	2,338	6,329	3,738	6,660	4,605
	10	1,831	18,629	14,438	1,085	14,883	11,091	39,874
Y	1	0,014	0,791	99,209	0,000	0,000	0,000	0,000
	3	0,026	3,719	86,124	8,114	1,873	0,010	0,157
	7	0,052	5,493	61,917	21,096	2,731	5,144	3,617
	10	0,108	3,300	29,728	6,252	13,043	10,742	36,932
RPIM	1	0,158	28,840	2,169	68,991	0,000	0,000	0,001
	3	0,268	50,343	2,556	43,260	1,502	0,864	1,474
	7	0,509	44,138	7,037	13,545	12,830	5,242	17,206
	10	1,504	11,195	18,152	1,691	15,366	10,332	43,262
DAP	1	0,049	27,957	10,175	12,990	48,876	0,000	0,000
	3	0,066	41,712	9,290	12,731	34,714	0,122	1,428
	7	0,224	12,765	19,838	2,625	17,679	8,040	39,050
	10	1,547	3,585	18,886	0,099	16,547	12,230	48,650
GODUM	1	0,124	2,837	29,578	2,424	7,913	57,246	0,000
	3	0,214	5,452	29,508	3,106	8,326	53,363	0,243
	7	0,324	9,867	32,440	4,506	12,451	32,767	7,967
	10	0,747	8,251	27,384	2,419	15,372	18,028	28,546
MRERV	1	0,002	3,031	21,634	0,523	21,182	6,981	46,647
	3	0,010	5,122	16,277	1,486	21,725	12,932	42,456
	7	0,075	13,125	11,814	2,843	14,481	17,637	40,098
	10	0,372	3,797	19,458	0,513	16,410	11,728	48,092

Table 7: Decomposition of forecast error variance k-years ahead for agricultural export supply

Forecast error in	k-years	SE	Shocks to
			X	RPEX	WGNP	CAPOR	MRERV	GODUM
X	1	0,242	100,00	0,000	0,000	0,000	0,000	0,000
	3	0,403	53,517	1,005	20,411	23,817	0,219	1,027
	7	0,518	50,286	2,285	24,983	19,368	0,406	2,670
	10	0,581	47,735	3,227	26,331	17,620	0,588	4,499
RPEX	1	0,181	77,567	22,432	0,000	0,000	0,000	0,000
	3	0,294	35,898	22,331	16,846	17,302	2,396	5,227
	7	0,403	28,121	25,431	20,206	10,050	4,556	11,634
	10	0,477	25,511	25,380	21,347	7,301	5,780	14,652
WGNP	1	0,041	11,228	24,682	64,000	0,000	0,000	0,000
	3	0,074	8,701	35,708	49,676	1,133	0,639	4,142
	7	0,124	11,220	38,051	39,967	0,771	2,244	7,744
	10	0,159	12,537	35,821	36,507	0,627	3,590	10,917
CAPOR	1	0,020	5,471	0,356	3,319	90,852	0,000	0,000
	3	0,046	10,007	1,589	4,133	83,705	0,506	0,057
	7	0,086	7,738	2,800	1,599	82,128	3,325	2,407
	10	0,111	5,326	2,427	0,997	78,460	6,378	6,411
MRERV	1	0,003	7,173	2,078	5,197	5,782	79,770	0,000
	3	0,014	9,854	2,808	7,846	8,282	62,310	8,899
	7	0,072	12,943	2,353	8,753	3,967	43,882	28,101
	10	0,140	14,431	2,399	8,791	2,251	37,517	34,608
GODUM	1	0,127	16,748	4,622	0,191	1,382	6,237	70,818
	3	0,164	11,600	7,561	0,283	7,911	4,158	68,486
	7	0,217	17,022	5,650	9,810	6,071	16,660	44,784
	10	0,299	22,464	3,017	13,574	3,238	21,071	36,634

The forecast error (SE) of each variable increases gradually with the length of the forecasting horizon; and reaches some upper limit. This behaviour of the forecast error confirms the presence of stationary processes in the variables. The forecast error comes from the variation in innovations over time.

5. Summary and conclusions

The past decade has witnessed the deregulation and implementation of extensive market-oriented agricultural policy reforms after decades of government intervention in the agricultural sector during the apartheid era. This paper used a VEC model to investigate the supply and demand relationships for agricultural trade flows in South Africa over the 1960–2004 period. In particular, two models were specified. They describe South Africa's agricultural export supply function and an import demand relationship for agricultural products in South Africa. The author's empirical findings show strong links between exchange rates, prices and other variables in the economy. In effect, changes in relative prices, real exchange rates, domestic production capacity, the change of government, trade policies and real incomes have a significant and persistent impact on agricultural trade in South Africa. The estimates show the potential impact of policy reforms and the fairly swift response by market participants to shocks on exchange rates and other variables that determine international agricultural trade in the South African economy.

Furthermore, trade and exchange rate policy reforms are facilitating the growth in agricultural exports and reintegration of the country into the global economy. The South African government participates in various trade agreements to promote international trade, economic growth and poverty reduction in the country. The author's results also reveal that exchange rate volatility (risk) faced by market participants cannot be viewed solely as an exogenous source of macroeconomic instability in the economy. Domestic policy actions are contributing to the volatility of exchange rates in South Africa. In particular, an expansionary monetary policy will depreciate the exchange rate and thus, at least temporarily, improve South Africa's international price competitiveness. Potentially, there are offsetting responses from monetary expansion. For example, imports might increase owing to an increase in domestic demand emanating from lower interest rates, thereby encouraging higher investment and consumption spending.

To sustain the benefits of economic reforms aimed at boosting agricultural exports, the government needs to continue implementing more competitiveness facilitating measures, such as increasing access to international agricultural markets for South African exporters, and encouraging labour market reforms that promote employment and job training aimed at increasing efficiency in resource use, thereby further reducing the inequities created during the apartheid era. The author's estimates of the potential response of domestic production capacity to various shocks, combined with the research on unemployment in South Africa reported here and by other analysts such as Simbi & Aliber (2000), would help in formulating policies that address these issues. Research has shown that agricultural development has the potential to increase employment and reduce poverty in rural South Africa. The government must continue to work with the farming community, agribusiness firms and various stakeholders to provide a favourable environment for agriculture to flourish, generate employment activities and reduce inequalities in ownership and access to land, thereby contributing to economic growth, poverty reduction and political stability. There are strong links between political instability and exchange rate behaviour in African and other developing countries.

The author's empirical evidence shows that the change of government in 1994 has been an important factor in influencing international agricultural trade in South Africa. Complementary monetary and macroeconomic policies must concentrate on the extent to which prevailing exchange rates and domestic policies are consistent with the simultaneous internal and external equilibrium of the South African economy over time. This requires an effective coordination of microeconomic (e.g. agricultural price, food and other farm support or household) policies with macroeconomic policies that deal with, for example, exchange rates, interest rates, trade, taxation, money supply, income and wage rates in the country. The mechanics of how this process works efficiently require ongoing research.

Acknowledgments

The comments and suggestions on earlier versions of the paper by Dr PJ Nomathemba Seme, Siphokazi Kargbo, the editor Caroline Kihato and anonymous journal referees are very much appreciated. The author's sincere thanks also go to various staff members at the South African Mission to the IMF/World Bank in Washington, DC, the National Department of Agriculture, Statistics South Africa, and the South African Reserve Bank for providing some of the data and materials used in this paper. The views expressed herein do not reflect the official position of Ameriprise Financial or its affiliates. The author accepts sole responsibility for any errors and omissions.