An Augmented grAvity model of South AfricA ’ S exportS of motor vehicleS , pArtS And AcceSSorieS

The study applies an “augmented” gravity model to South Africa’s exports of motor vehicles, parts and accessories to 71 countries over the period 1994 to 2004. A static panel data model is specified and estimated. Several conclusions are drawn from the study. First, a number of variables, namely, importer income, distance, level of import tariffs, government effectiveness, regulatory quality, use of right-hand drive vehicles are important determinants of bilateral trade flows for motor vehicles, parts and accessories. Second, solving the gravity model deterministically, we show that export potential exists in a number of countries like Malawi, Zambia, Kenya and Malaysia. A number of barriers hinder the members of the National Association of Automobile Manufacturers of South Africa (NAAMSA) from exploiting these export markets. These include very high import tariffs, lack of South Africa’s diplomatic mission in the trading partner and the uncertainty regarding what happens at the expiry of the Motor Industry Development Programme (MIDP) in 2012. Finally, the export potential identified by the gravity model should be regarded only as an indication since it is sensitive to the model specification and sample of countries.


Introduction
South Africa's automotive industry, which includes manufacturing, distributing and servicing of vehicles and components, contributes about 7 per cent to the gross domestic product (GDP).In 2005, it employed slightly more than 300 000 workers in manufacturing vehicle and automotive components, the tyre industry and distribution and servicing in the motor trade (Table 2).South Africa is ranked 19th in the world in vehicle production, accounting for about 0.7 per cent of the world's vehicle output in 2004 (NAAMSA, 2005).Vehicle production is the second largest industry in South Africa's manufacturing sector, and one of the fastest growing.The contribution of exports of motor vehicle parts and accessories and other transport equipment to South Africa's merchandise exports to the rest of the world grew from 2.8 per cent in 1994 to 9.2 per cent in 2004.South Africa's car assembly plants emerged during the 1920s and aimed at mainly servicing the needs of the local market.As the demand grew, the motor industry developed very small plants producing at very high unit costs.Cars were mainly assembled locally from imported components with about 20 per cent local content.During this period, the industrial policy was built around import substitution with high import tariffs.South Africa's automotive industry has experienced major policy reforms over the last four decades, which were aimed at improving competitiveness, increasing domestic value addition and promoting exports.During the period 1961-95 automotive manufacturers were obliged to observe local content requirements, which led to the emergence of the Original Equipment Manufacturers (OEM).However, in spite of the reforms, the pre-1995 automotive industry was characterised by many small assembly plants producing too many models at economically low volumes (NAAMSA, 2005).The dissatisfaction with the earlier reforms led to the introduction of the Motor Industry Development Programme (MIDP) in 1995, which complied with the General Agreement on Trade and Tariffs (GATT) and the World Trade Organisation (WTO).
The structural changes have led to a significant increase in exports of automotive products.Vehicle exports as a percentage of domestic production increased from 4 per cent in 1995 to 27 per cent in 2005 (Table 2).Exports of component parts increased from 12.6 billion rands in 2000 to 21.7 billion rands in 2004 (Table 3).The main export destinations are, among others, Germany, the United Kingdom, United States, Japan and Australia (Table 3).According to NAAMSA (2005), components accounted for 55 per cent while build-up vehicles contributed 45 per cent of the total automotive export revenues in 2004.Black (2003) shows that rapid export expansion, considerable foreign investment and productivity improvements in the motor industry have been strongly influenced by the MIDP programme.As pointed out by Franse (2006), export expansion remains crucial to the future sustainability of South Africa's automotive industry since the size of the domestic market is relatively limited.Indeed, Chenery and Shrinivasen (1988) argue that exports generate greater growth as a result of greater capacity utilisation; greater horizontal specialisation as firms concentrate on narrower range of products; increasing familiarity with new technologies; greater learning by doing and the stimulative effect of the need to achieve greater internationally acceptable quality standards.
Despite the reforms, South Africa's share of the world's trade in motor vehicles remains small.The industry also remains a net user of foreign exchange.There is therefore a clear need to enhance the volume of South Africa's exports of motor vehicles.Two sets of factors play a role in stimulating exports: demand-side and supplyside factors.The MIDP to a great extent focuses on the supply-side.This paper focuses to a great extent on the demand side factors due to data limitations.The main questions to be answered are as follows; (a) Which countries has South Africa exploited with its automobile export potential?
(b) Which countries has South Africa not reached with its export potential?(c) Are there barriers to exploiting the export markets?
We attempt to answer these questions using a gravity model of trade.In its basic form, a gravity model expresses that the amount of trade between South Africa and its trading partners increases with their size as measured by national incomes and diminishes with the cost of transportation between them, proxied by the distance between their economic centres.
We further employ the gravity model to predict within-sample potential export trade flows for motor vehicles, parts and accessories (SIC 381-383) given certain conditions.A gravity model has been constructed by the International Trade Centre (ITC) called TradeSim (International Trade Centre, 2003).It estimates bilateral trade flows of developing countries with any of their partner countries.With regard to South Africa, TradeSim shows that there is untapped trade potential in the United States for motor vehicles and other transport equipment.
The novelty of our paper lies in three areas.First, we employ a static two-way error component panel data model, which includes both time-specific fixed effects and crosssection specific fixed effects.Second, we solve the baseline gravity model deterministically and compute export potential for 71 countries.Finally, we attempt to identify some of the factors that tend to inhibit NAAMSA members from exploiting the identified unexploited export markets.
The rest of the paper is organised as follows.Section 2 presents the background of the South Africa's automotive industry as well as the gravity model.Section 3 presents the model, estimation framework and the data.Section 4 focuses on the results and estimation of export potential.The last section deals with the conclusions.(Franse, 2006).South African assemblers not meeting the local content requirements were subjected to prohibitively high import tariffs.This led to the emergence of a domestic components industry (Black & Mitchell, 2002).

Background of South Africa's automotive industry
In 1989, the South African government started to move away from import substitution towards an industrial policy promoting export.Local content requirement was reduced from 66 per cent to 50 per cent.The aim of the policy was to expand the size of the market for firms and thus force them to rationalise the completely builtup (CBU) vehicles and component markets.This led to an increase in exports of component markets with limited change in CBUs.In 1995 the government introduced the MIDP with a view to helping the motor industry adjust to South Africa's reintegration into the global economy (Flatters, 2005).The objectives of MIDP were to improve the international competitiveness of South Africa's automotive and associated industries; improve vehicle affordability in the domestic market; encourage growth in the vehicle market and in the component manufacturing industry particularly in the field of exports; stabilise employment levels in the automotive industry; and create a better balance between the industry's foreign exchange usage and foreign exchange earnings.There were four key features of the MIDP; (a) Reduce tariffs on light vehicles and components, with tariffs being phased down faster than required by the WTO obligations; (b) Removal of local content requirements; (c) Duty free import of components up to 27 per cent of the wholesale value of the vehicle; and (d) Duty-rebate credits to be earned on exports of vehicles and components and to be used for duty-free imports of vehicles and components.Thus, the MIDP grants a production-asset allowance to vehicle manufacturers that invest in new plants and equipment, giving them 20 per cent of their capital expenditure back in form of import duty credits over a period of 5 years.The MIDP has given the OEMs the opportunity to scale down on the number of models produced locally and to import models, which were less economically viable to produce.The MIDP has allowed auto makers to concentrate on manufacturing certain vehicles or components for export while importing other models.Since its inception, the MIDP has been subjected to two reviews: 1999 and in 2002.Phase I of MIDP was operational from September 1995 to June 2000.The first review extended the programme from July 2000 to 2007 while the second review extended it to 2012.

Table 3 South Africa's exports to some selected Countries
Source: Computed using exports data collected from Quantec research (http://ts.easydata.co.za).
Date accessed 21st August 2005

Automotive industry export performance
South Africa's export of motor vehicles and components has increased substantially since the introduction of the MIDP in 1995 (tables 2 and 3).Table 3 presents data on South Africa's export of motor vehicle, parts and accessories over the period 1988-2004.Column 2 of Table 3 shows the total exports over the period 1988 to 2004 in US dollars.Columns 3, 4 and 5 of Table 3 present rankings of export destinations using total exports as a criterion for ranking.Three different rankings are presented; the period 1988-1993, 1994-2004 and the average over the period 1988-2004.The ranking for the period 1988-1993 reflects South Africa's motor vehicle trade during the apartheid era.The countries that were important are Germany, United States, Zimbabwe, Zambia and United Kingdom, among others.During the post-apartheid era (1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004) the situation changed marginally with the emergence of Japan and the United Kingdom as major trading partners.The last column compares the ranking in the period 1988-1993 and 1994-2004.Positive figures indicate improvement in terms of ranking for that country.For instance Japan's ranking improved from 10 during the apartheid era to number 4 during the period 1994-2004.

The gravity model
The gravity model, first applied to international trade by Tinbergen (1962) and Pöyhönen (1963), has been used in the social sciences since the latter half of the nineteenth century to explain migration and other social flows in terms of the "gravitational forces of human interaction".
The basic form of the gravity model, true to its namesake, is analogous to the Newton's 1687 "Law of Universal Gravitation".The law holds that the gravitational force between two physical bodies (F ij in Newtons) is proportional to the product of each body's mass (M i and M j in kilograms) divided by the square of the distance between the respective centres of gravity (D 2 in metres); In 1962 Jan Tinbergen proposed that roughly the same functional form could be applied to international trade flows.Since then it has been applied to a whole range of "social interactions" including migration, tourism and investment with a lot of empirical success.The gravity law of social interaction can be expressed roughly in the same notation as equation 1 except that M i and M j are redefined as the economic sizes of the two locations.
The gravity model has gone from a theoretical orphan to being the favoured child of all main theories of international trade.Despite its use in many early studies of international trade by Tinbergen (1962), Pöyhönen (1963) and Linnemann (1966), the gravity model was considered suspect because it had no theoretical foundation.Leamer and Stern (1970) provided the first foundation with the "potluck assumption", that argues that nations produce their goods and throw them all into a pot; then each nation draws its consumption out of the pot in proportion to its income.They conclude that the expected value of nation i's consumption produced by nation j will equal nation i's share of world GDP times nation j's share of world GDP.Anderson (1979) was the first to provide a clear microfoundation but rested on an assumption that was viewed as ad hoc at that time, namely that each nation produced a unique good that was only imperfectly substitutable with another nation's goods.
Although, the gravity model fell into disrepute in the 1970s and early 1980s, an attempt was initiated by Bergrstrand (1985) to give it a sound microeconomic foundation.He developed a theoretical connection between the factor endowments theory and bilateral trade.However, he did not manage to reduce the complicated price terms to an empirically implementable equation.
A simple gravity model for South Africa's exports of motor products is Where X ij the value of bilateral exports of automobile products from South Africa to nation j.
In line with practice, i refers to origin country (South Africa) and j to destination nation Y i ,Y j South Africa and importer incomes ( ) A set of measures impacting either negatively or positively on trade flows from South Africa to country j but are positive so that logarithmic transformation can be applied A set of measures impacting either negatively or positively on trade flows from South Africa to country j but logarithmic transformation cannot be applied e.g.dummy variabales.ij f the random error term In this model four assumptions are made.Firstly, all goods are differentiated by place of origin (Armington Assumption).In other words automobile products from South Africa are distinguishable from others.Secondly, each country is specialised in the production of only one good.Thirdly, the supply of each good is fixed.Finally, preferences of consumers are homothetic (approximated through a constant elasticity of substitution utility function).
Equation 3 is generated by performing logarithmic transformation on Equation 2.

Remoteness variable approach
Bilateral trade between South Africa and country j depends on the bilateral trade barrier, t ij , relative to all other partners.As pointed out by Anderson (1979), bilateral trade flows are a function of the bilateral trade barrier relative to the average barrier of the two countries with all their partners after controlling for country size.Consequently a remoteness variable is included in Equation 2 to proxy for this average trade barrier.Remoteness of country i is usually computed as in Anderson and van Wincoop (2003: 173); Where d im is the distance of country i.This variable is intended to reflect the average distance of country i from all trading partners other than j.However, as argued by Anderson and van Wincoop (2003), this approach, which relies solely on distance, does not capture the entire range of factors impeding bilateral trade flows.Therefore the gravity model in Equation 3suffers from omitted variable bias.

Multilateral (price) resistance terms approach using price indices
Anderson and van Wincoop (2003) suggest a theoretical gravity specification based on the approaches of Anderson (1979), Bergstrand (1989) and Deardorff (1998), which explicitly takes into account "multilateral (price) resistance" terms.These multilateral resistance terms consist of country specific price indices, p i and p j .Consequently, bilateral trade between country i and country j is dependent on the following; Where  is the elasticity of substitution between all goods.Unfortunately, the multilateral trade barriers suggested in equation 5 are unobservable.However, it is possible to use a fixed effects model as suggested by Anderson and Van Wincoop (2003).
Equation 6 is an "augmented gravity model" of trade, which has been modified to include the multilateral terms.
3 The Model

Model specification
The gravity model has traditionally been estimated using cross-section data.However, this has been shown to generate biased results since typically heterogeneity among the countries is not appropriately controlled for.
With heterogeneity, South Africa may export different amounts to two different countries, even though the two export markets have the same GDPs and are equidistant from the South Africa.
The panel data approach attempts to solve the problem by permitting more general types of heterogeneity.Our study follows the panel data analysis approach.The basic static gravity model is specified as follows; Where: X ijt : Foreign price value (e.g.US dollars) of exports of goods by South Africa to country i; Y i : South Africa's GDP in US dollars; The geographical distance between South Africa's economic centre of gravity (capital city-Pretoria) and the trading partner's economic centre of gravity (capital city); p i : South Africa's price index using the GDP deflator; p j : Importer's price index using the GDP deflator; tar i : Most favoured nation (MFN) import tariffs applied on motor equipment by the trading partner.This is a normal non-discriminatory tariff charged on imports (excludes preferential tariffs under free trade agreements and other schemes or tariffs charged inside quotas); ge j : Index of government effectiveness in the trading partner; rq j : Index of regulatory quality in the trading partner; cc j : Index for control of corruption by the trading partner; Lang j : English language dummy.Trading partners, whose official language is English, are coded 1 and otherwise 0; In South Africa drivers are expected to keep left, which implies that production plants are designed to manufacture right hand-driven vehicles and components to be fitted on the same; and  ijt : captures all the factors that influence exports but not included in equation 6.  (ii) The country-invariant effect of the introduction of the MIDP in 1995; (iii) The different phases of MIDP; (iv) The effect of the government of national unity in South Africa since 1994; (v) The Asian crisis of 1997; (vi) Argentina/Mexican/Russian crisis of 1998-99; and (vii) The rand collapse in 2001 Equation 7 can now be re-specified as follows;

Estimation framework
In using the panel data approach, a number of challenges must be addressed.These include zero trade, estimating country fixed effects and time-invariant regressors, and endogeneity.

Dealing with zero trade flows
The log-linearised model in equation 8 is not defined for observations with zero trade.There are various approaches to handle the presence of zeros.These include discarding the zeros from the sample, adding a constant factor to each observation on the dependent variable, estimating the gravity in nonlinear form using the fixed effects Poisson maximum likelihood estimator and the use of unbalanced panel estimation techniques.

Heteroscedasticity and serial correlation
The log-linearised model in equation 8 may be both biased and inefficient in the presence of heteroscedasticity.This is corrected by using cross-section and period weights in the estimation.To avoid serial correlation, one of the models in Table 5 is estimated in dynamic form.

Estimating effects and time-invariant regressors simultaneously
The effects of time-invariant regressors like distance, language, etc. can be estimated using a random effects model (REM).REM assumes that the effects are randomly distributed across the different countries.However, the random effects coefficients cannot be interpreted since they cannot be attributed to a specific trading partner.
The alternative is to estimate country-specific fixed effects model (FEM), using WITHIN estimation (demeaning before OLS) in which the coefficients are interpreted as the unobserved effects of each country.The problem with estimating country-specific fixed effects and time-invariant regressors simultaneously is perfect multicollinearity.The problem emanates from the fact that both country-specific fixed effects and the time-invariant factors are captured by using dummy variables.
The study follows the approach of Cheng and Wall (2005).This approach uses two steps.The first step entails estimating a modified form of Equation 8excluding the time-invariant variables.The country-specific fixed effects ( j n t ) are regressed on other time-invariant variables as follows;

Data descriptions
Sample of countries: 71 countries are selected on the basis of importance to South Africa's automotive exports as well as availability of data for variables described in equations 8 and 9. Drive on left side dummy: About a quarter of the world, including South Africa, drives on the left.There are also countries in the sample whose inhabitants drive on the right-hand side of the road.A list of countries that keep left and right on the road is collected from http:// users.pandora.be/worldstandards/driving%20on%20the%20left.htm.The drive on left side dummy variable is created by coding those countries using right hand-drive vehicles as 1 and otherwise 0.

Importer price index:
The study uses GDP deflators as a proxy for import price index.The import price index, p j , captures a number of multilateral resistance variables p j =

!
. t ij are all the factors that restrict trade,  is the elasticity of substitution between automotive products,  i is the share of South Africa's GDP in world GDP, ! and  j is the share of importer GDP in world GDP.The data is collected from the International Financial Statistics of the IMF.

Regression results and discussions
Table 5 presents the estimation results from equations 8 and 9.The first, is the pooled model in which homogeneity in terms of cross-section and time is assumed.With adjusted R-squared above 0.6, the gravity equation explains more than half of the bilateral export trade of motor vehicles, parts and accessories.The low Durbin-Watson statistic shows that the models suffer from a problem of serial correlation. .
The following emerges from the results.First, the importer GDP has the expected positive and significant effect in all the models.The magnitude of the coefficient is close to unity in all models, which is consistent with the findings in McCallum (1995) and Anderson and Wincoop (2003).The high income elasticity indicates that South Africa's automobile exports could rise significantly if her trading partners maintain strong economic growth.
Second, distance has the expected negative and significant effect on South Africa's exports of motor vehicles, parts and accessories.It means that South Africa exports less to countries that are far off.Grossman (1998) reckons that if shipping costs are at the order of 5 per cent of the value of traded goods, the trade-distance elasticity should be around -0.03, which is much lower than the trade-distance elasticity in our study.Since distance is a proxy for transport costs, any policy that reduces costs for exporting will enhance exports of motor vehicles, parts and accessories.
Second, importer price index has the expected negative and significant effect on trade.This is not surprising since the importer price index captures a number of trade restricting factors.
Third, motor equipment MFN tariff has the expected negative sign but statistically insignificant in the pooled model.An attempt to include it in other models created perfect multicollinearity.
Fourth, importer government effectiveness, regulatory quality and control of corruption have the expected positive and significant effect.This indicates that governance issues in the importer country have a significant effect on South Africa's exports of automobile products.
Fifth, South Africa tends to export more motor vehicles, parts and accessories to Englishspeaking countries.Having the English language in common between South Africa and trading partner nations helps to build networks of trust.Institutions are also shared by increasing the degree of common cultural, literary and educational material, and by increasing the probability of migration.This affects the language policy in South Africa.
Sixth, being a member of the EU enhances South Africa's exports of automobile products to that country.This may be attributed to the positive effect of the EU-South Africa Trade Co-operation and Development Agreement of 1999 and the cultural heritage of South Africa with Europe.
Seventh, the fact that a country is a member of NAFTA enhances its imports of automobile products from South Africa.In other words there are some unobservable factors that enhance South Africa's exports of motor vehicles and parts to the United States, Canada and Mexico.
Eighth, belonging to the African continent tends to enhance South Africa's exports of automobile products to that country.This is to be expected since South Africa is perceived by many African states as the "big brother".It has played an important role in the African Union (currently hosting the African Parliament), NEPAD, solving conflicts and being the technological leader in the continent.
Ninth, membership to SADC tends to reduce trade barriers of South Africa's exports of automotive products.This is to be expected since South Africa is a key member of SADC, which implies that her exports of automotive products face preferential tariffs/treatment in member states.Additionally, some of the members within SADC belong to the South African Customs Union (SACU) and Southern Africa Monetary Union.
Tenth, membership in MERCOSUR enhances South Africa's trade in automobile products.Although trade relations between South Africa and MERCOSUR have been small, they have picked up in the recent past.South Africa signed a framework agreement with MERCOSUR in 2000, comprising Argentina, Paraguay, and Brazil (with Bolivia, Chile and Colombia, Ecuador, Peru and Venezueala as associate members) with the aim of forming an FTA.The agreement was eventually signed between SACU and MERCOSUR in 2004.
Eleventh, membership in Middle East and Asia has had a positive effect on South Africa's exports of automotive products.This shows that South Africa should attempt to enhance her trading relations with countries in the Middle East and Asia.
Twelfth, South Africa's export of motor vehicles, parts and accessories designed for vehicles driving on the left are obviously preferred by trading partners that drive on the left hand side.
Thirteenth, the positive (negative) effects show that these are unobservable characteristics that tend to enhance (inhibit) South Africa's exports of motor vehicles, parts and accessories to the countries in the sample.These time-specfic fixed effects reflect that initially the industry tended to struggle with high production costs.However, after 2001 the MIDP programme contributed to the sector's export performance.

Table 5
Estimation for motor vehicles, parts and accessories (SIC Notes: (i) Country-specific effects from stage 1 of Cheng-Wall model are presented in tables 6a and 6b (ii) ***, ** and * refer to significance at 1%, 5% and 10% respectively (iii) White cross-section standard errors used to correct for heteroscedasticity Finally, the estimates for country-specific effects are in column 7 of tables 6a and 6b.On the one hand, the positive coefficients indicate that there are unobservable characteristics in those countries (column 1 of tables 6a and 6b), which enhance South Africa's exports of motor vehicles, parts and accessories to those countries.Examples include Mozambique, Congo and United Kingdom.On the other hand, there are unobservable country-specific characteristics that tend to inhibit South Africa's exports of motor vehicles, parts and accessories to countries like Italy, Peru and Kuwait.From a policy perspective, it is imperative to conduct a survey on motor vehicle exporters to determine the other factors that may hamper trade to the countries that have negative country-specific effects.

Potential exports
There are basically two approaches to computing trade potential.The first approach obtains withinsample trade potential as the difference between estimated trade from actual trade relations between countries.The second approach derives out-of-sample trade potential estimates (e.g.Brülhart & Kelly, 1999).In this approach, gravity model parameters are estimated and applied to forecast "natural" trade relations between countries.The difference between the observed and predicted trade flows represent the unexhausted trade potential.
Whichever approach is used, the finding of untapped trade potential calls for proactive export promotion policies e.g.bilateral and multilateral agreements, trade facilitation etc.To the contrary, finding actual trade exceeding potential trade (successful partnership) implies that trade has reached its potential level and no social cost is anticipated from future integrations.
A model is constructed from stage 1 results of the Cheng-Wall model in Table 5 and solved deterministically to determine within-sample potential exports of motor vehicles, parts and accessories.The choice of this model is predicated by the fact that it addresses many of the problems in the gravity equation.These entailed calculating the predicted values of exports, XSit t i.e. export potential is equivalent to the within-sample predictions.
It is imperative to note that the export potentials are dependent on the nature of the sample specification employed in this study.This means that our potential exports should be understood in the context of the specified model we have used.Different specification and sample selection may lead to different results.
Column 1 of Table 6 presents the countries in the sample.Column 2 presents the export potential in US dollars, which is computed as follows; Potential = predicted -actual (10) Where predicted and actual represents the predicted and actual exports of motor vehicles, parts and accessories, respectively.Column 3 is the ranking of the countries on the basis of unexploited export potential.On the basis of this ranking, Mozambique has the highest untapped export potential that South Africa should explore ways of taking advantage of.The finding of untapped export in Table 6 calls for proactive export promotion policies e.g.bilateral and multilateral agreements, trade facilitation etc.To the contrary, if actual trade exceeds potential trade (successful partnership) it implies that trade has reached its potential level and no social cost is anticipated from future integrations.
Column 4 shows the ranking of the countries on the basis of actual exports.For instance despite the fact that Mozambique has the highest export potential, it is ranked number 8 in terms of importance to South Africa.Columns 7-12 present some important characteristics regarding the trading partners in question.The point is that even though export potential exists in a particular country, the characteristics presented in columns 7-12 could make it difficult for the National Association of Automobile Manufacturers of South Africa (NAAMSA) members to exploit that opportunity.
Column 7 indicates whether a trading partner is a WTO member state or not.If a country is a member of the WTO, it is obliged to observe some specific rules and principles such as most favoured nation (MFN), national treatment,

Table 6b
Potential exports and other selected Source: Compiled from various sources

Conclusion
This study applies an "augmented" gravity model South Africa's annual bilateral exports of motor vehicles, parts and accessories (SIC 381-383) to 71 of its trading partners over the period 1994 to 2004.A static panel data model is utilised to estimate the coefficients.A number of results emerge from the study.
First, certain characteristics of the trading partners enhance South Africa's exports of motor vehicles and parts and accessories: GDP, government effectiveness, regulatory quality, English language, membership to EU, Africa, NAFTA, Asia, and MERCOSUR (South America) as well as whether drivers in a country keep left or not.
Second, geographical distance, import tariffs levied on motor equipment and the fact that a country is located in the Middle East tend to inhibit the exports of motor vehicles and parts and accessories.The negative effect of distance implies that high transport costs inhibit South Africa exports of motor vehicles and parts and accessories.This is not surprising given the fact that the automotive export products are bulky.
Third, there are country-specific fixed effects, which tend to inhibit South Africa's exports.There is therefore a need to conduct a survey to determine these factors.
Fourth, the positive time-specific effects after 2002 show that the MIDP has been effective in enhancing the performance of the automotive industry exports.The question that needs to be addressed is what happens after the expiry of the MIDP in 2012?
Fifth, the gravity model shows that there are a number of countries where there is export potential that South Africa can exploit.However, there are some difficulties that exporters may face in exploiting these export destinations (Tables 6a and 6b).These include, among others, very high import tariffs, lack of South Africa's diplomatic mission in the trading partner, country not being a member of the WTO, lack of trade agreement with the country.There is therefore a need for the Department of Trade and Industry (DTI) to work on these issues with a view to making it easy for NAAMSA members to export.
AFR j : African countries dummy (African countries coded 1 and otherwise 0); SADC j : SADC member state dummy (SADC countries coded 1 and otherwise 0); EU j : European Union dummy (EU members coded 1 and otherwise 0); NAFTA j : North American Free Trade Agreement dummy (NAFTA members coded 1 and otherwise 0); ASIA j : Asian countries dummy (Asian countries coded 1 and otherwise 0); ME j : Middle East dummy (Middle East countries coded 1 and otherwise 0); MERC j : MERCOSUR FTA dummy (MERCOSUR members coded 1 and otherwise 0); Drive j : Keep left driving dummy (Countries where vehicles keep left are coded 1 and otherwise 0).

Table 1
(Black, 1994) key automotive industrial policy periods and key instruments.South Africa's automotive industry started with car assembly plants in 1920s, which mainly served the needs of the local market.Themotorindustrydevelopedvery small plants producing different models and in some cases different makes at very high unit costs(Black, 1994).Cars were mainly assembled locally from imported components with about 20 per cent local content.During this period, South Africa followed an import substitution industrial policy, which entailed imposing heavy tariffs that increased with valueadded(Barnes, Kaplinsky and Morris, 2004).

Table 1
Development of automotive policy in South Africa

Table 2 Selected indicators of South Africa's automotive industry
Source: The National Association of Automobile Manufacturers of South Africa (NAAMSA) Annual Reports of 2004 and 2005.