Can trade liberalisation in South Africa reduce poverty and inequality while boosting economic growth? Macro–micro reflections

Abstract

South Africa is trapped in a cycle of modest growth, unacceptable poverty levels and record unemployment. This has led to renewed interest on the relationship between macro (growth) and micro (poverty and distribution) issues. This paper uses a macro–micro tool that couples a computable general equilibrium model with microsimulation models to examine the impact of further unilateral trade policy reforms on growth, poverty and welfare. Trade liberalisation alone has very minimal short-run macroeconomic consequences while its long-term impacts are positive and magnified by technical factor productivity (TFP) effects. Trade liberalisation has no appreciable impact on poverty in the short run even if we allow for trade-induced TFP increases. In the long run, however, poverty reduces even in the case when we do not allow for TFP increases. Trade liberalisation policy has been found to be progressive despite the low level of tariff protection remaining in South Africa.

Keywords:

• trade
• poverty
• economic growth

JEL codes:

• C68
• D31
• D58

1. Introduction

Although South Africa has put in substantial effort aimed towards reducing poverty and inequality since 1994, there are many studies that show inequality has risen while poverty has fallen marginally since 2000 (StatsSA, 2002; Ozler, 2007). Disappointment with upward trends in inequality amidst positive economic growth, heightened especially by the 2009 global financial crisis, have all led to renewed interest in the relationship between macro (growth) and micro (poverty and distribution) issues. Increasingly, government economic programmes and policies are, with stronger emphasis than before, aiming at the dual objective of accelerating growth and fighting poverty and unequal access to opportunities. Recent examples of economic programmes and policies designed for this under the rubric of the New Growth Path and the National Development Plan include: public spending (shifts in size and composition); tax policy (transfers and subsidies); structural policies (trade liberalisation, regulation policy, labour market reforms); and monetary and exchange rate policies. It is clear that these macroeconomic pro-growth policies require rigorous analysis that assesses their effectiveness in terms of poverty and distributive effects.

With a view towards contributing to the literature on quantitative dimensions of growth, poverty and income inequality in South Africa, this paper examines the impact of further unilateral trade policy reforms on growth, poverty and inequality.³ The bulk of techniques used in the South African literature to carry out poverty and inequality analysis to date have been based on either the public finance (e.g. partial equilibrium microsimulations, structural econometric models, tax and benefit incidence analyses, etc.) or evaluation approaches of the randomised control trial type.⁴ While yielding extremely useful results in the appropriate context, these approaches cannot assess the poverty and income distribution effects of macroeconomic policies such as wholesale trade reform and trade liberalisation for two main reasons. The first is that such policy reforms are large and affect the whole economy so that the fiscal effects and the ensuing general equilibrium effects cannot be ignored. The second is that creation of a macro counterfactual that permits comparison of actual macro effects with what the economy would have looked like in the absence of policy reforms is critical. Macro–micro techniques of the type used in this paper are required for this.

To understand as well as quantify the impacts of trade liberalisation on the economy and on the poor, a sequential dynamic computable general equilibrium (CGE) model is used. The endogenous changes obtained from the sequential dynamic CGE model are then fed into national survey data for predicted household poverty effects. The CGE approach is appropriate because there are likely to be significant indirect effects that will have strong impacts on allocation of scare resources following trade liberalisation. There is a growing tradition of trade-focused CGE modelling in South Africa, starting with the work of Gelb et al. (1994) and followed by Cameron et al. (1994).⁵ Most of these CGE models have in common the fact that they are based on representative households. Although representative household models can give useful insights on the average impacts of trade liberalisation on household groups, poverty cannot be analysed explicitly in such models because the distribution of income within the group is not modelled. If intrahousehold distribution was modelled, then the model would be equivalent to a microsimulation model that works with actual as opposed to representative households. Even if one was to have detailed representative household disaggregation in the model, there will still be no way of capturing individual household heterogeneity because the distribution of income within groups is not modelled. Macro–micro empirical methods make it possible to estimate redistributive effects under a wide variety of policy scenarios because they account for individual household heterogeneity. We can draw similarities between our work and that of Annabi et al. (2005a, 2005b), Fofana et al. (2004, 2007) and Chitiga et al. (2007).

The rest of this paper is organised in the following way: Section 2 presents an overview of trade policy evolution in South Africa; while in Section 3 the model used for analysis is developed and also the policy simulations; Section 4 gives extended discussion of the results; and Section 5 presents concluding remarks and policy discussion.

2. Trade policy evolution in South Africa

At the policy-making level, South Africa's trade and investment policies are driven by the Department of Trade and Industry⁶ while poverty is handled by several multitude ministries because it is perceived to be a cross-cutting issue. According to Bell (1992, 1997), trade policy was broadly geared towards import substitution between 1925 and the 1970s. In the 1980s there were attempts to open the economy through export stimulation policies. Quantitative restrictions continued to be reduced throughout. The policy stance has been geared towards steering the country towards neutrality through a mix of rationalisation and liberalisation. By 1994 most of the quantitative restrictions had been removed.

South Africa has thus made a lot of progress towards trade liberalisation. These measures have resulted in substantial trade liberalisation. Average tariff rates have been halved. The largest absolute declines in protection have been on consumables. The total number of Harmonised System (eight-digit) commodity lines declined to 6700 in 2004. The Harmonised System eight-digit lines bearing formula duties declined from 1900 in 1993 to five in 2002 (WTO, 1998, 2002). The number of lines with specific tariffs fell from 500 in 1993 to 195 in 2002. Commodity lines with mixed non-ad valorem duties have fallen from 160 in 2000 to 60 in 2004. Despite these efforts towards a deep and sustained trade liberalisation strategy and far-reaching reforms, it is unclear whether this has been good for growth, equity and poverty reduction.

In recent years, the debate concerning what direction South African trade policy should take has tilted more towards greater protection. It has been argued that a more active role for the state in the form of increased protection is required to allow selected industries to grow their ability to export to a degree to which they will then be able to compete effectively in global markets without that protection or to save domestic jobs. The implicit argument that is behind the increased protection sentiments is centred on the notion that increased protection would be beneficial for the protected industry and the South African economy. This argument presents an interesting case of how a country might use safeguard trade barriers in order to better achieve a domestic policy objective.

There has been a debate around the effects of trade liberalisation on employment and wages (see Roberts & Thoburn, 2002). Many researchers have blamed unemployment on technology and rigidities in unskilled and semi-skilled wages. Using a model featuring perfect competition and efficient markets, Fedderke & Vase (2001) find that trade has led to higher returns to both labour and capital, while technology has induced lower wages. Given this, the main reason for high unemployment is suggested to be excessive wages (above what they would see as the ‘market clearing’ wage rate). This implies that they believe labour market outcomes are not unequal enough and that low-paid workers should be earning even less despite the wages of a large proportion being too little to meet basic housing, transport and nutrition requirements for an average family (Roberts & Thoburn, 2002).

The actual picture for trade, investment and inclusive growth for South Africa presents a mixed picture. Figure 1 shows South Africa's annual gross domestic product (GDP) growth for the period 1990 to 2010, and it can be seen that growth in South Africa has been positive from 1993 until 2008, and became negative in 2009 due to the 2008 global financial crisis, but slowly started to recover in 2010 with a growth rate of 2.9%.

Figure 1: South Africa's annual GDP growth: 1990–2010

Source: New York University (2009) and The South African Reserve Bank (2012).

Despite this mostly positive growth, the picture looks bleak for South Africa's unemployment. The trend from Figure 2 shows an increase in unemployment from 1994, with rates stubbornly remaining above the 20% threshold from 1996 onwards.

Figure 2: South Africa's unemployment rate: 1994–2010

Source: World Bank (2012) and CIA (2012).

There are indications of increased productivity in South Africa since the early 1990s, and this was especially so in the years from 2000 onwards but prior to the global financial crisis. Figure 3 presents South Africa's index of real production and it shows an upward trend for the period, with 2003 as the base period. This increased productivity has mostly come about through increased labour productivity, whose growth has been exceeding that of total factor productivity. This can be linked to growth in the capital/labour ratio and it can be deduced from this that growth in labour productivity has been at the cost of employment; that is, the capital/labour ratio has increased through the shedding of labour (McCarthy, 2005). To illustrate this, Figure 4 shows the indices of labour productivity and total factor production, as well as that of the capital/labour ratio. All indices show an upward trend, but the capital/labour ratio is in the lead, followed by labour productivity and then total factor productivity.

Figure 3: South African productivity index

Note: Year 2003 = 100.
Source: Statistics South Africa (2008).

Figure 4: South African indices for labour productivity, total factor productivity and capital/labour ratios

Source: Statistics South Africa (2008).
Note: Year 2003 = 100.

3. The analytical framework

This section presents the structure of the poverty-focused sequential dynamic CGE microsimulation model applied to South African data. The CGE model used is based on Annabi et al. (2005a, 2005b) and is documented fully in Mabugu & Chitiga (2006). The static part of the model is fairly standard and follows from the EXTER model (Decaluwé et al., 2001). Sequential dynamics is built into the EXTER model for a small, open economy so that the dynamics do not influence world prices and interest rates. Early recursive dynamic CGE models include the work of Bchir et al. (2002), Bourguignon et al. (1989) as well as Jung & Thorbecke (2003). Taking into account South African CGE literature, the model's dynamic structure is similar to that proposed by Thurlow (2004). Arndt & Lewis (2001) developed a similar model structure to analyse the consequences of AIDS on the economy. Rattso & Stokke (2005) analyse trade liberalisation in a dynamic Ramsey model, whose growth specification is of direct relevance to our model. By using our approach, we are able to study the impacts of increased protection on growth, poverty and inequality, as well as their transition path. The full model equations are presented in Mabugu & Chitiga (2006).

The static part of the model broadly has a production and demand side interacting simultaneously. Overall output is modelled using a Leontief production structure. In turn, value added is a CES combination of labour and capital. Total capital demand is derived from cost minimisation subject to the CES function. Labour is a CES aggregation of skilled and unskilled labour. All labour categories are assumed mobile across sectors, and wages are crucial for income distribution. Capital, on the other hand, is sector specific in the short run, implying rising supply curves on the real side, but is allowed greater mobility in the long run when dynamics set in. As a result of this asymmetry, we would expect greater volatility in the rental capital return in the short run and broad convergence in the long run. The choice between domestic and imported inputs is specified as a CES function. On the demand side, households maximise Stone Geary-type utility functions subject to their budget constraints, yielding linear expenditure system demands. The Armington assumption is used to model the choice between domestic and imported goods by households for final consumption. General equilibrium requires that the goods and factor markets are in equilibrium and the fundamental macroeconomic identity is satisfied. The goods market clears when demand and supply are equated via the material balance condition in each period. The fundamental macroeconomic identity requires the equality between investment and savings. The model has two options for revenue compensation in response to a liberalised trade scenario that may reduce tariff revenue. The adjustments could be on the indirect tax rate or on the direct tax rate. Finally, the nominal exchange rate is chosen to be the numéraire for each period.

The static model is made sequentially dynamic by a set of accumulation and updating rules from one year to the next. Growth in the total supply of labour is endogenous and is driven by an exogenous population growth rate. It is also assumed that minimal consumption in the linear expenditure system and all other nominally indexed variables such as transfers also grow at this same rate. Capital stocks are updated based on investment, capital stocks and depreciation in each period. A key question to resolve is how to allocate new investments among the different competing sectors. The literature suggests two approaches: using a capital distribution function (see Abbink et al., 1995) or using an investment demand equation. We opt for the investment demand approach that fits in well with the data we have available on investment by destination.

There are now a number of alternative specifications of the investment by destination functions in the literature (see for example Bchir et al., 2002). The most well known in dynamic CGE circles and one that we use in this study follows from the work of Bourguignon et al. (1989), and is later elaborated on in Jung & Thorbecke (2003) and Annabi (2003). It takes the following form:

where and are positive parameters calibrated on the basis of the investment elasticity and the investment equilibrium equation. The investment rate () is increasing with respect to the ratio of the rate of physical return to capital () and its user cost (). The user cost is the resulting dual price of investment multiplied by the sum of the depreciation rate and exogenous real interest rate. Investment by destination is used to satisfy the equality condition by being set equal to the investment by origin observations found in the benchmark data. It is also used to calibrate the sectoral capital stocks in the base run. The model is solved over a 20-year time horizon and is checked to confirm that it is homogeneous of degree zero in prices and satisfies the Walras Law.

The model addresses both comparative static and dynamic impacts of trade liberalisation. The dynamic effects so far captured are due to more efficient allocation of capital and labour to sectors over time, as factor supplies grow, and caused by trade liberalisation. In other words, it is the comparative static story of trade liberalisation repeated year by year as factor supplies grow. This channel usually leads to very small impacts. New trade theory has now moved beyond only looking at neoclassical market structures to consider things such as increasing returns to scale, imperfect competition, technology transfers and dynamic links such as those between trade liberalisation and total factor productivity (TFP). The model is extended so as to capture trade-induced TFP increases. There is some literature in South Africa that points to the importance of openness and domestic factors in inducing TFP growth that is used to inform this study. Jonsson & Subramanian (2001), based on econometric evidence, conclude that a one percentage point fall in nominal tariffs raises total factor productivity growth rate by 0.74 percentage points. They also find a role for machinery and equipment investment for TFP growth. In follow-up work, Harding & Rattso (2005) and Rattso & Stokke (2005) emphasise adoption and innovation factors in explaining endogenous TFP in South Africa, and offer econometric evidence supporting this claim. Ferdekke & Vase's (2001) work emphasises domestic factors in explaining TFP growth, highlighting a key role played by the ratio of skilled to unskilled labour for TFP growth. We explore, albeit in an ad hoc fashion, the likely influence of these trade-induced TFP changes on growth and poverty in South Africa.

The dynamic CGE model so far developed enables us to carry out trade policy simulations and work out the impacts on macroeconomic, sectoral and aggregated household categories. However, the model accounts for only representative household categories whereas indicators used for the analysis of poverty and inequality generally use household or individual-level data. Therefore, the CGE model needs to be coupled with microsimulation modelling that is essential in analysis of the distributional impacts of the macro shocks in order to reconcile the use of macro models with distributional impacts analysis. This is done following a two-layered macro–micro technique where the macro and micro modules are linked in a top-down fashion that does not account for the feedback (second-order) effects from the micro component to the macro component of the model. This procedure initially involves obtaining results summarising the effects of trade liberalisation from the sequential dynamic CGE model. In a second step, these results are fed into a microsimulation household model to obtain the predicted household effects. Therefore, one should interpret the results as a first-round (prices and quantities) distributive impact analysis of trade liberalisation. It is important to emhasise here that the CGE model is dynamic and is updated each period. However, we use top-down microsimulation as opposed to full microsimulation.⁷ The latter, as in Annabi et al (2005a), entails full integration of all household data from a representative survey to allow for dynamic changes as the CGE model is updated. The advantage of such a model is that it simultaneously takes into account accumulation effects and would allow for a decomposition of welfare into effects from distribution versus effects from growth. The approach of our paper is to map the dynamic results from the CGE onto the ‘once off’ household data to understand the welfare implications. In future, we will extend the paper to use the integrated full microsimulation.

Existing poverty lines for South Africa are used to calculate various poverty indices that help to characterise poverty.⁸ Non-parametric approaches are used based on the observed distribution of these households in the survey, their sample weights, the number of individuals in the household and their independent characteristics of ethnicity, skill type, and region. We have used the publicly available and efficient software called ‘Distributive Analysis/Analyse Distributive’ for poverty analysis (Duclos et al., 2002). This software allows us to compute many poverty-descriptive indicators. The ones in which we are interested for this particular study are the well-known Foster Greer and Thorbecke measures, which can be summarised thus (see Foster et al., 1984):

where j is a subgroup of individuals with consumption below the poverty line (z), N is the total sample size, y is expenditure of a particular individual j and α is a parameter for distinguishing between the alternative Foster Greer and Thorbecke indices.⁹ A distinct advantage of the approach used here – that of accounting for all households in a survey – is that we are able to translate the policy effects to national poverty effects. The policy relevance of the specific analytic tools used lies in being able to capture simultaneously the economy-wide effects and identifying with precision the distributive effects of trade policy, which is a crucial input when responding to losers and winners from the policy reform and when designing complementary policies to compensate those suffering losses.

We run two sets of scenarios that are assumed to commence in 2008. The two scenarios are as follows:

1. Unilateral trade liberalisation: The core simulation for this paper is a unilateral trade liberalisation involving a complete removal of all import tariffs. No dynamic trade-induced TFP increase is assumed.

2. Unilateral trade liberalisation coupled with dynamic trade induced TFP increases: This simulation is similar to the first but includes TFP effects induced by trade liberalisation.

In both simulations, the assumption made is that government budget equilibrium is arranged by an endogenous uniform increase in indirect taxes through the Euler price equations. Alternative compensatory tax mechanisms – direct income tax, sales tax and value-added tax – could also be used. An adjustment variable is introduced in the investment demand functions to handle savings-investment equilibrium. As pointed out in Annabi et al. (2005a), it is important to note that in dynamic analysis the economy is growing even without a shock. As a result, the relevant counterfactual to compare the results with is this ‘business as usual’ (BAU) growth path, unlike in static CGE analysis where the relevant counterfactual is the base year Social Accounting Matrix (SAM).

4. Results

4.1 Impact of a unilateral trade liberalisation

Table 1 summarises the macroeconomic effects of a full trade liberalisation scenario without including dynamic trade-induced productivity gains. Immediately we can see that trade liberalisation has a very small effect on the macroeconomy, an observation that is consistent with the observation that South Africa already has very low import tariffs so that their removal will not have major impacts on the economy. Taking 2009 as the short run, Table 1 shows that trade liberalisation increases GDP by only 0.02% in the short run and leads to small but positive increases in GDP over the rest of the policy period (2010–20) due mainly to accumulation effects. The minor short-run contraction in 2008 is explained by the contraction in previously highly protected sectors induced by increased import competition when the period is too short for capital to have relocated to the expanding export intensive sectors.¹⁰

Table 1: Macroeconomic effects of trade liberalisation (% change from BAU path)

	GDP path	Consumption	Investment	Exports	Imports	Skilled wage	Unskilled wage	CPI path	Price of capital	User cost of capital
2007	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
2008	–0.01	–1.75	0.67	2.74	2.87	–1.87	–1.80	–1.53	–2.61	–1.90
2009	0.02	–1. 63	–0.67	2.96	2.98	–1.76	–1.65	–1.046	–2.34	–1.84
2010	0.05	–1.53	0.69	3.14	3.08	–1.67	–1.53	–1.41	–2.06	–1.79
2011	0.07	–1.44	0.71	3.30	3.17	–1.58	–1.43	–1.36	–1.86	–1.76
2012	0.1	–1.37	0.73	3.44	3.25	–1.51	–1.33	–1.32	–1.72	–1.73
2013	0.12	–1.30	0.74	3.56	3.32	–1.45	–1.25	–1.28	–1.63	–1.70
2014	0.14	–1.24	0.74	3.66	3.37	–1.39	–1.17	–1.25	–1.57	–1.68
2015	0.17	–1.19	0.75	3.74	3.42	–1.33	–1.11	–1.23	–1.53	–1.66
2016	0.18	–1.14	0.75	3.82	3.46	–1.29	–1.05	–1.21	–1.52	–1.65
2017	0.20	–1.10	0.76	3.88	3.50	–1.25	–1.00	–1.20	–1.51	–1.64
2018	0.22	–1.07	0.76	3.93	3.53	–1.21	–0.96	–1.19	–1.51	–1.63
2019	0.23	–1.04	0.76	3.98	3.56	–1.18	–0.93	–1.18	–1.51	–1.62
2020	0.24	–1.02	0.75	4.01	3.58	–1.15	–0.90	–1.18	–1.51	–1.62

Note: CPI, Consumer Price Index.

Both the rental and the user cost of capital decline in both the short run and the long run, but the rental return to user cost ratio increases in the long run. As a result, we notice that full trade liberalisation leads to growth in investment by destination, with the long-run response being stronger than the short-run response. Similarly, the trade liberalisation-induced decline in domestic import prices leads to an increase in imports in the short run and the long run. The consumer price index also falls in the short run and the long run in response to reduced production costs made possible by lowering of tariffs. This coupled with the ensuing decrease in domestic costs of production and the real exchange rate depreciation induces exports to increase in the short run and the long run. Exports grow more than imports in the long run. Because of the volume movement in exports and imports, sales on the domestic market fall. Both skilled and unskilled wages decline throughout the period following reduced demand for labour from the contracting labour-intensive sectors. The short-run contraction is more severe than the long-run contraction since in the long run capital will have reallocated to the more efficient sectors compared with the short run. Also, unskilled wage rates contract much less than skilled wages. In line with GDP developments, welfare as measured by the dynamic equivalent variation also falls initially in the short run but increases thereafter. These welfare changes are consistent with the fall in consumer price index being less than the fall in consumption in the short run, while the fall in consumption in the long run is less than the fall in consumer price index.

The impact of trade liberalisation on poverty is captured by changes in the poverty indices reported in the last column of Table 2. The changes in poverty are largely in line with the changes in welfare. This is because the changes are largely driven by changes in the consumer price index and changes in household income or consumption. The impacts on poverty are very small. Using the percent change in average headcount index of poverty measure, the results in Table 2 suggest that a unilateral removal of tariffs has a very small but negative impact on poverty headcount. The burden of these negative impacts is shared almost evenly between urban and rural households. Indian unskilled households, in particular rural Indian households, shoulder a disproportionate amount of the poverty burden. This is largely because of their higher dependence on employment in textiles, the sector that faced the highest protection before the trade policy intervention. The average poverty gap and the squared poverty gap also follow a similar pattern.

Table 2: Impact of trade liberalisation on poverty (% of BAU)

	Short run = 2009			Long run =2020
	P0	P1	P2	P0	P1	P2
South Africa	0	0.55	0.83	–0.19	–2.15	–2.65
Residential area
Urban	0.0014	0.17	0.29	–0.01	–1.58	–1.98
Rural	0.0013	0.2	0.38	–0.52	–4.41	–4.65
Ethnic group
African	0.009	0.12	0.19	–1.1	–1.3	–1.66
Coloured	0.001	0.25	0.26	–5.2	–5.54	–6.1
Indian	0.026	0.19	0.43	–0.6	–1.82	–1.97
White	0.01	0.15	0.33	0	0	0
Region, ethnic and skill group
Urban African skilled	0	0	0	0	0	0
Urban coloured skilled	0	0	0	0	0	0
Urban Indian skilled	0	0	0	0	0	0
Urban white skilled	0	0	0	0	0	0
Urban African unskilled	0.007	0.22	0.45	–1.43	–1.44	–1.78
Urban coloured unskilled	0.001	0.54	0.84	–1.27	–2.35	–2.98
Urban Indian unskilled	0.03	0.55	0.65	–0.2	–0.67	–0.85
Urban White unskilled	0	0	0	0	0	0
Rural African skilled	0	0	0	0	0	0
Rural coloured skilled	0	0	0	0	0	0
Rural Indian skilled	0	0	0	0	0	0
Rural white skilled	0	0	0	0	0	0
Rural African unskilled	0	0	0	–1.34	–3.59	–3.92
Rural coloured unskilled	0	0	0	–1.4	–6.22	–7.05
Rural Indian unskilled	0.0031	0.72	0.78	0	0	0
Rural white unskilled	0.006	0	0	0	0	0

The picture reverses in the long run, with the incidence of poverty declining for the whole country by about 0.19%, which is still quite small. The reduction in poverty is as a result of the static and dynamic efficiency gains from trade liberalisation as well as accumulation effects. The main beneficiaries of reduced poverty are coloured households, followed by African households. Both supply a higher proportion of their labour endowment to the mining sector and other tradable sectors. They also consume disproportionately more foodstuffs whose cost has been reduced by trade liberalisation. Indian households also experience reductions in poverty, but by a relatively smaller margin. Rural households benefit more than urban households, given their higher dependence on the booming mining sector.

4.2 Impact of a unilateral trade liberalisation with TFP increases

As argued above, the impacts of trade liberalisation on the economy have tended to be very small, even after allowing for dynamic effects emanating from factor accumulations through time. One rationalisation used for this result was that the country has already reaped the gains from trade given that the country has undergone substantial trade liberalisation since 1994. In line with modern trade literature, we wish to explore in this section whether dynamic trade-induced TFP changes may lead to ‘bigger numbers’ from trade liberalisation.

According to Table 3, removing all tariffs under the assumption of trade-induced TFP increases has very pronounced and beneficial effects compared with trade liberalisation without productivity gains. We see that factoring TFP gains will raise GDP from about 1% in 2009 to over 6% in 2020. This in turn will positively impact on incomes, which in turn raises savings and consequently investment. Private consumption rises sharply compared with the scenario of no TFP change. The increase in GDP feeds into increased consumption both in the short run and the long run. The capital good price rises in the short run before falling in the long run. However, because of TFP increases, the user cost of capital falls from 2009 until 2020. Because of the rising rental to user cost of capital ratio coupled with the higher induced savings, there is a boom in investment by destination, with the long-run response being stronger than the short-run response. Imports increase dramatically not only due to the cost reducing effects of tariff cuts but also because TFP-induced growing economy requires a higher level of imports to meet higher production levels and increased household demands. Indeed imports rise much faster than exports in the short run, in part due to an induced real exchange rate appreciation. In the long run, exports grow more than imports. The consumer price index increases initially in the short run before declining in the long run. Skilled and unskilled wages increase in both periods following increased demand for labour to meet higher growth needs. Welfare rises dramatically in line with the observed consumer price index and consumption developments. Finally, trade-induced TFP increases and accumulation effects lead to reductions in poverty, both in the short run and the long run.

Table 3: Macroeconomic effects of trade liberalisation with induced TFP effects (% change from BAU path)

	GDP path	Consumption	Investment	Exports	Imports	Skilled wage	Unskilled wage	CPI path	Price of capital	User cost of capital
2007	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
2008	–0.06	8.52	17.50	3.36	10.32	7.87	8.65	0.35	12.94	1.27
2009	0.91	9.34	18.86	4.63	11.03	9.41	9.98	0.15	7.93	–0.93
2010	1.88	10.29	18.96	6.14	11.62	10.94	11.48	–0.17	5.60	–1.87
2011	2.76	11.17	18.57	7.63	12.07	12.31	12.85	–0.50	3.93	–2.43
2012	3.53	11.93	17.99	9.00	12.40	13.49	14.06	–0.79	2.55	–2.79
2013	4.18	12.58	17.36	10.22	12.66	14.49	15.09	–1.02	1.37	–3.03
2014	4.72	13.12	16.74	11.26	12.85	15.32	15.94	–1.20	0.38	–3.18
2015	5.16	13.55	16.16	12.14	12.98	16.00	16.64	–1.34	–0.44	–3.28
2016	5.51	13.90	15.63	12.87	13.08	16.53	17.19	–1.44	–1.11	–3.33
2017	5.78	14.17	15.15	13.46	13.15	16.94	17.61	–1.51	–1.66	–3.36
2018	5.99	14.37	14.72	13.93	13.18	17.24	17.93	–1.56	–2.09	–3.37
2019	6.15	14.51	14.34	14.30	13.20	17.46	18.16	–1.59	–2.43	–3.36
2020	6.26	14.60	14.00	14.58	13.19	17.61	18.31	–1.60	–2.70	–3.34

Note: CPI, Consumer Price Index.

The trade-induced TFP increase has a more significant impact on poverty reduction than trade liberalisation without induced TFP growth, as shown in Table 4. The poverty headcount ratio falls by 0.54% in the short run and by 5.34% in the long run. Most of the poverty reduction is felt amongst African and coloured households, while urban households benefit less than their rural counterparts from the ensuing fall in poverty. Once again the average poverty gap and the squared poverty gap also follow a similar pattern to the headcount ratio.

Table 4: Impact of trade liberalisation with induced TFP effects on poverty (% of BAU)

	Short run = 2009			Long run =2020
	P0	P1	P2	P0	P1	P2
South Africa	–0.54	–0.2	–1.3	–5.34	–4.3	–7.78
Residential area
Urban	–0.33	–0.15	–1.67	–5.22	–5.16	–6.45
Rural	–0.56	–0.46	–1.34	–7.46	–6.6	–8.3
Ethnic group
African	–1.15	–1.3	–2.2	–6.44	–5.42	–7.7
Coloured	–1.31	–0.78	–1.8	–7.92	–7.2	–8.2
Indian	–0.5	–0.1	–1.45	–5.77	–4.32	–7.2
White	–0.21	–0.1	–1.78	–4.25	–3.21	–5.4
Region, ethnic and skill group
Urban African skilled	0	0	0	0	0	0
Urban coloured skilled	0	0	0	0	0	0
Urban Indian skilled	0	0	0	0	0	0
Urban white skilled	0	0	0	0	0	0
Urban African unskilled	–0.7	–0.4	–1.45	–9.36	–11.4	–12.3
Urban coloured unskilled	–1.2	–0.54	–1.84	–9.9	–12.8	–16.6
Urban Indian unskilled	–0.12	–0.55	–1.65	2.3	–3.8	–5.6
Urban white unskilled	–0.13	–1.62	–1.92	–1.34	–3.3	–8.8
Rural African skilled	0	0	0	0	0	0
Rural coloured skilled	0	0	0	0	0	0
Rural Indian skilled	0	0	0	0	0	0
Rural white skilled	0	0	0	0	0	0
Rural African unskilled	–1.4	–1.3	–1.93	–10.52	–9.98	–11.2
Rural coloured unskilled	–1.86	–2.3	–2.23	–10.19	–8.87	–12.15
Rural Indian unskilled	–0.87	–0.72	–1.78	–4.19	–2.4	–6.4
Rural white unskilled	–0.67	–0.57	–1.83	–3.23	–3.2	–5.6

4.3. Impact of a unilateral trade liberalisation reversal

It is important to point out that there has been some shift in South Africa's trade policy recently. Following the aggressive trade liberalisation agenda in the 1990s that even exceeded the country's commitments under the Uruguay Round, this trajectory has been reversed. More recently a slow down (and tariff increases for key products such as clothing and textiles) reflects a much more inward-focused stance. What are the implications of this reversal? To assess these effects of such protection, in Mabugu & Chitiga (2010) we followed the top-down approach and used a nationally representative household survey to estimate the poverty effects of increasing protection on textile and clothing and assumed that there are no other changes in tariffs for other sectors. The underlying assumption under this simulation is that protectionist political pressures are sector specific so that we could analyse what would happen if a single sector succeeds in obtaining protection, with no other changes in trade policy. We also assumed that South Africa is a small, open economy and that there will not be retaliatory moves by other countries in response to the increase in textile protection. Poverty headcount increases in the short run. As expected, the protected sector experiences positive growth, taking along with it sectors that are highly dependent on the protected sector for selling its output. Other sectors, especially the export-oriented sectors, experience a contraction of output. In the long run, increased protection reduces investment and, as a result, capital accumulation. Wages fall compared with the short run while consumer prices continue to grow. More people are pushed into absolute poverty by increased protection both in the short run and the long run. These results suggest that the calls for protectionist pressures in textiles and poultry are understandable from the perspective that the sectors gain if they are able to achieve protection while other sectors do not simultaneously lobby for such support at home and abroad. In the short run, a limited number of households are actually pushed out of poverty, although this is not sufficient to reduce overall poverty. The move raises real costs of production for other sectors, especially the export-oriented sectors, which contract both in the short run and the long run. The measure reduces capital accumulation and GDP in the long run while prices continue to rise. The consequence is that welfare declines and poverty increases especially for African households who depend disproportionately more on employment in exportable sectors than other households.

5. Conclusion and policy discussion

South Africa has undergone significant trade liberalisation since the end of apartheid. Average protection has fallen while openness has increased. The macroeconomic performance in this era of liberalising trade has been unimpressive, with GDP growing by insufficient amounts to make inroads into the high unemployment levels. Poverty levels have also risen. In recent years, the debate concerning what direction South African trade policy should take has been increasingly reflecting a much more inward-focused stance. A more active role for the state in the form of protection is believed to encourage selected industries to grow their ability to export to a degree to which they will then be able to compete effectively in global markets. This argument presents an interesting case of how a country might use safeguard trade barriers in order to better achieve a domestic policy objective. However, our analysis suggests that trade protection is an inefficient means to achieve a domestic policy outcome. This is because such policies rarely give consideration to the impact that the policy has on other sectors of the economy. Much of the current discourse on textiles and poultry protection focuses on static effects of protection. The dynamic focus as in this paper is crucial in this debate because it enables us to capture not only the static effects of the policy but also its growth and accumulation effects that are shown to be less favourable. The study advances existing CGE work in South Africa in at least two ways. Firstly, it uses top-down approach-based household survey data to model explicitly poverty effects of policy. Secondly, it employs a sequential dynamic CGE model to carry out the sequential ‘top-down’ poverty microsimulation.

Trade liberalisation alone has very minimal short-run macroeconomic consequences. The outcome for the long-run macroeconomic developments is positive for tariff removal although the magnitude of the impacts is still very small. The sectoral results indicate that sectors which initially faced high protection levels tend to be the ones to lose out disproportionately more from trade liberalisation. The biggest winner is mining while the biggest loser is textiles. The picture reverses when we allow for trade-induced TFP increases, with bigger and positive impacts on the macroeconomy. Mining is no longer the main beneficiary of the reform. The welfare outcomes are initially negative in the short run but turn positive if we allow for trade-induced TFP increases. The welfare gains are positive in the long term in all scenarios. Although all households benefit in the long run, African and coloured poor households in general – and especially those residing in rural areas – reap the most benefits. In terms of poverty, trade liberalisation has no appreciable impact on poverty in the short run even if we allow for trade-induced TFP increases. In the long run, however, poverty reduces even in the case when we do not allow for TFP increases. Again, African and coloured households gain the most in the long run in terms of numbers being pulled out of absolute poverty, especially if the trade measure was to induce TFP increases. Trade liberalisation policy has therefore been found to be progressive – highest gains accrue to the poorest groups – despite the low level of tariff protection remaining in South Africa.

The policy relevance of the specific analytic tools used lies in being able to capture simultaneously the economy-wide effects and identifying with precision the distributive effects of trade policy, which is a crucial input when responding to losers and winners from the policy reform and when designing complementary policies to compensate those suffering losses. For instance, the findings of the paper imply that trade policy would need to be coupled with other policies that address poverty, fiscal outcome, and so on. It is thus a key area for government intervention. The question is whether the choice of focus areas is in line with the desired growth path. There is also the issue of skills shortages, which prevents the majority of the population from participating in and sharing the growth fruits. Some useful policy conclusions emerge from these results. Without exception, there is still substantial scope to lower prices and raise household welfare through stronger unilateral tariff liberalisation. However, there is an asymmetry in the timing of the welfare gains that can only be picked by dynamic analysis. If trade liberalisation induces TFP increases, the gains are magnified. These results point to a future role for trade policy in South Africa. They suggest that short-term temporary measures such as transfers to poor households may be justified to ameliorate the transitory negative effects on the poor before the long-term gains are realised. They also suggest that measures should be put in place so as to increase the chances that future tariff cuts generate substantial TFP growth. Such measures could include training programmes.

One possible extension of this paper is to integrate full household data into the CGE model and carry out a fully dynamic, welfare analysis. This would allow for an understanding of the contribution of distribution effects as compared with growth effects.

Notes

³While trade policy can have impacts on poverty, the main anti-poverty response in South Africa is in the form of social policy interventions that are delivered through provision of basic and free basic services, subsidies as well as assets that aim to benefit the poor. The package contributing to social wage includes free primary healthcare, no-fee-paying schools, social grants (such as old age pensions, and child support grants), free housing for indigents, provision of basic and free basic services in the form of reticulated water, electricity, sanitation and sewerage particularly for those categorised as indigent.

⁴These are usually based on drawing or identifying groups of micro units exposed to the policy reform being studied (‘treated’ groups) and groups of individuals not exposed to the reform (control groups).

⁵For a recent and concise review of trade-focused CGE modelling in South Africa, see Mabugu & Chitiga-Mabugu (2009).

⁶Trade policy is guided by multilateral arrangements as well as by bilateral and regional agreements. The Southern African Customs Union between South Africa, Botswana, Lesotho, Namibia, and Swaziland is the oldest Customs Union in the world. There are two Free Trade Areas between the European Union and the Southern Africa Development Corporation that the country has so far concluded. The country also benefits from the US African Growth and Opportunity Act. There are planned Free Trade Areas with India, the USA and MERCOSUR countries.

⁷For a review of different models on trade, see Mabugu & Chitiga-Mabugu (2009); and for a review of microsimulation models, see Davies (2004).

⁸Some examples are the Foster Greer and Thorbecke index, Watts's index, and the Clark, Hemming and Ulph index.

⁹When α = 0 the expression simplifies to J/N, or the headcount ratio. This is a measure of the incidence of poverty. When α = 1 the expression gives us poverty depth measured by the poverty gap. When α = 2 the expression gives us the severity of poverty measured by the squared poverty gap.

¹⁰Annabi et al. (2005b) find a similar effect in a study on Bangladesh.