The impact of tourism on poverty in South Africa

Abstract

This paper evaluates the potential impact of tourism on poverty in South Africa on the basis of recent survey data on international tourism spending patterns. It looks at three scenarios, using an applied general equilibrium model. The main finding is that the poor benefit very little in the short term from additional tourism income. A further finding is that domestic and international tourist expenditure affect the economy differently; both markets are therefore important. In essence, the research confirms that tourism receipts can be used as a tool to alleviate poverty, but in South Africa this must be supported by policies that focus on the labour market and human resource development.

Keywords:

• tourism
• poverty
• applied general equilibrium (AGE) modelling
• South African tourism industry

JEL classification:

• C68
• D31
• D58
• I32
• L83
• O24
• O50

1. Introduction

The South African tourism industry is valued at approximately US$10 billion a year, approximately 7% of GDP (Snyman, 2007). According to Rossetto et al., the potential for the tourism sector to help alleviate poverty has recently come under the spotlight of international and national tourism and development agencies (2007:49). However, apart from studies of specific projects and programmes that indicate how this industry can help (for example Blake et al., 2008), there is little economy-wide research evidence to suggest that tourism does reduce poverty and few studies that quantify the interactions between tourism and poverty (see Blake et al., 2008:108). Consequently, this paper seeks to offer some insight into who loses and who gains in South Africa when tourism expenditure increases.

Our findings show that in the short term the poor benefit very little, if at all, from additional tourism inflows to South Africa, and, most disturbingly, that a 10% increase in tourism would bring no significant benefit to the very poor, i.e. the lowest income households. This paper sheds some light on the significance of these findings by discussing some alternative scenarios.

2. Literature review

Until recently, the role of tourism in poverty alleviation has attracted only a small cohort of researchers within tourism studies, and the existing literature consists mainly of case studies (see for example Barling, 2005; Chok et al., 2007; Muhanna, 2007). In addition, related research has come from a diversity of sources, but few of these studies have contributed to the methodological development of this emerging field of research. Exceptions are Wanhill (1994), Adams & Parmenter (1995), Blake (2000), Dwyer et al. (2000) and Blake & Sinclair (2003). There is thus an urgent need for a more systematic, comprehensive and coherent approach to guide the enquiries in this field (Zhao & Ritchie, 2007:121).

Traditionally, the primary aim of local tourism development was seen as regional economic growth, while poverty alleviation was considered either a sub-goal or a natural outcome of growth (Deloitte et al., 1999; Ashley et al., 2000). A commonly held belief is that as long as the whole region gets wealthier, the benefits brought by economic growth will eventually trickle down to the local poor through multiple channels, such as employment, public welfare, health care and family networks (Zeng et al., 2005). This does not always seem to be the case. The benefits of tourism and spillovers to the poor depend on which model of tourism development is chosen (Christie, 2002). There are a number of examples, especially in developing countries, where big corporations enclose the products they invest in, with the result that tourists see little of the country around them. The tourist spending remains in the hands of a few, with high leakages to major corporations in developed countries (Saayman, 2007). McCulloch et al. estimate that between 55% and 75% of tourism spending leaks back to developed countries (2001:248). The leakage of foreign currency, particularly through imports, has long been recognised in the economic impact literature (see reviews by Fletcher, 1989; Wanhill, 1994; Archer, 1996). In most of these countries, which are mainly in sub-Saharan Africa, receipts are less than 5% of GDP (World Bank, 2005; WTO, 2005). Notable exceptions worldwide are Cambodia (10.4%), Eritrea (11.6%), the Gambia (18.6%), and Mongolia (12.1%). Tourism may increase government revenues, but the distribution of this income is uncertain (Blake et al., 2008:113). In a case where tourism revenue is used for purposes other than tourism, the local economy and the country do not necessarily benefit from tourism development. In fact, in most cases the host country has to pay a price in terms of additional costs or a negative social impact. However, as Christie (2002) observes, tourism can address poverty, not as a panacea but as a useful tool in the toolkit.

The idea of pro-poor tourism has recently been advanced as a solution to the leakage problem (Carbone, 2005). The idea is that the tourism sector can generate net benefits for the poor. Some of the benefits will be direct, such as an increase in individual wages through employment in the formal sector. Others will be indirect, such as improvements in roads, water and infrastructure, better levels of education and health and protection of the environment. Carbone (2005) argues that developing the tourism sector may be more conducive to pro-poor growth than developing other sectors because of greater multiplier effects and the opportunities the informal sector creates for unskilled workers and women. The question then becomes one of linkage rather than leakage.

The confluence of tourism and poverty, which were previously two separate domains (Bowden, 2005), reflects an essential change in the philosophy of tourism development and poverty alleviation. Tourism and poverty alleviation, according to Scheyvens (2007), are increasingly being linked. This is echoed in most developing countries' policy documents concerning economic growth – where tourism is seen as a tool for addressing poverty. According to a WTO report on the economic benefits of tourism, there are a number of ways to increase the overall economic benefits of tourism and thereby alleviate poverty, such as:

•	increasing tourist length of stay and spending,
•	spreading tourism development and its benefits geographically,
•	establishing stronger inter-sectoral linkages,
•	maximising employment of nationals in tourism,
•	encouraging locally owned and managed tourism enterprises, and
•	optimising tax revenues from tourism. (WTO, 2002:2)

If policy makers are to be convinced of the need to put in place initiatives to increase tourist spending, build inter-sectoral linkages and encourage local entrepreneurs, they will need evidence of the economy-wide impact on poverty of these drivers of the economic benefits of tourism, rather than just specific cases.

3. Poverty and tourism in South Africa

3.1 Poverty and inequality

South Africa is under pressure to increase economic growth and reduce poverty and inequality. When the first democratically elected ANC-led government came to power in 1994, approximately 58% of the population and 68% of blacks lived in poverty. A Gini coefficient of 0.56 showed South Africa to have one of the most unequal income distributions in the world (Özler, 2007). The new South Africa suffered from vast inequalities in education, health and infrastructure. At that time, government adopted the Reconstruction and Development Programme as its socio-economic policy framework and spelt out the requirements for delivery, including meeting basic needs and developing human resources. Since then, great strides have been made in redressing past social inequalities and realising the Reconstruction and Development Programme commitment.

When analysing what has been done and what is still do, we are somewhat restricted by the availability of data, but the literature is clear that much remains to be done. Leibbrandt et al.'s examination (2006) of the 1996 and 2001 Census data shows that inequality increased during the intervening five years, with the aggregate Gini coefficient increasing from 0.68 to 0.73. Inequality between the racial groups decreased slightly, but within them it increased significantly. Using income data from the 1995 and 2000 Income and Expenditure Surveys, Leibbrandt et al. (2005) find that real individual incomes deteriorated over the period. Using the same dataset, Özler (2007) finds that in 2000 more than one in three individuals were living on less than US$2 per day – that is R174 per capita per month in 2000 prices. Between 1995 and 2001, the annual per capita growth rate of household expenditures was 0.5% and expenditures at the bottom end of the distribution decreased. Changes in overall inequality were insignificant but there was an increase in inequality among the black population (Özler, 2007).

To offset this gloomy picture, however, some successes can be observed for this five-year period. Significant increases were observed in the mean expenditures of the coloured population, and their poverty headcount went down from 50% in 1995 to 35% in 2000. While there was no change for Asians and whites, there was at least no increase in poverty for these groups. It is also heartening to see that three provinces, the Western Cape, the Northern Cape and the Free State, experienced significant declines in poverty. The poverty headcounts also went down for all population groups in the Western and Northern Cape (Özler, 2007).

It is, however, clear that significant challenges remain. Özler (2007) ascribes the lack of progress against poverty to the low economic growth rate and the fact that what growth did materialise was not pro-poor.

3.2 Pro-poor tourism

It has been suggested that local economic development is ‘one of the most important post-apartheid development options’ (Rogerson, 2006:39). Local governments are responsible for the economic development of their localities, with a specific focus on the poorest members of society. Typical pro-poor interventions include improved regulatory frameworks, delivery of basic services, employment creation through public works programmes and the stimulation of local economic activities. Many South African small towns are also turning to the opportunities tourism offers. A particular focus is pro-poor tourism, which was showcased at the World Summit on Sustainable Development held in Johannesburg in 2002 and emphasised by the 2004 Washington Declaration on Tourism as a Sustainable Development Strategy (WTO, 2004). Rogerson (2006) outlines the ways that tourism can help reduce poverty.

The South African policy environment also emphasises the connection between tourism, poverty alleviation and local economic development. From early on, the tourism industry was seen as an engine of growth and poverty reduction, with key policy documents such as the 1996 White Paper on the Development and Promotion of Tourism, Tourism in GEAR and the Responsible Tourism Handbook: A Guide to Good Practice for Tourism Operators providing guidelines and targets (Rogerson, 2006). An example of a guideline for prioritising opportunities for local communities is that tourism businesses should monitor the proportion of goods and services that the enterprise sources from other businesses within a 50 km radius and aim for a 20% increase in local sources over three years. Another is that they should set targets for increasing the proportion of staff hired from, or the proportion of the enterprise's wage bill that goes to, communities within 20 km of the enterprise, or both. Commitments have been made that South African Tourism, by building partnerships with the private sector, will seek to ensure that 40% of all tourism spending in South Africa is for the benefit of previously disadvantaged groups, particularly women (Rogerson, 2006:49). Taken together, local economic development initiatives and tourism look set to make crucial contributions to poverty alleviation in South Africa.

To date, studies of the impact of tourism on poverty alleviation have been local rather than national. Nel & Binns (2002), Abrahams (2003), Binns & Nel (2003), Nel (2005) and Saayman & Rossouw (2011) present case studies of small towns that have attempted tourism-led development. Ndlovu & Rogerson (2004) examine rural community-based tourism projects. Saayman & Saayman (2010) analyse the influence of South Africa's national parks on regional development. Rogerson (2002, 2004) examines local economic development efforts by small towns and rural communities that have linked various attractions into themed routes for tourists. Hill et al. (2006) and Munthali (2007) look at nature-based tourism as a pro-poor development intervention. And studies by Saayman & Saayman (2006), Saayman et al. (2008) and Saayman & Rossouw (2010) examine the economic impact of events on local economies in South Africa.

4. Evaluating the economy-wide impact of tourism for the poor

4.1 The model

In this study we used an applied general equilibrium (AGE) model, which is ‘an economy-wide model that includes feedback between demand, income and production structure, and where all prices adjust until decisions made in production are consistent with decisions made in demand’ (Dervis et al., 1985:132). AGE models can be used to assess the effects on the economy of a specific year of shocks, such as a change in economic policy settings (e.g. tariffs on imports) or changes in international trading conditions (e.g. an expansion in international tourism). These models are therefore ideally suited to capturing the wider effects of tourism expenditure (Gillham, 2004). Moreover, AGE models are now well known in policy modelling and have, since 1993, been used with increased frequency in South Africa (see for example Naudé & Coetzee, 2004).

The model is applied (or computed) using economy-wide consistent data on a particular economy as is normally contained in a social accounting matrix (SAM). For this study we used the SAM for South Africa that had most recently been published at the time of writing (Stats SA, 2002, 2004). A South African adaptation of ORANI-G was used to solve the model. It is known as the ‘UPGEM’ and was developed for South Africa by the University of Pretoria. The UPGEM model used in these simulations distinguishes 32 sectors, six household types and four ethnic groups (Horridge, 2000).1

4.2 Structure of tourism in the South African social accounting matrix

An important consideration when modelling a tourism expansion (or recession) and formulating the simulations to be modelled is to consider the current tourism structure in South Africa based on the framework used in the AGE model. Using the South African Domestic Tourism Survey (Rule et al., 2001) and the latest South African SAM (see Stats SA, 2002, 2004), we were able to estimate the share of tourist expenditure in South Africa (as a percentage of domestic household expenditure). Table 1 shows the percentage of spending by tourists (domestic and foreign) on leisure and holidays. At provincial level, tourism in KwaZulu-Natal (33%) and the Western Cape (31%) are the largest contributors to overall tourism in South Africa.

Table 1: Percentage spending by tourists (domestic and foreign) on leisure/holiday, 2001

Province	Transport	Accommodation	Food	Entertainment	Gifts	R million	Weights
Eastern Cape	34	30	20	13	3	809	12.5%
Free State	48	24	21	6	1	106	1.6%
Gauteng	37	23	26	10	4	490	7.6%
KwaZulu-Natal	26	32	26	10	6	2130	32.9%
Mpumalanga	33	30	29	5	3	358	5.5%
Northern Cape	32	32	24	7	5	91	1.4%
Northern Province	31	40	25	3	1	272	4.2%
North-West	66	18	10	4	2	251	3.9%
Western Cape	32	29	26	9	4	1976	30.5%
RSA	32.2	29.7	24.6	9.2	4.2	6483	100%
Source of data: Rule et al. (2001).

These figures were further adjusted to more recent tourism data obtained from the South African Tourism Strategic Research Unit (SAT, 2007), as shown in Table 2. This table shows that transport (32%), accommodation (30%) and the food industry (25%) are the dominant tourism sectors in South Africa. Of the 94 commodities in the South African SAM (aggregated to 32 sectors in the UPGEM), 29 are classified as being tourism related, and both foreign and domestic tourism are classified as being expenditure on these 29 commodities (see Table 2). On the basis of the results shown in Table 2, it can be expected that the services industry will gain most from an inbound tourism expansion and lose most if there is a recession. On the other hand, the natural resources and manufacturing sectors will probably be the least affected by a recession.

Table 2: Estimated share of tourist expenditure in the South African SAM

Sectors	Industry type	Share based on domestic household expenditure	Tourist spending on major items					Final estimated share of tourist expenditure
			Transport (32.2%)	Accommodation (29.7%)	Food (24.6%)	Entertainment (9.2%)	Gifts (4.2%)
Agricultural products	Food	3.32%			1			2.17%
Meat products		7.72%			1			5.03%
Fish products		0.89%			1			0.58%
Fruit and vegetable products		1.41%			1			0.92%
Oils and fats products		1.14%			1			0.74%
Dairy products		2.43%			1			1.59%
Grain mill products		3.39%			1			2.21%
Bakery products		2.01%			1			1.31%
Sugar products		0.77%			1			0.50%
Confectionery products		0.78%			1			0.51%
Other food products		1.35%			1			0.88%
Beverages and tobacco products		12.60%			1			8.21%
Wearing apparel	Hardware	6.35%					1	2.54%
Handbags		0.15%					1	0.06%
Footwear		2.08%					1	0.83%
Published and printed products		1.15%					1	0.46%
Recorded media products		0.18%					1	0.07%
Petroleum products	Chemicals	5.63%	1					9.14%
Pharmaceutical products		3.15%		1				4.27%
Soap products		3.67%		1				4.98%
Rubber tyres	Spares	0.84%	1					1.36%
General hardware products		0.31%					1	0.12%
Motor vehicle parts		1.14%	1					1.84%
Jewellery		0.37%					1	0.15%
Accommodation	Services	4.62%		1				6.26%
Transport services		6.87%	1					11.15%
Communications		5.39%	1					8.76%
Financial services indirectly measured (FSIM) + insurance services		10.47%		1				14.20%
Other services / activities		9.80%				1		9.16%
Sum of household expenditure		299 539	19.86%	21.91%	37.83%	9.80%	10.60%	100.00%
Ratio adjusted			1.62	1.36	0.65	0.93	0.40
Source of data: Authors' calculations based on Rule et al. (2001) and Stats SA (2004).

4.3 Simulation set-up

The main issues examined in the model are the economic impacts and distributional effects of tourism expenditure. This paper is restricted to projecting the short-term comparative static effects.2 This implies that the impact reflects the change during a short time (approximately two to three years) before investment can react to the changed market conditions. According to tourism statistics of the South African Tourism Strategic Research Unit, the average annual growth rate of inbound tourism is around 10% (SAT, 2007). Thus the model assumes a 10% inbound tourism expansion in the first scenario. The next scenario assumes a 20% inbound tourism expansion as an ideal tourism strategy for the South African Government. Both simulations assume that tourism expands uniformly in all South Africa's provinces.

However, recent events such as the ongoing electricity crisis, increased levels of violent crime, the Zimbabwean crisis, a looming fuel shortage and rising oil prices have caused inbound tourism to be volatile. Consequently, the final simulation considers the effects of an inbound tourism recession. As a worst case scenario, the simulation assumes a 10% inbound tourism reduction in all regions. In this study we therefore implemented three simulations as indicated above with fixed supplies of all primary factors suggested by the neoclassical model closure.

All results presented in Tables 4 to 11 are in the form of percentage changes from the base case scenario. The simulation results are discussed in three groups: macro-economic results, sectoral results and household results. For simplicity and concise discussion, the focus is only on the simulation results of scenario 1. The simulation results of scenarios 2 and 3 are compared with those of scenario 1.

To implement the simulations, a number of explicit assumptions must be made about the macro-economic environment in which the additional tourism expansion is assumed to take place. This principally involves the selection of a set of exogenous variables, conventionally referred to as the closure of the model (Wattanakuljarus, 2006). An in-depth discussion about the closure of AGE models can be found in Horridge (2000). In the present case, the simulations were done using a modified short-term comparative static closure for the model (see Table 3). The following variables were assumed to be unaffected by the additional expansion of tourism: real wages, the rate of return on capital, real government consumption, and the public sector borrowing requirement. Moreover, it was assumed that there is no excess capacity in tourism facilities. Finally, the addition to the international tourist intake was assumed to exhibit expenditure patterns and a national distribution identical to those of the existing intake. The results of applying the shocks to tourism implied in scenarios 1, 2, and 3 in Table 3 are discussed in Section 4.4 below.

Table 3: Differences between scenarios 1, 2, and 3

Scenarios	Model	Policy interventions incorporated	Data used	Conclusions	Closure	Variable(s) shocked
Scenario 1	UPGEM03, based on UPGEM02; Van Heerden et.al. (2006)	Simulation experiment to evaluate the economy-wide impacts on South Africa of a 10% inbound tourism expansion with fixed supplies of all primary factors	Simulation experiments and alternative tourism expenditure scenarios; based on tourism statistics of the South African Tourism Strategic Research Unit	A 10% inbound tourism expansion leads to a 0.15% increase in real GDP, a 0.3% increase in total employment, a positive trade balance, and negligible changes in productivity	Modified DPSV standard short run	f4q(‘agriculture,… , CommunServs’) = shocks applied based on expenditure patterns and distribution in Section 4.2, to simulate a 10% inbound tourism expansion
Scenario 2	UPGEM03, based on UPGEM02; Van Heerden et.al. (2006)	Simulation experiment to evaluate the economy-wide impacts on South Africa of a 20% inbound tourism expansion with fixed supplies of all primary factors	Simulation experiments and alternative tourism expenditure scenarios; no data required	A 20% inbound tourism expansion leads to a 0.29% increase in real GDP, increased employment, a positive trade balance, and higher levels of inflation	Modified DPSV standard short run	f4q(‘agriculture,… , CommunServs’) = shocks applied based on expenditure patterns and distribution in Section 4.2, to simulate a 20% inbound tourism expansion
Scenario 3	UPGEM03, based on UPGEM02; Van Heerden et.al. (2006)	Simulation experiment to evaluate the economy-wide impacts on South Africa of a 10% inbound tourism reduction with fixed supplies of all primary factors	Simulation experiments and alternative tourism expenditure scenarios; no data required	A 10% inbound tourism reduction leads to a 0.15% decrease in real GDP, decreased employment, a negative trade balance, and lower levels of inflation	Modified DPSV standard short run	f4q(‘agriculture,… , CommunServs’) = shocks applied based on expenditure patterns and distribution in Section 4.2, to simulate a 10% inbound tourism reduction
Source: Framework adapted from Shoven & Whalley (1992:88–90).

Table 4: Impact on macro-economic variables (% change relative to the base case)

Annualised % change	Scenario 1	Scenario 2	Scenario 3
% Δ in real GDP^a (x0gdpexp)	0.15	0.29	–0.15
% Δ in aggregate employment (employ_iop)	0.30	0.60	–0.30
% Δ in GDP price index (p0gdpexp)	0.95	1.84	–1.00
% Δ in consumer prices (p3tot)	0.87	1.69	–0.93
% Δ in average nominal wage (p1lab_iop)	0.87	1.69	–0.93
% Δ in exports price index (p4tot)	0.60	1.17	–0.62
% Δ in total supplies of imported goods (x0imp_c)	0.88	1.72	–0.93
% Δ in export volume (x4tot)	1.23	2.41	–1.29
% Δ in import volume CIF (x0cif_c)	0.88	1.72	–0.93
% Δ in competitiveness (p0realdev)	–0.94	–1.81	1.02
% Δ in contribution of balance of trade to real GDP (contBOT)	0.15	0.29	–0.15
% Δ in the terms of trade (p0toft)	0.60	1.17	–0.62
Source: Authors' UPGEM model results
^aReal GDP from expenditure side.

Table 5: Sector results for scenario 1 (structural effects)

Sector		Value added		Exports		Imports		Employment
		Volume (x1tot)	Price (p1tot)	Volume (x4tot)	Price (p4tot)	Volume (x0imp)	Price (p0imp)	Volume (employ_op)	Price (p1lab_op)
1	Agriculture	0.15	1.70	2.89	1.68	3.04	0.00	0.54	0.87
2	Gold mining	–0.46	0.09	–0.47	0.12	0.78	0.00	–0.70	0.87
3	Other mining	–0.19	0.00	–0.11	0.03	–0.44	0.00	–0.47	0.87
4	Food processing	0.33	1.14	5.11	1.14	1.87	0.00	0.64	0.87
5	Beverages	–0.04	0.78	6.38	0.84	2.17	0.00	–0.17	0.87
6	Tobacco	–0.05	0.77	–2.59	0.66	1.71	0.00	–0.20	0.87
7	Textiles	–0.79	0.65	–2.51	0.64	0.90	0.00	–0.98	0.87
8	Clothing	0.84	0.75	6.73	0.76	2.26	0.00	0.96	0.87
9	Leather	–1.92	0.61	–2.37	0.60	0.86	0.00	–3.01	0.87
10	Footwear	–0.01	0.61	7.35	0.61	1.54	0.00	–0.02	0.87
11	Wood	–1.25	0.75	–2.85	0.73	1.09	0.00	–1.64	0.87
12	Paper	–0.73	0.54	–2.09	0.53	0.51	0.00	–1.54	0.87
13	Printing and publishing	0.12	0.77	6.61	0.79	0.90	0.00	0.20	0.87
14	Chemicals	–0.51	0.41	–1.53	0.39	0.34	0.00	–1.30	0.87
15	Rubber	0.99	0.99	5.72	1.00	1.25	0.00	2.03	0.87
16	Plastic	–0.49	0.63	–2.34	0.59	0.67	0.00	–0.62	0.87
17	Non-metallic minerals	–0.56	0.51	–1.96	0.50	0.78	0.00	–1.11	0.87
18	Basic metal products	–0.88	0.32	–1.31	0.33	0.61	0.00	–1.92	0.87
19	Fabricated metal products	–0.52	0.56	–2.15	0.54	1.05	0.00	–0.84	0.87
20	Machinery	2.50	0.96	5.85	0.97	0.53	0.00	3.71	0.87
21	Electrical machinery	–0.72	0.41	–1.61	0.41	0.26	0.00	–1.25	0.87
22	Transport equipment	1.85	0.81	6.51	0.81	1.74	0.00	2.80	0.87
23	Other manufacturing	0.80	1.52	3.57	1.52	2.15	0.00	2.84	0.87
24	Electricity	–0.07	0.67	0.00	0.00	0.00	0.00	–0.22	0.87
25	Building	0.02	0.63	0.00	0.00	0.00	0.00	0.03	0.87
26	Civil engineering	–0.15	0.65	–2.55	0.65	0.00	0.00	–0.27	0.87
27	Trade	0.19	0.93	–3.63	0.93	0.00	0.00	0.35	0.87
28	Accommodation & catering	0.46	1.66	2.71	1.73	1.47	0.00	1.89	0.87
29	Transport	0.08	0.83	6.41	0.83	1.03	0.00	0.18	0.87
30	Communication	0.11	0.87	6.28	0.86	0.00	0.00	0.25	0.87
31	Financial services	0.26	1.17	5.03	1.16	0.00	0.00	0.79	0.87
32	Communication services	0.53	0.88	6.17	0.89	1.29	0.00	0.63	0.87
	Industry average	0.00	0.77	1.66	0.73	0.95	0.00	0.05	0.87
Source: Authors' UPGEM model results.

Table 6: Sector results for scenario 2 (structural effects)

Sector		Value added		Exports		Imports		Employment
		Volume (x1tot)	Price (p1tot)	Volume (x4tot)	Price (p4tot)	Volume (x0imp)	Price (p0imp)	Volume (employ_op)	Price (p1lab_op)
1	Agriculture	0.29	3.30	5.52	3.27	5.93	0.00	1.03	1.69
2	Gold mining	–0.91	0.18	–0.91	0.23	1.50	0.00	–1.35	1.69
3	Other mining	–0.37	0.00	–0.22	0.05	–0.85	0.00	–0.92	1.69
4	Food processing	0.64	2.22	9.92	2.22	3.63	0.00	1.25	1.69
5	Beverages	–0.08	1.51	12.47	1.63	4.23	0.00	–0.33	1.69
6	Tobacco	–0.10	1.49	–4.95	1.28	3.33	0.00	–0.38	1.69
7	Textiles	–1.52	1.27	–4.81	1.24	1.75	0.00	–1.88	1.69
8	Clothing	1.65	1.46	13.18	1.47	4.40	0.00	1.89	1.69
9	Leather	–3.67	1.18	–4.54	1.17	1.67	0.00	–5.74	1.69
10	Footwear	–0.01	1.19	14.45	1.19	3.00	0.00	–0.02	1.69
11	Wood	–2.39	1.46	–5.44	1.41	2.12	0.00	–3.13	1.69
12	Paper	–1.39	1.05	–4.01	1.03	1.00	0.00	–2.95	1.69
13	Printing and publishing	0.24	1.51	12.93	1.53	1.74	0.00	0.39	1.69
14	Chemicals	–0.99	0.80	–2.95	0.75	0.66	0.00	–2.50	1.69
15	Rubber	1.93	1.94	11.11	1.94	2.43	0.00	3.96	1.69
16	Plastic	–0.94	1.23	–4.48	1.15	1.30	0.00	–1.18	1.69
17	Non–metallic minerals	–1.09	0.98	–3.77	0.97	1.52	0.00	–2.14	1.69
18	Basic metal products	–1.70	0.63	–2.53	0.64	1.20	0.00	–3.67	1.69
19	Fabricated metal products	–0.99	1.08	–4.12	1.06	2.05	0.00	–1.60	1.69
20	Machinery	4.88	1.87	11.39	1.88	1.02	0.00	7.27	1.69
21	Electrical machinery	–1.39	0.80	–3.09	0.79	0.50	0.00	–2.41	1.69
22	Transport equipment	3.64	1.58	12.72	1.58	3.39	0.00	5.51	1.69
23	Other manufacturing	1.53	2.95	6.86	2.94	4.17	0.00	5.50	1.69
24	Electricity	–0.12	1.29	0.00	0.00	0.00	0.00	–0.41	1.69
25	Building	0.04	1.22	0.00	0.00	0.00	0.00	0.06	1.69
26	Civil engineering	–0.29	1.27	–4.88	1.26	0.00	0.00	–0.51	1.69
27	Trade	0.37	1.81	–6.91	1.81	0.00	0.00	0.70	1.69
28	Accommodation & catering	0.88	3.22	5.18	3.35	2.85	0.00	3.64	1.69
29	Transport	0.17	1.62	12.52	1.62	2.00	0.00	0.37	1.69
30	Communication	0.21	1.69	12.25	1.68	0.00	0.00	0.49	1.69
31	Financial services	0.50	2.28	9.73	2.26	0.00	0.00	1.53	1.69
32	Communication services	1.04	1.72	12.04	1.73	2.52	0.00	1.24	1.69
	Industry average	0.00	1.49	3.27	1.41	1.85	0.00	0.12	1.69
Source: Authors' UPGEM model results

Table 7: Sector results for scenario 3 (structural effects)

Sector		Value added		Exports		Imports		Employment
		Volume ( x1tot )	Price ( p1tot )	Volume ( x4tot )	Price ( p4tot )	Volume ( x0imp )	Price ( p0imp )	Volume ( employ_op )	Price (p1lab_op )
1	Agriculture	–0.17	–1.82	–3.21	–1.80	–3.21	0.00	–0.60	–0.93
2	Gold mining	0.49	–0.09	0.49	–0.12	–0.84	0.00	0.74	–0.93
3	Other mining	0.20	0.00	0.11	–0.03	0.49	0.00	0.51	–0.93
4	Food processing	–0.35	–1.22	–5.46	–1.22	–1.99	0.00	–0.67	–0.93
5	Beverages	0.05	–0.83	–6.70	–0.90	–2.28	0.00	0.19	–0.93
6	Tobacco	0.06	–0.82	2.86	–0.70	–1.81	0.00	0.22	–0.93
7	Textiles	0.88	–0.70	2.77	–0.68	–0.94	0.00	1.09	–0.93
8	Clothing	–0.87	–0.80	–7.04	–0.81	–2.38	0.00	–1.00	–0.93
9	Leather	2.12	–0.65	2.62	–0.64	–0.91	0.00	3.35	–0.93
10	Footwear	0.03	–0.65	–7.61	–0.65	–1.64	0.00	0.04	–0.93
11	Wood	1.38	–0.81	3.16	–0.77	–1.16	0.00	1.81	–0.93
12	Paper	0.80	–0.57	2.28	–0.56	–0.53	0.00	1.70	–0.93
13	Printing and publishing	–0.12	–0.82	–6.93	–0.84	–0.95	0.00	–0.20	–0.93
14	Chemicals	0.56	–0.44	1.66	–0.41	–0.35	0.00	1.42	–0.93
15	Rubber	–1.05	–1.06	–6.08	–1.06	–1.32	0.00	–2.13	–0.93
16	Plastic	0.55	–0.67	2.57	–0.63	–0.70	0.00	0.68	–0.93
17	Non-metallic minerals	0.61	–0.54	2.15	–0.53	–0.83	0.00	1.22	–0.93
18	Basic metal products	0.96	–0.35	1.42	–0.35	–0.63	0.00	2.11	–0.93
19	Fabricated metal products	0.57	–0.59	2.35	–0.58	–1.11	0.00	0.92	–0.93
20	Machinery	–2.63	–1.02	–6.21	–1.03	–0.56	0.00	–3.86	–0.93
21	Electrical machinery	0.79	–0.44	1.75	–0.43	–0.27	0.00	1.37	–0.93
22	Transport equipment	–1.93	–0.86	–6.84	–0.86	–1.84	0.00	–2.89	–0.93
23	Other manufacturing	–0.88	–1.63	–3.91	–1.62	–2.29	0.00	–3.06	–0.93
24	Electricity	0.07	–0.71	0.00	0.00	0.00	0.00	0.24	–0.93
25	Building	–0.02	–0.67	0.00	0.00	0.00	0.00	–0.03	–0.93
26	Civil engineering	0.17	–0.69	2.80	–0.69	0.00	0.00	0.30	–0.93
27	Trade	–0.19	–0.99	4.04	–0.98	0.00	0.00	–0.35	–0.93
28	Accommodation & catering	–0.50	–1.78	–3.00	–1.86	–1.58	0.00	–2.06	–0.93
29	Transport	–0.08	–0.88	–6.74	–0.89	–1.09	0.00	–0.18	–0.93
30	Communication	–0.11	–0.92	–6.62	–0.92	0.00	0.00	–0.26	–0.93
31	Financial services	–0.27	–1.24	–5.42	–1.23	0.00	0.00	–0.84	–0.93
32	Communication services	–0.56	–0.94	–6.51	–0.95	–1.36	0.00	–0.66	–0.93
	Industry average	0.02	–0.82	–1.73	–0.77	–1.00	0.00	–0.03	–0.93
Source: Authors' UPGEM model results.

Table 8: Decomposition of demand for locally produced goods (percentage change)

Scenarios	Scn 1	Scn2	Scn 3	Scn 1	Scn2	Scn 3	Scn 1	Scn2	Scn 3	Scn 1	Scn2	Scn 3
Fandecomp	Local market			Domestic share			Export			Total
TranspEquip	0.27	0.53	–0.24	–0.93	–1.80	0.88	2.50	4.89	–0.58	1.85	3.62	0.06
Rubber	0.18	0.36	–0.01	–0.40	–0.79	0.21	1.22	2.37	–0.18	1.00	1.94	0.02
Clothing	0.06	0.12	–0.03	–0.55	–1.07	0.07	1.28	2.51	–0.12	0.79	1.55	–0.08
AccommCater	–0.08	–0.15	–0.17	–0.60	–1.15	0.40	1.19	2.27	–0.33	0.51	0.97	–0.11
CommunServs	0.01	0.02	–0.09	–0.01	–0.02	0.00	0.44	0.87	–0.02	0.45	0.87	–0.11
FoodPrcsing	–0.01	–0.02	–0.01	–0.27	–0.53	0.00	0.62	1.20	–0.26	0.33	0.65	–0.27
FinancServs	0.01	0.03	0.01	0.00	0.00	0.29	0.24	0.47	–0.66	0.26	0.50	–0.35
Source: Authors' UPGEM model results.

Table 9: Percentage change in real household consumption by population group

Households (x3tot_h)	Scenario 1	Scenario 2	Scenario 3
White	0.02	0.05	–0.03
Coloured	0.10	0.19	–0.11
Asian	0.07	0.13	–0.07
Black	–0.05	–0.10	0.05
Source: Authors' UPGEM model results.

Table 10: Percentage change in real household consumption by household type and population group (1)

x3totx (quintiles)	White	Coloured	Asian	Black
q1	0.05	0.09	0.06	–0.05
q2	0.05	0.09	0.06	–0.05
q3	0.03	0.10	0.07	–0.05
q4	0.02	0.11	0.07	–0.05
d9	0.02	0.10	0.07	–0.05
d10	0.02	0.08	0.06	–0.08
Source: Authors' UPGEM model results.

Table 11: Percentage change in household consumption price indices (Scenarios 1–3)

p3totx (quintiles)	White			Coloured			Asian			Black
	Scn 1	Scn 2	Scn 3	Scn 1	Scn 2	Scn 3	Scn 1	Scn 2	Scn 3	Scn 1	Scn 2	Scn 3
q1	0.85	1.66	–0.91	0.88	1.70	–0.94	0.88	1.72	–0.94	0.86	1.66	–0.92
q2	0.86	1.67	–0.92	0.87	1.70	–0.93	0.88	1.72	–0.94	0.86	1.66	–0.91
q3	0.88	1.70	–0.93	0.86	1.67	–0.92	0.88	1.71	–0.94	0.86	1.66	–0.92
q4	0.88	1.72	–0.94	0.86	1.67	–0.92	0.88	1.71	–0.94	0.86	1.66	–0.92
d9	0.89	1.72	–0.95	0.87	1.69	–0.93	0.88	1.71	–0.94	0.86	1.67	–0.92
d10	0.89	1.72	–0.95	0.88	1.71	–0.94	0.89	1.72	–0.94	0.88	1.71	–0.94
Source: Authors' UPGEM model results.

4.4 Modelling results

In interpreting the results from the AGE model, we followed Adams's proposal (2005) that results first focus on macro-economic impacts and then move down towards industry or sector level impacts and household impacts.

4.4.1 Macro-economic results

The simulation scenarios 1, 2 and 3 were implemented by shocking the variable f4q in the UPGEM, as described in Table 3. This alters the international trading conditions in the model to accurately simulate an inbound tourism expansion or recession. The tourism expansion scenarios 1 and 2 result in an increase in real GDP growth, whereas the recession scenario 3 causes industries to contract, and the trade balance to deteriorate – with an overall negative impact on GDP.

Under the assumptions discussed in Section 4.3, the principal macro-economic consequences of the expansion in tourism are:

•	A small increase in real GDP growth. With household consumption expenditure, investment and government consumption expenditure being fixed, the gain is purely as a result of an improvement in the trade balance.
•	A decrease in unemployment. The largest numbers of jobs are gained in the service industries catering directly to international tourists (e.g. transport, accommodation and catering, which are positively affected mainly because of changes in household real incomes, and substitution effects given the additional expansion in tourism).
•	A trade balance surplus. The immediate effect of the tourism expansion is an increase in export revenue, which drives the trade balance towards surplus. The increase in consumer prices, on the other hand, causes a contraction of non-tourism exports and a switch in domestic spending towards imports and away from domestic import-competing goods.

Table 4 summarises the impacts of the three scenarios on the main macro-economic variables. Columns 2 and 3 of the table show the results of a tourism expansion and column 4 the results of a tourism recession. It should be recalled, from the current tourism figures and arguments put forward in Section 4, that a tourism expansion can be expected to be generally associated with faster growth. This is evident from the difference in the percentage change in real GDP of the tourism expansion and recession scenarios (0.15% and 0.29% increase with a tourism expansion, compared to a decrease of 0.15% with a recession). However, these simulations were performed using a short-term comparative static closure and do not imply that greater tourism expenditure will improve long-term growth.

It should be noted that although the three tourism scenarios are implemented over one year the impact is simulated as a once-off event that plays itself out over a period of about two to three years. The results are then annualised and the impact can therefore be discounted back to reflect annual adjustments over the one-year period. From Table 4, it is evident that real GDP growth increases under both tourism expansion scenarios on an annualised basis, though the increase is more significant under scenario 2, which represents a 20% inbound tourism expansion (0.29% compared to 0.15% under scenario 1).

With an inbound tourism expansion (Table 4, second column), the reallocation of resources into capital goods-producing sectors reduces the unemployment rate (0.3%) and slightly increases the real GDP (0.15%). The reasons for the favourable effects of scenarios 1 and 2 (tourism expansion) on real GDP growth are the higher overall exports and a substantial improvement in the trade balance. Conversely, scenario 3 shows an unfavourable effect on real GDP growth, resulting from a deterioration in the trade balance. In these results, changes in relative consumer prices and their impacts on competitiveness are important. Changes in competitiveness will affect foreign countries' demand for South African exports and domestic demand for imports.

Under the closure rule (see Section 4.3), the external shock is balanced by a sharp expansion of other exports, whose share of GDP goes up by 0.41%. The reallocation of resources under scenario 2 further reduces the unemployment rate (0.6%) and also further boosts real GDP (0.29%). In both tourism expansion scenarios, changes in the contribution of the balance of trade to real GDP correspond to the actual change in real GDP growth. Scenario 3 shows clearly that a negative shock such as a fall in non-residents' demand results in both more unemployed workers (-0.3%) and increased unused capacity.

The relatively higher domestic prices, as indicated by the p3tot variable in Table 4, give rise to a slight surplus on the trade balance, mainly because of the increase in tourism into South Africa. The demand for imports has increased slightly, because of the substitution effect of the increase in prices of domestic goods, and the increase in output of various sectors in the economy. The net effect of changes in imports and exports of 0.15% corresponds to the percentage deterioration in the balance of trade (contBOT) variable shown in Table 4.

4.4.2 Sectoral impacts

As well as projecting the macro-economic effects of a tourism expansion, the UPGEM also provides estimates of the effects on the industrial structure. Four affected groups of industries can be distinguished:

•	Service industries catering directly to international tourists (e.g. transport, accommodation and catering). These are strongly stimulated by the additional expansion in tourism.
•	Industries indirectly supplying tourism-related activities (e.g. motor vehicle parts and construction). These are also stimulated by the additional expansion in tourism. Construction is an exception. In the base forecasts, investment growth is weak, especially residential investment. Hence additional expansion of tourism slightly eases the adjustment problems that the construction sector would otherwise have experienced.
•	Non-tourism exporters (e.g. agriculture, mining, food and metals processing). Growth prospects in these industries are reduced by the higher domestic prices.
•	Import-competing industries (e.g. transport equipment, chemicals, textiles, clothing and footwear). Prospects in these industries are reduced by the higher domestic prices.

Tables 5, 6 and 7 provide detailed results by individual sectors for scenarios 1, 2 and 3 respectively. In all three scenarios, the benefit (or brunt) of the adjustments falls on the main tourism sectors (see Section 4.2). Moreover, the effect under both tourism expansion scenarios is very similar for all of these sectors, with a more significant change under scenario 2. The effects on the remaining sectors vary considerably because capital goods-producing sectors do not always coincide with the major exporting sectors. In scenario 1, production in the construction and machinery sectors increases by 0.02% and 2.5% respectively. Scenario 1 also has a positive effect on all the major tourism sectors, although machinery (2.5%), transport equipment (1.85%), rubber (0.99%), clothing (0.84%) and other manufacturing (0.8%) register the largest gains.

Under scenario 2, there are three tourist-oriented sectors whose production falls much less (beverages and tobacco, textiles and footwear) or even goes up (accommodation and catering). This is not particularly surprising because the services of these sectors are used by both residents and non-residents, and residents' disposable income goes down when consumer prices increase. Production contracts in most non-tourist sectors and remains constant or expands slightly in the rest.

The third scenario (an inbound tourism recession) has exactly the opposite results. Production falls in those tourism sectors sensitive to domestic demand (food, clothing, accommodation and catering, and most of the services sectors) in contrast to the other two scenarios. These are also the sectors where most jobs are lost. The main difference, however, is in the performance of the non-tourist sectors, whose production levels go up quite considerably. This is true even in the capital goods-producing sectors, even though investment and capacity are kept fixed in these simulations.

It is also possible to decompose the change in demand for the seven top performing sectors under each scenario into (a) the local market effect (measured as the change in non-export demand for goods/services and other sectoral outputs), (b) the domestic share effect (measured as the change in domestic use/import ratio for the sectors' output demand) and (c) the export effect (measured as a change in demand for goods/services and output exports). In Table 8, the results of this decomposition show that in most of the sectors under scenarios 1, 2 and 3, changes in demand (negative or positive) come primarily through a change in exports (increase/decrease in inbound tourism).

4.4.3 Household impacts

This section examines in more detail the model results and findings, and summarises (in Tables 9 to 11) the different impacts of an inbound tourism expansion/recession on households in South Africa. As the focus is on the impact on poverty, a distinction is made in the model between white, coloured, Asian and black households (see Stats SA, 2002, 2004). Thus, from the household results the most insightful and revealing discovery is that the poor (i.e. lower income households) do not benefit much from additional tourist expenditures.

In the tourism expansion scenarios it is evident that there is an industry shift towards high skill occupations, due to the shift towards high skilled industries, such as services (and towards low skill occupations for the tourism recession scenario). This is in line with international experience (see for example Wood, 1998). As skill distribution is also racially biased in South Africa, the employment demand structure will have racial employment impacts (Naudé & Coetzee, 2004). Tables 9 to 11 show that blacks will benefit the least from an inbound tourism expansion because of the declining demand for less skilled workers (and the most from a recession).

Table 9 shows how these households' consumption (a rough measure of their welfare) is affected by either an inbound tourism expansion or recession. According to Naudé & Coetzee (2004), the net effect of the racial employment trends combined with the movements in wages should give an indication of changes in household income, which will have an impact on the racial private consumption expenditure patterns.

It can be seen from Table 9 that expenditure for all population groups increases under scenarios 1 and 2 (a tourism expansion), and declines under scenario 3 (a tourism recession), with the exception of blacks, whose expenditure decreases with a tourism expansion (scenarios 1 and 2) and increases under the recession scenario (which is in contrast to the three other population groups under all scenarios). This finding, that the poor suffer when tourism increases, is important and requires further clarification. How can this be explained? According to Naudé & Coetzee (2004), even though export growth and economic activity may be significantly boosted by an inbound tourism expansion, this may still be accompanied by increasing unemployment and inequality. Moreover, Bell & Cattaneo claim that this may be because ‘there has been a shift in the sectoral composition of exports, not only away from the more labour-intensive industries in general… but away from relatively black labour-intensive industries, and towards relatively white labour-intensive industries in particular’ (1997:14).

Table 10 also shows that an inbound tourism expansion has a moderate effect on the distribution of real consumption. Similarly, the table shows that there is some variation across households in the effects of the increase in consumption price indices that comes with an inbound tourism recession. This variation across households can be explained primarily by the effect of the change in the employment prospects of the households.

In terms of household consumption price indices, Table 11 shows that coloured and Asian households gain proportionally more than others from the inbound tourism expansion. In contrast, white and black households lose proportionally less than the others from an inbound tourism recession.

5. Implications

The following are some implications of this study.

Firstly, the results from an AGE model suggest that the welfare of lower income black households (which embody the bulk of the poor in this country) will be affected unfavourably compared to other households given a tourism expansion, mainly because of the labour market imperfections that exist in South Africa and which contribute to the allocative inefficiency and lack of equity that can be seen in the model results. The importance of this finding is that if tourism is to become an effective tool for decreasing unemployment (and ultimately alleviating poverty), it will have to be accompanied by more focused and improved labour market and human resource development policies. This policy improvement is necessary as a response to the inequality, unemployment and low growth experienced by the poorest South Africans in the tourism expansion scenarios. This corroborates Naudé & Coetzee's finding (2004) that domestic labour market policies and institutions can have the effect of increasing or decreasing inequalities between population groups.

If the aforementioned measures were to be put in place, then at a macro level and under the 20% growth in tourism scenario, the impact on poverty alleviation would be the greatest (although, of course, to achieve and sustain a 20% growth would not be easy). The opposite is also true: when tourist numbers (and hence expenditure) dwindle, then this also has a negative effect on employment. It is therefore important that tourist and destination marketing organisations should manage growth in a sustainable manner. In general the results of the present study support the research by Blake et al. (2008), whose similar exercise on data for Brazil shows that tourism can alleviate poverty, if managed correctly.

Secondly, arising from the discussion of the first implication, it remains of the utmost importance that destinations implement a sustainable growth plan. This implies a gradual increase in tourist numbers and especially tourist spending. The focus should remain on an increase in spending, which supports research by Uys (2003) and Saayman & Rossouw (2011). Barriers to growth – such as political instability, lack of proper infrastructure, uncontrolled crime, ineffective marketing, decreased budgets for marketing, and lack of tourism development and campaigns – must be identified and dealt with, especially by the various government structures.

Thirdly, it must be realised that expenditure by foreigners and by domestic tourists have different impacts on the various sectors. For example, foreign expenditure benefits accommodation more than domestic expenditure, perhaps because domestic tourists often stay with family and friends and do not always require paid accommodation. Both types of tourist are crucial for poverty alleviation. Tourism marketing organisations should therefore aim to develop both these markets and not focus on just one or the other.

Fourthly, factors raised by the UNWTO (see WTO, 2002:2) should be implemented as far as possible to maximise the impact of tourism on poverty alleviation. This entails an increase in tourists' length of stay, which increases spending in general, the spreading of tourism benefits, establishing stronger inter-sectoral linkages and optimising tax revenues from tourism. Attending to these factors could also assist in minimising leakages.

Lastly, tourism marketers must be more focused in segmenting markets. It is important to target markets with a high yield since these have the greatest impact on poverty alleviation. More research is needed in this area.

6. Concluding remarks

This paper provided an economy-wide analysis of the distributional effects of a tourism expansion and recession, thus offering a way to answer the question about how this industry can help to alleviate poverty in South Africa. This is the first time this type of modelling has been applied to tourism spending in South Africa. The most interesting and important finding is that, in the short term, the poor benefit very little from additional tourism inflows, if at all.

The economy-wide effects of an inbound tourism expansion (or recession) in South Africa include both benefits and costs. These are distributed unevenly across institutions and sectors. An inbound tourism expansion can stimulate higher real GDP growth and improve the balance of trade because the real exchange rate is exogenous. These results should not be seen in isolation but in conjunction with other recent findings that are consistent with ours, such as the study by Blake et al. (2008).

From a policy perspective, it would be important to foster both domestic and foreign tourist arrivals. It must be emphasised that it is not the number of tourists but the amount spent that counts. This article contributes to the discourse on tourism and poverty alleviation and emphasises that the mere fact that tourism takes place does not imply that the money trickles down to the poor – deliberate action is required. We recommend that future research should look at poverty alleviation at a provincial level, to see which provinces benefit the most.

Notes

¹ORANI-G (‘G’ stands for 'generic') is a version of ORANI which serves as a basis from which to construct new models. It has been applied to many countries, including China, Thailand, Korea, Pakistan, Brazil, the Philippines, Japan, Ireland, Vietnam, Indonesia, Venezuela, Taiwan and Denmark (Horridge et al., 1993). See for example www.monash.edu.au/policy/oranig.htm. For a more detailed exposition of the modelling approach followed in UPGEM, see Horridge et al. (1993) and for an application see Van Heerden et al. (2006).

²With comparative statics two different outcomes are compared, before and after a change in some underlying exogenous parameter (Mas-Colell et al., 1995:24; Silberberg & Suen, 2000). The model in its present form is also capable of long-term comparative statistics.