The economic contribution of tourism in Mozambique: Insights from a Social Accounting Matrix

Abstract

How much tourism contributes to the economies of developing countries is controversial and often not measured rigorously. Focusing on Mozambique, this study presents a simple accounting tool – a tourist-focused Social Accounting Matrix – which makes it possible to estimate the economic contribution of various tourism sub-types. Multiplier analysis is applied to evaluate the strength of backward linkages from tourism to the domestic economy. The results show the sector is moderate in size but has the potential to contribute significantly to aggregate economic development. However, potential weaknesses are already evident and careful attention must be paid to the full tourism value chain.

Keywords:

• economic contribution of tourism
• tourism development
• Social Accounting Matrix
• Mozambique

1. Introduction

Tourism is increasingly recognised as a valuable source of long-term growth for developing countries, including sub-Saharan Africa. Before the recent financial crisis, it was a robust growth sector in global terms and, on average, it was the less developed countries that saw the most rapid increases in tourist arrivals (Blanke & Chiesa, 2007). A growing body of literature affirms that tourism can be pro-poor and has the potential to develop strong backward linkages to the rest of the economy (Mitchell et al., 2007). However, these features are not guaranteed but depend on government policies as well as actions at the local level. For this reason, understanding the structure of tourism and its linkages with the domestic economy is critical to inform policy.

This paper considers the case of Mozambique, where tourism has been of historical importance and remains a strategic government priority. To date, economic analysis of the sector has been extremely limited. A primary objective of the paper is to provide an aggregate measure of the contribution of tourism to the economy, differentiated by distinct types of tourism such as foreign visitors and domestic residents. An additional objective is to estimate the multiplier effects associated with tourism, and to compare these with other economic sectors. This provides a means to evaluate the relevance of selecting tourism as a development priority. To do so we construct a detailed Social Accounting Matrix (SAM) for the economy, including separate treatment of six tourism sub-types. We find that the tourism sector is currently moderate in size, but has substantial potential. However, realising this potential depends on addressing constraints and giving due attention to the full tourism value chain, rather than simply targeting additional foreign visitor arrivals.

The remainder of the paper is structured as follows. Section 2 introduces the case of Mozambique, including an overview of the tourism sector, Section 3 describes the core methodology, namely the construction of a tourism-SAM for Mozambique and its subsequent analysis, and explains the motivation for using it, Section 4 presents the results and Section 5 concludes with a summary of the main findings and a discussion of policy implications.

2. Tourism in Mozambique

Prior to Independence in 1975, tourism was a jewel in Mozambique's crown. With around 2500 km of Indian Ocean coastline, as well as impressive game parks and historical sites from a long history of Portuguese settlement, the country attracted numerous visitors from across southern Africa and Europe. After Independence, however, the sector went into dramatic decline because of civil war and economic collapse. Since the end of conflict in 1992, however, Mozambique has achieved impressive peace, stability and economic growth (Arndt et al., 2007). As a result, there is now potential for substantial renewal of the tourism sector, despite weak infrastructure and widespread poverty.

Tourism has been identified by the Mozambican government as a key component of its development strategy. By 2020 the country aims to be the ‘most vibrant, dynamic and exotic’ destination in Africa, welcoming over 4 million visitors per year (República de Moçambique, 2004). The positive economic contribution of tourism is frequently emphasised, particularly as regards employment creation and intersectoral linkages (República de Moçambique, 2003, 2004). To a certain extent, investment in the sector has matched these expectations. From 1995 to 2006, investment worth over US$1.8 billion dollars was approved by the Investment Promotion Centre, equal in value terms to 14 per cent of all approved investments (both foreign and domestic). Press reports suggest that over US$200 million was invested in the sector in 2006 alone and the Ministry of Tourism (MITUR) also estimates that physical hotel capacity rose by over 50 per cent from 2000 to 2005.

A major difficulty in understanding the dynamics of the tourism sector in Mozambique is that data are limited and inconsistent. Jones (2007), for example, documents large differences in estimated total foreign tourist arrivals across various government and external sources. Apart from sporadic press announcements, neither MITUR nor the National Statistics Institute (INE) publishes regular, consistent tourism arrival numbers. Additionally, no study has considered the role of domestic tourism, despite the existence of a comparatively large cohort of expatriates, a growing domestic middle class and numerous workshops and conferences funded by international donors. More generally, minimal analysis has been undertaken that might contribute to moving from a high-level vision for the sector to a concrete strategy with associated policies and targets. For example, a study by the investment advisory arm of the World Bank concludes that Mozambique needs to flesh out the kind of tourism development that it desires over the longer term (Foreign Investment Advisory Service [FIAS], 2006).

Nevertheless, some idea of the structure of the sector comes from a foreign tourism expenditure survey carried out by the Mozambican MITUR in 2006. Although the survey was not perfectly representative, it shows that there are different segments of foreign (inbound) tourism. As shown in Table 1, it is helpful to distinguish between three kinds of foreign tourist in Mozambique – business visitors, self-drive visitors and other leisure visitors (which include people visiting friends and family). The second kind are visitors from (mainly) neighbouring countries, principally South Africa, who bring their own vehicle and often substantial food, cooking and camping equipment. Many South Africans have constructed beach accommodation facilities in Mozambique, including mid-budget to low-budget options where camping is permitted. These are extremely popular during public holidays in South Africa, but may be visited infrequently at other times of the year (see Jones et al., 2007).

Table 1: Profile of foreign tourism expenditure (2003)

Visitor type	Business	Leisure		Weighted average
		Self-drive	Others
Expenditure (daily US$ per person)
Accommodation	56.8	25.9	32.6	38.5
Food & restaurants	21.6	14.5	8.9	16.0
Other	19.6	13.8	18.6	16.8
Total	98.0	54.1	60.1	71.3
Average stay (number of days)	3.0	3.0	7.0	3.8
Spend per visit (US$ per person)	294.1	162.3	420.8	259.7
Total number of visitors (‘000s, annual)	161.6	195.6	83.8	−
Total expenditure (US$ million)	47.5	31.7	35.3	−
Notes: Expenditure values have been deflated to 2003 prices and converted to US$ at the average annual exchange rate for the period. Visitor types are as discussed in the text.
Sources: Author's estimates from 2006 MITUR expenditure survey, INE (2005) and MITUR (2007).

One implication of this segmented structure is that a focus on visitor numbers can be misleading. As Table 1 shows, on a per capita basis self-drive visitors contribute substantially less financially than business and other leisure visitors. Business visitors spend the most on a daily basis but typically do not stay long. Other leisure visitors stay much longer, thereby contributing more to the local economy. Consequently, despite their numbers being less than half those of the self-drive visitors, the value of their local spending is 2.5 times that of the self-drive variety. Partly reflecting the predominance of self-drive and short-stay visitors, the weighted average spend per (foreign) visitor is extremely low in Mozambique (at US$260) compared to other sub-Saharan African countries. According to World Economic Forum data (Blanke & Chiesa, 2007), the average spend is around US$1000 in South Africa, Mauritius and Tanzania. Notably, the World Economic Forum (WEF) also ranks Mozambique 119th of 124 countries in terms of cross-country tourism competitiveness, falling behind relevant peers such as Tanzania (80th) the Gambia (84th) and Zambia (94th).

3. A tourism-SAM for Mozambique

3.1 Motivation

The above suggests a number of important questions about tourism in Mozambique: (1) What is its overall economic contribution? (2) How does it compare with other economic sectors with respect to its current and potential economic contribution? (3) What is the relative economic contribution (both current and potential) of different types of tourism? Answering these questions would help to inform Mozambique's broader development strategy as well as to substantiate the government's tourism policy framework. Additionally, quantifying the current economic contribution of tourism can serve as a useful reference point for future assessments of the sector.

The three questions focus mainly on the aggregate economic contribution of tourism. This should be distinguished from its economic impact at either the macroeconomic or microeconomic level. The former is essentially an accounting problem. To measure the (past) economic contribution of a production sector there is no need for a theoretical framework that specifies the channels through which changes in tourism affect other areas of the economy. In contrast, the notion of economic impact refers precisely to such linkages and thus requires a model of economic behaviour. For example, Fayissa et al. (2008) estimate the impact of tourism on economic growth in Africa using a regression model applied to panel data. Although they do not make their theoretical framework explicit, their chosen specification can be interpreted as the reduced form of an augmented Solow–Swan growth model.

Although accounting for the economic contribution of tourism does not rely on any particular economic theory, it is not straightforward. Unlike most other economic sectors (e.g. agriculture), tourism does not correspond to a unique production activity; rather, it is characterised by a combination of commodity purchases from a range of domestic production sectors and imports. Thus, tourism is not treated in standard national accounts as a homogeneous (stand-alone) production sector or commodity (Organization for Economic Cooperation and Development [OECD], 2000; OECD et al., 2001). Some indication of tourism's economic contribution can be taken directly from national accounts data (where available), such as from the share of hotels and restaurants in total value added. Additional statistics, such as foreign tourist arrivals and length of stay, may also provide some guidance. However, these give an incomplete picture since tourism involves much more than food and beverages and accommodation.

One way to account for the tourism sector is to develop tourism satellite accounts (TSAs).1 These are extensions to the core system of national accounts (SNA) and represent a rigorous and internationally comparable means to account for tourism in a given economy. They are constructed from extensive information on tourism activity, collected separately from the core SNA data. This is used to construct an alternative and elaborate version of the national accounts classified according to tourism and non-tourism activities. There are a number of drawbacks associated with TSAs, however. A comprehensive set of TSAs depends on an extremely rich information base on both domestic and foreign tourism, as well as considerable national accounts expertise. Where statistical capacity is weak or poorly financed, investment of resources in TSAs may be hard to justify, especially where the underlying data are questionable, or where there is weak demand for comprehensive and regular TSA information (for discussion with respect to Tanzania see Sharma & Olson, 2005). In such cases, resources may be better focused on regularly collecting and publishing basic tourism information and/or producing a limited range of tourism accounting data.

These concerns are pertinent to Mozambique. As noted in Section 2, data on tourism are not extensive, the sector is relatively small (but growing) and national accounts resources are scarce. An alternative to comprehensive TSAs is to develop a tourism-focused SAM. An advantage of this approach is that SAMs form the information core for a range of economic models such as computable general equilibrium approaches. Thus, the construction of a SAM (with or without tourism) represents a valuable investment per se and can permit the analyst to move from economic accounting toward impact modelling. Moreover, SAMs are highly flexible tools and can be constructed to incorporate as much or as little sector specific detail as necessary. Thus, in contrast to the standardised approach of TSAs, it is possible to concentrate on specific dimensions of tourism that are of policy interest.

3.2 Construction of the tourism-SAM for Mozambique

SAMs are extensions of input–output tables that take into account the full structure of flows between production sectors, commodity groups and agents in the economy.2 A given SAM represents an accounting snapshot of relations between these different elements over a fixed period, usually 1 year. Specifically, elements are classified into individual accounts and arranged in a square matrix, with identical row (r) and column (c) totals. Each cell of the matrix represents the total value of payments (flows) from account c to account r over the chosen period. Thus, accounting relations between elements in the economy – that is, national accounts identities – can be read off directly from the matrix.

To facilitate SAM construction, existing SNA classifications typically dictate the choice of accounts. Broadly speaking, different accounts cover production activities, commodity groups, factors of production (e.g. labour and capital) and economic institutions (e.g. firms, households, government and the rest of the world). Accounts can be specified at different levels of aggregation depending on the objectives of the analysis and data availability. For example, various household and labour accounts will often be created in order to consider distributional questions. Both SAMs and input–output tables have been used extensively to analyse tourism (e.g. Wagner 1997; Kweka et al., 2001; Dwyer et al., 2004; Blake et al., 2008). These approaches have their limitations, however, as they typically follow SNA classifications and therefore do not provide an explicit treatment of tourism. Rather, the economic role of tourism must be inferred from ‘characteristic’ sectors such as hotels and restaurants. This may be informative, but no disaggregation by different tourism sub-types is possible; also ‘non-characteristic’ aspects of tourism may be ignored.

Before addressing this deficiency, however, the first step is to construct a ‘standard’ SAM (without tourism). Specific details regarding SAM construction cannot be given here due to space limitations. With respect to Mozambique, the methodology used to construct the ‘standard’ SAM is described in McCool et al. (2009) and is based on a detailed range of national accounts and household survey data; cross-entropy methods were used to balance the SAM as described in Robinson et al. (2001). The present exercise pertains to data for 2003, reflecting the significant time lag in availability of comprehensive and internally consistent national accounts data. Where more recent tourism information is employed (see below), this is deflated appropriately.

Table A1 in Appendix A presents an aggregated version of the 2003 Mozambique SAM, formed by compressing numerous separate accounts into single macroeconomic aggregates (e.g. 17 production activities are compressed into one production account). Nevertheless, it gives a sense of the general structure of the SAM and how it can be used for accounting purposes. For example, gross domestic product (GDP) at market prices can be calculated as the sum of net final commodity demand across different agents, conforming to the identity: Y ≡ C + I + G + X – M. This is read directly from Table A1 as the sum of the purchases of commodities by institutions, plus own production (households' direct purchase of activities), minus imports (sales of commodities by the rest of the world). This is equal to 111.86 trillion meticais (approximately US$5 billion).

Using this SAM, tourism can be included according to the method established in Dent et al. (2004). As such, tourism activities are incorporated by creating ‘phantom’ or artificial sectors. The starting point is the various commodity groups (baskets) distinguished in the national accounts. In the standard SAM, commodities are ‘sold’ either to production sectors (as intermediate goods) or to institutions such as households (as final consumption). Without altering these aggregate relations, it is possible to add a set of artificial activities which first purchase a range of commodities and then sell these as new composite commodities to institutions. These artificial accounts, which can be thought of as corresponding to different sub-types of tourism (e.g. domestic, foreign), are technical devices which only alter the composition of commodity baskets. The additional accounts have no associated labour, taxation or capital costs, and thus do not alter the real structure of the SAM. This is a natural way to treat tourism given that it is characterised by demand for various types of commodities.

Following the description in Section 2, six tourism sub-types are distinguished in the Mozambican tourism-SAM. Foreign tourism is disaggregated into business, self-drive and ‘other leisure’ types. Domestic tourism is disaggregated into household tourism, tourism undertaken by the government and domestic businesses (together) and tourism investment. To achieve this, we use the data available from tourism expenditure surveys as well as additional assumptions where necessary. Specifically, the 2006 MITUR tourism expenditure surveys and data on tourism arrivals (INE, 2005) is used to separate out foreign tourism. For the domestic disaggregation no similar data exist, so general household survey data are used as well as the weights of different kinds of domestic tourism expenditure in commodity demand taken from a Namibian domestic expenditure survey (Poonyth et al., 2001).3

To see how different commodities in the tourism-SAM are apportioned across (or sold to) tourism sub-types, it is helpful to follow the approach of TSAs and distinguish between commodities that are characteristic of tourism (e.g. hotel accommodation) and those that are not. Table 2 summarises the value of sales of different commodities to the newly created tourism accounts (for simplicity, only domestic and foreign tourism are shown). It shows that tourism activities constitute a core source of demand for hotel accommodation, restaurants and domestic air travel. This is not surprising given the definition of tourists as visitors who ‘travel to and stay in places outside their usual environment for not more than one consecutive year for leisure, business and other purposes’ (OECD et al., 2001). In contrast, non-characteristic tourism commodities are only weakly influenced by tourism demand, the most important being purchases of fuel, where tourism accounts for around 5 per cent of total demand. An interesting finding, discussed further below, is the notable difference in the pattern of spending between foreign and domestic tourists. For example, domestic tourism is less concentrated on characteristic tourism commodities (60 per cent) than foreign tourism is (81 per cent).

Table 2: Allocation of gross commodity sales by tourism types (10⁹ meticais)

	Tourism			Economy
	Foreign	Domestic	Total	Total	Percentage tourism
Characteristic products	1 733	3 368	5 101	17 467	29.2
Hotels	1 060	315	1 375	1 471	93.5
Restaurants	575	529	1 103	1 696	65.0
Air transport	98	561	659	1 213	54.3
Road transport	0	1 870	1 870	12 992	14.4
Travel agencies	0	93	93	94	99.0
Non-characteristic products	407	2 226	2 634	244 811	1.1
Agriculture	12	94	106	17 095	0.6
Fisheries	15	59	75	9 427	0.8
Food processing	53	214	268	24 800	1.1
Drinks processing	13	184	197	7 517	2.6
Manufactures (other)	116	105	220	15 120	1.5
Fuel	54	426	480	10 342	4.6
Construction	0	815	815	12 542	6.5
Transport	22	0	22	10 210	0.2
Services	123	329	452	32 503	1.4
Others	0	0	0	100 042	0.0
Total	2 140	5 594	7 735	257 064	2.9
Percentage characteristic	81.0	60.2	65.9	6.8	−
Percentage non-characteristic	19.0	39.8	34.1	93.2	−
Notes: Commodities are divided between characteristic and non-characteristic tourism products as per the categorisation in TSAs; cells represent values included in the detailed tourism-SAM, apportioning total commodity sales between tourism sub-types; given the average 2003 exchange rate of US$1 = MZN 23 500, each unit in the table is equivalent to approximately US$42 500.
Source: Author's estimates from the 2003 tourism-SAM.

With the tourism-SAM thus in place, analytical methods traditionally applied to input–output tables (e.g. Archer & Fletcher, 1996; Kweka et al., 2001) can be employed as a further stage in the analysis. The most simple of these involves estimating and decomposing various types of multipliers (see Pyatt & Round, 1985). In contrast to the accounting methods used to construct a SAM, the estimation of multipliers invokes a number of stronger behavioural assumptions. For example, an a priori distinction must be made between endogenous and exogenous accounts. In addition, multiplier analysis typically assumes either that prices are fixed or that technology and preferences are in fixed proportions. It also tends to assume that there are no binding supply-side constraints. Appendix B provides a more formal description of the multiplier analysis used in the present paper.

4. Results

4.1 Accounting for tourism

An aggregate version of the 2003 tourism-SAM is presented in Table A2 in Appendix A. This is similar in many respects to Table A1 but now includes tourism activity accounts and their corresponding composite commodity baskets. Again, for expositional simplicity, only domestic and foreign types of tourism are distinguished in the table. Certain aspects of the economic contribution of tourism can be read off (almost) directly from this modified matrix. For example, gross tourism demand is equal to MZN 7735 billion (approximately US$330 million), equal to 2.9 per cent of gross demand in the economy. These figures are taken directly from cells (row d, column b) and (row d, column c) in the table, which refer to commodity purchases by the tourism sub-types (also see Table 2). They show that total tourism demand is made up of foreign tourism (28 per cent) and domestic tourism (72 per cent). Figure 1 extracts data from the disaggregated tourism-SAM to replicate some of the principal outputs associated with TSAs. The figure classifies gross demand and supply of tourism into its main components.4 On the demand side the figure shows the main institutional sources of gross demand – overseas visitors (exports), resident businesses (intermediate consumption), resident households (personal consumption) and investment.

Figure 1: Components of gross tourism demand and supply (percentage, 2003)

On the supply side we need to estimate the value added associated with tourism. A simple approximation is to adjust final tourism demand, which includes domestic and imported goods, by one minus the ratio of imports to aggregate final demand. This gives an (upper bound) estimate that tourism accounts for approximately 4.2 per cent of GDP. However, this approach ignores the specific commodity composition of tourism. A preferred approach is to calculate the share of total value added embodied in each commodity, which will be zero for any commodity that is solely imported, and then apportion these shares across tourism activities according to their commodity composition. The results from this exercise are given in Table 3, showing that tourism accounts for 3.2 per cent of total value added (GDP at factor cost). The discrepancy with the initial estimate indicates a slightly higher import content of tourism relative to other sectors, in part reflecting the greater importance of transport services in tourism consumption. The table also gives a comparison of the ratio of value added to output by sector, which confirms the previous result. Agricultural activities correspond to the highest value added ratio while tourism more closely approximates the value added ratio of hotels and restaurants (56 per cent).

Table 3: Disaggregation of sector value added, including tourism and its sub-types

	GDP (10^6 US$)	Percentage of total	VA ratio (%)
Total economy	4 260.5	100.0	62.2
Agriculture	766.0	18.0	84.6
Industry	1 350.6	31.7	52.2
Services	2 143.9	50.3	60.5
Tourism	137.8	3.23	49.0
Foreign	43.1	1.01	54.0
Business	18.7	0.44	55.7
Leisure	12.7	0.30	53.3
Self-drive	11.7	0.27	51.9
Domestic	94.7	2.22	46.7
Households	60.6	1.42	50.1
Firms and government	18.8	0.44	32.5
Investment	15.4	0.36	50.5
Notes: ‘VA ratio’ gives value added (GDP) divided by the total output value for a given production sector.
Source: Author's estimates from the 2003 tourism-SAM.

In sum, the overall economic contribution of tourism in Mozambique is moderate but not insignificant. Similar sized industries from a value added perspective include fisheries, the construction sector and education (INE, 2007). Nevertheless, the thesis of segmentation between tourism sub-types is confirmed. The largest share of gross tourism demand derives from resident households; also, the overseas visitors market is highly differentiated. Overseas business visitors account for over 40 per cent of foreign tourism demand; self-drive visitors, who are largely residents of neighbouring countries (especially South Africa), account for a further 28 per cent. Thus, leisure tourists from the more lucrative markets of Europe and North America (as well as East Asia) currently represent a very small share of total tourism demand. The two categories of overseas leisure tourism account for only 17 per cent of total tourism value added (or half a percentage point of total GDP). Value added per overseas tourist thus varies dramatically across the three sub-types of overseas visitor. Indeed, self-drive leisure tourists contribute the lowest value added both in absolute terms (also Table 3) and, most dramatically, in per visitor terms (see visitor numbers from Table 1).

4.2 Multiplier analysis

Multiplier analysis provides a means to estimate the effect on (macro)economic accounts of exogenous changes in specific sectors. Numerous types of multipliers can be estimated from a SAM; those used here are described in Appendix B. Exogenous accounts are taken as investment spending and the rest of the world (accounts m and n of Table A2); thus, the multiplier analysis gives the final impact of changes in these accounts on the remaining accounts. Table 4 summarises the core results, encompassing all tourism commodity accounts and relevant aggregate commodity baskets. Each cell of the table gives a measure of the expected impact accruing to the specified column account arising from a unit exogenous increase in purchases of the row commodity account. To assist interpretation, these impact measures are normalised by the economy-wide average multiplier. Thus, a positive (negative) percentage indicates that the multiplier of interest is higher (lower) than the economy-wide average for the given column account. Thus, the cell referring to ‘foreign visitors’ and ‘production’ in the table indicates that the expected increase in domestic production accruing from an exogenous unit increase in foreign tourism purchases is 15.3 per cent higher than the economy-wide ‘average’ commodity multiplier.

Table 4: Normalised aggregate multipliers from the 2003 tourism-SAM

	(I)	(II)	(III)	(IV)	(V)
	All accounts	Production	HHld income^a	Poor income^b	Government revenue
Tourism commodities
Foreign visitors	12.3%	15.3%	14.5%	12.7%	3.5%
Business	13.5%	17.2%	16.1%	14.0%	2.5%
Leisure	11.6%	15.0%	13.5%	12.0%	−0.3%
Self-drive	11.1%	12.9%	13.1%	11.5%	8.9%
Domestic visitors	2.1%	2.9%	0.7%	−0.2%	17.5%
Households	8.2%	8.4%	7.8%	7.2%	34.8%
Firms and government	−5.4%	−4.5%	−8.9%	−10.6%	6.1%
Investment	−8.1%	−5.1%	−9.6%	−9.6%	−29.9%
Non-tourism commodities
Agriculture	12.8%	9.5%	18.4%	33.0%	−9.7%
Industry	−19.2%	−22.0%	−23.7%	−24.5%	−13.4%
Services	1.2%	3.1%	4.2%	1.7%	−5.2%
(Tourism) – (non-tourism)	8.0%	10.0%	6.6%	4.4%	22.2%
Notes: ^aRefers to average household income. ^bRefers to the lowest income quintile as disaggregated in the SAM. All calculations refer to an exogenous injection from a given row commodity account; the bottom row gives the weighted average difference in normalised multipliers between tourism and non-tourism commodities; see text for further discussion.
Source: Author's estimates.

Three main results can be highlighted. First, broadly speaking, tourism is a sector with comparatively strong backward linkages across production, household income and value added accounts. Specifically, among non-characteristic tourism sectors, growth in tourism would provide particular stimulus to food and beverages processing, agriculture (including fisheries) and service industries. The majority of normalised multipliers are positive and there also tends to be a positive difference in the multipliers for the average tourism commodity compared to the average non-tourism commodity (given in the bottom row of the table). Thus tourism represents a potentially valuable growth engine compared to many other sectors. In part, however, this is a consequence of the weak multipliers associated with the industrial sector, which reflects the predominance of capital-intensive industry in Mozambique's current industrial base. Second, comparing tourism sub-types, foreign tourism tends to display stronger multiplier effects than domestic types. Business tourism generally has the largest multipliers, reflecting the larger share of (high-end) hotels and restaurants in the expenditures of these visitors. Indeed, the weaker performance of domestic tourism can be traced to the relatively greater importance of transport expenditures, which largely go toward fuel imports. Third, for the same reason – namely, higher effective taxation on fuel purchases (especially land transport) – government revenue multipliers are strongest for domestic tourism subtypes as well as for the self-drive visitors among foreign tourists.

A more nuanced picture is gained from an analysis of value added and employment multipliers. These are shown in Table 5. The first column gives the unadjusted normalised value added multiplier, which is calculated and interpreted in the same way as the multipliers in Table 4 but measures the final impact on value added of exogenous changes in commodity demand. The adjusted value added multipliers give the ratio of the latter multiplier to the return to physical capital (‘K adjust’) or skilled labour (‘skilled L adjust’) necessary to sustain the multiplier process. As such, the value added multiplier is corrected for its dependence on comparatively more scarce factors of production; see Appendix B for further details.

Table 5: Normalised value added and employment multipliers from the 2003 tourism-SAM

	Value added
	Unadjusted	K adjust^a	Skilled L adjust^b	Employment
Tourism commodities
Foreign visitors	14.5%	0.9%	−10.9%	3.3%
Business	16.1%	0.7%	−11.5%	3.7%
Leisure	13.6%	0.6%	−9.9%	2.8%
Self-drive	13.0%	1.5%	−11.0%	3.4%
Domestic visitors	0.1%	6.8%	−8.4%	−7.4%
Households	7.0%	10.7%	−8.6%	−0.1%
Firms and government	−9.2%	3.9%	−11.3%	−17.3%
Investment	−9.3%	−3.3%	−2.4%	−17.4%
Non-tourism commodities
Agriculture	17.9%	9.5%	28.9%	79.8%
Industry	−23.3%	−7.5%	3.9%	−27.4%
Services	4.1%	3.5%	−4.6%	−2.6%
(Tourism) – (non-tourism)	6.2%	4.1%	−13.2%	−8.8%
Notes: ^aGives the normalised value added multiplier adjusted for returns to physical capital. ^bGives the normalised value added multiplier adjusted for returns to skilled labour. All calculations refer to an exogenous injection from a given row commodity account; the bottom row gives the weighted average difference in normalised multipliers between tourism and non-tourism commodities; see text for further discussion.
Source: Author's estimates.

The (normalised) adjusted multipliers show that foreign tourism is heavily dependent on scarce production factors, thereby indicating that strong potential multiplier effects from overseas tourism could become constrained by factor shortages. This is indicated for all foreign tourism sub-types by the substantially lower values of the adjusted multipliers compared to the unadjusted multipliers. The skilled labour adjustment is particularly striking – all tourism commodities have an average adjusted multiplier approximately 13.2 per cent lower than other commodities (see the bottom row of the table). Thus, shortages of appropriately skilled labour represent a potentially critical constraint on tourism growth, a finding that is consistent with experiences in other developing countries (Liu & Wall, 2006). Nevertheless, compared to foreign tourism, domestic tourism is somewhat less dependent on sustaining returns to physical capital (the adjusted value added multiplier for capital is 6.8 per cent), but remains dependent on skilled labour (skilled labour adjusted value added multiplier = –8.4 per cent).

The employment multipliers also point to a cautious final assessment of tourism's potential. These show that expected job creation associated with tourism growth is reasonable compared to other sectors, and is relatively weaker for domestic tourism variants. Specifically, for the same level of exogenous stimulus, job creation in tourism is similar to other service sectors. This means that it is much stronger than in industrial sectors but, unsurprisingly, much weaker than in agriculture and related sectors.

Finally, it is worth noting that the income multipliers for the poorest households are small in absolute terms – for a one-unit increase in purchases of the economy-wide average commodity, only 10 per cent of the final value ‘trickles down’ to the poorest households. Thus, even the moderately larger multipliers associated with foreign tourism do not necessarily make it a pro-poor industry. Indeed, with the exception of government revenue, agricultural commodities display by far the strongest multipliers on an adjusted and unadjusted basis. This is because the majority of (poor) households are active in this sector and it is least dependent on scarce factors of production. Consequently, any specifically pro-poor tourism agenda would be advised to focus on developing linkages between smallholder agriculture and (domestic and foreign) tourism.

5. Conclusion

This paper has presented a method for adapting SAMs to estimate the aggregate economic contribution of tourism. This is one of the first tourism-SAMs for a developing country and is able to approximate some of the main outputs of TSAs. An analysis of the tourism-SAM for Mozambique found that tourism remains a small but not insignificant sector in terms of gross demand and value added. Moreover, substantial segmentation between different kinds of tourism was quantified. Notably, foreign business tourists contribute almost half of the value added of foreign tourism, while leisure tourists from the lucrative European and North American markets make a much smaller contribution. A tourism-SAM constitutes the necessary information base from which to develop economic impact models. Applying a simple multiplier analysis to the Mozambican tourism-SAM, it was found that tourism has comparatively strong backward linkages to other sectors and therefore represents an important economic growth engine. However, this assessment is moderated by the finding that tourism is relatively dependent on scarce production factors, especially skilled labour.

Four policy implications stand out. First, greater attention should be paid to the quality rather than quantity of tourism in Mozambique. The analysis has shown that there are substantial differences between visitor types and that (on average) foreign leisure visitors to Mozambique spend relatively small amounts despite their growing numbers. Moreover, despite displaying smaller multiplier effects than foreign tourism, domestic tourism is (on aggregate) a key subcomponent of the tourism sector and should not be ignored. Second, since tourism growth depends on scarce factors of production, government policies to ease any such shortages should be considered. Local language training and investment in provision of core tourism skills would be highly relevant. Third, further analysis is required to understand how tourism quality can be increased along the value chain, thereby increasing spending and linkages to other sectors. This is particularly the case with regard to ‘self-drive’ visitors from the region as well as domestic residents. One attempt in this regard is Jones et al. (2007). These authors find that the absence of a range of leisure activities, poor development of local infrastructure and the extreme seasonality of visitors to tourism resorts remain major challenges. Thus, among other things, the development of Mozambique's core natural tourism attractions, which have qualities of public goods, must be a high priority.

Finally, coherent basic tourism statistics are not made available on a regular basis in Mozambique. This makes it almost impossible to pursue any form of evidence-based policy-making. This paper is an attempt to fill the gap, but ongoing efforts will be needed to monitor the performance of the sector and its changing structure. In sum, tourism shows much promise for Mozambique but must be carefully nurtured to fulfil its potential.

Acknowledgements

Thanks to Channing Arndt for valuable comments. The present study is based on work funded by the Danish Ministry of Foreign Affairs and undertaken in close collaboration with the Mozambican Ministry of Planning and Development and Ministry of Tourism. Especial thanks go to Hanifa Ibrahimo, Céu Matos, Hélio Neves and Paulo Nhampossa. The views expressed in this paper are those of the author; the usual caveats apply.

Notes

¹Further detail and discussion of TSAs can be found in OECD et al. (2001).

²For a general presentation and discussion of SAMs see Reinert & Roland-Holst (1997).

³This methodology has some similarities with the TSA simulations produced by the World Travel and Tourism Council (WTTC & Oxford Economicx [OE], 2009; for a critique see Smith & Wilton, 1997). However, these typically depend on a large number of assumptions taken from cross-country regressions or advanced country statistics. In contrast, we use a detailed range of country-specific data and distinguish between six tourism activities of relevance to Mozambique.

⁴These correspond to standard accounting identities: gross demand = (intermediate consumption + C + I + G + X) = gross supply = (Y + M + intermediate consumption).

Appendix A: Additional tables

Table A1: Aggregate 2003 SAM for Mozambique (10¹² meticais)

		Production, commodities & value added				Institutions
		i	ii	iii	iv	v	vi	vii	viii	ix	x	TOTAL
i	Production (all)	.	148.50	.	.	.	26.22	.	.	.	.	174.73
ii	Commodities (all)	76.77	42.02	.	.	.	65.93	14.70	.	26.85	30.79	257.06
iii	Labour	65.65	.	.	.	.	.	.	.	.	.	65.65
iv	Capital	35.67	.	.	.	.	.	.	.	.	.	35.67
v	Firms	.	.	.	29.22	.	.	0.15	.	.	.	29.36
vi	Households	.	.	65.65	6.46	22.79	.	0.41	.	.	1.34	96.65
vii	Government	.	.	.	.	0.10	.	.	13.67	.	.	13.77
viii	Taxes	−3.37	13.91	.	.	0.92	2.20	.	.	.	.	13.67
ix	Investment	.	.	.	.	1.67	2.29	−1.49	.	5.49	24.37	32.34
x	Rest of world	.	52.63	.	.	3.87	.	.	.	.	.	56.51
	TOTAL	174.73	257.06	65.65	35.67	29.36	96.65	13.77	13.67	32.34	56.51
Source: Author's estimates.
Notes: Given the average 2003 exchange rate of US$1 = MZN 23 500, each unit in the SAM is equivalent to approximately US$42.5 million.

Table A2: Aggregate 2003 tourism-SAM for Mozambique (10¹² meticais)

		Production activities			Commodities			Value added		Institutions
		a	b	c	d	e	f	g	h	I	j	k	l	m	n	TOTAL
a	Economy	.	.	.	148.50	.	.	.	.	.	26.22	.	.	.	.	174.73
b	Foreign tourism	.	.	.	.	2.14	.	.	.	.	.	.	.	.	.	2.14
c	Domestic tourism	.	.	.	.	.	5.59	.	.	.	.	.	.	.	.	5.59
d	Economy	75.22	2.14	5.59	42.02	.	.	.	.	.	62.69	14.70	.	26.03	28.65	257.06
e	Foreign tourism	.	.	.	.	.	.	.	.	.	.	.	.	.	2.14	2.14
f	Domestic tourism	1.55	.	.	.	.	.	.	.	.	3.23	.	.	0.81	.	5.59
g	Labour	65.65	.	.	.	.	.	.	.	.	.	.	.	.	.	65.65
h	Capital	35.67	.	.	.	.	.	.	.	.	.	.	.	.	.	35.67
i	Firms	.	.	.	.	.	.	.	29.22	.	.	0.15	.	.	.	29.36
j	Households	.	.	.	.	.	.	65.65	6.46	22.79	.	0.41	.	.	1.34	96.65
k	Government	.	.	.	.	.	.	.	.	0.10	.	.	13.67	.	.	13.77
l	Taxes	−3.37	.	.	13.91	.	.	.	.	0.92	2.20	.	.	.	.	13.67
m	Investment	.	.	.	.	.	.	.	.	1.67	2.29	−1.49	.	5.49	24.37	32.34
n	Rest of world	.	.	.	52.63	.	.	.	.	3.87	.	.	.	.	.	56.51
	TOTAL	174.73	2.14	5.59	257.06	2.14	5.59	65.65	35.67	29.36	96.65	13.77	13.67	32.34	56.51
Source: Author's estimates from 2003 tourism-SAM.
Notes: Only domestic and foreign tourism types are shown, which are created by aggregating the various tourism sub-types; given the average 2003 exchange rate of US$1 = MZN 23 500, each unit in the SAM is equivalent to approximately US$42.5 million.

Appendix B: SAM analysis methods

Consider a Social Accounting Matrix (SAM) with accounts given by the set S and with row and column accounts indexed by r and c respectively. Define the column vector y as total demand across row accounts, made up of exogenous and endogenous account totals (x and z). Thus:

where A represents a matrix of average expenditure shares (endogenous row entries divided by column totals) and M is a matrix of ‘accounting’ multipliers (see Round, 2003). Each element m_rc of M gives the expected final increase in the corresponding row total y_r arising from a unit increase in exogenous demand for column account c: m_rc  = ∂y_r / ∂z_c .

Four extensions help summarise the detailed information in M :

• Macroeconomic or aggregate multipliers are calculated by summing across selected row accounts within a given column. For a subset of accounts , we define an account-specific summation vector:

  The aggregate multiplier (m_Kj ) for row accounts K with respect to column j is then given by , where the last term is a vector that uniquely selects column from M .

• To compare (aggregate) multipliers across accounts, a simple normalisation is to divide the multiplier of interest by a benchmark, such as the average multiplier for a specific subset of column accounts. To benchmark against a subset of accounts , the normalised version of m_Kj thus becomes .

• Following Tarp et al. (2002), (backward) multipliers not only indicate the increment to a given row account arising from an exogenous injection, but also the rate of return necessary to sustain the multiplier process. If these rates of return are high for scarce factors of production, such as physical capital or skilled labour, then it may be the case that multiplier processes will be hindered by supply-side constraints. Adjusted multipliers are thus given by the ratio of a given (aggregate) multiplier to the ‘scarce factor’ multiplier for the same column account – i.e., . Again, these can be compared across column accounts either directly or after normalisation.

• Following Lima et al. (2004), we calculate employment multipliers. This first involves constructing a column vector of employment coefficients. For each production activity they consist in the number of employment posts divided by its total output. Defining the set of production activities as , it follows that the product of the employment coefficient vector (e_A ) and a vector of production-specific multipliers for a given column j, yields the total expected employment creation arising from a unit exogenous increase in demand for j. Formally, this is given by where the last term is the column vector taken from M corresponding to column j and row accounts A. To aid comparison across column accounts the resulting (total) employment multipliers also can be normalised.