ABSTRACT

The government of Mozambique has over the last 15 years relied on a range of policy instruments to regulate the prices of essential and strategic goods. Price controls in fuels and sugar are some of the most important, but other goods, specifically flour, tomato, tobacco, alcohol and soft drinks have also been directly subjected to fiscal policy instruments. Relatively little is known about who actually bears the burden of these taxes. This chapter aims at filling the gap using a countrywide household survey (IAF 2002-3) to trace the incidence of import tariffs, value added taxes (VAT) and excises (special consumption tax). Government expenditure is close to 26 per cent of GDP, 56 per cent of

which is self-financed by taxes and non-taxes on residents. The remaining 44 per cent is financed by external sources through grants and loan disbursements. Tax revenue is a key income source for the government. It is close to 12 per cent of GDP, and within that category it collects 57 per cent, or 3 per cent of GDP, from taxes on goods and services.1 The size and volatility of external funds threatens fiscal stability, and therefore redistribution of income by public spending. To avoid ineffective cuts in government expenditure with undesirable effects on equity, Mozambique, like many African countries, needs to stabilize its tax revenue either by enlarging the tax base or by increasing taxes on the current base. The history and political economy of government interventions since independence is described in Chapter 2 by Byiers, who also reviews government revenue. Arndt et al. (2007) give an overview of aid flows and their volatility. The fact that most people are employed in subsistence agriculture, and

that merchants typically try to avoid the public eye, makes the formal economy quite narrow. Thus the scope for levying taxes is very restricted, generally limited to product taxes at international borders or on the largest firms. This need not be a problem for anti-poverty tax reforms. To the extent that rich and poor households purchase different consumption bundles, it is possible that some product taxes can be made strongly progressive or regressive. The potential for levying taxes to promote equity therefore needs

careful examination (given information on consumption patterns). The analyst needs to know not only the relative importance of taxed goods in the consumption bundles of different households, but also whether the goods are actually likely to be subject to tax (e.g. purchases from a supermarket or petrol station will in general be taxed, but purchases of cassava from a market stall or street vendor will not). A tax transfers real purchasing power from individuals to the state, and

the incidence of taxes refers to whose real purchasing power is transferred to the state. In general, taxes are said to be progressive if individuals with a low living standard pay a proportionately smaller share of the tax than do people with a high living standard. For example, consumption taxes are generally assumed progressive if there are exemptions or low tax rates on goods heavily consumed by the poor, for example zero VAT on necessities, and higher rates on luxury items. This chapter has two objectives, one substantive and one methodological.

The substantive aim is to examine the effects of price interventions on the distribution of real incomes across different households. I do this by describing formal expenditure patterns for households in relation to their living standards. As a first approximation, a price increase hurts households in proportion to the amount of goods they purchase, so in order to know the distributional effects of a tax reform I depict who consumes taxed items and where consumers are situated in the overall welfare distribution. Such descriptions are important because they provide a first approximation of the immediate effects of a price change. Although such approximations contain a good deal less than we would ideally like to know, they are based on good data, and provide perhaps the only firm information we possess about the effects of change in pricing policy. Moreover, ‘[it] has been found to be a reasonably satisfactory shortcut for the study of a policy’s distributional impact’ (Sahn and Younger 2003: 29). The method, though, is not an appropriate tool for analysing the impact of composite tax categories, such as increasing or decreasing combined VAT or trade taxes. In such reforms there are many indirect effects. For such analysis computable general equilibrium (CGE) modelling is a more promising approach. Accordingly, this chapter focuses on single commodity taxes. The second, methodological, objective is to add a new dimension to the

graphical illustrations that have dominated the non-behavioural tax incidence literature. Lorenz-curves have been widely used in the literature to describe distributional inequality. Similarly, in the incidence literature (surveyed in Gemmell and Morrissey 2003) Lorenz-’type’ curves for different goods are used to compare inequality in welfare loss of different taxes. The intuition is straightforward. If, for example, poorer households tend to consume less of a particular good, say diesel, and more of, say, flour, then reducing taxes on flour and raising taxes on diesel will improve the distribution of welfare. To determine if two curves are different, one can visually inspect whether one curve lies above another, or use a statistical test

developed by Davidson and Duclos (1997) using a finite number of coordinates, to test if the vertical difference between two curves is statistically different from zero. My second, methodological objective is to combine these examination methods by creating confidence intervals for each curve. That done, one can visually inspect whether one curve is statistically different from another. I will do this by bootstrapping the calculations of curves (resampling observations from the data), which provides a way of estimating standard errors, and thus confidence intervals, for an infinite set of points on each curve. Bootstrapping provides a way of replacing analytic standard errors with computed standard errors. A non-technical description of the bootstrap procedure is given in Geweke et al. (2006). The organization of the chapter is as follows. I begin in Section 2 with the

assumptions and limitations of non-behavioural equity analysis. Section 3 provides the theoretical outline that motivates the empirical work, and in Section 4 I present the methodology of both the graphical illustration of tax concentration curves and their bootstrapped confidence intervals. Section 5 presents the data and in Section 6 the results of the substantive analysis are explained. Section 7 concludes.