A. Introduction
It is common for governments in developing countries to manipulate prices of goods and services using a range of policy instruments and institutional arrangements. The motivations behind these price manipulations reflect varying objectives, such as the need to raise revenue, the desire to redistribute income toward the poor or toward politically important groups, the desire to provide protection to domestic producers, or the desire to influence the levels of supply or demand in other related markets where prices cannot easily be influenced.17 For example, the major source of revenue in most developing countries is commodity taxation such as domestic sales and excise taxes and taxes on international trade (Burgess and Stern, 1993; and Keen and Simone, 2004); food prices are often kept artificially low for consumers in order to increase the real incomes of poor households (Pinstrup- Andersen, 1988; and Gupta and others, 2000); and public sector prices (e.g., of electricity, gas, petroleum, coal, other fuels, fertilizers) are also often controlled by governments, reflecting either the perceived strategic importance of these inputs for development or the need to provide these sectors with an independent source of revenue and thus greater financial autonomy (Julius and Alicbusan, 1986).
Reform of these indirect tax systems and publicly controlled prices is often an important component of many structural adjustment programs. Reform of indirect tax systems can take various forms, such as reducing high trade taxes and replacing lost revenue through other indirect taxes, replacing trade and sales taxes with a value-added tax, broadening the value-added tax base to include previously exempt goods and services, or simply raising existing tax rates (Abed and others, 1998; and Barbone and others, 1999). Reform of publicly controlled prices typically involves raising subsidized prices closer to world or cost-recovery prices or possibly replacing government price controls with market-determined prices (Gupta and others, 2000).
Governments and other stakeholders commonly express concerns regarding the potential adverse impact of these reforms on poverty. The desirability of these reforms is usually motivated primarily by efficiency and fiscal considerations, that is, the desire to raise revenue with the least distortion of economic activity. However, the associated price changes can potentially decrease the real incomes of households and thus possibly increase poverty. This potential for adverse effects on poverty may underlie the reluctance on the part of governments and other stakeholders to support such reforms. A credible reform strategy therefore requires an analysis of the likely impacts of proposed reforms on household real incomes and the distribution of these across households, with a particular emphasis on the impact on the poorest households. The insights from these analyses should influence program design (i.e., the structure of tax reforms as well as the speed and sequencing of their introduction), as well as inform the choice of alternative approaches to mitigating these adverse effects.
The objective in this chapter is to set out the various methodological approaches that can be used to analyze the impact of tax and price reforms on household real incomes, to explain how these are related and compare their resource requirements, and to identify general policy lessons from existing empirical studies. Section B describes the various methodological approaches used in the literature and identifies their time, data, and skill resource requirements. It supplements the more general discussion of alternative methodologies in Chapter II by providing a more detailed discussion of methodologies in the context of tax and price reforms. As in Chapter II, the alternative approaches are separated into three categories: partial equilibrium, limited general equilibrium, and general equilibrium approaches. This classification is motivated as follows. The total impact on household welfare can be separated into the direct effect on households arising from the price effects of the reforms and the indirect effect that results once households and firms respond by changing their demand for and supply of goods and services and factors of production (which result in efficiency and revenue impacts). The net distributional effect will depend on the magnitude of these indirect effects and how these indirect effects are distributed across households, such as how the extra revenue is spent. General equilibrium approaches allow for all commodity demand and factor supply responses and thus incorporate both the direct and indirect welfare effects of the reforms. Limited general equilibrium approaches typically focus on a subset of price reforms (e.g., agricultural price reforms) and/or allow for only a subset of household responses (e.g., responses in closely related markets or demand responses alone), thus incorporating only a subset of the indirect effects. Partial equilibrium approaches focus only on the direct effect of reforms on prices and household real incomes.
Section C reviews the findings of the empirical literature using these various approaches and identifies general lessons for policy reform. Reforms that replace trade and sales taxes with a value-added tax typically increase the progressivity of the tax system only if food is exempt or zero-rated. Relatively higher taxes on commodities consumed mostly by higher-income groups (e.g., through additional excise taxes on some energy products) can further increase tax progressivity. It is also important to allow for home consumption of unprocessed food by farm households when evaluating the distributional implications of agricultural pricing reforms because the pattern of market trades, not the distribution of total consumption, determines distributional implications. Incorporating the efficiency consequences of tax reforms through the use of more general equilibrium methodologies is especially important in the context of the reform of trade taxes owing to the narrower tax base (i.e., imports and exports as opposed to total consumption). In general, although narrower tax bases (e.g., in the case of trade taxes or agricultural taxes) may introduce greater potential for improving the progressivity of taxes through tax differentiation, a more inefficient tax structure usually results.
Section D concludes with a summary of the methodological and policy lessons suggested by the review. The argument is made that partial equilibrium analyses can provide very useful insights into the distributional implications of tax and price reforms but should be combined with at least a qualitative discussion of efficiency issues when policy advice is given. General equilibrium analyses are required to quantify the likely magnitude of the efficiency implications of such reforms and are likely to be especially important in the context of trade tax and agricultural pricing reforms.
A more comprehensive analysis of tax and price reforms would need to address other important determinants of successful reform strategies, particularly the administrative and political constraints on reforms. The fact that such issues are not addressed in this chapter should be interpreted not as an implicit assessment of their relative importance for policy advice but rather as a desire to keep the review manageable and focused. Such issues are addressed only indirectly insofar as they influence the set of tax and price reforms under consideration. Note also that the equity and efficiency implications of reforms can be expected to influence both the need for administrative reforms and the likely political economy of reforms.
B. Alternative Methodological Approaches
One can distinguish among three methodological approaches to the analysis of tax and price reforms: general equilibrium, limited general equilibrium, and partial equilibrium approaches. The total impact of a reform can be separated into its direct effect and indirect effect on household welfare.18 The direct effect captures the impact arising from the change in consumer prices owing to the reform, which affects household real incomes. The indirect effect captures the welfare impacts that result from demand- and supply-side responses to the reforms, which have implications for efficiency and revenue. The net distributional effect of a reform will therefore depend on the magnitude of these indirect effects and how they are distributed across households, such as how the extra revenue is spent. The three methodological alternatives differ according to the extent to which the indirect welfare effects are incorporated into the analysis. In addition, the data, time, and modeling resource requirements differ substantially across these methodological alternatives (see Appendix Table A3.1).19
Partial Equilibrium Approach
Partial equilibrium approaches focus solely on the direct effect of reforms on consumer prices and household real incomes. These studies therefore ignore all household and producer responses and focus on the first-order effect on the real incomes of households (or, equivalently, the effect on their cost of living). It is common to interpret these effects as the short-run impact of reforms, prior to household and producer responses. Household responses, such as switching consumption away from taxed goods or toward subsidized goods, tend to decrease adverse welfare impacts and increase beneficial welfare impacts. First-order effects are thus often interpreted as an upper bound on longer-term adverse impacts and a lower bound on beneficial impacts. Producer responses can affect the degree to which the incidence of taxes is pushed onto final goods prices or factor prices and thus also the overall distribution of the welfare impact.
Estimation of these first-order impacts requires household survey information on consumption of the relevant goods and services. Such surveys, which are now widely available for many developing countries, can be used to calculate the budget shares for goods and services. By multiplying budget shares by the proportional increase in the corresponding prices owing to the reforms, one gets an estimate of the proportional change in household real incomes. For example, if a household allocates 10 percent of its total expenditures to food and the price of food increases by 10 percent, a 1 percent decrease in household real income results. Where the prices of many goods are affected, this procedure can be carried out for each good and summed to get the total real income effect. One can then analyze the pattern of these real income changes across households at different levels of income, such as by income deciles.20 If the percentage decrease in income due to taxes is higher (lower) for higher-income deciles, then the incidence of the tax burden is said to be progressive (regressive).
Where reforms involve a change in the prices of intermediate goods (e.g., energy products), one needs to model the pass-through of these price changes to final goods prices. Such a model requires information on the input-output (IO) structure of the economy as well as information regarding which sectors are internationally traded, are nontraded, or have their prices directly controlled by the government. One typically assumes that such price increases are pushed forward onto output prices for nontraded goods but backward onto factor prices or quasi-fiscal deficits for traded or price-controlled sectors. Because the modeling of price shifting is relatively straightforward, subject to data availability, so too is the simulation of price-shifting outcomes.21
Partial equilibrium analyses can provide valuable input into policy dialogue and reform, especially when combined with a qualitative discussion of the likely efficiency effects of reforms. For example, switching taxes to products with inelastic demands and/or negative social externalities (e.g., petroleum products, tobacco, and alcohol), where these are initially small, can be expected to increase the overall efficiency of the tax system. Similarly, broadening the tax base to include previously exempt final-consumption goods and services is also generally expected to improve the overall efficiency of the tax system. Such gains can then be juxtaposed against the distributional implications of these reforms to identify possible welfare gains from reforms or any trade-offs between efficiency and distributional considerations.
The policy relevance of such analyses can be strengthened even further by using household survey data to simulate the likely effectiveness of existing or potential safety net expenditures at mitigating any adverse effects of reforms on poor households. Household survey data combined with knowledge of the design of any existing safety net programs can be used to simulate the potential for such programs to protect the poorest households during the reform process and the implications for the net revenue effects of the reforms. Where information on existing safety net programs is weak, one can construct the likely impact of (well-implemented) hypothetical programs as a way of focusing attention on the need for cost-effective programs. The aggregate first-order income effects also provide an estimate of the first-order revenue impacts of the reforms. Similarly, using household data, one can simulate the likely incidence of alternative social expenditures (e.g., increased education and health expenditures) that may be financed by the revenue gains from reforms. Such analyses help highlight the motivation behind and potential benefits from reforms.22
Limited General Equilibrium Approach
Limited general equilibrium approaches can be separated into two categories: multimarket models and demand-side models. Multimarket models typically focus on a limited set of price reforms (usually agricultural price reforms) and allow for only a subset of demand and supply responses (e.g., in closely related agricultural output markets). Demand-side models focus on demand-side responses only and implicitly assume fixed producer prices so that all tax and price reforms are shifted fully forward onto final goods prices. Both these approaches thus incorporate only a subset of the indirect welfare effects.
Multimarket models are useful when one is interested in price reforms in what are perceived as important markets, such as the rice market. These markets are often directly controlled by governments through a range of policy instruments (e.g., trade taxes or direct price controls) and institutional arrangements (e.g., marketing boards). These models attempt to identify important demand and supply responses in a subset of closely related markets, such as in rice and maize markets, which can have important implications for the efficiency and distributional implications of reforms. Identifying these responses in turn requires the estimation of a system of demand and supply elasticities for well-defined sectors using a combination of household survey information on the pattern of consumption, production, and prices across households and possibly time-series data on production and prices for important crops.
Demand-side models usually cover a broader range of goods and explicitly incorporate the efficiency implications of reforms by allowing for demand-side responses. The basic approach is to calculate the welfare (i.e., combined efficiency and distributional) impact of raising one unit of revenue via different indirect taxes. These welfare impacts can then be compared across commodities to identify revenue-neutral and welfare-improving reforms of the current system, by switching revenue raising from commodities with relatively high welfare costs per unit revenue to those with relatively low welfare costs. Alternatively, such an analysis can be used to determine how to raise extra revenue at the lowest welfare cost.
Incorporating the efficiency implications of taxation into the model essentially allows for the fact that the magnitude of the aggregate welfare loss increases when households respond by reducing consumption, such as by switching away from a taxed good.23 Partial equilibrium models assume away efficiency effects by assuming that demand is fixed so that a household’s share of the tax burden from taxing any commodity is given by its share of total consumption of that commodity. If, to take an extreme example, one cares only about the impact on poor households, then one should increase taxes on goods for which the poor have a relatively low share of total consumption, such as luxuries. However, the fact that households respond to a tax increase by decreasing their consumption of the taxed commodity means that taxes have to be higher to raise a given amount of revenue, thus increasing the total tax burden on households.24 The tax burden on poor households is calculated as their share in the tax burden (i.e., their share in total consumption) times the total tax burden. For example, if luxuries have high price elasticities, then raising revenue by taxing luxuries as opposed to necessities may actually result in a higher tax burden for the poor even though their share of the total burden is lower. Therefore, when elasticities differ sufficiently across commodities, partial equilibrium analyses can give a misleading picture of the best way to raise revenue.
Introducing efficiency implications into the analysis requires the additional estimation of a system of price elasticities. The data requirements for calculating the distributional effects are the same as under partial equilibrium approaches, that is, household-level data with consumption matched to tax categories and rates as well as an IO table where tax cascading is an issue. On the efficiency side, one needs to estimate the price elasticities of demand and/or supply, which can be done using information in the household survey. Typically there is a trade-off between simplifying assumptions regarding the structure of household demand and/or supply and the ease of calculation of the elasticities.25 But a useful approach in practice is to start off with a simple model, run through the analysis, and then increase the sophistication of the analysis over time. The calculation of elasticities within the standard utility and profit-maximizing framework also facilitates the estimation of so-called exact (as opposed to first-order) measures of the welfare impact of tax reforms, such as through the use of equivalent variation measures of the real income impact on households. These measures allow for the fact that households can avoid some of the tax burden by substituting away from highly taxed goods.
General Equilibrium Approach
General equilibrium approaches incorporate indirect effects arising from demand and supply responses in all commodity and factor markets. Implicit in the demand-side general equilibrium models discussed above is the assumption that producer prices, and thus production technology coefficients, are fixed. This assumption is then consistent with the incidence of all taxes being fully pushed forward onto consumers.26 However, in general, producer and factor prices cannot be assumed to be fixed, so some of the burden of taxation is pushed backward onto factor prices.
In the literature, one can find two approaches to capturing these wider general equilibrium effects of taxes. One approach is to use a computable general equilibrium (CGE) model, which involves setting out a fully articulated system of demand and supply functions for each of the various sectors of the economy. An alternative approach uses shadow prices, interpreted as summary statistics from a model that is not fully articulated, in place of producer prices, and the standard welfare analysis goes through as above but using shadow taxes (i.e., consumer prices minus shadow prices) in place of actual taxes.
Building computable general equilibrium models is time, data, and modeling intensive.27 However, once constructed, the model can be used to simulate a wide variety of reforms and market structures. One first sets out a system of commodity and factor demand and supply equations for each sector in the economy and specifies market-clearing and macroeconomic closure rules. The various model parameters are either specified exogenously or calibrated with existing data on consumption, production, and trade flows, leading to a set of equilibrium relative prices. Most of these models tend to be Walrasian in that all commodity and factor markets clear through adjustments in prices, although straightforward extensions are possible to allow for simple factor-market distortions.28 Demand functions are typically some modified version of the Linear Expenditure System whereas production functions are typically of the Constant Elasticity of Substitution variety. Domestic production of traded goods can also be differentiated according to the degree of substitutability in consumption between imported and domestically produced goods so that domestic prices for traded goods do not necessarily move one for one with world prices. Factor markets are often separated into skilled and unskilled labor, irrigated and nonirrigated land, and capital. Total factor supplies are usually fixed but can be reallocated between sectors and even segmented by region within a country. Note that although the model uses country-specific data on consumption, production, and international trade (e.g., from household surveys, manufacturing surveys, and IO tables), the production parameters are often guesstimates based on parameters available in the literature. In this sense, the models are part empirical and part analytical, and sensitivity analyses using different parameters are important.29
Once in place, the model is “shocked” with a tax reform, and a new set of equilibrium commodity and factor prices is calculated. These are used to calculate either first-order or exact welfare changes as above, such as by applying the price changes to household-level data. One can then decompose the total welfare impacts across households into those owing to changes in consumer prices and factor prices. One can also identify separately both the efficiency and distributional implications of the tax reform. Welfare-improving tax reform packages can be constructed by first examining the welfare cost of raising a fixed amount of extra revenue through manipulating individual tax rates and using the insights provided by such an analysis to construct reform packages. Given the sophistication and complexity of the models, with typically many commodity and factor markets interacting simultaneously in the presence of numerous distortions, it is important to decompose the source of the welfare changes for an initial set of narrow reforms so as to develop a clear understanding of the channels through which the welfare effects are operating.
The use of shadow prices provides a less resource-intensive approach for incorporating general equilibrium welfare effects of tax reforms. The parameters of the models underlying the calculation of shadow prices are smaller in number than for a typical CGE model. However, the use of a simpler model is traded off against a greater level of sectoral detail than is typical in a CGE model and the greater flexibility available when incorporating the sensitivity of results to alternative market structures and government policies. For example, the most widely used approach for specifying shadow prices is that of Little and Mirrlees (1974); their approach uses world prices as the basis for shadow prices for traded goods and the marginal social cost of production as the shadow price for nontraded goods. These shadow prices depend on underlying government policies as well as market structure; for example, binding trade quotas are captured by treating the relevant goods as being nontraded. Shadow wages are calculated based on a simple model of the labor market that adjusts for the fact that producer prices are not equated with shadow prices, that labor markets may not clear through wage adjustments but through some form of (often poorly understood) rationing process across sectors, and that the underlying distribution of income is not socially optimal. Other factor markets can similarly be adjusted to allow for price distortions (e.g., in agricultural output and input prices for land, or import prices or the interest rate for capital). These simple shadow-pricing rules are consistent with a wide class of second-best analytical models. The calculation of shadow prices is relatively straightforward, using data available in IO tables and household surveys.
A limitation of the shadow-pricing approach from the perspective of distributional analysis is the fact that the distribution of the indirect welfare effects is not typically analyzed in detail. This reflects in part the fact that the main channel through which indirect distributional effects are incorporated is through the shadow wage rate. This use of a single shadow wage rate compares with the CGE approach in which the changes in factor prices are modeled explicitly and these changes can be imposed on household survey data to analyze the distribution of factor income changes across households in different parts of the income distribution.
C. Lessons from the Empirical Literature
The discussion below of the empirical tax reform literature distinguishes among the literature in terms of the modeling approach used as well as the type of tax or price reform being considered. The first distinction is drawn, as above, in terms of the modeling approach used in the analysis, that is, between partial equilibrium, limited general equilibrium, and full general equilibrium models. Within each of these categories, a distinction is made between three categories of reforms: tax reform, trade liberalization, and reform of public sector prices.30
Indirect tax reforms include such reforms as the introduction of a value-added tax (VAT) system in place of existing sales and/or excise taxes. A VAT is often seen as superior to sales taxes because intermediate inputs are not taxed (thus avoiding distortion of production techniques) and a VAT often applied to a broader consumption base. Sales taxes are typically levied on both final and intermediate goods, resulting in tax cascading as tax rates are levied on output prices that have already been adjusted upward in response to higher production costs reflecting taxes on intermediate inputs. Other issues that arise include the inability to tax the informal sector (including household agriculture), the existence of tax-exempt and zero-rated goods and services, and the choice of differential or uniform VAT rates.
Trade liberalization refers to reforms that replace taxes on international trade with taxes levied on domestic consumption (including consumption of imported goods). Such reforms are often motivated by a desire to reduce the distortion in the domestic production and consumption of traded goods. Any revenue losses can be recouped by replacing such taxes with taxes levied on a broader base that does not differentiate between traded and nontraded goods.31 Because these reforms can be expected to result in relatively large changes in relative producer prices (e.g., between tradables and nontradables), one expects that modeling factor price changes is especially important for this set of reforms.
Public sector pricing reform includes reforms that adjust prices controlled by the government. Developing country governments often control the prices of a range of goods. This control could involve the use of marketing boards, combined with domestic trade and government procurement restrictions, to control the price of agricultural food production. These may be motivated by a combination of revenue (e.g., in the case of important agricultural exports) and distributional concerns (e.g., in the case of subsidized agriculture). In some cases, governments also sell processed foods at subsidized prices subject to quantity rationing. Governments also often control prices of energy, such as petroleum products or electricity, especially in the face of rapidly increasing world oil prices.
Before reviewing the literature, it is useful to distinguish between two distributional concepts, which are referred to as relative progressivity and absolute progressivity? 32 The concept of relative progressivity is commonly used in studies evaluating tax incidence, whereas the concept of absolute progressivity is more commonly used in studies evaluating the incidence of public expenditures. A tax system is relatively progressive if the percentage decrease in income is lower for low-income households. This will be the case if the share of low-income households in the tax burden is less than their share of total income. In this case, then, a neutral (or reference) tax system is one such that the percentage decrease in income owing to the tax burden is equal for all households. A tax system is absolutely progressive if the share of low-income households in the aggregate tax burden is less than their population share—for example, if the bottom 20 percent of the population pays less than 20 percent of the tax burden. The reference for neutrality here is thus a uniform absolute tax burden across all groups. It should be fairly obvious that, in the context of taxes, relative progressivity is a stricter definition of progressivity compared with absolute progressivity, because the former implies the latter but not vice versa. In the context of price subsidies or direct transfers, the opposite holds; that is, absolute progressivity implies relative progressivity but not vice versa. Below, unless specifically stated, the terms regressive and progressive refer to the relative concept.
Partial Equilibrium Studies
Distributional studies based on the consumption patterns in household-level data suggest that reforms that emphasize VAT and excise taxes are progressive. For example, Sahn and Younger (1999a and b) examines the likely incidence of various taxes in six African countries: Cóte d’lvoire, Ghana, Madagascar, South Africa, Tanzania, and Uganda. Their analysis of consumption patterns suggests that gasoline and diesel excise taxes are relatively progressive, followed by the VAT system. Kerosene excise taxes were found to be the most regressive, and export taxes also often appear to be regressive. Because excise taxes are typically levied on products that are thought to have relatively low price elasticities and are associated with negative consumption externalities (e.g., petroleum products, tobacco, and alcohol), and because a VAT is intended to be levied on a broader tax base, tax reforms that shift revenue raising to these tax instruments are typically assumed to improve both the equity and efficiency impacts of the tax system. Reforms that switch tax revenue from trade taxes (which distort both production and consumption) to a VAT (which, in principle, taxes only consumption) are similarly attractive.
The attraction of excises on petroleum products from a distributional perspective is not altered in studies that allow for the cascading effect of these taxes and the indirect effect on households through changes in other prices.33 The progressiveness of petroleum excise taxes above is based on the fact that low-income households directly consume relatively small amounts of petroleum products. However, a substantial proportion (e.g., over 50 percent) of petroleum product consumption is typically used as an intermediate input into transport and other production activities. Therefore, the net effect on households will depend on how these higher costs are passed on to consumer prices. Studies that use IO techniques to model these indirect impacts find that, although these taxes appear to be less progressive (because lower-income groups consume petroleum products indirectly), they are still more progressive than are other taxes.34 Import taxes are similarly found to be more progressive (or less regressive), consistent with the consumption of higher-income groups being more intensive in imported intermediate inputs.
The VAT often appears to be an even more progressive tax once one adjusts for the fact that agriculture and small-scale economic activities are typically VAT-exempt. For example, low-income households often consume food directly from their own production or from small outlets that fall outside the VAT system. This feature tends to make the VAT more progressive. In principle, these households may still pay some tax because the VAT on the inputs into these sectors is not rebated and is thus pushed onto consumer prices. However, the level of this taxation tends to be relatively low and, in any case, many agricultural inputs are often zero-rated.
A number of studies have found that replacing a sales tax with a VAT has made the tax system more regressive (or less progressive). Although the VAT is typically progressive it is often less progressive than the sales taxes it has replaced, mainly because sales taxes have not been imposed on basic foods.
-
Munoz and Cho (2004) evaluate the impact of replacing a system of sales taxes with a VAT in Ethiopia.35 Although the VAT was progressive, partly reflecting the importance of own-consumption for the poorest households (and in spite of exempt goods being disproportionately consumed by the nonpoor), it was less progressive than the sales taxes it replaced.
-
Hossain (1995 and 2003) undertook a similar analysis for Bangladesh, which introduced a VAT in 1991 to replace a system of excise taxes on domestic producers, import duties, and sales taxes levied on both imports and domestically produced goods. The analysis showedm, that a uniform, revenue-neutral VAT would be substantially more regressive with the reform resulting in more than a 2 to 3.5 percent decrease in the real incomes of the lowest income households compared with a 4.5 to 8.1 percent increase for the highest income households. 36 These rates reflected the inclusion of basic cereals and other food within the VAT system.
-
In 2005, an IMF Fiscal Affairs Department team analyzed the distributional impact of introducing a VAT in Bosnia and Herzegovina. The existing sales tax had a standard rate of 20 percent, a preferential rate of 10 percent, and a zero rate for exports and a number of basic food items and certain services. The sales tax was collected at the retail stage except for excisable products (i.e., alcohol and alcoholic beverages, soft drinks, coffee, oil and derivatives, and tobacco and tobacco products), for which it was collected at the importer or manufacturer stages. The proposed VAT was expected to involve a single rate of 17 percent, and the inclusion of previously exempt goods at this higher rate gave rise to concerns within the government regarding the impact of the reform on low-income households. The analysis found that the VAT was slightly less progressive than sales taxes. The tax reform would result in an increase in the average tax burden in the range of 1.9 to 2.6 percent across deciles, the higher percentage being for the lower-income deciles.
The progressiveness of the VAT can be improved by zero-rating basic foods. For example, Hossain (1995 and 2003) examines how zero-rating food in Bangladesh would affect the distribution of the tax burden. The existing system (which taxed food) is compared with a system that zero-rated food grains and vegetables, applied a uniform rate on other goods and services, and levied excise taxes on tobacco, energy goods, and sugar. This system is found to be less regressive, with the reform leading to lower losses for lower-income households, in the range of 1.2 to 2.7 percent, and lower gains to higher-income households, in the range of 0.8 to 6.6 percent.37 Note, however, that although the VAT can be restructured to enhance its progressivity, the substantial leakage of benefits to higher-income households makes such an approach much less attractive than, say, a well-designed and -targeted direct transfer program.
The direct effect of agricultural trade liberalization, which increases prices, appears to be to decrease overall poverty but increase extreme poverty.38 For example, Ravallion and Lokshin (2004) examine the likely distributional consequences of agricultural trade reform in Morocco. The authors take the output price changes generated by a CGE analysis of the removal of cereal import tariffs and apply them to a household survey to identify their first-order welfare effects, taking into account the fact that some households are net producers and others net consumers of cereals.39 They find that the consequent price decreases in cereals result in an increase in the rural poverty head count index, reflecting the fact that the rural poor are on average net producers of cereals. Note, however, that this scenario could also be consistent with the extremely poor benefiting if they were net consumers. For example, in their analysis of an increase in rice prices in Indonesia, Ravallion and van de Walle (1991) find that the extreme rural poor, who tend to be landless and net consumers of rice, suffered decreases in income whereas the moderately poor tended to be net producers and to gain from the reform. In addition, as shown below, the impact of the reforms on factor markets tends to reverse these effects so that these output price effects may be interpreted as short-run impacts prior to adjustments on factor markets (e.g., adjustments in unskilled wages).40
Limited General Equilibrium Models
One of the central findings of early empirical work on tax reform was the strong trade-off between efficiency and distributional concerns (Ahmad and Stern, 1984, 1987, and 1991). Commodities that were very attractive sources of revenue from the perspective of efficiency (e.g., food, which typically has a low price elasticity and thus a low deadweight loss associated with its taxation) were very unattractive from the perspective of their distributional impact (e.g., food is relatively more important in the budgets of the poor). In other words, taxing commodities for which low-income households have a relatively small share of the commodity tax burden will not necessarily lead to a smaller welfare loss for these households. This finding has two important implications for tax reform policy. First, ignoring the efficiency implications of taxes can give a misleading indication as to which commodities to tax more if one wants to minimize the welfare impact on poor households. Second, improved distributional outcomes via commodity taxation are typically bought at the expense of substantial inefficiency. Therefore, using the indirect tax system should generally be viewed as a short-term measure until more cost-effective redistributional policy instruments are developed, such as a well-designed and -implemented social protection program.41
However, it may be possible to identify commodities that are relatively attractive sources of tax revenue even when both equity and efficiency considerations are taken into account. For example, in the Ahmad and Stern (1991) studies of tax reform in Pakistan, the marginal social costs of raising revenue via taxes on rice, edible oils, housing/fuel/light, and clothing were always below the median. The attraction of taxation of cereals, on the other hand, depended on how concerned one was about distribution when setting tax levels. If effective direct redistribution instruments do not exist, then taxation of cereals is not desirable, even though the distributional gains come at a high efficiency cost.
The ranking of commodities may, however, be sensitive to the initial structure of taxes in a country. The marginal social cost of raising revenue through increasing a tax on a specific commodity depends on the existing level of taxes on commodities as well as on the patterns of own- and cross-price elasticities and consumption across income groups. Therefore, even if one expects that the latter patterns are similar across similar countries, the former may differ greatly. Therefore, one should be cautious when transporting policy lessons across countries.
Studies of agricultural price reforms (e.g., as part of trade liberalization) using multimarket models reinforce the partial equilibrium finding that rural extreme poor and urban poor lose from price increases but the urban moderate poor gain. Minot and Goletti (2000) evaluate the distributional impact of the removal of rice export quota controls in Vietnam using a multimarket model. Their model simulates the demand and supply responses in the markets for four staple foods (rice, maize, sweet potatoes, and cassava) in seven regions of the country. The resulting welfare impacts thus take account of important demand and supply responses of households. Their model also allows for the impact of higher rice production in terms of lower world prices. The results of their analysis of the welfare impact of rice price changes indicate that the poorest rural farmers lose from the higher domestic rice prices as do the urban poor, reflecting the fact that both are net consumers. Nonpoor rural households gain, reflecting their net-producer status. A similar but less pronounced pattern was observed when a simple first-order partial equilibrium analysis reflected the fact that the higher rice production reduces world prices so that domestic rice prices do not increase by as much.42
General Equilibrium Models
The results from tax reform analysis can differ substantially when one uses shadow taxes in place of effective taxes. For example, Ahmad and Stern (1990 and 1991) look at the implications of this in the context of tax reform in Pakistan.43 As one would expect given the implicit assumption in the calculation of effective taxes that all goods are nontraded, both sets of taxes differ substantially when goods are traded and subject to trade taxes or price controls. For example, when wheat and rice are treated as nontraded, their revenue collections were applied to the total consumption base, implying tax rates of-1.8 percent and 1.7 percent, respectively. When treated as traded, these revenues were applied to the smaller trade base to give substantially higher rates of -30.3 percent (reflecting import subsidies) and -10.8 percent (reflecting export taxes). But note that effective and shadow taxes can also differ for non-traded goods when factor markets are distorted or the prices of important traded inputs are distorted.
Because existing tax rates enter into the efficiency side of the analysis, not surprisingly, the differences that arise do so with regard to the efficiency implications of tax reform. For example, the large initial subsidy on wheat when shadow prices are used now makes it a much more attractive source of revenue (i.e., by reducing the subsidy) from an efficiency perspective. But such an increase is still as unattractive from a distributional perspective as before. The result is that the trade-off between efficiency and distributional concerns is increased, indicating a potential efficiency gain from having access to an effective transfer system that would enable efficiency considerations to take more prominence when tax rates are set.
The results from the shadow-pricing literature also highlight the fact that in the evaluation of reforms that liberalize agricultural prices, it is extremely important to capture the nature of and constraints on the policy instruments used as well as the precise consumption and production relationship between alternative agricultural commodities. Coady (1997a) extends the above results for Pakistan to allow for a more realistic and broader set of policy instruments, including the fact that households are both producers and consumers of agricultural commodities so that only net market trades (or marketed surplus) can be taxed. In this context, what matters for efficiency is the net trade (as opposed to total consumption) elasticities. When one allows for the fact that net trade is only a small proportion of total consumption and that commodities that are substitutes in consumption may be complements in production, not only are the net trade own- and cross-price elasticities substantially higher than consumption elasticities but their sign can also be different. For example, in Pakistan in the early 1980s, around 60 percent of wheat and 20 percent of rice were consumed on-farm. Reflecting this, and the fact that wheat and rice are sown in rotation on the same land and thus are production complements, the own- and cross-price net trade elasticities were very high and positive.
The constraint of being able to tax only net trades can therefore have substantial implications for both the distributional and efficiency implications of taxes. For example, in Coady (1997a), although rural households were, on aggregate, net producers of wheat, in rural areas poor households tended to be net consumers and nonpoor households tended to be net producers. As a result, low wheat prices (reflecting low procurement prices) acted as a subsidy to low-income households financed by a tax on high-income households, so that low prices were an extremely powerful redistributive instrument. However, net trade elasticities were also very high so that the efficiency costs of low prices were very large. Therefore, the constraint of being able to tax only net trade magnifies the trade-off between efficiency and distribution in setting taxes.
The corollary is that the efficiency gains from reforming the existing system of tax and price controls were also very substantial—in fact, the direct revenue-reducing effect of higher procurement prices for wheat was swamped by the positive indirect revenue effects through higher government procurement replacing more expensive wheat imports as well as increased production and export of rice resulting in higher export tax revenue. In other words, existing taxes were on the wrong side of the Laffer curve.44 This finding also reinforces the argument implicit in multimarket modeling that focusing on the price reform for one agricultural commodity in isolation may give misleading results. For example, the welfare impacts of lowering wheat prices were strongly influenced by the indirect revenue effects from lower rice exports. Note also that the presence of this indirect revenue effect through rice suggests that low wheat prices may be less inefficient if rice prices were higher. In other words, higher rice prices enable the distributional gains from low wheat prices to be achieved at a lower efficiency cost. This sequential approach to price reform could help mitigate the adverse distributional effects from a move to a more efficient price and tax system until an effective social protection system can be developed.
Analyses of tax systems using CGE models have supported the partial equilibrium findings that energy taxes are progressive and that the distributional effect of a VAT depends on how basic foods are treated.45 For example, Go and others (2005) test the partial equilibrium findings by Fourie and Owen (1993) that the VAT in South Africa was mildly regressive. Applying the same partial equilibrium method for 2001, they found similar results, with low income households paying over 5 percent of their income in VAT compared with only 3.5 percent for high-income groups. This situation occurred despite the fact that certain food items (e.g., brown bread, maize meal, milk and milk powder, rice, and unprocessed vegetables and fruits) are zero rated and small-scale firms are not required to register for the VAT.
To evaluate the welfare impact of the current VAT, they remove it completely and replace it with a proportional income tax in order to balance the government budget and not influence the overall incidence of indirect taxes. Whereas the overall tax system (including direct taxes, fuel and excise taxes, tariffs, and the VAT) is progressive, the VAT is found to be mildly regressive. Overall, the high-income groups pay over 20 percent of their income in taxes whereas low-income groups pay less than 10 percent. With the VAT, high-income groups pay less than 4 percent whereas low-income groups pay over 5 percent. When the VAT is removed and revenue replaced by scaling up sales taxes (which are levied on petroleum, beverages, and transport equipment) by 262 percent, the overall tax burden becomes slightly more progressive, indicating that a VAT is more regressive than sales taxes. This regressive incidence partly reflects the relatively high VAT rate on food, which is particularly important for poor households, and sales taxes are higher on goods disproportionately consumed by high-income households. Removing the VAT on food and increasing the base rate on other goods to 16.4 percent so as to keep VAT revenue constant transforms the VAT from a regressive to a progressive tax; low-income households pay less than 2 percent in VAT payments whereas high-income households pay more than 3 percent (compared with more than 5 percent and less than 4 percent, respectively, under the previous VAT structure).46
Analyses of the distributional impacts of trade liberalization using CGE models completely overturn the findings of studies that use partial equilibrium and limited general equilibrium models. The distributional impacts through factor markets are likely to be particularly important in the context of trade liberalization, which results in substantial changes in relative producer prices, that is, the relative prices of traded and nontraded commodities. Reimer (2002) and Hertel and Reimer (2004) provide surveys of the empirical findings from analyses of the distributional impact of trade liberalization. They find that a key channel for these impacts is the effect of reforms on factor markets, particularly labor markets. This finding is to be expected insofar as (1) classical trade theory shows that changes in output prices brought about by trade reforms lead to magnified changes in factor prices for intensively used factors, with the degree of magnification being higher in the short run when some sector-specific factors exist; and (2) households are typically more specialized in terms of sources of income compared with consumption patterns. For example, the removal of an export tax on rice will lead to an increase in domestic prices. If rice is relatively intensive in unskilled labor compared with other sectors, then the unskilled wage rate can be expected to increase proportionally more than rice prices. If the poor are net consumers of rice and receive most of their income from their unskilled labor, then the positive wage effect can be expected to dominate the negative effect of higher rice prices. Therefore, evaluations of the distributional effects of trade liberalization need to incorporate these factor price effects into their analysis through general equilibrium analysis.47
There is also evidence from CGE analysis that the use of optimal export taxes on agricultural exports can have adverse effects on poverty as well as on market share in the long run. Warr (2001) provides a very interesting and rigorous example of the use of CGE modeling to evaluate the distributional implications of controlling the domestic price of rice in Thailand using export taxes. Thailand is perceived as having some monopoly power in international rice markets so that, on the basis of first-best efficiency arguments, a positive export tax is optimal. With an export demand elasticity for rice of 0.25, the optimal tax from this perspective turns out to be about 42 percent and the net welfare gain is 0.63 of a percentage point of GDP compared with a situation with a zero export tax. However, this tax decreases the domestic price of rice so this aggregate gain comes at the expense of both rural and urban poor, reflecting lower prices for poor producers and lower unskilled wages for poor consumers. Incorporating even relatively modest distributional concerns into the analysis substantially changes this outcome, with the optimal situation quickly switching to a subsidy of 20 percent. The results highlight that (1) the main distributional effects come through factor-market prices, (2) higher rice prices are distributionally powerful in the long run even if poor net consumers lose in the short run before unskilled wages increase, and (3) any short-term efficiency gains from export taxes may come at the cost of higher poverty.
Results from CGE analysis also reinforce the finding that there may be substantial welfare gains from using more direct policy instruments to protect the incomes of low-income households in place of adjusting tax rates. For example, Coady and Harris (2004) look at the welfare impacts of using the revenue generated by efficiency-improving tax reforms to finance a (perfectly targeted) direct transfer program in Mexico. The initial indirect tax system was characterized by large agricultural food subsidies and a differentiated VAT structure with a low rate of zero applied to raw and processed food. The tax reforms considered are intended essentially to (1) remove agricultural subsidies, (2) keep the current VAT structure but scale up the rates, and (3) increase the VAT rate on food. When the revenue is raised by removing food subsidies, the cost to households is only 62 percent of the revenue raised (i.e., the cost of a unit of public funds is 0.62). Similarly, when revenue is raised through a single VAT applied to all sectors, the cost of public funds is only 0.95, so the gains from reforming the VAT structure outweigh the losses from the higher average rate required to finance the introduction of the transfer program. The other tax reforms considered all had costs of public funds in the range of 1.05 to 1.07. Although both low-income and high-income households bore a disproportionate amount of this higher tax burden, the existence of a well-targeted transfer program more than offset this negative effect on the poorest households. The gains from introducing such a program are thus twofold. First, the transfers are better targeted than are the subsidies inherent in the tax system. Second, the presence of the program enables one to focus on efficiency considerations when setting VAT rates. These findings were found to be very robust for alternative parameter values for the underlying consumption, production, and trade functions.
When CGE models are used to evaluate tax reforms, it is important to present the sources of the welfare impacts in a transparent manner both to have a clear understanding of the channels at work and to enhance the credibility of findings. For example, in Go and others (2005), the nominal wages of semiskilled and unskilled labor are fixed, reflecting unemployment of these types of labor. Therefore, tax reforms that increase demand in domestic sectors that are intensive in relatively unskilled labor will tend to increase welfare. For example, reducing taxes on a nontraded unskilled-labor-intensive sector can be expected to result in a substantial efficiency gain arising from the conventional decrease in deadweight loss but also the reduction in unemployment. In the language of the shadow-pricing model discussed above, the shadow tax (e.g., consumer price minus the social marginal cost of production) is high for these sectors, so decreasing them can lead to large welfare gains from the elimination of this shadow deadweight loss. It would therefore be useful to identify the various sources of the overall welfare impact, such as changes in output prices, changes in factor prices, and changes in unemployment. Whereas in the Walrasian model, with markets clearing through price adjustment, welfare impacts can be expressed solely in terms of price changes; in models where commodities or factors are rationed (e.g., with unemployment), welfare impacts depend on both price and quantity changes. Note also that results may in general be very sensitive to the mechanisms used to allocate rationed quantities across households with different socioeconomic characteristics, and the relatively simple allocation process used in a CGE may give very different results to a more sophisticated allocation process, say, similar to those being used in recent CGE-microsimulation models.
D. Conclusions
This section summarizes the main issues and findings discussed in the chapter, dealing first with lessons regarding the use of the alternative methodologies, then summarizes the findings from the empirical literature.
Methodological Lessons
(1) Simple partial equilibrium analyses can provide valuable information on the likely magnitude of the impacts of tax and price reforms on household real incomes as well as the distribution of the impacts across households. These studies have relatively low resource costs in terms of data, time, and modeling requirements and can therefore be undertaken on a routine basis. When combined with a qualitative discussion of their likely efficiency and fiscal implications as well as a quantitative analysis of the potential uses of revenue (e.g., mitigating measures or the financing of other social expenditures), these studies can be a very effective input into policy dialogue and the development of credible and acceptable reform strategies. As with all studies of tax reforms in developing countries, it is important that one incorporates the constraints on tax instruments into the analysis (e.g., the inability to tax consumption from own-production in rural areas or informal sector transactions) because these can greatly affect the distributional impacts of reforms.
(2) General equilibrium models are necessary when one wishes to evaluate and highlight the magnitude of the efficiency implications of reforms and the trade-off with distributional impacts. These studies can help highlight the fact that using indirect taxes to mitigate the adverse effects of taxation on the real incomes of poor households can be a very inefficient approach relative to a more direct approach to social protection through well-designed and well-implemented social safety net programs. Although the shadow pricing approach can provide a flexible and relatively low resource cost approach to explicitly incorporating the magnitude of the efficiency impacts, it is less useful when one wishes to disaggregate the distribution of these indirect effects. For this, a computable general equilibrium model is required. Building such a model from scratch is a resource-intensive activity. However, where one is available, the cost of adapting it for the analysis at hand is much lower.
(3) The use of the computable general equilibrium model is particularly valuable when analyzing the distributional impact of reforms that involve significant changes in producer prices, such as in trade liberalization. The distributional impacts of such reforms arise mainly through changing factor prices as opposed to changing consumer prices, and the relative distributional impacts can differ substantially across these two channels. Typically the consumer price effects are interpreted as short-run impacts until factor prices have time to adjust.
Empirical Lessons
(1) Typically, the introduction of a relatively broad-based VAT in place of sales taxes has reduced the progressivity of the tax system, reflecting the broadening of the tax base to include previously exempt goods and services, which are usually relatively more important in the budgets of the poor, and/or the reduction of taxes on goods that are relatively more important in the budgets of higher-income households. Therefore, revenue-neutral reforms will generally lead to gains by upper-income groups at the expense of lower-income groups.
(2) Excise taxes on petroleum products (except kerosene), tobacco, and alcohol are highly progressive, even after allowing for their indirect effects, which tend to be less progressively distributed compared with their direct effects.
(3) Because both these tax instruments are often associated with a more efficient collection of tax revenue—reflecting their broader base, lower price elasticities, and negative consumption externalities—opportunities exist for improving both the efficiency and equity effects of tax reform. The introduction of the VAT and the use of excise taxes on petroleum products and tobacco play an important role in realizing these welfare gains. For example, the use of excise taxes can help to generate sufficiently large revenues, enabling the VAT to be introduced at a lower rate. The introduction of differential rates, with lower rates on goods consumed disproportionately by the poor, would further improve the progressivity of the VAT.
(4) In practice, distributional gains from tax reform often come at the expense of efficiency, and these may be particularly large in the context of agricultural commodities where households are both consumers and producers and the tax base is limited to net market trades of these commodities. It is therefore useful to have some indication of the relative magnitude of these trade-offs across commodities.
(5) Similarly, although differentiating excise taxes within aggregate commodity groups, such as low taxes on kerosene combined with higher taxes on gasoline and diesel, may help to mitigate the impact on the real incomes of the poor, this is likely to come at a high efficiency and revenue cost given the relatively high degree of substitutability between these different commodities, especially over the long term.
(6) In general, manipulating commodity taxes to mitigate the impact on poor households is a very blunt second-best approach to protecting the real incomes of the poor given the substantial leakage of benefits to higher-income households and the potentially large efficiency costs. In cases where price manipulations provide a very effective approach to distribution—for example, low prices for agricultural goods both produced and consumed by rural households—the efficiency cost is very high. The introduction of well-designed and well-implemented social safety net programs provides a more effective way of protecting the poor and can generate substantial efficiency gains by allowing taxes to be raised more efficiently. In this regard, the paucity of information often available on the design, implementation, and performance of targeted social programs is a major constraint on policy advice in this area. If an effective safety net system is not in place then knowledge of the magnitude and pattern of the equity-efficiency trade-off can guide the choice of tax-mitigating mechanism that should be used as a short-term social protection measure.
Appendix 3.1. Theoretical Approach to Evaluating the Welfare Impact of Price Reforms
The empirical literature evaluating the welfare impact of commodity tax or price changes (or reforms) covers a broad range of methodological approaches, which can be usefully classified according to whether they are general equilibrium, partial equilibrium, or somewhere in between (often referred to as limited general equilibrium or multimarket models). This appendix sets out a general equilibrium model that captures the key ingredients of any analysis of the welfare impacts of commodity tax or price reforms, recognizing the three different roles played by commodity taxation and price controls, namely, resource mobilization (i.e., government revenue), resource reallocation (or efficiency), and resource redistribution (or equity). Although it focuses primarily on the distributional consequences of price changes, a comprehensive evaluation of such reforms must recognize the other dimensions of these reforms because these may be the main factors motivating the reforms in the first place and can be expected to have important implications for household welfare. For example, the price changes may reflect the desire to reform the structure of commodity taxes to raise either the same or greater revenue more efficiently. Or the price changes may reflect a desire to make public sector prices better reflect the true cost of meeting demand. Where revenues increase, these may be used to increase coverage of social safety nets among the poorest households or to expand access to valuable publicly supplied services, such as education, health and nutrition, or various infrastructures. A comprehensive evaluation of price effects, therefore, needs to take all these implications into account.
The appendix starts by describing the model and deriving analytical equations identifying the main ingredients in the general equilibrium analysis of the welfare impacts of marginal reforms. This analysis is then used to interpret partial equilibrium analyses as a special case, making explicit the assumptions behind these analyses. This is followed by a brief discussion of how the analysis needs to be adapted for the evaluation of non-marginal tax and price reforms. Finally, the various data and modeling requirements of each approach and the trade-offs inherent in these choices are set out.
The Model 48
Consider an economy made up of households (denoted by superscript h =1, H), producers (denoted by superscript g = 1, G), and the government. Households choose consumption bundles (xh) based on the following constrained maximization problem:
where
where
Producers choose net supply vectors yg (with positive entries for outputs and negative entries for inputs) to maximize the following profit maximization problem:
where
where Vh(.) is the household’s indirect utility function and W(.) is a Bergson-Samuelson social welfare function. If V* (s;to) denotes the maximum value function of this problem, then, from the envelope theorem, the gradient of V* is the same as the gradient of the following Lagrangian:
where
The first term on the right-hand side is the direct effect on social welfare and the second term is the indirect effect capturing the social value of the additional excess demands generated by the proposed reform. Note that the shadow prices will also depend on the specification of choice variables.
So far the government’s budget constraint has not been explicitly introduced. However, by Walras’ law, if commodity markets balance then so too does the remaining government budget constraint. Then, as shown by Drezé and Stern (1987), using Walras’ law the above Lagrangian can be equivalently rewritten as
where R is the shadow revenue of the government, defined as
where
The relationship between shadow prices and market prices will depend on both the structure of markets and government policy. For example, for a small open economy the shadow price of traded goods is the world price, and a tariff will drive a wedge between market prices and
shadow prices with
Price Reforms
The above model is now used to derive analytical equations for evaluating the welfare impact of a marginal change in the consumer price of commodity i (i.e., dqt), for example, owing to a change in the tax rate or a change in public sector pricing. Differentiating the Lagrangian with respect to qt (and assuming that producer prices are fixed52) results in
where
The first term in Equation (3.3) gives the direct effect on household welfare of the price change and just says that, for marginal price changes, the level of household consumption gives a money measure of the welfare loss from a price increase—this loss is valued using the social welfare weight of each household. The second term in brackets gives the indirect effect on social welfare arising from the change in consumer demands brought about by the price change, which leads in turn to a change in revenue reflecting a more or less efficient pattern of consumption and production.
For example, consider the case in which the price reform involves an increase in the price of a publicly supplied nontraded commodity that was previously priced below its marginal cost of production; so τ< 0 for this commodity, leading to too high a level of consumption. The price increase will lead to an efficiency gain when it results in a decrease in the demand for this commodity, which will show up in higher government revenue. If the price increase leads to a shift in the demand toward other commodities that have consumer prices above (below) their shadow value, then the increase will also lead, other things being equal, to a more (less) efficient pattern of consumption and production and this efficiency gain (loss) will again show up as an increase (decrease) in revenue. So increases (decreases) in revenue are associated with more (less) efficient consumption and production patterns. Note that the indirect effect is likely to be relatively large when demand responses are large and/or existing price distortions (i.e., shadow taxes) are large.
The above equation provides a useful focus for classifying the various methodologies used in the empirical literature evaluating the welfare impact of price reforms. These can be categorized into three different types of approaches: (1) general equilibrium, (2) limited general equilibrium, and (3) partial equilibrium.53 Equation (3.3) gives the full general equilibrium welfare impact of a price change and requires one to allow for responses in all commodity and factor markets. Limited general equilibrium approaches focus on a few key markets or on just the demand side of the economy. For example, multimarket models typically focus only on key agricultural sectors when analyzing agricultural price reforms. Demand-side models implicitly assume producer prices are fixed and focus on household demand responses, ignoring producer responses. Both approaches ignore factor-market responses. Partial equilibrium approaches ignore responses completely.54 Chapter III reviews the empirical literature in detail using the above three classifications.
The above discussion has also focused on marginal price reforms when in practice price reforms are often sizable so that second-order (and higher-order) welfare effects are likely to be important. However, it is straightforward to incorporate this aspect into the above analysis. For the direct welfare effect on households (i.e., partial equilibrium approaches) the first expression is simply replaced with an estimate of the compensating or equivalent variation of a price change (Triest, 1990). This, of course, involves estimating demand responses, at least in the relevant market. The size of the price response determines how much the direct welfare impact of the price reform differs between marginal and nonmarginal reforms, with the former tending to be an overestimate (underestimate) of the latter, reflecting households’ abilities to switch away from (toward) a commodity whose price has increased (decreased). Note also that these responses may in principle differ substantially across income groups and that this could have important implications for the distribution of welfare changes.
For nonmarginal reforms one should in principle also replace local estimates of responses with estimates of responses over the related price change. In addition, the assumption of constant producer prices is less likely to be valid so one may want to model the supply side of the economy more explicitly. One also needs to consider the implications for what constitutes the appropriate shadow price and whether it changes with the reform, such as the extreme case where the price increase results in the commodity changing from being imported to being exported.
The obvious advantage of analyzing marginal reforms is that data requirements are less demanding because one needs only information on the pattern of consumption across households (for the direct effect) and estimates of aggregate price elasticities (for the indirect effect). Cross-sectional household surveys, which are now widely available, typically provide sufficient data for these purposes.
Note, however, that in the above model it is implicitly assumed that all consumption could be taxed. In many developing countries, especially for agricultural commodities, households consume some or all of what they produce (so-called own-consumption) so that this proportion of consumption cannot be taxed. In this case it is useful to treat these producers as being part of the household sector and replace total consumption with net trades and consumption elasticities with net trade elasticities in Equation (3.3). 55 Note that the pattern of net trades across households, such as with poor and landless households being net consumers of food and large-landholder households being net producers, may increase the distributional power of price controls because low prices are effectively a subsidy to the poor financed by a tax on the rich. However, because net trade elasticities are likely to be relatively large, such price controls are likely to be extremely distortionary so that more efficient transfer instruments probably exist.
It is also common for studies to focus explicitly on the distribution of direct and indirect welfare impacts across households in different parts of the income distribution. To focus on income distribution it is useful to rewrite Equation (3.3) above as
where
direct income effect.56 Fully differentiating the social welfare function, assuming lump-sum incomes and government revenue are constant and the price of commodity i is controlled by the government, results in
Equating the last term on right-hand side of Equation (3.4) to that in Equation (3.5) results in
where λD and λl capture the distributional impact of the direct and indirect income effects respectively (arising solely from changes in commodity and factor prices) and μ captures the efficiency impacts of the price reform. The distributional parameters are essentially a weighted average of welfare weights, with the weights being the share of each household in the total direct and indirect income effects, respectively. For example, in the absence of any efficiency effects, welfare will increase only if the direct income effect is distributed more (less) progressively (regressively) than the indirect income effects, that is, if λD > λl The efficiency parameter μ is greater than one if there are additional efficiency gains from the reforms.
Analyses of the distributional impacts of reforms typically capture these effects through indices similar to λD above (Coady and Skoufias, 2004). For example, if welfare weights are {1,0} for {poor, nonpoor} then this distributional index is equivalent to the share of the aggregate income effect accruing to the poor. Or one could use the concentration coefficient, which aggregates income shares based on household rankings in the income distribution (regardless of the size of income differences). Or one may look at the posttransfer distribution of income and compare it with the pretransfer distribution using inequality indices such as the Gini coefficient, the Atkinson index, or the General Entropy Family of inequality indices. These indices either explicitly or implicitly assume some underlying welfare weights, so it is important to undertake sensitivity analyses. Alternatively, one could just present numbers on the shares accruing to households in the various income deciles or plot-related concentration curves.
One may also wish to focus on the size of the income effects (and not just their distribution) and the resulting effects on income poverty in order to inform a policy discussion on the design of appropriate compensating policy measures. Or one may wish to understand the distributional effects across socioeconomic groupings (e.g., by region or ethnic group) to understand implications for horizontal (as opposed to vertical) equity or for political economy.
Appendix 3.2. Alternative Price-Shifting Models
In general, tax and price reforms will involve changes in the prices of intermediate goods. The extent to which these price changes are passed forward onto output prices or backward onto factor prices will depend on such things as the structure of the economy, such as how substitutable different commodities are with internationally traded goods, as well as on the degree of control the government has over prices in general. To the extent that taxes are pushed forward onto output prices, the actual tax content of the final equilibrium price, that is, the effective tax (te), will exceed that of the nominal or statutory tax on the sector (t). Obviously, subsidies on intermediate goods may mean that te < t.
In addition, in practice, existing tax systems in developing countries include a range of different tax components (e.g., excise taxes, trade taxes, and value-added taxes) and some sectors fall outside of the tax sector (e.g., agriculture and the informal sector and the existence of sectors exempt from value-added taxation). It is therefore important when modeling the price effects of tax reforms to capture these important features of the tax system, which can be expected to have important implications for the equity and efficiency effects of tax reforms.
This appendix sets out a simple price-shifting model that can be used to capture these different features of tax systems and that simply requires use of available data on the input-output (10) structure of the economy. The model allows for different degrees of tradability within each aggregate sector of the 10 table. Note that to the extent that price changes are not pushed forward onto output prices they must be pushed backward onto factor prices. The model presented does not follow through the implications of these factor price changes or of these changes on output prices. The model is very similar in spirit to that presented in Hughes (1986) and also has much in common with the effective tax model of Ahmad and Stern (1984 and 1991).57 Once price changes have been calculated, one can apply these to household-level consumption data (as discussed in the main text of this chapter) to evaluate the real income effect on households from the equilibrium output price changes.
A Price-Shifting Model for Energy Taxation
The implications of higher costs for output or factor prices will, of course, depend on the structure of the economy: for example, whether commodities are traded internationally or nontraded, the nature of commodity taxes, and the extent to which prices are controlled by the government. Therefore commodities are first grouped into three broad classifications reflecting the assumed relationships between higher production costs and output prices:
(1) Cost-Push Sectors. These are sectors in which higher input costs are pushed fully onto output prices. They therefore can (loosely) be thought of as nontraded commodities.
(2) Traded Sectors. These are sectors that compete with internationally traded goods and whose output prices are determined by world prices and the import or export tax regime. Higher input costs are not pushed forward onto output prices, so the brunt of these higher costs is borne through lower factor prices or lower profits.
(3) Controlled Sectors. These are sectors in which output prices are controlled by the government. Therefore, the relationship between output prices and production costs depends on if and how the government adjusts controlled prices. If controlled prices are not adjusted then the burden of higher costs will be borne by factor prices, profits, and/or government revenue.
When modeling price changes, it is useful to think of aggregate commodity categories (e.g., the aggregate categories available from an IO table) as made up of a certain proportion of cost-push, traded, and controlled commodities, with these proportions given by a, P, and y, respectively. These proportions should obviously sum to unity and never be negative:
Alternatively, one could interpret these proportions as capturing the degree of tradability of a single commodity.
The technology of domestic firms is captured by a standard IO coefficient matrix, A, with typical element aij denoting the cost of input; in producing one unit of output j—think of units of output defined such that they have a user price of unity so that price changes can be interpreted as percentage changes. Consistent with the interpretation of A as capturing an underlying Leontief (i.e., fixed coefficient) production technology, aij can be interpreted as the change in the cost of producing a unit of j owing to a unit change in the price of input i.
For traded sectors, user prices, q*, are determined by world prices, pw, and by trade taxes (including tariffs and sales taxes), t*:
and
and both terms on the right-hand side will be specified exogenously by the reform package under consideration.
For controlled sectors, producer prices are determined by pricing controls (say,
Alternatively, one could think of the difference between user prices and average unit production costs as an implicit tax with the revenue accruing to the public sector enterprise and thus entering government revenue through the quasi-fiscal side of the budget. The formula for price changes is then given simply as
where the right-hand side is specified exogenously in the reform package.
For cost-push sectors, the relationship between user prices and producer prices is given by
where qc is the price paid by users of a commodity and pc is the price received by producers, the difference between these being any sales or excise taxes, tc imposed by the government. Producer prices are, in turn, determined as follows:
where q are the user costs of intermediate inputs and w are factor prices. For these sectors, cost increases are assumed to be fully pushed forward onto user prices so that factor payments are fixed. From Equation (3.9) one gets
Using Equation (3.10) and the IO coefficient matrix, and assuming factor prices are fixed, the change in producer prices is derived as
where A signifies a price change and all price changes are interpreted as n x 1 row vectors where n is the number of commodity groups, (α, β,γ) are now n x n diagonal matrices, and A is an n x n IO coefficient matrix. Substituting in from Equation (3.10)’’ Δqc and using Equation (3.7)’ for Δq* results in
so that
where K—(I- αA)-1 with I being an n x n identity matrix. The typical element of the inverse matrix K, kij, captures the combined direct and indirect use of cost-push sector i used to produce one unit of cost-push sector j. Notice that if the only price changes are changes in controlled prices, then
The change in sector aggregate prices is then given by
This price change is analogous to the effective tax calculations by Ahmad and Stern (1984 and 1991), except their model assumes imported goods to be perfect complements to domestically traded goods, whereas the current model assumes different degrees of substitutability with imported goods.
A Price-Shifting Model for VAT Reforms
For the purpose of deriving the price effects of tax reforms, the study categorizes goods and services (or sectors) into four groups as follows:
(1) Vatable Sectors (V). These are sectors that fall within the VAT system and include both sectors subject to positive VAT rates as well as those that are zero-rated. These sectors receive rebates of the VAT paid on any inputs but not of other indirect taxes.
(2) Exempt Sectors (E). These sectors are exempt from the VAT and therefore do not receive rebates of the VAT or other indirect taxes paid on their inputs.
(3) Excisable Sectors {X). These sectors have an excise tax imposed on their output and are also exempt from the VAT so that they do not receive rebates of any indirect taxes paid on their inputs.
(4) Vatable/Excisable Sectors (Z). These sectors have both an excise tax and a VAT imposed on them. They therefore receive rebates on the VAT paid on inputs. In addition, the VAT may be levied on the excise-tax-inclusive price.
(5) Traded Sectors (T). These sectors have to compete with imported goods. They may be subject to the VAT or other sales taxes.
For each of these sectors, one can define two sets of prices: the prices received by producers (p) and the prices paid by users (q). These differ in the presence of indirect taxes.58 For each sector, the following holds:
where superscripts are used to denote sectors and t denotes the relevant indirect taxes. Taxes are set exogenously through government policy decisions. User and producer prices are determined endogenously, except for the traded sector where these are exogenously determined by world prices and taxes. User prices are determined by producer prices and exogenously fixed taxes. Producer prices are determined endogenously through the following system of equations:
where Aij is an IO matrix with typical element αij denoting the amount of sector i used in producing one unit of sector j. The main difference between Equations (3.19) through (3.22) is that whereas Vand Z pay producer prices on V inputs (because VAT on inputs is rebated), other sectors pay user (i.e., VAT-inclusive) prices. In other words, for E and X, VAT operates just like sales or excise taxes. Note also that all sectors pay tax-inclusive prices for E, X, and T inputs 59 and, to the extent that ad valorem taxes are levied on producer prices that already include these taxes, the tax system is subject to cascading. This system of equations essentially treats sectors V, E, X, and Z as being nontraded sectors with all costs (including taxes) being pushed fully forward onto user prices. This is not the case for traded goods whose prices are determined by border world prices and any taxes levied on imports (including VAT). Therefore, the greater the importance of traded goods, the less the overall tax system is subject to tax cascading.
In order to solve out for producer prices, it is useful to think of sectors as aggregate sectors made up of different components of vatable, exempt, excisable, vatable/excisable, and traded goods and services. Then, substituting in for consumer prices, one can rewrite Equations (3.19) through (3.22) as follows:
where
Taking all terms with p to the left-hand side of Equation (3.23) and solving out for p gives60
where
Equation (3.24) solves out for the level of producer prices as a function of world prices and indirect taxes. One can rewrite Equation (3.24) in terms of percentage price changes (dp) by simply replacing t with dq and use Equations (3.14) through (3.19) to calculate dt. Note that by interpreting existing taxes as changes (i.e., dt = -t), one can derive basic prices, defined as the prices that would exist in the absence of taxes, as 62
Effective taxes, defined as the difference between current tax-inclusive and basic prices, can then be derived as
As written, fe is a percentage of current tax-inclusive prices, that is, the proportion of taxes in the final user price. If one prefers to have effective taxes expressed as a percentage of basic prices then one can simply derive these as fe0 = fe/ (1 + dq).
The main objective of such an analysis is to evaluate the effect of the tax reforms on household real incomes. The effect on household real income can be calculated by multiplying proportional price increases, dq or dq*, for each good or service by the corresponding budget share for the household and aggregating this product across goods and services. One can then evaluate the magnitude and distribution of this real income effect for different tax reforms, including a reform that involves a change in tax regime.
Appendix 3.3. General Lessons from Tax Theory
This appendix briefly identifies key insights provided by the theoretical literature in addressing the issue of how to design or reform a structure of commodity taxes. One can find three potential roles for commodity taxation in the literature:63 (1) raising revenue to finance government activities when lump-sum taxation is not available to the government; (2) reallocating resources to bring about a more efficient reallocation, such as Pigovian taxes; and (3) redistribution of income when lump-sum taxes and transfers are not available to the government. The evolution of the theoretical literature in terms of the extension of tax rules to incorporate each of these roles is briefly discussed below. The theory of optimum taxation is distinguished from the theory of tax reform; the former focuses on what a system of optimum taxes would look like whereas the latter focuses on the identification of welfare-improving reforms. There is obviously an intimate link between these two strands of the literature because an optimum tax system is one from which no welfare-improving reforms are possible.
Optimum Commodity Taxation
One of the earliest contributions to the formal literature on the structure of optimum commodity taxes was Ramsey (1927). This paper examines the optimum structure of commodity taxes when the government had a fixed revenue requirement that could not be financed through lump-sum taxation. The main insight from this paper is captured by the so-called Ramsey Inverse-Elasticity Rule, which indicates that taxes should be higher on commodities with low price elasticities of demand. The basic intuition behind this result is straightforward: optimum taxes require that the reduction in (compensated) demand be the same for all commodities, which requires relatively high tax rates on commodities with relatively low price elasticities.
Probably the definitive papers in the area of optimum commodity taxation are by Diamond and Mirrlees (1971a and 1971b), which also focus on the revenue-raising role of commodity taxes. These papers developed the basic analytical framework used in the modern optimum commodity tax literature. A key contribution of this work was the identification of the conditions under which production efficiency was desirable, implying that taxation of intermediate goods was undesirable. In particular, intermediate goods taxation is potentially desirable only when some final consumption goods cannot be optimally taxed (e.g., because of the presence of consumption from own-production in rural areas or the existence of an untaxable informal sector) but can be taxed indirectly via taxes on inputs. Similarly, input taxes may be desirable as a way of taxing economic profits where these cannot be taxed directly via an optimum profits tax. If all consumption by households can be taxed and an optimal profits tax exists, then all revenue should be raised via taxes of final consumption (e.g., a value-added tax, or VAT), without distinguishing between commodities according to whether they are traded or nontraded. Note that these assumptions also imply that trade taxes are not desirable.
Diamond (1975) extends the above model to allow for the redistributive role of commodity taxes. This paper shows that, relative to the tax-efficient structure identified above, taxes should be lower on commodities that are relatively more important in the budgets of low-income households. The fact that these commodities (e.g., necessities) are often those with relatively low price elasticities suggests the existence of a trade-off between equity and efficiency when taxes are set; that is, reducing the burden of taxation on lower-income households is likely to come at the cost of a higher aggregate burden.
Drezé and Stern (1987), building on an earlier model by Guesnerie (1979), develop a fairly general model that captures a wide range of second-best worlds and incorporates a resource reallocation role for commodity taxes in addition to the revenue-raising and redistributional roles. They derive a Generalized Ramsey Rule, which shows that the previous rules go through as before but now with actual taxes replaced by shadow taxes defined as the difference between consumer prices and shadow prices. This rule incorporates, for example, the standard Pigovian argument for commodity taxation as well as other second-best departures from the standard Ramsey Rule.
Reform of Commodity Taxes
The earliest work in the area of tax reform was undertaken in the context of trade taxation where revenue requirements were not considered and optimal lump-sum taxes and transfers were implicitly assumed to be available to the government to raise and redistribute revenue. Therefore, this literature was concerned primarily with efficient resource reallocation. Dixit (1975 and 1985), which played an important role in integrating a separate body of theory on trade taxation into the standard analytical approach employed in public finance theory, examines the conditions under which radial reforms (i.e., an equiproportionate reduction of commodity taxes) and concertina reforms (i.e., reducing only the highest taxes) are welfare improving. Such reforms are obviously less interesting in the context of a government revenue requirement.
Ahmad and Stern (1984) introduced the general theory of tax reform in the context of a government revenue requirement and a redistributive role for commodity taxes. This theory was extended by Dréze and Stern (1987) to incorporate a resource allocation role. Their approach is to identify the marginal social cost (MSC) of raising a unit of revenue using alternative commodity taxes while keeping all other taxes fixed. Revenue-neutral and welfare-improving tax reforms can then be identified by decreasing the tax rates on commodities with high MSC and replacing the lost revenue by raising taxes on those with low MSC. Although this approach helps identify potentially welfare-improving and revenue-neutral reform strategies, the focus on single tax rates keeping all other rates fixed introduces limitations to this approach from a policy perspective.
Another related strand of the tax reform literature identifies the conditions under which specific (and commonly discussed) tax reform packages are welfare improving. A corollary of the Diamond-Mirrlees (1971a) production efficiency theorem is simply that replacing trade taxes with optimal commodity taxes, keeping revenue constant, will always be welfare improving under the conditions discussed above (i.e., that all consumption can be taxed and that production exhibits constant returns to scale or, if not, that optimal profit taxes are available to the government).64 Keen and Ligthart (2002) examine the conditions under which replacing any trade taxes with consumption tax reforms that keep consumer prices constant and increase production efficiency are unambiguously welfare improving (i.e., they improve production efficiency, increase revenue, and keep household welfare unchanged). A corollary is that eliminating all tariffs and adjusting domestic consumption taxes to keep consumer prices unchanged will unambiguously increase both household welfare and government revenue. Note that the Diamond-Mirrlees result keeps revenue unchanged and requires optimal commodity taxes.
A key assumption in the above literature is that all the consumption base can be taxed. However, this assumption is implausible for developing countries where many agricultural households consume out of their own food production and the informal sector (including small-scale rural and urban enterprises) is a substantial proportion of the economy. An important finding in Diamond and Mirrlees (1971a and 1971b) is that if the total consumption of a given commodity cannot be taxed then production efficiency is desirable only among the fully taxed sectors. The informal sector can then be incorporated as part of the household sector, with production inputs treated in the same way any household purchases are. In addition, the efficiency implications of taxation will then depend on net trade elasticities (Newbery and Stern, 1987). Note that under these conditions there is no presumption that taxation of inputs is undesirable and such taxes may be a desirable way of taxing the informal sector indirectly (Stiglitz and Dasgupta, 1971; Heady and Mitra, 1982; and Newbery, 1986). This presumption is essentially the issue addressed by Emran and Stiglitz (2005), who show that, in the presence of an informal sector not subject to VAT, replacing trade taxes with revenue-neutral VAT reforms is not necessarily welfare improving.
Tax Reform and Redistribution
A key assumption in all of the above literature is that the government has access to optimal lump-sum transfers. Once one introduces a distributional role for indirect taxes, for example, because direct transfer systems are unavailable or ineffective, then one can say very little regarding the welfare impacts of the tax reform packages conventionally discussed in the literature (Coady and Drzée, 2002).65 These welfare impacts will depend on the specific patterns of production and consumption found in a country, particularly their pattern across income groups, as well as the range of policy instruments available for redistributing income (e.g., see Deaton and Stern, 1986). However, a common finding in the empirical literature is that using indirect taxes (including price controls) is a very ineffective instrument for protecting poor households from the adverse effects of taxation (e.g., reflecting the fact that these households account for a small proportion of the consumption base of most commodities) or, when such taxes are distributionally attractive they come at very high efficiency costs (Coady, 1997a). One therefore expects substantial gains from developing effective direct transfer programs, that is, the gains from having better targeted transfers plus those from not having to incur the standard trade-off between equity and efficiency when using taxes to redistribute income (Coady and Harris, 2004).
Alternative Approaches for Evaluating the Welfare Impacts of Tax and Price Reforms
Alternative Approaches for Evaluating the Welfare Impacts of Tax and Price Reforms
| Characteristics | Resource Requirements | Modeling Requirements | |
|---|---|---|---|
| Partial equilibrium | Incorporates only the direct effect of reforms, focusing only on welfare impact arising through changes in consumer prices. Ignores efficiency effects resulting from demand and supply responses. Can also incorporate revenue effect and alternative mitigating measures. | Data: Requires information on tax and price system and reforms as well as household survey data on consumption of relevant commodities. Input-output tables are required when evaluating changes in prices of intermediate goods. Time: Basic analysis can be completed in around two person-weeks once relevant data have been collected and processed. | Simple models capturing the key features of the tax and public price system are relatively easy to construct and implement using household survey and tax data. Typically only welfare effects through consumer price changes are captured. |
| Limited general | Incorporates direct effect and a subset of indirect effects, e.g., demand and supply responses in a subset of (typically agricultural) markets or just demand effects in all final product markets. Ignores factor-market responses, Can also address alternative mitigating measures. | Data: As above but now requires detailed information on sectors being analyzed and demand and supply elasticities. Time: Basic analysis can be completed in around eight person-weeks once relevant data have been collected and processed. | Needs to model sector supply and demand responses explicitly as well as interaction between sectors. Relevant modeling skills take longer to acquire. |
| equilibrium | |||
| (Multimarket models; demand-side models) | |||
| General equilibrium | Incorporates direct effects and indirect effects through product and factor markets. Can address equity and efficiency implications of a wide range of policy scenarios including mitigating measures. | Data: Very data-intensive approach requiring detailed information on consumption and income patterns of households, factor intensities of all relevant sectors, and trade statistics. Typically have to make assumptions about wide range of consumption and production response parameters. Time: Can take up to six person-months to organize data and get basic analysis completed. | Approach is very modeling intensive and therefore requires strong modeling skills. |
| (Shadow-pricing approach; computable general equilibrium models) |
Alternative Approaches for Evaluating the Welfare Impacts of Tax and Price Reforms
| Characteristics | Resource Requirements | Modeling Requirements | |
|---|---|---|---|
| Partial equilibrium | Incorporates only the direct effect of reforms, focusing only on welfare impact arising through changes in consumer prices. Ignores efficiency effects resulting from demand and supply responses. Can also incorporate revenue effect and alternative mitigating measures. | Data: Requires information on tax and price system and reforms as well as household survey data on consumption of relevant commodities. Input-output tables are required when evaluating changes in prices of intermediate goods. Time: Basic analysis can be completed in around two person-weeks once relevant data have been collected and processed. | Simple models capturing the key features of the tax and public price system are relatively easy to construct and implement using household survey and tax data. Typically only welfare effects through consumer price changes are captured. |
| Limited general | Incorporates direct effect and a subset of indirect effects, e.g., demand and supply responses in a subset of (typically agricultural) markets or just demand effects in all final product markets. Ignores factor-market responses, Can also address alternative mitigating measures. | Data: As above but now requires detailed information on sectors being analyzed and demand and supply elasticities. Time: Basic analysis can be completed in around eight person-weeks once relevant data have been collected and processed. | Needs to model sector supply and demand responses explicitly as well as interaction between sectors. Relevant modeling skills take longer to acquire. |
| equilibrium | |||
| (Multimarket models; demand-side models) | |||
| General equilibrium | Incorporates direct effects and indirect effects through product and factor markets. Can address equity and efficiency implications of a wide range of policy scenarios including mitigating measures. | Data: Very data-intensive approach requiring detailed information on consumption and income patterns of households, factor intensities of all relevant sectors, and trade statistics. Typically have to make assumptions about wide range of consumption and production response parameters. Time: Can take up to six person-months to organize data and get basic analysis completed. | Approach is very modeling intensive and therefore requires strong modeling skills. |
| (Shadow-pricing approach; computable general equilibrium models) |