Overview of Methodological Approaches

Generally, three broad types of methodological approaches are employed to analyze the poverty impact of trade liberalization and devaluation in empirical studies: general equilibrium, limited general equilibrium, and partial equilibrium.⁶⁸’⁶⁹ Studies with general equilibrium approaches evaluate both the direct as well as the indirect effects that arise from a specific reform on household welfare. The direct effect captures the impact on welfare arising from the change in consumer prices resulting from the reform that affects the household’s real income. The indirect effect captures the welfare impacts that result from demand- and supply-side responses to the reforms, which generate efficiency and revenue impacts. Limited general equilibrium approaches (sometimes referred to as multimarket) similarly evaluate the direct effect but incorporate only a subset of market responses, for example, by focusing on responses in key agricultural markets related to an agricultural commodity whose price is changed or only on final demand responses, ignoring factor demand responses. Partial equilibrium approaches ignore all responses and focus solely on the direct effect.

A comprehensive study of the distributional impact of trade liberalization and devaluation ideally requires a general equilibrium approach. Because these policies have significant effects on factor prices in addition to changing consumer prices, general equilibrium approaches are necessary to incorporate relevant indirect welfare effects into the results and avoid serious partial equilibrium biases.

In practice, however, data limitations and time constraints typically determine the level of analysis in empirical studies related to distributional effects of trade liberalization and devaluation. Information requirements for the estimation of fully fledged general equilibrium models are quite high for the average developing country. Therefore, many empirical studies of trade liberalization and devaluation are conducted at the partial equilibrium level with implicit or explicit assumptions regarding the significance of general equilibrium effects that could significantly alter their conclusions.

Another practical difficulty facing studies is being able to separate empirically the impact of each policy and define the relevant counterfactual. On many occasions several economic policies are implemented at the same time the policies of interest are implemented. Therefore, disentangling the separate effects may prove very difficult. The issue of the counterfactual refers to defining the initial position to which to compare the outcome postimplementation of the economic policy to be analyzed. For example, in the case of devaluation, comparing the situation after devaluation with an unsustainable initial situation would not be appropriate. Ideally, the comparison should be with the situation that would have prevailed in the absence of devaluation.

The remainder of this section discusses the main economic mechanisms linking trade policies and devaluation to poverty use in empirical studies. The discussion below follows very closely that in Winters, McCulloch, and McKay (2004).⁷⁰

Economic Impact Mechanisms of Trade Liberalization and Devaluation on Poverty

Winters, McCulloch, and McKay (2004) refer essentially to two types of mechanisms: those of a macroeconomic nature and those of a mainly microeconomic nature. Macroeconomic mechanisms are those that have a mostly indirect impact on poverty whereas microeconomic mechanisms tend to affect the household’s budget constraints in a more direct fashion.

The main mechanism of a macroeconomic nature that is common between devaluation and trade liberalization is the impact of these policies on growth and then from growth to poverty. The specific channels by which growth is affected by these policies differ, though. Trade liberalization can affect growth through its impact on productivity. Devaluation can correct a real exchange rate overvaluation and stimulate growth through renewed export growth that reflects improved competitiveness resulting from lower real wages. Another shared mechanism is the impact of the policies on the overall level of revenues and expenditures of the government. However, the specific impacts on revenues and expenditures and therefore the poverty impact of this mechanism will differ across countries, depending on the characteristics of revenues and expenditures.⁷¹

Two additional macroeconomic mechanisms of impact on poverty are applicable to trade liberalization specifically. One is the increase in the openness of the economy. Increased openness may increase macroeconomic volatility because of the increased possibility of economic fluctuations related to the exposure to new risks such as terms of trade shocks, which may create difficulties or opportunities for different types of economic agents. The second mechanism is the creation or destruction of markets. Significant changes in relative prices resulting from trade liberalization may cause the disappearance of markets with a potentially adverse poverty impact and/or the creation of markets for goods and services that were previously unavailable, with a potentially positive poverty impact.

On the microeconomic mechanisms side, trade liberalization and devaluation are likely to impact goods and factor income prices directly and indirectly through responses in revenue and expenditure policies of the government. Regarding factor price changes, the impact on wages is of particular interest because in some countries most of the poor rely on labor markets for the bulk of their income. When labor market rigidities preclude the full adjustment of wages to a new equilibrium in the short term, the impact on unemployment also becomes relevant because the adjustment in the labor market will occur partly through changes in wages and partly through changes in unemployment depending on the degree of flexibility in the labor market.⁷² With respect to changes in government policies, for example, these may come as a response to the possibility that trade liberalization may lower trade revenues that need to be replaced to maintain fiscal sustainability.

The microeconomic transmission mechanism of border price shocks (created by trade liberalization or devaluation) to local prices and the ability of households to respond to price changes are very important to incorporate into any assessment of the poverty impact. Border shocks are typically not fully reflected in local prices. The extent of transmission may be limited by a number of factors, including transport costs and other costs of distribution; the extent of competition between traders and the functioning of markets more generally; and infrastructure, domestic taxes, and regulation. In addition, the ability of households to respond to price changes seems to vary and may significantly alter the welfare impact of the shocks generated by devaluation and trade liberalization. Important determinants of the level of price response of households seem to be the level of access to key inputs, markets, and infrastructure because these determine the extent to which households can protect themselves from the negative consequences or take advantage of the opportunities provided by trade liberalization or devaluation.⁷³

Economic Growth

Most theoretical papers highlight different channels by which increased openness positively affects the growth of productivity and therefore economic growth in the long term. Productivity growth can be spurred by improved access to technology, improved access to intermediate and capital goods, benefits of scale and competition, and constrained government incompetence and corruption (Lucas, 1988; Romer, 1990; Grossman and Helpman, 1991; and Rivera-Batiz and Romer, 1991). The most notable exception to this general view is that increased openness, and therefore increased importance of comparative advantage in determining production patterns, may lead countries into less dynamic sectors (e.g., primary extraction) and therefore less economic growth (e.g., Rodriguez and Rodrik, 2001).

Empirical papers broadly support the view that increased openness is positively associated with growth in the long term.⁷⁴ In addition, some empirical studies highlight the need for complementary policies such as adequate investment policies for the benefits of trade liberalization to materialize (e.g., Taylor, 1998; and Wacziarg, 2001). Methodological difficulties in empirical papers relate to problems in measuring the trade stance, the endogeneity of the variables used to measure openness in studies,⁷⁵ the possibility that openness may be correlated to other policies that could be causing growth, and that most of the results are based on the use of cross-country regressions with their many conceptual shortcomings.⁷⁶

Economic growth is found to reduce poverty on average in empirical studies.⁷⁷ Recent research has confirmed the long-maintained position that economic growth is on average positively correlated with poverty reduction (Ravallion, 2001; and Dollar and Kraay, 2002). However, there may be considerable variations in the impacts on different vulnerable groups. A recent study (Ravallion, 2004) concludes that there is considerable heterogeneity in the welfare impacts of trade liberalization among the poor, with both gainers and losers. Notwithstanding the general positive impact on poverty, it is not difficult to imagine situations in which there may be economic growth and increases in poverty in the short term. In particular, if productivity increases as a result of improved access to technology, rationalization gains defined as the shrinking or elimination of inefficient firms may reduce employment and increase unemployment in the short term if there is little labor market flexibility. If the specific sectors that are to undergo the rationalization process are where most of the poor work, poverty may increase in the short term.

Overall, the evidence on this channel of impact suggests that trade liberalization is likely to promote economic growth and reduce poverty on average in the medium to long term. Even if the poor do not benefit directly from increased demand generated by trade liberalization, they may do so indirectly, as those who do benefit directly increase their demand for inputs and consumption of goods and services. For example, some studies such as Delgado and others (1998) argue that these spillover effects for several African countries may be significant. The effectiveness of spillover effects in increasing the income of the poor depends on the ability of local business to respond to the increased demand. However, on some occasions, as highlighted above, it is possible that trade liberalization could have, through the rationalization process that may accompany it, an adverse short-term impact on the poor. However, evidence in the literature supporting these kinds of situations is limited. ⁷⁸ A recent survey on the relationship between trade, growth, and poverty by Berg and Krueger (2003) is also consistent with these results.

Macroeconomic Volatility

Increased openness could theoretically increase the exposure of the economy to output volatility. A more open goods market may imply higher specialization and reduced risk of spreading opportunities in production, implying higher output volatility⁷⁹ (Razin and Rose, 1994). In addition, the more open the economy, the larger the impact of a given terms of trade volatility on output. If the increased terms of trade volatility increases output uncertainty, then investment could be reduced with a negative impact on growth (Basu and McLeod, 1991).

The empirical support for the increased output volatility resulting from increased openness hypothesis is inconclusive. Easterly and Kraay (2000) found that small states that are generally more open than larger states have more volatile growth, albeit at higher averages. However, other studies found no correlation or negative correlation between openness and output volatility (Razin and Rose, 1994; and Lutz and Singer, 1994, respectively).

Even if it were accepted that increased openness resulting from trade liberalization contributed to increased output volatility, this increased openness would not necessarily imply increased household vulnerability to external shocks and risk of falling into poverty. First, as shown by Easterly and Kraay (2000), although volatility may be higher in more open economies, mean income also tends to be higher, so the chances of a household falling into poverty will depend on the relative sizes of these shifts. Second, a large body of literature on poor households shows that these households take steps to insure themselves against bad outcomes or to protect themselves from the effects of negative shocks.⁸⁰

However, there is evidence that the poor are much less insured and much less able to cope with negative shocks than are the nonpoor. Evidence is provided in Jalan and Ravallion (1999). In particular, some mechanisms to cope with shocks, such as reducing education and health investments, are likely to have negative long-run consequences for the poor. This suggests that if trade liberalization is expected to increase the variability of the income of vulnerable groups, attention should be paid to the effectiveness of mechanisms available to the poor to smooth consumption.

Creation and Destruction of Markets

Greater openness arising from trade liberalization can result in a wider variety of commodities being available and new opportunities for production. Examples of these positive effects are illustrated in Booth and others (1993) and Gisselquist and Grether (2000). In the first study, the greater availability of goods at international prices was regarded as beneficial by the rural poor in Zambia even though it entailed an increase in prices. In the second study, the increased availability of inputs resulting from the liberalization provided substantial benefits to agricultural producers in Bangladesh.

There have been cases in which changes in domestic marketing arrangements accompanying trade liberalization measures have destroyed markets. A subset of households can become completely isolated and suffer substantial income losses in these circumstances. An example in Winters, McCulloch, and McKay (2004) refers to the elimination of marketing boards that imposed artificially low prices on farmers for their produce. If marketing boards were providing financial services to small farmers preliberalization by allowing them to secure inputs against future output and if such services are discontinued postliberalization, small farmers could experience significant income losses even if prices for their output have risen significantly. Winters (2000) discusses the case of rural Zambia where the abolition of the official maize-purchasing monopoly in the early 1990s led to the abandonment of purchasing maize from poor farmers in remote areas and the suspension of the financing arrangements. Transportation costs were partially responsible for this situation. Heavily deteriorated roads made purchasing maize from farmers in remote areas unviable. For more details on these issues, see Chapter V.

Prices of Commodities

The impact of trade liberalization on the prices of commodities and the resulting impact on household welfare shows mixed results depending on household characteristics. Barrett and Dorosh (1996) and Minot and Goletti (2000) analyze the impact of rice market liberalization in Madagascar and Vietnam, respectively. The first study estimates that one-third of poor households could lose from higher prices because they were net consumers of rice. The second study instead reports a more mixed picture depending on the type of households. Whereas poor rural households were net sellers of rice and benefited from the reform, urban poor households were net consumers and suffered adverse effects.

Evidence suggests that changes in border prices following liberalization may not be fully reflected in changes in local prices. The transmission mechanism tends to smooth the price changes at the border for different reasons.⁸¹ In agriculture, many export crops, especially those of small farmers, are sold through public or private marketing agencies that set the prices below the f.o.b. export prices. The differential reflects transport, marketing, and other costs of the agencies, including monopsonistic profits. For example, increases in international prices may not be fully reflected owing to increased marketing costs or a change in the profit margin required by the agencies. Another example is the existence of high transaction costs, such as high transport costs. These costs may reduce the impact of changes in border prices on local prices. In the extreme, as Goetz (1992) suggests, high transport costs prevent some households from trading in many parts of sub-Saharan Africa, particularly poor households in remote rural areas. A more recent example of high transport costs is Nicita (2005), which concludes that the Doha trade liberalization round will not significantly affect Ethiopia in part because the poor are isolated from markets residing in remote areas and engaging in pervasive subsistence.

There is substantial evidence on the responses of households to price shocks that affect them as consumers or producers.⁸² In particular, this evidence could also be applied to trade liberalization if it is thought of as an external price shock. Responses relate to households protecting themselves from adverse effects as well as taking advantage of opportunities. Coping strategies could include reorganizing households to locate dependents in low-cost locations and workers in households that could employ them, increasing hours of work, postponing deferrable consumption, reducing investment in human capital, and dissaving. In particular, the responses in savings and investment highlight that the trade liberalization impact may also have an intertemporal dimension depending on the magnitude of these responses. Frankenberg, Smith, and Thomas (2003) illustrates these coping mechanisms for the Indonesian crisis over the 1997-98 period and how they could reduce decreases in total family income to only about half of the fall in individual real earnings. With respect to illustration of production responses and taking advantage of opportunities, agricultural producers seem to be quite responsive to price incentives when they have access to the necessary inputs, information, and credit as illustrated in McKay, Morrissey, and Vaillant (1997).

Evidence from the agricultural sector suggests that the nonpoor are in general better placed to respond to price shocks and opportunities than are the poor. This is due to better access to land, physical and human capital, key productive assets, and infrastructure. The better-endowed nonpoor also tend to have better access to credit, which is typically related to endowments, given the need to provide collateral. Examples of the impact of insufficient capital or key productive assets on poor small farmers include Deininger and Olinto (2000) for Zambia; López, Nash, and Stanton (1995) for Mexico; and Heltberg and Tarp (2002) for Mozambique and Madagascar.

The asymmetric capacity of the poor and nonpoor to respond to shocks is generally interpreted as a justification for complementary policies to trade liberalization. These policies would help poor households enhance their access to key inputs, education, and infrastructure and take advantage of the opportunities brought about by trade liberalization or protect themselves for short-term adverse impacts, if any. A recent study by Porto, Brambilla, and Balat (2004), which analyzes the relationship between trade and poverty in Zambia, finds that complementary policies (i.e., extension services in agriculture and job programs supporting employment opportunities to heads of households) could be important to allow households to take full advantage of trade liberalization. Another complementary policy deemed necessary for the poor to take advantage of trade liberalization, also specifically in the agricultural sector, is liberalizing domestic product and factor markets (Anderson, 2004). Hertel and Winters (2005) also stress the need for complementary policies and the need to liberalize services trade and investment to allow the poor to take advantage of the new opportunities brought about by the Doha trade liberalization round.

The flexibility of vulnerable households to respond to shocks and the capacity to provide adjustment support are important considerations that need to be factored in when deciding on the pace of trade liberalization. Although it is sometimes politically convenient to push for trade liberalization in a nongradual fashion, abrupt liberalizations could sometimes exacerbate short-term adverse negative effects because households do not have sufficient time to respond and prepare adequately for the change in relative prices. On the other hand, the pace of liberalization could be sped up if adequate social safety net instruments that facilitate the targeting of adjustment support to vulnerable groups are in place.

Factor Prices and Unemployment

The impact of trade liberalization on factor prices,⁸³ particularly on wages, is found to be more important in magnitude as a channel of impact on poverty than are commodity prices. Studies that highlight this point are Coxhead and Warr (1995) for the Philippines, Harrison and others (2003) for Turkey, Warr (2001) for Thailand, and Porto, Brambilla, and Balat (2004) for Zambia. This result is also highlighted in a recent survey (Hertel and Reimer, 2004). This empirical result is consistent with the well-known Stolper-Samuel son theorem in classical trade theory.⁸⁴ The Heckscher-Ohlin theorem establishes that a country has a comparative advantage in the good that intensively uses the relatively abundant factor. Because the price of the good for which the country has a comparative advantage will tend to be relatively lower than in other countries in autarky, trade liberalization will tend to increase the price of that good. The price increase of this good not only will generate an increase in the real return of the abundant factor used in the production of the good but also implies that the increase will be larger in percentage terms than the increase in the price of the good. Another strand of literature has focused on the effect of trade liberalization on industry-level wages as tariffs are frequently applied at an industry level (Goldberg and Pavcnik, 2007).

The skill premium—the difference between skilled and unskilled wages—has been increasing in developing countries at the same time that trade liberalization episodes occurred. The skill premium result is based mostly on the Latin American experience. Several empirical papers surveyed in Goldberg and Pavcnik (2004) provide empirical support for several hypotheses linking the increasing skill premium to trade liberalization episodes.⁸⁵ This finding could be relevant for poverty analysis in countries where the main source of income for the poor is unskilled wages and where such wages have been falling. The emphasis of this literature, though, is on explaining the reasons for the increased demand for skilled labor and the resulting increase in income inequality. The attention to cases in which the wages of unskilled declined and poverty was increased is limited. However, in a recent paper related to the literature, Topalova (2006) analyzes the regional impact of trade liberalization in India by looking directly at poverty measures such as the head count ratio and the poverty gap. Although the study did not assess the India-wide effects of trade liberalization on poverty, Topalova finds that rural areas with a high concentration of industries that were disproportionately affected by tariff reductions experienced slower progress in poverty reduction. This result reflects the limited labor mobility across regions, industries, and districts in India documented in Topalova (2004). The limited labor mobility reflects, in part, labor market rigidities fostered by inadequate regulations.

There is little evidence on adjustment costs of liberalization related to transitional unemployment and if the poor are disproportionately affected. Some studies, such as Matusz and Tarr (1999), discuss transitional unemployment generated in the manufacturing sector and find relatively minor costs and short unemployment durations. Attanasio, Goldberg, and Pavcnik (2004) find that increases in the probability of unemployment before and after tariff reductions were not larger in the manufacturing sector (where tariff cuts were the largest) than they were for workers with the same observable characteristics in the nontraded sectors. Evidence of transitional unemployment and of whether the poor are disproportionately affected on specific sectors of interest, such as the agricultural sector, the service sector, or the informal sector seems to not be available. However, the limited evidence of adjustments costs of liberalization does not imply they are negligible. Adjustment costs are likely to be greater the more protected the liberalized sectors were originally, the greater the magnitude of the implied price shock, and the more inflexible the factor markets, particularly the labor market.

Government Revenues and Expenditures

The reduction in trade tax rates that occurs with trade liberalization does not necessarily need to be accompanied by a reduction in revenues from trade taxes. This is especially the case for countries with a high level of protection. Some measures associated with liberalization, such as the simplification of tariff structures, conversion of nontariff barriers into tariffs that give the same level of trade, and the elimination of exemptions on the purchase of imported inputs used by sectors protected with a high tariff on their outputs, may be revenue enhancing. Another consideration is that a reduction in tariffs could potentially be revenue increasing if tariffs were beyond the revenue-maximizing point.⁸⁶ In the medium term, the fall in consumer prices of tradable goods and the expansion in the revenue base associated with increased real income and improved growth rates are likely to imply improved revenues from a value-added tax (VAT), profit and income taxes, and other taxes. Examples of cases in which trade liberalization enhanced revenue collection can be found in Ebrill, Stotsky, and Gropp(1999).

If risks of falling trade revenues generate the need for fiscal adjustment, vulnerable groups could be affected depending on the instruments used to replace the trade revenues or to reduce expenditures. To the extent that trade liberalization ultimately entails reduced revenues from trade taxes, measures may be needed to compensate for revenue losses or to reduce expenditures. In particular, on the revenue side, indirect taxes and, to a lesser extent, strengthening of income taxes of both individuals and companies are possible ways to replace the lost revenues. Distributional concerns are typically focused on indirect tax replacement, which is deemed regressive. This is particularly the case with VAT, especially where foods are incorporated at the standard rate. On the expenditure side, cutting of social expenditures is another area of concern regarding impact on the poor.

Recent empirical evidence suggests that although revenue from trade taxes fell in many developing countries, recovering them through indirect taxes posed no significant difficulties. Trade revenues were recovered in many cases by introducing a VAT. Keen and Simone (2004) show that, on average, countries in all income groups managed to raise indirect tax revenue by about as much as trade tax revenue fell. The exception is countries in sub-Saharan Africa where, although trade taxes were reduced, indirect tax collections remained mostly constant.

The risks of impact on vulnerable groups of increasing indirect taxation or reductions in social spending as a response to reduced trade revenues should not be overplayed. Indirect taxes such as the VAT may not necessarily be regressive. A Poverty and Social Impact Analysis (PSIA) study regarding the replacement of a sales tax with a VAT in Ethiopia (Munoz and Cho, 2004) shows that in spite of being less progressive than the sales tax, the VAT is still progressive. In addition, the income loss to the most vulnerable groups did not exceed 1 percent of their consumption. The limited impact on the poor was fundamentally due to the high ratio of in-kind transactions that are not taxed and because many of the exempt goods under the sales tax were not disproportionately consumed by the poor. This result is consistent with Sahn and Younger (1999b) on the progressivity of VAT in African countries. More arguments belying the regressivity of the VAT can be found in Ebrill and others (2001). With respect to reductions in social spending, Winters, McCulloch, and McKay (2004) survey several papers suggesting that there is consensus on the fact that social spending has in general been relatively protected, especially when compared with capital expenditures.

General Implications for Policy Analysis

The literature seems to suggest that given its positive impact on economic growth, overall trade liberalization tends to reduce poverty in the long run but short-term negative effects are possible. In particular, the aggregate gain in the welfare of the poor in the long run would come from the impact of trade liberalization on productivity and growth and the negative correlation between growth and poverty. However, there will likely be winners and losers in the short run. Specific characteristics of the countries’ economies and how these are related to poverty will determine if the poor tend to be among the winners or losers. Important characteristics of the countries’ economies are (1) the relationship between the different economic sectors, labor intensity, and the degree of initial economic protection and (2) the extent to which factor markets, especially the labor market, are flexible. Even if poverty were to decrease in the short term, there may still be winners and losers among the poor if the poor are a heterogeneous group.

The possibility of adverse short-term impacts on poverty should not be an argument for not pursuing trade liberalization. Rather, the likelihood of positive long-run effects on the poor argues for adequately deciding the pace of reform, identifying the losing poor, and considering possibilities to compensate these groups for negative short-term impacts. The rationale for compensation is also given by the fact that the poor tend to be less able to cope with shocks than the nonpoor and that some mechanisms the poor use to respond to shocks may be detrimental for their well-being in the long run. Compensation mechanisms should take into account the magnitude of the welfare loss faced by poor households, factoring in the transmission of border price shocks to local prices, the ability of poor households to smooth and respond to shocks, and the government’s resource constraint consistent with macroeconomic stability.

The degree of access that the poor have to key assets, information, and infrastructure should be an important consideration when assessing the poverty impact of trade liberalization. Numerical estimates of poverty impacts of trade liberalization on the positive or negative side tend to be quite modest. A frequent explanation for these small impacts in the literature is that the impact of trade policies often depends on complementary policies for trade liberalization. These policies help the poor gain access to key assets, infrastructure, and information that may be essential to take advantage of the benefits of liberalization or protect themselves from short-term adverse effects. The relevance of these policies will depend on the variety of country contexts.

More generally, in thinking about poverty impacts of trade liberalization it is important to identify which channels are going to be emphasized and the rationale for the choice. Papers in the literature typically tend to use models that capture only a subset of channels at a time in their poverty impact analysis and do not always justify their choice of channels. Reasons for the different choices are likely to be data limitations, differences in the economic structure, heterogeneous characteristics of poverty, and differences in implementation of trade liberalization policies. In addition, the need to keep models tractable is likely to be another important constraint. However, justifying the choice of channels explicitly is important in making a convincing argument that the model underlying the poverty impact estimates captures the main relevant effects and is therefore adequate for policy analysis.

Economic Growth

The theoretical impact of devaluations on short-term output growth is generally ambiguous.⁸⁸ In Agénor and Montiel (1999), the many different channels by which aggregate demand and supply may be impacted by devaluation and could potentially conflict with each other are illustrated in the context of the dependent economy model. For example, although nominal wage stickiness resulting from fixed nominal contracts could imply an expansionary supply effect in the short run as a result of reduced real wage costs, an income redistribution effect (where income is transferred from individuals with high marginal propensity to consume, such as wage workers, to individuals with low marginal propensity to consume) could imply a contractionary effect on the demand side. On the investment side, increasing prices for capital and intermediate inputs could also have contractionary effects on aggregate demand. In addition, structural characteristics of the economies may change the direction of impact of a specific effect. For example, in economies where private sector asset holdings are not indexed to the domestic price level, a devaluation reduces the real value of existing wealth and therefore negatively affects aggregate demand. If the private sector instead holds mostly foreign assets, the opposite result would hold.

The empirical evidence on the short-term impact of devaluations on output growth is also inconclusive. This is due in part to the many possible conflicting theoretical channels of impact that may vary depending on country structures and the several different empirical methodologies used to test the theoretical models. Agénor and Montiel (1999) reviewed the empirical evidence in this area. Reviewed macro simulation studies such as Gylfason and Schmid (1983) find that devaluations are expansionary in 8 out of 10 countries in their sample because of the prevalence of expansionary expenditure switching effects. Other studies of the same type, such as Gylfason and Radetzki (1991), find that although devaluations are expansionary in developed countries, they are contractionary in developing countries. Econometric studies, such as Edwards (1986 and 1989), find that devaluation is contractionary in the short run on a sample of 12 developing countries.⁸⁹

A strong supply response is a necessary but insufficient condition for a growth-enhancing devaluation to be poverty reducing in the short term. The structure of the economy and the characteristics of the poor are key determinants of the impact of devaluation on poverty. For example, if the economy’s tradables sector is relatively large and mostly composed of mineral production (which makes intensive use of natural resources and capital), the poor are mostly employed in the labor-intensive nontradables sector (i.e., services), and their wage is their main source of income, a devaluation may increase poverty. The relative price shift toward tradable goods could imply an increase in output growth resulting from increased tradable production. However, this same change in relative price would imply a reduction in real wages as a result of Stolper-Samuelson effects and reduced employment in the nontradables sector in a neoclassical setting. If labor market rigidities, such as wage inflexibility, say, resulting from fixed labor contracts and/or limited sectoral mobility of labor, are assumed, a devaluation may imply increased unemployment as a result of the contraction of the nontradables sector and only a partial absorption by the expanding tradables sector in the short term. The increase in unemployment could then have a negative poverty impact. For a more detailed discussion see Stewart (1995).

In economies with vulnerable balance sheets, a devaluation could generate significant balance sheet effects and trigger a financial crisis with strong and persistent negative effects on growth and indirect effects⁹⁰ on poverty. Balance sheets may be vulnerable, for example, if the private sector, including the financial system, and/or the government have large unhedged liabilities in foreign currency. Countries with a history of inflation and that experienced extensive currency substitution as a result are more likely to be in this type of position. As suggested in Allen and others (2002), the transmission mechanism by which such weaknesses in the balance sheets could trigger an external balance of payments crisis typically goes through the domestic banking system. For example, concerns about the government’s ability to pay its debt as a result of the devaluation due to the its foreign currency denomination of debt could quickly destabilize the confidence in the banks holding this debt and lead to a deposit run. Alternatively, a change in the exchange rate coupled with unhedged foreign liabilities in the corporate sector can also undermine the confidence in the banks that lent to that sector. The run on the banking system could then take the form of a withdrawal of cross-border lending by nonresident creditors, or the withdrawal of deposits by domestic residents. A banking crisis typically tends to exert strong negative short-term effects on output through financial accelerator channels, and these can be quite persistent.

Prices of Commodities

The impact of devaluation on inflation is an important element to consider when analyzing poverty impacts. If a devaluation leads to high and variable inflation as indicated in Gunter, Cohen, and Lofgren (2005), this is likely to have a negative impact on poverty through the negative impact on economic growth and the lack of protection of the real value of the incomes and assets of poor people from inflation.⁹¹

The impact of devaluation on inflation depends on several characteristics of the economies. Burstein, Eichenbaun, and Rebelo (2004) isolate some characteristics of economies that can dampen the response of inflation to a devaluation by generating a slow adjustment of nontradable goods prices. These characteristics are low share of tradable goods in consumption (which is related to the openness of the economy), high domestic distribution margin for imported goods, low elasticity of demand for exports, and low elasticity of substitution between tradables and nontradables. In addition, wage-setting mechanisms of the economy will play a crucial role in determining the magnitude of the price responses of nontradable goods to devaluation.⁹²

The impact of relatively large devaluations on inflation seems to have been relatively low in contractionary devaluations during the 1990s. When analyzing nine large contractionary devaluations in the 1990s, Burstein, Eichenbaum, and Rebelo (2002) suggest that this has been the case. The reasons presented are that tradable goods require nontradable distribution services to be sold and that households substitute higher cost imports with low-quality nontradable goods. Both arguments predict an increased share for nontradables in the price index. Because nontradables prices typically tend to adjust slowly⁹³ or are adversely affected in devaluation episodes, the direct impact of the devaluation on the price of tradable goods in the price index is at least partially offset by the impact on nontradables prices.

The impact of devaluation on cost of living for the poor depends on the share of tradable and nontradable goods in their consumption baskets. The larger the share of tradable goods, the larger the possibility of a negative impact. For example, Minot (1998) shows for the case of Rwanda that increased prices of clothing account for almost half of the negative effect on low-income rural households.

There is evidence of several types of household responses to large devaluation episodes to mitigate adverse effects. McKenzie (2001) finds that the main smoothing mechanism used by households during the 1994 Mexican crisis was a change in the composition of their consumption. The paper shows that households have increased their expenditure share on certain food items even more than Engel’s law would predict⁹⁴ by reducing expenditure on luxury goods. Labor supply response was found to be weak. Fallon and Lucas (2002) described how households have smoothed their incomes in the context of the Asian crisis. Labor supply responses seem to have been stronger, increasing labor force participation in particular. Increased reliance on transfers has been another important smoothing mechanism.

The increase in the relative price of tradables implied by devaluation seems to have been poverty reducing in African economies.⁹⁵ In African economies, poverty is mostly rural and most of the rural poor rely on earnings from the labor-intensive agricultural sector for their livelihoods (Sahn, Dorosh, and Younger, 1997). Studies such as Dorosh and Sahn (2000) typically conclude that because poor rural farmers⁹⁶ are net producers of agricultural tradables, they tend to benefit because devaluation increases the relative price of tradables. The urban poor, on the other hand, tend to be negatively affected because they are net consumers of tradables. More generally, these studies highlight the role of devaluation in reducing or eliminating the negative impact of overvalued exchange rates on agriculture as a source of poverty reduction.⁹⁷ Impacts, however, may be of limited magnitude because in many African countries the poor have only limited access to markets and the cash economy and rely on self-production for subsistence.

The results on the poverty impact of devaluation for the African economies cannot be easily extended to Latin American countries. There are key structural differences in the economies as suggested in Agénor (2004) and Fallon and Lucas (2002): The poor are more evenly split between urban and rural sectors, a larger fraction of the poor depend on salaried employment for their income rather than direct production of agricultural crops (formal sectors are larger), and greater real wage rigidity prevails in labor markets. These conditions generally tend to increase the chance of negative poverty effects of devaluation.

Factor Prices and Unemployment

The impact on factor prices of a real devaluation⁹⁸ on the poor will depend on factor intensity characteristics of the tradables and nontradables sectors. The poor typically sell their labor and own few assets. Therefore it is assumed that their well-being will change with the price of labor at the highest level of aggregation. If the tradables sector is the labor-intensive sector, in the context of a two-sector Heckscher-Ohlin model the poor would be expected to benefit from devaluation. The opposite would be expected to hold if the nontradables sector were labor intensive. Finer levels of disaggregation differentiate between several types of tradable and nontradable activities, skilled labor and unskilled labor, and the degree of formality of the activities. Because the poor tend to be unskilled workers working in informal activities, factor intensities are sometimes defined in these more disaggregated sector definitions. Results are quite sensitive to the structure of the models.

The impact of devaluation on the labor market tends to be especially important in determining the poverty impact of devaluation. As highlighted in Agénor (2004), the poor often generate a significant share of their income from labor services, and, because of their precarious condition, can be particularly affected by labor market imperfections that prevent an efficient allocation of resources. In particular, lack of labor mobility across sectors and nominal wage rigidity may lead to increased unemployment and poverty when the relative price change implied by devaluation requires resource reallocation to the tradables sector. Corbacho, Garcia-Escribano, and Inchauste (2003) highlight the importance of changes in employment status as a significant source of vulnerability during the 1999-2002 Argentine crisis.

Evidence from large devaluations in financial crises episodes suggests that labor market flexibility could be important in reducing the poverty impact of the crises. Fallon and Lucas (2002) show that cross-country evidence reveals strong positive associations between depreciation of the exchange rate and the cut in real wages and between the cut in real wages relative to the decline in GDP and loss in employment. Where cuts in real wages did not materialize as a result of lack of labor market flexibility, high unemployment levels could have exacerbated the negative poverty impact of the crises. The great deal of turnover in employment that accompanied the crises, more casual wage employment, and self-employment were all deemed to have been critical in sustaining or even increasing employment.

The response of nominal wages to devaluation is a critical factor in the determination of its inflationary impact. Because wages are typically a large share of production costs, especially for nontradable goods, their response will be a key determinant of the response of prices in the economy and the behavior of real wages. The stabilization literature suggests that the higher the degree of anticipation of devaluation, the stronger the power of labor unions in wage negotiations, and that a history of high inflation tends to increase the response of nominal wages to inflation and reduce the reduction in real wages that can be achieved by a devaluation. A history of inflation may imply the presence of indexation mechanisms on wage contracts that automatically adjust wages on the basis of price changes. For a survey and more detailed discussion of the stabilization literature, see Agénor and Montiel (1999).⁹⁹

Government Revenues and Expenditures

Devaluation can also impact poor households by changing the real value of tax revenues and the real value of government expenditures. If devaluation increases the amount of net resources available to the government, then the final outcome on the poor will depend on how the additional resources are allocated (reduce revenues, increase spending, or pay off debt). If devaluation reduces the net resources available to the government instead, the poverty impact will be affected by the way the additional resource gap is financed (increasing revenue, reducing expenditure, or adding more foreign or domestic financing).

The effect of devaluation on revenues will depend on their composition and whether rates are specific or ad valorem. If revenues are dominated by trade taxes and the rates are ad valorem, revenues are likely to increase in real terms (i.e., the tax burden increases) in the short term because the rate of devaluation will exceed inflation. If instead rates are mostly specific and there are collection lags, the Olivera-Tanzi result holds and revenues will decrease in real terms. If revenues are from domestic economic activity, the impact on real revenues is likely to depend on the composition of the tax base in terms of taxed tradables and nontradables sectors. For example, after a devaluation, tradables firms’ profits are likely to increase whereas nontradables firms’ profits are likely to decrease in the short term. Therefore, profit tax collections will depend on the composition of the tax base in terms of tradables and nontradables firms.

Interest spending may increase significantly if debt is denominated in foreign currency. If debt is denominated in foreign currency, a devaluation will increase the value of debt expressed in domestic currency. Because the price-level reaction tends to lag the increase in the exchange rate, the real value of interest payments may increase significantly.

The effect of devaluation on non-interest expenditures will depend on the tradable content of the spending. The more tradable the content of spending, the more likely non-interest expenditures will go up in real terms. Burnside, Eichenbaum, and Rebelo (2003) suggest that because the rate of devaluation will be larger than inflation in the short term and that a large fraction of expenditures is nontradable (such as the wage bill in health and education), the reduction in real value of spending may be large.

Recent empirical evidence on large devaluations suggests devaluation has tended to imply a contractionary impact on the fiscal stance. Burnside, Eichenbaum, and Rebelo (2003) suggest that in recent large devaluations, fiscal policy was rather contractionary because the eroding Olivera-Tanzi effect on tax revenues of devaluation could be more than offset by the reduction in the real value of non-interest expenditures.

An increase in the tax burden may not necessarily imply that its incidence will fall on the poor. In developing economies, the poor tend to work in informal sectors that by definition are mostly outside of the tax net. A relatively small share of formal sector contributors is typically responsible for a large share of tax revenues. Under these circumstances, it is conceivable to imagine cases in which the increase in the tax burden will not significantly affect the poor.

Reduced real social spending does not necessarily mean a negative impact on the poor. Government social expenditures are on many occasions being captured by nonpoor groups and are therefore poorly targeted (Agénor, 2004). In some countries these nonpoor groups may be providing services in the sectors. For example, in the education sector, determination of teacher salaries is influenced by powerful unions and is weakly related to opportunity cost or productivity. If expenditures are reduced and the composition of spending changes toward areas of spending that are more beneficial for the poor, such as an effective safety net, there may be a positive poverty impact.

General Implications for Policy Analysis

The impact of devaluation on the poor in the short term depends mainly on the economic structure of countries and its relationship with the characteristics of the poor through the channels of impact. In particular, the empirical literature suggests that devaluations are likely to have a positive poverty impact in countries where in the context of an overvalued exchange rate the poor work mostly in rural areas, are net producers of tradable products such as the ones related to agriculture, and are not landless workers who sell their labor. In countries with overvalued exchange rates where the poor work mostly in urban areas in an informal nontradable service sector, the income of the poor depends significantly on salaried employment, and factor mobility between sectors (labor mobility in particular) is limited, a devaluation may have negative poverty effects.

Analyzing the impact of devaluation on overall inflation and the labor market is crucial to the analysis of poverty effects. Important characteristics of economies to take into account when analyzing the impact of devaluation on inflation are the domestic distribution margin for imported goods, the elasticity of demand for exports, the elasticity of substitution between tradables and nontradables, the share of tradable and nontradable goods in the household consumption basket, and the wage-setting mechanisms. In the analysis of labor market adjustment, the degree of labor market flexibility in terms of labor mobility, possibilities for more casual work employment, unionization, and the presence of indexed labor market contracts should be considered.

Even though a devaluation can imply a contractionary impact on the fiscal stance, it should not be assumed that the impact of this fiscal adjustment on the poor will necessarily be negative. For example, if part of the savings coming from the devaluation impact on government accounts were to be used to increase spending in the areas that are better targeted to the poor, such as in a well-functioning safety net, the net impact on the poor could be positive.