Policy Research Working Paper 9045 Household Impacts of Tariffs Data and Results from Agricultural Trade Protection Erhan Artuc Guido Porto Bob Rijkers Development Economics Development Research Group December 2019 Policy Research Working Paper 9045 Abstract How do trade reforms impact households in different a stylized model of the first-order impacts of import tariffs parts of the income distribution? This paper presents a on household real income, this paper quantifies the welfare new database, the Household Impacts of Tariffs data set, implications of agricultural trade protection. On average, which contains harmonized household survey and tariff unilateral elimination of agricultural tariffs would increase data for 54 low- and lower-middle income countries. The household incomes by 2.50 percentage points. Import tar- data cover highly disaggregated information on household iffs have highly heterogeneous effects across countries and budget and income shares for 53 agricultural products, within countries across households, consumers, and income wage labor income, nonfarm enterprise sales and transfers, earners; the average standard deviation of the gains from as well as spending on manufacturing and services. Using trade within a country is 1.01 percentage points. This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at eartuc@worldbank.org, guido.g.porto@gmail.com, and brijkers@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Household Impacts of Tariffs: Data and Results from Agricultural Trade Protection∗ Erhan Artuc† Guido Porto‡ Bob Rijkers§ The World Bank Dept. of Economics The World Bank DECTI UNLP DECTI ∗ We thank J. Angbazo, S. Fernandez, N. Gomez Parra, W. Kassa, H. Liu, A. Luo, M. Saleh and N. Santos Villagran for providing excellent research assistance. This research has been supported by the World Bank’s Research Support Budget, the ILO-World Bank Research Program on Job Creation and Shared Prosperity and the Knowledge for Change Program. It has also been supported by the governments of Norway, Sweden, and the United Kingdom through the Multidonor Trust Fund for Trade and Development, and by the UK Department for International Development (DFID) through the Strategic Research Partnership on Economic Development. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank of Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the countries they represent. All errors are our responsibility. † Development Economics Research Group, Trade and Integration, The World Bank. email: eartuc@worldbank.org ‡ Universidad Nacional de La Plata, Departamento de Economia, Calle 6 e/ 47 y 48, 1900 La Plata, Argentina. email: guido.porto@depeco.econo.unlp.edu.ar § Development Economics Research Group, Trade and Integration, The World Bank. email: brijkers@worldbank.org 1 Introduction The recent backlash against globalization and resurgence of protectionist tendencies have renewed interest in the distributional impacts of trade protection. To inform trade and social protection policy reform, identifying who gains and who loses from trade and quantifying by how much is of crucial policy interest. Trade reform impacts households as consumers, producers, wage earners, and, possibly, taxpayers. As a consequence, how a particular household is impacted by a trade reform depends on its income and consumption portfolios. This implies that trade reforms typically have very heterogeneous effects on household well-being. These heterogeneous impacts operating at the household level are difficult to measure with readily available data. This paper presents a new cross-country household survey data set, the Household Impacts of Tariffs (henceforth HIT) data set, that enables researchers to investigate how tariff changes impact the real incomes of households across the income distribution. The data set covers 54 developing countries, and was constructed by harmonizing representative household surveys with import tariff data from UNCTAD. The sample comprises all low-income countries for which relevant nationally representative household survey data—i.e., data with information on both households incomes and consumption spending is available—and a number of middle-income countries. In addition, we use the HIT database to assess trade policy and establish stylized facts about agricultural trade protection using a simple agricultural household model and a first-order effects approach as in Deaton (1989). In mostly agrarian economies, as the ones covered in our data set, agriculture is a major source of gains and losses from trade, especially for the poor. We find substantial gains from own agricultural tariff liberalization, amounting to 2.50 percentage points of real household income per capita across our sample of 54 countries. Because of differences in consumption and income portfolios as well as in initial tariffs, there is huge heterogeneity in the gains from trade both across countries and across households within countries. For example, the average standard deviation of the gains from trade across countries is 1.01 percentage points, but it can be as high as 2.68 percentage points. Furthermore, in 29 countries, agricultural tariff liberalization would be pro-rich in the sense 1 that the top 20% richest households would gain proportionately more than the bottom 20%. Yet, the poor would benefit more than the rich in 25 countries. We also demonstrate the importance of having very disaggregated data by showing that using more aggregated data yields biased estimates of the gains from trade. The mean absolute difference in average gains estimated using disaggregated data versus aggregated data is 0.75 percentage points, or 30% of the average gains from trade across countries. Granularity and heterogeneity are among the key features of the HIT data set, which can be downloaded from http://www.worldbank.org/en/research/brief/hit. The website also contains an online tariff reform simulation tool. The online data appendix describes in more detail how the data were harmonized, and how the tariff data can be updated. The remainder of the paper is organized as follows. Section 2 presents the data and harmonization procedures, as well as some descriptive statistics both on the structure of protection as well as households’ income and consumption portfolios. Section 3 presents a simple framework for assessing the first-order impacts of trade reform on household welfare. Section 4 presents the results from agricultural tariff cuts. A final section concludes. 2 Harmonizing Household Survey and Trade Data 2.1 Harmonizing household surveys Household surveys are the predominant instrument for analyzing poverty and income inequality and are thus a natural starting point for evaluating the distributional impacts of trade policy. In this paper we introduce a harmonized household-level information data set, which is designed to assess trade policy, that covers 54 low- and middle-income countries (see Table 1 for a list of all surveys included in the HIT data set).1 These data fill an important gap in the toolbox of policy makers and researchers, because these types of household-survey-based data are not usually readily amenable to analyzing the impacts of trade-reforms and can be hard to access. The list of countries, household surveys, year of data collection and sample sizes are reported in Table 1. 1 Household surveys are typically collected by national statistical agencies. 2 A challenge for those interested in assessing how different households are impacted by trade policy is that tariff data are typically classified using the Harmonized System, whereas household survey classifications have historically been somewhat ad hoc. To render them compatible and comparable across countries, we aggregate goods in the household surveys to common 4- and 2-digit categories using separate expenditure, autoconsumption and income templates. We cover spending on, income derived from, and autoconsumption of 53 4-digit agricultural and food items. These include Staple Agriculture, such as corn and rice, and Non-Staple Agriculture, such as oils, cotton and tobacco. We also categorize spending on five classes of manufacturing items. In addition, we keep track of spending on five (non-tradeable) services and on four other expenditures. The fact that we have much more granular data on agricultural products than on manufacturing services reflects the nature and structure of the household surveys we are standardizing. Note that not all categories are populated in all surveys, which reflects both survey design and local consumption patterns (e.g., pork not being consumed in the majority of predominantly Muslim countries). The expenditure template is shown in Figure 1. On the income side, we keep track of income derived from the sales of the same 53 food items we cover on the expenditure side. In addition, whenever the survey design allows it, we also split wage income by sector, defined roughly at the 1 digit level, and keep track of non-farm household enterprise sales across 10 sectors, as well as various types of transfers. The income template is shown in Figure 2. We also keep track of production for home consumption using the autoconsumption template, which is shown in Figure 3 and contains 53 agricultural products and a select few categories for other goods. Since many of the surveys are subject to confidentiality agreements, we aggregate the households and offer statistics for each percentile of the household per capita real income distribution. The database thus has 5,400 observations (54 countries and 100 observations per country), but is based on an underlying data set of 521,639 households which are, in turn, representative of approximately 1.8 billion people. The HIT database is best suited for country-specific analysis. In order to facilitate cross-country comparisons, we converted incomes to their constant 2010 USD equivalent by setting the survey mean of real expenditure 3 equal to the 2010 GDP per capita from the World Development Indicators. It should be noted that this is only an approximation to more proper international comparisons (see e.g. Deaton and Dupriez, 2011) and that the HIT database is not the World Bank Group’s official poverty data, which can be found in PovcalNet.2 Figure 4 pools all the data and shows how the aggregate spending categories identified in our data vary with (the log of) household income per capita (in constant 2010 USD). To start, households spend a large share of their income on food and agricultural products. Across all countries, the average household in our sample spends 44.7% of its income on food items, 17.4% on manufacturing goods, and 15.1% on services. Another 16.9% of expenditure is accounted for by goods households have produced themselves, which highlights the importance of dealing with home consumption in the analysis. This also implies that the total expenditure share in agriculture is 61.6%. While there is huge heterogeneity in spending patterns both across and within countries, the graph shows that, as households get richer, the share of income spent on food decreases, especially for the richest households. Spending on manufacturing goods, services, and other goods first declines with income but then increases sharply. The opposite happens with home consumption. The implication is that the tariff burden on different households will vary as a result of these different consumption patterns. Agriculture is the most important source of income, accounting for 38.5% of total household income on average. This total comprises a 20.9% share of production for household consumption and 17.6% of sales of agricultural products in the market. Labor income represents 29.6% of household income, on average, and non-farm enterprises account for another 12.7%. Figure 5 shows how these aggregate income shares vary by income (in constant 2010 USD), demonstrating that for the very poorest households in our sample, agricultural income and home production tend to be the most important sources of earnings. As households get richer, wage income becomes a progressively more important source of 2 Note also that the HIT consumption aggregates may differ from those in PovcalNet because the methodology used to calculate aggregate consumption differs from that of PovcalNet. To give a few examples, the HIT consumption aggregate includes expenditures on durable goods, while PovcalNet aggregates typically aim to capture the rental value of durables. As another example, the HIT data include all health expenditures, whereas health expenditures are not uniformly treated in PovcalNet. Note also that we are scaling up average expenditure per capita to match the national accounts estimates. Specifically, we set mean expenditure per capita equal to GDP per capita in constant 2010 US dollars. 4 income on average. As a consequence, on the income side too, tariff incidence will vary across households. Having disaggregated and nationally representative household-level data with information on both consumption and income portfolios is a major advantage since it enables policymakers and researchers to quantify the significant heterogeneity in the impacts of trade reforms across households. At the same time, it is worth bearing in mind some of the limitations of household surveys. For instance, they typically fail to capture households at the very top of the household income distribution. In addition, they do not adequately capture capital incomes. Moreover, they suffer from measurement error, especially in incomes, which are often underreported. To minimize the role of measurement error, we dropped households in the top and bottom 0.5% of the expenditure distribution prior to aggregating. 2.2 Harmonizing tariff data The next step in the analysis is to convert tariffs at the HS6 level to the standardized product classifications from the household surveys. Each group i from the household surveys contains many finer product groups from the HS classification. To arrive at a product level average, we computed weighted average tariff rates τi for each of the groups in the survey classification. mc,n (1) τi = τc,n , c,n∈i c,n∈i mc,n where n is an HS-category that belongs to survey-category i and m(c,n) are imports of good n from country c. To calculate (1), we use tariffs from the latest year for which data are available. Tariffs vary both across countries and across products. The average tariff across countries is 14.2%. Tariffs are highest on average in Bhutan, notably 48.4% on average, and lowest in Iraq (5% on average), and tend to be lower in countries with higher levels of GDP per capita. There is also significant variation in tariffs across the different products in our data. For agriculture, the focus of our applications, Figure 6 depicts both the mean and the maximum tariff for each of the 53 food items in our data. On average, the highest tariff is 39.4%, but 5 this masks considerable heterogeneity across countries: Sri Lanka levies a 125% tariff on cigarettes, while in Jordan the tariff on beer is 200%.3 3 Tariffs and Household Welfare To derive the welfare effects of agricultural tariffs at the household level, we adopt the standard framework of Deaton (1989). In this setting, the indirect utility of household h is: (2) V h (p, y h ) = V p, xh 0 + h πj (p) , j h where p is a vector of prices, πj is the profit derived from agricultural activity j , and xh 0 comprises fixed sources of income (such as gifts, remittances, transfers, and so on). Thus, household income y h is the sum of fixed income and agricultural income and is assumed to be equal to household expenditures.4 This is the simplest possible setting to study the welfare effects of tariffs and we adopt it here in part as homage to Deaton’s (1989) original formulation to study rice export taxes in Thailand, but mostly because an extended agricultural household model is appropriate given that we will be focusing on agricultural tariffs. Furthermore, this setting does not require strong structural assumptions. Adding more structure, the literature has extended this framework to allow for effects on wages, transfers, and non-traded consumer prices as well as non-traded family enterprise income. See Porto (2005, 2006), Nicita, Olarreaga and Porto (2014), Atkin, Faber and Gonzalez-Navarro (2018) and Artuc, Porto and Rjkers (2019). The welfare effects can be calculated with a first-order approximation. Differentiation of (2) with respect to the price of good i, pi yields: dVih ∂V h (3) h = h φh h i − si d ln pi , y ∂ ln y where φh h i is the income share derived from the sales of good i and si is the share of good i in 3 Note that we ignore tariffs on alcoholic products in the Arab Republic of Egypt, as they are clear outliers given that they are 1200% or higher. 4 This rules out saving, debt and dynamic considerations. 6 the consumption bundle of household h. Following Deaton (1989, 1997), we can safely ignore the private marginal utility of income ∂V h /∂ ln y h in policy evaluation. In addition, letting τi be the instrument of tariff protection for sector i and assuming perfect price pass-through elasticities,5 so that d ln pi = −τi /(1 + τi ), the estimable welfare effects are given by: dVih τi (4) h = − φh h i − si . y 1 + τi The interpretation of this equation is straightforward. After a price change caused by tariff cuts d ln pi = −τi /(1 + τi ), the first-order effects on real income can be well-approximated with the corresponding income and expenditure shares. In the language of Deaton (1989), because we are working with tariff cuts and price declines, net-consumers benefit while net-producers suffer. Though this is well-known, it is important to clarify that the Deaton first-order approximation captures direct, short-term effects of tariff liberalization. In particular, it does not take into consideration second-order effects such as consumption and production adjustments (which may become progressively more important over time), labor or investment decisions, or any dynamic effect more generally. The goal is to calculate the welfare effects generated by the entire structure of tariff protection. To obtain a measure, we sum the changes in welfare in (4) over all traded goods i to get: dV h dVih (5) Vh = = , yh i yh where V h is the proportional change in household real income. Next, we use this expression to assess trade policy with the HIT database. 5 Note that we are making the assumption of perfect pass-through for convenience and ease of exposition. The data in principle accommodate richer and more realistic pass-through assumptions, see e.g. Nicita (2009), Ural-Marchand (2012), and Bergquist (2017) for pass-through estimates for selected commodities in selected countries. 7 4 HIT Data in Action: Agricultural Tariff Liberalization In order to illustrate the use of the database, we explore the welfare and distributional impacts of agricultural tariffs. To do this, we use our simple framework and simulate what would happen to average incomes, and their distribution, if a country were to eliminate its own agricultural import tariffs. This amounts to setting all agricultural tariffs to zero. Note that we work with unilateral tariff cuts and thus we run independent simulations for each of the 54 countries separately.6 The outcome of this exercise comprises a set of results on the welfare effects for households at different levels of well-being for each country. As a first step in the analysis, these effects will be aggregated to study issues related to the gains from trade (Arkolakis, Costinot and Rodriguez-Clare, 2012; Costinot, Donaldson and Komunjer, 2012; Costinot and Rodriguez-Clare, 2014; Artuc, Lederman and Porto, 2015; Caliendo and Parro, 2015; Melitz and Redding, 2015; Arkolakis, Costinot, Donaldson and Rodriguez-Clare, 2019; Caliendo, Dvornik and Parro, 2018). The results are presented in Table 2 for all 54 countries in our database. The gains from eliminating tariffs amount to 2.50% of real household income on average across countries (column 1). This means that, from a global standpoint, developing countries would win from a unilateral agricultural tariff liberalization.7 There are two broad mechanisms at play. On the one hand, lower food prices create consumption gains, which amount to a real income gain of 4.64 percentage points on average (column 2). On the other hand, the loss of protection implied by the tariff cuts create income losses, which amount to 2.15 percentage points on average (column 3). Consequently, the overall gains from trade are driven by lower prices and consumption gains that outweigh the income losses from tariff de-protection. A major distinctive feature of our data is that they allow analysis of heterogeneity in the welfare effects of trade. Heterogeneity takes several forms. There is heterogeneity across 6 Although a multilateral liberalization scenario would be feasible, this requires different modeling assumptions about the functioning of global markets. 7 Note that we ignore impacts on global prices, which we assume to be exogenous given that the countries we are considering only account for a very small share of global trade. 8 countries, heterogeneity across households, heterogeneity across goods, and heterogeneity across sources of gains (namely, consumption and income effects). We illustrate these varying forms of heterogeneity by slicing the results from our simulations in different ways. To begin, Table 2 shows that the aggregate gains are highly heterogeneous across countries. While the gains can exceed 5% of real household income, as in Zambia (6.93%), Bhutan (6.53%), Jordan (6.55%), Cameroon (6.25%), and Ecuador (5.05%), there are two countries with income losses, namely Burundi (-3.23%) and Ghana (-0.50%) and a few other countries with small gains (e.g., Cambodia, 0.17%). Figure 7 clearly illustrates the cross-country heterogeneity with a map of the relative magnitudes of the gains from trade. We can also use these data to show an interesting result: there is a positive correlation between the aggregate gains from trade and the log of per capita GDP (Figure 8). Roughly speaking, this suggests that richer countries stand to gain more from unilateral agricultural liberalization than the poorest ones. Second, we examine household heterogeneity and explore distributional impacts, which is one of the main advantages of the Household Impact of Tariffs database. There is a burgeoning literature on this topic, including Porto (2006) and more recent contributions from Fajgelbaum and Khandelwal (2015), Atkin, Faber and Gonzalez-Navarro (2018), Faber (2014) and Atkin and Donaldson (2015), Antras, de Gortari and Itskhoki (2018), and Galle, Rodriguez-Clare, and Yi (2017), and Artuc, Porto and Rijkers (2019). We start by plotting the developing world distribution of the gains from trade to show how HIT data can be used to analyze global implications of trade with a specific focus on farm households.8 The variability in the gains from trade across countries is sizable; the average standard deviation of the income gains across the income distribution is 1.01% across countries. In Ukraine, the country with the lowest variance in the gains from trade, the standard deviation is 0.19 percentage points while in Burundi, the country with the highest variability, the standard deviation is 2.68 percentage points. It is interesting to note that the variability 8 Artuc, Porto and Rijkers (2019) investigate the trade-off between the income gains and the inequality costs of liberalization for countries separately using raw data. Here, we show how related policy problems can be tackled at the global level using conveniently organized HIT data. 9 in the gains from trade across households is negatively correlated with the log of per capita GDP (Figure 9). This is because, in the HIT data, poorer countries tend to have more heterogeneous income and consumption household portfolios.9 Pooling all countries in our database, we plot the household-level gains from agricultural trade against the initial level of per capita household expenditure using a kernel non-parametric regression. Figure 10 shows the result of this exercise. In line with the stylized facts above, we find that the kernel slope is positive and steeps upward until the top percentiles of the developing world distribution of income and then becomes negative. Still, richer households tend to gain more from liberalization than poorer households. But it is the upper-middle class that stands to gain the most from agricultural tariff liberalization. a-vis the rich, To quantify the extent to which the effects of trade vary for the poor vis-` we calculate the pro-poor bias in agricultural trade policy. This is the difference between the average welfare effect for the poor (defined as the bottom 20% of the income distribution) and the average welfare effect for the rich (the top 20%).10 On aggregate, agricultural tariff liberalization would be slightly pro-rich, with the richest households gaining 2.64 percent and the poor 2.20 percent (see columns 4-6 in Table 2). There are 29 of the 54 countries in which tariff liberalization would have a pro-rich bias. Using data from six countries which are also included in HIT data set, Nicita, Olarreaga and Porto (2014) show that tariff protection is pro-poor. In this study, however, we find that their results do not generalize to all developing countries. In 24 countries in the HIT data set tariff protection is pro-rich (and liberalization, pro-poor). One simple way to assess how much we can gain by exploiting the heterogeneity in the household surveys is to compute the gains from trade using aggregate price changes and aggregate income and budget shares. To do this, we aggregate the staple agriculture income and budget shares from the 4-digit classification up to staple and non-staple agriculture. We then compute the welfare gains from the elimination of aggregate tariffs, as before. The 9 Note that the HIT data have limited heterogeneity in expenditures on manufacturing and services. This may also play a role in our results. 10 In the online Appendix we show that our results are very robust to using alternative cutoffs. In addition, we show that both average gains and the pro-poor bias are correlated with GDP per capita. The pro-poor bias is also larger in countries with higher levels of poverty. 10 results are in Table 3. There is indeed a bias when calculating the welfare effects with aggregated data. The size and direction of the aggregation bias calculated by comparing measures derived from aggregated data (presented in Table 3) instead of disaggregated data (presented in Table 2), are summarized in Table 4. At the developing world level, for instance, the welfare gains would be 2.39%, instead of the 2.50% obtained when we use disaggregated data; using aggregate data thus leads to an underestimation of the average gains across countries by approximately 4 percentage points. For individual countries the biases are much larger and can be positive (in 25 countries) or negative (29 countries). The average absolute difference between average gains estimated using aggregated vs disaggregated data is 0.75 percentage points, or approximately 30 percent of the average gains from trade across countries. In the Central African Republic, Pakistan, Indonesia, or the Republic of Yemen, the welfare gains using aggregate data are roughly half of the estimates using disaggregated data. More extreme differences appear when we inspect countries with positive biases. In Cambodia and Ethiopia, the welfare gains are overestimated by a factor of 9 or 4, respectively. There are many instances where the estimates are biased by a factor of close to or above 2 (Burkina Faso, Cambodia, Ethiopia, Guinea-Bissau, Madagascar, Niger). Using more aggregated data also gives different estimates of the distributional impacts of the gains from trade. Using the aggregated data trade policy is estimated to be pro-rich in 29 of the 54 countries, whereas it is estimated to be pro-rich in 23 of the 54 countries when using disaggregated data. More aggregated data tend to underestimate the gains accruing to rich households slightly more than they underestimate the gains accruing to poor households. We now turn to examining product heterogeneity in the welfare effects. This is the heterogeneity that arises when we decompose the gains from trade into the consumption and income effects and, fundamentally, into disaggregated effects across goods. Since our templates build up from granular 4-digit categories of goods, we can exploit this feature of the data to showcase this heterogeneity. To do this, we select one country, Vietnam, and report the aggregate effects, the consumption effects and the income effects of each one of the 53 4-digit goods in the template (see Figures 1 and 2). Results are in Table 5. 11 The aggregate gains from trade in Vietnam are 1.99 percent. These gains are mostly explained by gains from lower tariffs on Other Processed Food (0.92 percent), Bananas (0.88 percent) and Cigarettes (0.61 percent). These are mostly consumption gains (with minor or even absent income gains). Fish contributes 0.37 percentage points to the aggregate gains, with a consumption gain of 0.50 percent and a lower income loss of 0.13 percent. The case of Rice is interesting because lower tariffs (and lower prices) create large income losses of –1.57 percent. Other important products are Corn (with income losses of –0.13 percent), Alcohol (with consumption gains of 0.17 percent), Pork (0.28 percent) and Tea (0.13 percent). The pro-poor index in Vietnam is –1.70. This anti-poor bias of agricultural liberalization is due to the fact that the gains for the rich (2.74 percent) are higher than the gains for the poor (1.04 percent). Tariff liberalization in products such as Bananas and Cigarettes yields the largest pro-poor biases, but the tariff cuts in Corn, Other Processed Food and Rice are distinctly anti-poor. Note in particular the role of the rice income effect: because of lower prices and because the poor are major producers and sellers of rice, the net rice income loss for the poor is –2.35 percent, much higher than the net rice income loss for the rich (of only –0.58 percent). To end, we report in Table 6 the results from scenarios where tariffs are eliminated only in cereals (namely, corn, wheat, rice and other cereals—sorghum, barley—in the templates). This scenario is interesting because it combines a simulation for a set of related goods that accounts for more than 50% of global calorie intake,11 across all countries.12 Tariff liberalization in cereals would bring aggregate developing world gains of 0.42 percent, but there is a lot of variation in the gains from trade not only among the winners but also among the losers. For example, the gains from cereals liberalization can be as high as 1.92 percent in Guinea Bissau or 1.83 percent in Bhutan to as low as being almost negligible in Armenia, Georgia, and South Africa. There are also countries that would incur large losses, such as Vietnam and the Central African Republic (–1.70 and –0.99 percent, respectively), as well as countries in which losses are almost negligible (e.g., Ghana, the Arab Republic of Egypt, 11 See World Health Organization (2003). 12 Note that we are running country-by-country unilateral tariff reductions rather than multilateral liberalization scenarios. In particular, we ignore spillovers of tariff reductions in one country on another country, because the low-income countries in our sample only account for a small share of world trade. 12 Malawi). This heterogeneity reflects the fact that some countries are more agrarian than others and that some are large net-consumers, others are large net-producers, and yet others either consume and produce little cereals or do so in similar magnitudes so that the net effects tend to cancel out. A similar story can be told about the pro-poor bias of liberalization. There are countries where cereals liberalization would be clearly pro-poor (Tanzania, Bhutan, Guinea Bissau, Ecuador) and countries where it would be clearly pro-rich (Vietnam). There are also cases where the liberalization would be neutral and this can in turn be because of little direct consumption and production of grains (Armenia, Georgia) or because of the offsetting consumption and income effects (Madagascar, Nigeria). As this analysis illustrates, the HIT data and our framework enable researchers to examine and exploit the extensive household heterogeneity in incomes and expenditures in many dimensions. There are some limitations to the framework that are important to note for accurate interpretation of our results. To make it operational, we need to impose some structure, in particular perfect competition, constant returns to scale, and product homogeneity. Since we work with household surveys and with an agricultural household a la Deaton, there are some limitations in the scope of the analysis as well. The model ` household surveys do not collect reliable information on the returns to capital and on profits and often fail to capture very rich households. Also, there are marked differences between consumption and production aggregates from the household surveys and those from the national accounts. As a result, there might be some discrepancies between the household welfare effects from the HIT data and the aggregate welfare effects stemming from more general trade models. In addition, some relevant impacts, such as heterogeneous varieties, two-way trade or labor and investment effects, may require additional modeling assumptions as well as additional data (such as demand and supply elasticities and so on). In any case, most of the discussion about poverty, inequality and household welfare is typically based on household surveys and these surveys and the HIT data set are thus a natural starting point for analyzing the distributional impacts of trade policy. 13 5 Conclusion Quantifying who benefits and who loses from trade reform and by how much is of crucial policy interest, but often challenging because of a lack of suitable data. The Household Impacts of Tariffs (HIT) data introduced in this paper are a publicly available household survey based data set covering 54 developing countries that enables researchers to analyze the distributional impacts of tariffs. It contains granular data for each percentile of the income distribution on both the income derived from and consumption of 53 agricultural products. In addition, it keeps track of spending on five different types of manufacturing goods and services, as well as transfers, and wage income disaggregated by 1-digit sector, 10 different types of non-farm household enterprise sales and various types of transfers. Using a stylized agricultural model and a first-order effects approach we have illustrated potential applications of the data and shown that the prevailing structure of agricultural tariffs represses household incomes by 2.50% percentage points across countries. However, the costs of protectionism vary enormously across countries and across households within countries, because households in different parts of the income distribution tend to have very different income and consumption portfolios; the average standard deviation of the gains from trade within a country is 1.01 percentage points. We also show that using disaggregated data is important, because using more aggregated data yields biased estimates of the average gains from trade. While we have focused on the elimination of agricultural tariffs, the HIT data set has a much wider set of potential applications and can accommodate richer and more sophisticated modeling assumptions. Examples of potential applications include assessing how EU and U.S. agricultural tariffs or regional trade-agreements, such as AGOA, impact households in low-income countries.13 An analysis of poverty and inequality impacts of food price shocks is also possible. Moreover, the data can be used to study issues that are not related to trade reform, such as food subsidies or value-added tax reforms. 13 In the online Appendix we analyze the impact of non-tariff barriers. 14 References Antras, P., A. de Gortari, and O. Itskhoki (2017). “Globalization, Inequality, and Welfare,” Journal of International Economics, 108, pp. 387-412. Arkolakis, C., A. Costinot, and A. Rodriguez-Clare (2012). “New Trade Models, Same Old Gains?,” The American Economic Review, vol. 102(1), pp. 94–130. Arkolakis, C., A. Costinot, D. Donaldson, and A. Rodriguez-Clare (2019). “The Elusive Pro-Competitive Effects of Trade,” Review of Economic Studies, 86, 46–80 Artuc, E., D. Lederman, and G. Porto (2015). “A Mapping of Labor Mobility Costs in the Developing World,” Journal of International Economics, vol. 95(1), pp. 28–41. Artuc, E., G. Porto, and B. Rjkers (2019). “Trading Off the Income Gains and the Inequality Costs of Trade Policy,” Journal of International Economics, forthcoming. Atkin, D. and D. Donaldson (2015). “Who’s Getting Globalized? The Size and Implications of Intranational Trade Costs,” mimeo. Atkin, D., B. Faber and M. Gonzalez-Navarro (2018). “Retail Globalization and Household Welfare: Evidence from Mexico,” Journal of Political Economy, 126(1): pp.1–73. Benjamin, D. and A. Deaton (1993). “Household Welfare and the Pricing of Cocoa and Coffee ote d’Ivoire: Lessons from the Living Standards Surveys,” The World Bank Economic in Cˆ Review vol. 7, pp. 293–318. Bergquist, L. (2018). “Pass-Through, Competition, and Entry in Agricultural Markets: Experimental Evidence from Kenya” mimeo. Caliendo, L. and F. Parro (2015). “Estimates of the Trade and Welfare Effects of NAFTA,” The Review of Economic Studies, vol. 82(1), pp. 1–44. Caliendo, L., M. Dvorkin, and F. Parro (2015). “Trade and Labor Market Dynamics,” NBER Working Paper No. 21149. 15 Costinot, A., D. Donaldson, and I. Komunjer (2012). “What Goods Do Countries Trade? A Quantitative Exploration of Ricardo’s Ideas,” The Review of Economic Studies, vol. 79(2), pp. 581–608. Costinot, A., A. Rodriguez-Clare (2014). “Trade Theory with Numbers: Quantifying the Consequences of Globalization,” Handbook of International Economics, Edited by G. Gopinath, E. Helpman and K. Rogoff, Chapter 4, pp. 197–262. Deaton, A. (1989). “Rice Prices and Income Distribution in Thailand: a Non-parametric Analysis,” Economic Journal, 99 (Supplement), pp. 1–37. Deaton, A. (1997). The Analysis of Household Surveys - A Microeconometric Approach to Development Policy. Baltimore: Johns Hopkins Press. Deaton, A. and O. Dupriez (2011). “Purchasing Power Parity Exchange Rates for the Global Poor” American Economic Journal: Applied Economics, 3 (2):137-66. Faber, B. (2014). “Trade Liberalization, the Price of Quality, and Inequality: Evidence from Mexican Store Prices,” mimeo, Department of Economics, University of California at Berkeley. Fajgelbaum, P. and A. Khandelwal (2016). “Measuring the Unequal Gains from Trade,” Quarterly Journal of Economics, vol. 131, pp. 1113–1180. Galle, S., A. Rodriguez-Clare, and M. Yi (2017). “Slicing the Pie: Quantifying the Aggregate and Distributional Effects of Trade,” mimeo Berkeley University. Nicita, A. (2009). “The Price Effect of Trade Liberalization: Measuring the Impacts on Household welfare,” Journal of Development Economics, vol. 89(1), pp. 19–27. Nicita, A., M. Olarreaga, and G. Porto (2014). “Pro-Poor Trade Policy in Sub-Saharan Africa,” Journal of International Economics, Vol. 92(2), pp. 252–265. Melitz, M. and S. Redding (2015). “New Trade Models, New Welfare Implications,” American Economic Review vol. 105(3), pp 1105–1146. 16 Porto, G. (2005). “Informal Export Barriers and Poverty,” Journal of International Economics Vol. 66, pp. 447–470. Porto, G. (2006). “Using Survey Data to Assess the Distributional Effects of Trade Policy,” Journal of International Economics 70, pp. 140–160. Ural Marchand, B. (2012). “Tariff Pass-Through and the Distributional Effects of Trade Liberalization,” Journal of Development Economics, 99(2), pp. 265–281. World Health Organization (2003). “Diet, nutrition and the prevention of chronic diseases,” Technical Report Series, Geneva, Vol. 916, pp 1–150. 17 Figure 1 Expenditure Template Expenditure 1. Agriculture/Food 11. Staple Food 111. Cerals 112. Legumens 113. Fruits 114. Vegetables 115. Oils/Fats 116. Fish 117. Meat/Livestock 118. Dairy/Eggs 119. Other staple food 1111. Corn 1121. Beans 1131. Banana 1141. Tomato 1151. Vegetable Oils 1161. Fish 1171. Pork (Pig) 1181. Milk 1191. Other staple food 1112. Wheat 1122. Other 1132. Grapes 1142. Potato 1152. Animal Fats 1162. Shrimp 1172. Beef (Cattle) 1182. Eggs 1192. Other processed food 1113. Rice 1133. Citrus 1143. Greens 1153. Other oils/fats 1163. Other Crustacean 1173. Poultry (Chicken) 1183. Cheese 1114. Other Cereals 1134. Apples 1144. Other 1174. Other meat/animals 1184. Other Dairy 1135. Other Fruits Vegetables 12. Non Staple 121. Alcohol 122. Tobacco 123. Oil seeds 124. Spices/herbs 125. Coffee/tea/cocoa 126. Nuts 127. Cotton 128. Other non-staple food 1211. Wine 1221. Cigarettes 1231. Soya 1241. Cloves 1251. Coffee 1261. Cashew 127. Cotton 1281. Sugar (any kind) 1212. Beer 1222. Other tobacco 1232. Other oil seeds 1242. Pepper 1252. Tea 1262. Coconut 1282. Other non-staple 1213. Other alcohol 1243. Vanilla 1253. Cocoa 1263. Other nuts 1244. Saffron 1245. Qat (chat) 1246. Other spices 2. Manufacturing/Household Items 21. Energy 22. Textiles/Apparel 23. Electric/Electronics 24. Household items/Furniture 25. Other physical goods 3. Services 31. Transportation 32. Health 33. Education 34. Communication 35. Other Services 4. Other Expenditures 41. Remittances/transfers given 42. Investment of any sort 43. Festivities 44. Other Disbursement Notes: Template use to harmonized household expenditures. Own elaboration. 18 Figure 2 Income Template Income 1. Agriculture/Food 11. Staple Food 111. Cerals 112. Legumens 113. Fruits 114. Vegetables 115. Oils/Fats 116. Fish 117. Meat/Livestock 118. Dairy/Eggs 119. Other staple food 1111. Corn 1121. Beans 1131. Banana 1141. Tomato 1151. Vegetable Oils 1161. Fish 1171. Pork (Pig) 1181. Milk 1191. Other staple food 1112. Wheat 1122. Other 1132. Grapes 1142. Potato 1152. Animal Fats 1162. Shrimp 1172. Beef (Cattle) 1182. Eggs 1192. Other processed food 1113. Rice 1133. Citrus 1143. Greens 1153. Other oils/fats 1163. Other Crustacean 1173. Poultry (Chicken) 1183. Cheese 1114. Other Cereals 1134. Apples 1144. Other 1174. Other meat/animals 1184. Other Dairy 1135. Other Fruits Vegetables 12. Non Staple 121. Alcohol 122. Tobacco 123. Oil seeds 124. Spices/herbs 125. Coffee/tea/cocoa 126. Nuts 127. Cotton 128. Other non-staple food 1211. Wine 1221. Cigarettes 1231. Soya 1241. Cloves 1251. Coffee 1261. Cashew 127. Cotton 1281. Sugar (any kind) 1212. Beer 1222. Other tobacco 1232. Other oil seeds 1242. Pepper 1252. Tea 1262. Coconut 1282. Other non-staple 1213. Other alcohol 1243. Vanilla 1253. Cocoa 1263. Other nuts 1244. Saffron 1245. Qat (chat) 1246. Other spices 2. Wages 20. Agriculture, forestry, and fishing 21. Mining, oil, and gas extraction 22. Manufacturing 23. Construction 24. Transportation, communications, electric, gas, and sanitary services 25. Wholesale and retail trade 26. Finance, insurance, and real estate 27. Entertainment Services (Restaurant, entertainment, hotels, etc.) 28. Professional Services (Education, health, other professional occupations) 29. Public Administration 3. Sales of Goods/Services 31. Mining, oil, and gas extraction 32. Manufacturing 33. Construction 34. Transportation, communications, electric, gas, and sanitary services 35. Wholesale and retail trade 36. Finance, insurance, and real estate 37. Entertainment Services (Restaurant, entertainment, hotels, etc.) 38. Professional Services (Education, health, other professional occupations) 39. Public Administration 4. Transfers 41. Remittances/transfers received (friend, relative) 42. Profits of investment (rent, interests) 43. Government transfers 44. Non-governmental transfers 45. Other Notes: Template use to harmonized household incomes. Own elaboration. Figure 3 Auto-Consumption Template Autoconsumption 1. Agriculture/Food 11. Staple Food 111. Cereals 112. Legumens 113. Fruits 114. Vegetables 115. Oils/Fats 116. Fish 117. Meat/Livestock 118. Dairy/Eggs 119. Other staple food 1111. Corn 1121. Beans 1131. Banana 1141. Tomato 1151. Vegetable Oils 1161. Fish 1171. Pork (Pig) 1181. Milk 1191. Other staple food 1112. Wheat 1122. Other 1132. Grapes 1142. Potato 1152. Animal Fats 1162. Shrimp 1172. Beef (Cattle) 1182. Eggs 1192. Other processed food 1113. Rice 1133. Citrus 1143. Greens 1153. Other oils/fats 1163. Other Crustacean 1173. Poultry (Chicken) 1183. Cheese 1114. Other Cereals 1134. Apples 1144. Other 1174. Other meat/animals 1184. Other Dairy 1135. Other Fruits Vegetables 12. Non Staple 121. Alcohol 122. Tobacco 123. Oil seeds 124. Spices/herbs 125. Coffee/tea/cocoa 126. Nuts 127. Cotton 128. Other non-staple food 1211. Wine 1221. Cigarettes 1231. Soya 1241. Cloves 1251. Coffee 1261. Cashew 127. Cotton 1281. Sugar (any kind) 1212. Beer 1222. Other tobacco 1232. Other oil seeds 1242. Pepper 1252. Tea 1262. Coconut 1282. Other non-staple 1213. Other alcohol 1243. Vanilla 1253. Cocoa 1263. Other nuts 1244. Saffron 1245. Qat (chat) 1246. Other spices 2. Other goods 21. Energy (wood, coal) 22. Gathering (forest, mushrooms, berries, etc.) 23. Other goods collected for free 24. Other goods produced and consumed within the household Notes: Template use to harmonized household home production. Own elaboration. 19 Figure 4 Expenditure Shares Across the Income Distribution .5 .4 expenditure share .2 .1 0 .3 2 4 6 8 10 12 log per capita expenditure Agriculture and Food Manufacturing Services Other Home consumption Notes: non-parametric kernel regressions of budget shares on the log of per capita household expenditure across the 54 countries included in the HIT database. 20 Figure 5 Income Shares Across the Income Distribution .8 .6 Income share .4.2 0 2 4 6 8 10 12 log per capita expenditure Agricultural sales Wages Non-farm enterprise Other/transfers Home production Notes: non-parametric kernel regressions of income shares on the log of per capita household expenditure across the 54 countries included in the HIT database. 21 tariff rate 0 50 100 150 200 Bhutan Cameroon Kenya Rwanda Uganda Tanzania Ethiopia Jordan Burundi Zambia Sri Lanka Malawi Vietnam Ghana Gambia, The Sierra Leone Ecuador Burkina Faso Bangladesh Pakistan Benin Cote d'Ivoire Nigeria Mali average tariff Mozambique 22 Central African Republic Uzbekistan Notes: Average tariffs and tariff dispersion in Agriculture. Figure 6 Madagascar Niger Togo Guinea-Bissau Papua New Guinea Cambodia Guinea Mauritania Nicaragua Bolivia Tariff Dispersion in Agriculture Guatemala Yemen, Rep. Nepal Azerbaijan Egypt, Arab Rep. Liberia Comoros Tajikistan maximum tariff Moldova Armenia South Africa Mongolia Indonesia Georgia Ukraine Kyrgyz Republic Iraq Figure 7 The Developing World Distribution of the Gains from Trade (6,7] (5,6] (4,5] (3,4] (2,3] (1,2] (0,1] (-1,0] 23 (-2,-1] (-3,-2] [-4,-3] Notes: The developing world distribution of the aggregate gains from the elimination of own agricultural import tariffs. Figure 8 The Gains from Trade and per capita GDP 10 5 gains 0 -5 5 6 7 8 9 log per capita GDP Notes: scatter plot and linear fit of the aggregate gains from agricultural liberalization and the log of per capita GDP. 24 Figure 9 Variability in the Gains from Trade and per capita GDP 3 standard deviation of gains 1 0 2 5 6 7 8 9 log per capita GDP Notes: scatter plot and linear fit of the standard deviation in the gains from agricultural liberalization and the log of per capita GDP. 25 Figure 10 The Gains from Trade and Household Income 10 5 gains 0 -5 -10 2 4 6 8 10 12 log per capita expenditure Notes: non-parametric kernel regression of the household-level gains from trade and the log of per capita household expenditure. 26 Table 1 Household Surveys Country Year Obs Survey Benin 2003 5296 Questionnaire Unifi´ ˆ e sur les Indicateurs de Base du Bien-Etre Burkina Faso 2003 8413 Enquˆ ete sur les Conditions de Vie des M´ enages Burundi 1998 6585 Enquˆ ete Prioritaire, Etude Nationale sur les Conditions de Vie des Populations Cameroon 2001-2002 10881 Deuxi` eme Enquˆ ete Camerounaise Aupr` es des M´ enages Central African Republic 2008 6828 Enquˆ ete Centrafricaine pour le Suivi-Evaluation du Bien-ˆ etre Comoros 2004 2929 Enquˆ egrale aupr` ete Int´ es des M´enages Cˆote d’Ivoire 2008 12471 Enquˆ ete sur le Niveau de Vie des M´ enages Egypt, Arab Republic 2008-2009 23193 Household Income, Expenditure and Consumption Survey Ethiopia 1999-2000 16505 Household Income, Consumption and Expenditure Survey Gambia, The 1998 1952 Household Poverty Survey Ghana 2005-2006 8599 Living Standards Survey V Guinea 2012 7423 Enquˆ ete L´ ere pour l’Evaluation de la Pauvret´ eg` e Guinea Bissau 2010 3141 Inquerito Ligeiro para a Avalic˜ ao da Pobreza Kenya 2005 13026 Integrated Household Budget Survey Liberia 2014-2015 4063 Household Income and Expenditure Survey Madagascar 2005 11661 Permanent Survey of Households Malawi 2004-2005 11167 Second Integrated Household Survey Mali 2006 4449 Enquˆ ete L´eg` ere Int´ ee aupr` egr´ es des M´enages Mauritania 2004 9272 Enquˆ ete Permanente sur les Conditions de Vie des M´ enages Mozambique 2008-2009 10696 Inqu´erito sobre Or¸ camento Familiar Niger 2005 6621 Enquˆ ete Nationale sur les Conditions de Vie des M´ enages Nigeria 2003-2004 18603 Living Standards Survey Rwanda 1998 6355 Integrated Household Living Conditions Survey Sierra Leone 2011 6692 Integrated Household Survey South Africa 2000 25491 General Household Survey Tanzania 2008 3232 Household Budget Survey Togo 2011 5464 Questionnaire des Indicateurs de Base du Bien-ˆ etre Uganda 2005-2006 7350 National Household Survey Zambia 2004 7563 Living Conditions Monitoring Survey IV Notes: List of household surveys included in the HIT databse. 27 Table 1 (cont.) Household Surveys Country Year Obs Survey Armenia 2014 5124 Integrated Living Conditions Survey Bangladesh 2010 12117 Household Income and Expenditure Survey Bhutan 2012 8879 Living Standards Survey Cambodia 2013 3801 Socio-Economic Survey Indonesia 2007 12876 Indonesian Family Life Survey Iraq 2012 24895 Household Socio-Economic Survey Jordan 2010 11110 Household Expenditure and Income Survey Kyrgyz Republic 2012 4962 Intergrated Sample Household Budget and Labor Survey Mongolia 2011 11089 Household Socio-Economic Survey Nepal 2010-2011 5929 Living Standards Survey Pakistan 2010-2011 16178 Social and Living Standards Measurement Survey Papua New Guinea 2009 3776 Household Income and Expenditure Survey Sri Lanka 2012-2013 20335 Household Income and Expenditure Survey Tajikistan 2009 1488 Tajikistan Panel Survey Uzbekistan 2003 9419 Household Budget Survey Vietnam 2012 9306 Household Living Standard Survey Yemen, Republic of 2005-2006 12998 Household Budget Survey Azerbaijan 2005 4797 Household Budget Survey Georgia 2014 10959 Household Integrated Survey Moldova 2014 4836 Household Budget Survey Ukraine 2012 10394 Sampling Survey of the Conditions of Life of Ukraine’s Households Bolivia 2008 3900 Encuesta de Hogares Ecuador 2013-2014 28680 Encuesta de Condiciones de Vida Guatemala 2014 11420 Encuesta Nacional de Condiciones de Vida Nicaragua 2009 on de Niveles de Vida 6450 Nicaragua - Encuesta Nacional de Hogares sobre Medici´ Notes: List of Household Surveys in the HIT database. 28 Table 2 The Gains From Agricultural Tariff Liberalization Country Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Developing world 2.50 1.01 4.64 -2.15 -0.44 2.20 2.64 Armenia 3.12 0.45 3.70 -0.58 -0.87 2.79 3.66 Azerbaijan 1.73 0.83 4.37 -2.64 1.64 2.59 0.95 Bangladesh 0.64 0.94 4.87 -4.23 -2.15 -0.37 1.78 Benin 1.50 1.81 4.25 -2.75 -4.75 -0.78 3.97 Bhutan 6.53 1.56 10.61 -4.08 1.80 6.51 4.71 Bolivia 3.56 1.11 5.01 -1.45 -0.00 2.70 2.70 Burkina Faso 0.72 0.64 3.79 -3.08 -0.09 0.73 0.83 Burundi -3.23 2.68 7.82 -11.05 -4.80 -4.59 0.21 Cambodia 0.17 1.25 4.95 -4.79 0.11 0.47 0.36 Cameroon 6.25 1.02 9.60 -3.35 -1.38 5.42 6.80 Central African Republic 3.05 0.99 7.29 -4.24 1.45 3.57 2.13 Comoros 0.46 0.64 1.59 -1.13 -1.41 -0.35 1.06 Cˆote d’Ivoire 1.66 1.53 4.62 -2.97 -3.75 -0.76 2.99 Ecuador 5.05 0.97 6.98 -1.93 2.35 5.90 3.56 Egypt, Arab Rep. 3.20 0.63 4.95 -1.75 -1.31 2.43 3.74 Ethiopia 0.41 0.73 3.68 -3.26 -0.20 0.63 0.83 Gambia, The 4.51 2.38 6.08 -1.57 -5.72 1.07 6.79 Georgia 1.49 0.27 2.16 -0.68 -0.27 1.25 1.53 Ghana -0.50 0.52 1.19 -1.69 -0.84 -1.02 -0.19 Guatemala 2.82 0.33 3.78 -0.96 -0.19 2.57 2.76 Guinea 2.56 0.85 4.93 -2.38 -1.63 1.82 3.45 Guinea-Bissau 1.47 2.06 5.89 -4.42 -4.62 -1.69 2.93 Indonesia 2.88 0.38 3.13 -0.25 0.63 3.05 2.42 Iraq 1.33 0.25 1.79 -0.46 0.57 1.48 0.91 Jordan 6.55 1.06 6.87 -0.32 2.64 7.58 4.94 Kenya 2.83 1.25 7.64 -4.82 -2.24 1.61 3.85 Kyrgyz Republic 1.10 0.30 1.82 -0.72 -0.21 1.01 1.22 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the elimination of country’s own agricultural import tariffs. 29 Table 2 (cont.) The Gains From Agricultural Tariff Liberalization Country Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Developing world 2.50 1.01 4.64 -2.15 -0.44 2.20 2.64 Liberia 2.22 0.72 3.22 -1.00 -1.41 1.28 2.69 Madagascar 0.53 0.78 3.08 -2.55 -1.80 -0.24 1.55 Malawi 0.55 1.33 3.36 -2.81 -3.40 -0.72 2.67 Mali 2.30 1.52 3.59 -1.29 3.65 4.25 0.60 Mauritania 3.27 0.96 4.94 -1.67 1.99 4.13 2.14 Moldova 0.78 0.43 1.44 -0.66 0.50 1.18 0.68 Mongolia 2.37 0.49 2.93 -0.55 1.25 2.87 1.63 Mozambique 3.18 1.97 4.98 -1.81 -4.93 0.91 5.84 Nepal 2.66 0.60 3.19 -0.52 1.52 3.37 1.85 Nicaragua 3.85 1.06 5.42 -1.57 1.80 4.60 2.80 Niger 0.99 1.04 4.06 -3.07 -2.41 -0.16 2.25 Nigeria 3.22 1.65 5.93 -2.71 -2.80 0.89 3.69 Pakistan 2.31 1.90 3.58 -1.27 4.92 4.45 -0.47 Papua New Guinea 2.24 0.65 4.90 -2.65 0.46 2.36 1.90 Rwanda 1.57 1.39 4.33 -2.76 2.89 3.40 0.51 Sierra Leone 2.40 1.48 5.40 -3.00 -3.36 0.98 4.34 South Africa 1.71 0.42 1.75 -0.04 0.60 1.66 1.06 Sri Lanka 3.02 1.52 4.74 -1.72 4.21 4.97 0.75 Tajikistan 3.08 0.41 3.24 -0.16 -0.38 2.93 3.31 Tanzania 3.68 1.65 5.94 -2.26 3.04 5.26 2.22 Togo 3.70 1.38 5.39 -1.69 -3.21 1.62 4.83 Uganda 3.26 1.09 5.44 -2.18 0.72 4.39 3.67 Ukraine 3.44 0.19 3.64 -0.20 -0.42 3.17 3.59 Uzbekistan 3.97 0.69 4.68 -0.71 -1.71 3.25 4.96 Vietnam 1.99 0.82 5.52 -3.53 -1.70 1.04 2.74 Yemen, Rep. 3.68 0.49 4.64 -0.96 0.67 3.84 3.17 Zambia 6.93 0.46 8.07 -1.14 0.75 7.35 6.60 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the elimination of country’s own agricultural import tariffs. 30 Table 3 The Gains From Agricultural Tariff Liberalization Aggregated Data Country Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Developing world 2.39 0.85 4.33 -1.94 -0.33 2.16 2.48 Armenia 3.52 0.32 4.12 -0.60 0.07 3.70 3.62 Azerbaijan 1.32 0.56 2.96 -1.64 1.25 2.02 0.77 Bangladesh 1.17 0.27 3.55 -2.38 0.20 1.29 1.09 Benin 1.28 1.87 4.29 -3.01 -4.86 -1.13 3.72 Bhutan 6.47 1.57 10.45 -3.98 1.77 6.38 4.62 Bolivia 3.73 1.17 5.21 -1.48 0.03 2.84 2.82 Burkina Faso 1.70 0.57 4.43 -2.73 0.27 1.94 1.67 Burundi -3.12 2.98 9.70 -12.82 -5.70 -5.20 0.50 Cambodia 1.53 0.82 4.33 -2.81 0.40 1.81 1.41 Cameroon 4.65 0.69 6.73 -2.08 -1.08 4.05 5.13 Central African Republic 1.34 0.99 9.28 -7.94 -1.31 0.75 2.05 Comoros 0.89 0.40 1.70 -0.81 -0.59 0.56 1.15 Cˆote d’Ivoire 1.72 0.99 3.70 -1.97 -2.31 0.16 2.46 Ecuador 4.35 0.81 5.84 -1.49 1.89 4.96 3.07 Egypt, Arab Rep. 2.11 0.69 4.09 -1.98 -1.61 1.22 2.83 Ethiopia 1.78 0.38 3.16 -1.38 0.10 2.02 1.92 Gambia, The 3.18 1.24 4.16 -0.98 -2.76 1.25 4.01 Georgia 1.86 0.23 2.39 -0.53 0.15 1.78 1.64 Ghana -0.56 0.47 1.23 -1.79 -0.77 -0.95 -0.18 Guatemala 3.15 0.32 4.02 -0.87 0.15 3.04 2.89 Guinea 3.15 0.84 5.90 -2.75 -1.57 2.46 4.03 Guinea-Bissau 3.10 1.62 6.88 -3.78 -3.09 0.71 3.79 Indonesia 1.57 0.24 1.87 -0.30 0.48 1.77 1.28 Iraq 1.33 0.25 1.79 -0.46 0.57 1.48 0.91 Jordan 4.49 0.56 4.95 -0.46 1.38 5.07 3.69 Kenya 2.54 1.11 6.63 -4.09 -2.18 1.39 3.57 Kyrgyz Republic 1.97 0.32 2.77 -0.79 0.18 2.02 1.84 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the elimination of country’s own agricultural import tariffs. 31 Table 3 (cont.) The Gains From Agricultural Tariff Liberalization Aggregated Data Country Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Developing world 2.39 0.85 4.33 -1.94 -0.33 2.16 2.48 Liberia 2.39 0.52 3.18 -0.79 -0.87 1.81 2.67 Madagascar 1.13 0.52 3.49 -2.36 -1.03 0.67 1.70 Malawi 0.74 0.88 2.98 -2.25 -2.16 -0.05 2.10 Mali 2.29 1.50 3.62 -1.33 3.58 4.18 0.60 Mauritania 3.84 1.00 5.14 -1.29 2.30 4.79 2.49 Moldova 0.78 0.35 1.40 -0.62 0.44 1.15 0.70 Mongolia 2.42 0.47 2.94 -0.53 1.20 2.90 1.69 Mozambique 2.32 1.88 4.32 -2.01 -4.76 0.30 5.05 Nepal 2.29 0.58 2.76 -0.47 1.51 2.99 1.48 Nicaragua 3.41 0.77 4.84 -1.43 1.04 3.85 2.81 Niger 2.75 1.06 5.17 -2.42 -2.48 1.45 3.93 Nigeria 2.39 1.79 5.25 -2.85 -3.60 -0.33 3.27 Pakistan 1.07 0.93 1.57 -0.50 2.43 2.13 -0.30 Papua New Guinea 1.63 0.45 2.99 -1.36 -0.35 1.42 1.76 Rwanda 2.24 1.63 5.19 -2.95 3.60 4.39 0.79 Sierra Leone 3.34 1.01 6.09 -2.75 -1.98 2.52 4.50 South Africa 2.52 0.99 2.57 -0.04 2.67 3.65 0.98 Sri Lanka 1.99 1.26 3.77 -1.78 3.50 3.75 0.25 Tajikistan 2.73 0.38 2.89 -0.15 -0.21 2.66 2.88 Tanzania 2.68 1.11 4.66 -1.99 1.56 3.26 1.70 Togo 3.35 1.25 5.28 -1.93 -2.91 1.44 4.35 Uganda 2.54 0.83 4.21 -1.67 0.43 3.36 2.93 Ukraine 2.49 0.14 2.62 -0.13 -0.31 2.30 2.60 Uzbekistan 4.23 0.60 5.22 -0.99 -1.37 3.67 5.04 Vietnam 1.80 0.75 4.13 -2.33 -1.76 0.82 2.58 Yemen, Rep. 2.10 0.28 3.07 -0.97 0.14 2.14 2.00 Zambia 7.44 0.46 8.61 -1.17 0.72 7.84 7.12 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the elimination of country’s own agricultural import tariffs. 32 Table 4 Bias arising from using aggregated instead of disaggregated data Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Average bias -0.10 -0.16 -0.31 0.20 0.11 -0.04 -0.16 Countries with positive bias 25 11 23 32 29 26 24 Countries with negative bias 29 43 31 22 25 28 30 Minimum bias -2.06 -1.14 -2.87 -3.71 -2.75 -2.83 -2.78 Maximum bias 1.76 0.57 2.00 1.98 2.96 2.40 1.67 Average absolute bias 0.75 0.22 0.84 0.53 0.65 0.94 0.55 Notes: Authors’ calculations based on HIT data. The table presents the bias associated with using aggregated instead of disaggregated data, by comparing results from Tables 2 and 3. 33 Table 5 The Gains From Agricultural Tariff Liberalization Across Commodities Vietnam Product Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Aggregate 1.99 0.82 5.52 -3.53 -1.70 1.04 2.74 Animal fats 0.17 0.07 0.17 0.00 0.20 0.28 0.08 Apples 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Banana 0.88 0.23 0.92 -0.04 0.59 1.16 0.57 Beans 0.01 0.01 0.01 -0.00 0.00 0.01 0.00 Beef 0.02 0.07 0.09 -0.06 -0.15 -0.06 0.10 Beer 0.12 0.05 0.12 0.00 -0.10 0.05 0.16 Cashew -0.02 0.03 0.00 -0.02 0.01 -0.01 -0.02 Cheese 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Cigarettes 0.61 0.17 0.61 0.00 0.31 0.70 0.39 Citrus 0.00 0.06 0.04 -0.04 -0.04 -0.01 0.02 Cloves 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Cocoa 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Coconut -0.01 0.01 0.00 -0.01 0.00 -0.01 -0.01 Coffee -0.15 0.12 0.03 -0.18 0.01 -0.13 -0.14 Corn -0.13 0.15 0.01 -0.14 -0.35 -0.36 -0.01 Cotton -0.00 0.00 0.00 -0.00 -0.00 -0.00 -0.00 Eggs 0.04 0.06 0.09 -0.05 0.04 0.06 0.02 Fish 0.37 0.12 0.50 -0.13 0.16 0.42 0.26 Grapes -0.00 0.01 0.00 -0.00 -0.00 -0.00 0.00 Greens 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Milk 0.02 0.01 0.02 -0.00 -0.01 0.01 0.02 Other 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Other Dairy 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Other Processed Food 0.92 0.26 0.92 0.00 -0.66 0.55 1.22 Other Spices 0.09 0.08 0.13 -0.04 0.17 0.18 0.01 Other alcohol 0.17 0.07 0.17 0.00 0.18 0.26 0.08 Other cereals 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Other crustaceans -0.02 0.06 0.03 -0.05 -0.04 -0.03 0.01 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the elimination of Vietnam’s own agricultural import tariffs. 34 Table 5 (cont.) The Gains From Agricultural Tariff Liberalization Across Commodities Vietnam Product Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Aggregate 1.99 0.82 5.52 -3.53 -1.70 1.04 2.74 Other fruit 0.05 0.13 0.22 -0.17 -0.20 -0.05 0.15 Other meat -0.06 0.06 0.05 -0.11 -0.07 -0.09 -0.02 Other non-staple 0.06 0.01 0.06 0.00 -0.00 0.06 0.06 Other nuts -0.03 0.05 0.02 -0.06 -0.05 -0.04 0.00 Other oil seeds -0.00 0.00 0.00 -0.00 -0.00 -0.00 -0.00 Other oils/fats 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Other staple foods -0.00 0.00 0.00 -0.00 -0.00 -0.00 -0.00 Other tobacco -0.00 0.03 0.01 -0.01 0.01 0.00 -0.01 Other vegetables -0.00 0.10 0.19 -0.19 -0.12 -0.07 0.05 Pepper -0.05 0.07 0.00 -0.05 0.04 -0.02 -0.06 Pork 0.28 0.12 0.56 -0.28 0.20 0.37 0.18 Potato 0.00 0.02 0.01 -0.01 -0.01 -0.01 0.00 Poultry 0.03 0.05 0.09 -0.07 -0.08 -0.02 0.05 Qat (chat) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Rice -1.57 0.73 0.04 -1.61 -1.77 -2.35 -0.58 Saffron 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Shrimp -0.05 0.05 0.00 -0.05 -0.01 -0.04 -0.03 Soya 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Sugar 0.09 0.11 0.19 -0.10 -0.04 0.04 0.08 Tea 0.13 0.06 0.18 -0.04 0.07 0.15 0.08 Tomato 0.03 0.02 0.04 -0.00 0.01 0.03 0.02 Vanilla 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Vegetable oils 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Wheat 0.01 0.00 0.01 0.00 0.00 0.01 0.01 Wine 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the (country by country) elimination of Vietnam’s own agricultural import tariffs. 35 Table 6 The Gains From Cereals Tariff Liberalization Country Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Developing world 0.42 0.26 0.67 -0.25 0.53 0.30 0.23 Armenia 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Azerbaijan 0.38 0.13 0.46 -0.08 0.44 0.24 0.20 Bangladesh 0.03 0.03 0.08 -0.05 0.01 0.05 -0.04 Benin 0.30 0.20 0.58 -0.28 0.07 0.46 -0.39 Bhutan 1.83 0.73 2.39 -0.56 2.61 0.90 1.71 Bolivia 0.89 0.28 0.99 -0.11 1.19 0.53 0.66 Burkina Faso 0.63 0.15 0.79 -0.16 0.67 0.62 0.05 Burundi -0.22 0.33 0.13 -0.35 -0.34 -0.02 -0.32 Cambodia 0.07 0.13 0.28 -0.21 0.15 0.07 0.08 Cameroon 0.96 0.45 1.36 -0.41 1.31 0.46 0.85 Central African Republic -0.99 0.37 0.30 -1.28 -1.03 -0.72 -0.31 Comoros -0.36 0.25 0.03 -0.39 -0.55 -0.17 -0.38 Cˆote d’Ivoire 0.54 0.18 0.72 -0.17 0.37 0.45 -0.08 Ecuador 1.07 0.64 1.34 -0.27 2.04 0.30 1.73 Egypt, Arab Rep. -0.00 0.00 0.00 -0.00 -0.00 -0.00 -0.00 Ethiopia 0.49 0.09 0.49 -0.01 0.59 0.44 0.15 Gambia, The 0.81 0.38 0.86 -0.05 0.41 0.94 -0.53 Georgia 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Ghana -0.02 0.18 0.32 -0.34 -0.10 0.10 -0.20 Guatemala 0.75 0.26 0.85 -0.10 1.09 0.41 0.68 Guinea 0.54 0.21 1.07 -0.52 0.55 0.43 0.12 Guinea-Bissau 1.92 0.70 1.97 -0.05 2.31 1.00 1.31 Indonesia -0.00 0.03 0.03 -0.03 -0.02 0.00 -0.03 Iraq -0.03 0.08 0.16 -0.20 -0.17 0.03 -0.20 Jordan 0.00 0.00 0.00 -0.00 0.00 0.00 0.00 Kenya 0.26 0.59 2.23 -1.97 -0.28 0.55 -0.83 Kyrgyz Republic 0.08 0.04 0.11 -0.03 0.12 0.05 0.07 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the (country by country) elimination of import tariffs on cereals. 36 Table 6 (cont.) The Gains From Cereals Tariff Liberalization Country Aggregate St.Dev. Consumption Income Pro-Poor Gains Gains Gains Gains Effects Effects Bias Poor Rich Developing world 0.42 0.26 0.67 -0.25 0.53 0.30 0.23 Liberia 0.38 0.09 0.43 -0.06 0.43 0.30 0.13 Madagascar 0.04 0.09 0.12 -0.09 0.00 0.13 -0.13 Malawi -0.01 0.07 0.07 -0.08 -0.07 0.09 -0.16 Mali 0.39 0.38 0.72 -0.33 0.74 0.06 0.68 Mauritania 1.05 0.33 1.16 -0.12 1.39 0.61 0.78 Moldova 0.22 0.08 0.26 -0.04 0.31 0.15 0.15 Mongolia 0.78 0.18 0.79 -0.00 1.04 0.56 0.49 Mozambique 0.07 0.15 0.20 -0.13 -0.06 0.26 -0.33 Nepal 0.63 0.32 0.81 -0.18 1.06 0.24 0.82 Nicaragua 1.03 0.52 1.33 -0.30 1.75 0.43 1.31 Niger 1.17 0.34 1.25 -0.08 0.72 1.44 -0.72 Nigeria 0.40 0.24 0.78 -0.38 0.11 0.59 -0.48 Pakistan 0.33 0.55 0.60 -0.28 1.03 -0.37 1.41 Papua New Guinea 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Rwanda 0.37 0.29 0.71 -0.33 0.66 0.19 0.46 Sierra Leone 0.67 0.21 0.95 -0.28 0.79 0.52 0.27 South Africa 0.00 0.00 0.00 -0.00 0.00 0.00 0.00 Sri Lanka 0.52 0.21 0.61 -0.09 0.74 0.21 0.53 Tajikistan 0.69 0.19 0.69 -0.01 0.66 0.65 0.00 Tanzania 1.29 1.03 1.73 -0.44 2.61 0.42 2.18 Togo 0.53 0.11 0.69 -0.16 0.55 0.46 0.10 Uganda 0.29 0.60 0.81 -0.52 1.19 0.01 1.17 Ukraine 0.46 0.10 0.47 -0.01 0.60 0.33 0.26 Uzbekistan 0.34 0.08 0.55 -0.21 0.34 0.33 0.01 Vietnam -1.70 0.83 0.05 -1.75 -2.70 -0.59 -2.11 Yemen, Rep. 1.11 0.34 1.13 -0.02 1.47 0.67 0.80 Zambia 1.66 0.19 1.78 -0.12 1.82 1.48 0.33 Notes: Authors’ calculations based on HIT data. The table presents the gains associated with the elimination of import tariffs on cereals. 37