The World Bank Economic Review, 37(2), 2023, 331–349 https://doi.org10.1093/wber/lhad002 Article The Double Dividend of a Joint Tariff and VAT Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 Reform: Evidence from Iran Kowsar Yousefi and Mohammad Vesal Abstract This paper provides empirical evidence on a novel complementarity between VAT and trade taxes. Downstream domestic firms require VAT receipts from importers to claim VAT on purchases, increasing incentives for honest reporting of imports. Trade gap, the difference between mirror and domestic trade reports in Iran at 6-digit HS disaggregation, is used to measure this complementarity. Iran introduced VAT in 2008 and, since then, has increased its rate from 3 to 9 percent. Difference-in-differences estimates show that a 1 percentage point increase in the VAT rate reduces the trade gap by about 2 percent. Consistent with the compliance mechanisms for VAT, a smaller effect for consumer products that have a shorter value chain is observed. Findings suggest that replacing tariffs with VAT results in a double dividend. Tax revenue might increase due to better tariff compliance and a broader VAT base. JEL classification: H26, F13, F14 Keywords: value added tax, trade liberalization, tariffs, chains effect, tax compliance 1. Introduction Developing countries raise a small amount of tax revenue due to low state capacity and high informality. They also rely on highly distortionary but easier to enforce tax instruments such as tariffs and corporate taxes. The advent of value added tax (VAT) and its adoption in many developing countries changed this pattern. VAT provided an avenue for revenue neutral tax reforms, like tariff reduction, that improve efficiency in developing countries. Furthermore, VAT is collected along the production chain often with an invoice-credit system that creates opposite evasion incentives for the two sides of a transaction. It also leaves a paper trail along the production chain which enables tax authorities to cross check tax reports. The invoice-credit system also withholds taxes at upstream firms, which creates an incentive for honest reporting (Keen 2008; Waseem 2022). Given the enforcement properties of VAT, a reform package that replaces tariffs with VAT could maintain tax revenue and improve economic efficiency. Kowsar Yousefi is an associate professor of economics at the Institute for Management and Planning Studies and adjunct professor at Sharif University of Technology, Tehran, Iran; her email address is k.yousefi@imps.ac.ir. Mohammad Vesal (cor- responding author) is an assistant professor of economics at Sharif University of Technology, Tehran, Iran; his email address is m.vesal@sharif.edu. We would like to thank seminar participants at Sharif University of Technology, University of Tehran, and the Institute for Management and Planning Studies for helpful comments and discussions. We specially acknowledge helpful comments by Ali Taiebnia, Mazhar Waseem, Ali Motavasseli, Arash Nekoei, Mostafa Beshkar, conference partici- pants at the Iran Economic Forum (2021), and the International Association for Applied Econometrics Annual Conference (2021). This article has benefited from partial funding from the Institute for Management and Planning Studies. All remaining errors are ours. © The Author(s) 2023. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com 332 Yousefi and Vesal The theoretical literature on the effectiveness of a joint tariff and VAT reform is inconclusive. Keen and Ligthart (2002) argue that the broad base of VAT allows for a smaller increase in the VAT rate to offset a reduction in tariffs, which creates further efficiency gains. Furthermore, in a destination-based system, VAT is imposed on imports, which creates a burden on the informal firms purchasing from formal ones (Keen 2008). This mechanism provides a more efficient burden on the informal sector compared to the de facto burden of tariffs discussed by many scholars, including Emran and Stiglitz (2005). Thus, an Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 under-investigated aspect of a joint tariff and VAT reform is the bilateral spillovers. Formal firms pur- chasing from importers would require VAT receipts to reclaim purchases VAT. They would also need valid receipts to claim expenses for corporate tax purposes. This creates a backward linkage that incen- tivizes importers to truthfully report imports (De Paula and Scheinkman 2010). However, De Paula and Scheinkman (2010) show potential formation of informal chains after the imposition of VAT when collu- sion along the value chain happens.1 Therefore, theoretically, the introduction of VAT has an ambiguous effect on trade tax evasion. When formation of formality chains is dominant then tariffs and VAT are complementary instruments in improving economic efficiency and raising tax revenue.2 In this paper, we empirically test the existence of a complementary relation between tariffs and VAT in Iran. Specifically, we aim to measure the impact of a VAT rate increase on trade evasion. Hypothetically, there are two mechanisms in place. First, imposing VAT increases the cost of honest reporting like any other tax. Hence, illegal imports and the trade gap should increase. Second, VAT creates chain effects and incentivizes honest reporting. We aim to measure which of these opposing mechanisms is dominant. Fur- thermore, we provide direct evidence on the importance of chain effects by looking at consumer products with shorter value chains compared to non-consumer products that have longer value chains. In our con- text, the government introduced VAT in 2008 without any significant trade liberalization. Many product codes were exempted from VAT, creating a natural control group. Initially, the VAT rate was 3 percent and then gradually increased to 9 percent. At the same time, the government has been continuously changing tariffs before and after the introduction of VAT at individual product categories without any clear lib- eralization trend. Therefore, we have an ideal setting to separate the impact of tariffs and VAT on trade evasion. In the literature, the empirical evidence on the effectiveness of a joint tariff and VAT reform is sparse. In one of the few empirical studies, Baunsgaard and Keen (2010) show that tax revenue losses due to trade liberalization are offset by domestic taxes like VAT for middle- and high-income countries but not for poor countries. We did not find any studies on the complementarity between tariffs and VAT. Empirical evaluation of this complementarity is impossible under a joint reform because a uniform tariff reduction also affects trade evasion (Fisman and Wei 2004). To construct a measure of trade evasion, we subtract reported exports to Iran by the World Integrated Trade System (mirror reports) from reported imports by Iran Customs Agency at the Harmonic System (HS) 6-digit level. This is called the trade gap in the literature (Fisman and Wei 2004). We employ a difference-in-differences strategy and control for HS6 product and year fixed effects to estimate the impact of VAT on the trade gap. The quasi-random variation created by the introduction of and changes in VAT on a subset of HS6 codes allows us to compare the evolution of the trade gap for VATable products to others over 2005–2016. The identification assumption is that, in the absence of VAT, the evolution of the trade gap for the two groups would be identical. 1 Collusion between exporters and importers is unlikely because (a) the Iran market is small, (b) in the absence of export taxes, exporters do not have an incentive to misreport, (c) Iran Customs Administration does not check mirror reports for evasion detection. Even if collusion happens, this would bias our estimates against finding a significant impact of VAT. 2 Three additional reasons support the formation of formality chains. First, basic utilities charge VAT on their sales and hence many firms that use these common inputs are incentivized to register for VAT to claim back input VAT (Waseem 2022). Second, valid VAT receipts are required for reported costs under the corporate tax. Finally, VAT collection is easier at the port of imports, and hence a formality chain is triggered at this point. The World Bank Economic Review 333 Our results indicate that a 1 percentage point increase in the VAT rate reduces the trade gap by about 2 percent. This estimated elasticity is robust to multiple tests such as controlling for mean reversion and exclusion of sanction years.3 We also find evidence consistent with stronger chain effects for products with longer value chains. Consumer goods have shorter value chains compared to non-consumer ones (e.g. raw, intermediary, and capital goods4 ). Therefore, we expect weaker chain effects of VAT on consumer goods. Our results corroborate this conjecture. A 1 percentage point increase in the VAT rate increases the trade Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 gap by 4.6 percent for these goods. For non-consumer goods, for which the chain effect is stronger, the overall effect of VAT is −3.1 percent. This finding is consistent with the theoretical result of De Paula and Scheinkman (2010) on the coexistence of formality and informality chains. We observe two interesting non-linearities for this complementarity. For products with a larger initial trade gap, VAT is more effective in reducing the trade gap. Interpreting the initial trade gap as a measure of informality for a given product, we can claim that VAT has a stronger effect for products with higher degrees of informality. A higher tariff results in a smaller VAT effect, so that for products with tariffs above 35 percent, an increase in VAT actually increases the trade gap. Our paper makes five contributions. First, we provide evidence on a novel complementary mechanism that links VAT and trade taxes. Studying the inter-relation of taxes is rare in the literature and involves real empirical challenges. The richness of our context allows us to identify this complementarity. Second, there is a nascent literature that studies the impact of VAT on tax evasion. Most of these studies rely on independent audits and experimental variation to tease out evasion responses (e.g. Pomeranz 2015; Carrillo, Pomeranz, and Singhal 2017). These studies often require mobilization of significant resources and are hard to conduct. The existence of mirror reports for trade flows allow us to easily construct evasion proxies. Therefore, we are able to add empirical support to the idea that VAT can reduce tax evasion but from a novel angle, i.e. its impact on trade tax evasion. Third, we shed light on the significance of VAT chain effects and the coexistence of formality and informality chains by looking at products with shorter and longer value chains. Fourth, we provide suggestive evidence that VAT is even more effective for product codes that have higher initial informality. Finally, we show that the complementarity between VAT and tariffs fades away as the tariff rate increases. The rest of the paper is organized as follows. First, we briefly discuss the relevant literature. Next, we provide a short description of the Iranian context. Then, we describe our data and present the summary statistics. The method section presents the estimation strategy. Results are presented next followed by a discussion of their implications. The final section concludes. 2. Literature Our work relates to five branches of the literature. First is the theoretical literature that discusses the pros and cons of a joint tariff and VAT reform. Keen and Ligthart (2002) argue that VAT is more efficient than trade taxes because of a broader base and no distortion of trade flows. In contrast, Emran and Stiglitz (2005) claim that the presence of a large informal sector erodes the revenue gains of VAT. Therefore, the results derived in the earlier literature are “unhelpful at best and potentially misleading as the basis of indirect tax policy reform in developing countries.” This argument relies on the fact that tariffs are 3 There might be concerns that pass-through of taxation affects either trade flows or illegal importation. While VAT pass-through is complicated and possibly heterogeneous for different products, we do not expect a role from it to alter our results. First, Iran imports constitute a small share of imports in the Middle East. Therefore, internal policies do not affect world-level prices. Therefore, the importers would bear the full burden of border taxes. Second, within the country we might have different degrees of pass-through, which seems inconsequential for the incentives of firms along the value chain to collude or hide taxes. 4 The categorization of consumption, raw, intermediary, and capital goods, is referred to as the Broad Economic Classifi- cation (BEC). 334 Yousefi and Vesal better suited for imposing a de facto burden on the informal firms. Davies and Paz (2011) use a simu- lation method with a model of heterogeneous firms to show that VAT does not necessarily expand the informal sector but that tariff reduction reduces the informal sector. To summarize, the theoretical debate is inconclusive as there are special cases under which the joint tariff and VAT reform are not welfare improving. The second literature which relates to our study is the empirical studies on the intersection of tariffs Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 and VAT. This part of the literature is surprisingly thin. Baunsgaard and Keen (2010) is the only study that looks at the impact of domestic tax reforms in offsetting tariff revenue losses after trade liberalization. Their findings show that poor countries are unable to recoup the lost revenue, while the rich countries actually improve revenue after trade liberalization. For middle-income countries there does not seem to be a decrease in revenue after tariff reduction. The empirical literature provides no guidance on the potential mechanisms that might become operative under a joint reform. This is where our paper fits in. We provide evidence on the complementary effect of VAT on trade evasion. We empirically quantify the role of VAT in reducing trade evasion and hence improving the collection efficiency of tariffs. The third literature related to our study is the growing literature on the enforcement properties of VAT. We contribute to this literature by looking at the spillover of VAT on trade taxes. This literature identifies three mechanisms that improve the enforcement properties of VAT (Keen and Lockwood 2010; De Paula and Scheinkman 2011, 2010). First, VAT is collected through the production chain. Receipts from seller and purchaser of a product could be cross-checked. These receipts leave a paper trail for tax agencies that reduces fraud. Second, VAT creates opposite incentives for the two sides of a transaction, which reduces the likelihood of collusion. Third, purchases VAT is effectively withheld at the upstream firm. Even if the downstream firm evades VAT, it bears some of the VAT burden on purchases. The literature has shown that the proper operation of these mechanisms relies on the extent of the informal sector (De Paula and Scheinkman 2011, 2010. Recently, a series of well-identified empirical studies have shown that VAT chain effects and mechanisms are quite strong in improving compliance. For example, Pomeranz (2015) reports on an audit experiment in Chile and shows firms subject to the VAT paper trail respond less to additional audit threats. Waseem (2022) provides strong evidence in support of the withholding effect of VAT from Pakistan. Hoseini and Briand (2020) also report on the importance of chain effects in the Indian context and show forward and backward linkages matter for the efficiency gain of replacing sales taxes with VAT. Li and Wang (2020) use the expansion of VAT in China to show better compliance with VAT than a turnover tax, especially for business-to-business transactions. Finally, Waseem (2020) report on extensive VAT evasion in Pakistan which is weaker for upstream stages of the production chain. Our paper also contributes to the VAT literature by looking at the compliance effect of VAT on trade taxes. The fourth line of research that relates to our work is the literature on the impact of VAT on trade flows. Theoretically, a comprehensive uniform VAT system should not affect trade flows (Feldstein and Krugman 1990). Corroborating this theoretical prediction, Benzarti and Tazhitdinova (2021) show that in the EU, VAT does not affect trade flows. However, there are papers (e.g. Sharma 2020) that find an impact of VAT on trade flows in developing countries with weaker enforcement regimes. Our focus is on the impact of VAT on import evasion and not trade flow per se. But, we do find a negative impact from VAT on imports to Iran. The fall in imports reported by Iran importers is, however, smaller than the fall in imports reported by exporters to Iran, which creates our main result on trade gap. Finally, our work speaks to the fiscal externality literature. This literature is quite diverse and ranges from inter-jurisdictional spillovers of taxes (e.g. Marion and Muehlegger 2018) to cross-tax-base exter- nalities (e.g. Goolsbee 2004 and Waseem 2018). One key aspect of cross-tax-base externalities arises due to enforcement spillovers that might be positive or negative. Li et al. (2021) show that strengthening enforcement of VAT in China results in more evasion in payroll taxes due to firms’ optimization and lim- ited government resources. López-Luzuriaga and Scartascini (2019) find a positive enforcement spillover from property taxes on sales taxes in Argentina. They also discuss the theoretical possibility of a negative The World Bank Economic Review 335 Figure 1. Tax to GDP, and Share of Consumption and Import Taxes in Tax Revenue Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 Source: Authors’ analysis based on data from the Iran National Tax Administration and the Central Bank of Iran. Note: The solid black line shows tax revenue as a percentage of gross domestic product (GDP) over 1971–2016. The vertical axis for this series is the left axis. The gray solid line shows the share of import tax revenue in total tax revenue. The dashed gray line shows the share of consumption tax revenue in total tax revenue. Both these series are shown on the right vertical axis. The vertical dashed line shows the introduction of VAT in 2008. Tax revenues include only central government levies. spillover due to the interplay of perceptions about penalties and detection probabilities. Our work adds to this literature by showing that the superior enforcement properties of VAT create an incentive for truthful reporting of imports. 3. Context The Iranian tax system features five main instruments: corporate income tax, personal income taxes, consumption taxes, import taxes, and wealth taxes. Iran does not have a comprehensive personal income tax system and treats various sources of income separately. Tax law has been reformed several times during the past four decades, with little impact on government tax revenue. Iran collects a very small share of GDP in taxes compared to other developing countries. Figure 1 shows that tax revenue as a percentage of GDP in Iran never exceeds 10 percent. If anything, there is a gentle declining trend over time. The figure also shows the share of tax revenue from consumption and import taxes. Consumption taxes include sales taxes, excise taxes, and VAT. In 2008, sales tax was replaced by VAT. However, other excise taxes are available both before and after 2008. Prior to 2008, consumption taxes had a share of between 5 and 22 percent of total tax revenue. But after the introduction of VAT, we see a peculiar increase 336 Yousefi and Vesal Figure 2. Evolution of Maximum and Average VAT Rates 9 8 Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 7 6 Tax rate (%) 4 3 2 1 05 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year Max VAT rate Mean VAT rate Source: Authors’ analysis based on data from the Iran Customs Administration. Note: The solid black line shows the maximum applicable VAT rate in the given year. VAT was introduced in mid-2008 and hence the reported maximum rate is half of the maximum rate. The black dashed line with hollow markers shows the mean statutory VAT rate for each year. The average rate is the simple mean of VAT rates for HS6 codes available in our data in each year. in the share of consumption taxes. After 2012, consumption taxes raise a higher share of revenue relative to import taxes. The reform seems to have a small effect on total tax revenue, as tax to GDP increases from around 6 percent to 7 percent. But the importance of import taxes declined around the time of VAT introduction. Figure 2 shows the evolution of the maximum statutory VAT rate and its mean over the sample period. In mid-2008, the government introduced VAT with a 3 percent rate.5 In the three following years, the rate stayed at 3 percent and then it saw 1 or 2 percentage point increases in the following years. The figure also shows the mean statutory VAT rate, which is always below the maximum, because many products are either exempt or zero rated. VAT exemption means that product is outside the VAT net. Therefore, producers of that product cannot reclaim any purchases VAT paid on their inputs and bear some burden of VAT. Zero-rated products are eligible for VAT, and hence their producers can reclaim any VAT paid on inputs while they do not pay VAT on their sales. Thus, zero-rated producers receive no burden of VAT. This distinction was not well understood in Iran and the VAT law passed in 2008 effectively leaves zero rating only for exports. Since we focus on imports, zero rating does not apply to any of our goods. We only have VATable products that are subject to the statutory VAT rate or non- VATable products that are exempt from VAT. This translates to a zero tax rate at the port of imports and no outstanding purchases VAT for downstream firms. Reading the political debate around VAT in Iran and talking with politicians in the field, we believe that the main motive for the exclusion of some products from VAT was to mitigate the regressiveness of this tax. In other words, commodities that weighed heavily 5 We take the VAT rate to be 1.5 percent in 2008 because VAT was not collected for the first half of the year. The World Bank Economic Review 337 for the poor were exempted. Therefore, ease of trade evasion was not a factor in setting VAT. Finally, Iran’s VAT does not have a registration threshold.6 4. Data Our data come from three different sources. First, we use Iran Customs Administration (ICA) reported Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 imports and collected tariffs at 8-digit harmonized system classification (HS8). We merge these data with the world’s reported exports to Iran from the World Integrated Trade Solution (WITS) database at 6- digit HS codes (HS6). Finally, statutory tariffs and VAT rates at HS8 disaggregation are taken from the Handbook of Customs Regulations (also known as the Iran Imports and Exports Tariff Book), published annually by the Ministry of Industry, Mine and Trade. Since WITS data are at HS6 level, we collapse the two other sources to this level. We trim the data by excluding observations with trade values below the 1st and above the 99th percentile of its distribution in each year. Furthermore, observations with negative tariff rates are excluded as data errors. We also exclude observations affected by the Preferential Trade Agreements (PTAs) between Iran and a few countries, e.g. Pakistan and Bosnia. In PTAs, there are limited lists of goods that are subject to discounted tariff rates. The final cleaned data set contains 37,448 observations from 2005 to 2016. For the purpose of aggregation and merging, VAT and tariff rates are simply averaged for HS8 codes inside a given HS6 code.7 The trade gap is calculated as the difference between the log of world reported exports to Iran and the log of reported imports by ICA: TradeGap := ln (exportit ) − ln (importit ).8 Iran imports might differ from mirror reports due to reporting conventions, noise, or intentional misreporting. Exports are often based on free-on-board (FOB) prices, while imports are based on cost-insurance-freight (CIF) prices. This should result in a negative trade gap, because CIF prices include cost, insurance, and freight charges. We do not expect the changes in this part of the trade gap to be correlated with the introduction of VAT across goods (difference-in-differences term). Trade gap might also contain noise due to unintentional errors, which could result in attenuation bias in our estimates. Finally, trade gap reflects intentional misreporting to evade export or import duties. Since there are very few examples of export duties around the world, exporters do not have an incentive to misreport. But importers face substantial tariffs, especially in Iran. They would like to report smaller values to pay lower tariffs. Changes in this part of the trade gap are due to changes in border taxes or the enforcement regimes in the destination. Therefore, we do not take trade gap as a measure of trade evasion, but we assume that the correlations between the trade gap and the tariffs and VAT are due to trade evasion. Trade evasion could happen through the three channels of (a) under- declaration of quantity, (b)) under-declaration of prices, (c) misclassification of imports, i.e. reporting the consignment under a different HS6 category. To control for tariff exemptions we rely on the tariff gap, which is defined as the gap between statutory and effective tariff rates. The latter is the sum of collected tariffs for a given HS6 code divided by the sum of reported value of imports by Iranian importers (Yousefi, Vesal, and Pilvar 2020). Table 1 reports summary statistics for the main variables. Average reported exports stand at around USD 6.5 million, while average reported imports are USD 4.2 million. There is significant standard 6 During the implementation of VAT, Iran Tax Administration issued notices between 2008 to 2015 which called firms in specific sectors or above specific turnover thresholds for VAT registration. By 2015 all firms were called to register. This procedure is not reflected in our data set because notices cannot be mapped to HS6 product codes. Furthermore, the import VAT, which is the topic of this study, was enforced from the first day. 7 In 2012, following the UN sanctions and economic downturn in Iran which caused a foreign currency crisis in the country, the government limited importation of so-called luxury products by doubling their statutory tariffs. Accordingly, we have doubled the tariffs of those goods in our data set. 8 All observations with zero exports or imports are excluded from the analysis. 338 Yousefi and Vesal Table 1. Summary Statistics Standard deviation Number of observations Average Median Overall Within Between (1) (2) (3) (4) (5) (6) Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 Exports to Iran ($, WITS) 37,448 6,439,324 1,396,956 23,347,761 12,431,867 25,001,497 Imports ($, ICA) 37,448 4,110,559 1,059,896 7,293,567 3,713,054 6,919,346 VAT rate (%) 37,448 2.93 1.50 3.30 2.75 2.09 VAT dummy 37,448 0.55 1.00 0.50 0.37 0.37 Statutory tariff rate (%) 36,886 21.37 12.00 23.67 9.47 28.91 Trade gap 37,448 0.17 0.17 1.93 1.21 1.90 Tariff gap (%) 36,886 4.93 0.29 14.11 10.14 15.86 Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The table shows summary statistics of variables reported in the rows. Column (1) reports the number of non-missing observations. Column (2) reports the average of the variables. Column (3) shows the median. Columns (4) to (6) respectively report the overall, within, and between standard deviation of variables. ICA, Iran Customs Administration; WITS, World Integrated Trade Solution. deviation both within and between HS6 codes. The VAT rate also shows plausible within and between variations due to exemptions and temporal changes in the max statutory rate. Average statutory tariff is relatively high and stands at 21.4 percent. This variable has a large between standard deviation and a significant within variation. The average trade gap is 0.18, which means that reported exports to Iran are on average 18 percent more than reported imports. This is despite the fact that import values are based on CIF prices, while exports are based on FOB. Figure 3 shows the distribution of the trade gap for VATable and non-VATable HS6 codes. Apart from the greater dispersion of the trade gap for VATable goods, the two distributions are similar. They are slightly right-skewed, with more mass on the positive numbers of the trade gap, but they both extend well to the negative numbers. Figure 4 plots the evolution of the average tariffs for VATable and non-VATable goods. The average tariffs are smaller for VATable goods but this has been the case from the start of the sample. The gap widens toward the end of the sample years but seems fairly stable between 2005 to 2014. 5. Method We use the difference-in-differences (DID) strategy to identify the impact of VAT on the trade gap. The introduction of VAT in 2008 and the rate increases afterward create quasi-random variations that allow us to tease out the impact of VAT from tariffs. Figure 5 shows the evolution of the average trade gap for VATable and non-VATable HS6 codes. The trade gap for both categories is decreasing prior to 2008, but starts to diverge after 2008. Table 2 summarizes the averages of trade gap for VATable and non-VATable products before and after the introduction of VAT to complement fig.5. Average trade gaps for both product categories increase over time; however, the rise for VATable products is much smaller than non- VATable ones. The table reveals that the trade gap for VATable products shrinks by 23 percent relative to non-VATable products after the implementation of VAT. This differential change is significant at the 1 percent level. Figure 5 and table 2 do not control for confounders but are still suggestive of parallel trends prior to VAT introduction and a divergence between VATable and non-VATable products after VAT. Next we discuss several regression specifications that control for potential confounders to establish causality. The first regression specification we estimate is s s e ln(exportit ) − ln(importit ) = αi + δt + β VATit + γ τit + ψ (τit − τit )+ it , (1) The World Bank Economic Review 339 Figure 3. Kernel Density of the Trade Gap for VATable and Non-VATable Goods, 2005–2016 .4 Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 Probability density function .1 .20 .3 −6 −5.5 −5 −4.5 −4 −3.5 −3 −2.5 −2 −1.5 −1 −.5 0 .5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 Trade gap non VATable goods VATable goods Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The solid black line shows the kernel density estimation of the distribution of the trade gap for VATable goods. The dashed line shows a similar distribution for non-VATable goods. We have excluded observations below the 1st percentile and above the 99th percentile of the trade gap distribution in this plot. where, the dependent variable is the trade gap (ln (exportit ) − ln (importit )) for product i (HS6) in year t. The variable VATit is either a dummy variable that equals 1 if commodity i is VATable after 2008, or the VAT rate for commodity i in year t. The latter variable has more variation and is preferred. The variable τit s is the statutory tariff rate for product i in year t and τit s − τit e reflects tariff exemptions, i.e. the difference between statutory and effective tariff rates. The variables α i and δ t indicate HS6 and year fixed effects. The coefficient of interest is β . When VATit is a dummy, β shows the average percentage change in the trade gap for VATable commodities relative to non-VATable ones after VAT introduction. Similarly, when we use the VAT rate as the explanatory variable, this coefficient shows the percentage change in the trade gap in response to a 1 percentage point increase in the VAT rate. The identification assumption in (1) is that in the absence of VAT the evolution of the trade gap for VATable and non-VATable products would be similar. The inclusion of HS6 fixed effects removes the influ- ence of any time-invariant product characteristics that correlate with VAT and the trade gap. For example, the government might have exempted necessities that are harder to misreport. In fact, the distributional concerns around VAT influenced the decision to include products in this tax and the government did not target ease of trade tax evasion. To the extent that this targeting results in time-invariant differences between VATable and non-VATable products, our results are valid. Year fixed effects control for flexible global trends in the trade gap that are common to both VATable and non-VATable products. For example, the ICA might have been implementing enforcement reforms that prevent misreporting. Furthermore, the parallel trend of the trade gap for VATable and non-VATable products supports our identification ( fig.5). Albeit, all these specifications (e.g. (1)) are vulnerable to events that occur after 2008 and affect VATable and non-VATable products deferentially. 340 Yousefi and Vesal Figure 4. Evolution of Average Tariff Rates for VATable and Non-VATable Goods 50 Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 40 Average tariff 30 20 10 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year non VATable goods VATable goods All goods Source: Authors’ analysis based on data from the Iran Customs Administration. Note: The figure shows the average tariff rates for VATable goods (thick solid line), non-VATable goods (dashed line), and all goods (thin solid line). The mean rate is the simple mean of tariff rates at the HS6 level in each year. The tariff does not include the VAT rate. We have excluded observations below the 1st percentile and above the 99th percentile of the trade gap distribution in this plot. To study the heterogeneous effect of VAT on products with a short and long value chain, we use the BEC, which splits HS codes into capital, intermediate, and consumer goods. Since consumer goods are products that are intended for the final consumer, they have the shortest value chain. Specification (2) measures this heterogeneity: ln(exportit ) − ln(importit ) = αi + δt + β VATit + ηVATit × 1[cons]i s s e +γ τit + ψ (τit − τit )+ it , (2) where 1[cons]i shows whether product i is categorized as a consumer good under BEC. The omitted category is non-consumer products. Therefore, the coefficient estimate for η reflects the additional impact of VAT on consumer goods relative to non-consumer ones. The variable η is expected to be negative if the effect of VAT is stronger for longer value chains. Finally, we look at the heterogeneity of the VAT effect in HS6 codes with a large initial trade gap by estimating the following specification: ln(exportit ) − ln(importit ) = αi + δt + β VATit + ζ VATit × TradeGap0,i s s e +γ τit + ψ (τit − τit )+ it , (3) where TradeGap0, i reflects the average of the trade gap for HS6 code i from 2005 to 2007 (before VAT introduction). Some products have higher trade evasion during this time period. The impact of VAT on these products might go either way. Entrenchment of informal links might be the cause of high initial trade gaps. If these links have no formal downstream connections, VAT would not be effective The World Bank Economic Review 341 Figure 5. Evolution of the Trade Gap for VATable and Non-VATable Products .6 Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 .4 Average trade gap .2 0 −.2 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Year non VATable goods VATable goods Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The figure shows the average trade gap for VATable goods (solid line) and non-VATable goods (dashed line). Observations below the 1st percentile and above the 99th percentile of the trade gap are excluded. The vertical dashed line shows the introduction of VAT in 2008. Table 2. Average Trade Gap for VATable and Non-VATable Products, before and after VAT Introduction Before VAT (<2008) After VAT (≥2008) Col. (2) − Col. (1) (1) (2) (3) (1) VATable products 0.015 0.20 0.19*** (0.02) (0.012) (0.02) (2) Non-VATable products −0.08 0.33 0.42*** (0.04) (0.028) (0.05) (3) Row (2) – row (1) 0.098*** 0.13*** −0.23*** (0.04) (0.027) (0.05) Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The table shows average and standard error of trade gap for various product and year groups. Rows (1) and (2) correspond to VATable and non-VATable products, respectively. Columns (1) and (2) correspond to years prior and after the introduction of VAT. Column (3) shows the difference between columns (2) and (1). Row (3) shows the difference between rows (2) and (1). Standard errors are reported in parentheses. in luring them into the formal sector. On the other hand, a large initial trade gap might reflect better trade evasion opportunities due to product characteristics or ineffective customs regulations. The im- position of VAT in this case would result in a stronger effect, as VAT enforcement technology would shift large parts of imports to the formal sector. An alternative interpretation is that products with a small trade gap are those with a small share of informal importers that are most likely linked with the 342 Yousefi and Vesal Table 3. Regression Results for the Impact of VAT on the Trade Gap VAT dummy VAT rate Dependant variable: trade gap (1) (2) (3) (4) VAT −0.12* −0.10 −0.019* −0.020* Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 (0.064) (0.075) (0.01) (0.0099) τs – 0.026*** – 0.026*** (0.0028) (0.0027) τs − τe – −0.016*** – −0.016*** (0.0033) (0.0032) Observations 37,448 36,886 37,448 36,886 2 R 0.56 0.57 0.56 0.57 Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The table shows coefficient estimates and standard errors from four regressions. In all regressions the dependent variable is the trade gap (difference between log of exports to Iran reported by WITS and log of imports reported by ICA). Columns (1) and (2) include a VAT dummy that is equal to 1 for VATable goods after 2008. Columns (3) and (4) include the VAT rate in the regression instead. The term τ s is the statutory tariff rate and τ s − τ e is tariff exemption, defined as the difference between the statutory and effective tariff rates. All regressions include constant variable, HS6 and year fixed effects. Standard errors are corrected for two-way clustering at HS4-year level and are reported in parentheses below coefficients. There are 1,109 HS4 clusters. ICA, Iran Customs Administration; WITS, World Integrated Trade Solution. *, **, and *** respectively show significance at 10, 5, and 1 percent level. informal domestic producers. The sign of ζ would determine which of these stories is relevant in our context. 6. Results This section provides the estimation results for the impact of VAT on illegal trade. We first look at the main results, including the average effect of VAT on the trade gap and its heterogeneity across consumer and non-consumer products. Then we conduct several robustness checks to see the sensitivity of the results. 6.1. Main Results Table 3 shows the coefficient estimates from regressions of the trade gap on VAT and other covariates. Columns (1) and (2) use the VAT dummy which is equal to 1 for VATable products after 2008 and 0 otherwise. Column (1) shows that the trade gap for VATable products decreases by about 15 percent (significant at 1 percent) after the introduction of VAT. In column (2), we control for statutory tariffs and tariff exemption. As a result, the VAT dummy coefficient reduces to 10 percent and becomes insignificant. Columns (3) and (4) use the VAT rate, which varies from 0 to 9. These columns exploit a richer variation by using the introduction and rate changes of VAT. The coefficient estimate from our preferred specifi- cation in column (4) suggests that a 1 percentage point increase in the VAT rate results in a reduction of 2 percent in the trade gap. The effect of the VAT rate is the opposite of statutory tariffs. A 1 percentage point increase in statutory tariffs increases the trade gap by 2.6 percent. This means that a uniform 1 percentage point reduction in tariffs, combined with a 1 percentage point increase in VAT, has a double dividend in reducing the trade gap. The combined effect is 4.6 percent, with VAT playing the main role in reducing the trade gap.9 Table 4 looks at the heterogeneity of the VAT impact across different product categories. Again we report estimates from the VAT dummy and the VAT rate regressions. In columns (1) and (5) we add the interaction between the VAT measure and a dummy for consumer products. The coefficient estimate for 9 We also run regressions that include the average tariff of other HS6 codes within a given HS4 to test for misclassification as a mechanism for trade tax evasion. However, the coefficient estimate for this variable is very small (0.00057) and insignificant at conventional levels. The World Bank Economic Review 343 Table 4. Heterogeneity of the Impact of VAT on the Trade Gap VAT dummy VAT rate Dependant variable: trade gap (1) (2) (3) (4) (5) (6) (7) (8) VAT −0.16* −0.064 −0.21 −0.25**−0.031** −0.013 −0.049** −0.044** Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 (0.077) (0.12) (0.14) (0.091)(0.01) (0.013) (0.013) (0.016) VAT × 1[cons] 0.42*** – – 0.31*** 0.077*** – – 0.037** (0.12) (0.088) (0.019) (0.015) VAT × TradeGap0 – −0.16*** – −0.17*** – −0.027*** – −0.030*** (0.03) (0.032) (0.0043) (0.0043) VAT× τ s – – 0.0074** 0.0048* – – 0.0014** 0.0013*** (0.0024) (0.0024) (0.00039) (0.00039) τs 0.025*** 0.025*** 0.022*** 0.022*** 0.024*** 0.026*** 0.020*** 0.019*** (0.0027) (0.0027) (0.0026) (0.0024) (0.0023) (0.0028) (0.0021) (0.0019) τs − τe −0.016*** −0.017*** −0.016*** −0.017*** −0.016*** −0.017*** −0.017*** −0.017*** (0.0032) (0.0033) (0.0031) (0.0032) (0.0029) (0.0034) (0.0026) (0.0027) Observations 36,551 27,630 36,551 27,630 36,551 27,630 36,551 27,630 2 R 0.56 0.59 0.56 0.6 0.57 0.6 0.57 0.6 Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The table shows coefficient estimates and standard errors from eight regressions. All regressions include τ s (statutory tariff rate), τ s − τ e (tariff exemption), and HS6 and year fixed effects. In columns (2–4) and (6–8), the average initial trade gap is controlled, and observations in those years (2005–2006) are excluded. For the columns including factor interactions, the factors are controlled (but not reported). Standard errors are corrected for two-way clustering at HS4-year level and are reported in parentheses below coefficients. There are 1,109 HS4 clusters. *, **, and *** respectively show significance at 10, 5, and 1 percent level. this term is positive and statistically significant at the 1 percent level in both regressions, meaning that consumer products receive a significantly different impact from the implementation of VAT. The trade gap for consumer products actually increases, whereas non-consumer products show a reduction. Based on column (5), a 1 percentage point increase in the VAT rate results in an increase of (−3.1 + 7.7 =) 4.6 percent in the trade gap for consumer products, and a reduction of 3.1 percent for non-consumer products (coefficients are significant at 1 percent and 5 percent, respectively). This confirms the earlier conjecture that products with a longer value chain receive a larger impact from the VAT implementation. Columns (2) and (6) of table 4 look at the role of the initial trade gap in mediating the effect of VAT. The initial gap is defined as the average trade gap in the early years of the data set, 2005–2006.10 The negative coefficient estimate suggests that products with a larger initial trade gap receive a stronger impact from VAT. Based on column (6), a 100 percent increase in the initial trade gap intensifies the impact of VAT on the trade gap by 5.4 percentage points. Evaluated at the mean of initial trade gap (12 percent), the estimated impact of VAT on the trade gap is 2.45 percent, which is close to the average effect in column (4) of table 3. Columns (3) and (7) of table 4 look at the impact of tariffs on the VAT effect by including an interaction of tariffs and VAT in the regression. The interaction term is positive and significant. Based on column (7), a 10 percentage point increase in tariffs lowers the marginal impact of VAT on the trade gap by 1.4 percent. This is consistent with the idea that higher tariffs create stronger evasion incentives and lower the benefit of reclaiming input VAT. If tariffs go above 35 percent then VAT would actually increase the trade gap. Including the three interaction terms in a single specification does not change the pattern of results (columns 4 and 8). 10 In models that include the initial gap, we exclude the years defining initial trade gap to avoid any potential endogeneity between the initial trade gap and the dependent variable. However, including such observations does not affect the sign and magnitude of results but improves the significance of coefficients. 344 Yousefi and Vesal Table 5. Robustness of the Impact of VAT on the Trade Gap to Exclusion of Specific Years Excluding 2012–2013 Excluding 2008–2014 Dependant variable: trade gap (1) (2) (3) (4) (5) (6) (7) (8) VAT rate −0.028** −0.013 −0.045** −0.045** −0.026* −0.014 −0.040** −0.048*** Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 (0.01) (0.013) (0.012) (0.016) (0.01) (0.01) (0.011) (0.009) VAT × 1[cons] 0.066*** – – 0.029** 0.055** – – 0.030** (0.016) (0.012) (0.016) (0.01) VAT × TradeGap0 – −0.027*** – −0.030*** – −0.054*** – −0.057*** (0.0043) (0.0041) (0.01) (0.01) VAT× τ s – – 0.0013** 0.0014** – – 0.00098** 0.0014** (0.0004) (0.00042) 0.00033) (0.00034) τs 0.027*** 0.029*** 0.023*** 0.020*** 0.027*** 0.028*** 0.023*** 0.019*** (0.0025) (0.0029) (0.0021) (0.0026) (0.0038) (0.0053) (0.0033) (0.0037) τs − τe −0.015*** −0.017*** −0.016*** −0.019*** −0.015** −0.015* −0.016** −0.016** (0.0023) (0.0025) (0.0022) (0.002) (0.005) (0.0068) (0.0049) (0.0053) Observations 30,661 22,230 30,661 22,230 15,328 14,694 15,328 14,694 2 R 0.57 0.6 0.57 0.6 0.55 0.57 0.55 0.57 Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The table shows coefficient estimates and standard errors from 10 regressions. All regressions include τ s (statutory tariff rate), τ s − τ e (tariff exemption), and HS6 and year fixed effects. In columns (2–4) and (6–8), the average initial trade gap is controlled, and observations in those years (2005–2006) are excluded. Standard errors are corrected for two-way clustering at HS4-year level and are reported in parentheses below coefficients. *, **, and *** respectively show significance at 10, 5, and 1 percent level. 6.2. Robustness Regressions In order to check the robustness of the results we conduct three tests. First, during 2012 and 2013 the tightening of the economic sanctions resulted in a recession and a significant exchange rate depreciation. Year fixed effects would absorb the effect of sanctions as far as they are uniform across VATable and non-VATable products. However, it might be that luxury products, which are VATable, are hit harder by the UN sanctions in 2012. To see the sensitivity of our results to this event, we exclude 2012 and 2013 from the sample and reestimate the coefficients. Columns (1) to (4) in table 5 show the results for this sample restriction. Coefficient estimates and their significance show small changes. Second, as discussed in the context, the VAT law implementation was phased in through the issuance of six notices by the Iran National Tax Administration. These notices applied to domestic firms; notably, the importers were required to register for VAT from the beginning. Even though this is unlikely to affect imports, it might weaken the chain effects. To address this issue, we exclude 2008 to 2014 (phase-in years) and estimate coefficients with the remaining years in columns (5) to (8) of table 5. This is a stringent test as we are left with five years. However, the new results are sometimes stronger and patterns do not change. The heterogeneity of the results remains the same as well; products with a shorter value chain and an initially smaller trade gap receive a smaller impact from VAT. Third, the trade gap might be mean reverting in the sense that when the trade gap is high it tends to follow a declining trend. This threat is particularly important for us because the VAT rate increases almost linearly after 2008. In other words, any differential trends in the trade gap might be confounded with the effect of the VAT rate increases, because the variation in the latter is simply uniform over the years. We conduct two robustness checks to alleviate mean reversion concerns. First, we drop observations with outlier trade gaps which are more likely to be far from the mean. The results from this regression are shown in column (1) of table 6. Coefficient estimates and their significance are unchanged.11 Sec- ond, we allow the trade gap to have a heterogeneous linear trend with respect to the initial trade gap. 11 The patterns of results are also robust to stricter exclusion of extreme observations. The World Bank Economic Review 345 Table 6. Robustness of the Impact of VAT on the Trade Gap Gap ∈ [p1, p99] Mean reversion Dependant variable: trade gap (1) (2) (3) (4) VAT rate −0.018 −0.026* −0.026* −0.026* Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 (0.012) (0.014) (0.014) (0.014) VAT × 1[cons] 0.069*** 0.075*** 0.075*** 0.075*** (0.017) (0.02) (0.02) (0.02) VAT × TradeGap0 −0.027*** −0.004 −0.004 −0.0044 (0.0045) (0.0071) (0.0072) (0.0068) τs 0.022*** 0.024*** 0.024*** 0.024*** (0.0025) (0.0025) (0.0025) (0.0025) τs − τe −0.016*** −0.016*** −0.016*** −0.016*** (0.0029) (0.0032) (0.0032) (0.0031) Linear trend × TradeGap0 Y Y Y Quadratic trend × TradeGap0 Y Y Cubic trend × TradeGap0 Y Observations 27,294 27,727 27,727 27,727 2 R 0.59 0.6 0.6 0.6 Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: The table shows coefficient estimates and standard errors from four regressions. All regressions include τ s (statutory tariff rate), τ s − τ e (tariff exemption), and HS6 and year fixed effects. In all regressions the average initial trade gap is controlled, and observations in those years (2005–2006) are excluded. Standard errors are corrected for two-way clustering at HS4-year level and are reported in parentheses below the coefficients. *, **, and *** respectively show significance at 10, 5, and 1 percent levels. Column (2) shows the coefficient estimates from this specification. Coefficient estimates for the VAT rate and its interaction with the consumer dummy are similar to the main results. However, the coeffi- cient estimate for the interaction of VAT and the initial trade gap changes sign and become smaller. This change is due to the fact that we do not have a rich variation in the VAT rate. Therefore, it is virtually impossible to interpret this coefficient estimate once we include the interaction of the linear trend with the initial trade gap. The two other coefficients are, however, robust even to quadratic and cubic trends (columns 3 and 4). We also conduct three placebo regressions to further validate the results. First, remembering that 56 percent of observations in the data set are VATable, here we randomly assign a mock VAT dummy to 56 percent of the HS6 products irrespective of whether they are actually VATable or not. Estimating model (1) indicates that the impact of the mock VAT dummy is 0.006 (standard deviation 0.017). Second, we restrict the sample to non-VATable HS6 codes and randomly assign a mock VAT rate to 56 percent of them. The assignment is completely random but we set the mock VAT rate equal to the actual VAT rate in each year. Again, we estimate model (1) and find that a 1 percentage point change in the mock VAT rate is associated with an insignificant 0.7 reduction in trade gap. The third placebo experiment in- cludes keeping VATable products, with their actual VAT rate, and randomly assigning a zero VAT rate to 44 percent of them, remembering that 44 percent of observations in the real data are non-VATable prod- ucts. Again, we observe a very small and insignificant coefficient. The estimated impact based on model (1) is equal to 0.00018. These three tests show that our results are not based on random differences between or within VATable and non-VATable products. 7. Discussion This study investigates the impact of VAT on illegal imports measured through the trade gap variable, the difference between reported export and reported imports. However, there might be concerns that VAT 346 Yousefi and Vesal Table 7. Regression Results for the Impact of VAT Rate on Trade Flow ln(exports) ln(imports) Trade gap Dependant variable (1) (2) (3) VAT −0.054*** −0.034*** −0.020** (0.012) (0.008) (0.0099) Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 τs 0.0022 −0.024*** 0.026*** (0.0026) (0.0018) (0.0027) τs − τe 0.0021 0.018*** −0.016*** (0.0023) (0.002) (0.0032) Observations 36,886 36,886 36,886 2 R 0.8 0.79 0.57 Source: Authors’ analysis based on data from the Iran Customs Administration and the World Integrated Trade Solution. Note: This table shows coefficient estimates from three separate regressions. Column (1) reports coefficient estimates from a regression of log of exports reported by Iran trade partners on covariates. Column (2) reports coefficient estimates from a regression of log of imports to Iran reported by the Iran Customs Administration. Column (3) reproduces results from column (4) of table 3 for comparison. affects the trade flows. While it is theoretically documented that a comprehensive well-administered VAT would not affect trade flows (Feldstein and Krugman 1990), the real-world VAT system has differences from such baselines. Specifically, in Iran, a large share of products are exempt from VAT, which could create trade distortions. In order to decompose our trade gap effect and relate to the well-established literature on the impact of VAT on trade flows, we estimate two regressions in table 7. In this table we regress the reported exports to Iran by exporters (column 1), the reported imports by Iran (column 2), and the trade gap (column 3) on covariates. Based on the estimation results in column (1), exports decrease by 5.4 percent as a result of a 1 percentage point increase in the VAT rate. Reported exports by Iran trade partners are seen as real imports to Iran, because there is no incentive on the exporters’ side to misreport. At the same time, the reported imports by Iran decrease by 3.4 percent (column 2), which is smaller than the decline in reported exports. This means that the trade gap, our measure of illegal imports, declines by 2 percent (column 3). In other words, the VAT system in Iran affected the real flow of imports and the reporting of imports. The fact that reported imports decline less compared to real imports is evidence on lower trade evasion. Thus, in contrast to Benzarti and Tazhitdinova (2021), but consistent with Sharma (2020), we find evidence on the sensitivity of real imports to VAT in Iran. This contrast is probably due to a close-to-ideal VAT system in the sample of Benzarti and Tazhitdinova (2021) and a different VAT system in Iran. Nevertheless, our findings show that even a less-ideal VAT system brings in revenue gains by improving the overall enforcement of tariffs. Our investigations may require more follow-up studies. For example, there are concerns that VAT not only alters the intensive margin of trade but also the extensive margin. If the importation of some goods stops after imposing VAT, then our results reflects the sum of extensive and intensive margins. In contrast, if there are some products that are replaced by others, it might be worth distinguishing the two effects. Regarding full identification, our study faces two limitations. First, the variation in the VAT rate is close to linear. Therefore, we do not have enough statistical power to control for product-level differential trends. Had the VAT rate been reduced or increased dramatically in our context, we would have had stronger identification to support our findings. While we have conducted several robustness tests and showed similar pre-trends, our results might still suffer from non-VAT-related differential trends that are specific to VAT products. Second, we do not observe tax revenue (for VAT and tariffs). Therefore, we cannot show the impacts on the final variable of interest. Here, aggregate figures support the idea that the tax revenue has gone up after VAT introduction, but more evidence is required for full identification. Besides the microeconomic aspects of the mechanisms at work, the macroeconomic condition of a country has a firsthand effect on trade pattern. While we control for year dummies in our regressions, Iran’s The World Bank Economic Review 347 macroeconomy may have heterogeneous impact on the pattern of traded goods. This is another reason why studying extensive margins of trade matters as a future study. During the course of our research, Iran experienced a severe set of international sanctions, starting from 2007, heavily pronounced by the UN multilateral sanctions in January 2012, followed by the stop of the SWIFT service to Iran in March 2012. These sets of sanctions hit the economy hard and limited oil export revenue.12 In 2015, the Joint Comprehensive Plan of Action (JCPOA) was signed between Iran and the P5 + 1. However, President Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 Trump withdrew from the JCPOA in May 2018 and reinstated the previous sanctions. This caused a sharp currency depreciation, followed by a surge in inflation and stifled nascent economic growth. As a response to these situations, and to manage the demand for foreign goods, in 2012 Iran set 10 different levels of priorities in order to allocate foreign exchange to importation, with the first priority devoted to essential goods (such as cereals), and the 10th to the least essential ones (such as luxury sport tools). This system was revised several times, but the spirit was to limit the importation of so-called inessential goods and to manage the exchange rate crisis (Garshasbi 2014). In 2017 and 2018, when the US unilateral withdrawal from JCPOA worsened the economic situation in Iran, the government introduced a dual exchange rate system, with the lower amount allocated to importation of essential goods and the higher rate to the others. Notably, the government used the lower exchange rate (called the reference rate; in Farsi, Nerkh-e Marja’) to calculate the import tariffs, VAT, and any other customs non-tariff payments. This causes a bias when comparing the tariff and VAT rates after 2017 with the years before, which is obviously not the case in the current study, but needs to be addressed in future research if the data set is extended beyond 2017. 8. Conclusions In this paper we identify a novel interaction between VAT and tariffs. We find that the introduction of VAT significantly reduces trade evasion. In our preferred specification, a 1 percentage point increase in the VAT rate reduces the trade gap by about 2 percent. Three points are worth emphasizing. First, the introduction of VAT has complementary effects on other tax bases. In our context, the backward linkages activated by VAT incentivize importers to truthfully reveal their imports. The fact that consumer products with a shorter value chain are less affected by the introduction of VAT further confirms this hypothesis. This channel could be extended to other types of taxes, including corporate tax. VATable firms would require VAT invoices for purchases and sales. These receipts are available from other trading partners as well (paper trail). Therefore, it becomes harder to evade corporate tax. Such inter-tax-base mechanisms are yet to be studied in the empirical literature. Second, the magnitude of the VAT effect on the trade gap is as large as that of the tariff rate. The trade evasion literature has largely focused on tariffs (Fisman and Wei 2004) or other product characteristics (Javorcik and Narciso 2008; Mishra, Subramanian, and Topalova 2008). Our finding shows that the overall enforcement environment, including the existence of efficient tax instruments such as VAT, could create large spillovers. Third, we add support for a joint tariff VAT reform by shedding light on the complementarity between VAT and tariffs, which has not been studied in the literature (Baunsgaard and Keen 2010; Emran and Stiglitz 2007, 2005; Keen 2008; Keen and Ligthart 2002;). Our findings suggest a double dividend from replacing tariffs by VAT. Tax revenue increases by reduced imports misreporting and by a broader VAT base. VAT is an effective tax instrument because it creates opposite evasion incentives for the two sides of a transaction, leaves a paper trail, and withholds taxes at the upstream (Keen and Lockwood 2010; Waseem 2022). Such properties are operative at the point of imports as well, because this is the first point that VAT 12 Many studies try to measure the impact of sanctions on Iran’s economy. For example, see Esfahani and Yousefi (2017), Farzanegan, Khabbazan, and Sadeghi (2016), Madanizadeh, Karimirad, and Rahmati (2019), Rahmati, Ebrahimian, and Madanizadeh (2021), and. 348 Yousefi and Vesal is levied for production chains that rely on imported commodities (Keen 2008). Therefore, VAT creates an incentive for honest reporting of imports. In the presence of VAT, import tax evasion affects both importers (risk of getting caught) and downstream firms. If downstream firms have some exposure to the formal sector, e.g. through VAT registration or corporate taxes, they would need valid VAT receipts. This backward pressure creates incentives for honest reporting of imports. Theoretically speaking, VAT may push firms into either formal or informal sectors, depending on how large the informal sector is (De Paula Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 and Scheinkman 2010, 2011). The larger the informal sector, the higher the likelihood of pushing firms into the informal sector. However, many empirical studies suggest that the overall effect is in favor of the formal sector (Pomeranz 2015; Carrillo, Pomeranz, and Singhal 2017). Our findings add support to the effectiveness of VAT in a middle-income country with pervasive tax evasion. Data Availability The replication data for this article is available the journal’s website. References Baunsgaard, T., and M. Keen. 2010. “Tax Revenue and (or?) Trade Liberalization.” Journal of Public Economics 94(9–10): 563–77. Benzarti, Y., and A. Tazhitdinova. 2021. “Do Value-Added Taxes Affect International Trade Flows? Evidence from 30 Years of Tax Reforms.” American Economic Journal: Economic Policy 13(4): 469–89. Carrillo, P., D. Pomeranz, and M. Singhal. 2017. “Dodging the Taxman: Firm Misreporting and Limits to Tax Enforcement.” American Economic Journal: Applied Economics 9(2): 144–64. Davies, R. B., and L. S. Paz. 2011. “Tariffs versus Vat in the Presence of Heterogeneous Firms and an Informal Sector.” International Tax and Public Finance 18(5): 533–54. De Paula, A., and J. A. Scheinkman. 2010. “Value-Added Taxes, Chain Effects, and Informality.” American Economic Journal: Macroeconomics 2(4): 195–221. —— . 2011. “The Informal Sector: An Equilibrium Model and Some Empirical Evidence from Brazil.” Review of Income and Wealth 57(S1): S8–S26. Emran, M. S., and J. E. Stiglitz. 2005. “On Selective Indirect Tax Reform in Developing Countries.” Journal of Public Economics 89(4): 599–623. —— . 2007. “Equity and Efficiency in Tax Reform in Developing Countries.” Available at SSRN 1001269. Esfahani, H. S., and K. Yousefi. 2017. “Rash Credit Injection, Hasty Job Creation, and Firm Bifurcation in Iran’s Manufacturing.” Presented at International Iran Economic Association 2018, available from https://www.econ.cam.ac.uk/people-files/faculty/km418/IIEA/IIEA_2018_Conference/Papers/Yousefi_Rash%20 Credit%20Injection,%20Hasty%20Job%20Creation,%20and%20Firm%20Bifurcation%20in%20Iran%E2% 80%99s%20Manufacturing.pdf. Farzanegan, M. R., M. M. Khabbazan, and H. Sadeghi. 2016. “Effects of Oil Sanctions on Iran’s Economy and House- hold Welfare: New Evidence from a CGE Model.” In Economic Welfare and Inequality in Iran, 185–211. New York: Palgrave Macmillan. Feldstein, M., and P. Krugman. 1990. “International Trade Effects of Value-Added Taxation.” Chapter 7. In A. Razin and J. Slemrod, eds. Taxation in the Global Economy, University of Chicago Press. Fisman, R., and S.-J. Wei. 2004. “Tax Rates and Tax Evasion: Evidence from Missing Imports in China.” Journal of Political Economy 112(2): 471–96. Garshasbi, A. 2014. “Managing the Importation during Sanctions by Prioritization.” Barresi-haye Bazargani(Farsi) 12(6): 1–12. Goolsbee, A. 2004. “The Impact of the Corporate Income Tax: Evidence from State Organizational Form Data.” Journal of Public Economics 88(11): 2283–99. Hoseini, M., and O. Briand. 2020. “Production Efficiency and Self-Enforcement in Value-Added Tax: Evidence from State-Level Reform in India.” Journal of Development Economics 144: 102462. The World Bank Economic Review 349 Javorcik, B. S., and G. Narciso. 2008. “Differentiated Products and Evasion of Import Tariffs.” Journal of International Economics 76(2): 208–22. Keen, M. 2008. “VAT, Tariffs, and Withholding: Border Taxes and Informality in Developing Countries.” Journal of Public Economics 92(10–11): 1892–906. Keen, M., and J. E. Ligthart. 2002. “Coordinating Tariff Reduction and Domestic Tax Reform.” Journal of Interna- tional Economics 56(2): 489–507. Downloaded from https://academic.oup.com/wber/article/37/2/331/7028408 by Joint Bank-Fund library user on 04 September 2023 Keen, M., and B. Lockwood. 2010. “The Value Added Tax: Its Causes and Consequences.” Journal of Development Economics 92(2): 138–51. Li, J., and X. Wang. 2020. “Does VAT Have Higher Tax Compliance than a Turnover Tax? Evidence from China.” International Tax and Public Finance 27(2): 280–311. Li, L., K. Z. Liu, Z. Nie, and T. Xi. 2021. “Evading by Any Means? VAT Enforcement and Payroll Tax Evasion in China.” Journal of Economic Behavior & Organization 185: 770–84. López-Luzuriaga, A., and C. Scartascini. 2019. “Compliance Spillovers across Taxes: The Role of Penalties and Detection.” Journal of Economic Behavior & Organization 164: 518–34. Madanizadeh, S. A., A. Karimirad, and M. H. Rahmati. 2019. “Business Cycle Accounting of Trade Barriers in a Small Open Economy.” Quarterly Review of Economics and Finance 71: 67–78. Marion, J., and E. Muehlegger. 2018. “Tax Compliance and Fiscal Externalities: Evidence from US Diesel Taxation.” Journal of Public Economics 160(4): 1–13. Mishra, P., A. Subramanian, and P. Topalova. 2008. “Tariffs, Enforcement, and Customs Evasion: Evidence from India.” Journal of Public Economics 92(10–11): 1907–25. Pomeranz, D. 2015. “No Taxation without Information: Deterrence and Self-Enforcement in the Value Added Tax.” American Economic Review 105(8): 2539–69. Rahmati, M. H., M. Ebrahimian, and S. A. Madanizadeh. 2021. “Dynamic Effects of Trade Barriers with Speculation on Foreign Currency: The Case of Iran.” International Economics 167(3): 151–73. Sharma, R.R.. 2020. “Does the VAT Tax Exports?” Economic Inquiry 58(1): 225–40. Waseem, M. 2018. “Taxes, Informality and Income Shifting: Evidence from a Recent Pakistani Tax Reform.” Journal of Public Economics 157(1): 41–77. —— . 2023. “Overclaimed Refunds, Undeclared Sales, and Invoice Mills: Nature and Extent of Noncompliance in a Value-Added Tax.” Journal of Public Economics 218: 104783. —— . 2022. “The Role of Withholding in the Self-Enforcement of a Value-Added Tax: Evidence from Pakistan.” Review of Economics and Statistics 104(2): 336–54. Yousefi, K., M. Vesal, and H. Pilvar. 2020. “Import Tax Evasion and Avoidance: Evidence from Iran.” Quarterly Review of Economics and Finance 75: 31–39.