Explaining Customs Tax Evasion in Ethiopia: The Effect of Trade Tax,Law Enforcement,and Product Characteristics

Abstract

Low-income African countries face significant customs tax evasion due to undervaluation and smuggling of imports by registered traders. This study uses trade policy, law enforcement, product misclassification, and product differentiation to examine customs tax evasion in Ethiopian imports from Kenya. Using “missing imports” data disaggregated at the harmonized system 6-digit level, the econometric model employs the trade gap as a measure of evasion. This study seeks to explain two types of evasion: undervaluation-based evasion, and smuggling- and product misclassification–based evasion. According to the econometric estimates, 1% increase in the taxes imposed on imports increases evasion by 1.12% for undervaluation-based evasion and 2% for entirely “missing imports”. Based on a quantitative measure of law enforcement, the findings show that expecting higher fee as a penalty for tax evasion is negatively associated with both forms of evasions. Regarding the extent of evasion, customs tax evasion is significantly higher for differentiated items than for homogeneous goods. Furthermore, mislabeling products from higher to lower tariff categories explains a large portion of evasion in goods that are either entirely smuggled or misclassified. The findings validate the importance of reducing evasion by judiciously lowering tariffs and enforcing the law at the border.

Keywords

Tax rates customs tax evasion law enforcement smuggling product differentiation Ethiopia

Introduction

In many developing countries, illegal transactions account for a significant portion of foreign trade (Pitt, 1981). This is particularly pervasive in Africa, where cross-border trade among African countries is mainly unofficial and informally conducted across borders. As a result, much of the inter-regional trade in Africa is unrecorded and takes the form of smuggling (Bensassi et al., 2019; Golub & Mbaye, 2009). Studies confirm that informal trade in Africa is sizeable and volatile (Bouet et al., 2018). Given that many low-income countries rely heavily on revenue from border taxes, smuggling in intra-African trade results in the loss of tax revenue for African states. However, because of the lack of tax collection capabilities, and as the border taxes are a relatively easy source of revenue, taxes on international trade accounts for a sizeable part of government revenue (Jean & Mitaritonna, 2010; van Dunem & Arndt, 2009). For instance, trade taxes continue to account for between 30% and 50% of the total state revenue in Africa (Cantens & Raballand, 2017). Import tax evasion may weaken the potential for economic growth and development in the long run through its impact on investment in the formal sector, public revenues, and food safety (Bouet et al., 2018).

This study analyzes customs tax evasion and its responsiveness to tax rates, the strength of law enforcement, product differentiation, and product misclassification in the case of Ethiopia, a representative low-income African country. Considering the importance of border taxes for public revenue and ramifications of the loss of trade tax revenue for development financing, it is vital to understand the major drivers of import tax evasion in low-income African countries. Ethiopia is an interesting case study for examining customs tax evasion for two crucial reasons. First, besides undervaluation and misclassification of imports at official border crossings, considerable illegal imports (i.e., smuggling) occur along Ethiopia’s porous territorial borders shared with neighboring countries. For instance, imports from Kenya alone were undervalued by over 65% in 2017, suggesting a greater degree of tax evasion in imports from the neighboring countries. Second, understanding border tax evasion is essential for a country where trade taxes contribute to 28% of total tax revenues (Harris & Seid, 2021). The issue of import tax evasion is particularly vital for a low-income country with a narrow tax base yet aspiring to be a middle-income country by 2030. This study analyzes the major explanations for the significant discrepancies in cross-border trade between Ethiopia and its neighboring trade partners using mirror statistics.

Theoretically, customs tax evasion is primarily associated with the prevalence of trade barriers at the border, notably higher import tax rates (Bensassi et al., 2019; Fisman & Wei, 2004; Mengistu et al., 2022). In particular, to reduce the tax burden stemming from higher customs duties and tax rates, traders resort to informal cross-border trade (ICBT; Mitaritonna et al., 2017) and evade border taxes through undervaluation, smuggling, and product misclassification. This study is mainly related to the literature linking smuggling and customs tax evasion with trade policy in developing countries using mirror trade data. Bhagwati (1964) pioneered this field of study by arguing that tariff evasion is practiced through undervaluation of goods. Following Bhagwati’s (1964) work, mirror trade data, that is, relying on data from the exporting country to identifying missing corresponding data in the importer’s records, are often used to assess the extent of smuggling and evasion (Fisman & Wei, 2004; Levin & Widell, 2007, 2014; Mengistu et al., 2022; Mishra et al., 2008; van Dunem & Arndt, 2009). Literature mainly analyzes the responsiveness of evasion in the form of undervaluation at customs to tariff rates. Despite the variation in the magnitude of elasticity of evasion to tariff rates, studies find that higher tariff rates are positively associated with import tax evasion (Bouet & Roy, 2012, for Kenya, Mauritius, and Nigeria; Fisman & Wei, 2004, for China; Levin & Widell, 2007, for Kenya and Tanzania; Mengistu et al., 2022, for Ethiopia; Mishra et al., 2008, for India). These studies measure evasion using trade gap (the discrepancy between a country’s reported imports and the corresponding exports reported by its trading partners) and explain deliberate undervaluation or product misclassification at the border to reduce the tariff burden stemming from higher custom duty rates.

Aside from policy-related trade barriers, this study is also linked to a handful of studies documenting the role of law enforcement and the quality of governance in explaining evasion. Literature suggests that the extent of smuggling is greater in those countries where there is weak law enforcement (Lesser & Moisé-Leeman, 2009). Some studies that explored the role of law enforcement in evasion find that the presence of strict law enforcement and governance discourages smuggling and evasion (Javorcik & Narciso, 2008; Miskam et al., 2013).

This study is also broadly linked to the literature studying the responsiveness of evasion to product characteristics, where evasion is reported to be higher for differentiated products relative to homogenous goods. Along this line, Javorcik and Narciso (2008), using trade data between Germany and 10 Eastern European countries, find that a 1% increase in the tariff rate is associated with a 1.7% increase in evasion for differentiated products and a 0.4% increase in evasion in the case of homogenous goods. Likewise, Mishra et al. (2008) using different proxies for product differentiation, also find that differentiated products exhibit a higher evasion elasticity than homogenous products in India.

The practice of ICBT on a larger scale leads to considerable customs tax evasion and thus erodes the tax collection capacity of the African states. Notwithstanding the significant loss of trade tax revenue, only recently has there been an empirical attempt to study the responsiveness of evasion to major contributory factors in Africa (Bensassi et al., 2019; Bouet et al., 2018; Levin & Widell, 2014; Mengistu et al., 2022; van Dunem & Arndt, 2009). Besides the limited empirical evidence on border tax evasion and the underlying drivers in low-income African countries, some significant gaps in the extant literature need to be addressed. Despite the pervasiveness of ICBT between Ethiopia and neighboring partners, there is, to my knowledge, limited evidence for the extent and drivers of evasion in Ethiopia’s imports from its neighboring trade partners.

Though there are different forms of evasion (for instance, evasion at customs entry and evasion outside official customs checkpoints), previous studies mainly focused on analyzing one form of smuggling or evasion. Most studies examined evasion practiced at customs entry using mirror statistics (Fisman & Wei, 2004; Javorcik & Narciso, 2008; Levin & Widell, 2014; Mengistu et al., 2022). However, other studies examined tax evasion for goods bypassing official border checkpoints using small-scale survey data collected over a short period (Bensassi et al., 2019; Mitaritonna et al., 2017). But, given the illicit nature of the business, survey data capture only a tiny fraction of smuggling. Thus, both categories of studies fail to consider entirely “missing imports”—goods reported by the exporting side but recorded with zero value on the importer’s side. In short, the extant studies that examine only one form of evasion offer an incomplete picture of customs tax evasion. Given the scant literature, it is vital to provide a comprehensive analysis of evasion by considering both forms of evasion practiced by registered or licensed importers. For this purpose, official data of the trade partners may be used to examine the evasion due to undervaluation, smuggling, and mislabeling of imports.

Most importantly, the focus of the extant studies appears to be mainly on the role of trade policy and product characteristics in explaining evasion (Bensassi et al., 2019; Fisman & Wei, 2004; Mengistu et al., 2022). They fail to consider the theoretical tax evasion models in identifying the drivers of evasion. In this regard, the role of institutional capacity and the quality of law enforcement at the border, such as strictly punishing the tax evaders, is overlooked. Theoretically, however, it is indicated that the quality of law enforcement by customs authority determines the extent to which taxpayers intend to evade tax. In this regard, Allingham and Sandmo’s (1972) theoretical model of tax evasion, perhaps based on a risk-averse taxpayer, and a study by Shimeles et al. (2017) predict that a higher penalty rate and fines discourage tax evasion. Thus far, very few studies, such as Javorcik and Narciso (2008) and Miskam et al. (2013), have examined the empirical significance of the quality of law enforcement in explaining evasion. Studies that incorporate a quantified measure of the extent of law enforcement in Africa are particularly scant. It seems that very little is known about the effect of law enforcement on import tax evasion. Against these backdrop, this study empirically analyzes the explanations for import tax evasion for Ethiopia using “missing imports” disaggregated at the harmonized system (HS) 6-digit product level.

This study addresses the gaps mentioned above and offers three significant contributions to the literature. First, few studies have examined customs tax evasion in Ethiopia using “missing imports” from the neighboring countries. Except for a recent study by Mengistu et al. (2022) that relies entirely on Ethiopia’s imports from major non-African countries, there are, to my knowledge, limited studies that make use of the data from Ethiopia’s major neighboring trade partners. This study contributes to the literature by explaining import tax evasion among contiguous countries in Africa using Ethiopia’s imports from Kenya. In contrast to imports from non-African sources and goods from developed countries, imports from neighboring African countries are more likely to be undervalued at the border. Additionally, the probability of smuggling is higher from sharing a common border (Jean & Mitaritonna, 2010), resulting in more significant evasion. As Ethiopia is characterized by widespread smuggling in cross-border trade, understanding the underlying reasons for customs tax evasion is indispensable for informed policymaking.

Second, aside from offering new evidence from Ethiopia, the study contributes to the debate on customs tax evasion by examining the drivers of two distinct forms of customs tax evasion using “missing imports” data. In particular, the analysis considers tax evasion both at customs entry points (i.e., partial evasion) and evasion outside customs checkpoints. For this purpose, products crossing the official border checkpoints that may evade part of their tax burden and those bypassing official customs checkpoints (i.e., entirely “missing imports”) by registered traders are considered. Thus, it sheds light on the explanations for evasion using trade gaps in the mirror data of partners for two categories of border tax evasion. The first one takes the form of undervaluation, a terminology used by Lesser and Moisé-Leeman (2009), which is ICBT category C, where traders can partially evade tax. The second is ICBT category B goods, completely smuggled and/or misclassified goods, where traders may entirely or partially avoid border taxes. Hence, evasion through the channel of undervaluation and smuggling and/or product misclassification is separately analyzed. Since cross-border trade among neighboring African countries is mainly characterized by widespread evasion both at official customs entry points and outside customs points, using Ethiopian data is essential to explain evasion behavior in a greater depth. Moreover, the magnitude of the trade gap and its responsiveness to key underlying drivers depend on the type of evasion practiced by traders.

Third, and importantly, its major contribution lies in incorporating a quantitative measure of customs penalties which is expected to be levied on importers for non-compliance with customs law and regulations. This study notably differs from prior studies because it considers an overlooked yet a crucial contributory factor of evasion—the role of customs law enforcement. To bridge this gap, the study introduces a quantitative measure of the strength of law enforcement at the border in the econometric model. In particular, the monetary fine expected to be faced by tax evaders is used to capture the effect of law enforcement at the border in explaining evasion behavior. For simplicity, it is assumed that there is a high probability of detecting evasion by customs officers, and punishment is applied to evaders as per the law. Such penalties, if enforced practically, are expected to increase the costs of evasion and affect the decision to evade border taxes. Thus, the estimation is mainly motivated to find out the actual responsiveness of the trade gap if there is a high probability of detecting evasion and importers expect to face additional costs in the form of monetary fine for non-compliance with customs regulations.

In short, this study adds new evidence to the body of literature on the combined effects of tax rates, penalties for evasion, product heterogeneity (i.e., product differentiation), and product misclassification on both forms of import tax evasion in Ethiopia.

The study draws important insights from the empirical results. The analysis reveals that import tax evasion increases with increased tax rate. The trade gap is significantly higher for differentiated products relative to homogenous goods, suggesting that the intention to evade tax also depends on the ease of misrepresenting the true value of imports at the customs checkpoint. Moreover, when traders expect a higher penalty or monetary fine for non-compliance with the law, the proportion of actual imports declared tends to increase, suggesting that appropriate enforcement of the law could have resulted in minor evasions. Interestingly, mislabeling products from higher to lower tariff categories appears to significantly explain the trade gap only for those goods considered to be either completely smuggled or misclassified. This suggests that the underlying drivers of different forms of evasion (i.e., evasion through undervaluation, smuggling, and/or product misclassification) can be distinct. Hence, estimating a separate model for each type of evasion is appropriate. The findings inform the role of trade policies, quality of law enforcement, nature of products, and product misclassification in determining evasion and smuggling.

The rest of the article is organized as follows. Section two presents the methodology of the study. Section three describes the data and descriptive statistics of the variables used in the model. Section four reports the results, discusses the findings, and highlights the study’s policy implications. The final section concludes the article.

Methodology

This study uses an econometric model to examine the discrepancies between the reported export value on the Kenyan side and the corresponding mirror data on the Ethiopian side. Several factors can lead to discrepancies in mirror statistics (Yeats, 1995). The causes of discrepancies in mirror statistics are broadly categorized into three: unavoidable factors, structural reasons, and deliberate misreporting by traders. Unavoidable factors are related to discrepancies arising mainly from transportation costs. On the other hand, structural factors refer to structural differences between two customs offices, such as differences in the coverage of goods recorded by the two customs offices, transit time/time lag in recording transactions, and exchange rate fluctuations (Federico & Tena, 1991; Hamanaka, 2011). Similarly, on the potential origins of trade gap causing factors, Carrère and Grigoriou (2015) isolated the deliberate errors from the structural or “logistic errors.” They pointed out that while the latter is more or less unavoidable or due to structural reasons, the former is motivated by illicit operations, from tax evasion to capital flight, resulting in deliberate misreporting or misclassification.

Trade gaps arising from deliberate errors are the focus of this study. Deliberate misreporting could significantly explain a part of the discrepancies in mirror statistics. This may be practiced through a false declaration of value or misclassification of the imported goods, notably to evade tariffs and taxes (Carrère & Grigoriou, 2015). Given that the ultimate goal of deliberate misreporting is customs tax evasion, a more significant export-import gap indicates greater evasion (Fisman & Wei, 2004). Taking a higher trade gap as a possible signal of one form of evasion will not bias the measure. Though structural factors contribute to discrepancies, they are not motivated by illicit operations. For this reason, such gaps do not signal smuggling or undervaluation. Moreover, aside from being more or less unavoidable, discrepancies caused by structural factors are considered insignificant in this study.¹

Measurement of Customs Tax Evasion: Trade Gap

Tax evasion is measured by the discrepancy in mirror trade data of partners. Mirror statistics indicate the appropriate ratio of cost-insurance-freight (CIF) imported value and FOB (free on board) exported value in a reasonable range from 1.05 to 1.1 (Hien & Viet Hung, 2017). In the literature, import–export discrepancies ranging from 5% to 10% are an acceptable gap. On the other hand, import–export gaps that fall outside this range are viewed as arising from deliberate misreporting, signaling some form of evasion. According to Golub (2015), if the value of exports declared by the source country exceeds the CIF value of imports reported in the destination country, smuggling is inferred.

As a measure of evasion, the trade gap is quantified using the disparities in the export value declared by Kenya and import value reported by Ethiopia across products at the HS 6-digit level of disaggregation. The econometric estimation focuses mainly on two forms of import tax evasion: evasion by undervaluation and evasion through smuggling and/or product misclassification. While evasion from undervaluation of imports is captured using the gap in products reported by both partners (i.e., matched products), evasion arising from smuggling, product misclassification, or both are captured using entirely “missing imports.” This is important since the nature of smuggling, the magnitude of tax evasion, and the determinants vary across both categories of evasion. Accordingly, the econometric model uses two separate dependent variables to analyze the underlying determinants of both forms of evasion.

The first measure is the trade gap in matched trade flows which is the difference in import and export values for goods reported by both partners. Following Fisman and Wei (2004), the dependent variable, the reporting gap in official trade statistics in value (Gap_value_k), is defined as follows:

G a p_{k} = \log (e x p o r t_{k}) - \log (i m p o r t_{k})

(1)

where import _k denotes Ethiopia’s imports of product k from Kenya as reported to Ethiopia’s customs, and export _k denotes the exports of good k from Kenya to Ethiopia as declared by Kenya at the HS 6-digit level.

This measure of trade gap indicates the value of disparity in product lines reported by both partners (matched products) at the 6-digit level. This measure of trade gap in value is used as a measure of partial import tax evasion. It is instrumental in capturing the extent of quasi-smuggling or undervaluation in monetary terms. This sample is restricted to 310 products appearing in both exporter and importer’s records in the analysis. The evasion in quantities is measured in the same way as the gap in value, albeit quantities of exports and imports are used instead of their monetary value.

The second measure of evasion considers entirely “missing imports”—goods declared by Kenya with non-zero values but unrecorded by Ethiopian customs officers. Mishra et al. (2008) make the extreme assumption of “complete smuggling” for India to capture such goods. Following this study, the second definition estimates evasion by the gap in the value of goods declared by the Kenyan side but unrecorded by Ethiopian customs agents. Thus, to account for entirely missing imports in the estimation, 1 is added to both exports and imports data as follows:

{Gap_Entirely Missing}_{k} = \log (1 + {export}_{k}) - \log (1 + {import}_{k}) .

(2)

Though entirely missing imports are taken as a measure of the size of smuggling in imports practiced by licensed traders bypassing official customs checkpoints as in Mishra et al. (2008), another possibility for such goods is product misclassification. Meaning that the products may have entered the country of destination under a different product classification—in which case misclassification may be deliberate in favor of a less heavily taxed product (Jean & Mitaritonna, 2010). Thus, the second measure may capture evasion due to smuggling or product misclassification, or both. If entirely missing imports are due to smuggling of goods, this will result in a complete evasion of customs tax. On the other hand, if missing rows arise from deliberate product mislabeling, the gap captures partial import tax evasion. The sample of entirely missing rows in Ethiopia’s imports contains 285 products.

Model Specification

This study attributes part of the observed discrepancies in the partner’s transactions to some form of illicit operations. Deliberate misreporting may take the form of undervaluation/overvaluation, complete smuggling, and/or product misclassification. Such illegal practices are often motivated by trade policy factors, the nature of products, and the strength of law enforcements among others. This study considers the key factors that could explain the trade gap arising from illicit operations. Though structural reasons are a valid source of error, such gaps do not occur from a deliberate act of economic agents targeting evasion. Importantly, as they do not help explain discrepancies arising from deliberate misreporting, they are not considered in the econometric analysis. Thus, the major variables explaining the discrepancies associated with deliberate misreporting are tax rates, product differentiation, a proxy for the extent of law enforcement, and product misclassification.

The Benchmark Model: Trade Gap and Tax Rates

The primary interest in the econometric analysis is to test whether the trade gap is increasing with the tariff rates due to evasion. In the benchmark model specification, the estimation focuses on whether the variation in trade gap across imported goods from Kenya at the 6-digit level is systematically related to tax applied across products. Using the evasion gap, the following model is used to estimate the effect of tax rates on the trade gap:

G a p_{k} = α + β_{1} t a x_{k} + ε_{k} .

(3)

Gap stands for evasion both in value and quantities for products reported by both partners (i.e., the gap in matched product level transactions). The trade gap in value and quantities is regressed on the same set of explanatory variables. Similarly, the baseline model for entirely “missing imports” in Ethiopia is specified as follows:

Gap_entirely missing g_{k} = α + β_{1} t a x_{k} + ε_{k} .

(4)

The disturbance term ε is assumed to be white noise. Tax evasion through undervaluation and complete smuggling occurs when export values are greater than import values. In both cases, the trade gap is greater than zero, albeit the magnitude is expected to be larger for goods assumed to be “completely smuggled” relative to undervalued ones. Besides smuggling, the other potential reason for completely missing imports is differences in how countries classify products. In the case of measurement errors, such honest misclassification of products should not be systematically related to tax rates (Mengistu et al., 2022). More importantly, if the missing imports are due to the trader’s deliberate misclassification of products from high to low tariff categories, such misclassification should be associated with the tax rates.

Regardless of whether imports with zero values represent complete smuggling or deliberately misclassified products, the analysis emphasizes a sample of goods for which the Kenyan side records exports but for which the Ethiopian customs record no import transaction. Notwithstanding the difficulty in clearly identifying entirely missing goods, the interest is whether this measure is systematically correlated with taxes. In other words, the exercise here mainly aims to test whether more “value” is missing or lost when taxes are higher.

Extended Empirical Model

Aside from product-specific tax applied on imports indicated in the benchmark specification, this study also examines whether the trade gap in products is systematically correlated with the strength of law enforcement, product differentiation, and product misclassification. Hence, the benchmark model is augmented with additional key regressors as follows:

G a p_{k} = α + β_{1} t a x_{k} + β_{2} P e n a l t y_{k} + β_{3} D_{k} + β_{4} M i s c l a s s_{k} + ε_{k} .

(5)

Where Gap is the value of disparity in trade flows, which is a measure of the evasion gap, tax is the actual tax applied on imports from Kenya at the 6-digit level; Penalty is a proxy for the strength of law enforcement or state capacity; D is the interaction between the dummy for product differentiation and tariff; Misclass is product misclassification, and ε is assumed to be white noise.

Description of Study Variables and Hypothesis

Tax: This is the tax paid at the border as a proportion of the import value of each product at the HS 6-digit level. The tax rate is the sum of tariff, VAT, excise, and surtax rates applied on imports at the HS 6-digit level. Since both Ethiopia and Kenya are members of COMESA, the analysis uses the COMESA preferential tariff rate.² While other factors may contribute to discrepancies, a positive association between trade gap and tax rates suggests the presence of import tax evasion. As customs duties and taxes increase across products, the trade gap also increases. Thus, if the prevalence of higher tax rates induces evasion, the coefficient of tax is expected to be signed positively.

Strength of law enforcement: The extent of import tax evasion depends on the quality of law enforcement (whether smuggling is detected by customs officers and importers are punished for evasion as per the law). The intention to evade depends on the probability of detection by customs agents, the magnitude of punishment for evasion (or penalty rate), and the agent’s risk aversion. The aim is thus to empirically examine if customs officers or authorities strictly apply the law on tax evaders (i.e., evasion is detected and importers are punished for bypassing the law), and what would be the response to evasion. To simplify the analysis of the trade gap effect of law enforcement, the approach uses two key assumptions. First, there is a higher probability of detection at the border, where importers declaring import values lower than the actual value would possibly be inspected by the customs officers. Second, and more importantly, customs officers will fine importers who understate the true value of the imports, smuggled and/or misclassified goods. Though the importers might offer the customs officers a bribe to overlook the understatement, it is assumed that the officer would decline the offer to avoid the risk of being caught by the customs administration or authorities (see Jean & Mitaritonna, 2010, for a theoretical and empirical analysis of the issue).

The study quantifies the effect of law enforcement on import tax evasion using the hypothetical penalty that could have been levied on importers when they fail to comply with the law or evade taxes. As a proxy for the quality of law enforcement, the ideal monetary fine or penalty (measured in thousands of dollars), expected to be levied by Ethiopia’s Customs Commission for tariff evasion at the border, is used. The Ethiopian customs proclamation suggests that any form of tax evasion is subject to punishment. In particular, the Customs Proclamation states that

Importing and exporting without paying duties or taxes, not correctly stating in a declaration, or paying understated duty or taxes will be subject to a fine equivalent to twice the amount of such duty and tax, on top of the settlement of the duty and tax payable. However, if the remaining duties and taxes payable are not more than 10 percent of the total duty and tax payable, the penalty will be waived for the importer. If concealed items account for more than 50% of the value of the goods described in the declaration, the matter will be criminally investigated under Article 169 of the Customs Proclamation. (Ethiopian Revenue and Customs Authority [ERCA], 2017, p. 131)

If the Customs Commission strictly implements the proclamation, the expected fine increases with the increase in the magnitude of evasion. The value of the desired fine as punishment for evasion is taken as the sum of the unpaid duty to be settled and twice the value of the actual duty and tax that should have been paid for those goods undervalued by more than 10% of their reported FOB value. However, for entirely “missing imports,” whether they capture smuggled or misclassified goods, the square of the duty and tax payable on the imports is used. The tax payable is squared because, besides imprisonment, the penalty charged against smuggling and misclassifying goods for contravening customs law and the confiscation of the goods by the Customs Commission makes the ideal monetary value of the penalty higher.³

Theoretically, since fining import tax evaders increases the cost of evasion, strict law enforcement at the border is expected to discourage tax evasion. Theoretical models of tax evasion by Shimeles et al. (2017) suggest that the effect of higher detection rates and fines on tax evasion is unambiguously negative. The empirics on the link between the quality of law enforcement and tax evasion also suggest that as enforcement quality enhances or penalty increases, tax evasion decreases (Javorcik & Narciso, 2008; Mishra et al., 2008; Miskam et al., 2013). Thus, if vigorous law enforcement discourages evasion, this variable is hypothesized to be signed negatively.

Product differentiation: Degree of differentiation/homogeneity is one of the intrinsic characteristics of products that may affect the ease of enforcement (Bouet & Roy, 2012). For homogenous products, the prices are widely known and quoted in exchanges, making compliance easier as underreporting or misclassification can be more easily detected from available information. However, differentiated products may lend themselves more readily to tariff evasion than homogeneous goods. Their price depends on many attributes, some of which may not be easily verifiable by a person unfamiliar with the product (Javorcik & Narciso, 2008). Following Javorcik and Narciso (2008) and Mishra et al. (2008), the applied tariff is interacted with the product differentiation dummy to estimate the effect of this product characteristic on evasion. The product differentiation dummy takes the value 1 if the product is differentiated according to Rauch’s (1999) classification and zero otherwise. The Rauch classification distinguishes goods by whether they are homogenous goods (whose prices are widely known or quoted in exchanges) or differentiated goods (whose prices are less well known and determined more by specific transactions). Since it may be easier for differentiated products to evade tariffs through undervaluation, the trade gap is expected to be more responsive to tariffs for differentiated goods relative to homogenous goods.

Product misclassification: Besides evasion through underreporting the value of imports, the trade gap may be explained by deliberate misclassification of imports from higher-tariff to lower-tariff categories. The importers may deliberately misclassify goods and report a higher taxed product as a lower one to reduce the tax burden. To investigate this channel of evasion, studies use the average tariff on other similar products in the same HS 4-digit category as an additional regressor (Fisman & Wei, 2004). Following the literature, I used the average tariff on other similar products in the same HS 4-digit category. If product misclassification exists, the coefficient of tariff on similar products is expected to sign negatively, which would signal that holding the own rate constant, a lower tariff on similar products creates more opportunities for misreporting (Bouet & Roy, 2012).

Both models are estimated using the ordinary least squares method. After estimating the benchmark model, stepwise regression is followed, where each explanatory variable is introduced as an additional regressor into the baseline model.

Data and Descriptive Statistics

Data

The analysis of import tax evasion relies on secondary data sources. Evasion can be inferred from examining officially recorded Ethiopia-Kenya bilateral trade statistics. I employed discrepancy in mirror trade statistics of partners to gauge tax evasion in imports using the HS-disaggregated product-level data. The HS classifies bilateral trade flows and tariff rates at different levels of product disaggregation. The study uses product-level mirror statistics and tariff and tax rates facing imports at the 6-digit level of disaggregation.

The primary data source for bilateral trade flows of partners used for the discrepancy in trade was World Bank’s World Integrated Trade Solutions (WITS), which is derived from the United Nation’s COMTRADE database. The database reports trade statistics at different levels of disaggregation. Other important data sources used on preferential tariff rates and tax rates were the WITS database and the Ethiopian Customs Commission (ECC). The WITS database reports information on Most Favored Nation (MFN) and preferential duty rates specific to the pairs of countries included in the sample at 6-digit HS product classification. On the other hand, the ECC database contains information on duty rates, VAT, surtax, and excise tax at the much disaggregated 8-digit HS product classification.

In the regression, tax rates are taken as the sum of preferential tariff rates applied, value-added tax, excise tax, and surtaxes. Data on preferential tariff rates specific to the pairs of countries obtained from WITS is combined with the other tax rates applied across products obtained from ECC. To combine the preferential tariff information and trade data available at 6-digit with the 8-digit HS tax data, the tax data at the 8-digit is converted into 6-digit level by taking the average within each 6-digit product category.

Regarding the sample of neighboring countries included in the analysis, the product level information on Somalia and Djibouti’s mirror exports to Ethiopia is not available in the WITS or in the COMTRADE database; therefore, they are excluded from the sample. Moreover, Sudanese goods have preferential access to the Ethiopian market with zero duty rates. Since they are imported duty-free, the data is of little importance to analyzing customs tax evasion. As a result, the tax evasion analysis emphasizes products imported from Kenya. Kenya provides a unique opportunity to examine both forms of customs tax evasion. It appears to be the most important import origin compared to the other neighboring partners regarding the number and diversity of products imported and smuggled into Ethiopia. The approach facilitates capturing evasion through the channels of undervaluation and product misclassification at customs and smuggling at unofficial land borders by formally registered or licensed traders.

Descriptive Statistics

The study uses data on imported goods and tariff rates applied at the 6-digit HS level of product disaggregation. Products and tariffs at the 6-digit level are taken for a single year (2017) and a single country (Kenya). Other scholars followed similar approaches; for instance, Fisman and Wei (2004) used the 6-digit HS for Chinese imports from Hong Kong for a single year, and van Dunem and Arndt (2009) applied similar techniques for Mozambique’s imports from South Africa.

The sample on traded products of partners originally contained 857 products from Kenya. In theory, Ethiopia’s reported imports from Kenya should match Kenya’s reported exports to Ethiopia. However, there were missing observations for 547 product lines imported from Kenya in 2017. In the trade transaction, a significant proportion of product lines (63.8%) was reported exclusively by either the exporter or importer only and found to be unmatched by the type of goods reported by the other partner. Goods reported either by the exporter only or the importer only that are unmatched with the record available on the other side of trade flow are considered either lost exports or orphan imports. The share of product lines with lost export and orphan imports account for 33.3% and 30.4% of the product lines reported by the partners, respectively. On the other hand, trade in only 310 product lines matched up, and 36.17% of the product lines were reported by both the exporting and importing sides at the 6-digit level. This suggests that there is a significant discrepancy both in the number as well as the value of goods reported by the partners. While some proportion of the disparity could be attributed to the undervaluation of imports, the remainder of the missing values is primarily due to unrecorded imports—representing either complete smuggling or product misclassification cases. As is the case for many African countries, the disparity indicates that trade between Ethiopia and countries in the Horn of Africa is mainly unofficial and unrecorded.

The analysis focuses on 310 matching trade flows, that is, goods reported by both partners and 285 goods reported by Kenya, with positive trade flows that went completely unrecorded by Ethiopia’s customs. The analysis explores the country pair trade disparity at the 6-digit product level. The summary statistics in Table 1 show that the mean trade gap is positive, indicating that reported exports are about 7% higher relative to the corresponding value of imports recorded by the Ethiopian customs. Here positive trade gap is not only a measure of evasion but also a proxy for the forms and magnitude of informality in imports. In other words, the mean positive trade gap suggests the prevalence of some degree of undervaluation (i.e., quasi-smuggling), smuggling, and misclassification resulting in import tax evasion.

Table 1.

Summary Statistics.

	Mean	Standard Deviation	Observation
Gap value	0.070	2.611	310
Log export value	8.308	3.283	310
Log import value	8.238	2.954	310
Gap_quantity	–0.119	2.540	310
Log export quantity	6.400	3.540	310
Log import quantity	6.519	3.21	310
Tariff rates	16.970	9.904	310
Tax rates	40.87	9.826	310
Gap_unit value	0.0612839	1.8466	310
Log export unit value	1.907882	1.861369	310
Log import unit value	1.846599	1.785101	310
Expected penalty (‘000 USD)	257.644	980.62	310
Gap value_entirely missing	6.12	2.47	285
Log export value_entirely missing	7.75	2.47	285
Tariff rates__entirely missing	14.4	10.10	285
Tax rate__entirely missing	37.51	15.3	285

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Notes: Gap value = log(export value) – log (import value) for matched products; Gap_ quantity = log (export quantity) – log (import quantity) for matched products; Gap value_entirely missing = log (1 + export value – log (1 + import value); Gap_unit value = Log export unit value-Log import unit value for matched products; Expected penalty is monetary fine applied on tax evaders, measured in thousands USD; Tariff rates_entirely missing and Tax rates_entirely missing stand for tariff and tax applied on entirely “missing imports”, respectively.

In addition, Table 1 shows that the mean per unit value gap is positive. It may appear that the trade gap, on average, is associated with underpricing of the unit value of imports. Undervaluation through underreporting the per-unit values of imports is one of the channels of tax evasion. On the other hand, the mean quantity gap is negative, suggesting that the quantities recorded by Ethiopia are higher than those recorded by Kenya. Such negative value seems to be consistent with the stylized fact that countries tend to monitor their imports more carefully than exports (Javorcik & Narciso, 2008). However, depending on the extent and nature of the false declaration of value (undervaluation, overvaluation), the mean trade gap in unit value and quantity change signs from positive to negative and vice versa for some classes of goods.

The magnitude of the trade gap may vary depending on the tax rates (lower-relative to higher-tax categories) and the nature of the products imported (differentiated versus homogeneous goods). For this purpose, the notion that the trade gap is larger for differentiated goods relative to homogenous goods, and higher-tax categories (those with tax rates above the median tax rate of 43%) relative to lower-tax ones (those with tax rates below the median tax rate of 43%) is assessed in Table 2. On average, the trade gap is positive for homogeneous goods while negative for differentiated goods. The evidence shows that evasion is greater among homogeneous goods than differentiated ones. However, in line with the theoretical expectation, the mean evasion gap for differentiated goods is higher when measured in unit values rather than quantities, suggesting that evasion mainly takes the form of underreporting of unit values. For differentiated goods, on average, the unit price of imports is underreported by 6%. Evasion occurs primarily through under-invoicing unit values, which is in line with the notion that concealing the actual value of imports is easier for differentiated goods relative to their homogenous counterparts. For homogeneous goods, on the other hand, the evasion gap is higher when measured in quantity rather than values, suggesting that evasion mainly takes the form of underreporting of quantities for such goods. It might be easier to underreport quantities for homogenous goods, mainly when imported in bulk, relative to concealing unit prices.

Table 2.

Mean Trade Gap Across Product Differentiation and Tax Categories for Matched Products.

	Nature of Product		Tax Category
	Differentiated	Homogenous	Higher Tax	Lower Tax
Trade gap	–0.061	0.27	0.061	0.065
Quantity gap	–0.17	0.286	–0.102	–0.0073
Unit value gap	0.060	–0.064	0.108	0.017
Observation	226	84	112	138

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Notes: Trade gap = log (export value) – log(import value); Quantity gap = log(export quantity) – log(import quantity); Unit value gap = Log (export unit value)-Log (import unit value).

In contrast to prior expectations, Table 2 shows that the mean trade gap is comparable in both cases, indicating no significant difference in the trade gap between higher-tax and lower-tax categories. However, the unit value gap is higher for higher-tax goods than lower-tax category goods. For higher-taxed goods, unlike the unit value gap, which is positive, the mean quantity gap is negative. This suggests that the tendency for evasion through underreporting unit values is greater for higher-tax category goods.

The mean tax rate applied on all imports is 39.2% for the whole sample. Though Kenyan-origin goods enjoy preferential treatment by it being a COMESA member country, duties and taxes remain significantly high. However, Table 1 shows that the tax rates facing goods vary with the forms of evasion. While the mean tax rates for goods reported by both partners (i.e., matching trade flows) is 40.87%, it is 37.5% for products assumed to be either smuggled into Ethiopia bypassing the official customs checkpoints or misclassified. For matching flows, the mean tariff rates and tax rates applied on products from Kenya at the 6-digit level are 16.97% and 40.87%, respectively. The higher tax rates are due to most goods facing multiple taxes such as VAT of 15%, surtaxes, while some selected goods also face excise duties. Moreover, Figure 1 also reveals considerable variation in the customs duties and taxes applied on imports across products. The average tariff rate ranges from 0% to 35% for some goods, the effective tax rate being as high as a staggering 131.5% (see Figure 1).

Figure 1.

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Ideally, goods that are smuggled through unofficial border crossings are expected to face higher tax rates compared to products partially evading tax at the official customs entry. In this regard, goods to be higher than the goods partially evading tariff. Mishra et al. (2008), find that the mean tax rates imposed on entirely “missing imports” are higher than the tax applied on goods with matching flows in India. However, in the Ethiopian case, the mean tax rates facing both categories of goods are relatively comparable.

Taxes do not appear to be higher for entirely missing products for two possible reasons. First, it seems that “entirely missing imports” capture both smuggled goods and deliberately mislabeled products. In this regard, the relatively lower average tax rate possibly signals product mislabeling from higher- to lower-tax categories, resulting in lower average tax rates. Second, entirely “missing imports” might not be explained by higher tax rates alone but also by other non-tax trade protections applied to imports at the border. Perhaps, a cursory observation of the statistics shows that the prevalence of non-tariff measures (NTM) facing product categories or specific sectors is relatively higher for completely smuggled goods. Moreover, entirely “missing imports” are noted to have a higher per-unit price and are smaller in volume than goods crossing official borders.

The distribution of the trade gap in Figure 2 resembles the normal distribution. In addition, the dispersion in the evasion gap and shape suggests a considerable degree of random noise in the data.

Figure 2.

Source: Computation based on UN COMTRADE database.

Empirical Results and Discussion

The econometric analysis examines tax evasion in Ethiopia’s imports from Kenya at the 6-digit level by comparing Kenya’s declared exports and Ethiopia’s reported imports in the same product lines. Aside from examining the effect of Ethiopia’s import tax rates applied on goods imported from Kenya, the model estimates the role of law enforcement, product differentiation, and product misclassification in explaining evasion.

Baseline Model Results: Trade Gap and Tax Rates

This section estimates the benchmark model specification, which links the trade gap with the tax-facing imports. The estimation result in Table 3 shows that the trade gap is positively and significantly associated with tax rates applied to imports from Kenya. The magnitude of the coefficient suggests that a 1% increase in the tax rate across products increases import tax evasion by 1.12% and 2.09% for matched products and entirely missing imports (i.e., smuggling and misclassified import cases). The evidence suggests a systematic positive association of tax rates with the trade gap.

The finding is in accord with the theoretical literature that associates undervaluation with the prevalence of high customs duties and tax rates (Bhagwati, 1964). Evasion most often requires understating the import value at the customs clearance, that is, as reported by the importer, while it does not require faking the exporter’s declaration in its own country. While several reasons may explain the discrepancies between the values recorded by the importer and the exporter (Hamanaka, 2011), evasion is the only one originating a positive relationship between tariffs and the gap between the value declared by the exporter and the importer (Jean & Mitaritonna, 2010). When firms pay high rates of customs duties or VAT on imports, they have an incentive to understate the actual value of imports (Patnaik et al., 2010).

The semi-elasticity of evasion concerning tax rates is above unity in both measures of evasion. Empirically, despite variations in the magnitude of the coefficients, similar findings have also been reported using African and non-African data and employing cross-sectional regressions for a single year. In this regard, van Dunem and Arndt (2009), using imports of Mozambique from South Africa in the year 2003, find a much similar point estimate of 1.4. Likewise, Bouet and Roy (2012), using a sample of African countries, find estimates for 2004 for Kenya and for 2001 for Nigeria that are nearly identical to the estimates of this study. On the other hand, Levin and Widell (2007) estimate a much higher coefficient of 3.8 for Tanzania. Moreover, for non-African countries, Mishra et al. (2008), using Indian imports, and Fisman and Wei (2004), using China’s imports from Hong Kong, reported coefficient estimates of 2.7 and 0.9, respectively.

Table 3.

Trade Gap and Tax Rates.

	Gap Value	Quantity Gap	Gap_Entirely Missing
Tax	1.12***(0.409)	1.19***(0.397)	2.09***(0.2258)
Constant	–0.1498	–0.216	5.83
Observation	310	310	285
Adjusted R²	0.020	0.025	0.23
F-stat.	7.51***	9.03	86.29***

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Notes: Standard errors are in parenthesis; *** is significant at 1% levels. Quantity gap stands for evasion in quantities. Gap_value = log(export value) – log(import value) for matched products; Quantity gap = log(export quantity) – log(import quantity) for matched products; Gap_entirely missing = log(1 + export value – log(1 + import value).

Aside from the trade gap measured in value, the analysis uses evasion in quantities as a dependent variable. The results in Table 3 (column 3) show that there is a positive and statistically significant effect of taxes on the quantity gap. The estimation result indicates that the underreporting of total quantities imported across all tax levels is significant, suggesting that underreporting of import quantities is prevalent for some classes of goods. Moreover, the magnitude of the coefficient reported in this case is not any different from the one estimated when evasion is defined in value terms. Whether it is measured in values or quantities, the trade gap is associated positively and significantly with tax rates. This suggests that tax evasion is driven primarily by the prevalence of higher tax rates on imports.

Extended Model Results

Tax Evasion and Product Differentiation

The baseline specification in Table 3 reports a positive and statistically significant effect of tax on import tax evasion. Yet aside from tax rates, the trade gap is also explained by product characteristics that make law enforcement and the probability of detecting evasion either easy or difficult at the border. To test the hypothesis that differentiated products are more likely to be subject to undervaluation and then evasion, the trade gap is regressed on the tax rate and the interaction between the product differentiation dummy and the tariff rate.

Table 4.

Trade Gap, Tax Rates, and Product Differentiation.

	Gap_Value	Quantity Gap	Gap_Entirely Missing
Tax	0.736*(0.418)	1.018**(0.412)	2.71***(0.35646)
Tariff*differentiation dummy	0.127***(0.038)	0.057(0.037)	–0.019***(0.00873)
Constant	–0.684	–0.459	5.81
Observation	310	310	285
F-stat.	9.43***	5.71***	46.19***
Adjusted R²	0.0518	0.0296	0.246

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Note: Standard errors are in parenthesis; *, ** and *** are significant at 10%, 5% and 1% levels, respectively. Gap_value = log(export value) – log(import value) for matched products; Quantity gap = log(export quantity) – log(import quantity) for matched products; Gap_Entirely Missing = log(1 + export value – log(1 + import value); Tariff*differentiation dummy is the product of a dummy for product differentiation and the tariff rates facing imports.

Table 4 shows that when product differentiation is included as an additional regressor, the effect of taxes on evasion drops from 1.12 in the benchmark model estimation to 0.73 for matched products. It is significant only at the 10% level. This suggests that the responsiveness of the trade gap to tax rates is smaller for homogenous goods relative to differentiated products. The interaction between tariff and the differentiated product dummy is positive and statistically significant at the 1% level when evasion (using the first definition) is measured in value. The estimate suggests that for a 1% increment in the tariff, the responsiveness of the trade gap to the tariff level is larger by 0.127% for differentiated goods relative to homogenous products. The result is in line with the notion that with more product differentiation, it is easy to conceal the true per unit value of the good, making the probability of detecting evasion difficult for customs officers. In particular, Javorcik and Narciso (2008) noted that differentiated products might lend themselves more readily to tariff evasion than homogenous products. Since there is greater variation in the prices of differentiated goods, it is relatively easier for importers or corrupt customs officers to misrepresent the price of the imports.

The finding that import tax evasion is more likely to be greater for differentiated goods than non-differentiated classes of goods is also corroborated by the results of other studies. Mishra et al. (2008) using different proxies for product differentiation find that more differentiated products exhibit a higher evasion elasticity than homogenous products in India. Likewise, Javorcik and Narciso (2008), using trade data between Germany and 10 Eastern European countries, find that a 1% increase in the tariff rate is associated with a 1.7% increase in evasion for differentiated products and a 0.4% increase for homogenous goods. However, the result contradicts Mengistu et al.’s (2022) study for Ethiopia, which finds no substantial evidence that the effect of the effective tax rate is different for homogenous and differentiated goods, and product differentiation appears to be statistically insignificant. On the other hand, product differentiation is associated significantly and negatively with evasion for entirely missing imports. The responsiveness of the gap in entirely “missing imports” to tax rates is lesser for differentiated goods relative to homogenous ones. This might be because it is easier for differentiated goods to evade tax at customs rather than being smuggled or misclassified.

Tax Evasion and Strength of Law Enforcement: The Effect of the Expected Penalty on Evasion

This section tests whether evasion decreases with strict law enforcement or with an increase in the cost of the expected penalty for tariff evasion. In the literature, the association between the strength of law enforcement, for instance, in levying fine on smugglers or punishing traders for evasion, and tax evasion is overlooked. The estimation here uses a quantitative measure of penalty for all traders evading more than 10% of the total duty payable; traders are expected to be fined as per the Customs Proclamation. In line with the theoretical a priori, the coefficient of penalty reported in Table 5 is negative and statistically significant at the 1% level. The estimated coefficient shows that the effect of increases in tax rate is larger when the quality of law enforcement is controlled, suggesting that punishment for evasion can serve to explain evasion.

The inclusion of the expected fine, as a proxy for the presence of strict law enforcement as an additional regressor in the benchmark model, increases the magnitude of the coefficient of tax rates considerably from 1.12 in Table 3 to 3.17 in Table 5. The result shows that an increase in the expected fine as a penalty for evasion tends to decrease the intention to evade tax significantly. In line with this study, Javorcik and Narciso (2008) show that the quality of governance in the importing country is a critical determinant of tariff evasion. They find that the extent of evasion is negatively related to the quality of governance in the importing country for bilateral trade flows within Central and Eastern Europe. Interestingly, in countries with a better quality of governance, for instance, in the case of German imports from Central and Eastern Europe, they find no evidence of evasion. Miskam et al. (2013) also find that penalty for evasion is negatively associated with tax evasion in imported vehicles in Malaysia.

Table 5.

Trade Gap, Tax Rates, and Law Enforcement.

	Gap_Value	Quantity Gap	Gap_Entirely Missing
Tax	3.17***(1.028107)	4.980***(0.977)	4.60***(0.4262)
Expected penalty	–0.001***(0.0003)	–0.0015***(0.0003)	–0.00021***(0.00003)
Constant	–0.106	–0.137	5.66
Observation	310	310	285
Adjusted R²	0.0324	0.0757	0.33
F-stat.	6.17***	13.66***	72.77

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Note: Standard errors are in parenthesis; *** is significant at 1% level. Gap value = log(export value) – log(import value) for matched products; Gap_ quantity = log(export quantity) – log(import quantity) for matched products; Gap_entirely missing = log (1 + export value – log(1 + import value); Gap_unit value = Log export unit value-Log import unit value for matched products; Expected penalty, which is a proxy for law enforcement, is fine applied on tax evaders, measured in thousands USD.

The finding shows that if traders expect to face a higher penalty for smuggling and circumventing tax, the intention to smuggle goods tends to decline. Perhaps, the literature suggests that when choosing to trade informally, traders usually compare the costs of formal trade relative to trading through the informal route (Bouet et al., 2018). Equivalently, smugglers usually undertake a cost-benefit analysis of evasion and evade tax so long as it is profitable to do so. The negative coefficient suggests that the benefit of evasion could decrease more rapidly than the cost as tax rates increase and resorting to smuggling would not be profitable for importers. When the costs of informal trade are lower, notably due to weak law enforcement, smuggling is profitable. Smuggling occurs when, due to inadequate law enforcement, a high degree of corruption and the requirement of facilitation payments prevail along with the official border posts (Golub, 2012; Lesser & Moisé-Leeman, 2009). However, strong law enforcement in the form of fines increases evasion costs, making smuggling less profitable and evasion less prevalent.⁴

The result is in accord with the theoretical literature where an increase in the penalty rate will always increase the fraction of the actual value declared (Allingham & Sandmo, 1972) and discourages evasion (Shimeles et al., 2017). In particular, Allingham and Sandmo’s (1972) theoretical model of tax evasion, which is perhaps based on a risk-averse taxpayer, predicts that a higher penalty rate or a higher probability of detection tends to discourage tax evasion. It should be noted that the magnitude of the penalty facing smugglers and the decision to evade tax depends on the probability of detecting evasion by customs and the prevalence of corruption among customs officers and authorities. For risk-averse traders (i.e., smugglers), since the expected fine is the cost of evasion, the finding that the trade gap decreases with the rise in the cost of evasion is reasonable. The higher the expected cost of evasion, which is conditional to strict law enforcement at the border that identifies deliberate underreporting at the official border and fines as per the law, the lower is customs tax evasion. In other words, an increase in the expected magnitude of fine payment (i.e., the penalty rate) will increase the proportion of actual imports declared which discourages smuggling and enhances the collection of import tax revenue. The result shows the crucial role of enforcing the law at the border, such as punishing smugglers to reduce the magnitude of smuggling and evasion. The finding implies that had the law been strictly enforced in Ethiopia, the cost of evasion would have been higher for smugglers, reducing evasion in Ethiopia.

Tax Evasion and Product Misclassification

Literature suggests that the lower the tariff on similar products, the greater is the incentive to misclassify imports (Fisman & Wei, 2004; Mishra et al., 2008). The empirical model tests if evasion is practiced by misclassifying high tariff products as a lower tariff product category. Accordingly, using both measures of trade gap, the average tariff on similar products classified at the 4-digit level is included as an additional regressor in the benchmark model.

The estimation results reported in Table 6 show that the average tariff on similar products has a statistically insignificant effect on the discrepancy in matching trade flows. This suggests no evidence of import tax evasion through the channel of product misclassification for undervaluation-based evasion. This finding corroborates the findings of Javorcik and Narciso (2008) and Levin and Widell (2007), who report no evidence of evasion through product misclassification. On the other hand, there is strong statistical evidence that tariff evasion in unrecorded imports (imports recorded with zero values) in Ethiopia is practiced by misclassifying goods from higher to lower tariff categories. Holding the own tax constant, a 10% decrease in the tariff on similar products leads to a 0.5% increase in tax evasion for goods reported to be entirely missing (see column 4 of Table 6 for the statistical significance of the average tariff variable). Unlike the case for the first measure of evasion, product misclassification significantly explains the discrepancy in entirely “missing imports.” This suggests that the explanations for evasion could vary depending on the type of evasion practiced by traders. Using a similar definition of evasion, Mishra et al. (2008) also find negative coefficient estimates for India.

Table 6.

Trade Gap, Tax Rates, and Product Misclassification.

	Gap_Value	Quantity Gap	Gap_Entirely Missing
Tax	1.11***(0..410)	1.177 ***(0.396)	2.23***(0.22)
Average tariff	–0.013(0.015)	–0.021(0.015)	–0.057***(0.0129)
Constant	0.072	0.1353	6.66
Observation	309	309	281
Adjusted R²	0.0198	0.0285	0.28
F-stat.	4.1**	5.1***	55.48***

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Notes: Standard errors are in parenthesis; ** and *** are significant at 5% and 1% levels, respectively. Gap value = log(export value) – log(import value) for matched products; Gap_ quantity = log(export quantity) – log(import quantity) for matched products; Gap_Entirely Missing =log(1+export value)- log (1+ import value); Average tariff is average tariff applied on similar products at the 4-digit level.

The literature considers such a systematic negative association between tariffs on similar products and evasion as evidence of deliberate misclassification of imports. Perhaps, aside from capturing “complete smuggling” cases, entirely missing imports are also apprehended as evidence of deliberate misclassification of goods by the importers (Mengistu et al., 2022). The tax statistics demonstrate that the mean tax rate for entirely missing imports is considerably lower than the tax rate for matched items, indicating that high tariff products may be mislabeled as lower tariff goods. Moreover, about 7% of the misclassified goods face no customs duty at all. Thus, the appearance of duty-free goods and goods facing lower tariffs in the entirely “missing imports” supports the notion that part of the missing values is due to product misclassification.

Finally, all the explanatory variables were jointly included in the model, and the responsiveness of the trade gap to tax rates, expected penalty for evasion, and product characteristics were evaluated. Except for the changes in the magnitude of the coefficients, the results reported in Table 7 are similar to those reported using stepwise regression earlier. The estimation reveals that the trade gap in value is associated positively with tax rates and product differentiation and negatively with the expected penalty rate. When evasion in quantity is employed as a dependent variable, the results are similar. However, in this case, product differentiation is statistically insignificant. As discussed earlier, this result is acceptable since under-reporting of quantities is not the primary form of evasion for differentiated products. For such goods, evasion is easier through underreporting the per-unit value of imports. As seen earlier, while the average tariff on similar goods turns out to be insignificant for partial evasion cases, it is statistically significant for the second definition of evasion.

Table 7.

Drivers of Trade Gap: Tax, Product Differentiation, Law Enforcement, and Product Misclassification.

	Gap Value	Quantity Gap	Gap_Entirely Missing
Tax	2.90***(1.016)	4.874***(0.9766863)	4.60***(0.41245)
Average tariff	–0.0116(0.0153)	–0.0220(0.0147)	–0.055***(0.0118)
Tariff*differentiation	0.128***(0.038)	0.0578(0.0366615)	0.022**(0.0098)
Expected penalty	–0.00086**(0.0003714)	–0.00154***(0.0003568)	–0.00025***(0.000037)
Constant	–0.318	–0.0101	6.44
Observation	309	309	281
Adjusted R²	0.0635	0.0848	0.39
F-stat.	6.22***	8.13***	46.15***

Source: Computation based on UN COMTRADE, World Bank’s WITS, and ECC databases.

Notes: Standard errors are in parenthesis; ** and *** significant at 5% and 1% levels, respectively.

Policy Implications

Evasion in Ethiopia is practiced through undervaluation, smuggling, and product misclassification. Though this study draws evidence from a single country using its primary neighboring import source, the results can contribute to discussions on trade policy and law enforcement in cross-border trade.

First, import tax evasion is explained by the prevalence of higher duty and tax rates for Ethiopian imports from Kenya. Since importers seek to minimize trade costs through the formal channel, they resort to evasion to reduce or altogether avoid the tax burden at the border. Due to import tax evasion, the government faces considerable tax revenue losses, which could have been otherwise used for financing public expenditure. From a policy perspective, as long as evasion is primarily the outcome of higher customs duty and taxes, policy measures should aim to reduce those costs arising mainly from high trade taxes applied on imports from the neighboring partners. To reduce the costs of formal trade, there is a need to embark on selective and evidence-based reduction of tax rates imposed on various categories of goods. Such selective reductions are expected to encourage traders to choose the legal (or formal) trade route that will reduce illicit operations.

Lowering duties and tax rates, however, will have ramifications for generating tax revenue and the development financing capacity of the state. If the majority of the import tax rates are to the left of the peak of the Laffer curve, reductions in tax rates may entail a loss of customs tax revenue. However, a reduction in tariff levels might increase the tax revenue if increases in the tax rate above a certain level tend to reduce the collection rate. This is an empirical issue involving the estimation of the turning point tax rate. To estimate the turning point, a spline regression is performed.⁵ The regression result reveals that the turning point tax rate is about 35%, suggesting that a further increase in the tax rate beyond this level would result in a reduction in the customs tax collection rate and tax revenue. Thus, the fact that the average tax rate for imports from Kenya is on the wrong side of the Laffer curve entails that a reduction in the mean tax rates below 35% may increase the tax revenue.

Second, aside from trade policy factors, this study quantifies the effect of law enforcement on import tax evasion using the (hypothetical) penalty that could have been applied to importers when traders fail to comply with the law or evade taxes. The finding suggests that a higher level of expected penalty for evading taxes should discourage evasion. The presence of pervasive smuggling and other forms of illicit operations in Ethiopia likely suggests that weak law enforcement at the border is the main reason for evasion. Given that laxity of law enforcement contributes to evasion, strong law enforcement is crucial to reducing evasion practiced by registered traders. In this regard, enhancing the effectiveness of detecting undervaluation and product mislabeling by using an up-to-date price database of goods, strictly punishing tax evaders, controlling the corrupt behavior of customs officers and authorities through appropriate punishment, and improving the capacity of the Customs Commission to control smuggling at unofficial borders will be effective interventions for reducing smuggling and evasion.

Third, the study suggests that the magnitude, significance of variables, and explanations for evasion depend on the form of evasion practiced by traders. For instance, the differential association of product misclassification and differentiation with the different measures of evasion suggests that the explanations for the trade gap can vary depending on the type of evasion practiced by traders. Thus, rather than focusing on one dimension of evasion, or aggregating both forms of evasion, it is important to analyze each form of evasion separately.

Conclusion

Intra-African trade is largely informal and mostly takes the form of undervaluation and product mislabeling at customs and smuggling through unofficial borders. In particular, undervaluation and smuggling practiced by registered traders lead to considerable customs tax evasion and loss of tax revenue for low-income African countries. This study examines the major explanations for evasion using “missing imports” data from Kenya disaggregated at the HS 6-digit product level. Two distinct forms of customs tax evasion practiced by registered traders were considered: evasion in matched products and evasion through smuggling and/or product misclassification. The trade gap arising from deliberate misreporting was explained using tax rates, a quantitative measure of the strength of law enforcement at the border, product differentiation, and misclassification of imports as major explanatory variables.

The econometric estimation results show several interesting findings. Increasing taxes levied on a given product by 1% leads to an increase in import tax evasion by 1.12% and over 2% for products reported by both partners and entirely “missing imports,” respectively. Import tax evasion is primarily explained by the prevalence of higher tax rates facing imports from Kenya. From a policy perspective, as long as traders resort to illicit operations owing to higher tax rates, reducing customs tax evasion through judicious tariff reduction needs to be the government’s essential trade policy objective.

Aside from higher taxes, evasion is also affected by the degree of homogeneity or differentiation of the imported good. Evasion is significantly higher for differentiated products than homogenous goods, suggesting that the intention to evade tax increases since misrepresenting import prices at customs is easier for differentiated products. Moreover, expecting a higher penalty for evasion is associated negatively with the trade gap, suggesting that appropriate enforcement of the law could have discouraged evasion. On the other hand, the observed evasion probably arises from the lower penalty that importers expect to face when they circumvent the customs law, that is, traders evade tax owing to the laxity of law enforcement by the Ethiopian Customs Commission.

To sum up, tax evasion in imports from Kenya is explained primarily by higher customs duties and tax rates, weak law enforcement at the border, and product differentiation that makes it easier to evade tax. There is strong evidence that tariff evasion in goods that are entirely unrecorded on the Ethiopian side is practiced by deliberately mislabeling higher-taxed products as lower-taxed categories.

This study draws conclusions using evidence from a single country. However, to get a generalized picture of tax evasion and ICBT in Africa, further studies should consider using data both from official trade statistics and border surveys. Moreover, future studies may focus on analyzing the role of disparities in trade policy and the quality of law enforcement across countries in explaining evasion using data from multiple countries.

Footnotes

Declaration of Conflicting Interests

The author declared no conflict of interest with respect to research, authorship, and /or publication of this article.

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work has been funded by Dilla University as a part of its “large-scale research grant” scheme. The funding falls under the category of “regular research grants” at the university. The grant was awarded in 2019 with the grant number DU/LS/CoBE/01/2019.

Notes

ORCID iD

Tasew Tadesse

References

Allingham

M. G.

, & Sandmo

(1972). Income tax evasion: A theoretical analysis. Journal of Public Economics , 1(3–4), 323–338.

Bensassi

, Jarreau

, & Mitaritonna

(2019). Regional integration and informal trade in Africa: Evidence from Benin’s borders. Journal of African Economies , 28(1), 89–118.

Bhagwati

(1964). On the underinvoicing of imports. Oxford Bulletin of Economics and Statistics , 27(4), 389–397.

Bouet

, Pace

, & Glauber

(2018). Informal cross-border trade in Africa: How much? Why? And what impact? (IFPRI Discussion Paper 01783). https://ideas.repec.org/p/fpr/ifprid/1783.html

Bouët

, & Roy

(2012). Trade protection and tax evasion: Evidence from Kenya, Mauritius, and Nigeria. Journal of International Trade and Economic Development , 21(2), 287–320.

Cantens

, & Raballand

(2017). Cross-border trade, insecurity and the role of customs: Some lessons from six field studies (International Center for Tax and Development Working Paper No. 67). Institute of Development Studies. https://www.ictd.ac/publication/ictd-working-paper-67-cross-border-trade-insecurity-and-the-role-of-customs-some-lessons-from-6-field-studies-in-post-conflict-regions

Carrère

, & Grigoriou

(2015). Can mirror data help to capture informal international trade? (Working Paper 123). FERDI Development Policies. https://ferdi.fr/en/publications/can-mirror-data-help-to-capture-informal-international-trade

Ethiopian Revenue and Customs Authority (ERCA). (2017). Ethiopian customs guide .

Federico

, & Tena

(1991). On the accuracy of foreign trade statistics 1909–1935: Morgenstern revisited. Explorations in Economic History , 28(3), 259–273.

10.

Fisman

, & Wei

(2004). Tax rates and tax evasion: Evidence from “missing imports” in China. Journal of Political Economy , 112(2), 471–496.

11.

Golub

(2015). Informal cross-border trade and smuggling in Africa. In Morrissey

(Ed.), Handbook on trade and development in Africa (pp. 179–209). Edward Elgar Publishing.

12.

Golub

S. S.

(2012). Entrepôt trade and smuggling in West Africa: Benin, Togo, and Nigeria. World Economy , 35(9), 1139–1161.

13.

Golub

S. S.

, & Mbaye

A. A.

(2009). National trade policies and smuggling in Africa: The case of the Gambia and Senegal. World Development , 37(3), 595–606.

14.

Hamanaka

(2011). Utilizing the multiple mirror technique to assess the quality of Cambodian trade statistics (Working Paper Series No. 88). ADB. https://www.adb.org/publications/utilizing-multiple-mirror-technique-assess-quality-cambodian-trade-statistics

15.

Harris

, & Seid

(2021). 2019/20 Survey of the Ethiopian tax system. https://ifs.org.uk/publications/15402#

16.

Hien

, & Viet Hung

(2017). Mirror study of Vietnam’s international merchandise trade: Findings and implications on cross-border trade administration (Working Paper No. 26/2017). https://www.wti.org/research/publications/1157/mirror-study-of-vietnams-international-merchandise-trade-findings-and-implications-on-cross-border-trade-administration/

17.

Javorcik

B. S.

, & Narciso

(2008). Differentiated products and evasion of import tariffs. Journal of International Economics , 76(2), 208–222.

18.

Jean

, & Mitaritonna

(2010). Determinants and pervasiveness of the evasion of customs duties (Working Paper No. 2010-26). CEPII. https://ageconsearch.umn.edu/record/115432/?ln=en

19.

Lesser

, & Moisé-Leeman

(2009). Informal cross-border trade and trade facilitation reform in sub-Saharan Africa (Trade Policy Working Papers No. 86). https://www.oecd-ilibrary.org/trade/informal-cross-border-trade-and-trade-facilitation-reform-in-sub-saharan-africa_225770164564

20.

Levin

, & Widell

L. M.

(2007). Tax evasion in Kenya and Tanzania: Evidence from missing imports (School of Business Working Paper No. 8). https://ideas.repec.org/p/hhs/oruesi/2007_008.html

21.

Levin

, & Widell

L. M.

(2014). Tax evasion in Kenya and Tanzania: Evidence from missing imports. Economic Modelling , 39, 151–162. https://doi.org/10.1016/j.econmod.2014.02.021

22.

Mengistu

A. T.

, Molla

K. G.

, & Mascagni

(2022). Trade tax evasion and the tax rate: Evidence from transaction-level trade data. Journal of African Economies , 31(1), 94–122. https://doi.org/10.1093/jae/ejab005

23.

Mishra

, Subramanian

, & Topalova

(2008). Tariffs, enforcement, and customs evasion: Evidence from India. Journal of Public Economics , 92(10–11), 1907–1925. https://doi.org/10.1016/j.jpubeco.2008.04.017

24.

Miskam

, Noor

R. M.

, Omar

, & Aziz

R. A.

(2013). Determinants of tax evasion on imported vehicles. Procedia Economics and Finance , 7, 205–212. https://doi.org/10.1016/S2212-5671(13)00236-0

25.

Mitaritonna

, Bensassi

, & Jarreau

(2017). Regional integration and informal trade in Africa: Evidence from Benin’s borders (Working Papers No. 2017-21). CEPII. https://ideas.repec.org/p/cii/cepidt/2017-21.html

26.

Patnaik

, Gupta

, & Shah

(2010). Determinants of trade misinvoicing (NIPFP Working Paper No. 2010-75). https://econpapers.repec.org/paper/npfwpaper/10_2f75.htm

27.

Pitt

M. M.

(1981). Smuggling and price disparity. Journal of International Economics , 11(4), 447–458. https://doi.org/10.1016/0022-1996(81)90026-X

28.

Pritchett

, & Sethi

(1993). Tariff rates, tariff revenue, and tariff reform: Some new facts (Policy Research Working Paper Series 1143). https://ideas.repec.org/p/wbk/wbrwps/1143.html

29.

Rauch

J. E.

(1999). Networks versus markets in international trade. Journal of International Economics , 48(1), 7–35.

30.

Shimeles

, Gurara

D. Z.

, & Woldeyes

(2017). Taxman’s dilemma: Coercion or persuasion? evidence from a randomized field experiment in Ethiopia. American Economic Review , 107(5), 420–424.

31.

van Dunem

, & Arndt

(2009). Estimating border tax evasion in Mozambique. Journal of Development Studies , 45(6), 1010–1025.

32.

Yeats

(1995). Are partner-country statistics useful for estimating “missing” trade data? Policy research (Working Paper No. 1501). The World Bank. https://documents.banquemondiale.org/fr/publication/documents-reports/documentdetail/256801468764415488/are-partner-country-statistics-useful-for-estimating-missing-trade-data