Policy Research Working Paper 10339 Natural Resource Dependence and Monopolized Imports Rabah Arezki Ana Fernandes Federico Merchán Ha Nguyen Tristan Reed Development Economics Development Research Group March 2023 Policy Research Working Paper 10339 Abstract Countries with greater commodity export intensity have efficiency. Hydrocarbon fuel exporting economies especially more concentrated markets for imported goods. Within have higher tariffs, tariff evasion, and non-tariff measures countries over time, import market concentration is associ- that concentrate markets. These results suggest a novel ated with higher domestic prices, suggesting that markups channel for the resource curse stemming from the monop- due to greater concentration outweigh any potential cost olization of imports. This paper is a product of the Development Research Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at afernandes@worldbank.org and treed@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Natural Resource Dependence and Monopolized Imports an, Ha Nguyen, Tristan Reed∗ Rabah Arezki, Ana Fernandes, Federico Merch´ JEL Classification: D2; F1; L1; O1; Q0 Keywords: imports, market concentration, natural resources, resource curse. ∗ Rabah Arezki is a Senior Fellow at the Harvard Kennedy School, Ana Fernandes and Tristan Reed are economists at the World Bank Development Research Group, Ha Nguyen is an economist at the International Monetary Fund (IMF), and Federico Merch´ an is a Ph.D. student at Kiel University (Germany). We thank Leila Baghdadi, Olivier Blanchard, Shanta Devarajan, Simeon Djankov, Jeffrey Frankel, Caroline Freund, Penny Goldberg, Jean Imbs, Daniel Lederman, Rick van der Ploeg, Bob Rjikers, Gregoire Rota-Graciozi, and Tony Venables for valuable comments and suggestions, as well as participants at the 2nd Annual Central Bank Conference on Development Economics in the Middle East and North Africa. Mayra Monroy, Jan Oledan, and Gaston Nievas provided excellent research assistance. This paper benefited from support from the Umbrella Facility for Trade trust fund financed by the governments of the Netherlands, Norway, Sweden, Switzerland and the United Kingdom and the Multi-Donor Trust Fund (MDTF) from the World Bank’s Middle East and North Africa Region. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the World Bank and the IMF, the Executive Directors of the World Bank and the IMF or the governments they represent. 1 Introduction Dependence on natural resources for exports creates a variety of macroeconomic challenges known collectively as ‘the resource curse’ (Sachs and Warner, 2001; Van der Ploeg, 2011; Frankel, 2012; Venables, 2016; Arezki, Ramey and Sheng, 2017). One challenge is rent- seeking wherein natural resource rents controlled by the state increase the return to state capture, leading to inefficient policy choices in the absence of strong institutions. Another challenge is the so-called Dutch disease wherein a natural resource discovery or price appre- ciation is accompanied by an increase in the real exchange rate, which in turn shrinks the non-resource tradable sector. In principle, both challenges could interact. Foreign exchange receipts from natural resources increase domestic demand for imports, increasing the value of the domestic import market. By making the import market larger, natural resources raise the return to effort by importers towards capturing the state and directing state power to shield them from competition. Yet, existing theoretical models of state capture in natural resource-dependent economies (Tornell and Lane, 1999; Robinson, Torvik and Verdier, 2014) do not emphasize profits in the import market as a source of rents. Anecdotal evidence is consistent with this import monopolization effect, as the wealth of many of the richest businesspeople in natural resource dependent economies is linked to profits in import markets. Prominent billionaires in Nigeria, Saudi Arabia, and the Russian Federation accumulated their wealth for instance as a fuel importer, an exclusive distributor for a car manufacturer, and an importer of cigarettes, food, and alcohol.1 This paper moves beyond anecdotes and provides systematic evidence that natural resource dependence causes the monopolization of imports, and that this monopolization can account for higher price levels in the non-resource tradable sector of natural resource dependent economies. The term ‘monopolization’ is used to describe a shift in market structure toward one that is more concentrated. The analysis exploits a novel database of all firm-level import transactions in 29 devel- oping and emerging market economies. These data reveal that natural resource dependent economies have more concentrated markets for imported products. This basic pattern is illus- trated in Figure 1a, which shows a positive association across countries between commodity exports as a share of total merchandise exports and the average Herfindahl–Hirschman index (HHI) across all imported product markets in a country. The HHI for an imported prod- uct market is the sum of the squared market shares of every firm importing that product. Econometric estimates show this relationship is robust to controlling for GDP per capita, 1 See Freund (2016) for an account of the origin of billionaires’ wealth in emerging markets. Other than imported product markets, ownership of firms in the telecom and logistic sectors, which can be natural monopolies, are important sources of such wealth. 2 a crucial test since smaller markets could mechanically be more concentrated under fixed import costs. A panel specification with country-product fixed effects reassures that the re- lationship is causal, as does a specification that uses exogenous increases in world commodity prices instead of the commodity share of exports. Import monopolization can account for a stylized fact about commodity exporting coun- tries, which is that their prices of tradable goods in a common currency are higher than those in other countries. While this fact is typically attributed to the Dutch disease, an alternative mechanism suggested by our results is that prices are elevated due to higher markups ensu- ing from monopolistic or oligopolistic pricing by importers. Data on domestic price levels from the International Comparison Program (ICP) show that within countries over time greater import market concentration can account for higher prices of tradable goods relative to the United States (U.S.), ICP’s benchmark economy. This result is not obvious ex-ante: if higher market concentration is associated with a higher fixed cost but lower marginal cost of importing, import market concentration could be associated with lower prices, even if it contributes to higher markups. Two additional pieces of evidence suggest that capture of trade policy is a mechanism for import monopolization in natural resource dependent economies. First, commodity export intensive economies place higher tariffs on imports, as shown in Figure 1b. The seminal literature on rent-seeking suggested tariff evasion could explain the persistence of high tariffs, as elites that enjoy the advantage of evasion are a constituency in favor of tariffs remaining high (Tullock, 1967; Krueger, 1974). State capture allows elites to evade tariffs (Rijkers, Baghdadi and Raballand, 2017) and it is well documented, including in our sample, that import underinvoicing, a method for tariff evasion, increases with the tariff rate (Bhagwati, 1964; Fisman and Wei, 2004; Yang, 2008; Mishra, Subramanian and Topalova, 2008; Sequeira, 2016; Javorcik and Narciso, 2017). Greater tariff evasion can create market concentration when a subset of firms evades tariffs. In this case, firms obtain a cost advantage that allows them to increase their market share relative to those that cannot evade tariffs. Second, within countries over time, tariffs and tariff evasion have a stronger effect on import market concentration during oil export booms compared to booms of other com- modities. Isham, Woolcock, Pritchett and Busby (2005) document that among natural resource dependent economies, governance is especially weak in those economies dependent on commodities whose extraction is point-based, meaning revenues typically transit directly through government coffers, as opposed to commodities with a more diffuse production base. Oil is the quintessential point-based commodity, and fuel export intensive economies have weaker control of corruption. Non-tariff measures in the form of import quotas and price 3 restrictions based on Kee and Nicita (2022) are associated with import monopolies, and are also more prevalent in fuel exporting economies. Our paper is the first to systematically explore differences across countries in import market structure, contributing to several literatures beyond that about the resource curse. While the export sector has been the traditional focus of the trade and development litera- ture, in developing and emerging markets the value of imports is about as large as the value of exports, and many exported goods are made using imported inputs (UNCTAD, 2021). We identify patterns in import market structure that contrast starkly with those in studies examining export market structure. Fernandes, Freund and Pierola (2016) use the same customs transactions data to document that higher-income economies have more exporting firms, but also more concentrated export markets dominated by “superstars,” or firms with especially large market shares, whose characteristics are described by Freund and Pierola (2015, 2016).2 The pattern in import markets is the opposite. Higher-income economies have less concentrated import markets, independent of their commodity export intensity. More generally, our paper informs a macroeconomics literature interested in measuring the association between market structure and welfare. While contributions by Edmond, Midrigan and Xu (forthcoming) and Aghion, Bloom, Blundell, Griffith and Howitt (2005) show that increased markups stemming from high market concentration may (though need not) harm welfare, less is known empirically about differences in market concentration across economies, and their causes and implications. A recent paper by Leone, Macchiavello and Reed (2021) describes how high market concentration leading to high markups has raised prices in Africa’s domestic cement industry, though they argue that the source of these markups is small national market size in the presence of fixed costs, rather than higher entry costs that are unique to African economies. In contrast, the present paper provides evidence of entry costs in importing that are unique to commodity export intensive economies, and which can account for higher costs in these economies. Finally, we demonstrate that the association between exports of commodities and import market concentration is strongest for inputs like primary goods, parts, and semi-finished materials. This implies that import market concentration could shape firms’ international input sourcing decisions, a topic of recent research using U.S. data (Antr` as, Fort and Tin- telnot, 2017; Goldberg and Reed, 2023). To the extent that import monopolization raises costs of input procurement in global value chains, it may impede efforts to diversify exports away from natural resources. 2 Freund and Pierola (2015) show national revealed comparative advantage is shaped by the presence of superstar exporters. Gaubert, Itskhoki and Vogler (2021) discuss the policy implications of such “granular” comparative advantage in exports. Our evidence highlights that import markets can also be granular, with implications for the price level. 4 2 Commodity export intensity and import market con- centration 2.1 Measuring import market concentration Our analysis relies on a novel database of all firm-level import transactions recorded by customs authorities in 43 countries that are geographically and institutionally diverse and broadly representative of middle-income economies (see Table A1). The database has the same source as the World Bank Exporter Dynamics Database described by Fernandes et al. (2016) but includes import rather than export transactions. The sample period covers 1998-2018 but with different year coverage for each country. We eliminate observations in HS27 (oil, petroleum, natural gas, and coal) as their trade is not uniformly reported across countries’ customs data. Country-year total non-oil imports in our data are very similar to the corresponding total non-oil imports reported by COMTRADE (the average difference is 5.6%). Measuring market concentration requires defining a relevant market, or the set of products over which the firms in question have market power. While relevant market definition is often the object of intense debate in antitrust litigation, a general principle is that it should include the set of goods that are close substitutes for the same set of consumers (Davis and Garc´ es, 2009). Benkard, Yurukoglu and Zhang (2021) note that economic census data whereby firms are classified into industries are collected at the point of production rather than consumption and so may be less useful for analyzing the relation between market concentration and consumer welfare. In contrast, the Harmonized System (HS) product categories defined by the United Nations used in trade data classify goods with a similar end-use, and so are conceptually like the relevant product markets in antitrust analysis. As a first pass, in Figure 1a, we define imported product markets at the HS 4-digit level. These markets are quite specific though in principle could capture a market that is broader than a relevant consumer market. For instance, HS2101 includes “extracts, essences and concentrates of coffee, tea or mate and preparations thereof.” While coffee and tea are substitutes, oligopoly power could be most relevant within the markets for coffee or tea, as some consumers drink only coffee, while others drink only tea. Hence, in subsequent analysis we also define markets at the HS 6-digit level, separating for instance HS210111 “Extracts, essences and concentrates, of coffee,” which includes Nescafe instant coffee; from HS210120 “Extracts, essences and concentrates, of tea or mate, and preparations with a basis of these extracts, essences or concentrates, or with a basis of tea or mate,” which includes Lipton tea bags. Though the six-digit classification ultimately provides more specificity, it is reassuring 5 that the qualitative patterns are similar when using the four-digit categories and do not depend on relevant market definition. Following Fernandes et al. (2016), we use a time- consistent consolidated classification that concords and harmonizes product codes across the HS 1996, 2002, 2007, and 2012 versions (present in the raw data). Import market concentration is measured as the Herfindahl-Hirschman index or I 2 Mc,i,s,t HHIc,s,t = 100 × i i Mc,i,s,t where Mc,i,s,t is the import value in country c of firm i in relevant market s in year t, and I is the number of firms in the relevant market, country, and year. Values are measured including cost of freight and insurance (CIF) and are either reported by customs agencies in US dollars or converted from local currency to US dollars using the current exchange rate from IMF International Financial Statistics. Table A1 reports the distribution of HHI for HS 6-digit products by country. 2.2 Panel evidence Considering Figure 1a, one might be worried about omitted variables at the country level such as economic size that could mechanically influence concentration in the presence of fixed costs. To discipline the analysis further, we specify a fixed effects panel regression using about 10 years of data for each country: HHIc,s,t = αc,s + τs,t + β1 ExpComc,t + β2 log(GDP P Cc,t ) + ϵc,s,t (1) where αc,s is a country-product market fixed effect that captures unobserved market char- acteristics that may explain concentration (e.g., market size, consumer preferences) and τs,t controls for global product-year specific factors that do not vary across countries (e.g., tech- nological fixed costs, per-unit good value, logistics network requirements). The independent variable of interest is the percent of commodity exports in total merchandise exports, de- noted by ExpComc,t , a measure of natural resource dependence from the World Development Indicators. The regression includes economic size, measured by (the log of) GDP per capita, log(GDP P Cc,t ) since in the presence of fixed costs of importing, smaller markets could be mechanically more concentrated (population is slower moving over time, and so its effect is subsumed into the country-product fixed effect). Moreover fixed costs of importing could increase as income grows, and wages and the price level increase. The coefficient β1 is the main parameter of interest. The term ϵc,s,t is an error. 6 Panel A of Table 1 reports estimates of Equation 1. Within country-product markets over time an increase in commodity exports is positively associated with import market concentration, similarly to the cross sectional relationship in Figure 1a. Column 1 of Panel A defines relevant product markets at the HS 4-digit level within a country. The estimate of β1 implies that a 1 percentage point increase in commodity export intensity is associated with a statistically significant increase in the HHI of 4.97 (standard error = 1.97), everything else equal. US Department of Justice (DOJ) guidelines consider markets with an HHI in excess of 2,500 to be highly concentrated. The average market in column 1 has an HHI of 3,224. DOJ guidelines further suggest that an increase in the HHI of 200 should be expected to increase market power in a highly concentrated market.3 Consequently, an increase in commodity export intensity by 200/4.97 = 40.24 percentage points would be expected to increase market power according to these guidelines.4 Economically significant variation in importer market power is present in our sample, where commodity export intensity ranges from 8.5% in Bangladesh to 92% in Zambia (Table A1). Column 2 defines sectors at the HS 6-digit level, which is more narrow and likely closer to a relevant market in antitrust litigation. Here, as expected, average concentration according to the HHI is higher compared to column 1, at 4,101. A 1 percentage point increase in commodity export intensity is associated with a statistically significant increase in the HHI of 6.31 (2.33). Column 3 uses another measure of concentration, the largest firm concentration ratio (market share of the largest importer). Quantitatively moving from the commodity export intensity of Bangladesh to Paraguay increases the largest importer’s market share by 4.2 percent (0.0005 × 84). One interesting pattern across these columns is that an increase in GDP per capita significantly reduces import market concentration. This is consistent with richer countries having larger import markets and therefore being able to sustain more entrants. This result is in contrast to the findings of Fernandes et al. (2016) that exporter concentration within a country rises with GDP per capita and suggests potential scale economies in importing. Exploiting only international commodity price variation An alternative measure of natural resource dependence relies only on fluctuations in world commodity prices. These prices are plausibly exogenous since the economies in our sample are small relative to the world economy and do not have major export shares in key commodity groups. For instance, 3 See https://www.justice.gov/atr/herfindahl-hirschman-index. Nocke and Whinston (2022) come to a similar conclusion in their analysis of the potential price effects of mergers. 4 This magnitude of 40 percentage points is roughly the difference in commodity export intensity be- tween Gabon, which exports almost exclusively oil, and Mauritius, which exports food commodities but also manufactures. 7 on oil exports, our sample does not include Russia, Saudi Arabia, and the United States, the 3 largest exporters. The time-varying measure of commodity export intensity is the commodity export basket price index from Gruss and Kebhaj (2019). For country c in year t this is index is J j =1 log(Pj,t )ωc,j where Pj,t is the world price of commodity j in year t, and ωc,j is the weight that equals to the average value of commodity j ’s exports as a share of GDP across the 1980-2020 period. The index is scaled for each country so 100 equals the price index in 2012. Column 4 of Table 1, Panel A shows results using this index in place of the commodity export share. A 1% increase in the commodity price index, which is in the 75th percentile of year-on-year changes in the index, increases the HHI by 31.17 percent (7.70). This effect is smaller in magnitude than the effect measured using the commodity export share. Though capturing different variation, these results exploiting exogenous international price variation give us confidence the association between natural resource booms and import market concentration is causal. Heterogeneity across product type We explore the heterogeneous association between commodity export intensity and import market concentration by splitting the sample be- tween goods with different end uses: capital goods, consumption goods, materials (parts and semi-finished goods), and primary goods according to the Broad Economic Categories (Revision 5). Materials represent 48% of total import value in our sample, capital goods 27%, consumption goods 12%, primary goods 12%, and the remainder are not classified. Common examples of intermediate goods are electronic circuits; examples of capital goods are transmission apparatuses, data processing machines, and airplanes; examples of con- sumption goods are medicaments, small vehicles, and televisions; and examples of primary goods are iron ore, raw sugar, soybeans, and wheat. Table 1, Panel B reports estimates of Equation (1) restricting the sample to capital, consumption, intermediates, or primary goods. The association between commodity export intensity and concentration is smallest for capital goods in column 1. In contrast, markets for primary goods, in column 4, where average HHI is 5,964 and higher than average, a 1 percentage point increase in commodity export intensity is associated with a statistically significant increase in the HHI of 14.85 (2.55). Effects on consumption goods and intermedi- ates in Columns 2 and 3 are in between these extremes, and economically significant based on DOJ guidelines. An explanation for this result could be that primary goods are often the focus of trade policies restricting entry. For example, raw sugar and wheat imports are often subject to tight government control. In Nigeria, refined sugar imports are banned. Imports of raw sugar are dominated by two firms that import raw sugar into the country where it is refined 8 and sold (Premium Times, 2021). Wheat imports in many countries are handled by state monopolies (Ackerman and Dixit, 1999). In contrast, capital goods are typically less subject to entry restrictions as developing countries often do not produce them and thus rely mostly on foreign supply. In addition to primary goods, trade policies restricting entry can focus on consumption goods and materials when the intention is to substitute away from imports. 3 Import market concentration and welfare The resource curse manifests in higher costs and lower per capita consumption expenditure in commodity exporting economies.5 The question is whether import monopolization can account for this phenomenon. In theory, the relationship between concentration, costs, and expenditure is not obvious. If higher concentration is associated with higher fixed costs but lower marginal costs of importing, it could be associated with lower prices, even if also associated with higher markups (e.g., as in a differentiated products Nash-in-prices game). Alternatively, higher concentration could be associated with higher prices, if the associated markups outweigh any marginal cost savings. To distinguish between these hypotheses, we relate commodity export intensity and import market concentration to International Com- parison Program (ICP) data on domestic prices and per capita consumption expenditure. An advantage of the ICP is that measured prices and consumption are measured in local markets, and so capture a potential mitigating role of competition from domestic supply in product markets. We emphasize that this exercise is an accounting decomposition, not an attempt to estimate a causal relationship between prices and concentration. The industrial organization literature has long argued that such causal effect is not well-defined, because a variety of economic mechanisms can cause a (positive or negative) correlation between prices and concentration (see Miller, Berry, Scott Morton, Baker, Bresnahan, Gaynor, Gilbert, Hay, Jin, Kobayashi et al., 2022). Our exercise is simply to estimate the correlation between prices and concentration to infer whether the marginal costs associated with concentration outweigh the markups associated with concentration in determining equilibrium prices. The ICP reports the purchasing power parity price for a product category s in country c LCU (P P Pc,s ) as the ratio of the domestic price denominated in local currency units (Pc,s ) to $ LCU $ the price in the United States denominated in US dollars (PU SA,s ): P P Pc,s = Pc,s /PU SA,s . For every dollar spent in category s in the U.S., P P Pc,s local currency units are needed to purchase the same product in country c. We define the relative price level of the product 5 Sachs and Warner (2001) show, for example, that commodity export intensive economies had higher price levels relative to the global average in 1979. 9 LCU $ category in U.S. dollars as RelativeP ricec,s ≡ P P Pc,s /Ec = Pc,s /PU SA,s Ec , or the local price in U.S. dollars at the market exchange rate between country c’s local currency and the U.S. dollar (Ec ), divided by the U.S. price in U.S. dollars. The ICP also reports con- sumption expenditure per capita by country, which we use to test whether import market concentration is associated with lower consumption, as would be expected if it is associated with higher prices. Using these data we construct for each country c and product category LCU $ s a measure of relative per capita consumption RelativeConsumptionc,s ≡ Cc,s /CU SA,s Ec , LCU $ where Cc,s is local per capita consumption in local currency units and CU SA,s is U.S. per capita consumption in U.S. dollars. We focus on the 2011 and 2017 rounds of ICP data as those two years overlap available import concentration measures in most countries in our sample. The product categories for which the ICP reports PPP prices and consumption are much broader than the HS 4-digit classification. For instance, the ICP contains prices and consumption for two categories called “general purpose machinery” and “special purpose machinery,” whereas our data contain 135 unique HS 4-digit product categories within the HS 2-digit chapters related to machinery: HS84 “nuclear reactors, boilers, machinery and mechanical appliances; parts thereof” and HS85 “electrical machinery and equipment and parts thereof; sound recorders and repro- ducers, television image and sound recorders and reproducers, and parts and accessories of such articles.” Therefore, we relate relative prices and consumption in ICP broad categories to the HHI recalculated using the market shares of all firms importing any products in that broad category. We confirm that relative prices are higher and per capita consumption is lower in com- modity exporting countries. We estimate the regressions LCU Pc,s,t log $ = αc + τ {t = 2017} + β3 ExpComc,t + ϵc,s,t (2) PU SA,s,t Ec,t and LCU Cc,s log $ = αc + τ {t = 2017} + β3 ExpComc,t + ϵc,s,t (3) CU SA,s Ec where s is an ICP product category, β3 is the coefficient of interest, τ {t = 2017} is a fixed effect for year 2017, and αc are country fixed effects.6 Columns 1 and 2 in Table 2 confirm the resource curse: countries with a 1 percent higher commodity share of exports have 0.25 6 Country fixed effects ensure the identifying variation is relative prices and consumption within countries, as commodity prices change. We experimented with including country-product fixed effects. Point estimates were similar though standard errors were larger. Taking the log of the dependent variables also produced better fit due to right skew in the variables, especially relative consumption. 10 (0.10) percent higher prices and 0.45 (0.17) percent lower consumption per capita. We now examine whether import market concentration, which is higher in commodity exporting countries, can account for the higher prices and lower consumption observed in these countries. Cross-product correlations of concentration and prices or consumption can be misleading, since the relationship between concentration and marginal cost, an omitted variable, varies across products. To avoid this issue, we restrict our analysis to changes in concentration within product categories using two regressions LCU Pc,s,t HHIc,s,t = αs,c + τ {t = 2017} + β4 $ + ϵc,s,t (4) PU SA,s,t Ec,t and LCU Cc,s HHIc,s,t = αs,c + τ {t = 2017} + β4 $ + ϵc,s,t (5) CU SA,s Ec where αs,c is a country-product fixed effect. The coefficient β4 describes alternatively the correlation between relative prices or relative consumption per capita and import market concentration within country product markets. Column 3 of Table 2 reports estimates of equation 4 showing that rising domestic prices relative to the U.S. are associated with rising import market concentration between 2011 and 2017. The results are economically significant, with a 100 percent increase in price being associated with higher HHI of 197.47 (93.52), off a mean of 1,821. Consistent with this result, column 4 reports estimates for equation 5 that confirm increases in concentration are associated with lower consumption in addition to higher prices. The within-product market correlations indicate that markups associated with higher concentration appear to, on average, outweigh any lower marginal costs of importing. This is consistent with a model of a competitive world price and importers who are price-takers but have market power in local distribution. Importing firms in India, for example, have been shown to conform to this model (De Loecker, Goldberg, Khandelwal and Pavcnik, 2016). The panel association between import concentration and the symptoms of the resource curse (higher prices and lower expenditure) suggest import concentration could be a channel through which the curse materializes. 4 Trade policy mechanisms for import monopolization It remains to illustrate the mechanisms through which commodity exports cause import market concentration. Our argument is that vested interests in the import sector in some commodity exporting countries are able to capture the state and direct trade policy in ways 11 that reinforce import monopoly. Since this likely occurs in secret, there is little direct ev- idence of state capture for this purpose beyond anecdotes. Therefore we take an indirect approach. First, we demonstrate the direct effect of trade policy on import market concen- tration. Second, we demonstrate how specific types of export commodities that coincide with institutional weakness mediate the effect of trade policy on import market concentration. Isham et al. (2005) argue that rent-seeking associated with the resource curse is greatest when countries export ‘point-based’ resources, whose revenues transit directly through gov- ernment coffers, as opposed to ‘diffuse’ resources whose revenues flow to many small holders. Oil is the quintessential point-based commodity whose extraction is often controlled by the state and is associated with state capture (see, e.g., Ross, 2012). In contrast, production of food crops is diffuse outside of plantations. Ores and metals are an ambiguous case; although industrial extraction can be capital-intensive and thus point-based, labor-intensive artisanal mining with diffuse ownership can account for a substantial portion of output in some coun- tries due to variation in geography (Rigterink, Ghani, Lozano and Shapiro, 2022).7 Building on these ideas, our strategy to pinpoint a role for state capture in import monopolization is to examine whether the relationship between trade policy and concentration is stronger in hydrocarbon fuel exporting economies. To confirm the theory that ‘point-based’ resource exports lead to corruption in our sam- ple, Table 3, column 1 reports results from a regression of the control of corruption score from the Worldwide Governance Indicators (Kaufmann, Kraay and Mastruzzi, 2010) on the share of exports in three types of commodities: hydrocarbon fuels; ores and metals; and food.8 The regression uses the country-product-year sample from Table 1, but does not include country fixed effects to capture both the short-run and long-run relationships of commodity exports and control of corruption. Consistent with the theory of ‘point-based’ commodity exports, control of corruption is lower in fuel export intensive economies. In contrast, ores, metals and food exports are associated with stronger control of corruption. Next, we consider trade policies that could influence import market concentration. The obvious candidates are non-tariff measures (NTMs) that specifically restrict entry or pric- ing among importers: Chapters E “Non-automatic import licensing, quotas, prohibitions, quantity-control measures” and F “Price-control measures, including additional taxes and charge” as defined by UNCTAD (2019). Table 3, column 2 estimates a linear probability 7 These authors show that after positive commodity price shocks there is more violence in locations where industrial mining faces competition from artisanal mining compared to locations suitable for industrial mining only. 8 The control of corruption score is based on expert assessments and measured in standard deviations from the mean. The export shares in each type of commodity are measured in differences from the mean, without dividing by the standard deviation, so they are interpretable as the effect of a 1 percentage point change in exports of that commodity. 12 model where the left hand side variable equals one if such an NTM is present and zero otherwise. Very few products have such an NTM, as shown by the mean of the dependent variable. An increase in the fuel export share by 1 percentage point increases the likelihood of an NTM by 0.07 percentage points off a mean frequency of 1.18 percent. In contrast, increases in ores, metals and food exports reduce the likelihood of NTMs. Column 3 establishes the association between trade policy and market concentration. The specification includes HHI as dependent variable and export shares of the different commodities, a trade policy variable, the NTM in that market, and interactions of the NTM with export shares of the different commodities as regressors in 2018. Strikingly, the coefficient on the NTM is 17,516 (4,212) above the HHI variable range, whose maximum is 10,000 when a single monopolist imports a good (market share is 100, so HHI = 100 × 100). NTMs are associated with import monopolies, not just concentration of multiple firms. The interaction effects demonstrate that NTMs are associated with even greater concentration in commodity exporting countries. This could reflect different implementation of NTMs in commodity exporting countries. NTMs are gazetted regulations imposed by local authorities that create import monop- olies. The avoidance of regulation rather than adherence to it may also help to concentrate markets. It is well documented that tariff evasion is greater in the presence of higher tariffs (Bhagwati, 1964; Fisman and Wei, 2004; Mishra et al., 2008; Javorcik and Narciso, 2017), and that politically connected elites are uniquely able to evade tariffs (Rijkers et al., 2017). When a subset of firms evades tariffs, these firms obtain a cost advantage that allows them to increase their market share, increasing concentration. In column 4 one plus the ad- valorem import tariff is the dependent variable, with the shares of exports of different types of commodities as regressors. Unlike NTMs, tariffs are observed in all years. Even so, no country fixed effects are included in this specification for comparison with columns 1 and 2. Larger fuel and food export shares are associated with higher tariffs, but ores and metals are associated with lower tariffs if anything, suggesting a more nuanced picture than that implied by Figure 1b which showed a positive relationship between tariffs and the share of all commodities in total exports. To explore whether tariff evasion specifically could lead to importer concentration, column 5 uses the HHI as dependent variable and the three types of commodity exports, the tariff, and its interaction with the three types of commodity exports as regressors. Two findings emerge. First, tariffs are positively associated with importer concentration, with a 100 percent tariff being associated with an increase in HHI of 748 (157). This is not consistent with a perfectly competitive import market. If tariffs change the price for all firms equally, and demand is inelastic, nothing needs to trigger concentration of market share. However, if 13 demand is elastic, some firms may exit, increasing concentration. Alternatively if, under the tariff regime an individual firm is able to gain a cost advantage through evasion, its market share could increase, increasing concentration. The interaction terms suggest the positive effect of tariffs on concentration is arising from fuel exporting economies. A 1 percentage point increase in the fuel export share (relative to the mean) and a 100 percent increase in the tariff increases concentration by 6.8 (1.4). In contrast, these interaction terms are negative for ores, metals, and food. While NTMs have similar associations with importer concentration across economies, the association between tariffs and importer concentration is special to fuel exporting economies. Tariff evasion can be measured by comparing import and export mirror statistics. We use the measure of Fisman and Wei (2004) that allows for zeros in the value of exports, or the (exports − imports) evasion gap = (exports + imports) where exports are as reported by other countries in COMTRADE, and imports are as re- ported in our customs data. Column 6 uses HHI as dependent variable and the evasion gap term and its interaction with the export shares of different commodities as regressors, again including country fixed effects. As expected, more evasion is associated with greater importer concentration, as was the case for tariffs.9 A 100 percent increase in the evasion gap increases concentration by 2,407.4 (55.5), more than enough to warrant scrutiny under DOJ guidelines. More interesting is the significant interaction between the evasion gap and fuel exports equal to 22.2 (3.5). In contrast, the interaction with food exports is 11.4 (2.6) and with ores and metals exports 1.8 (2.9). These patterns match the pattern of interactions with tariffs, where the interaction has the largest positive coefficient for fuel exports. This suggests that tariff evasion can explain the positive association between tariffs and import concentration, especially in fuel export intensive economies. Taking stock, there is a nuanced relationship between import concentration, trade policy, and the varieties of commodities exported by countries. Non-tariff measures lead to import monopolies, and are most common in fuel exporting economies. Tariffs contribute to import market concentration, especially in fuel exporting economies and because of greater tariff evasion. Since control of corruption is weaker in fuel exporting economies, these results suggest that weak institutions mediate the effect of trade policy on import concentration. 9 As a sense check, we confirm that tariffs and evasion are positively correlated. In a similar specification with the evasion gap as dependent variable and tariffs as regressor controlling for product-year and country fixed effects and the different commodity export shares, a 100 percent increase in tariffs is associated with a 0.013 (0.0031) increase in the evasion gap. 14 5 Concluding remarks This paper identifies a novel channel for the ‘resource curse,’ the monopolization of imports. Commodity export intensity causes concentration of import markets, which can account for the higher price levels typically attributed to the Dutch disease. Trade policy measures and tariff evasion are mechanisms through which imports are monopolized. While economies’ export orientation has been the focus of the literature on trade and development, the role of imports and import market structure has been overlooked. While openness to imports is generally thought to increase competition in an economy, this effect could attenuate severely in the presence of importer market power, with implications for welfare. Further research could explore which domestic value chains could emerge from more competitive import markets, especially in fuel export intensive economies. As natural resource dependent economies such as those in Africa embark on regional integration, the lever of de-monopolization of imports could be of relevance as a method to increase the benefits of integration, and develop their domestic productive base. 15 Exhibits Figure 1: Import market concentration and trade protection in natural resource dependent countries (a) Import market concentration (b) Trade protection 7000 30 UGA PRY ZMB KEN TLS CPV BEN LAO BDI 6000 Average tariff in import markets MAR Average HHI in import markets 25 GAB CMR ETH GAB RWA ZAF SEN KHM MWI MDG URY ECU 5000 BWA NPL SEN TZA CPV BGD CMR ETH UGA COL RWA 20 MUS ZMB SLV BDI BGR GEO KEN CHN SLV ALB ALB TZA MWI MKD NPL PER 4000 BGD DOM GTM URY PRY KHM MUS ECU MDG CHL LKA BGR MAR PER GEO SRB EGY COL CHL BWA EGY LAO HRV 15 LKA 3000 ZAF DOM MKD MEX SRB TLS ROU GTM 16 MEX IND ROU HRV 2000 CHN 10 0 20 40 60 80 0 20 40 60 80 Commodity exports (as % of total merchandise exports) Average commodity exports (as a % of total merchandise exports) Notes: Values for each country are the simple mean across all years of the sample in Table A1. The slope of the best fit line in Panel A is 13.5 (standard error = 5.45), with an R-squared of 0.13. The slope of the best fit line in Panel B is 0.099 (0.026) with an R-squared of 0.26. Table 1: Import market concentration, commodity exports, and GDP per capita (1) (2) (3) (4) Panel A) HHI HHI Market share HHI of largest importer (%) Commodity export share ∈ [0, 100] 4.97** 6.31*** 0.0005*** (1.97) (2.33) (0.0002) Export commodity price index (100=2012) 31.17*** (7.70) Log(GDP per capita) -969.90*** -1,184.56*** -0.1103*** -1,305.17*** (182.37) (217.31) (0.0191) (226.97) R-squared 0.76 0.79 0.74 0.76 Observations 1,470,225 1,470,225 1,470,225 1,470,225 Dependent variable mean 3,224 4,101 52.06 4,101 HS digit product category 4 6 6 6 Country-product fixed effects Yes Yes Yes Yes Product-year fixed effects Yes Yes Yes Yes Panel B) HHI HHI HHI HHI Product type Capital Consumption Materials Primary Commodity export share ∈ [0, 100] 2.167 4.289*** 6.546*** 14.85*** (2.139) (1.095) (2.388) (2.550) Log(GDP per capita) -704.3*** -723.8*** -753.7*** -690.6*** (51.66) (28.55) (45.91) (47.88) R-squared 0.44 0.42 0.43 0.35 Observations 226,051 412,230 763,084 73,053 Dependent variable mean 3,480 3,528 4,459 5,964 HS digit product category 6 6 6 6 Country-product fixed effects Yes Yes Yes Yes Product-year fixed effects Yes Yes Yes Yes Notes: HHI is the sum of squared percentage import market shares for each country-product market with a maximum value of 10,000. Product markets are classified by either the 6 digit or 4 digit Harmonized System (HS) groupings. Product types correspond to the Broad Economic Categories of the HS. In Panel A, column 4, the export commodity price index is a country-year index of world commodity prices, where each commodity price is weighted by the long-run average share of the commodity’s exports in GDP (Gruss and Kebhaj, 2019). Standard errors clustered at the country-year level are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. 17 Table 2: Import market concentration, prices, and per capita consumption (1) (2) (3) (4) P LCU C LCU log P $E log C$E HHI HHI Commodity export share ∈ [0, 100] 0.0025** -0.0045** (0.0010) (0.0017) P LCU /P $ E 197.47** (93.52) C LCU /C $ E -40.71 (38.36) R-squared 0.1243 0.3202 0.90 0.90 Observations 3,883 3,869 2,716 2,722 Dependent variable mean -0.247 -2.608 1,821 1,819 Year fixed effects Yes Yes Yes Yes Country fixed effects Yes Yes No No Country-product fixed effects No No Yes Yes Notes: E is the exchange rate of local currency units per U.S. dollar. HHI is calculated pooling within a country importers of all goods in an ICP product category. Standard errors clustered at the country-year level are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Products are International Comparison Project broad product categories. 18 Table 3: Trade policy mechanisms for import market concentration (1) (2) (3) (4) (5) (6) Control of corruption NTM (=1) HHI Tariff HHI HHI Fuel export share ∈ [0, 100] -0.0075*** 0.0007*** -15.1 0.0004*** -1.8 28.3*** (0.0014) (0.0002) (14.1) (0.0001) (2.6) (4.0) Ores and metals export share ∈ [0, 100] 0.0104*** -0.0002 -10.1 -0.0001 12.2*** 11.8** (0.0023) (0.0001) (11.6) (0.0001) (4.5) (5.1) Food export share ∈ [0, 100] 0.0015 -0.0004* 20.6 0.0007*** 12.1*** 20.0*** (0.0017) (0.0002) (13.5) (0.0001) (3.5) (4.1) NTM (=1) 17,516.9*** (4,212.2) Fuel export share × (NTM = 1) 79.6*** (15.4) Food export share × (NTM = 1) 65.9*** (16.7) Ores and metals export share × (NTM = 1) 1,726.6*** (412.4) Tariff 748.5*** (157.2) Fuel export share × Tariff 6.8*** (1.4) Food export share × Tariff -2.7*** (1.0) Ores and metals export share × Tariff -1.2 (1.2) Evasion gap 2,407.4*** (55.5) Fuel export share × Evasion gap 22.2*** (3.5) Food export share × Evasion gap 11.4*** (2.6) Ores and metals export share × Evasion gap 1.8 (2.9) R-squared 0.1406 0.2192 0.5 0.9431 0.5 0.5 Observations 1,430,239 12,843 12,843 1,504,443 1,504,443 1,504,443 Dependent variable mean -0.227 0.0118 3,880 1.200 4,169 4,169 Product-year fixed effects No Yes Yes Yes Yes Yes Country fixed effects No No No No Yes Yes Notes: HHI is calculated pooling within a country importers of all goods in an HS 6 digit product category. Standard errors clustered at the country-year level in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Columns (2)-(3) are based on data for 2018 given that NTMs are available for that year only. 19 Appendix Table A1: Import market structure by country Commodity Fuel HHI for HS 6-digit products export export Country Start End min p25 p50 p75 max share (%) share (%) Albania 2007 2018 39.37 1673 3355 6505 10000 58.76 31.00 Bangladesh 2004 2015 32.07 1213 2950 6333 10000 8.570 1.873 Benin 2016 2018 103.8 3273 6046 10000 10000 37.32 2.579 Botswana 2004 2010 105.5 2101 4275 8214 10000 26.36 0.357 Bulgaria 2002 2006 36.29 1576 3363 6599 10000 42.03 13.28 Burundi 2010 2016 23.45 3416 6146 10000 10000 88.96 2.324 Cambodia 2016 2016 145.6 2463 4756 8736 10000 4.730 0.00786 Cameroon 2007 2017 13.81 1808 4039 7997 10000 76.94 58.68 Cabo Verde 2010 2018 62.21 1990 4086 8212 10000 87.09 0.000589 Chile 1998 2018 31.89 1164 2348 4821 10000 82.30 2.904 Colombia 1998 2018 37.51 1197 2374 4759 10000 80.14 69.64 Croatia 2007 2015 38.41 1038 2178 4739 10000 31.90 13.91 Dominican Republic 2002 2017 49.06 1470 3137 6154 10000 37.07 8.209 Ecuador 2002 2018 35.40 1331 2756 5540 10000 89.41 61.72 Egypt, Arab Rep. 2005 2016 6.380 1129 2431 5078 10000 65.31 56.39 El Salvador 2006 2018 91.59 1602 3313 6386 10000 29.59 3.554 Ethiopia 2008 2016 52.04 1847 4081 7994 10000 84.57 6.646 Gabon 2009 2009 87.33 2372 4851 9131 10000 86.92 83.13 Georgia 2000 2018 22.45 1461 3371 6968 10000 68.32 8.446 Guatemala 2005 2013 20.08 1525 3094 6021 10000 59.67 8.806 India 2016 2018 23.46 705.0 1529 3346 10000 28.72 14.93 Kenya 2006 2018 25.13 1709 3514 6675 10000 58.12 7.141 Lao PDR 2014 2016 355.2 3406 6116 9916 10000 69.40 0.251 Macedonia, FYR 2009 2017 91.91 1587 3229 6308 10000 31.33 8.715 Madagascar 2007 2017 116.3 2283 4633 8777 10000 67.93 6.960 Malawi 2005 2017 15.62 2417 4826 8810 10000 89.98 0.186 Mauritius 2000 2018 74.57 1705 3550 6817 10000 40.44 1.575 Mexico 2011 2016 42.50 783.8 1564 3337 10000 26.57 16.31 Morocco 2002 2010 46.50 1196 2459 5151 10000 35.06 2.819 Nepal 2011 2014 129.2 2145 4341 8207 10000 29.02 0.0127 Paraguay 2012 2018 78.28 1543 3111 6090 10000 90.17 31.21 Peru 2000 2018 53.32 1267 2574 5078 10000 87.78 16.06 Rwanda 2005 2016 50.60 2422 4942 8987 10000 94.49 0.240 Senegal 2000 2018 44.50 2180 4379 8375 10000 71.13 34.31 Serbia 2006 2007 50.35 1106 2383 4888 10000 33.84 3.507 South Africa 2010 2018 28.20 1080 2109 4214 10000 51.84 12.59 Sri Lanka 2016 2017 69.54 1336 2772 5424 10000 29.39 2.571 Tanzania 2003 2012 11.27 1829 4128 8122 10000 71.22 2.931 Uganda 2011 2018 27.33 1778 3806 7494 10000 74.47 6.678 Uruguay 2002 2018 53.54 1542 3128 5995 10000 67.56 4.855 Zambia 2010 2018 3.247 1797 3588 6766 10000 92.39 1.913 Note: US antitrust authorities generally consider markets in which the HHI is between 1500 and 2500 points to be moderately concentrated, and consider markets in which the HHI is in excess of 2500 points to be highly concentrated. 20 References Ackerman, Karen Z and Praveen M Dixit, “An introduction to state trading in agri- culture,” Agricultural Economic Reports 33909, USDA Economic Research Service 1999. Aghion, Philippe, Nick Bloom, Richard Blundell, Rachel Griffith, and Peter Howitt, “Competition and innovation: An inverted-U relationship,” The Quarterly Jour- nal of Economics, 2005, 120 (2), 701–728. as, Pol, Teresa C Fort, and Felix Tintelnot, “The margins of global sourcing: Antr` Theory and evidence from US firms,” American Economic Review, 2017, 107 (9), 2514– 2564. Arezki, Rabah, Valerie A Ramey, and Liugang Sheng, “News shocks in open economies: Evidence from giant oil discoveries,” The Quarterly Journal of Economics, 2017, 132 (1), 103–155. Benkard, C Lanier, Ali Yurukoglu, and Anthony Lee Zhang, “Concentration in product markets,” Working Paper 28745, National Bureau of Economic Research 2021. Bhagwati, Jagdish, “On the underinvoicing of imports,” Bulletin of the Oxford University Institute of Economics & Statistics, 1964, 27 (4), 389–397. Davis, Peter and Eliana Garc´ es, Quantitative techniques for competition and antitrust analysis, Princeton University Press, 2009. der Ploeg, Frederick Van, “Natural resources: curse or blessing?,” Journal of Economic Literature, 2011, 49 (2), 366–420. Edmond, Chris, Virgiliu Midrigan, and Daniel Yi Xu, “How costly are markups?,” Journal of Political Economy, forthcoming. Fernandes, Ana M, Caroline Freund, and Martha Denisse Pierola, “Exporter be- havior, country size and stage of development: Evidence from the exporter dynamics database,” Journal of Development Economics, 2016, 119, 121–137. Fisman, Raymond and Shang-Jin Wei, “Tax rates and tax evasion: evidence from “missing imports” in China,” Journal of Political Economy, 2004, 112 (2), 471–496. Frankel, Jeffrey A, “The natural resource curse: A survey of diagnoses and some prescrip- tions,” Commodity price volatility and inclusive growth in low-income countries, 2012, pp. 7–34. 21 Freund, Caroline, Rich People Poor Countries: The Rise of Emerging-Market Tycoons and their Mega Firms, Peterson Institute for International Economics, 2016. and Martha Denisse Pierola, “Export superstars,” Review of Economics and Statis- tics, 2015, 97 (5), 1023–1032. Gaubert, Cecile, Oleg Itskhoki, and Maximilian Vogler, “Government policies in a granular global economy,” Journal of Monetary Economics, 2021, 121, 95–112. Goldberg, Pinelopi Koujianou and Tristan Reed, “Is the global economy deglobal- izing? If so, why? And what is next?,” Brookings Papers on Economic Activity, March 2023. Gruss, Bertrand and Suhaib Kebhaj, Commodity terms of trade: A new database, International Monetary Fund, 2019. Isham, Jonathan, Michael Woolcock, Lant Pritchett, and Gwen Busby, “The vari- eties of resource experience: natural resource export structures and the political economy of economic growth,” The World Bank Economic Review, 2005, 19 (2), 141–174. Javorcik, Beata S and Gaia Narciso, “WTO accession and tariff evasion,” Journal of Development Economics, 2017, 125, 59–71. Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi, Worldwide Governance Indicators, World Bank, 2010. Kee, Hiau Looi and Alessandro Nicita, “Trade fraud and non-tariff measures,” Journal of International Economics, 2022, 139, 103682. Krueger, Anne O, “The political economy of the rent-seeking society,” The American Economic Review, 1974, 64 (3), 291–303. Leone, Fabrizio, Rocco Macchiavello, and Tristan Reed, “The Falling Price of Ce- ment in Africa,” Policy Research Working Paper 9706, World Bank 2021. Loecker, Jan De, Pinelopi K Goldberg, Amit K Khandelwal, and Nina Pavcnik, “Prices, markups, and trade reform,” Econometrica, 2016, 84 (2), 445–510. Miller, Nathan, Steven Berry, Fiona Scott Morton, Jonathan Baker, Timothy Bresnahan, Martin Gaynor, Richard Gilbert, George Hay, Ginger Jin, Bruce Kobayashi et al., “On the misuse of regressions of price on the HHI in merger review,” Journal of Antitrust Enforcement, 2022, 10 (2), 248–259. 22 Mishra, Prachi, Arvind Subramanian, and Petia Topalova, “Tariffs, enforcement, and customs evasion: Evidence from India,” Journal of Public Economics, 2008, 92 (10- 11), 1907–1925. Nocke, Volker and Michael D Whinston, “Concentration Thresholds for Horizontal Mergers,” American Economic Review, 2022, 112 (6), 1915–48. Premium Times, “How Nigeria’s sugar production collapsed for decades, allowing Dangote, BUA imports to thrive,” Apr 2021. Rigterink, Anouk, Tarek Ghani, Juan Lozano, and Jacob Shapiro, “Mining Com- petition and Violent Conflict in Africa: Pitting Against Each Other,” University of Wash- ington mimeo, 2022. Rijkers, Bob, Leila Baghdadi, and Gael Raballand, “Political connections and tariff evasion evidence from Tunisia,” The World Bank Economic Review, 2017, 31 (2), 459–482. Robinson, James A, Ragnar Torvik, and Thierry Verdier, “Political foundations of the resource curse: A simplification and a comment,” Journal of Development Economics, 2014, 106, 194–198. Ross, Michael L, “The oil curse,” in “The Oil Curse,” Princeton University Press, 2012. Sachs, Jeffrey D. and Andrew M. Warner, “The curse of natural resources,” European Economic Review, May 2001, 45 (4-6), 827–838. Sequeira, Sandra, “Corruption, trade costs, and gains from tariff liberalization: Evidence from Southern Africa,” American Economic Review, 2016, 106 (10), 3029–63. Tornell, Aaron and Philip R Lane, “The voracity effect,” American Economic Review, 1999, 89 (1), 22–46. Tullock, Gordon, “The welfare costs of tariffs, monopolies, and theft,” Economic inquiry, 1967, 5 (3), 224–232. UNCTAD, “International Classification of Non-tariff Measures,” Technical Report 2019. , “Key Statistics and Trends in International Trade 2020,” 2021. Venables, Anthony J, “Using natural resources for development: why has it proven so difficult?,” Journal of Economic Perspectives, 2016, 30 (1), 161–84. Yang, Dean, “Can enforcement backfire? Crime displacement in the context of customs reform in the Philippines,” Review of Economics and Statistics, 2008, 90 (1), 1–14. 23