Structural Change and Economic Dynamics 59 (2021) 63–78 Contents lists available at ScienceDirect Structural Change and Economic Dynamics journal homepage: www.elsevier.com/locate/strueco The distributional dimension of the resource curse: Commodity price shocks and income inequality Soran Mohtadi a, David Castells-Quintana b,∗ a Department of Applied Economics, Universidad Autónoma de Barcelona, 08193 Bellaterra, Barcelona, Spain b Department of Applied Economics, Universidad Autónoma de Barcelona, 08193 Bellaterra, Barcelona, Spain a r t i c l e i n f o a b s t r a c t Article history: How does high dependence on natural resources affect income inequality? Surprisingly little is known Received 30 March 2021 about the impact of dependence on primary goods on income distribution. Building on insights from Revised 19 July 2021 the resource curse literature, this paper studies the relationship between income shocks through changes Accepted 7 August 2021 in commodity prices and income inequality in a panel of 80 countries from 1990 to 2016. We analyze Available online 15 August 2021 differentiated effects of commodity price shocks depending on the type of commodity (labor vs. capital- JEL classification: intensive). We also study differences across world regions and explore potential mechanisms by looking O13 at different types of inequality (pay-driven vs. capital-rents-driven). Results show that commodity price O15 shocks have an impact on income inequality. However, this impact depends on the type of commodity Q33 and inequality. © 2021 The Authors. Published by Elsevier B.V. Keywords: This is an open access article under the CC BY-NC-ND license resource curse commodity price shocks (http://creativecommons.org/licenses/by-nc-nd/4.0/) inequality natural resources resource booms development 1. Introduction ity in terms of trade and foreign direct investment, low rates of economic growth and higher socio-political instability (see A high dependence on natural resources is characteristic of Acemoglu et al., 2003; Van der Ploeg and Poelhekke, 2009; Sala-i- many developing countries, especially the least developed ones, Martin and Subramanian, 2013; Ferraro and Peretto, 2018).3 Never- and has been of interest for both academics and policy makers.1 theless, little is known about the potential impact of changes in in- Research into the (natural) “resource curse” suggests that high de- ternational commodity prices on the evolution of income inequal- pendence on natural resources, especially under low institutional ity. levels, can slow structural change and hinder economic develop- In this paper, we analyze the distributional dimension of the re- ment (see for instance, Caselli and Tesei, 2016; Boschini et al., source curse by studying the connection between resource booms 2013; Bazzi and Blattman, 2014).2 High volatility in commodity due to changes in international commodity prices and the evolu- prices has been found to (partially) explain many of the prob- tion of income inequality within countries. In doing so, we build lems associated with the “resource curse”, including high volatil- a unique dataset looking at 23 commodities and the evolution of their international prices, as well as export shares of these com- modities for 80 countries worldwide from 1990 to 2016 (based ∗ Corresponding author. on data availability). With these data, we construct country-period E-mail addresses: soran.mohtadi@e-campus.uab.cat (S. Mohtadi), specific commodity price shock, and relate these shocks to the david.castells.quintana@uab.cat (D. Castells-Quintana). evolution of income inequality in each of the 80 countries. We 1 For recent studies on income inequality on developing countries see for in- stance Addison et al., 2017; Castells-Quintana, 2019; Gradin et al., 2021. 2 According to data, in least developed countries primary goods (i.e., commodi- 3 ties) still represent a big fraction of the economy (in some case up to 78%). Simi- For some regions, in the last decades, this volatility has represented a sig- larly, for three-quarters of all states in Sub-Saharan Africa and two-thirds of those in nificant cycle of boom and collapse. For instance, in Latin America, commodity Latin America, the Caribbean, North Africa, and the Middle East, primary commodi- export prices increased during first decade of 21st century, but have declined ties still represent around half of their export income. In some cases, the figure goes sharply recently, which may be contributing to current social unrest in the region up to 96%. (Haslam, 2016, Siegel, 2016). https://doi.org/10.1016/j.strueco.2021.08.002 0954-349X/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 differentiate between commodities based on their factor inten- Section 5 concludes and derives policy implications and avenues sity (i.e., labor vs. capital). Positive price shocks on labor-intensive for further research. commodities are expected to reduce inequality by potentially in- creasing demand for (low-skilled) labor. By contrast, positive price 2. Commodity price shocks and inequality: A brief literature shocks on capital-intensive commodities could increase inequality review by potentially favoring rent-seeking. In this regard, we consider two types of inequality – pay- vs. capital-rents-driven inequality – 2.1. Natural resources, commodity price shocks and income inequality to study these differentiated mechanisms for price shocks to affect income inequality differently based on the type of commodity. The socio-economic consequences of high dependency on nat- In relation to existing studies, this paper is linked to several ural resource have been widely investigated (see, for instance, strands in the economics literature. First, our work relates to those Acemoglu et al., 2003; Van der Ploeg and Poelhekke, 2009; in the traditional study of the “resource curse” (see Sachs and Fum and Hodler, 2010; Arezki and van der Ploeg, 2010; Sala- Warner, 1999; Papyrakis and Gerlagh, 2004; Arezki and van der i-Martin and Subramanian, 2013; Carmignani and Avom, 2010; Ploeg, 2010; Papyrakis, 2014). None of these papers focuses on in- Bazzi and Blattman, 2014; Kim and Lin, 2018; Behzadan et al., come distribution. Second, our paper is closely linked to studies on 2017). The connection between high specialization in primary the relationship between natural resources and income inequality goods and patterns of economic development has been exten- (Gylfason and Zoega, 2002; Fum and Hodler, 2010; Carmignani and sively studied (Carmignani and Avom, 2010; Kim and Lin, 2018; Avom, 2010; Parcero and Papyrakis, 2016; Behzadan et al., 2017; Behzadan et al., 2017). In a globalized world, high dependency on Kim and Lin, 2018; Davis, 2019; Kim et al., 2020 for world natural resources translates into high dependency on international samples, Leamer et al., 1999, for Latin America, Farzanegan and commodity prices. And in recent decades, international commod- Krieger, 2018, for Iran). While these papers study the natural- ity prices have shown high volatility (Van der Ploeg and Poel- resources-inequality relationship, these papers usually look at one hekke, 2009). or two resources and do not consider commodity prices. Finally, For commodity-dependent countries, these changes in interna- our paper relates to those in the conflict literature focusing on tional prices can represent massive shocks. For every country, the commodity price shocks and showing differentiated effects on extent of a shock depends on the array of commodities exported conflict and civil war depending on the factor intensity of the and on the share of each commodity in the countryś total exports. commodity (see Dube and Vargas, 2013, for Colombia, Bazzi and Thus, the interplay of changes in international commodity prices Blattman, 2014; Ciccone, 2020, for world samples).4 and each countryś export shares defines country-specific com- To the best of our knowledge, only two previous papers have modity price shocks. Shocks have the potential to influence sev- analyzed the connection between commodity price shocks and dis- eral socio-economic outcomes, including foreign direct investment, tributional issues, specifically: Goderis and Malone (2011), looking trade flows, economic growth, and even socio-political stability at pay inequality in manufacturing sectors for the period 1965- (see Acemoglu et al., 2003; Van der Ploeg and Poelhekke, 2009; 1999, and Bhattacharyya and Williamson (2016), looking at income Sala-i-Martin and Subramanian, 2013). Recently, resource rents inequality in Australia. We contribute to the literature by i) ana- from higher commodity export prices have also been associated lyzing the effects of commodity price shocks on income inequality with populism and authoritarian institutions (Seghezza and Pit- looking at 23 commodities and taking a global view, ii) providing taluga, 2018). Commodity price shocks can also have potential ef- evidence of opposing effects of commodity price shocks on the dis- fects on employment, its distribution across different sectors, and tribution of income depending on the type of commodity, and iii) thus on structural change, as well as on wages across the econ- studying potential mechanisms for these differentiated effects to omy. Consequently, commodity price shocks could be expected to take place. have potential effects on the distribution of income within coun- We show that non-agricultural (capital-intensive) price shocks tries. But the direction of the impact of a commodity price shock are significantly associated with increases in inequality. This result on income inequality is not straightforward. is robust to a long list of controls, different specifications and es- On the one hand, higher commodity prices can lead to less in- timation techniques, as well as to multiple robustness checks. We come inequality; rising prices for commodity exports can increase find that the inequality-increasing impact of non-agricultural price the demand for (low-skilled) labor, leading to higher wages and a shocks is mostly felt in countries with high initial levels of inequal- more equal distribution of income. On the other hand, commodity ity and low institutional quality, as is the case of many countries price shocks can lead to more inequality; higher commodity prices in Sub-Saharan Africa (SSA) and Latin America (LA). We also find generate rents that can be appropriated by few, usually already suggestive evidence that agricultural (labor-intensive) commodity rich, individuals. Previous papers have already noted that natural price shocks reduce pay inequality, while non-agricultural (capital- resource rents can increase the gap between the rich and the poor intensive) price shocks increase capital-income inequality. (Ross, 1999), deteriorating the distribution of income (Kim and The remainder of this paper is organized as follows. Lin, 2018). However, global evidence in this regard is very limited Section 2 reviews the relevant literature. Section 3 describes to date (i.e., Goderis and Malone, 2011).5 the data used to study the relationship between commodity price shocks and inequality. Section 4.1 presents the main em- pirical approach, including descriptive and econometric analysis, 2.2. Commodity price shocks and the “opportunity cost” and while section 4.2 explores mechanisms for differentiated ef- “rapacity” effects fects of commodity price shocks on income inequality. Finally, Given potential differences in the mechanisms for higher com- modity prices to influence income distribution, we can expect 4 This conflict literature suggests that inequality might be a key factor con- 5 necting commodity shocks and higher risk of conflict (see for instance Bazzi and Goderis and Malone (2011) provide some cross-country evidence, looking at the Blattman, 2014). In this and related literature, institutions are shown to play a relationship between natural resource booms (commodity exports) and pay inequal- relevant role (see Mehlum et al., 2006; Cabrales and Hauk, 2011; Musayev, 2014; ity for the period 1965-1999. They find that oil and mineral booms reduce inequal- Caselli and Tesei, 2016; Krieger and Meierrieks, 2016). By performing a heterogene- ity in the year of the boom. Lessman and Steinkrauss (2019) study the association ity analysis at the global level, we also connect to these papers. between the spatial distribution of natural resources and spatial inequality. 64 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 that the impact of commodity price shocks on inequality will de- been able to appropriate most of the rising rents from natural re- pend on the type of commodity, in particular, on its factor in- sources (Bjorvatn and Naghavi, 2011). The role of political institu- tensity. Omitting differentiated impacts depending on the type of tions has been widely studied in the resource curse literature (see, commodity may explain the inconclusive results to date. If the for example, the literature cited in Van der Ploeg, 2011). Countries inequality-decreasing effect of higher commodity prices is associ- with weak institutions are more prone to conflict (Musayev, 2014; ated with higher employment opportunities and wages, we can Caselli and Tesei, 2016), and weak institutional settings also tend expect to see the inequality-decreasing effect mostly when posi- to correlate with high levels of inequality (Krieger and Meier- tive shocks take place in labor-intensive commodities. By contrast, rieks, 2016). In these countries, natural resource booms tend to if the inequality-increasing effect is associated with higher rents, lead to lower levels of development (Caselli and Tesei, 2016; we can expect to see the inequality-increasing effect mostly when Bazzi and Blattman, 2014). Similarly, in countries with weak insti- positive shocks take place in capital-intensive commodities. tutions, the tax system is usually also weak. It is therefore normal The relevance of differentiating commodities depending on to expect that a potential inequality-increasing impact of commod- their factor intensity has already been highlighted in the resource ity price shocks to affect countries with weak institutions more curse literature. Boschini et al. (2013) suggest that “point-source severely. resources” (such as plantation crops, minerals and fuels), compared To sum up, we expect that the effect of commodity price shocks to “diffused resources” (such as other labor-intensive agricultural on income inequality will depend on i) the type of commodity, ii) products), are expected to be more centrally controlled and there- the type of inequality, and iii) the institutional context. For labor- fore generating rents that are more easily appropriable by few intensive commodities, higher prices are expected to reduce in- individuals. Resource rents have also been associated with lower equality through lower pay inequality. For capital-intensive com- fiscal capacity (Crivelli and Gupta, 2014), and therefore the abil- modities, higher prices are expected to increase inequality through ity of the state to redistribute income. In this line, Auty and Fur- higher capital-rents inequality. And this last effect is expected to longe (2019) discuss the “so-called” rent curse; the idea that high be more pronounced in countries with a weak institutional setting rents encourage rent siphoning for immediate enrichment, which (and higher initial levels of inequality). can lead to higher inequality and lower long-run growth. Contrary to what happens with the relationship between com- 3. Data modity price shocks and income inequality, the relationship be- tween price shocks and conflict has been more extensively studied Our empirical analysis is based on a unique (unbalanced) panel (see for instance Dube and Vargas, 2013; Bazzi and Blattman, 2014; dataset consisting of 80 countries over the period from 1990 to Ciccone, 2020). According to Dube and Vargas (2013), a shock 2016. These 80 countries concentrate most of the world exports which raises wages will reduce conflict by decreasing labor sup- of the commodities studied (up to 82% in the case of coffee and plied to appropriation activities. This wage mechanism is defined 89% in the case of oil). We study 23 highly traded commodities, as an “opportunity cost effect”. By contrast, a shock which raises collecting data on the evolution of their international prices, and the return to appropriation will increase conflict by increasing la- looking at what these commodities represent as a share of total bor supplied to the conflict sector and rents captured by a few in- exports for each of our 80 countries.6 Using these data, we con- dividuals. This appropriation mechanism is defined as a “rapacity struct commodity price shocks for every country-year observation, effect”. and combine these with data on the evolution of inequality in each Inequality and conflict are interrelated. Greater income in- country in our sample. equality has been associated with higher risks of civil war on- set (Fearon and Laitin, 2003), state violence and coups d’état 3.1. Inequality (Galbraith and Purcell, 1999). Hence, workers in a conflict-prone society may choose between a productive sector and a criminal or Our main dependent variable is income inequality. Data for in- illegal one. Therefore, in countries with abundance of natural re- come inequality for several countries and for a long time span is sources and lower quality of institutions we may observe higher scarce. To overcome this limitation, we use Gini coefficients from levels of inequality as well as higher propensity of conflict (Le Bil- the SWIID (Standard World Income Inequality Database) version lon, 2014). Our analysis is therefore connected to that on conflict, 6.1 (Solt, 2016).7 The SWIID dataset is the most comprehensive and the impacts of commodity price shock on inequality can be dataset on inequality providing a very wide coverage of compa- understood as another dimension of the “resource curse”. rable inequality data across countries.8 For our empirical analysis, and as discussed in Section 2, we 2.3. Two types of inequality and the role of institutions further consider two components of income inequality. First, we Beyond differentiating commodities based on their factor inten- 6 The commodities that are analyzed includes, oil, gas, coal, gold, diamond, silver, sity, it may also be relevant to distinguish between the types of zinc, aluminum, iron, copper, tin, nickel and lead, coffee, cocoa, rice, sugar, banana, inequality: pay vs. capital-rents-driven inequality. If the reduction wheat, cotton, wool, wood and rubber, which contain more than 75% of all com- modities that have been exported in year 2016, according to International Trade of inequality due to commodity price shocks is related to higher Statistics Yearbook (2016). employment opportunities and higher wages for the low-skilled la- 7 SWIID uses a custom missing-data multiple-imputation algorithm to standard- bor, we should expect this to be reflected in lower pay inequality. ize observations collected from multiple sources (i.e., UTIP; OECD Income Distribu- By contrast, if the increase in inequality due to commodity price tion Database; The Socioeconomic Database for Latin America and the Caribbean shocks happens due to higher rents, we should expect this to be generated by CEDLAS and the World Bank; Eurostat; the UN Economic Commis- sion for Latin America and the Caribbean; national statistical offices around the reflected mainly in a more unequal distribution in capital rents. world, and many other sources). SWIID recently updated the database to version According to the literature, the potential inequality-increasing 9.0. We check our results to using this latest version. Recent studies using SWIID effect of higher commodity prices is also likely to depend on the data include Parcero and Papyrakis (2016); Kim and Lin (2018); Gylfason (2019); institutional context. According to Ross (1999), the connection be- and Hartwell et al. (2019). 8 tween natural-resource rents and increasing gaps between the rich We check the results using household disposable post-tax income data. Inequal- ity after the political processes of rent-seeking and redistribution may be more rel- and the poor heavily depends on the presence of weak institu- evant for policy makers. However, it is important to note that, for developing coun- tions. Indeed, in many resource-rich countries with weak qual- tries, surveys suggest little difference between pre- and post-tax inequality. Results ity of institutions, local elites, together with foreign capital, have are similar using either pre- or post-tax data. 65 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 consider pay inequality using data from the University of Texas Inequality Project (UTIP).9 Based on the United Nations Industrial Development Organization (UNIDO), pay inequality measures in- equality in wages and earnings in the manufacturing sector (as used for instance by Drehen and Gaston, 2008; Tan and Law, 2012; Doerrenberg and Peichl, 2014; Kim and Lin, 2018; Kim et al., 2020; Law et al., 2020). Using this data enables us to analyze the inequal- ity among the employed individuals, and observe how commodity price shock impact on labor wages inequality depending on the type of commodity (labor vs capital intensive). For inequality in capital-rents there is no available data for international compar- isons. What we do is therefore to calculate the difference between the household disposable income inequality and pay inequality. This gives us a (crude) measure to study the impacts of commodity shocks on (a proxy of) capital-income-driven inequality.10 3.2. Commodity Price Shocks Figure 1. The annual evolution of commodity prices Our key explanatory variable is a country- period-specific mea- sure of resource booms, which we calculate as an export-share- weighted commodity price shock. We construct this measure lowing Ciccone, 2020).11 Annual shocks are calculated as the log using a similar methodology to previous papers, as Bazzi and difference of the price index Pit , and scaled by the weight of to- Blattman (2014), Musayev (2014) and Castells-Quintana (2017). To tal commodity exports to GDP -a time-invariant measure of the calculate these country- period-specific commodity price shocks, importance of commodity prices in the economy for country i; we rely on i) data on international commodity prices for every more commodity-dependent nations are obviously more sensitive year in our period of analysis, ii) commodity exports for every one to commodity price shocks. 12 of our 23 commodities from every country and in every year in Hence, in Equation (2), Sit is calculated as the annual difference our sample, and iii) total GDP for every country-period observa- in each country’s log commodity export price index: tion. Data for international prices for our 23 commodities is col- XiT Sit = (log Pit − log Pit−1 ) ∗ (2) lected from the IMF-IFS International Financial Statistics, the World GDPiT Bank, the FRED Federal Reserve Economic Data and the World Gold The measurement of commodity price shocks using export Council. For commodity exports, we use data from the UNCTSD shares of commodities has several advantages, including wide data (United Nations Commodity Trade Statistics Database). With the availability and the fact that international commodity prices are UNCTSD data, we calculate shares of the 23 different commodities typically not affected by individual countries, and therefore are not in total exports by country and year. For GDP, we rely on data from likely to be endogenous with respect to the growth of individual the World Development Indicators (World Bank). countries (something we also consider in our robustness checks). The commodity price shock is calculated from a commodity ex- To test our prediction that the effect of commodity price shocks port price index, Pit , as a geometrically-weighted index of interna- depends on the factor intensity of the commodity, and follow- tional export prices for country i in period t: ing the literature, we distinguish between agricultural vs. non- J w agricultural commodities. As shown by previous papers, agricul- j=1 Pjt i jt−k Pit = (1) tural commodities are, on average, labor-intensive - for instance cpit coffee, cocoa, rice, banana, cotton, wool, wood and rubber - while non-agricultural commodities are, on average, capital intensive - Where Pjt captures prices on international markets for commod- for instance hydrocarbons and minerals (see Goderis and Mal- ity j in period t (normalized to 100 in 2010). Since prices are one, 2011; Musayev, 2014; Bazzi and Blattman, 2014). dollar-denominated, the index is deflated by the US consumer Figure 1 illustrates the annual evolution of agricultural and non- price index, cpit . Each commodity price is weighted by wi jt−k , agricultural global prices from 1990 to 2016. As shown, the main its average share in total national exports (excluding re-exports) increase in global prices happened between 2002 and 2011 (with a from t-2 to t-4, to reduce potential reverse causality (Following previous increase for agricultural commodities in 1995). From 2011 Bazzi and Blattman, 2014). For robustness, we also check using onwards, commodities display a decrease in their global prices. time-invariant (fixed) export shares as commodity weights (fol- 3.3. Conflict and other controls 9 The Theil index as a measure of pay inequality is calculated from industrial As the literature has mainly focused on the impacts of com- sector statistics provided by the United Nations Industrial Development Organiza- modity prices shocks on conflict, and we have seen how conflict tion (for more information on the underlying data and construction of the index, see Conceição et al., 2001; Galbraith and Kum, 2005, and Galbraith et al. 2014). may be associated with income inequality, we consider intentional 10 We follow the idea from recent analysis of income sources using micro data. homicides (per 10 0,0 0 0 people), from the UN Office on Drugs and We calculate the difference between the SWIID Gini (0-100) and UTIP-UNIDO, mul- Crime’s Intentional Homicide Statistics database. tiplied by 100. Our proxy for capital-income-driven changes in inequality relies on the within-country evolution of this difference. We acknowledge that this is a very 11 basic proxy and results using this measure should be taken with caution. For ro- Ciccone (2020) claims that time-varying export shares partly reflects changes in bustness, as explained below, we check our results calculating our proxy for capital- the quantity and variety of countries’ exports, which can jeopardize causal estima- income inequality in three different ways: i) as the difference between SWIID dis- tion. 12 posable income inequality and UTIP-UNIDO pay inequality, ii) as the difference be- The average of the ratio is taken in 1990 to 2016 to calculate X/GDP for each tween SWIID market income inequality and UTIP-UNIDO pay inequality, or iii) as country. This scaling increases the expected size and precision of any impact of the difference between EHII income inequality and UTIP-UNIDO pay inequality. prices on growth and political instability (Bazzi and Blattman, 2014). 66 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table 1 Table 2.b Descriptive statistics, main variables Conflict and commodity price shocks Variables Mean Std. Dev. Min Max Observations No Year FE Year FE Income overall 38.62 8.09 20.21 58.45 N=1869 No country FE -0.054∗ -0.053 in- countries=80 Country FE -0.054∗ -0.051 equal- T=23.36 ∗ ity Income overall 0.01 0.42 -2.09 2.30 N=1789 Note: The panel includes 947 observations. (levels) in- countries=80 p<0.1. equal- T=22.36 ity Commodity overall 0.2 0.2 -1.21 2.99 N=1760 Table 2.c (changes) price countries=80 Inequality and commodity price shocks shocks T=22 No Year FE Year FE Conflict overall 8.17 13.41 0.1 93.2 N=1044 (levels) countries=80 No country FE 0.042∗ 0.031 T=13.05 Country FE 0.042∗ 0.031 Conflict overall -0.04 2.21 -14.9 36.20 N=924 (changes) countries=68 Note: The panel includes 1528 observations. ∗ T=13.58 p<0.1. Quality overall 66.33 13.72 19.16 96.08 N=1680 of countries=71 Table 3.a Institutions T=23.66 Inequality and agricultural price shocks No Year FE Year FE Table 2.a No country FE 0.008 -0.005 Inequality and conflict Country FE 0.008 -0.005 No Year FE Year FE Note: The panel includes 1528 observations for 80 countries. ∗ p<0.05. No country FE 0.045 0.056∗ Country FE 0.045 0.055∗ Table 3.b ∗ Inequality and non-agricultural price shocks Note: The panel includes 889 observations. p<0.1. No Year FE Year FE No country FE 0.049∗ 0.043∗ Country FE 0.049∗ 0.043∗ We also consider other several variables relevant to explain in- Note: The panel includes 1528 observations come inequality at the country level, including economic growth for 80 countries. ∗ p<0.05. rates, income per capita (in logs), the share of investment, the share of government spending, and the average years of school- ing. For robustness, and following the literature, additional con- fects.13 The introduction of country fixed effects allows us to con- trol variables are also included: total population, fertility rates, and trol for country-specific characteristics, while the introduction of the quality of institutions. Finally, variables that may correlate with year fixed effects allows us to control for global shocks.14 A pos- commodity price shocks, such as trade openness and the inflation itive correlation between commodity price shocks and inequality rate, are also considered. Table A.1 in the Appendix, lists all vari- is found; price shocks are positively associated with higher in- ables definitions and sources, while descriptive statistics for main equality. There is a positive association between inequality and variables, as well as list of countries included in the analysis, can conflict, in line with the literature (Collier and Hoeffler, 2004; be found in Tables A.2 and A.3 in the Appendix. Fearon and Laitin, 2003; Fearon, 2008; Blattman and Miguel, 2010; Esteban and Ray, 2011). Higher quality of institutions is associ- ated with lower income inequality (in line with Parcero and Pa- pyrakis, 2016) and lower risk of civil war and conflict (in line with 4. Inequality and commodity price shocks: an empirical Caselli and Tesei, 2016; Musayev, 2014). analysis Following our discussion in Section 2, in Table 3, the com- modities considered are disaggregated into agricultural and non- 4.1. Descriptive analysis agricultural.15 Table 3.a shows a negative association between in- equality and agricultural price shocks, while Table 3.b shows a Table 1 provides descriptive statistics for income inequality, positive association between inequality and non-agricultural price commodity price shocks, conflict, and the quality of institutions. In shocks, even after controlling for country and year fixed effects. our sample, the average level of inequality, measured by the Gini Finally, we also explore differences in the relationship between coefficient (from 0 to 100) is 38.6. For commodity price shocks, commodity price shocks and income inequality across world re- a value above zero indicates that the country faces higher com- gions: Europe (EU), North America (NA), Asia (A), Oceania (OC), modity export prices. According to Table 1, the mean of overall Latin America (LA), Sub-Saharan Africa (SSA), and the Middle East commodity price shocks for our sample is 0.2 (or 20 per cent), (ME). Table A.6 in Appendix A shows average values for income indicating that over the 1990-2016 period our 80 countries faced, distribution and quality of institutions for each of these world re- on average, more positive shocks in commodity export prices than gions. As expected, countries in SSA and LA have, on average, the negative ones. Looking at specific countries, we find a connection between high inequality levels and low quality of institutions (see 13 Table A.4), as previously highlighted in the literature (Chong and Table A.4 in the Appendix, shows simple correlations (i.e., without fixed effects) for our key variables. Gradstein, 2007). 14 The binned scatter plots have been applied based on all data points to purge Tables 2.a, 2.b, and 2.c show the pairwise correlations be- from year and country fixed effects. Here, all observations are grouped into 20 bins. tween our main variables (commodity price shocks, inequality, We also checked the data and these points are not the individual outlier countries. conflict, and institutions) controlling for year and country fixed ef- 15 Table A.5 lists all commodities. 67 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Figure 2.a. Inequality and agricultural price shocks in SSA and LA Figure 3.a. Inequality and agricultural price shocks in the rest of the world highest inequality levels and the lowest quality of institutions. As data (1990-2016), we consider alternatively year-to-year variation Figures 2.a and 2.b show, in SSA and LA, the regions with the worst as well aggregating over 3-year periods. αi controls for time- income distribution and lowest institutional quality, commodity invariant country-specific characteristics. Period-specific effects, δt , price shocks in agricultural commodities are negatively associated are also included to control for common global shocks. The key with changes in inequality, while shocks in non-agricultural com- independent variable is Sit , the annual commodity prices shock in modities are positively associated with changes in inequality. By country i in period t. Finally, εit is an idiosyncratic error term. The contrast, we find no significant association in the rest of the world coefficient of interest is β1 , which captures the relationship be- (see Figures 3.a and 3.b). tween commodity price shocks and income inequality. In a second step, we differentiate commodities based on their 4.2. Econometric analysis labor intensity. Thus, we analyze whether changes in prices of agri- cultural (i.e., labor-intensive) commodities affect income inequality In this section, we consider a simple empirical model that al- differentially than non-agricultural (i.e., capital-intensive) ones: lows us to test the relationship between resource booms and in- come distribution. In particular, the association between commod- ity price shocks and income inequality is analyzed using the fol- inequalit yit = αi + δi + γ1 agriit + γ2 nonagriit + it (4) lowing specification: inequalit yit = αi + δt + β1 Sit + εit (3) where inequalit yit stands for changes in household income in- where agr iit stands for agricultural price shocks and nonagr iit equality in country i in period t. Given the long coverage of our stands for non-agricultural price shocks. 68 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table 4 Main results (1) (2) (3) (4) (5) (6) (7) (8) Commodity price shocks 0.07∗ ∗ 0.05 0.02 0.03 (0.03) (0.03) (0.04) (0.05) Agricultural price shocks -0.01 -0.01 -0.12 -0.18∗ (0.08) (0.1) (0.07) (0.1) Non-agricultural price shocks 0.12∗ ∗ 0.09∗ 0.1∗ ∗ 0.11∗ ∗ (0.04) (0.04) (0.04) (0.04) Controls No Yes Yes No Yes Yes Yes Yes Year FE No No Yes No No Yes Yes Yes Country FE No No Yes No No Yes Yes Yes Additional controls No No No No No No Yes Yes Observations 1528 1153 1153 1528 1153 1153 953 953 No. of countries 80 75 75 80 75 75 66 66 Notes: Dependent variable: the change in Inequality (Gini coefficient). All control variables are lagged one year. Controls include: income (logs), economic growth, investment, government consumption and secondary schooling. Additional controls include: population (logs), fertility rate, openness, quality of institutions, and inflation (logs). The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. All left-hand-side variables are included one period before, and ual countries, our estimates could suffer from endogeneity if a ma- we cluster errors at the country level.16 Our identification strategy jor commodity exporter has an influence on world prices. To ad- relies on the fact that international commodity price shocks are dress this concern, we do two things: i) exclude countries which not affected by individual countries (below we test for this) and exports represent more than 10% of total world exports, and ii) ex- therefore are expected to be exogenous to changes in inequality clude countries where a single commodity represents more than levels within countries. We also perform several robustness checks 50% of the country’s total exports. In the first case, we exclude 12 to reinforce the validity of our empirical strategy and our results. countries. In the second case, we exclude 11 countries. Our results do not seem to be affected by major exporters (see Tables A.8 and 4.3. Main results A.9, in the Appendix). To further address endogeneity concerns, we check the robustness of our results to using time-invariant Table 4 presents our main results. Column 1 presents pooled- export shares in the construction of our commodity price shock. OLS estimates. The regression yields a positive and significant co- Main results are not affected.20 Second, our results are also ro- efficient, indicating that the higher export-share-weighted aver- bust to controlling for initial levels of inequality, considering a dy- age of commodity prices, the higher inequality. Column 2 intro- namic model estimated using different techniques, including GMM duces some controls (at the expense of losing observations), while estimations, as standard with dynamic models (see Table A.10).21 column 3 introduces country fixed effects to control for time- Third, our main results do not change significantly when we aggre- invariant country characteristics.17 The coefficient for commodity gate commodity price shocks over 3-year periods to reduce short- price shocks yields a positive but no longer significant result.18 run noise in the data (see Table A.11).22 In all cases, we find that A non-significant coefficient for commodity price shocks could non-agricultural commodity price shocks significantly increase in- be masking differentiated effects of commodity price shock de- equality. pending on the type of commoditiy, as discussed in Section 2. Finally, we check whether our main results are robust to a dif- Consequently, in columns 4 to 6 of Table 4, we differentiate be- ferent specification considering whether increases in global non- tween agricultural and non-agricultural commodities (i.e., capital- agricultural commodity prices were relevant or not for each coun- intensive vs labor-intensive commodities). In all three columns, try in our sample. To do so, we classify countries into “treated” or while price shocks in agricultural commodities yield a negative co- “untreated” depending on whether they exported non-agricultural efficient, price shocks in non-agricultural commodities yield a pos- commodities in the whole period (>1) or not (=1). For each group, itive and significant coefficient. In columns 7 and 8, additional con- treated and untreated, we calculate the average Gini coefficient for trols are included (i.e., population in logs, fertility rate, openness, every period and plot its evolution over time, beside that of the av- quality of institutions, and inflation in logs). The main results hold, erage non-agricultural price index for countries in our sample (see suggesting that inequality increases in response to resource booms Figure A.2 in the Appendix). Looking at the evolution of inequal- associated with non-agricultural (i.e., capital-intensive) commodi- ity for the two groups, we see a sharp increase in the average Gini ties.19 of our treated group along a sharp increase in the non-agricultural price index from around 20 07-20 09 to 2014, something that we do not see for the untreated group. With this classification be- 4.4. Robustness tween treated and untreated countries, we run a simple Difference- in-Differences (Diff-in-Diff) estimation where we consider the sharp Our main results are robust to several additional checks. First, increase in prices between 2009 and 2014 as treatment, and find they are robust to the exclusion of major exporters of one com- a positive and significant “effect” of treatment (see Table A.12). In modity. Although world prices are typically unaffected by individ- 16 20 As data to measure income inequality comes from SWIID, our regressions use Results are available upon request. multiple imputation estimates (100 imputations) as suggested by Solt (2016). 21 The lagged level of inequality enters with a negative sign and is significant at 1 17 Table A.7 in the Appendix shows the coefficients for all controls. percent. 18 22 The results were also checked using pre-tax income inequality and the results The change in commodity prices is the average over last 3 years. The average were unchanged. of price changes over 3 years is taken to reduce the role of extremely transitory 19 The results remained unchanged using the latest version of SWIID database, shocks as well as to control for measurement error. However, using rolling windows version 9.0. can introduce serial correlation. 69 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table 5 Results by world regions OECD Non-OECD SSA & LA Non-OECD, Non- SSA & LA (1) (2) (3) (4) (5) (6) (7) (8) Commodity price shocks -0.15 0.07 0.03 0.1 (0.12) (0.04) (0.06) (0.21) ∗ Agricultural price shocks 0.31 -0.19 -0.34 -0.36 (0.22) (0.14) (0.19) (0.26) Non-agricultural price shocks -05 0.15∗ ∗ ∗ 0.13∗ ∗ 0.25 (0.32) (0.03) (0.06) (0.28) Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Yes Yes Additional controls Yes Yes Yes Yes Yes Yes Yes Yes Observations 375 375 578 578 356 356 257 257 No. of countries 21 21 45 45 29 29 18 18 Note: Dependent variable: the change in Inequality (Gini coefficient). All control variables are lagged one year. Note that two countries in our SSA & LA sample are OECD members. Controls include: Income (logs), economic growth, investment, govern- ment consumption and secondary schooling. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. other words, given a sharp increase in the price of non-agricultural bor intensive) commodity prices leads to lower inequality. These commodities, countries that exported non-agricultural commodi- effects seem particularly relevant in SSA and LA. This may be ex- ties experienced an increase in inequality compared to countries plained by the fact that countries in these two regions are charac- that did not export non-agricultural commodities. These results re- terized by high reliance on commodity exports, compared to the inforce our main finding of a positive impact of non-agricultural rest of the world. Additionally and as already highlighted, coun- price shocks on the evolution of inequality.23 tries in SSA and LA tend to show significantly higher levels of in- come inequality and lower levels of institutional quality than coun- tries elsewhere. As discussed in Section 2, the level of institutional 4.5. Results by world region quality is a key factor explaining potential effects of natural en- dowments and commodity price shock.25 In Table 5, we let the effects of commodity price shock to vary across different world regions, to analyze potential differences in the relationship between commodity price shocks and inequal- 4.6. Potential mechanisms ity in different contexts. In columns 1 to 4, we differentiate be- tween developed (22 OECD members) and developing countries Finally, and following our discussion in Section 2, we study po- (53 non-OECD members). For developed countries, we find non- tential mechanisms for different commodity price shocks to af- significant coefficients (columns 1 and 2). For developing countries, fect income inequality differently. According to our expectations, by contrast, we find a negative (but non-significant) coefficient for agricultural (i.e., labor-intensive) price shocks lower inequality, and agricultural commodities and a positive and highly significant co- this may happen due to an increase in wages (i.e., the ‘opportunity efficient for non-agricultural commodities (column 4). These re- cost’ effect). By contrast, non-agricultural (capital-intensive) price sults suggest that the positive short-run effect of higher capital- shocks will increase inequality, and this may happen due to an intensive prices on inequality occurs only in resource-rich devel- increase in capital-rent revenues (i.e., the ‘rapacity effect’). In this oping countries, but not in resource-rich developed countries. line, we consider two different types of inequality – pay vs. capital- According to Figures 2 and 3 in section 4.1, the relationship be- rents-driven inequality – as in Equation (5): tween commodity price shocks and income inequality is stronger  git = αi + δi + γ1 agriit + γ2 nonagriit + it (5) in SSA and LA than in rest of the world. Consequently, in columns 5 and 6 of Table 5, we consider only countries in SSA and LA, where git is either pay inequality or our proxy for capital-income while in columns 7 and 8 we consider non-OECD countries not in inequality of country i in period t. The hypothesis is that while LA or SSA. For countries in SSA and LA, we find a negative coef- higher prices of agricultural (i.e., labor-intensive) commodities re- ficient for agricultural price shocks and a positive coefficient for duce pay inequality, higher prices of non-agricultural (i.e., capital- non-agricultural price shocks, being both significant (see column intensive) commodities increase capital-income inequality. 6). For the rest of non-OECD countries, we find non-significant co- Results are presented in Table 6. In columns 1 to 4 we con- efficients (see column 7 and 8).24 sider the whole world sample. In columns 5 to 8 we only con- Results so far support the hypothesis that, in developing coun- sider countries in SSA and LA, following results in Table 5. Columns tries, a rise in non-agricultural (i.e., capital intensive) commodity 1, 2, 5 and 6 consider pay inequality as the dependent variable, prices lead to higher inequality, while a rise in agricultural (i.e., la- while columns 3, 4, 7 and 8 consider our proxy for capital-rents inequality. For agricultural (i.e., labor-intensive) price shocks, we find non-significant coefficients. However, for the SSA-and-LA sub- 23 As our key variables (inequality and price shocks) are continuous and show sample, the coefficients are negative in sign, as expected. For non- yearly variability, our preferred specifications and results are those presented in agricultural (i.e., capital-intensive) price shocks, results show a Table 4, using the full information in our panel data. Diff-in-Diff results should only be taken as robustness to our main results, and offering an insight into a long-run negative and highly significant coefficient on pay inequality, but impact of a rise in international prices. A more detailed Diff-in-Diff analysis is out a positive and highly significant coefficient on capital-income in- of the scope and aim of the paper. equality. These results suggest that while non-agricultural price 24 Note that two countries in our SSA & LA sample are OECD members. Results in columns 5 and 6 of Table 5 are robust to excluding these two countries. Results are 25 also robust to including additional controls as in column 7 of Table 4 (i.e., popula- Table A.13 in the Appendix reports the suggestive evidence of a potential role tion in logs, fertility rate, openness, quality of institutions, and inflation in logs). of initial levels of inequality and quality of institutions. 70 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table 6 The mechanisms the whole sample SSA & LA (1) (2) (3) (4) (5) (6) (7) (8) Dependent variables: pay inequality capital rents inequality pay inequality capital rents inequality Commodity price shocks -0.18 0.18 -0.94∗ ∗ 0.91∗ (0.23) (0.27) (0.39) (0.48) Agricultural prices shocks 0.6 -0.87 -0.39 -0.23 (0.57) (0.58) (1.04) (1.24) Non-agricultural price shocks -0.46∗ ∗ 0.56∗ ∗ -1.04∗ ∗ 1.11∗ ∗ ∗ (0.18) (0.17) (0.4) (0.45) Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Yes Yes Additional controls Yes Yes Yes Yes Yes Yes Yes Yes Observations 669 669 668 668 149 149 148 148 No. of countries 50 50 50 50 17 17 17 17 Note: all control variables are lagged one year. Controls include: Income (logs), economic growth, investment, government con- sumption and secondary schooling. The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. shocks reduce pay inequality they increase capital rents inequality, We also found evidence of the heterogeneities across countries; as expected.26 the inequality-increasing effects of non-agricultural price shocks We perform two robustness checks to results in Table 6. First, are mostly felt in countries with high initial levels of inequal- we consider an alternative measure of income inequality relying ity and low institutional quality, as is the case of many coun- on the Estimated Household Income Inequality (EHII) dataset.27 Ta- tries in Sub-Saharan Africa and Latin America. Looking at potential bles A.14 and A.15 in the Appendix, present results for the whole mechanisms, we have also found suggestive evidence that agricul- sample and for our SSA-and-LA subsample, respectively. Main re- tural (labor-intensive) commodity price shocks reduce pay inequal- sults hold. Second, we consider total natural resource rents (as % ity, while non-agricultural (capital-intensive) price shocks increase of GDP). As mentioned in Section 2, in developing countries, local capital-income inequality. elites together with foreign capital have been able to appropriate The results in this paper have important policy implications. In most of the rents from natural resources. As natural resource rents particular, results highlight important heterogeneities in the role of mostly relate to non-agricultural commodities, such as oil, gas and different price shocks for economic development. As international minerals, we can expect i) non-agricultural prices shock to be as- commodity prices have shown high volatility in recent decades, it sociated with natural resource rents, and ii) natural resource rents is important for policy makers to understand how changes in dif- to be associated with income inequality. This is precisely what we ferent prices can have different effects, including on the internal find in Table A.16 in the Appendix, especially significant for our distribution of income. SSA-and- LA subsample. 5. Conclusions CRediT authorship contribution statement In this paper, we have analyzed a relatively neglected dimen- Soran Mohtadi: Conceptualization, Methodology, Software, In- sion of the resource curse, namely the impact of commodity booms vestigation, Writing – original draft, Visualization. David Castells- on the evolution of income distribution. In doing so, we have taken Quintana: Conceptualization, Methodology, Validation, Formal a global look studying the evolution of inequality for more than 80 analysis, Writing – review & editing, Supervision. countries during more than two decades (1990-2016), and combin- ing this with the evolution of international prices for 23 globally Acknowledgments traded commodities. This has allowed us to study how commod- ity price shock influence income inequality, considering differenti- We thank James K. Galbraith, Oriol Roca Sagales, Vicente ated effects depending on the type of commodity (labor vs. capital- Royuela, Sambit Bhattacharyya, Pedro Trivin, Motasam Tatahi and intensive) and different components of income distribution (pay vs. Julian Messina for their valuable discussions and comments. We capital-rents inequality). are also grateful for comments received at the UAB-Applied Eco- Our results show that commodity price shocks have an im- nomics seminars (2018), at the 2nd Catalan Economic Society Con- pact on income inequality. However, this impact depends on the ference (2019), at the UB-AQR-IREA seminars (2019), at the Uni- type of the commodity, with non-agricultural (capital-intensive) versity of Buenos Aires (2019), and at the 44th symposium of the price shocks significantly increasing inequality. This result is found Spanish Economic Association (SAEe). David Castells-Quintana ac- to be robust to a long list of controls, different specifications knowledges the support of the Spanish Ministry of Science and and estimation techniques, as well as multiple robustness checks. Innovation (Project ID: PID2019-104723RB-I00) and the Agency for Management of University and Research Grants (Project ID: 26 Results are robust controlling for wages and capital tax following Dube and Var- 2017SGR1301). gas (2013). 27 The EHII provides an index ranging from 0 to 100 and is built by combining information from the Deninger and Squire Gini coefficients with a more precise Supplementary materials Theil-index-based measure of dispersion of pay within the industrial sector, and using data from the UTIP-UNIDO database (see Galbraith and Kum, 2005). The EHII has been used by Meschi and Vivarelli, 2009; Goderis and Malone, 2011; Kim and Supplementary material associated with this article can be Lin, 2018; Kim et al., 2020, among others. found, in the online version, at doi:10.1016/j.strueco.2021.08.002. 71 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Appendix Figure A.2. Inequality vs. price index. Table A.1 Variable names, definitions and sources Main variables Description Source Inequality (changes) Income inequality measured by the Gini coefficient SWIID v6.1 (Solt, 2016) Sit Commodity price shocks Constructed with data from the IMF-IFS and from the UNCTSD (2017)-Comtrade Conflict Intentional homicide, Number of deaths purposely inflicted by another person, per 100,000 World Bank-World Development Indicators (WDI) population Pay inequality (changes) Calculate measures of industrial pay inequality and provides a wage inequality Theil UTIP-Unido measure. Capital rents inequality The difference between income inequality and pay inequality Our calculation (changes) Wage (changes) Wage and salaried workers, total (% of total employment) WDI Capital tax rents (changes) Taxes on income, profits and capital gains (% of total taxes) WDI GDP per capita Per capita GDP (in logs) WDI Growth rate of GDP per capita Cumulative annual average per capital GDP growth rate WDI Institutions The sum of the political risk components including: Government stability, Socioeconomic ICRG International Country conditions, Investment Profile, Internal conflict, External conflict, Corruption, Military in Risk Guide politics, Religious tensions, Law and Order, Ethnic tensions, Democratic accountability and Bureaucracy quality. Investment Investment share (%GDP) WDI Government consumption Government consumption (%GDP) WDI Schooling Gross enrolment ratio, secondary, both sexes (%) WDI Population Total population (in logs) WDI Fertility Fertility rate, total (births per woman) WDI Openness Trade openness, measured as the sum of exports and imports of goods and services (%GDP) WDI Inflation Inflation, consumer prices (annual %) WDI Natural resource rents Total natural resources rents(%GDP) are the sum of oil rents, natural gas rents, coal rents WDI (hard and soft), mineral rents, and forest rents. 72 Table A.2 Descriptive statistics S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Variable Obs. No. of countries Mean Std.Dev. Min Max Income inequality (changes) 1789 80 -0.01 0.42 -2.09 2.30 Commodity price shocks 1760 80 0.02 0.21 -1.21 3.05 Agricultural commodity price shocks 1760 80 0.002 0.11 -0.83 1.19 Non-agricultural commodity price shocks 1760 80 0.013 0.15 -1.21 2.61 Conflict (changes) 947 73 -0.04 2.18 -14.9 36.2 Pay inequality (changes) 1101 66 0.07 1.45 -10.93 14.71 Capital rents inequality (changes) 1070 62 -0.02 1.32 -9.1 10.59 Wage (changes) 1975 79 0.23 1.13 -9.3 7.58 Capital tax rents (changes) 1243 66 0.15 4.08 -59.55 56.43 GDP per capita (logs) 2130 80 8.4 1.67 5.09 11.42 GDP per capita (growth) 2046 80 2.02 4.44 -67.80 36.98 Investment 2084 80 22.66 7.78 0 61.46 Government consumption 2072 80 15.74 6.2 2.04 76.22 Schooling 1642 79 75.43 34.08 5.21 163.93 Population 2157 80 16.65 1.5 11.15 21.04 Fertility 2160 80 3.25 1.81 0.91 7.72 Openness 2108 80 75.7 55 0 442.62 Institutions 1704 72 66.37 13.68 19.16 96.08 Inflation 1975 80 1.72 1.41 -4.09 10.1 Natural resource rents (changes) 2055 80 -0.12 2.75 -25.11 21.51 Table A.3 List of countries 1 Algeria 29 Honduras 57 Peru 2 Angola 30 Hong Kong 58 Philippines 3 Argentina 31 Hungary 59 Poland 4 Australia 32 India 60 Qatar 5 Belgium 33 Indonesia 61 Romania 6 Benin 34 Iran 62 Russia 7 Bolivia 35 Iraq 63 Rwanda 8 Brazil 36 Ireland 64 Senegal 9 Bulgaria 37 Italy 65 Singapore 10 Burkina Faso 38 Japan 66 South Africa 11 Burundi 39 Kazakhstan 67 Spain 12 Cameroon 40 Kenya 68 Sweden 13 Canada 41 Korea 69 Switzerland 14 Central African Republic 42 Lesotho 70 Thailand 15 Chile 43 Lithuania 71 Togo 16 China 44 Madagascar 72 Turkey 17 Colombia 45 Malawi 73 Uganda 18 Costa Rica 46 Malaysia 74 United Kingdom 19 Czech Republic 47 Mali 75 United States 20 Dominica 48 Mauritania 76 Uruguay 21 Ecuador 49 Mexico 77 Venezuela 22 Ethiopia 50 Netherlands 78 Viet Nam 23 France 51 New Zealand 79 Zambia 24 Gambia 52 Nicaragua 80 Zimbabwe 25 Germany 53 Niger 26 Ghana 54 Nigeria 27 Guinea 55 Norway 28 Haiti 56 Paraguay Table A.4 Correlation matrix, main variables Income inequality (changes) Commodity price shocks Conflict (changes) Institutions Income inequality (changes) 1 Commodity price shocks 0.042∗ 1 Conflict (changes) 0.045 -0.054∗ 1 Institutions 0.178∗ -0.005 -0.067∗ 1 ∗ Note: P<0.1 73 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table A.5 List of Commodities Non-Agricultural Oil Zinc Nickel Natural Gas Aluminum Diamond Coal Iron Lead Gold Copper Silver Tin Agricultural Coffee Cotton Rubber Banana Wool Rice Wheat Wood Sugar Cocoa Table A.6 The average rate of inequality and quality of institutions in different regions of the sam- ple Inequality Institutions Asia 39.8 68.42 Europe 30.45 78.09 Latin America 45.92 63.28 Middle East 38.84 57.18 North America 33.51 83.55 Oceania 32.16 85.23 Sub-Saharan Africa 41.04 55.98 Total 38.62 66.37 Table A.7 The coefficients of Table 4 (controls) (1) (2) (3) (4) (5) (6) (7) (8) Dependent variable: the change in Inequality (Gini coefficient) Commodity price shocks 0.07∗ ∗ 0.05 0.02 0.03 (0.03) (0.03) (0.04) (0.05) Agricultural price shocks -0.01 -0.01 -0.12 -0.18 (0.08) (0.10) (0.1) (0.11) Non-agricultural price shocks 0.12∗ ∗ 0.09∗ 0.1∗ ∗ 0.11∗ ∗ (0.04) (0.04) (0.04) (0.04) Log (income) 0.04∗ ∗ -0.11 0.04∗ ∗ -0.1 -0.11 -0.09 (0.01) (0.23) (0.01) (0.23) (0.2) (0.26) Economic growth 0.004 -0.003∗ ∗ 0.004 -0.003 -0.002 -0.002 (0.003) (0.004) (0.003) (0.004) (0.004) (0.004) Investment 0.0001 -0.003 0.00009 -0.003 -0.003 -0.003 (0.01) (0.004) (0.001) (0.004) (0.004) (0.004) Government spending 0.008∗ ∗ 0.01∗ ∗ 0.008∗ ∗ 0.01∗ ∗ 0.009 0.01 (0.002) (0.006) (0.002) (0.006) (0.007) (0.007) Schooling -0.0005 -0.003∗ -0.0005 -0.003∗ -0.005 -0.005 (0.0007) (0.002) (0.0007) (0.002) (0.002) (0.002) Total population 0.51 0.49 (0.33) (0.32) Fertility 0.17 0.162∗ (0.1) (0.1) Openness -0.001 -0.001 (0.001) (0.001) Quality of Institutions 0.002 0.002 (0.005) (0.004) Inflation -0.01 -0.01 (0.03) (0.03) Year FE No No Yes No No Yes Yes Yes Country FE No No Yes No No Yes Yes Yes Observations 1528 1153 1153 1528 1153 1153 969 969 Number of countries 80 75 75 80 55 78 67 67 74 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table A.8 Exclude countries which exports represent more than 10% of total world exports (1) (2) (3) (4) (5) (6) (7) (8) ∗∗ Commodity price shocks 0.07 0.05 0.009 0.01 (0.03) (0.03) (0.04) (0.05) ∗ Agricultural price shocks 0.007 0.01 -0.14 -0.23∗ ∗ (0.09) (0.11) (0.07) (0.1) Non-agricultural price shocks 0.11∗ ∗ 0.07 0.08∗ 0.09∗ (0.05) (0.04) (0.04) (0.05) Controls No Yes Yes No Yes Yes Yes Yes Year FE No No Yes No No Yes Yes Yes Country FE No No Yes No No Yes Yes Yes Additional controls No No No No No No Yes Yes Observations 1284 965 965 1284 965 965 786 786 No. of countries 68 64 64 68 64 64 55 55 Note: Dependent variable: the change in Inequality (Gini coefficient). All control variables are lagged one year. Con- trols include: income (logs), economic growth, investment, government consumption and secondary schooling. Ad- ditional controls include: population (logs), fertility rate, openness, quality of institutions, and inflation (logs). The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. Table A.9 Exclude countries where a single commodity represents more than 50% of the country’s total exports (1) (2) (3) (4) (5) (6) (7) (8) Commodity price shocks 0.05 0.04 0.01 0.02 (0.03) (0.03) (0.04) (0.05) Agricultural price shocks -0.05 -0.08 -0.12 -0.18 (0.09) (0.11) (0.08) (0.11) Non-agricultural price shocks 0.11∗ ∗ 0.09∗ 0.08∗ 0.09∗ ∗ (0.05) (0.04) (0.04) (0.04) Controls No Yes Yes No Yes Yes Yes Yes Year FE No No Yes No No Yes Yes Yes Country FE No No Yes No No Yes Yes Yes Additional controls No No No No No No Yes Yes Observations 1336 1034 1034 1336 1034 1034 858 858 No. of countries 69 65 65 69 65 65 58 58 Notes: Dependent variable: the change in Inequality (Gini coefficient). All control variables are lagged one year. Controls include: income (logs), economic growth, investment, government consumption and secondary schooling. Additional controls include: population (logs), fertility rate, openness, quality of institutions, and inflation (logs). The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. Table A.10 Controlling for inequality in levels (1) (2) (3) (4) (5) (6) (7) (8) ∗∗∗ ∗∗∗ ∗ ∗∗∗ ∗∗∗ ∗ ∗ Inequality t-1 -0.007 -0.008 -0.02 -0.007 -0.008 -0.02 -0.02 -0.02∗ ∗ (0.001) (0.001) (0.01) (0.001) (0.001) (0.01) (0.01) (0.01) Commodity price shocks 0.08∗ ∗ 0.07∗ 0.03 0.03 (0.03) (0.04) (0.04) (0.05) Agricultural price shocks -0.02 -0.04 -0.12 -0.2∗ (0.11) (0.1) (0.1) (0.08) Non-agricultural price shocks 0.14∗ ∗ 0.12∗ ∗ 0.1∗ ∗ 0.1∗ ∗ (0.04) (0.05) (0.04) (0.04) Controls No Yes Yes No Yes Yes Yes Yes Year FE No No Yes No No Yes Yes Yes Country FE No No Yes No No Yes Yes Yes Additional controls No No No No No No Yes Yes Observations 1528 1153 1153 1528 1153 1153 953 953 Number of countries 80 75 75 80 75 75 66 66 Notes: Dependent variable: the change in Inequality (Gini coefficient). All control variables are lagged one year. Controls include: Income (logs), economic growth, investment, government consumption and secondary schooling. Additional controls include: pop- ulation (logs), fertility rate, openness, quality of institutions, and inflation (logs). The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. 75 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table A.11 3-year commodity price shocks (1) (2) (3) (4) (5) (6) (7) (8) ∗∗∗ ∗∗ 3-year commodity price shock 0.17 0.12 0.03 0.03 (0.04) (0.04) (0.07) (0.08) 3-year agricultural price shock 0.05 0.008 -0.25 -0.36∗ ∗ (0.11) (0.14) (0.19) (0.17) 3-year non-agricultural price 0.22∗ ∗ 0.16∗ ∗ 0.15∗ 0.19∗ ∗ shock (0.06) (0.06) (0.08) (0.07) Controls No Yes Yes No Yes Yes Yes Yes Year FE No No Yes No No Yes Yes Yes Country FE No No Yes No No Yes Yes Yes Additional controls No No No No No No Yes Yes Observations 1613 1205 1205 1613 1205 1205 1034 1034 Number of countries 80 75 75 80 75 75 66 66 Notes: Dependent variable: the change in Inequality (Gini coefficient). All control variables are lagged one year. Controls include: Income (logs), economic growth, investment, government consumption and secondary schooling. Additional controls include: population (logs), fertility rate, openness, quality of institutions, and inflation (logs). The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. Table A.12 Difference-in-Difference estimation Dependent variable: the change in Inequality (Gini coefficient) untreated treated Diff-in-Diff Period ∗ treated 0.009 1.138 1.12 Std. Error 0.057 0.444 0.448 P>|t| 0.88 0.011∗ ∗ 0.012∗ ∗ Number of observations 618 238 856 Notes: we take the 2009-2014 as the treatment period. Treated countries are those hav- ing Pit>1 for the period of analysis. ∗ ∗ p<0.05 Table A.13 The role of initial level of inequality and quality of institutions (1) (2) Agricultural price shocks 0.04 ∗ OECD (0.21) Non-agricultural price -0.25 shocks∗ OECD (0.28) Agricultural price shocks ∗ -0.11 Non-OECD (0.08) Non-agricultural price shocks∗ 0.13∗ ∗ ∗ Non-OECD (0.03) Agricultural price shocks ∗ -0.09 Non-OECD, Non-SSA& LA (0.32) Non-agricultural price shocks∗ 0.29 Non-OECD, Non- SSA& LA (0.29) Agricultural price shocks ∗ SSA -0.11 & LA (0.08) Non-agricultural price 0.12∗ ∗ ∗ shocks∗ SSA & LA (0.02) Controls Yes Yes Year FE Yes Yes Country FE Yes Yes Observations 1153 1153 No. of countries 75 75 76 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 Table A.14 Robustness to Table 6, the whole sample (1) (2) (3) (4) (5) (6) Dependent variables: income inequality (EHII) pay inequality capital rents inequality Commodity price shocks 0.12 -0.18 0.29 (0.07) (0.23) (0.24) Agricultural prices shocks -0.41 0.6 -1.11∗ ∗ (0.59) (0.57) (0.44) Non-agricultural price shocks 0.28∗ ∗ -0.46∗ ∗ 0.79∗ ∗ ∗ (0.11) (0.18) (0.18) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Additional controls Yes Yes Yes Yes Yes Observations 703 669 669 623 623 No. of countries 54 50 50 50 50 Note: all control variables are lagged one year. Controls include: Income (logs), economic growth, investment, government consumption and secondary schooling. The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. Table A.15 Robustness to Table 6, Sub-Saharan and Latin-American countries (1) (2) (3) (4) (5) (6) Dependent variables: income inequality (EHII) pay inequality capital rents inequality Commodity price shocks -0.02 -0.94 0.94∗ (0.12) (0.35) (0.41) Agricultural prices shocks -1.63∗ -0.39 -1.77 (1.13) (1.04) (1.15) Non-agricultural price shocks 0.28∗ -1.04∗ ∗ 1.46∗ ∗ ∗ (0.15) (0.41) (0.45) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Additional controls Yes Yes Yes Yes Yes Observations 209 149 149 142 142 No. of countries 20 17 17 17 17 Note: all control variables are lagged one year. Controls include: Income (logs), economic growth, investment, government consumption and secondary schooling. The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. Table A.16 Robustness to Table 6 using natural resource rents the whole sample SSA & LA (1) (2) (3) (4) (5) (6) (7) (8) income capital rents income capital rents Dependent variables: natural resource rents inequality inequality natural resource rents inequality inequality Commodity price shocks 0.83∗ ∗ 0.18 (0.31) (0.32) Agricultural prices shocks -1.36∗ ∗ -2.67∗ ∗ (0.59) (1.14) Non-agricultural price shocks 1.67∗ ∗ 1.01∗ ∗ (0.59) (0.42) Natural resource rents 0.007∗ 0.03 0.01∗ ∗ 0.1∗ ∗ (0.004) (0.02) (0.005) (0.03) Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Yes Yes Yes Additional controls Yes Yes Yes Yes Yes Yes Yes Yes Observations 995 995 1071 717 393 393 394 171 No. of countries 67 67 66 54 30 30 29 19 Note: all control variables are lagged one year. Controls include: Income (logs), economic growth, investment, government consumption and secondary schooling. The time span goes from 1990 to 2016. All estimations are done with multiple-estimation regressions (100 imputations). Robust standard errors (clustered by country) in parentheses. ∗ ∗ ∗ P<0.01, ∗ ∗ P<0.05, ∗ P<0.1. 77 S. Mohtadi and D. Castells-Quintana Structural Change and Economic Dynamics 59 (2021) 63–78 References Galbraith, J.K., Halbach, B., Malinowska, A., Shams, A., Zhang, W., 2014. UTIP Global Inequality Data Sets 1963-2008: Updates, Revisions and Quality Checks (Work- Acemoglu, D., Johnson, S., Robinson, J., Thaicharoen, Y., 2003. Institutional causes, ing Paper No. 68). University of Texas at Austin. macroeconomic symptoms: volatility, crises and growth. Journal of Monetary Galbraith, J.K., and Purcell, G. 1999. Inequality and state violence: A preliminary Economics 50 (1), 49–123. report. UTIP Working Paper No. 4. Addison, T., Pirttilä, J., Tarp, F., 2017. Inequality: Measurement, Trends, Impacts and Goderis, B., Malone, S.W, 2011. Natural resource booms and inequality: theory and Policies. Review of Income and Wealth 63, 603–607. evidence. The Scandinavian Journal of Economics 113 (2), 388–417. Arezki, R., Van der Ploeg, F., 2010. Trade policies, institutions and the natural re- Gylfason, T., Zoega, G., 2002. Inequality and economic growth: do natural resources source curse. Applied Economics Letters 17, 1443–1451. matter? Center for Economic Studies and ifo Institute (CESifo) Working Paper Auty, R.M., Furlonge, H.I, 2019. The Rent Curse: Natural Resources, Policy Choice, 712 (5). and Economic Development. Oxford University Press, USA. Gylfason, Thorvaldur, 2019. Inequality Undermines Democracy and Growth. Center Bazzi, S., Blattman, C., 2014. Economic shocks and conflict: Evidence from commod- for Economic Studies and ifo Institute (CESifo Working Paper 7486. ity prices. American Economic Journal: Macroeconomics 6 (4), 1–38. Hartwell, C.A., Horvath, R., Horvathova, E., Popova, O., 2019. Democratic institutions, Behzadan, N., Chisik, R., Onder, H., Battaile, B., 2017. Does inequality drive the Dutch natural resources, and income inequality. Comparative Economic Studies 61 (4), disease? Theory and evidence. Journal of International Economics 106, 104–118. 531–550. Bhattacharyya, S., Williamson, J.G, 2016. Distributional consequences of commodity Haslam, P.A., 2016. Overcoming the resource curse: Reform and the rentier state in price shocks: Australia over a century. Review of Income and Wealth 62 (2), Chile and Argentina, 1973–20 0 0. Development and Change 47 (5), 1146–1170. 223–244. Kim, D.H., Lin, S.C, 2018. Oil abundance and income inequality. Environmental and Bjorvatn, K., Naghavi, A., 2011. Rent seeking and regime stability in rentier states. Resource Economics 71 (4), 1–24. European Journal of Political Economy 27, 740–748. Kim, D.H., Chen, T.C., Lin, S.C, 2020. Does oil drive income inequality? New panel Blattman, C., Miguel, E., 2010. Civil war. Journal of Economic Literature 48 (1), 3–57. evidence. Structural Change and Economic Dynamics 55, 137–152. Boschini, A., Pettersson, J., Roine, J., 2013. The resource curse and its potential re- Krieger, T., Meierrieks, D., 2016. Does Income Inequality Lead to Terrorism? CESifo versal. World Development 43, 19–41. Working Paper Series (5821). Cabrales, A., Hauk, E., 2011. The quality of political institutions and the curse of Law, S.H., Naseem, N.A.M., Lau, W.T., Trinugroho, I., 2020. Can innovation improve natural resources. The Economic Journal 121 (551), 58–88. income inequality? Evidence from panel data. Economic Systems 44 (4), 100815. Carmignani, F., Avom, D., 2010. The social development effects of primary commod- Le Billon, P., 2014. Natural resources and corruption in post-war transitions: matters ity export dependence. Ecological Economics 70 (2), 317–330. of trust. Third World Quarterly 35 (5), 770–786. Caselli, F., Tesei, A., 2016. Resource windfalls, political regimes, and political stability. Leamer, E.E., Maul, H., Rodriguez, S., Schott, P.K, 1999. Does natural resource abun- Review of Economics and Statistics 98 (3), 573–590. dance increase Latin American income inequality? Journal of Development Eco- Castells-Quintana, D., 2017. Malthus living in a slum: Urban concentration, infras- nomics 59 (1), 3–42. tructure and economic growth. Journal of Urban Economics 98, 158–173. Lessmann, C., Steinkraus, A., 2019. The geography of natural resources, ethnic in- Castells-Quintana, D., 2019. Urban Inequality in Latin America: All the Horizon. Re- equality and civil conflicts. European Journal of Political Economy 59, 33–51. gional and Urban Economy Review. Armand Colin 239–257. Mehlum, H., Moene, K., Torvik, R., 2006. Institutions and the resource curse. The Chong, A., Gradstein, M., 2007. Inequality and Institutions. The Review of Economics Economic Journal 116 (508), 1–20. and Statistics 89 (3), 454–465. Meschi, E., Vivarelli, M., 2009. Trade and income inequality in developing countries. Ciccone, A. 2020. International commodity prices and civil war outbreak: new evi- World development 37 (2), 287–302. dence for Sub-Saharan Africa and beyond. Musayev, V. 2014. Commodity price shocks, conflict and growth: the role of institu- Collier, P., Hoeffler, A., 2004. Greed and grievance in civil war. Oxford Economic Pa- tional quality and political violence. MPRA Paper No. 59786. pers 56 (4), 563–595. Papyrakis, E., Gerlagh, R., 2004. The resource curse hypothesis and its transmission Conceição, P., Galbraith, J.K., Bradford, P., 2001. The theil index in sequences of channels. Journal of Comparative Economics 32 (1), 181–193. nested and hierarchic grouping structures: implications for the measurement Papyrakis, E., 2014. A development curse: formal vs. informal activities in re- of inequality through time, with data aggregated at different levels of industrial source-dependent economies. International Journal of Social Economics 41 (3), classification. Eastern Economic Journal 27 (4), 491–514. 244–264. Crivelli, E., Gupta, S., 2014. Resource blessing, revenue curse? Domestic revenue Parcero, O.J., Papyrakis, E., 2016. Income inequality and the oil resource curse. Re- effort in resource-rich countries. European Journal of Political Economy 35, source and Energy Economics 45, 159–177. 88–101. Ross, M.L., 1999. The political economy of the resource curse. World Politics 51 (2), Doerrenberg, P., Peichl, A., 2014. The impact of redistributive policies on inequality 297–322. in OECD countries. Applied Economics 46 (17), 2066–2086. Sachs, J.D., Warner, A.M, 1999. The big push, natural resource booms and growth. Dreher, A., Gaston, N, 2008. Has globalization increased inequality? Review of Inter- Journal of Development Economics 59 (1), 43–76. national Economics 16 (3), 516–536. Sala-i-Martin, X., Subramanian, A., 2013. Addressing the natural resource curse: An Dube, O., Vargas, J.F, 2013. Commodity price shocks and civil conflict: Evidence from illustration from Nigeria. Journal of African Economies 22 (4), 570–615. Colombia. The Review of Economic Studies 80 (4), 1384–1421. Seghezza, E., Pittaluga, G.B, 2018. Resource rents and populism in resource-depen- Esteban, J., Ray, D., 2011. Linking conflict to inequality and polarization. American dent economies. European Journal of Political Economy 54, 83–88. Economic Review 101 (4), 1345–1374. Siegel, K.M., 2016. Fulfilling Promises of More Substantive Democracy? Post-neolib- Farzanegan, M. R. and Krieger, T. 2018. Oil rents shocks and inequality in Iran. CESifo eralism and Natural Resource Governance in South America. Development and Working Paper, No. 6876. Change 47 (3), 495–516. Fearon, J., 2008. Economic development, insurgency, and civil war. Institutions and Solt, F., 2016. The standardized world income inequality database. Social Science Economic Performance 292, 328. Quarterly 97 October 2017, SWIID Version 6.1. Fearon, J.D., Laitin, D.D, 2003. Ethnicity, insurgency, and civil war. American Political Tan, H.B., Law, S.H, 2012. Nonlinear dynamics of the finance-inequality nexus in de- Science Review 97 (1), 75–90. veloping countries. The Journal of Economic Inequality 10 (4), 551–563. Ferraro, D., Peretto, P.F, 2018. Commodity prices and growth. The Economic Journal Van der Ploeg, F., 2011. Natural resources: curse or blessing? Journal of Economic 128 (616), 3242–3265. literature 49 (2), 366–420. Fum, R.M., Hodler, R., 2010. Natural resources and income inequality: The role of Van der Ploeg, F., Poelhekke, S., 2009. Volatility and the natural resource curse. Ox- ethnic divisions. Economics Letters 107 (3), 360–363. ford Economic Papers 61 (4), 727–760. Galbraith, J.K., Kum, H., 2005. Estimating the inequality of household incomes: A statistical approach to the creation of a dense and consistent global data set. Review of Income and Wealth 51 (1), 115–143. 78