Inequality and market power in Latin America and the Caribbean^†

Abstract

This study estimates firm-level markups (product market power) and markdowns (labor market power) for formal firms in 16 Latin American and Caribbean (LAC) countries and 28 economically comparable but less unequal countries. Objectives include (1) contrasting market power in LAC firms with peers; (2) evaluating its effect on firm-level labor revenue share; and (3) assessing its influence on the overall labor income share, considering market power’s distribution, labor share and firm size. Findings reveal LAC firms’ average price markup is 87% over marginal costs, and wages are 38% below labor’s marginal revenue product. The firm-level negative link between labor share and market power mainly stems from labor market power, more pronounced in larger firms. This influences the total labor share and income distribution. Yet, LAC’s market power intensity and distribution don’t surpass its peers, nor do they contribute more to inequality.

labor market power, product market power, markups, markdowns, Latin America and the Caribbean

Introduction

Firms matter for earnings inequality (Card et al. 2018). Recent trends in wage inequality are well traced by differences in pay across firms. In developed economies, including the USA and Germany, labor earnings inequality increased in recent decades hand in hand with differences in pay across firms (Card et al. (2013); Song 2019; Criscuolo 2020). In Latin America, by contrast, earnings inequality sharply declined during 2002-2015 (Messina and Silva 2018). Narrowing between-firm pay differentials in the formal sector was a fundamental driver of the observed trends (Alvarez et al. 2018; Messina and Silva 2019).

A prominent reason why firms may matter for inequality is market power. Firms with power in product markets charge higher prices, depressing demand and employment. Firms’ power in the labor market takes the form of reduced wages for given marginal revenue productivity of labor, thus further depressing the part of revenue that goes to workers. Moreover, dispersion in market power across firms implies wage dispersion not explained by workers’ productive capacity. In sum, product and labor market power are fundamentally linked to inequality through their influence on the labor share in income and wage dispersion (Autor et al. 2020; Brooks et al. 2021).

Across developed countries, firms exhibit substantial market power in product and labor markets (Loecker et al. 2019; Manning 2021; Sokolova and Sorensen 2021). The evidence is more scattered for Latin America, although a rapidly growing literature yields results for the formal labor market of specific countries that are surprisingly similar to those in the literature for advanced economies (Tortarolo and Zarate 2020; Amodio and de Roux 2021; Amodio et al. 2022; Casacuberta and Gandelman 2023). We provide a broad picture of how the level and distribution of product and labor market power among formal firms in Latin America relate to the share of labor in revenue and to excess wage dispersion beyond what can be explained by productivity differentials. Understanding this relationship is particularly important for a region where inequality is high. Our main contribution is to provide basic stylized facts for a wide range of countries in the region and contrast them with economies with similar levels of development but lower levels of inequality.

Building on the work of Eslava et al. (2023), this study provides an overview of the extent of product and labor market power across Latin American and Caribbean (LAC) countries and its relationship to wages and the labor share of income. We provide stylized facts that address three questions: 1) The extent and dispersion of market power across firms; 2) the relationship between market power and the labor share of revenue at the firm level; and 3) the implications of that relationship for the aggregate labor share of income, which depends on the joint distribution (across firms) of market power and firms’ size.

For comparison, we also examine other middle-income countries with similar GDP per capita, which we label as ‘peers’. This comparative analysis is not only of intrinsic interest but also holds significance due to the considerably lower levels of inequality observed among the peer group of countries. The average Gini coefficient in the LAC countries of our sample is 47.6, compared to 38.6 among the peers.¹ Thus, our analysis sheds light on the extent to which higher inequality in LAC is associated with the exacerbated exercise of market power by firms when compared to peer economies.

We capture a firm’s product market power through its markup, i.e. the margin between the price it charges and its marginal cost. In turn, our measure of the firm’s labor market power is its markdown: the (inverse of the) wedge between the wage received by its workers and those workers’ contribution to the firm’s revenue, measured by the marginal revenue product of labor. To compute firm-level measures of markups and markdowns in each country, we build on the recent work of Loecker and Eeckhout (2018), Brooks et al. (2021), and Yeh et al. (2022). Once measures of market power are obtained, we explore the relationship at the firm level between market power and firms’ labor shares out of revenue. Macroeconomic implications are derived by studying the associations between micro measures of labor shares, market power and firms’ size.

The main source of information for the analysis is the World Bank’s Enterprise Survey (ES) conducted, depending on the country, in any year between 2009 and 2019. These firm-level surveys are representative of formal firms with more than five employees in manufacturing and selected service industries and are collected using a consistent methodology across countries and years. The wide coverage among developing countries renders this survey unique to benchmark Latin America and the Caribbean against other middle-income regions. It should be noted that the construction of markups and markdowns requires detailed firm-level data on the cost shares of production inputs, including labor, energy, intermediaries and capital. Because these firm expenditure shares are seldom reported by service firms in the ES, the analysis in this paper is restricted to the manufacturing sector.

We find that product and labor markets in LAC are far from perfectly competitive. The average manufacturing firm charges a markup of 87% over the marginal cost and pays wages 38% below the marginal revenue product of labor. However, average markups and markdowns in the region do not differ much from those observed in peer economies with significantly lower inequality. In fact, markups and markdowns in our sample of peer countries are higher than in Latin America: the average firm markup is 136% over the marginal production cost, and the average markdown is such that wages are 52% below the marginal revenue product of labor.

We find that the negative correlation at the firm level that arises by construction between our combined measures of market power and the labor share of revenue is primarily driven by labor market power, represented by markdowns, rather than product market power, characterized by markups. At the aggregate level, the influence of market power on functional inequality is exacerbated by larger firms extracting higher markdowns. While these two dimensions of market power are undoubtedly significant contributors to inequality in the region, they are not unique to Latin America. Evidence from peer countries mirrors our findings in LAC, suggesting that market power among formal firms is not the sole culprit for the excessive levels of inequality observed in the region.

Organization of the paper. Section 2 outlines the production function approach methodology we apply to derive firm-level measures of markups and markdowns. Section 3 discusses the central features of the data. Section 4 analyzes the properties of the estimated markups and markdowns and studies the extent to which markups and markdowns correlate to firms’ sizes and labor shares. Section 5 concludes.

Market power in product and labor markets

Deriving firms’ markups and markdowns

Firms operating in perfectly competitive markets can sell the products or services they produce but cannot influence the price at which those products are sold. They can hire as many workers as they need to produce those goods or services but cannot set their wages. Wages, instead, are determined by market dynamics through the interplay of supply and demand forces.

When product markets are imperfectly competitive, firms can influence prices and market outcomes. Firms exerting market power in product markets distort competition, leading to higher prices and reduced employment and output. We quantify this phenomenon using the concept of markups, which measure the gap between the price at which a product is sold and the cost the firm incurs to make one more unit of it; more formally, the ratio between the selling price of goods or services and the marginal cost of producing the last unit. Markups exceeding 1 indicate the presence of market power within product markets.

In labor markets, companies wielding market power keep wages at lower levels to maximize their profits. To gauge this, we use the concept of markdown, which reflects the relationship between the wage a company pays and the value a worker contributes to the company’s productivity. More precisely, we measure a firm’s markdown as the ratio between its marginal revenue product of labor and the wage paid to its workers. If the markdown exceeds 1, it means that the least productive worker (the marginal worker) is generating more value for the company than they are receiving in labor compensation. Markdowns don’t always have to be greater than 1; they can also be less than 1, indicating that the company shares some of the extra rents obtained in product markets with its employees.

The methodology for estimating markups and markdowns at the firm level follows the production approach proposed by Hall et al. (1986) for markups, further refined by Loecker and Warzynski (2012), and subsequently expanded to markdowns by numerous researchers (e.g. Yeh et al. 2022). In this approach, expressions used to calculate markups and markdowns are derived relying on the relatively weak assumption that a firm seeks to minimize the cost of producing its optimal quantity of product. The approach permits calculating markups and markdowns using information about a firm’s output and inputs use that appears in standard firm-level surveys. This is an advantage relative to methods that would require stronger assumptions about the structure of markets, or very detailed data on a firm’s individual products and the cost of inputs used to produce each product. As we discuss further below, the method still requires estimating production function elasticities, which demands additional assumptions.

In this framework, firms choose the quantity they produce and the optimal mix of inputs to minimize their production costs. They have enough market power that the price at which they sell is not taken as given but varies as a function of the quantity produced. They also influence the wages paid to their employees within the constraints of labor supply schedules. One key assumption is that in the market for at least one input of production, such as material inputs, the firm takes the price of that input as given, ensuring a degree of competitive forces within the production process.

The firm’s cost minimization problem is the following:

$$ \begin{align} \min_{\{X_{i}^{k}\}_{k=1}^{K}\in \mathbb{R}_{+}} \sum_{k=1}^{K} V_{i}^{k}(X_{i}^{k}) X_{i}^{k} \;\; \text{s.t.}\;\; F\big(X_{i}^{1},\ldots,X_{i}^{K};\omega_{i}\big)\geq Q_{i}\end{align} $$

(1)

where we allow the price of input |$k$|⁠, |$V_{i}^{k}$| to potentially depend on the level of the input used by the firm, |$X_{i}^{k}$|⁠. The production function |$F(\cdot )$| is twice continuously differentiable. Assuming input |$k^{\prime}$| is fully flexible, static and not subject to monopsony forces, the first-order condition with respect to input |$k^{\prime}$| allows deriving Loecker and Warzynski (2012) markups formula for firm |$i$| at time |$t$|⁠:

$$ \begin{align} \underbrace{\mu_{i}}_{\textrm{Markup}} &\equiv \frac{P_{i}}{MC_{i}} = \underbrace{\left(\frac{\partial F(\cdot)}{\partial X_{i}^{k^{\prime}}}\frac{X_{i}^{k^{\prime}}}{Q_{i}}\right)}_{\textrm{Output Elasticity}}\cdot \underbrace{\left(\frac{V_{i}^{k^{\prime}}X_{i}^{k^{\prime}}}{P_{i}Q_{i}}\right)^{-1}}_{\textrm{Factor Share}}\end{align} $$

(2)

The markup, |$\mu _{i}$|⁠, equals the ratio of the output elasticity of the flexible and perfectly competitive input with respect to the cost share of that input in total revenues. When the firm does not have market power in its product market this ratio equals 1, implying that the contribution of each flexible factor of production to output equals its cost share in revenues.

In addition, the firm can have market power in other factors of production. We consider imperfect competition in labor markets. If labor is a flexible input not subject to adjustment costs, but firm |$i$| has the power to affect wages, then the first-order condition for labor can be rearranged to derive the wage markdown for firm |$i$| at time |$t$|⁠:

$$ \begin{align} \underbrace{\nu_{i}}_{\textrm{Markdown}} &\equiv \left[\frac{\partial w_{i}(l_{i})}{\partial l_{i}}\frac{l_{i}}{w_{i}(l_{i})}+1\right] = \frac{1}{\mu_{i}} \left[ \underbrace{\left(\frac{\partial F(\cdot)}{\partial l_{i}}\frac{l_{i}}{Q_{i}}\right)}_{\textrm{Output Elasticity}} \underbrace{\left(\frac{w_{i}(l_{i})l_{i}}{P_{i}Q_{i}}\right)^{-1}}_{\textrm{Labor Share}}\right]\end{align} $$

(3)

Equations (3) and (2) are very similar. In the absence of monopsony power in labor markets or any other frictions preventing labor from adjusting to their perfectly competitive levels instantaneously, the term in brackets on the right-hand side of eq. (3) equals the markup, and the markdown would be equal to 1. However, because of labor market power, the price of labor is not given to the firm. Departures from the competitive price of labor introduce a wedge in labor markets between the output elasticity of employment and the factor share that leads to markdowns different from 1.

Note that labor market power, as captured by the markdown, is not independent from product market power. Indeed, equation (3) shows that for given output elasticity to labor and labor share, there is a negative direct relationship between markups and markdowns. This reflects that when markups are high for a given labor share of income, the firm is willing to accept less productive workers for a given wage.

A key advantage of the production function approach to markups and markdown estimation is that it only requires balance-sheet data. Total revenue and the costs incurred for variable production inputs such as labor and materials are readily available in typical firm-level datasets, as they are in our data. Input-output elasticities can be estimated using the same information. The next section describes in more detail the data, as well as the implementation of (2) and (3) to estimate markups and markdowns.

Data and key variables

We rely on the World Bank Enterprise Surveys (WBES) to derive measures of markups and markdowns. The WBES is a firm-level survey of a representative sample of an economy’s formal private sector. The surveys cover a broad range of business environment topics and firm-level outcomes. Relevant to our analysis, it includes information on firms’ revenues and a series of costs incurred to hire production inputs. The questions are standardized and formulated in a similar format across countries, guaranteeing comparability. Although the WBES has several limitations compared to other national firm-level datasets, the key advantage of the WBES for our analysis is the availability of (comparable) firm data representative of the entire formal segment of production for a wide range of developing economies. The coverage of these countries is limited and far from representative in other data sets that cover firms in a wide range of countries in a comparable fashion, such as Compustat.

The WBES has a sample design stratified by sector of economic activity, firm size and geographical location. The sampling frame excludes firms with less than 5 employees and fully government-owned firms. Sample sizes and sectoral coverage vary by country depending on the size of the economy, as measured by Gross National Income (GNI). In all countries, the sample is designed to be representative of manufacturing (Group D, ISIC Rev 3) and retail trade (ISIC 52). In larger economies, as represented by their GNI, other sectors of particular relevance for the country are included.

The main variables we exploit in our analysis are sales, expenditures on materials and labor costs. In the WBES database, the values for these variables are reported by firms for the last fiscal year. Sales consider the value of all produced goods or services purchased by third parties over the period. Material expenditures include the overall cost of raw materials and intermediate goods used in production. Finally, the payment to labor in the survey corresponds to overall employee compensation, including wages, salaries and bonuses.

While the WBES is rich in covering a wide range of countries with a standard methodology, it has certain limitations that we deal with by performing data cleaning procedures. First, we only consider firms in the manufacturing sector with information for sales, material input and labor payroll. This leads to the deletion of 23.6% of the surveyed firms. Second, we drop data from firms reporting arbitrary or unreliable figures (4.4% of the sample) and only consider data coming directly from books or estimated with some precision by managers.² Finally, we drop observations with materials and labor cost and revenue shares (the sum of the labor and material cost shares in total revenues) below the 1st or above the 99th percentiles of the respective distribution across all countries and sectors.

The main analysis considers information for the last surveyed year for LAC and peer countries. We define peers as countries with a real per capita income between USD$3000 and USD$15 000 in 2019, as measured by the World Bank’s World Development Indicators data. This income range roughly mimics the distribution of per capita income in our sample of LAC countries. Table A.1 presents the final set of countries included in the analysis. The final sample encompasses data for a year between 2009 and 2019 for 16 LAC nations and 28 Peer countries. Once expansion factors are used, the total number of firms represented by the survey in these countries is 335 728, with 31% located in LAC. LAC comprises information for Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Jamaica, Mexico, Nicaragua, Paraguay, Peru, Trinidad and Tobago, and Uruguay. The set of peer countries considers data for emerging economies in Africa (4 countries), Asia (13 countries) and Europe (11 countries).

Markups and Markdowns. To estimate markups and markdowns using (2) and (3), we assume that intermediate material inputs are fully flexible and not subject to market power by the firm. We also assume that labor is subject to no other friction than labor market power. This last assumption is somewhat strong as contracts and firing costs may prevent firms from automatically adjusting labor to their optimal level. However, available evidence (see Yeh et al. 2022,for US evidence) suggests that the resulting bias in the estimation of markups and markdowns is, most likely, of second-order.

The estimation of markups and markdowns requires values for the input-output elasticities and their corresponding expenditure share in firms’ revenues. While we can compute the latter directly from the WBES data, input-output elasticities need to be estimated. For this, we specify a constant-returns-to-scale Cobb Douglas production function, using labor and materials as inputs (as well as other inputs, such as the capital stock) to produce output. We assume that the same technology is available to all firms in the sector (in this case, manufacturing), independent of the country where the firm operates. These assumptions allow us to compute input-output elasticities for the manufacturing sector as the average cost share across all firms in the WBES dataset.³ By averaging cost share across firms, we avoid the estimated input-output elasticities to reflect the choices of individual firms, which, in turn, are likely to be affected by market power. This choice limits the variability of input-output elasticities across firms, effectively shutting down one potentially important source of variability in markups and markdowns. In Eslava et al. (2023), we show that the findings in this paper for Latin America continue to hold when allowing for variable returns to scale or when utilizing production function estimation techniques that allow for cross-firm variation in output elasticities to inputs (i.e. translog production functions estimated following Ackerberg et al. 2015).

Results

Markups and markdowns in latin america

We begin by providing a basic characterization of firms’ market power in Latin American and peer economies. We are interested in both the average level of markups and markdowns and their dispersion across firms within a country. The former is informative about the overall presence of market power in the country or region and the functional distribution of income. The dispersion of market power across firms within a country has implications for inter-personal inequality, as it provides more space for unequal earnings of equally productive workers across firms.

Table 1 displays average estimated markups and markdowns in manufacturing for each LAC country in our sample and the average country in our LAC and peer samples. The findings are broadly consistent with significant market power in product and labor markets in LAC, with average levels of markups and markdowns considerably above 1. However, the comparison with peer countries also indicates that market power in the region is not systematically higher. Indeed, market power appears less pronounced in LAC than in peer countries when considering the average estimated levels of markups and markdowns.

Table 1

Average markups and markdowns by countries and regions

	Markups		Markdowns
Elasticities:	Pooled	Country-specific	Pooled	Country-specific
	(1)	(2)	(3)	(4)
Argentina	1.753	1.715	1.271	1.526
Bolivia	2.292	1.992	1.432	1.875
Brazil	2.300	2.332	1.870	1.745
Chile	1.775	1.923	2.055	1.634
Colombia	1.774	1.918	2.320	2.021
Costa Rica	1.561	1.624	1.891	1.609
Dominican Republic	1.420	1.639	2.074	1.429
Ecuador	2.154	2.193	1.741	1.742
El Salvador	1.739	1.780	1.616	1.576
Jamaica	1.216	1.422	1.780	1.268
Mexico	1.727	1.704	1.842	1.660
Nicaragua	2.386	2.338	1.327	1.293
Paraguay	1.818	1.742	1.445	1.443
Peru	1.503	1.762	3.004	2.089
Trinidad and Tobago	1.600	1.545	1.119	1.162
Uruguay	2.399	2.350	1.742	1.878
Unweighted average of country values
Latin America	1.839	1.874	1.783	1.622
Peers	2.446	2.358	2.418	2.102

Notes: The figure shows the country-level average markups and markdowns for different input-output elasticities (pooled across all countries, or country-specific) for countries in LAC covered in the WBES. Peer countries are defined as countries with a real per capita income between US$3000 and US$15 000 in 2019, as measured by the World Bank’s World Development Indicators database. The list of peer countries is available in Appendix Table A1.

Focusing first on product market power (column 1 of table 1, our baseline estimates suggest an average markup in LAC of 1.8. The range of estimated markups across countries remains limited, varying from around 2.3 in Brazil, Nicaragua and Uruguay, to as low as 1.2 in Jamaica and 1.4 in the Dominican Republic. Turning to power in the labor market, the average markdown in LAC is 1.78, ranging from 1.11 in Trinidad and Tobago to 2.09 in Peru.

Despite the fact that both average markups and markdowns are estimated to be significantly above 1 in the region, they are lower in LAC than in peer-group countries. The average markup in peer economies is 2.4 (compared to 1.8 in LAC), while the average markdown is 2.4 in middle-income countries outside Latin America, compared to 1.7 in LAC. These differences between peer countries and Latin America are remarkably similar if we allow for country-specific elasticities, as shown in columns (2) and (4) of Table 1.⁴  Eslava et al. (2023) report similar average levels of markups and markdowns across middle-income regions (including LAC and peer economies) when focusing on a single sector—Food, Beverages and Textiles. This suggests that compositional effects are not distorting the comparison between LAC countries and their peers.

We now turn our attention from the average levels of markups and markdowns to their dispersion across firms within countries. Figure 1 displays the distribution across firms of markups and markdowns in LAC and peer countries. While there is a sizable dispersion in both measures of market power, the variance is particularly large across markdowns. In LAC, for instance, the standard deviation of markdowns (0.89) is about 1.6 times larger than for markups (0.55). Such dispersed exercise of labor market power is consistent with large cross-firm wage dispersion not explained by differences in worker productivity. Therefore, this dispersion serves as a source of inequality that cannot be accounted for by differences in productivity. At the same time, Fig. 1 shows that markups and markdowns are even more dispersed in peer economies than in LAC (st.dev=1,08 for markdowns and 0.70 for markups). In that sense, there is no broad indication that greater wage inequality in LAC arises from greater dispersion in market power.

Figure 1

Markup and Markdown Distributions. Notes: The figure shows the distribution of (log) markups (left panel) and (log) markdowns (right panel) for countries in LAC covered in the WBES. See the notes to Table 1 for detailed information about countries in the LAC and peer samples of countries.

It is worth noting that a sizable fraction (48%) of LAC firms have markdowns below 1. Taken at face value, this would indicate that a non-negligible part of the firms’ spectrum pays wages above the marginal contribution of workers to their revenues. Markups below 1 are also not rare (17%), but they remain close to 1 in most cases. Markdowns below 1 are possible in models of efficient bargaining if workers hold sufficient bargaining power (Mertens 2022). Markups could also be below 1 temporarily. Although these facts may explain the presence of below-one markups and markdowns, the substantial fraction of observations for which this occurs in studies that use the production function approach suggests that the estimates of markups and markdowns may be subject to considerable measurement error.

Market power and the labor share: firm-level evidence

One of the reasons why market power in the product and labor markets is important for inequality is that a firm’s labor revenue share will be lower if its market power is higher. By construction, estimated markups and markdowns combined are closely related to the share of firms’ revenues that go to pay workers. Equation (3) directly captures this relationship. Rewriting:

$$ \begin{align} s_{i}=\frac{w_{i}(l_{i})l_{i}}{P_{i}Q_{i}}= \frac{\alpha_{i}}{\nu_{i}\mu_{i}},\end{align} $$

(1)

where |$\alpha _{i}=\frac{\partial F(\cdot )}{\partial l_{i}}\frac{l_{i}}{Q_{i}}$|⁠, and |$s_{i}$| is the labor revenue share in firm |$i$|⁠. In other words, holding constant the elasticity of product to labor, |$\alpha _{i}$|⁠, the share of labor in the firm’s |$i$| revenue is inversely proportional to the product between the firm’s |$i$| markup and markdown. The intuition is simple: a higher markup means that the firm is extracting more revenue for each dollar of marginal cost (considering all variable inputs in that cost), while a higher markdown means that specifically for labor, each worker is paid a wage lower keeping the marginal revenue generated constant. Both mechanisms contribute to a lower labor share in the firm’s revenues.

The question we tackle in this section is what separate roles product and labor market power play in shaping levels of the labor share across firms. To this end, Fig. 2 depicts (the logarithm of) the labor revenue share against the (logarithms of) markups and markdowns across firms operating in LAC (panel a) and peer countries (panel b). All variables are demeaned with respect to the country-level mean. The bin scatter plots show these relationships across quantiles of the distribution of the logs of markups, markdowns and the firms’ labor revenue share. As evident from inspection, markdowns dominate the negative relationship with the labor share. Although higher markups reduce the labor share for a given markdown, reflecting the extraction of higher revenue per unit of marginal cost, this is empirically irrelevant. The labor share is fundamentally lower in those firms that extract higher rents from their workers in the form of wage payments below their contribution to the firm’s revenue.

Figure 2

Micro-level markups, markdowns and labor revenue share in LAC and peers. Notes: The figure shows binned-scatter plots between the (log) markups and the (log) labor revenue share (left panels) and between the (log) markdown and the (log) labor revenue share (right panels). All variables are demeaned with respect to the corresponding country-level mean. See the notes to Table 1 for detailed information about countries in the LAC and peer samples of countries.

The labor share and markdowns exhibit a strong negative correlation at the firm level. The beta coefficient in this relationship for LAC is −0.99 and holds high statistical significance. This negative correlation holds not only within LAC but also among firms in the peer group of countries. The beta coefficient for the binned scatter plot in the peer group reflects a similar relationship (represented by a beta coefficient of −0.995). This cross-regional consistency suggests that the connection between the labor share and markdowns is a fundamental aspect of firm behavior and not a peculiarity of Latin America.

Conversely, when examining market power in product markets (as measured by markups), no significant correlation is observed with the labor revenue share. Again, this absence of association isn’t unique to LAC. Firms in peer countries exhibit a very similar pattern. Essentially, firms with varying price markups over marginal costs seem to be extracting this higher margin from factors of production other than labor.

Summarizing the evidence presented so far, market power in LAC is substantial overall, exhibiting significant variation across firms, particularly in the labor market. The dispersion of markdowns across firms is closely associated with the dispersion of labor shares: Firms with considerable power in the labor market tend to pay their workers a smaller proportion of their revenue compared to other firms. Thus, high firm’s market power in the formal sector contributes to income inequality in LAC through a lower labor share. However, it’s important to note that market power in LAC is not notably higher or more dispersed than in comparable regions, and the negative correlation between markdowns and labor shares at the firm level is similar in peer countries as in LAC. Consequently, the high levels of inequality in the region do not appear to be primarily explained by exacerbated market power effects, at least at the firm level. We now investigate if these conclusions change at the aggregate level (of the formal sector covered by our data).

Figure 3

Micro-level markups, markdowns and market revenue share in LAC and peers. Notes: The figures show binned-scatter plots between the (log) labor share of revenues and the (log) market revenue share (panel a), the (log) markdown and the (log) market revenue share (panel b), and between the (log) markups and the (log) market revenue share (panel c). Variables are demeaned with respect to the corresponding country-level mean. See the notes to Table 1 for detailed information about countries in the LAC and peer samples of countries.

From micro to macro: market power and firms’ size

At the aggregate level, market power matters for inequality not only as a direct reflection of the firm-level dimensions that we have already explored. In particular, the influence of firms’ market power on the aggregate labor share of revenue is exacerbated when the firms that hold more power also happen to be large. Intuitively, because large firms contribute more to the total revenue in the economy and have more weight in aggregate payments to workers, the aggregate labor share is low not only if the average labor share is low but also if the largest firms (in terms of revenue) are also those with the lowest labor shares. Since we know that market power, particularly labor market power, is inversely related to the labor share, this also means that markdowns will reduce the labor share more if the largest firms exhibit the highest markdowns. This is the issue we investigate in this section.

Figure 3 shows that the impact of market power is amplified at the aggregate level due to a strong positive correlation between markdowns and firms’ revenue shares. The figure presents several binned-scatter plots for LAC (left panels) and peer countries (right panels), consistently displaying the market share of each firm, that is, the (log) firms’ shares in aggregate manufacturing revenue, on the horizontal axis.⁵ As we move from top to bottom, the top panels illustrate the relationship with the firm’s labor share, while the middle and bottom panels replicate the analysis for markups and markdowns, respectively.

The share of revenue accruing to the firm’s employees, i.e. the firm’s labor share, tends to be lower among those firms that possess a larger market share (i.e. a higher share of revenue in the total market revenue). This negative relationship is highly significant across LAC countries (−0.09) and even slightly more negative in the case of peer countries (−0.16). In a purely accounting sense, the primary driver of this negative association is markdowns. The association between markdowns and firms’ revenue share is both positive and significant across LAC countries (0.12), and it is even stronger in the case of peer countries (0.18). Markups, on the other hand, exhibit no relationship with the firm’s share of revenue in the manufacturing sector.

This indicates that market dynamics significantly magnify the impact of firms’ market power on the labor share. This is so because larger firms exercise more market power over their workers and pay them wages significantly below their contribution to firms’ revenue. However, this multiplier effect alone does not account for the exceptionalism in inequality observed within the LAC region. Intriguingly, the relationship between firms’ labor market power and size appears even more pronounced when examining peer countries.

Conclusions

This study contributes to the understanding of market power in LAC shedding light on its potential implications for income distribution. Our findings underscore a negative association between market power among regional firms, as measured by markdowns and markups, and the labor share. This phenomenon is compounded by the fact that formal firms in the region are predominantly owned by the wealthiest families. Consequently, the increased profits and capital returns linked to market power serve to exacerbate existing economic inequality.

Our paper distinguishes between two distinct manifestations of market power: labor market power and product market power. Notably, it is labor market power—specifically, the ability of firms to set wages below the marginal contributions of workers to firm revenue—that predominantly drives our results. We observe a significant negative correlation between labor market power and the labor share at the firm level. Furthermore, we identify a multiplier effect through the interplay with the firm size distribution. Given that labor market power is more pronounced among firms with higher revenues, which exert greater influence on the aggregate labor share, the impacts of market power on income distribution are magnified.

Our research also contributes to benchmarking LAC by comparing regional market power with countries possessing similar GDP per capita but higher levels of inequality. The evidence gleaned from this analysis is enlightening. While market power undeniably plays a significant role in regional inequality, LAC does not distinctly stand out in any of the dimensions we’ve explored. Thus, the roots of LAC inequality exceptionalism likely lie elsewhere.

AUTHORS' CONTRIBUTIONS

Conceptualization: Eslava, Garcia-Marín and Messina Data curation: Garcia-Marín Formal analysis: Garcia-Marín Methodology: Eslava, Garcia-Marín and Messina Project administration: Eslava, Garcia-Marín and Messina Visualization: Garcia-Marín Writing—original draft: Eslava, Garcia-Marín and Messina Writing—review & editing: Eslava, Garcia-Marín and Messina

FINANCIAL SUPPORT

This article was written for the Latin American and Caribbean Inequality Review, funded by the International Inequalities Institute at the London School of Economics and Political Science, the Inter-American Development Bank, Yale University, and the Institute for Fiscal Studies. The views expressed are those of the authors and not necessarily of the funders. Messina also acknowledges financial support from Project PID2021-124237NBI00 (financed by MCIN/ AEI /10.13039/501100011033/ and by FEDER Una manera de hacer Europa) and from Generalitat Valenciana, Consellería de Innovación, Universidades, Ciencia y Sociedad Digital through project Prometeo CIPROM/2021/068.

CONFLICT OF INTEREST

All authors have no conflict of interest regarding this manuscript.

DATA AVAILABILITY

Data available upon request.

Footnotes

1

We rely on the Gini series from the World Income Distribution Database (WIID) based on percentiles to compute these numbers. WIID pools information from various sources and definitions of income inequality. These may lead to differences between the official numbers reported by each country and the information in the WIID in some cases. However, as discussed in Gradín (2021), the inequality trends in WIID and other sources, such as ECLAC or SEDLAS are highly consistent.

2

Interviewers are required to complete a brief questionnaire after the WBES interview. This questionnaire assesses whether managers’ responses regarding quantitative information requested during the interview were checked with their books or obtained through appropriate inquiries to ensure the trustworthiness of their responses. We only consider firm-level observations where interviewers believed these necessary checks were conducted.

3

The baseline estimates compute the production cost as the sum of labor payroll, expenditure in materials, energy consumption and the cost of capital services. To compute this last variable, we rely on the book value of fixed assets and assume a rental rate and depreciation rate of capital equal to 5% each. When capital stock is missing in the data, we impute it as the predicted value from country-specific log-log regressions, using sales, employment and two-digit industry-year fixed effects as predictors.

4

The conclusions in the paper are unaffected by the introduction of country-specific elasticities. Hence, we report in the rest of the study the pooled results for simplicity.

5

All variables in the graphs are demeaned with respect to the country average.

Appendix I

Table A1

Observations per country

	Country	Period	Obs.		Country	Period	Obs.
Latin America		—	102 063	Peers (cont’d)		—
1	Argentina	2017	14 242	23	Indonesia	2015	25 536
2	Bolivia	2017	864	24	Iraq	2022	549
3	Brazil	2009	47 420	25	Jordan	2019	1191
4	Chile	2010	3036	26	Kazakhstan	2019	5952
5	Colombia	2017	6566	27	Lebanon	2019	1622
6	Costa Rica	2010	682	28	Malaysia	2019	10 801
7	Dominican Republic	2010	1,685	29	Mauritius	2009	659
8	Ecuador	2017	2918	30	Moldova	2019	1341
9	El Salvador	2016	1590	31	Mongolia	2019	882
10	Jamaica	2010	492	32	Morocco	2019	2992
11	Mexico	2010	16 947	33	North Macedonia	2019	872
12	Nicaragua	2016	1102	34	Philippines	2015	8000
13	Paraguay	2017	976	35	Romania	2019	5987
14	Peru	2017	1939	36	Russia	2019	43 582
15	Trinidad and Tobago	2010	797	37	Serbia	2019	2605
16	Uruguay	2010	807	38	Sri Lanka	2011	8895
Peers		—	231 392	39	Thailand	2016	8677
17	Belarus	2018	6654	40	Tunisia	2013	4892
18	Bosnia and Herzegovina	2019	1418	41	Turkiye	2019	23 230
19	Bulgaria	2019	6353	42	Uzbekistan	2019	4617
20	Croatia	2019	3774	43	Vietnam	2015	16 108
21	Egypt, Arab Rep.	2016	30 735	44	West Bank and Gaza	2019	2971
22	Georgia	2019	495		Total		333 450

Author notes

We thank Raquel Fernández, Francisco H. G. Ferreira, Joana Silva and seminar participants at the 2022 LACIR Workshop for useful comments and suggestions. Jorge Jara provided superb research assistance. Messina acknowledges financial support from Project PID2021-124237NB-I00 (financed by MCIN/ AEI /10.13039/501100011033/ and by FEDER Una manera de hacer Europa) and from Generalitat Valenciana, Consellería de Innovación, Universidades, Ciencia y Sociedad Digital through project Prometeo CIPROM/2021/068. Contact: [email protected], [email protected], [email protected].