4. Monopoly’s neglected twin? The effect of labour market concentration on wages and inequality

Large companies with monopoly power can boost their profits by imposing high prices on consumers. But large companies may also be able to suppress wages if workers have few alternative employment options within reasonable commuting distance, i.e. if local employment is highly concentrated. The resulting redistribution of income from workers to company owners hurts workers and reduces overall economic efficiency, as companies paying low wages generally employ fewer workers and curtail output. In many OECD countries, industry sales have become more concentrated (Bajgar et al., 2019[1]) while the share of wages in total income has declined (Autor et al., 2017[2]; Schwellnus et al., 2018[3]), raising the question whether increased sales concentration has gone together with increased labour market concentration and wage-setting power.

This chapter analyses the links between labour market concentration, wages and inequality. The analysis is based on linked employer-employee data from seven OECD countries for which relevant and comparable measures of labour market concentration can be constructed.2 The main focus is on labour market concentration at the level of detailed industries and regions. The paper presents comparable descriptive evidence on the degree of local labour market concentration across industry groups (manufacturing and services), geographical areas (rural and urban areas) and worker groups (low- and high-qualified workers), as well as its changes over time. It further presents econometric evidence on the causal effect of labour market concentration on wages, distinguishing between low- and high-qualified workers and testing whether the wage effects have changed over time.

A growing literature studies wage-setting power in single-country contexts, but comparable estimates of wage-setting power and its implications for wage growth and wage inequality across OECD countries are still missing. One strand of the literature analyses concentration in local labour markets and its relation to wages in the United States (Rinz, 2020[4]; Benmelech, Bergman and Kim, 2020[5]), France (Marinescu, Ouss and Pape, 2020[6]), the United Kingdom (Abel, Tenreyro and Thwaites, 2018[7]), Austria (Jarosch, Nimczik and Sorkin, 2019[8]), and Portugal (Martins, 2018[9]).3 Some recent studies have pointed to the key role of job mobility in shaping workers’ outside options and thus the relationship between concentration and wages (Caldwell and Danieli, 2018[10]; Jarosch, Nimczik and Sorkin, 2019[8]; Schubert, Stansbury and Taska, 2020[11]; Berger, Herkenhoff and Mongey, 2019[12]).A limitation of these studies is the lack of a unified definition of local labour markets and concentration measures, which limits cross-country comparability.4 A second strand of literature attempts to estimate the labour supply elasticity to the individual firm, a key theoretical determinant of wage-setting power (Sokolova and Sorensen, 2020[13]; Manning, 2011[14]; Bassier, Dube and Naidu, 2021[15]). While most empirical estimates of the labour supply elasticity to the individual firm suggest significant potential wage-setting power, the extent to which firms actually exercise it has not yet been clearly established (Manning, 2020[16]).

This chapter makes three main contributions. First, it analyses developments in labour market concentration since the early 2000s from a cross-country perspective, drawing on comprehensive administrative data. The data treatment and the definitions of wages and local labour market concentration are harmonised across countries as much as possible, improving comparability. Second, the chapter provides econometric estimates of the impact of labour market concentration on wages based on instrumental variable techniques, holding constant a large number of potential confounding factors, including unobserved productivity differences between local labour markets. Third, the chapter puts the analysis of labour market concentration into the broader context of firms’ wage-setting power (by providing estimates of the labour supply elasticity to the individual firm) and recent trends in product markets (by providing evidence on the links between sales and employment concentration).

The main results of the chapter are as follows. On average across the covered countries, around 20% of the workforce is employed in highly concentrated local labour markets, with the share being even higher for rural and manufacturing workers. The consequent reduction in workers’ job options puts significant downward pressure on wages, especially those of low-qualified workers, thus raising overall wage inequality. However, local labour market concentration has tended to slightly decline over the period 2003-17 despite rising sales concentration. These results imply that labour market concentration is a relevant issue from the perspective of public policies aiming to address inequality but cannot explain broader economic trends related to wage stagnation and the decline in the labour income share experienced by number of countries over the past two decades.

Wage-setting institutions such as minimum wages and collective bargaining can counter-balance negative wage effects from labour market concentration, and integrating labour market power considerations into merger control can prevent firms from reaching dominant positions in the first place. In many cases, reforming policy settings in product and labour markets that limit competition would curb market income inequality while at the same time raising economic growth and employment. For instance, reducing regulatory barriers to worker mobility (such as professional licencing or non-compete clauses) and business dynamism (such as regulatory obstacles to firm entry and growth) would raise workers’ wages relative to productivity while allowing high-performing firms to expand more easily.

The next Section describes a conceptual framework linking public policies, wage-setting power and wages. Section 4.3 describes the linked employer-employee data used in the empirical analysis, as well as the methodology used to construct the measures of labour market concentration used in the descriptive and econometric analysis. Section 4.4 reports the descriptive evidence on labour market concentration across countries, industries, geographical areas and worker groups and over time. Section 4.5 presents the econometric results on the effects of labour market concentration on wages across worker groups and over time. Section 4.6 concludes and discusses the implications of the analysis for public policy.

Workers’ wages are determined by their individual characteristics, such as qualifications, experience and gender, but also by the degree of firms’ wage-setting power (Figure 4.1). Wage-setting power arises when workers have only limited job options due to a lack of available jobs in their relevant labour market and/or when there is limited labour mobility. A lack of available jobs may arise when hiring is concentrated among a small number of potential employers, or potential employers post few vacancies relative to the number of job seekers, for example during economic downturns. Even when there is no lack of job vacancies, workers may nonetheless have little options outside their current job if there is low job mobility. Job mobility may be limited for different reasons, e.g. because workers incur monetary costs (including search costs related to gathering information on job opportunities and their suitability) or due to preference-related non-monetary costs from moving to a different firm, occupation, industry or region. Firms with wage-setting power can afford paying workers lower wages than other firms because only a fraction of its workers would quit to take up higher-paying jobs. This reduces average wages and can contribute to wage inequality.

While limited job availability and/or labour market frictions give employers the power to set wages, they may choose or be legally constrained not to exercise it. For instance, firms may anticipate that employees perceive low wages as unfair and that they may consequently cut back on effort. This would reduce the firm’s output and limit the gains from paying lower wages. In other words, the firm may not fully exercise its wage-setting power out of efficiency-wage considerations (Akerlof and Yellen, 1990[17]). Even if setting wages below workers’ (marginal) productivity is optimal from the firm’s point of view, institutional constraints such as minimum wages or collectively-agreed wage floors may prevent it from doing so (Azar et al., 2019[18]).5 The measurement of wage-setting power thus faces the challenge of distinguishing between firms’ potential to set wages below marginal productivity and the extent to which they actually use it.

The literature has traditionally measured wage-setting power by the labour supply elasticity to the individual firm. The rationale is that firms ultimately derive their wage-setting power from the fact that workers do not switch jobs in response to small wage differentials between firms. The labour supply elasticity encompasses both wage-setting power deriving from a limited number of employers in a given labour market (“classical monopsony”) and from frictions in labour markets related to search and hiring costs (“modern monopsony”) (Manning, 2020[16]). When there is a large number of effectively available employers, i.e. employment is not concentrated among a small number of firms and frictions are low, the labour supply elasticity is expected to be high, theoretically approaching infinity in the case of a perfectly-competitive labour market without frictions. By contrast, a much smaller labour supply elasticity is expected when there is only a small number of effectively available employers, indicating the presence of wage-setting power. On average across countries where data to estimate the labour supply elasticity are available, its estimated value is around 2, which is consistent with estimates from previous studies (Sokolova and Sorensen, 2020[13]) and implies significant wage-setting power (Annex Figure 4.B.1). However, estimates of the labour supply elasticity to the individual firm may be affected by significant measurement and endogeneity issues.

Complementing the traditional approach based on the labour supply elasticity, an emerging literature has approximated firms’ wage-setting power by local labour market concentration (Schubert, Stansbury and Taska, 2020[11]; Azar et al., 2020[19]; Marinescu, Ouss and Pape, 2020[6]; Rinz, 2020[4]). Unlike the labour supply elasticity, labour market concentration is a partial measure of wage-setting power that does not account for search and hiring frictions. However, it can be directly observed in the data and allows analysing whether wage-setting power is actually exercised by employers by relating concentration to wages at the local labour market level, which is infeasible using the labour supply elasticity.6 The remainder of the chapter therefore focuses on local labour market concentration.

In contrast to product market concentration, which is often measured at the national level, labour market concentration is typically measured at the local level (Rinz, 2020[4]). Adding the geographical dimension accounts for the fact that there are large barriers to worker mobility across regions, with workers typically searching for new jobs in a local area within commuting distance from their home (Manning and Petrongolo, 2017[20]). By contrast, competition in product markets often takes place at the national or international levels. Indeed, in most OECD countries even local services (e.g. physical retail; hotels and restaurants) are often provided by national and multinational chains.7

The definition of a local labour market is too narrow if many workers can find alternative employment in another labour market (i.e. there is a high degree of worker mobility across local labour market boundaries), whereas it is too broad if many jobs within the local labour market are actually not accessible to workers. Ideally, boundaries of local labour markets are defined such that most jobs inside the same market are available to all workers in the market, while worker flows across markets are minimal (Nimczik, 2020[21]). Most of the literature has defined the relevant labour market at the level of occupation by commuting zone (Martins, 2018[9]; Marinescu, Ouss and Pape, 2020[6]; Schubert, Stansbury and Taska, 2020[11]) – the rationale being that there are fewer barriers to job mobility within occupations and within commuting zones. Another common definition of the local labour market is at the level of industries by commuting zones (Benmelech, Bergman and Kim, 2020[5]; Rinz, 2020[4]), reflecting the fact that worker mobility is typically much higher within industries than between them.8

The preferred definition of the local labour market used in this chapter is at the level of 3-digit industries (around 230) and TL3 regions (generally comparable to French départements or Spanish provincias). TL3 regions overlap with commuting zones but do not always coincide with them. The chosen definition of the local labour market represents a compromise between country coverage and a sufficiently narrow definition of local labour markets (Box 4.1). The main measure of local labour market concentration used in the analysis is the Herfindahl-Hirschman-Index (HHI).9 The HHI can take values between 0 (when a large number of small firms accounts for very small shares of total hiring) and 10,000 (in the extreme case when a single firm dominates the entire market). Larger values thus indicate higher levels of concentration, with values above 2500 typically considered as indicating high concentration (Marinescu and Hovenkamp, 2019[22]; OECD, 2020[23]; OECD, 2019[24]).10

In order to analyse the degree to which firms exercise their potential wage-setting power, wages are related to local labour market concentration based on the following equation:11

where w denotes wages; HHI the Herfindahl-Hirschman index of local labour market concentration; and ${\epsilon }_{ijmt}$ the error term. Subscripts i, j, m and t denote, respectively, workers, firms, local labour markets and years; and ${\rho }_{t}$ year fixed effects. Worker fixed effects ${\mu }_{i}$ control for all time-invariant, individual determinants of wages, both observable and unobservable. This ensures that the estimated ${\beta }_{1}$ can be interpreted as the effect of concentration on the wages of similar workers. It further removes any potential endogeneity arising from a correlation between worker characteristics and concentration, such as a higher prevalence of low-qualified workers in highly concentrated regions and industries.12

Another econometric concern that needs to be addressed is the possible spurious correlation between concentration and wages at the level of local labour markets. For example, urban areas might attract a larger number of firms – leading to lower concentration – and may at the same time be more productive, for instance due to agglomeration effects (Glaeser, 2010[26]). The inclusion of local labour market fixed effects allows controlling for time-invariant omitted factors that may be correlated with both wages and concentration at the local labour market level. In other words, labour market fixed effects allow isolating the pure market power effect of labour market concentration from the effect of other factors that may be correlated with concentration and also affect wages, such as average productivity or average firm size in the local labour market.

By construction, the inclusion of local labour market fixed effects cannot address endogeneity issues related to time-varying omitted factors, such as productivity shocks (rather than productivity levels) that may be correlated with both concentration and wages at the local labour market level. For instance, an unobserved positive productivity shock in a local labour market may lead to market entry of new firms, reduce concentration and raise wages. This would bias the estimated coefficient on local labour market concentration down, leading to an estimated coefficient that is more negative than the true wage effect of concentration and thus overstating the effect of concentration on wages.

The potential bias from unobserved productivity shocks is addressed by using an instrumental variable for local labour market concentration. Following seminal studies in the academic literature (Martins, 2018[9]; Marinescu, Ouss and Pape, 2020[6]; Azar et al., 2019[18]), the average inverse number of firms in the same year and industry but in other regions, weighted by industry-employment shares of each region, is used as an instrumental variable. The rationale is that the number of firms in a market is strongly and inversely related to concentration but unrelated to productivity shocks to individual firms. Unlike potential instrumental variables that are a function of firm size (such as average concentration in the same industry but other regions), this variable has the advantage of being invariant to productivity shocks to individual firms.13

The analysis is conducted separately on individual-level data for each country in a distributed micro data approach. In contrast to individual-level data that are subject to strict confidentiality restrictions in many countries, aggregate and semi-aggregate descriptive statistics and regression results based on the micro data can generally be distributed. Country-level estimates are averaged following established procedures for the statistical aggregation of regression estimates.14

The analysis in this paper is based on a newly created harmonised cross-country dataset based on linked employer-employee data that provide information on employees and the firms where they work. The data cover the universe of workers (or a large representative sample) in each country and are of very high quality (Criscuolo et al., 2020[27]), which allows calculating precise measures of labour market concentration. The availability of employee information furthermore allows controlling for worker characteristics when estimating the effect of concentration on wages. In particular, linked employer-employee data allow accurately measuring concentration not only in employment but also in hiring, which is not possible with firm-level data alone.15 New hires are identified from workers switching firms for their main job. All country datasets contain a core set of comparable information on workers (wage, gender, age and location) and firms (industry and size). Most datasets also contain a number of additional relevant variables, such as hours worked, occupation and education, but there are large differences in availability and detail. The main results presented in this paper rely on the core set of comparable characteristics. Additional analysis as well as a large set of robustness checks exploit the more detailed information for different subsets of countries.

The analysis requires making a number of data harmonisation choices. A basic prerequisite for measuring labour market concentration in a specific local labour market is the availability of information on the firm’s industry (at the 3-digit level) and information on the location of the worker at the level of TL3 regions, corresponding roughly to provinces (e.g. provincias in Spain or départements in France) or groups of smaller units such as counties or districts (e.g. in Austria).16 The wage regressions are based on monthly wages.17

The main analysis based on the local labour market definition at the 3-digit industry and TL3 region level covers seven countries over a period from the early 2000s up to 2017.18 Where these detailed levels are unavailable, a number of descriptive results are reported at a coarser level of aggregation (2-digit industry or TL2 region) for a maximum of 11 countries.19 The analysis covers dependent employees in all sectors of the private economy other than agriculture, mining and utilities. This covers on average 97% of total private sector employment.20 Industries are classified according to the International Standard Industrial Classification (ISIC), revision 4. TL3 regions are classified into rural and urban according to a harmonised classification by Fadic et al. (2019[28]).21

On average across countries, local labour markets are moderately concentrated but around 20% of workers are employed in markets with high levels of concentration (Figure 4.2). The cross-country average of local labour market concentration for the average worker as measured by the employment-weighted HHI is around 1600 (Panel A), which is the threshold conventionally used in merger reviews to indicate moderate sales concentration (US Justice Department and the Federal Trade Commission, 2010[29]). Moreover, around 20% of workers are employed in highly concentrated labour markets (based on the conventional threshold of an HHI above 2500). This share is substantially higher in Austria, Denmark, Finland and France (Panel B).

Cross-country differences in local labour market concentration can reflect structural differences between countries, but may also be due to differences in the average size of TL3 regions. For instance, in some countries, such as Austria and Finland, average employment of TL3 regions is lower than in other European countries; whereas in Costa Rica, it is higher. This may introduce an upward bias in measured concentration in Austria and Finland relative to the other countries. Given the measurement challenges characterising cross-country comparisons of concentration levels, the remainder of the paper focuses on within-country differences in concentration across geographical areas, industries, worker groups and over time rather than on cross-country differences in local labour market concentration.

The share of workers exposed to high local labour market concentration in rural areas (around 30%) is twice as large than in urban ones (around 15%) (Figure 4.3). In rural areas, there are fewer job opportunities for workers in any given industry due to a limited number of potential employers. In urban areas, there are significantly more job opportunities, as firms generally tend to locate close to large population centres to access a larger pool of workers and consumers and benefit from agglomeration economies (Glaeser, 2010[26]). In some industries, firms also tend to co-locate with firms in the same or closely related industries, generally in urban areas, thereby expanding job opportunities for workers with industry-specific skills (Moretti, 2013[30]). While the rural-urban differential in local labour market concentration holds for most countries covered by the analysis, in Denmark and in the Slovak Republic concentration is lower in rural than urban areas. This may reflect the fact that in these countries a number of very large employers, in particular multi-national firms, account for a very large share of employment in the capital region.

A significantly higher share of manufacturing workers (around 40%) than services workers (around 15%) is employed in highly-concentrated labour markets (Figure 4.4). Manufacturing is generally more geographically concentrated than services, which can be explained by larger scale economies in manufacturing and higher tradability (Gervais and Jensen, 2019[31]). For instance, the benefits for an automobile firm to locate close to its customers is small since scale economies are large and automobiles can be shipped to the location of final demand at low cost. By contrast, even though in some digitally-intensive services sectors economies of scale are becoming increasingly important and remote provision is becoming more feasible, in many services sectors economies of scale remain limited and provision still requires physical presence.22

The rural-urban differential in local labour market concentration holds within industries and the manufacturing-services differential holds within regional groups (Annex Figure 4.A.1). This suggests the concentration differentials reported above cannot be explained by a higher tendency of manufacturing firms to locate in rural areas with higher levels of concentration.

Low-qualified workers tend to face significantly higher concentration within their local labour markets (Figure 4.5).23 The lack of a link between a local labour market’s workforce composition and its degree of employer concentration partly reflects the fact that a high share of low-qualified workers are employed in low-concentrated urban services sectors. By contrast, within their local labour markets, i.e. within a given industry and geographical area, low-qualified workers generally have a smaller number of job options than their medium and high-qualified peers. Exposure to local labour market concentration appears to be lowest for medium-qualified workers. The ratio of concentration for low-qualified workers relative to the mean reported in Figure 4.5 implies that the average low-qualified worker is employed in a moderately concentrated labour market with an HHI of around 2400, whereas the average medium-qualified worker is employed in a low-concentrated local labour market with an HHI of around 1300.24

The patterns in local labour market concentration documented at the 3-digit industry by TL3 region local labour market level are robust to alternative definitions. Defining local labour markets in terms of employment rather than hiring has no quantitative effect of at the level of concentration (Annex Figure 4.A.5). Data available for less detailed aggregations of industry and/or region for a larger sample of 11 countries are reported in Annex Figure 4.A.6. While the cross-country pattern of concentration documented above is broadly similar to the one in Figure 4.2, the measured levels of the HHI decrease at this more aggregate local labour market definition, reflecting a mechanical increase in the number of firms when regions or industry boundaries are expanded.

There is no evidence of an increase in local labour market concentration over the period 2003-2017. Averaging across countries, there is a slight decline until 2008 and a broadly stable trend since (Figure 4.6). The initial decline in local labour market concentration is mainly driven by Finland and Spain, which may partly be explained by the rapid shift from manufacturing to lower-concentrated services in these countries in the run-up to the global economic crisis of 2008-09. But changes in industry composition do not appear to be the sole explanation for this initial decline, given that concentration has declined even within services.25

The trend decline in local labour market concentration has occurred despite an increase in sales concentration over the past two decades. Available measures of sales concentration typically refer to industry sales at the national level, whereas local labour market concentration is measured as hiring or employment. This suggest that there are two possible explanations for the observed decoupling of local labour market concentration from national sales. Firstly, national employment concentration may have decoupled from national sales concentration if firms are increasingly able to scale up production without increasing employment, including through domestic and international outsourcing. Secondly, local employment concentration may have decoupled from national employment concentration, which may for instance be the case if large national employers increasingly enter each other’s local labour markets (Rinz, 2020[4]).26 Box 4.2 suggests that the main explanation is the decoupling of national employment concentration from national sales concentration.

Local labour market concentration has a significantly negative effect on wages, even after accounting for differences in workforce composition and productivity across local labour markets (Figure 4.8). In other words, a worker employed in a highly-concentrated local labour market earns a significantly lower wage than a worker with similar characteristics in a low-concentrated market with similar average productivity. On average across countries, the mean reduction in wages from a 1,000 point increase in the HHI is around 2%, which is broadly in line with existing estimates from country-level studies relying on occupation-by-region based local labour market definitions (Martins, 2018[9]; Marinescu, Ouss and Pape, 2020[6]).27 While all country-level coefficients estimated from Equation 4.1 are negative as predicted by theory, some of them are estimated with large error, which precludes direct cross-country comparisons.28 Consequently, the remainder of the section focuses on the average cross-country effect.

Based on the estimated average cross-country effect of concentration, wages at the 90th percentile of employment-weighted local labour market concentration (i.e. the 90th percentile of workers rather than the 90th percentile of local labour markets) are 7% lower than at the 10th percentile (Annex Table 4.A.2). On average across countries, for workers at the 90th percentile the value of the HHI is about 4000, whereas for workers at the 10th percentile it is around 150. Based on the average wage effect reported in Figure 4.8, this difference in labour market concentration translates into an economically significant wage difference of 7%. The implied wage difference would be even larger (around 16%) between workers at the 90th and 10th percentiles of the unweighted concentration distribution, given that concentration is typically highest in small markets with low employment.

The wage effects of local labour market concentration tend to be driven by low-qualified workers (Annex Table 4.A.3). On average across the three countries for which disaggregated coefficients by skill group can be estimated, the wage effect for low- and medium- qualified workers of a 1,000 point increase in the HHI is about 2% whereas the effect for high-skilled workers is close to zero.29 At the same time, low-qualified workers face about 40% higher local labour market concentration than high-qualified ones. Combining the effects on low-qualified workers’ wages from the stronger wage response to concentration with the higher exposure to concentration suggests that labour market concentration reduces low-qualified workers’ wages by around 6% relative to those of high-skilled ones.

The negative wage effect of labour market concentration has tended to become stronger over time (Figure 4.9). The estimated wage effect is about twice as strong in 2015-2017 than in 2003-2005, with the difference being statistically significant at the 5% level. The increasingly negative wage response to concentration suggests that firms are increasingly exercising their wage-setting power. To some extent, this could reflect the weakening of workers’ bargaining position due to changes in wage-setting institutions such as minimum wages and collective bargaining, or increased exposure to domestic and international outsourcing (Abel, Tenreyro and Thwaites, 2018[7]).

The analysis in this chapter covers the degree of labour market concentration, the extent to which it varies across different segments of the labour market and over time, as well as its effects on wages. The main results are that (1) on average across the covered countries a significant share of workers (around 20%) are employed in highly-concentrated labour markets, especially in manufacturing and rural areas; (2) labour market concentration has negative effects on wages; (3) wage effects from labour market concentration tend to be particularly negative for low-qualified workers; and (4) wage effects have tended to become more negative over time. These results can potentially inform a range of public policy areas.

The high degree of labour market concentration for a significant share of workers and the increasingly negative effect of concentration on wages suggest that wage-setting policies may play a useful role in counter-balancing wage-setting power. In a labour market where firms have a high degree of wage-setting power, statutory or collectively-bargained wage floors can increase wages without reducing employment by limiting firms’ scope to reduce wages below workers’ reservation wages (Card and Krueger, 1994[36]; Manning, 2020[16]; OECD, 2019[37]; OECD, 2018[38]).30 In a number of OECD countries, the real value of the minimum wage and the share of workers covered by collective bargaining agreements have tended to decline over the past decades, suggesting room for policy action.

Wage-setting policies may become particularly relevant in the context of the emergence of digital platforms that have gained dominant positions in some local labour markets. Many digital platforms, including in ride-hailing, food delivery and retail, rely mainly on low-skilled self-employed workers for whom the wage effects of local labour market concentration are particularly negative.31 Collective bargaining over wages and working conditions on the part of these self-employed workers should not be prevented by the undue application of non-collusion clauses in competition law (OECD, 2020[23]).

A high degree of wage-setting power may also indicate the need for public policies to directly address labour market concentration, especially in a context where collective bargaining is under pressure, trade union density is declining and “winner-takes-most” dynamics are emerging in some sectors of the economy. In many jurisdictions, competition authorities already have the legal mandate to include labour market power as a consideration in reviews of mergers and acquisitions (OECD, 2020[23])). One way to operationalise labour market power in merger reviews is to define a threshold above which a labour market is considered to be highly concentrated, which would then trigger further investigation (Marinescu and Hovenkamp, 2019[22]). Even though the analysis in this chapter suggests that increasing product-market concentration does, on average, not imply higher labour market concentration, anecdotal evidence nonetheless suggests that, in some sectors of the economy, increased product market concentration has been associated with increased labour market concentration. For instance, large digital platforms in the transport and retail sectors have become dominant employers in some local labour markets.

Excessive wage setting power may further be tackled by policies to promote voluntary job mobility, which would increase the job options effectively available to workers. While job mobility is partly determined by individual preferences over non-wage job characteristics, monetary costs to mobility can be influenced by public policies. Such costs could, for instance, be reduced by strengthening active labour market policies; improving the portability of benefits; regulatory action that reduces legal or contractual barriers to job mobility (occupational licensing, non-compete and non-poaching agreements, portability of workers’ ratings across digital platforms); and through housing and transport policies. The uptake of telework has effectively expanded the geographical boundaries of worker’s job options but teleworkable jobs and occupations are typically located at the top of the skill distribution (OECD, 2021[39]; Espinoza and Reznikova, 2020[40]). While policies to support telework would thus tend to raise average wages, they may further widen the gap between workers at the top and the rest of the wage distribution.

References

[7] Abel, W., S. Tenreyro and G. Thwaites (2018), Monopsony in the UK, https://ssrn.com/abstract=3270944.

[41] Abowd, J., F. Kramarz and D. Margolis (1999), “High Wage Workers and High Wage Firms”, Econometrica, Vol. 67/2, pp. 251-333, https://about.jstor.org/terms (accessed on 2 December 2019).

[17] Akerlof, G. and J. Yellen (1990), “The Fair Wage-Effort Hypothesis and Unemployment”, The Quarterly Journal of Economics, Vol. 105/2, p. 255, http://dx.doi.org/10.2307/2937787.

[33] Andrews, D., C. Criscuolo and P. Gal (2016), The Best versus the Rest, https://doi.org/10.1787/63629cc9-en.

[14] Ashenfelter, O. and D. Card (eds.) (2011), Imperfect Competition in the Labor Market, Elsevier.

[2] Autor, D. et al. (2017), “Concentrating on the Fall of the Labor Share”, American Economic Review, Vol. 107/5, pp. 180-185, http://dx.doi.org/10.1257/aer.p20171102.

[32] Autor, D., L. Katz and J. Van Reenen (2020), “The Fall of the Labor Share and the Rise of Superstar Firms”, The Quarterly Journal of Economics, Vol. 135, pp. 645–709, https://doi.org/10.1093/qje/qjaa004.

[18] Azar, J. et al. (2019), Minimum Wage Employment Effects and Labor Market Concentration, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w26101.

[42] Azar, J., I. Marinescu and M. Steinbaum (2019), “Measuring Labor Market Power Two Ways”, AEA Papers and Proceedings, Vol. 109, pp. 317-321, http://dx.doi.org/10.1257/pandp.20191068.

[19] Azar, J. et al. (2020), “Concentration in US labor markets: Evidence from online vacancy data”, Labour Economics, Vol. 66, p. 101886, http://dx.doi.org/10.1016/j.labeco.2020.101886.

[1] Bajgar, M. et al. (2019), “Industry Concentration in Europe and North America”, OECD Productivity Working Papers, No. 18, OECD Publishing, Paris, https://dx.doi.org/10.1787/2ff98246-en.

[25] Bassanini, A. et al. (2022), Labour market concentration in Europe.

[15] Bassier, I., A. Dube and S. Naidu (2021), “Monopsony in Movers: The Elasticity of Labor Supply to Firm Wage Policies”, Journal of Human Resources, pp. 0319-10111R1, http://dx.doi.org/10.3368/jhr.monopsony.0319-10111r1.

[5] Benmelech, E., N. Bergman and H. Kim (2020), “Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?”, Journal of Human Resources, pp. 0119-10007R1, http://dx.doi.org/10.3368/jhr.monopsony.0119-10007r1.

[12] Berger, D., K. Herkenhoff and S. Mongey (2019), Labor Market Power, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w25719.

[34] Berlingieri, G., S. Calligaris and C. Criscuolo (2018), “The Productivity-Wage Premium: Does Size Still Matter in a Service Economy?”, AEA Papers and Proceedings, Vol. 108, pp. 328-33, http://dx.doi.org/10.1257/pandp.20181068.

[49] Bound, J., D. Jaeger and R. Baker (1995), “Problems with Instrumental Variables Estimation When the Correlation Between the Instruments and the Endogeneous Explanatory Variable is Weak”, Journal of the American Statistical Association, Vol. 90/430, p. 443, http://dx.doi.org/10.2307/2291055.

[10] Caldwell, S. and O. Danieli (2018), Outside Options in the Labor Market.

[36] Card, D. and A. Krueger (1994), “Minimum Wages and Employment: A Case-Study of the Fast-Food Industry in New Jersey and Pennsylvania”, American Economic Review, Vol. 84/4, pp. 772-793, https://www.jstor.org/stable/2118030.

[48] Causa, O., N. Luu and M. Abendschein (forthcoming), Labour market transitions across OECD countries: stylised facts.

[27] Criscuolo, C. et al. (2020), “Workforce composition, productivity and pay: the role of firms in wage inequality”, OECD Economics Department Working Papers, No. 1603, OECD Publishing, Paris, https://dx.doi.org/10.1787/52ab4e26-en.

[50] DerSimonian, R. and N. Laird (1986), “Meta-analysis in clinical trials”, Controlled Clinical Trials, Vol. 7/3, pp. 177-188, http://dx.doi.org/10.1016/0197-2456(86)90046-2.

[43] Dube, A., A. Manning and S. Naidu (2018), Monopsony and Employer Mis-optimization Explain Why Wages Bunch at Round Numbers, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w24991.

[40] Espinoza, R. and L. Reznikova (2020), “Who can log in? The importance of skills for the feasibility of teleworking arrangements across OECD countries”, OECD Social, Employment and Migration Working Papers, No. 242, OECD Publishing, Paris, https://dx.doi.org/10.1787/3f115a10-en.

[28] Fadic, M. et al. (2019), “Classifying small (TL3) regions based on metropolitan population, low density and remoteness”, OECD Regional Development Working Papers, No. 2019/06, OECD Publishing, Paris, https://dx.doi.org/10.1787/b902cc00-en.

[31] Gervais, A. and J. Jensen (2019), “The tradability of services: Geographic concentration and trade costs”, Journal of International Economics, Vol. 118, pp. 331-350, http://dx.doi.org/10.1016/j.jinteco.2019.03.003.

[26] Glaeser, E. (ed.) (2010), Agglomeration Economics, The University of Chicago Press, http://www.nber.org/books/glae08-1.

[46] Haltiwanger, J. (2021), Rising between Firm Inequality and Declining Labor Market Fluidity: Evidence of a Changing Job Ladder, National Bureau of Economic Research.

[8] Jarosch, G., J. Nimczik and I. Sorkin (2019), Granular Search, Market Structure, and Wages, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w26239.

[47] Langella, M. and A. Manning (2021), The measure of monopsony, https://cep.lse.ac.uk/pubs/download/dp1780.pdf.

[16] Manning, A. (2020), “Monopsony in Labor Markets: A Review”, ILR Review, Vol. 74/1, pp. 3-26, http://dx.doi.org/10.1177/0019793920922499.

[45] Manning, A. (2003), Monopsony in Motion, Princeton University Press, http://dx.doi.org/10.1515/9781400850679.

[20] Manning, A. and B. Petrongolo (2017), “How Local Are Labor Markets? Evidence from a Spatial Job Search Model”, American Economic Review, Vol. 107/10, pp. 2877-2907, http://dx.doi.org/10.1257/aer.20131026.

[22] Marinescu, I. and H. Hovenkamp (2019), “Anticompetitive Mergers in Labor Markets”, Indiana Law Journal, Vol. 94, p. 1031.

[6] Marinescu, I., I. Ouss and L. Pape (2020), Wages, Hires, and Labor Market Concentration, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w28084.

[9] Martins, P. (2018), Making their own weather? Estimating employer labour-market power and its wage effects.

[30] Moretti, E. (2013), The New Geography of Jobs.

[21] Nimczik, J. (2020), Job Mobility Networks and Data-driven Labor Markets.

[51] Nordhaus, W. and A. Moffat (2017), A Survey of Global Impacts of Climate Change: Replication, Survey Methods, and a Statistical Analysis, National Bureau of Economic Research, Cambridge, MA, http://dx.doi.org/10.3386/w23646.

[39] OECD (2021), Inclusiveness during the Pandemic: Gender and Skills Differences in Exposure to Employment Effects, OECD Publishing, Paris.

[23] OECD (2020), Competition in Labour Markets, https://www.oecd.org/daf/competition/competition-in-labour-markets-2020.pdf.

[24] OECD (2019), Labour market regulation 4.0: Protecting workers in a changing, OECD Publishing, Paris, https://doi.org/10.1787/9ee00155-en.

[37] OECD (2019), Negotiating Our Way Up, OECD, http://dx.doi.org/10.1787/1fd2da34-en.

[38] OECD (2018), Good Jobs for All in a Changing World of Work: The OECD Jobs Strategy, OECD Publishing, Paris, https://doi.org/10.1787/9789264308817-en.

[4] Rinz, K. (2020), “Labor Market Concentration, Earnings, and Inequality”, Journal of Human Resources, pp. 0219-10025R1, http://dx.doi.org/10.3368/jhr.monopsony.0219-10025r1.

[44] Rossi-Hansberg, E., P. Sarte and N. Trachter (2021), “Diverging Trends in National and Local Concentration”, NBER Macroeconomics Annual, Vol. 35, pp. 115-150, http://dx.doi.org/10.1086/712317.

[11] Schubert, G., A. Stansbury and B. Taska (2020), “Monopsony and Outside Options”, SSRN Electronic Journal, http://dx.doi.org/10.2139/ssrn.3599454.

[3] Schwellnus, C. et al. (2018), “Labour share developments over the past two decades: The role of technological progress, globalisation and “winner-takes-most” dynamics”, OECD Economics Department Working Papers, No. 1503, OECD Publishing, Paris, https://dx.doi.org/10.1787/3eb9f9ed-en.

[13] Sokolova, A. and T. Sorensen (2020), “Monopsony in Labor Markets: A Meta-Analysis”, ILR Review, Vol. 74/1, pp. 27-55, http://dx.doi.org/10.1177/0019793920965562.

[35] Stanley, T. (2001), “Wheat From Chaff: Meta-Analysis As Quantitative Literature Review”, Journal of Economic Perspectives, Vol. 15/3, pp. 131-150, http://dx.doi.org/10.1257/jep.15.3.131.

[29] US Justice Department and the Federal Trade Commission (2010), Horizontal Merger Guidelines, https://www.justice.gov/atr/horizontal-merger-guidelines-08192010#5c.

[52] Yeh, C., C. Macaluso and B. Hershbein (2021), Monopsony in the U.S. Labor Market.

[TABLE CONTINUED ON NEXT PAGE]

The labour supply elasticity to the individual firm can be obtained empirically by estimating the elasticity of job separations to wages, where wages relate to the component of wages that is due to pay differences between firms for similar workers (Manning, 2011[14]). Following Bassier, Dube and Naidu (2021[15]), the labour supply elasticity is estimated in two stages. The first stage isolates the firm component of wages from other worker-related components by estimating a two-way fixed effects model based on Abowd, Kramarz and Magnolis (1999[41]):

where ${w}_{ijt}$ is the wage of individual i in firm j in year t; ${\phi }_{j}$ is a firm fixed effect; ${\mu }_{i}$ is a worker fixed effect; ${ϵ}_{ijt}$ is the error term; and ${x}_{it}^{\mathrm{\text{'}}}$ are time-varying worker control variables.32 Based on the results from the first stage, the second stage then estimates the elasticity of worker separations to the firm component of wages:

where ${s}_{ijt}$ is a dummy indicating separation of worker $i$ from firm $j$ in year $t$; ${\stackrel{^}{\phi }}_{j}$ is the estimated firm fixed effect; $\gamma$ is the elasticity of separations to wages; and ${\nu }_{ijt}$ is the error term.33 The fact that separations are estimated using only the component of wages that corresponds to firm pay premia (see Chapter 2) and not the component that corresponds to worker characteristics mitigates concerns of endogeneity of wages to the quit rate.34

On average across the covered countries, of which only Costa Rica is non-European, the estimated labour supply elasticity is around 2 (Annex Figure 4.B.1). This translates into a potential wage loss of about 30% compared to a worker’s market wage in the absence of wage-setting power.35 To some extent, the cross-country pattern of the estimated labour supply elasticity may reflect structural differences, e.g. related to cross-country differences in job mobility. But it may also be explained by differences in measurement error or the severity of endogeneity issues related to omitted factors that influence both wages and quit rates. For instance, higher-paying firms may be more likely to offer better non-wage working conditions (e.g. flexible hours, telework) that would have a direct effect on the quit rate. But firms may also pay higher wages to compensate workers for difficult or harsh working conditions. The severity, direction, and relative importance of such endogenous non-wage determinants of quits could additionally vary across countries, implying that care needs to be taken when interpreting the cross-country pattern in Annex Figure 4.B.1.

Overall, these results suggest a substantial degree of potential wage-setting power. However, they do not address the question of the extent to which firms actually exercise their power. This question is addressed using local labour market concentration as a partial indicator of wage-setting power.

Notes

← 1. This chapter has been written by an OECD team consisting of Michael Koelle, Nathalie Scholl and Cyrille Schwellnus with contributions of: Antoine Bertheau (University of Copenhagen, DENMARK), Chiara Criscuolo (OECD), Antton Haramboure (OECD), Alexander Hijzen (OECD), Balazs Murakőzy (University of Liverpool, HUNGARY), Satu Nurmi (Statistics Finland/VATT, FINLAND), Vladimir Peciar (Ministry of Finance of the Slovak Republic, SLOVAK REPUBLIC), Kevin Rinz (US Census Bureau, UNITED STATES), Catalina Sandoval and Jonathan Garita (Costa Rica Central Bank, COSTA RICA). Matej Bajgar, Chiara Criscuolo and Jonathan Timmis kindly provided the sales concentration data. For details on the data used in this chapter please see the standalone Data Annex and Disclaimer Annex.

← 2. The seven OECD countries that form the core of the analysis are Austria, Costa Rica, Denmark, Finland, France, Slovak Republic, and Spain. Comparable labour market concentration measures from the United States (based on establishment-level employment data) are additionally available for part of the descriptive analysis. Data for the Slovak Republic are available only for a short timespan (2014-2017), precluding any analysis which relies on the time series dimension of the data (including wage regressions).

← 3. Early studies on labour market concentration studied particular non-standard market niches, such as postings on online job boards (Azar, Marinescu and Steinbaum, 2019[42]; Azar et al., 2020[19])

← 4. A notable exception is ongoing work by Bassanini et al. (2022[25]) that analyses labour market concentration in a number of European countries.

← 5. Firms may also refrain from exercising wage-setting power because of costs related to setting optimal wages. Dube, Manning and Naidu (2018[43]), for instance, find strong evidence for bunching of wages at round numbers, suggesting the presence of optimisation costs.

← 6. Estimating the labour supply elasticity at the level of a narrowly defined local labour market is challenging due to the limited number of worker transitions observed in smaller partitions of the data. A sufficiently high number of separations – i.e., workers switching between different firms – is crucial for the precise estimation of firm pay premia and the elasticity of separations to cross-firm wage differences. Previous work relating wages to estimated labour supply elasticities did so either for larger labour markets (such as the entire US state of Oregon (Bassier, Dube and Naidu, 2021[15])) or for elasticities of online job applications to wages rather than actually observed separations (Azar, Marinescu and Steinbaum, 2019[42]).

← 7. However, trade costs may also imply that product markets, at least in some industries, are local (Rossi-Hansberg, Sarte and Trachter, 2021[44]).

← 8. A third approach to measure firms’ wage-setting power that is not further explored in this chapter relies on firm-level data and the estimation of firm-level production functions to infer the mark-down of wages below marginal costs (Yeh, Macaluso and Hershbein, 2021[52]). The drawback of this approach is that firm-level data generally do not allow to control for workforce composition and the inference of mark-ups relies on a set of theoretical assumptions.

← 9. The HHI consists in the sum of the squared market shares (in percent) of individual firms: $HHI=\sum _{i=1}^{k}{S}_{i}^{2}$.

← 10. When reporting the average of local labour market concentration at the national or industry level, each local labour market is weighted by its employment, such that national averages reflect the concentration faced by the average worker in the economy rather than concentration in the average local labour market.

← 11. The analysis is done at the preferred level of local labour markets, i.e. 3-digit industry by TL3-region, but is tested on a smaller subset of countries also for alternative labour market definitions for robustness.

← 12. Such sorting could arise either as an optimal worker response to wage penalties from concentration – with high-skill, high-wage workers more likely to overcome costs to mobility – or it could be driven by a third factor, such as the sorting of high-skilled workers to cities, where concentration is lower because of the higher density of markets. In some alternative specifications reported in Annex Table 4.A.1, observable worker characteristics (flexible gender-age interactions and a dummy for marginal workers) substitute for worker fixed effects.

← 13. This instrumental variable identifies the causal of effect of concentration under the assumption that changes in the average number of firms in other regions affect wages only through their effect on concentration, which may for instance be the case of changes in regulatory barries to entry.

← 14. Aggregation of estimates of single studies follows the methodology of “meta analysis” that is commonly used in economics (Stanley, 2001[35]; Nordhaus and Moffat, 2017[51]). It follows long-established statistical procedures that originate from applications in public health, medical science and adjacent fields (DerSimonian and Laird, 1986[50]). The aggregate coefficient is a weighted average of the individual country-level estimates, with weights taking into account both the estimation error within each country, and the between-country variation in estimates (so-called random effects meta analysis).

← 15. It is not possible to study labour market concentration from worker-level data, such as labour force surveys (LFS), due to lack of information that would allow grouping workers in the same firm. Firm-level data provide information on total employment at the firm level, which allows measuring firm-level employment concentration if sufficiently large and representative samples are available. But firm-level data lack information on individual workers, which precludes the measurement of concentration in hiring for different types of workers. Firm-level data also lack information on individual wages.

← 16. If a dataset does not provide information on worker location, establishment location is used instead.

← 17. Wages can be harmonised to the hourly level in about half of all countries where information on hours worked or equivalent (e.g. full-time equivalent rates) is available, which allows checking the robustness of the results obtained with monthly wages.

← 18. In many countries, the first year of observation is 2002, which implies that hiring concentration (which requires observing worker transitions between firms) is available from 2003.

← 19. The seven countries for which local labour market concentration is available at the 3-digit industry and TL3 region level are: Austria, Costa Rica, Denmark, Finland, France, Spain and Slovakia. Partial data based on national US classifications (4-digit NAICS industries and Commuting Zones) are available for the United States. Additional countries for which labour market concentration is available only at higher levels of aggregation (2-digit industry or TL2 region) are: Estonia, Hungary, Italy and Portugal.

← 20. The public sector (public administration and defence) is, by definition, not part of the market economy and not subject to market competition, and is therefore excluded from the analysis. The geographical distribution of agriculture and mining, and to some extent utilities, depends on natural geography, which large differences across countries and little relation to policy and economic structure.

← 21. Urban regions are equivalent to those classified as metropolitan regions in Fadic et al. (2019[28]). A metropolitan region is a TL3 region which contains a functional urban area – a single agglomeration or a group of agglomerations with strong cross-commuting patterns – of at least 250,000 people.

← 22. The cross-sectional differences in concentration levels may partly also be explained by the smaller size of the average manufacturing industry compared with the average services industry.

← 23. However, low-skilled workers do not systematically work in local labour markets where employer concentration is high Annex Figure 4.A.6.

← 24. Due to data confidentiality issues, local labour market concentration by skill group could not be obtained at the 3-digit industry by TL3 region level for a sufficient number of countries. The ratios of local labour market concentration by skill group relative to the mean reported in Figure 4.5 are therefore obtained at the 2-digit industry by TL3 region level, with the calculations to obtain skill group-specific HHIs assuming that the ratios are similar across different levels of industry aggregation.

← 25. By contrast, the evolution of concentration in rural and urban markets is very similar, suggesting no increasing divergence across geographical areas (Annex Figure 4.A.3).

← 26. A number of studies for the United States suggest that local sales and employment concentration decrease despite increasing national concentration (Rossi-Hansberg, Sarte and Trachter, 2021[44]). In this case, local sales and employment may be closely linked despite the apparent disconnect between national sales concentration and local labour market concentration.

← 27. Estimates using a measure of local labour market concentration based on 3-digit occupation by TL3 region yield similar results where this alternative measure of concentration is available. In France, Marinescu et al (2020[6]) find an effect of occupation-region concentration on hourly wages equivalent to a semi-elasticity of -5%; Martins (2018[9]) estimates a semi-elasticity of -1% on monthly wages in Portugal. This suggests that definitions based on 3-digit industry and 3-digit occupation by TL3 region provide similarly-performing approximations of local labour market concentration.

← 28. In the case of Costa Rica, the estimate is almost one order of magnitude larger (in absolute terms) than in other countries, but also comes with very large standard errors and a low Kleibergen-Papp (KP) first stage statistic of around 9. A similar combination of large standard errors and low KP statistic of around 3 is observed for Denmark. Since the critical value of the KP statistic is around 16, this indicates that the instrumental variable is weak in the case of Costa Rica and Denmark, which could bias the estimate in addition to rendering them imprecise (Bound, Jaeger and Baker, 1995[49]). The procedure to aggregate coefficients across countries preserves these country-level estimates instead of completely removing them, but assigns a very low weight to each of them (1-2%) to reflect the lower quality and precision. Alternatively, results obtained removing the coefficients from Costa Rica and Denmark are very similar to those reported in the main text.

← 29. For the purpose of this analysis, low and medium-skilled workers are grouped together since differences in the estimated wage effect of concentration between these groups are statistically insignificant.

← 30. In the limiting case of a monopsonist that chooses wages in order to maximise profits, both equilibrium wages and employment are below the social optimum (Manning, 2020[16]). Raising wages from the monopsonistic wage reduces the monopsonist’s profits (the mark-down relative marginal productivity) but raises employment by drawing workers unwilling to work at the monopsonist’s wage into the labour market.

← 31. Gig economy workers are found by Caldwell and Oehlsen (2018[10]) to have a fairly low labour supply elasticity, despite the absence of institutional constraints to hours worked.

← 32. The flexible controls consist in age group dummies, interacted with gender. Any time-constant worker characteristics are, by construction, controlled for through the worker fixed effects.

← 33. The estimated elasticity of separations to wages can be translated into the labour supply elasticity to the individual using the formula $LSE=-2\cdot \gamma /\left(1/n{\sum }_{i}{s}_{ijt}\right)$.

← 34. An endogenity probem could arise due omitted worker characteristics that determine both wages and quit rates. However, the two-way fixed effects model of Abowd, Kramarz and Magnolis (1999[41]) allows isolating the firm-level component of wages, which mitigates some of the endogeneity concerns. A different econometric concern concers measurement error in the estimated firm fixed effects, which could be addressed using a split-sample instrumental variables strategy (Bassier, Dube and Naidu, 2021[15]). However, unreported empirical analysis suggests that such estimates are very similar to the OLS estimates presented here.

← 35. In a simple monopsony model, the wage is marked down from the worker’s marginal product (MRPL) as a function of the labour supply elasticity ϵ: w= 1/(1+ϵ ) MPRL.