OECD Employment Outlook 2022

Building Back More Inclusive Labour Markets

Report

OECD Employment Outlook

9 September 2022

Available in:

English
français

Download PDF

3. Monopsony and concentration in the labour market

Satoshi Araki

Andrea Bassanini

Andrew Green

Luca Marcolin

Abstract

There is evidence that monopsony power is pervasive and substantial in OECD economies. Monopsony is the situation that arises where firms have the power to set wages unilaterally, leading to inefficiently low levels of employment and wages. This chapter reviews the causes, incidence, consequences and policy responses to labour market monopsony, focusing especially on labour market concentration, which is a key determinant of monopsony because in concentrated markets, few firms offer employment opportunities for workers. Using a harmonised dataset of online job vacancies, the chapter provides the largest cross-country comparison of the incidence of labour market concentration to date. It also presents original estimates of the consequences of labour market concentration on job quality, using employer-employee data. The chapter concludes by reviewing policy responses available to address monopsony and help labour markets function closer to the competitive ideal.

In Brief

Monopsony describes the situation in which employers possess unilateral wage‑setting power, and use it to set wages and employment below the levels that would prevail in a competitive market, where firms have to pay workers a “market rate” aligned with their productivity. Monopsony does not just imply lower wages for affected workers, but also a misallocation of resources: wages, employment and social welfare are lower when firms have monopsony power, compared to competitive labour markets.

This chapter explains why some firms have wage‑setting power, in particular in the case of labour market concentration where only a few employers compete in a market for workers. It provides novel statistics on the incidence of employment in concentrated labour markets, as well as the implications for job quality. Finally, the chapter discusses policies that directly reduce monopsony, improve the job quality of workers in uncompetitive labour markets and help labour markets function better for all workers.

Employers in monopsonistic labour markets are likely to depress employment and pay lower wages in order to reap higher profits. Labour market frictions that make it difficult to reallocate labour, employers offering unique sets of working conditions that tie workers to their workplace and highly concentrated markets (very few employers) are all reasons why firms may exercise monopsony power.

Key empirical findings include:

The empirical literature suggests that firm monopsony power is pervasive and substantial in OECD economies. One popular approach for measuring monopsony is to estimate the labour supply elasticity a given firm faces, namely the percentage reduction in the number of workers willing to work for the firm if it lowers the offered wage by 10% independently of other firms. Estimates of this elasticity found in the literature are often quite low. In other words, firm-level employment is far less responsive to wage changes than it would be if there were perfect competition. Even in markets where one would expect high competition – such as online labour markets – employer wage‑setting power is often substantial. However, while these estimates suggest the existence and pervasiveness of monopsony power, they do not identify the channels through which it affects the labour market.
This chapter finds that labour market concentration, one of the key determinants of monopsony power, is pervasive in a wide range of OECD countries. Using harmonised data of online job vacancies and the same labour market definitions across countries, this chapter finds that 16% of business-sector workers in 15 OECD countries are in labour markets that are at least moderately concentrated – according to the conservative definition frequently used by antitrust authorities – and 10% are found in highly concentrated markets. These figures can be considered a lower bound to the share of workers in concentrated markets.
Workers are not evenly distributed across concentrated markets. Workers in rural labour markets are more likely to be in concentrated labour markets, for example.
Workers who have been on the front line during the COVID‑19 crisis – those with substantial contact with colleagues or customers and thus with a higher than average risk of infection (see Chapter 1) – are more likely to work in concentrated labour markets. By contrast, workers in occupations amenable to telework tend to be found in much less concentrated markets.
A year into the COVID‑19 pandemic, labour market concentration was 10% higher, on average, in OECD labour markets. The rise was sharp at first, and probably driven by a significant drop in job openings at most firms during the first lockdown, with the remaining vacancies being posted by a few resilient employers. Since then, concentration has begun to fall back towards pre‑pandemic levels, in line with a progressive normalisation of hiring by firms.
Available evidence suggests that monopsonistic labour markets tend to be associated with lower employment, but more research is needed. Studies based on mergers tend to find that employment falls after a merger. The few studies using measures of labour market concentration also find that mergers reduce overall employment in the affected labour market. However, quantitative estimates remain very heterogeneous between studies, and there are unresolved methodological issues in the literature.
New evidence provided in this chapter relying on harmonised linked employer-employee data for a number of European OECD countries, confirms results from the literature, showing in particular that more concentrated markets result in lower wages. Estimated elasticities of wages to concentration are similar in Denmark, France, Germany and Portugal. A 10% increase in concentration from the average level is estimated to reduce daily wages of full-time workers by 0.2% to 0.3%. These estimates imply that the 10% of workers who are employed in the most concentrated labour markets experience a wage penalty of at least 5% compared with a worker in a labour market with the median level of concentration.
Other aspects of job quality are also affected. Regression analysis suggests that labour market concentration tends to increase the use of flexible contracts. In France, Germany and Portugal increasing concentration is estimated to reduce the probability of being offered an open-ended contract at hiring. The effect of a 10% increase in concentration can be up to 2.3% (for both Portuguese men and women).
Evidence from Spain and Italy suggests that, for those hired on a temporary contract, labour market concentration clearly depresses their chances of accessing a more stable position within the calendar year after hiring. The effect appears particularly large for Italy, where a 10% increase in concentration reduces the conversion rate by 2.5% for both men and women.
Another dimension affected by monopsony power is skill requirements. Monopsonistic employers tend to curb their labour demand, allowing them to be more selective when hiring. Regression analysis on online job postings finds that labour market concentration often increases skill requirements in posted job vacancies, both in the number of skills that are required, and the frequency with which cognitive and social skills are expected.

Policy can help make labour markets more competitive:

Existing evidence from the United States and Austria suggests that facilitating the enforceability of Non‑Compete Agreements (NCAs) unambiguously reduces job mobility and often depresses wages. NCAs are clauses in contracts that prevent workers from working to a competitor after they separate from their employer. There is evidence that employers frequently use NCAs to limit the outside options of their workers, even when they have no access to the employers’ confidential information or other intangible assets. The chapter discusses several options governments could consider to limit the spread of NCAs.
Other areas of regulatory and enforcement interventions concern occupational licensing, labour market collusion and horizontal mergers. In all these areas, regulators should devote more attention to the consequences of employers’ actions for the competitiveness of the labour market. In many cases, interventions could be undertaken by antitrust authorities, as well as labour authorities.
Interventions to promote collective bargaining could have a strong impact on monopsony power. Under collective bargaining, if workers have sufficient countervailing power, the parties may internalise the position of the firm in the product market so that negotiations may lead to a more efficient labour market outcome, with the greater rents generated in this way shared among the parties. In fact, the negative impact of concentration on wages has been found to be smaller where trade unions are stronger.
Minimum wages can also be used to curb the negative effects of monopsony and concentration. Under monopsony, minimum wages, if set at a reasonable level, lower the marginal cost of hiring at the lower end of the wage distribution. Therefore, minimum wages can raise both employment and wages in monopsonistic labour markets. Consistently, available evidence shows that existing minimum wages have minimal disemployment effects in concentrated markets.
Policies to promote telework may help workers in concentrated markets. For workers in occupations suitable for teleworking, increased telework may allow them to accept positions in a wider geographic radius, increasing the set of employers who can bid for their labour. A simulation performed for this chapter suggests that opening jobs to full-time telework could decrease labour market concentration by about 20%, on average.
A simulation performed in the chapter suggests that aggregate labour market concentration would decrease by 18% on average across the OECD countries considered if workers could retrain and seek employment in alternative occupations. Reskilling and training policies therefore can play a role in improving labour market conditions when markets are monopsonistic.

Introduction

If your employer threatens to lower your compensation, would you be able to quit, and quickly find a new job elsewhere with similar working conditions? For many workers, in the absence of policy intervention or some form of collective action, the ability to credibly quit for a higher-paying job is the main bargaining power they have. Moving from one employer to another is one of the strongest sources of wage growth because it allows workers to move up the “job ladder” to higher-paying firms (Topel and Ward, 1992[1]; Haltiwanger, Hyatt and McEntarfer, 2018[2]; Wang, 2021[3]) – see also Chapter 4. The ability to shop around easily for a new employer is a core mechanism for ensuring pay keeps pace with productivity, and it is one of the foundations of competitive labour markets.

When a worker is confronted with a labour market with many similar workers (sellers of labour) but only one or a few employers (buyers of labour), their bargaining power is greatly diminished, and it may be difficult to find a new employer. A classic and extreme example is a coal company that employs miners and also owns the only mine within a reasonable commuting distance. In such “company towns” there is really only one employer, so workers seeking a different employer would need to move to a different town often at considerable expense. The firm knows this, and unless there are some countervailing forces, it leaves workers at a disadvantage when bargaining for wages or working conditions. In labour markets, a worker’s compensation is not solely determined by their skills or productivity, but what they have the power to negotiate.

Monopsony is the situation that arises when competitive markets break down and workers cannot easily find enough suitable employment offers. The term encapsulates the situation of markets where few employers exist – labour market concentration. However, monopsony is more general, and arises even in markets with many employers. For example, a single mother whose employer provides subsidised childcare and flexible working hours may find many potential employers, but few offering the same set of working conditions tailored to her personal situation. Alternatively, a low-wage worker who has multiple jobs to make ends meet may simply not have the time to search for jobs effectively and attend interviews with prospective employers. In both of these situations, workers cannot profit from a market of many competing employers to bid wages up to their level of productivity, and instead must negotiate with a limited set of employers who therefore retain some unilateral wage‑setting power.

In line with the literature, this chapter defines monopsony as the situation that arises when firms retain discretion in setting wages and working conditions as opposed to the case of competitive markets where firms must pay workers the “market rate”, which aligns with their productivity.1 Employers in monopsonistic labour markets may use their bargaining power to inefficiently lower wages and depress employment in order to reap larger rents. This not only affects the distribution of rents between workers and firms, but the economy-wide allocation of resources. Monopsonistic labour markets should lead to lower employment and output than what would prevail if labour markets were perfectly competitive.

Policy, however, can directly address the misallocation of resources wrought by monopsony through direct interventions to realign bargaining power (regulation, antitrust policy, the role of social partners), as well as other, more indirect, policy tools (such as minimum wages). In addition, uncompetitive markets themselves may have important implications for other, only tangentially related, labour market policies such as employer-sponsored training. The resulting misallocation of resources from monopsony also justifies policy interventions regardless of labour market conditions: labour market tightness may result in better salaries and working conditions – although there is little evidence of this in the current recovery (see Chapter 1) – but it is unlikely to restore the outcomes of a competitive labour market.

Although research on monopsony goes back decades, the assumption that labour markets are competitive has persisted in most policy circles. The recent availability of high-quality data covering the near-universe of workers, firms, online job vacancies or mergers is forcing a re‑examination of this assumption. Researchers are now better able to compute the concentration of firms in a well-defined labour market directly. Concentration is a key source of monopsony power because in concentrated markets, workers’ outside options are limited. Concentration indexes are used by antitrust authorities as a rough proxy for market power to identify markets where action may be required – see e.g. US Department of Justice and Federal Trade Commission (2010[4]). However, much of the current research on labour market concentration looks at the United States, and focuses on the effect of concentration on employment, wages or earnings while neglecting other job characteristics such as job insecurity, opportunities for promotion and progression, commuting distance and training. The cross-country studies that do exist, furthermore, have light country coverage and often use data which make cross-country comparisons challenging.

This chapter fills the gaps in this evolving literature by providing a cross-country evaluation of labour market concentration with an emphasis on policy. The first question of interest is the proportion of a country’s workers who are employed in concentrated labour markets. Using harmonised data on both online vacancies and matched employee‑employer data, the chapter provides the largest cross-country comparison to date of the share of workers facing concentrated labour markets. In addition to national averages, the chapter shows how certain vulnerable groups such as front-line workers may be disproportionately working in concentrated markets, as well as how market concentration has evolved over the COVID‑19 pandemic. The chapter then analyses the effect of concentrated markets on various aspects of labour market performance including employment, earnings, job security and skill demand.

Finally, the chapter reviews the current literature around policy responses to labour market monopsony and concentration. The discussion touches on policies that have a direct impact on the relative power of workers and employers, as well as on policies that can be mobilised to counteract the negative effects of power imbalances on labour market outcomes.

The chapter begins with the definition and examples of monopsony, before proceeding to considerations of measurement, economic consequences and policy responses. Section 3.1 defines monopsony first broadly as the likelihood of a worker quitting when faced with a reduction in wages – monopsonistic competition (Manning, 2003[5]) – but then more specifically for the case of concentrated labour markets characterised by a limited number of employers for many workers. Section 3.2 then presents cross-country estimates of concentration including a focus on key occupations and demographic groups, and an examination of how concentration has evolved over the COVID‑19 pandemic. Section 3.3 shows the effects of concentrated labour markets on employment, wages, job security and skill demand.2 Section 3.4 then reviews some direct policy responses as well as other policies which may have indirect consequences for monopsony in the labour market. Section 3.5 offers concluding remarks and identifies avenues for further policy research.

3.1. The general case of monopsony

In monopsonistic labour markets, employers depress labour demand in order to reduce labour costs and reap higher profits from paying works less than their marginal productivity – see e.g. Boal and Ransom (1997[6]), Manning (2003[5]), Ashenfelter, Farber and Ransom (2010[7]) and Blair and Harrison (2010[8]). In other words, employment and wages are set at a lower level than what would be achieved in competitive markets, where employers must pay workers the market rate which aligns with their productivity. The misallocation induced by unilateral employer power suggests a role for governments to intervene and limit the scope of monopsonistic labour markets (see Section 3.4).

Monopsony encompasses the case of firms that have large discretion in setting wages (and, by extension, employment). Technically, the term characterises markets with one buyer of labour (employers) but many sellers (workers). However, at least as used in labour economics, the term monopsony encompasses a more general definition, where employers have wage‑setting power over workers and labour markets therefore deviate from the competitive ideal. In competitive labour markets, firms take wages as given by the market, and if any firm attempted to offer wages lower than the market-determined price, all of their workers would quit and/or hiring would be rendered impossible. In practice, labour markets exist on a spectrum between purely competitive (firms take wages as given), and completely monopsonistic (firms have complete wage‑setting power). Research on labour market monopsony concerns itself with theorising why firms may exercise unilateral wage‑setting power, and measuring its extent.

There are three broad reasons why firms may have unilateral wage‑setting power in the labour market. First, there may simply be too few firms relative to available workers. In a simple model in which few employers compete in a market with each other, firms employ fewer workers than in the competitive equilibrium and offer a lower wage (Boal and Ransom, 1997[6]). This is analogous to product markets where there are few buyers and many sellers. Labour markets of this type are often concentrated – they have too few employers. The measurement of labour market concentration constitutes one avenue of research on monopsony.

While this model is widely used in industrial organisation and retains salience for empirical work in labour economics, it does not take into account the specific characteristics of the labour market. Employers can have monopsony power even if markets are not concentrated, e.g. because of clauses in labour contracts which limit workers’ ability to look for alternative jobs (such as non-compete agreements, see Section 3.4.1).3 Alternatively, workers may have preferences for specific job attributes provided by the employer (see below).

For these reasons, another strain of thought, referred to as “Dynamic Monopsony” or “Modern Monopsony”, posits that workers must search for suitable employment opportunities. In these models, workers cannot immediately quit an employer and instantaneously find a new one, or instantaneously find a new job if unemployed. These search frictions imply that workers must wait for a suitable job offer, which provides firms with some wage‑setting power (Burdett and Mortensen, 1998[9]; Manning, 2003[5]). In addition to such “natural” search frictions, firms may actively introduce additional frictions in their labour market (e.g. through collusion among employers and non-compete agreements), thereby increasing their monopsony power vis-à-vis their workers (see Section 3.4.1).4

A third explanation derives monopsony power from workers’ preferences for firms besides the offered wage. For example, if employers offer different health insurance plans, or access to childcare, which vary in their generosity, workers may prefer certain firms even if they offer identical wages. Such preferences also extend to amenities such as “company culture”, or employer attributes like commuting distance (Card et al., 2018[10]). With such differentiated employers, workers may find it difficult to quit and find a suitably similar firm. Regardless of why firms offer different amenities, models that rely on workers’ preferences for differentiated firms result in wage‑setting power for firms.5

In all of the explanations for monopsony, firms obtain the ability to decrease wages while retaining most of their workers and their ability to hire. This relationship, the change in available workers for a given change in the wage offered, is called the elasticity of labour supply to a firm. Research on these elasticities represents one classical way to measure firm monopsony power. When this elasticity is low – changes in offered wages result in small changes in hires, quits or employment – this is evidence of some degree of monopsony power.

3.1.1. Firms retain ample wage‑setting power, even in what should be competitive markets

What is becoming increasingly evident from the literature is that monopsony is far more prevalent in labour markets than previously expected. In the language of labour economics, estimates of the labour supply elasticity facing the firm are small. In the ideal case of perfect competition own-firm labour supply elasticities would be infinite. Normally, single‑digit estimated elasticities, or lower, are considered to be evidence of monopsony power – see e.g. Manning (2003[5]).

In one of the largest reviews of the literature so far, the consensus estimates of the own-firm labour supply elasticity are in the single digits. Sokolova and Sorensen (2021[11]) examine 1 320 recent estimates of labour supply elasticities reported in 53 studies. They report own-firm labour supply elasticities around 3 for women and 4.2 for men on average among the most rigorous estimates. This corresponds to a 22% wage markdown from the worker’s marginal productivity, on average.

Reported own-firm elasticities tend to be lower in Australia, Canada and the United States than in Europe (Sokolova and Sorensen, 2021[11]). An OECD analysis of linked employer-employee data of 10 OECD countries6 finds the weighted average of the country-level estimates to be 2 (OECD, 2021[12]), which again, implies pervasive monopsony in the labour markets of these countries. Finally, Webber (2016[13]) finds significant variation in the firm’s wage‑setting power across the wage distribution, with the elasticity in the lowest quartile being only 0.22 (against an average estimate of 1.08).

Firms may hold unilateral wage‑setting power even when search frictions are considered to be a priori minimal, and the labour market in question appears to be perfectly competitive. Dube et al. (2020[14]) conclude that the elasticity of labour supply facing the requester on Amazon Mechanical Turk (“MTurk”) – a prominent online job market matching task requesters and workers – amounts to 0.14, suggesting substantial market power of requesters (firms) despite the apparent absence of search frictions – see also OECD (2019[15]), which dealt extensively with issues on monopsony for the own-account self-employed. Similarly, Caldwell and Oehlsen (2018[16]) run a field experiment in which they randomly assign higher wage rates to Uber drivers in the United States for one week. They estimate labour supply elasticities below 1 for both those who can work for a rival platform and those who cannot. Overall, there is growing evidence that monopsony power is pervasive, even in what one would assume to be the most competitive labour markets.

3.1.2. Monopsony may fall disproportionately on certain groups

Monopsony power may affect women more than men, on average. The estimates of labour supply elasticities generally find a lower elasticity for women, and the wage markdown to a worker’s marginal productivity is around 6 percentage points higher for women than men (Sokolova and Sorensen, 2021[11]). There are plausible reasons for that. For example, there is evidence that women have different and more marked preferences for certain job amenities, especially in the case of mothers with young children, which reduces their bargaining power (Mas and Pallais, 2017[17]; Wiswall and Zafar, 2017[18]). In addition, women tend to search for jobs closer to their home, and they are ready to accept a significant wage penalty for a closer job, which exposes them to greater monopsony power (Le Barbanchon, Rathelot and Roulet, 2020[19]; Jacob et al., 2019[20]). Lastly, women’s caregiving responsibilities are also linked to their occupation choices. Women may choose occupations with less working hours and more flexibility (Goldin, 2014[21]), which may lead them into more concentrated labour markets.

More generally, one could expect that historically disadvantaged groups (such as youth, migrants and ethnic/racial minorities) are more exposed to monopsony power than insiders. Monopsony models predict that employment should be below what would prevail in a competitive market. Firms may therefore have their choice of workers and they may have discretion on whom they choose to hire for the jobs they make available. This could mean that they may prefer to employ workers with more labour market experience which would disadvantage youth (see Section 3.3.2). They can also choose to pay workers with comparable productivity but worse prospects for employer-to‑employer job mobility less than others, as shown for non-white workers in Brazil by Gerard et al. (2021[22]). In addition, in models of dynamic monopsony where even a small fraction of firms may discriminate against certain groups, workers in these groups are penalised with larger wage markdowns even in non-discriminating firms (Lang and Lehmann, 2012[23]; Cahuc, Carcillo and Zylberberg, 2014[24]). In most models of discrimination (assuming competitive or monopsonistic markets), firm entry should drive discriminating firms out of the market. However, concentrated labour markets likely have some barriers to firm entry, and one should expect they therefore contain a lower share of disadvantaged groups.

3.2. How concentrated are OECD labour markets?

As mentioned in the previous sub-section, labour market concentration – i.e. the situation wherein labour markets are dominated by a few firms – is expected to result in monopsony power for these firms. When few firms dominate a given labour market, they may be able to affect wages through their own labour demand. It also means that workers are less likely to find similar suitable employers, or are more likely to meet the same firms while searching for suitable jobs (Manning, 2020[25]). Lastly, fewer employers are more likely to implicitly (or explicitly) co‑ordinate their wage setting – see Section 3.4.1. To the extent that the variety of suitable job offers depends on the number and relative size of the firms in a market, the elasticity of own-firm labour supply can be seen as a decreasing function of labour market concentration (Jarosch, Nimczik and Sorkin, 2019[26]).

In short, labour market concentration is likely one of the major sources of monopsony, and it therefore makes for an imperfect, easy-to-measure, empirical proxy for employer wage‑setting power.7 Namely, there should be a positive correlation between labour market concentration and employer wage‑setting power across markets (Jarosch, Nimczik and Sorkin, 2019[26]; Azar, Marinescu and Steinbaum, 2019[27]; Boal and Ransom, 1997[6]).

For this reason, among others, the use of concentration as an empirical measure of monopsony has exploded. In just the last few years, studies using labour market concentration to measure monopsony have appeared, using data from the United States (Azar et al., 2020[28]; Benmelech, Bergman and Kim, 2022[29]; Yeh, Hershbein and Macaluso, forthcoming[30]; Qiu and Sojourner, 2019[31]; Rinz, 2022[32]), the United Kingdom (Abel, Tenreyro and Thwaites, 2018[33]), France (Marinescu, Ouss and Pape, 2021[34]), Austria (Jarosch, Nimczik and Sorkin, 2019[26]), Portugal (Martins, 2018[35]), Norway (Dodini et al., 2020[36]), and more recently, cross-country studies (OECD, 2021[12]; Bassanini et al., 2022[37]) for a limited number of countries.8

One open question is whether the results of these studies reflect differences in data or methodology, or if they reflect real cross-country differences in the competitiveness of labour markets. This chapter builds on this previous work by presenting the largest cross-country coverage of labour market concentration in OECD countries with the greatest uniformity in the definition of a labour market.

Using data on the universe of online job vacancies, this section reports estimates of the share of workers in concentrated labour markets for 15 OECD countries, as well as for Singapore.9 This is the largest cross-country study of labour market concentration to date, and it is the only cross-country study to use a large harmonised dataset and labour market definition for cross-country comparability. In addition to country-level averages, the section shows how concentration impacts certain segments of the labour market including specific occupations, gender and youth, among others. The section concludes by analysing concentration dynamics over the COVID‑19 pandemic.

3.2.1. A sizeable share of workers are in concentrated labour markets

Whether a labour market is concentrated depends on how one defines the local labour market where a potential worker can reasonably expect to quickly find a suitable job. The literature typically defines labour markets with the combination of detailed economic classes (industry or occupation), and geography. In theory, the local labour market is an area that captures all employers to which a potential worker could reasonably commute. Some studies of labour market concentration use commuting zones or functional urban areas which are often designed empirically to capture observed home‑to-work flows (Foote, Kutzbach and Vilhuber, 2021[38]). Due to data limitations, this chapter uses Territorial Level 3 (TL3) regions, which are a higher level of geographic aggregation than commuting zones (see Box 3.1). Designed by the OECD, TL3 regions cover every OECD country, are generally stable over time, and are designed to be roughly comparable across OECD countries (OECD, 2016[39]).

In addition to TL3 regions, this chapter defines the relevant labour market using occupations instead of industries. Industries are designed based on the economic activity carried out in an establishment. Occupations are classified based on the skills and qualifications required of the worker, and are therefore portable across industries in most cases. Occupations are thus more suitable to define workers’ job search patterns, and to measure labour market concentration as a consequence. Using occupations is also consistent with evidence presented in certain famous cases of unlawful no-poaching agreements in the United States in the mid‑2000s (Koh, 2013[40]), which show that companies can produce different products while competing for the same workers. Hovenkamp and Marinescu (2019[41]) provide further examples. For Continental European countries, this chapter uses 4‑digit ISCO‑08, and 6‑digit SOC‑2010 for Anglophone countries.10

The standard measure of concentration in the labour market is the Herfindahl-Hirschman Index (HHI) of either vacancies, new hires or employment in a local labour market. This is defined as the sum of the squared percentage shares of each firm in the market. The index ranges from 0, no market concentration, to 10 000, the case of a single firm controlling the entire market.11 Markets are considered concentrated according to the threshold for action used by antitrust authorities for product market concentration, which are typically very conservative (Nocke and Whinston, 2022[42]; Affeldt et al., 2021[43]). According to US antitrust authorities, high concentration markets display an HHI of 2 500 and above, and moderately concentrated markets an HHI of 1 500 to 2 500 – see e.g. US Department of Justice and Federal Trade Commission (2010[4]).12 These can be considered to yield a lower bound to the share of workers in concentrated markets.

This chapter uses data on online job postings from Emsi Burning Glass (EBG) to measure labour market concentration. EBG collects online job postings in many OECD countries, which contain information on the posting’s occupation, geography and firm (including industry), in addition to other characteristics such as skills and educational requirements. The data have been shown to have an almost full coverage of vacancies, and is increasingly representative of overall employment in the United States (Hershbein and Kahn, 2018[44]; Azar et al., 2020[28]). This chapter then validated the data coverage on the remaining OECD data for which EBG data are available and Singapore. Fifteen OECD countries and Singapore were assessed to have suitable coverage for inclusion in the chapter.13 With the exception of the analysis of concentration dynamics during the pandemic, the analysis in this section uses data from 2019.

After calculating HHI at the occupation by TL3 level, the cells were aggregated to the ISCO‑3 level using job posting weights and then they were weighted to employment using the occupation distribution in the business sector (omitting industries where public employment is sizeable)14 in each country available from labour force surveys (see Annex 3.B for a full description of data validation, construction and analysis). The final country-level estimates are adjusted to account for heterogeneity in the average population size of TL3 regions across countries (see Box 3.1).

This chapter finds a sizeable share of workers in OECD countries work in markets that are moderately to highly concentrated. Figure 3.1 shows the share of workers in moderately concentrated labour markets (light blue) and the subset of those who are in highly concentrated labour markets (dark blue), as derived from estimates of HHI at the national level (Annex Figure 3.A.1). Just over 16% of workers find themselves in labour markets that are at least moderately concentrated, on average across OECD countries in the sample. Of those, more than half, or about 10% of the total, work in highly concentrated labour markets. The highest shares of workers in markets that are at least moderately concentrated are found in Estonia and Latvia with shares above 24%, while the smallest shares are found in Belgium and Switzerland with shares under 10%. The results in this section confirm that cross-country differences in labour market concentration are not simply due to differences in data or labour market definitions.15

Over a longer time series, concentration tends to be stable across OECD countries. The Emsi Burning Glass job posting data used in this chapter do not allow for the comparison of HHI over a long time period. However, using administrative data on new hires, OECD (2021[45]) finds that HHI is relatively stable from 2003 to 2017 in an average of 7 OECD countries.16 There is likely variation across countries in this trend, however. For example, Rinz (2022[32]) finds a modest decrease in local labour market concentration in the United States from around 2000 to 2009, and then a modest increase during the financial crisis.

These results are relevant because, all else equal, workers in these markets are likely paid wages below what their productivity would suggest in a competitive market. A similar argument can be made for other measures of job quality (see Section 3.3.2). While this applies to all workers whether they find themselves in concentrated markets or not, one should be especially careful for workers in markets which are moderately or highly concentrated.

In addition, one needs to place these estimates in their proper context. First, the thresholds used in this analysis to determine whether a market is concentrated are high (see the discussion of HHI above), and these estimates can therefore be viewed as a lower bound of workers in concentrated markets. Second, these estimates are of labour market concentration, and they therefore only represent one source of monopsony power. Even in markets that do not meet the thresholds for concentration, workers may still be subject to other sources of monopsony power (see Section 3.1). Finally, this chapter does not analyse the causes of the reported cross-country differences in labour market concentration. Countries differ, for example, in the composition of the labour market in terms of occupations, sectors, and commuting patterns which can directly affect concentration (see below). A structured analysis of the determinants of labour market concentration is left to future analyses.

Figure 3.1. The share of business sector employment in moderately to highly concentrated labour markets, 2019

Box 3.1. Defining the relevant labour market to measure labour market concentration

The demarcation of local labour markets to identify monopsony power is challenging, especially in a cross-country context, and a consensus on methodology is yet to be reached (Azar et al., 2020[28]; Manning, 2020[25]; Naidu, Posner and Weyl, 2018[46]; Hovenkamp and Marinescu, 2019[41]). Too narrow a market restricts the set of workers’ outside options and inflates firms’ wage‑setting power, while the opposite holds true for too large a market. The definition of a local labour market involves a labour market statistics interacted with the combination of geographical units and economic units (occupations or industries).

Frequently used geographic units are commuting zones – e.g. Azar et al. (2020[28]), Marinescu, Ouss and Pape (2021[34]), Benmelech et al. (2022[29]), Berger et al. (2019[47]), Rinz (2022[32]) – or administrative units – e.g. Modestino, Shoag and Ballance (2016[48]). While administrative units may not fully capture travel-to-work flows in an area, definitions of commuting zones are not necessarily comparable across countries. EU-OECD functional urban areas (FUA) are defined using the same methodology for all countries as urban centres and catchment areas thereof (Dijkstra, Poelman and Veneri, 2019[49]). As such, FUAs leave out rural areas. Ascheri et al. (2021[50]) use FUAs, but their analysis is therefore limited to urban areas.

In light of these considerations and the availability of information in the EBG dataset, HHIs in this chapter are calculated based on TL3 regions, unless otherwise specified. TL3s correspond to sub-national administrative units1 that are roughly comparable across countries (OECD, 2021[51]), even though their size and number can differ across countries. In order to improve their comparability further, however, an adjustment factor is obtained by regressing aggregate concentration statistics on the logarithm of the country-specific population average of TL3 regions. This adjustment factor is then applied to each statistic in order to obtain figures for an average regional population of 200 000 people, which roughly corresponds to commuting zones in the United States, and allows therefore an easy comparison with figures obtained by Azar et al. (2020[28]) – see also Annex 3.B.

As far as the economic unit is concerned, Berger et al. (2019[47]), Benmelech et al. (2022[29]), Rinz (2022[32]) and OECD (2021[45]) calculate HHIs by industry, whereas Azar et al. (2020[28]), Martins (2018[35]), Marinescu, Ouss and Pape (2021[34]), and Azar, Marinescu and Steinbaum (2022[52]) do so by occupation.2 This chapter calculates HHIs by occupation3 for two reasons. First, empirical evidence shows occupation switches imply a wage penalty even controlling for employer and industry switches – see Kambourov and Manovskii (2009[53]), Gathmann and Schonberg (2010[54]) – they cause losses in occupation-specific human capital. Second, the use of industries is likely to conflate product market and labour market concentrations, even though one can exist without the other – see Manning (2020[25]), Hovenkamp and Marinescu (2019[41]), Redding and Rossi-Hansberg (2017[55]).4 In fact, there is evidence that firms operating in different industries can still collude to control the labour market of the same occupation (Hovenkamp and Marinescu, 2019[41]; Gibson, 2021[56]).

Two additional elements need to be chosen to compute the HHI. The variable over which firm shares are computed (usually employment, hires or vacancies), and the relevant time period. Due to data availability, the analysis in this chapter is based on quarterly vacancies, except in Section 3.3.2. An HHI based on employment seems to be a reasonable measure of concentration both in a classical, static model of monopsony and in a stationary search and matching model with granular search, where concentration affects workers’ outside options (Boal and Ransom, 1997[6]; Jarosch, Nimczik and Sorkin, 2019[26]). However, in a non-stationary environment, downsizing firms may have a positive share of employment without hiring so that they do not effectively contribute to the number of outside options in the labour market. In this case, a measure based on job vacancies or new hires better captures the fact that labour market concentration is a key determinant of monopsony power (Marinescu, Ouss and Pape, 2021[34]; Bassanini, Batut and Caroli, 2021[57]; Azar, Marinescu and Steinbaum, 2022[52]).

Finally, this chapter computes HHIs on a quarterly basis. Many papers compute flows over annual intervals due to data availability. However, Azar et al. (2020[28]) compute HHI quarterly, arguing that that an annual interval is manifestly too long to capture outside options. This chapter follows on that lead and computes HHIs on a quarterly basis.

1. For Australia, Canada and the United States, TL3 corresponds to groups of sub-national administrative units. For Luxembourg, there is only one TL3 region assigned to the whole country. One TL3 region is assigned to the whole of Singapore for the current analysis.

2. Other dimensions are sometimes explored in some studies – see e.g. Azar et al. (2020[28]) Dodini et al. (2020[36]).

3. Four-digit ISCO‑2008 is used for European countries (excluding the United Kingdom) and 6‑digit US SOC‑2010 is used for Australia, Canada, Singapore, the United Kingdom and the United States.

4. For example, evidence exists that product market concentration has a negative impact on productivity. Neglecting to take this into account when estimating the impact of labour market concentration on wages may underestimate the effect of concentration on wages.

Some blue‑collar and health-related occupations are in more concentrated labour markets

A few blue‑collar occupations and health-related labour markets tend to be more concentrated. Figure 3.2 depicts the average share in concentrated markets by 2‑digit ISCO occupation.17 The occupations which are the most concentrated, on average, are handicraft and printing workers, and health professionals, where over 50% of business-sector employment in these occupations is found in concentrated markets.18 In addition to those two occupations, the top five most concentrated occupations include other blue‑collar occupations – such as agricultural, forestry and fishing labourers and refuse workers.

The least concentrated occupations are information and communication technology professionals, sales workers and business administration professionals where less than 7% of workers in these occupations are found in concentrated markets. The least concentrated occupations are not confined to high-skill, high-wage professionals. General cleaners and helpers and sales workers are also present in the least concentrated occupations, likely because workers in these occupations are typically employed in numerous small establishments and shops. In short, occupations in the least concentrated markets appear to be employable in a wide variety of industries, which would grant them more employment options.

The analysis in this chapter also finds that workers in middle‑skill occupations are the most likely to be in concentrated labour markets. Low-skill workers face the lowest concentration and high-skill workers the next highest after middle‑skill workers. This pattern is not particularly robust across countries, however (Annex Figure 3.A.2). The declining employment share of middle‑skill jobs, and the rise in job polarisation and deindustrialisation is a well-documented fact across many OECD countries (OECD, 2017[58]; OECD, 2020[59]). As the employment shares of middle‑skill jobs shrink, the remaining workers may face a smaller and smaller pool of potential employers who continue to use the production technologies to employ them.

Concentration is lower in urban areas

In addition to occupation, the other key dimension of a labour market is geography. Larger labour markets, in particular cities, have long been hypothesised (with increasing empirical evidence) to allow more efficient matches between firms and workers (Petrongolo and Pissarides, 2006[60]; Andersson, Burgess and Lane, 2007[61]; Bleakley and Lin, 2012[62]; Dauth et al., 2018[63]). A worker searching for job is more likely to find a suitable employer when there are many potential employers, and vice versa. Labour markets are more efficient when they are thick. The same logic applies to market concentration as measured by HHI: workers should find it easier to quit and find a new employer when there are more potential employers.

Urban areas are less concentrated than rural geographies in all countries for which data are available. Figure 3.3 uses the OECD definition for metropolitan regions, which includes TL3 regions that have more than 50% of their population living in a functional urban area of over 250 000 people (Fadic et al., 2019[64]). On average across OECD countries in the sample, rural regions (29%) have about two and half times more people working in moderately concentrated markets than urbanised regions (11%). The largest differences are in Canada and Australia, two countries with large urban centres but also geographically large, but sparsely populated provinces including remote areas.

Figure 3.2. The occupations facing the most and least concentrated labour markets, 2019

The finding confirms results from the literature that rural labour markets are more concentrated. Azar et al. (2020[28]) and Bassanini, Batut and Caroli (2021[57]) find a decrease in HHI as the size of commuting zones increases in the United States and France, respectively. Using the same urban-rural definition as this chapter (but different data and definition of labour market), OECD (2021[45]) similarly finds a large urban-rural difference in the share of workers in concentrated labour markets.

Figure 3.3. Rural regions are more concentrated than urban ones

3.2.2. Concentration does not vary substantially by demographic groups

This chapter finds little evidence that women are more likely than men to work in concentrated labour markets. Figure 3.4 (Panel A) shows the share of men and women in moderately to highly concentrated labour markets. On average, 16.6% of women are in labour markets that are at least moderately concentrated, compared to 17.2% of men. Estonia, Latvia and Sweden have the highest shares of women in markets that are at least moderately concentrated, each with shares over 20%. However, in those countries, the share of men in concentrated markets is also high and even exceeding the share of women.

Just as with women, there is little difference in the share of youth in moderately to highly concentrated labour markets compared to other adults. The share of youth and other adults in labour markets that are at least moderately concentrated is around 17% for both, on average (Figure 3.4, Panel B). The highest shares of youth in concentrated labour markets are also in Latvia, Estonia and Sweden.

Figure 3.4. The share of employment in moderately to highly concentrated labour markets in the business sector by selected demographic characteristics, 2019

While the share of workers in concentrated markets does not differ appreciably by gender or age (or level of education – see Annex Figure 3.A.3), concentration is only one measure of monopsony power. As discussed in Section 3.1, there are other aspects of monopsony apart from concentration that may differentially affect vulnerable groups. Furthermore, concentration may still impact some labour market outcomes unevenly across groups of workers, as shown in Section 3.3.2.

3.2.3. Labour market concentration and the COVID‑19 pandemic highlighted existing inequalities in the labour market

The onset of the COVID‑19 crisis saw workers split into three groups: those who were able to work from home (telework), those who found themselves unemployed or on reduced working hours, and those who continued to work in their physical workplace and in proximity of other people during the pandemic, or front-line workers – see Chapter 1. The gradual abatement of lockdowns and the recovery of the labour market have greatly diminished the ranks of the unemployed and those on short-time work (OECD, 2021[65]). However, more than two years after the onset of the pandemic, the dichotomy between those who must work in person, and workers who may work from home, is still relevant – see Chapter 1.

Labour market concentration may degrade occupational safety if investing in a safe work environment is costly for employers. Employers in concentrated markets may not need to offer a safe work environment to attract and retain good workers. If front-line workers are in concentrated markets, therefore, they could face a heightened risk of infection. Many OECD governments have instituted various protective measures for workers who are required to work, and therefore stand a chance of infection (OECD, 2020[66]) – see also Chapter 2. In countries or regions where such precautions are not mandated, or where workers nonetheless find them inadequate, often one’s only recourse is to quit, and find a job with an employer with better safety measures. Moreover, the ease with which a worker can credibly quit can by itself spur greater safety measures. Whether front-line workers face monopsonistic labour markets is, all things equal, an important aspect of their occupations’ safety and job quality.

Figure 3.5 depicts the share of workers in highly concentrated labour markets by whether their occupation is required to work in person, and whether, because of close contacts with colleagues or customers, they have a high risk of infection with COVID‑19 on the job compared to those who do not (Basso et al., 2022[67]). On average, about 11% of these workers at significant risk of COVID‑19 infection are found in highly concentrated labour markets compared to a little over 9% of those who are not. The largest gaps are found in Australia, the Netherlands and the United Kingdom. In contrast, there is little difference in the shares in highly concentrated markets in the United States and Singapore. Women, the low-educated and workers on temporary contracts among other more economically vulnerable groups are over-represented among at-risk workers (Basso et al., 2022[67]; DOL, 2022[68]).

The other defining feature of labour markets during the pandemic were workers who had the option of working from home. Workers who are able to telework are those in occupations where one can work from home without physically interacting with co-workers or customers, based on the tasks that are typically performed on their job according to the US Occupational Information Network database (Dingel and Neiman, 2020[69]; Basso et al., 2022[67]).

Figure 3.5. Occupations where workers face significant risk of COVID‑19 infection tend to be more concentrated

Compounding the a priori occupational health disparity with front-line workers, workers who are able to telework are found in less concentrated labour markets. On average, 9% of workers in occupations amenable to telework are in highly concentrated markets on the eve of the COVID‑19 crisis, compared to 11% of those workers who cannot telework (Figure 3.6).

In addition to protecting workers from the virus, the shift to telework during the pandemic may have improved the labour market prospects of these workers. Workers who can telework may search in a wider labour market than simply their local living area. This has the potential to further lower local employers’ monopsony power, and increase bargaining power for workers who can telework (Section 3.4.2).

Figure 3.6. Workers who can telework face less concentrated labour markets

3.2.4. Concentration rose at the onset of the crisis, but it has largely returned to pre‑COVID‑19 levels

One year into the pandemic, labour market concentration increased. Figure 3.7 shows the change in concentration from 2019 to the average of 2020Q2‑2021Q1.19 Concentration increased by 10% over this time period on average across the OECD countries in the sample, with the United Kingdom, Latvia and Estonia recording the largest growth. In France, Austria, Australia and Belgium, the HHI in the year following the onset of the pandemic was on average below its pre‑crisis level.

Figure 3.7. A year into the pandemic, concentration settled slightly above pre‑COVID‑19 levels

These average values of concentration one year into the pandemic mask dynamics at the quarterly level. The average HHI grew by over 20% year over year in the second quarter of 2020 (Annex Figure 3.A.4). By the third and the fourth quarter of 2020, HHIs continued to grow, on average, but many countries were experiencing year-over-year declines in their aggregate HHI. By the first quarter of 2021, HHIs had decreased sharply in most countries.20 This suggests that the run-up in concentration at the beginning of the pandemic is starting to abate, and the HHI is converging back toward pre‑pandemic levels.

These patterns likely reflect the limited number of firms posting job vacancies during the acute stages of the pandemic, and the progressive normalisation of hiring in more recent periods. For example, some firms not hit by mandatory foreclosures, kept posting job openings even during the peak of the pandemic, causing a temporary increase in labour market concentration. The dynamics could also reflect an initial stark increase in concentration in certain sectors that represented a larger share of employment in 2019,21 e.g. retail. However, as a large share of workers who would have otherwise sought new job opportunities refrained from doing so because of the pandemic (OECD, 2021[65]), it is not clear that the described movements in labour market concentration translated into actual changes in the wage‑setting power of employers.

3.3. The effects of labour market concentration on labour market performance

The analysis in the previous section finds that labour market concentration is pervasive across OECD countries. However, if labour market concentration leads to monopsony power, one should expect concentration to be associated with changes in employment and wages. This section presents evidence of the effect of concentration on job quantity (employment) and quality (wages). The section begins by reviewing the literature on changes in employment in more concentrated markets, as well as how concentration affects wages. In addition to the literature, this section provides new cross-country empirical estimates of the effect of concentration on earnings, job security and job stability using matched employer-employee data. The estimates also disentangle the effects of concentration on different groups including youth and women. The section concludes by showing how labour market concentration affects the skill composition of labour demand.

3.3.1. Labour market concentration tends to reduce employment

Monopsony in the markets for inputs (including labour) can have a negative impact on prices (wages and benefits) and quantities (overall employment). In principle, one would expect to find a clear relationship between measures of monopsony or labour market concentration and employment (see Section 3.1). In practice, however, few studies have documented this relationship due to the difficulty of identifying the effect of labour market concentration independently from other confounding factors while simultaneously solving potential reverse causality issues.

Most of the studies in the literature focus on plant takeovers and mergers. These studies typically find a negative effect of mergers or takeovers on employment at merged firms or acquired plants. A number of early studies have looked at takeovers and found negative effects on employment – for example Lichtenberg and Siegel (1990[70]). Takeovers, however, may not result in greater concentration and market power if they are simply the result of a change in ownership with the acquiring entity operating in other, unrelated markets. More recent studies have focused directly on horizontal mergers, which are more likely to result in increased concentration, with similar results – that is, negative effects of mergers on employment levels of merged firms, see Conyon et al. (2001[71]), for the United Kingdom, Lehto and Böckerman (2008[72]) for Finland, Siegel and Simons (2010[73]) for Sweden, Arnold (2021[74]) for the United States, and the cross-country study of Gugler and Yurtoglu (2004[75]), covering European countries and the United States.22

The limit of merger studies is that they usually cannot disentangle changes in product market competition and, often, efficiency gains from mergers from changes in labour market competition. Policy responses are obviously different when the effect on employment derives from efficiency gains instead of inefficient demand restraints. In one of the very few studies trying to isolate directly the economy-wide effect of labour market concentration on employment, Marinescu, Ouss and Pape (2021[34]) examine its impact on new hires in France, controlling for both productivity and product market concentration. Relying on a standard leave‑one‑out instrumental variable strategy for identification (see Box 3.2 below), they find a very large negative effect of concentration on new hires: taking their estimates at face value, increasing the concentration index at the sample mean by 10% would imply a reduction in the number of new hires in a given local labour market by as large as 3%.23 However, such large effects could suggest a problem of misspecification, related for example to the fact that the number of new hires is indirectly an input into the measure of concentration.24 For this reason, these results should be taken with some caution.

Overall, these results suggest that labour market concentration tends to have a negative impact on employment, although more research is needed to establish the magnitude of this effect. However, job quantity is only one aspect of labour market performance and job quality is equally important. The next section analyses the possible effects of concentration on job quality.

3.3.2. Labour market concentration also has an adverse impact on job quality

Labour market concentration reduces earnings

There is a large empirical literature that has tried to estimate the effect of employers’ market power on job quality, although most available studies focus only on the impact on wages and earnings. The literature on the effects of mergers on wages in the merging firms has yielded mixed results – see e.g. Lichtenberg and Siegel (1990[70]), Currie, Farsi and Macleod (2005[76]) and Siegel and Simons (2010[73]). More recent studies have shown that the impact of mergers on wages in a labour market tend to be larger, the greater the impact of the merger on labour market concentration – see e.g. Prager and Schmitt (2021[77]) and Arnold (2021[74]). Recent studies have also looked at the impact of reforms leading to enhanced firm entry, divestitures or greater outside options, thereby unambiguously increasing competition, and have typically found positive effects of these reforms on wages – see e.g. Hensvik (2012[78]), Hafner (2021[79]), Thoresson (2021[80]) and the literature on non-compete agreements discussed in Section 3.4.1 below.

A large recent literature has estimated directly the impact of local labour market concentration on wages in the United States.25 There is also a growing body of recent evidence covering other OECD countries.26 Most of these studies use instrumental variable techniques to deal with potential endogeneity issues (see Box 3.2). The estimated elasticity of wages to concentration typically ranges between ‑0.01 and -0.05. That is, when concentration doubles the wage falls by between 1% and 5%, with larger estimates being found only in a few of the US studies.27 However, the heterogeneity of the measures of concentration and the differences in the specifications used make it difficult to compare point estimates across countries.28

In order to present comparable cross-country estimates, this section relies on Bassanini et al. (2022[37]), who analyse the impact of labour market concentration on wages and job security using harmonised linked employer-employee data for a number of European OECD countries (see Box 3.2 for a detailed discussion of the specification).29

In the four countries for which comparable wage data are available (Denmark, France, Germany and Portugal) the estimated elasticity of wages to labour market concentration varies between ‑0.02 (in Germany) and ‑0.03 (in Denmark) in the case of daily wages for full-time workers (Figure 3.8).30 In other words, at the sample average, increasing labour market concentration by 10% lowers the daily wage by 0.2% to 0.3%.31 This may seem low at first glance, but these results must be interpreted considering that concentration distributions are quite dispersed. In all these four countries, the ratio of the 9th decile of the distribution of HHIs to the median HHI is between 6.7 (in Denmark) and 8.8 (in Germany and France, see Annex Figure 3.A.5). Taken at face value, these estimates therefore imply that, all other things equal, the 10% of workers who are employed in the most concentrated labour markets experience a wage penalty of at least 5% with respect to the median worker. And a few of them, those in markets with concentration well above the 9^th decile, suffer from a much greater wage penalty.32

Overall estimated elasticities for different countries remain close to one another. This is remarkable, given the significant differences across the labour markets of these countries – see e.g. OECD (2018[81]). These estimates also appear close to most of the other estimates in the literature, including for countries not included in our sample.33 These two observations, taken together, cautiously suggest that the pattern presented in Figure 3.8 is likely to be more general, and rigorously estimated average wage elasticities are likely to belong to this range in other OECD countries not shown in the chart.

Box 3.2. Estimating the impact of concentration on wages and job security

Bassanini et al. (2022[37]) estimate the effect of labour market concentration on wages and job security on samples of linked employer-employee data on the following groups: all workers, full-time workers and new hires. They use the following specification:

$Y_{i, j, f, l, s, t} = β \log ({H H I}_{l, t}) + γ X_{i, j, f, l, s, t} + μ_{f t} + μ_{l} (+ μ_{i}) + ε_{i, j, f, l, s, t}$

where $Y$ stands for the dependent variable, $X$ is a vector of individual and plant-level controls, $μ$ are fixed effects (with parentheses indicating fixed effects that are not included when the equation is estimated only on the sample of new hires), $i$ indexes the worker, $j$ the plant, $f$ the firm-by-municipality couple,1 $l$ the local labour market, $s$ the industry and $t$ is the year. $H H I$ stands for the Herfindahl-Hirschman Index calculated using the share of each employer in new hires in the local labour market defined by 4‑digit occupation and cross-country comparable functional geographical areas, so that $l = (o, z)$ , where $o$ is the occupation and $z$ is the geographical area.2 The dependent variables include: the logarithm of daily and hourly wages; and dummy variables for having started an open-ended contract at hiring, or having the contract converted into an open-ended one within one year. Due to data limitations, wage equations are estimated only for Denmark, France, Germany and Portugal, while job security equations are estimated for France, Germany, Italy and Spain. In each country, household workers, self-employed, and those working in agriculture and outside the business sector are excluded from the sample.

Ordinary least squares (OLS) cannot consistently estimate the above equation if there is a time‑varying factor that is correlated with both the local HHI and the dependent variables and is not proxied for by existing control variables. For example, positive or negative shocks to local labour supply are likely to affect the wage offers that workers are ready to accept and the number of firms that find it attractive to operate in the local labour market, thereby biasing OLS estimates of the above equation. To solve this problem, many papers3 resort to a leave‑one‑out instrument à la Hausman, which is popular in the trade and industrial organisation literatures.4 In practice, $\log (H H I)$ in local labour market $l = (o, z)$ at time $t$ is instrumented with the average of $l o g ({1 / N}_{o, z', t})$ in all other functional areas $z ’$ for the same occupation $o$ and time period $t$ – where $N_{o, z^{'}, t}$ is the number of firms with positive number of hires in a given year. The same strategy is followed as regards estimates presented in this chapter.

1. The firm-by-municipality fixed effect plays a key role as it allows controlling for labour productivity and product market competition at both the national and local level. The only other study using the same fine‑grained control for productivity and product market competition is Bassanini, Batut and Caroli (2021[57]). Qiu and Sojourner (2019[31]), Marinescu, Ouss and Pape (2021[34]) and Benmelech, Bergman and Kim (2022[29]) include labour productivity, as measured by accounting data, as a control variable without, however, addressing its endogeneity.

2. In the main specification functional geographical areas are composed of OECD functional urban areas (OECD, 2012[82]) and remaining large portions of NUTS3 regions, the latter being added to ensure a mixture of urban and rural areas. Results are however robust to using either functional urban areas or NUTS3 regions only.

3. Azar, Marinescu and Steinbaum (2022[52]), Rinz (2022[32]), Martins (2018[35]), Qiu and Sojourner (2019[31]), Marinescu, Ouss and Pape (2021[34]), Bassanini, Batut and Caroli (2021[57]), OECD (2021[12]) and Popp (2021[83])

4. See e.g. Hausman, Leonard and Zona (1994[84]), Nevo (2001[85]), and Autor, Dorn and Hanson (2013[86]).

Source: Bassanini et al. (2022[37]), “Labour Market Concentration, Wages and Job Security in Europe”, https://docs.iza.org/dp15231.pdf.

Figure 3.8. Estimated wage effect of labour market concentration

The similarity of the estimated wage elasticities across countries, and their close alignment with the literature, suggest that it is possible to use the literature to infer how these elasticities might have changed over time. The estimates in this chapter were obtained on a limited number of years which does not allow studying trends in the wage elasticity over time. Given the close conformity of these estimates, one can use the wider literature as a guide as to how these elasticities may have evolved over time. For example, using a different concentration measure, OECD (2021[12]) find that this elasticity has become on average more negative in the last two decades. In other words, even though labour market concentration has not increased – see Section 3.2.3, its impact has become stronger over time. One possible explanation might be the concomitant reduction of collective bargaining and the weakening of trade unions (OECD, 2019[15]), which may be increasingly less able to act as a countervailing power – see Section 3.4.1.

In the four countries for which the analysis is possible, there is no systematic gender difference in wage elasticities to labour market concentration. This may appear surprising in view of the literature on separation elasticities that has tended to find smaller elasticities for women than for men (Manning, 2003[5]; Hirsch, Schank and Schnabel, 2010[87]; Webber, 2016[13]; Vick, 2017[88]). The estimates in Figure 3.8, however, should not be interpreted as implying that women are exposed to the same degree of monopsony power as men. As discussed in Section 3.1, women tend to search for jobs closer to their home and are ready to accept a significant wage penalty for a closer job. As a result, the same level of concentration implies fewer acceptable alternative jobs for women, and therefore lower wages. But increasing concentration may still have a similar percentage effect on the rarefication of available alternatives for both men and women, consistent with the gender pattern shown in Figure 3.8.

The negative impact of labour market concentration on wages as presented above is the aggregation of the average effects on two different groups of employees: those who have been hired over the previous year (the new hires) and those who were already employed by the firm the year before (the incumbents). It has been conjectured in the literature that the effect on new hires should be larger than that on incumbents (Marinescu, Ouss and Pape, 2021[34]) since the latter’s wage is considered to be less sensitive to changing labour market conditions (Pissarides, 2009[89]; Haefke, Sonntag and van Rens, 2013[90]; Kudlyak, 2014[91]). Disaggregating the effect of labour market concentration between new hires and incumbents, the effect on the former’s wages, while always significant, does not appear systematically larger than that on incumbents’ (Figure 3.9).34 For incumbents, one can conjecture therefore that the impact of concentration on wages occurs mainly through reduced rates of promotions and lack of wage increases – that is upward wage rigidity, rather than downward wage flexibility, which might more easily concern new hires. This is consistent with recent findings by Grigsby, Hurst and Yildirmaz (2021[92]), who suggest that incumbents’ wages appear no less flexible than those of new hires once the characteristics of the latter are properly accounted for. From a policy perspective, this is important since incumbents represent a large share of employment and their wage dynamics have been found to be driving aggregate wage growth in recent years (Hahn, Hyatt and Janicki, 2021[93]; Hijzen, Zwysen and Lillehagen, 2021[94]).35

Figure 3.9. Estimated wage effect of labour market concentration for incumbents and new hires

Employers in more concentrated markets tend to use less stable contracts

There is a large literature showing that workers consider wages and working conditions together when evaluating jobs and job offers, and are ready to trade off part of their wage for terms and conditions of employment that they consider to be better – see e.g. Mas and Pallais (2017[17]), Taber and Vejlin (2020[95]) and Kesternich et al. (2021[96]) for recent evidence. If delivering better terms and conditions of employment is costly for employers, it can be expected that monopsonistic employers will tend to offer jobs with worse terms and conditions (Manning, 2003[5]). Yet, there is surprisingly little literature on the effect of labour market concentration on the terms and conditions of employment. Qiu and Sojourner (2019[31]), who find a negative effect of concentration on the probability of being covered by employer-provided health insurance, represents one of the few exceptions.

There is evidence that, all things equal, employees often have a preference for stable jobs and firms need to offer a wage compensation for more unstable or insecure jobs – see e.g. Bassanini et al. (2013[97]) and Albanese and Gallo (2020[98]). Dynamic monopsony theory would predict, therefore, that employers in more monopsonistic labour markets would be more likely to offer temporary contracts at the margin in an attempt to shift most of the labour adjustment onto workers.36 One could expect this mechanism to be particularly important in countries with stringent employment protection rules, which imply higher termination costs for employees on open-ended contracts than on temporary contracts (OECD, 2020[99]; 2021[100]).

Regression analysis suggests that labour market concentration tends to increase the use of flexible contracts. Figure 3.10 reports estimates of the effect of labour market concentration on the probability of starting a permanent contract at the time of hiring, and the probability of having the contract converted into an open-ended one if hired on a temporary contract.37 In Germany and France, increasing concentration by 10% from the average level is estimated to reduce significantly the probability of being offered an open-ended contract at hiring, with effects that vary between 0.35% (for French men) and 0.7% (for French women). In other words, in these two countries, taking into account the dispersion of the distribution (see Annex Figure 3.A.5), the 10% of workers in the most concentrated markets are estimated to be at least 10% less likely to be hired on a permanent contract than those in a labour market with median concentration. The estimated effect of a 10% increase in concentration is much higher in Portugal (about 2.3% for both men and women), but it is imprecisely estimated. By contrast, the effect is insignificant in Italy and Spain. The latter finding probably reflects the fact that most employees are first hired on temporary contracts in these two countries,38 even in low concentration labour markets, which reduces the scope for further increasing temporary contracts for firms with market power.

Figure 3.10. Estimated effect of labour market concentration on contract type

Box 3.3. Labour market concentration and job stability: Evidence from France and Germany

What is the expected effect of monopsony and labour market concentration on job stability? On the one hand, to the extent that employees have fewer outside options, it can be expected that job spells become longer as employees find it more difficult to quit for another job. On the other hand, to the extent that employers with market power impose lower wages and worse working conditions, a greater share of their employees might be tempted to quit (Manning, 2003[5]). The overall effect is therefore ambiguous.

Figure 3.11 shows estimated percentage point effects of labour market concentration on the probability of working with the same employer 12 months after hiring in France and Germany. In the case of workers hired on permanent, open-ended contracts the impact of labour market concentration is insignificant (and very close to 0 in Germany), reflecting the offsetting mechanisms outlined above. A negative effect emerges for those hired on temporary contracts. In this case, a 10% increase in labour market concentration results in lower retention rates 12 months after hiring by 0.2 to 0.4 percentage points. Whatever the factors behind this pattern,¹ in France and Germany, labour market concentration appears to depress job security by both affecting contract characteristics and job spells.

Figure 3.11. Estimated effect of labour market concentration on job stability

Available data also allow the examination of the impact on contract conversion in Italy and Spain. For those hired on a temporary contract, labour market concentration clearly depresses their chances of accessing a more stable position within the calendar year after hiring (Figure 3.10). The effect appears particularly large for Italy where a 10% increase in concentration reduces the conversion rate by 2.5% for both men and women.39 In France, Germany and Portugal, the structure of the data prevents this type of analysis,40 but one can have an indication of the impact of concentration on the precariousness of temporary contracts in these countries by looking at retention rates one year after hiring (Box 3.3).41 Overall, the findings presented in Figure 3.10 and Box 3.3 provide evidence that labour market concentration has a negative effect on job security. Employers with market power tend to shift adjustment costs onto workers by either hiring more workers on temporary contracts or reducing the conversion rate of these contracts.

Is labour market concentration particularly bad for youth?

The evidence presented above suggests a strong effect of labour market concentration on job quality, and it is consistent with the idea that employers with market power inefficiently reduce labour demand in order to dampen the cost associated with wages and non-wage attributes. However, employers depressing their labour demand could also be expected to become more selective in hiring. For example, they may prefer job candidates with work experience – whose competences are therefore less noisily signalled by their resume – to those with more uncertain productivities, such as young labour market newcomers. Figure 3.12 shows that, in three out of four of the countries for which data are available, the wage elasticity to labour market concentration is stronger for youth than for older adults. In particular, in France and Denmark, the wage elasticity difference between employees aged 24 years or less42 and their older peers is estimated to be at least 50% larger, in absolute terms, than the economy-wide wage elasticity (cf. with Figure 3.8 above).

Germany is the only country for which the estimated wage elasticity is smaller for youth than for their older peers. While the positive estimated wage elasticity differential in the case of Germany may be surprising, it cannot be interpreted without looking at the effect of concentration on other dimensions of job quality. Labour market concentration appears to have consistently a stronger negative impact on the probability of starting a permanent, open-ended contract at the time of hiring for youth than for other adults in all the countries for which this effect can be estimated, with the exception of Portugal. But the impact is particularly large for Germany where the differential effect between youth and older adults is as large as the economy-wide effect (cf. Figure 3.12 with Figure 3.10 above). In other words, this pattern is consistent with German youth trading off wages and type of contract differently with respect to other countries. This could be due to the fact that a large share of temporary contracts in Germany are apprenticeships, which tend to be considered better than other temporary contracts.43 This points to the importance of examining several different margins together in order to appreciate the full impact of labour market concentration. Looking only at wage effects may indeed provide a biased picture.

Taken together and interpreted with caution, these results suggest that employers in concentrated labour markets tend to become more selective, and that this may have a particularly negative impact on the job quality of young workers. The next section analyses the possible effects of this increased selectivity on skill demand.

Figure 3.12. Estimated differential effect of labour market concentration between youth and adults

3.3.3. The effect of concentration on skill demand

Employers operating in monopsonistic markets may leverage the presence of limited outside options to become more selective in their recruitment, hire workers with greater skills, or become more demanding vis-à-vis incumbent employees. The limited existing literature finds supporting evidence of a positive association between labour market concentration and demand for skills in US online job advertisements (Macaluso, Hershbein and Yeh, 2019[101]). Other studies on the United States data also find that skills requirements within occupations increase (respectively, decrease) when labour markets are slack (tight) and more (less) talent is available on the market (Modestino, Shoag and Ballance, 2016[48]; Modestino, Shoag and Balance, 2020[102]), but they do not associate the phenomenon to changes in the relative market power of employers.44

This section provides new empirical evidence of the effect of labour market concentration on skill demand, as reported in online job advertisements. It expands on the existing literature by covering countries other than the United States, and by presenting first-time causal estimates of the relationship of interest, as opposed to correlations.

Multiple proxies of (online) skill demand are used: the number of distinct skill categories demanded in a job posting, an indicator of the fact that the job posting requires cognitive skills, and a similar indicator for social skills. The focus on cognitive and social skills is motivated by their importance in explaining wage heterogeneity across labour markets and firms (Deming and Kahn, 2018[103]), and the fact that they are largely in demand, both separately and jointly (Deming, 2017[104]; Deming and Kahn, 2018[103]). The construction of these skill demand indicators restricts the analysis to four countries for which data are available (Australia, Canada, the United Kingdom, the United States). More information on the construction of the skill indicators and the empirical specification is proposed in Box 3.4.

The regression analyses tend to confirm that employers in more concentrated markets demand more skilled workers (Figure 3.13). In Australia, Canada and the United States, a 10% increase in labour market concentration from its average level in the country significantly increases the number of skill categories mentioned as requirements in the job advertisement by 1.2% (United States), 1.4% (Australia) or 1.8% (Canada) of the average number of skill categories in the country’s job postings. The same change in concentration would increase the probability that an employer requires at least one social skill by 1.1% (United Kingdom) and 1.8% (Canada) of the initial average probability. The result for cognitive skills, when significant, is larger (although more imprecisely estimated), reaching 4.2% (Canada) to 5.5% (United States) of the average probability for a posting to require at least one such skill. In the remaining instances, the estimated effects are statistically insignificant and with magnitude close to zero, but there is no consistent country-specific pattern. The lack of a significant finding could reflect the absence of a relationship between concentration and skills demand in those countries, conditional on the large number of fixed effects included. However, it could also reflect a context where firms offering low wages can only attract low-skill workers, if their wage elasticity is lower than that of high-skill workers.

These exceptions notwithstanding, the analysis documents that monopsony power, as proxied by labour market concentration, induces not only a reduction in wages and job security, but also an increase in the skills that workers are required to mobilise while performing their job.

Figure 3.13. Estimated effect of labour market concentration on skill demand in online job advertisements

Box 3.4. Estimating the impact of concentration on skill demand in job advertisements

This chapter estimates the effect of labour market concentration on indicators of demand for skills, as reported in job advertisements. The following specification is used:

$Y_{i, f, l, s, t} = β \log ({H H I}_{l, t}) + γ X_{i, f, l, s, t} + μ_{f, z, t} + μ_{l} + ε_{i, f, l, s, t}$

where $Y$ stands for the dependent variable, $X$ is a vector of posting-level controls, $μ$ are fixed effects, $X_{i, f, l, s, t}$ stands for classes of education attainment and work experience that are used as extra controls, and $H H I$ stands for the Herfindahl-Hirschman Index as calculated using the share of each employer in total postings in the local labour market defined by 4‑digit occupation and TL3 regions, as in Section 3.2.The letter $i$ indexes the posting, $f$ the employer, $l$ the local labour market, $s$ the industry and $t$ is the quarter in a year. $l = (o, z)$ , where $o$ is the occupation and $z$ is the geographical area. In light of the fixed effects used, empirical results are identified by differences in job postings of the same establishment across occupations, and their variation over time.

The dependent variables of interest are alternatively the number of skill categories mentioned in the job advertisement, an indicator variable signalling that the posting requires at least one cognitive skill, and a similar indicator for one social skill. The database used for the analysis lists thousands of distinct skills, which cannot be easily or meaningfully described, if not grouped in an appropriate way. Furthermore, some of this variation is fictitious, as it originates from synonyms, differences in spelling, or from country-, occupation- or employer-specific practices in writing the advertisements. The skill keywords reported in the database are therefore grouped in 61 mutually exclusive skill categories following Lassébie et al. (2021[105]), ahead of performing the analysis. This allows identifying cognitive skills from job advertisements containing keywords related to quantitative abilities, reasoning, problem solving, learning and originality; and social skills via keywords related to co‑ordination, decision-making, persuasion and negotiation, social perceptiveness, speaking, writing, communication, or active listening. In the sample considered for the analysis, 20 to 28% of job advertisements explicitly require at least one cognitive skill, and 40 to 45% require at least one social skill.

The availability of the classification of skills into skill categories according to Lassébie et al. (2021[105]) constrains the analysis to Australia, Canada, the United Kingdom and the United States. Outliers and postings from firms operating in agriculture, household production, and the public sector (education, health care and social security, public administration and defines) are excluded.1 Within each country, the analysis is performed on a panel of 5 000 randomly chosen employers covering the years 2017‑19, where the sampling enables significant savings in computing power and time. The time series is kept short to limit possible biases emerging from changes in the representativeness of a dataset of online job postings over long periods.

Obtaining consistent estimates of the effect of labour market concentration on skill demand requires, however, an instrumental variable approach. Following a standard practice in the literature (see Box 3.2), a given market’s HHI is therefore instrumented by the inverse of the number of firms posting ads in the same occupation but in all other TL3 regions, averaged over all these alternative regions.

1. Outliers are postings that alternatively (i) report more than 20 skill requirements, or (ii) are advertised for an occupation that accounts for less than 1% of total postings in the year.

3.4. Monopsony and labour market policy: Direct and indirect policy considerations

The previous sections have shown that employers’ market power is likely to be large in many labour markets. In particular, the analysis has shown that about one sixth of workers in the business sector of 15 OECD countries find themselves in moderately or highly concentrated labour markets and that this is likely to have a negative impact on both job quantity and quality. This section discusses labour market policy in light of labour markets with monopsony power.

The discussion will touch on both direct and indirect policy considerations. Direct policy considerations concern first-order levers to counteract the power imbalance between workers and firms. The two main direct policy interventions include the role of competition and labour policy with regard to labour market concentration and abuses of monopsony power, and the role of trade unions to counteract firms with strong bargaining power. Other policy considerations concern how policy makers should view labour market policies in a world where labour markets are not perfectly competitive and how these policy levers can be mobilised to counterbalance the adverse effects of monopsony. This section will discuss three examples of such indirect policy levers: minimum wages, geographical mobility and teleworking, and reskilling.

3.4.1. Direct policy interventions

At its simplest, monopsony in labour markets tilts the balance of power towards employers and away from workers. Logically, policies which directly limit concentration or counteract uneven employer power in the employment relationship can improve labour market outcomes for incumbent workers in the firm as well as job seekers and similar workers employed at competing firms. This section discusses two categories of such policy interventions: explicit regulation to limit employer concentration and fight abuses of monopsony power, and the role of trade unions.

Regulation to address labour market monopsony

Historically, legislators and enforcement authorities (including labour inspectorates and antitrust authorities) have paid little attention to employers’ market power in the labour market. However, this issue, as well as how to enhance competition in the labour market, is receiving increasing attention – see e.g. US Department of Justice; Federal Trade Commission (2016[106]; 2022[107]), US Department of Treasury (2022[108]), and Vestager (2021[109]). There are four areas of action, in which a more active role of regulators could be considered. Ranking them in terms of their links with the competences of labour authorities, they are: i) non‑compete agreements; ii) occupational licensing; iii) no‑poaching agreements and wage‑setting collusion; and iv) mergers.45

Non-compete agreements

Non‑compete agreements or covenants (NCAs) are clauses in contracts that prevent workers from working for a competitor after they separate from their employer.46 In most countries, NCAs are lawful and justified by the need to protect trade secrets and specific investment in the employment relationship by the employer (such as investment in knowledge).47 Where statistics on NCA use are available, they suggest that they are widespread. According to an establishment survey of 2019, between 28% and 47% of US private‑sector workers are subject to NCAs (Colvin and Shierholz, 2019[110]). Vuorenkoski (2018[111]) reports that 45% of workers belonging to the Finnish trade union Akava are bound by an NCA. A 2015 Dutch survey shows that about 19% of employees were covered by NCAs (Streefkerk, Elshout and Cuelenaere, 2015[112]). Young (2021[113]) reports that over 35% of private sector workers in Austria were bound by an NCA in 2005‑06.

NCAs are often considered to have a positive impact on employers’ investment in intangible capital and training, in particular when incumbent companies cannot protect their investment in knowledge through patents or other types of contracts (such as training pay-back clauses). However, NCAs may lead to negative spillovers resulting in a tax on future employers, thereby discouraging market contestability, firm entry and entrepreneurship. This explains why there is no unambiguous evidence on their impact on innovation and productivity – see e.g. Starr, Balasubramanian and Sakakibara (2018[114]); Shi (2020[115]); Lavetti (2021[116]); and Jeffers (2021[117]). By discouraging entrepreneurship and firm entry, NCAs use tend to increase labour market concentration – see Hausman and Lavetti (2021[118]).

There is also evidence that employers tend to use NCAs to limit the outside options of their employees. They are in fact frequently used in many countries even when the employee has no access to the employers’ trade secrets or other intangible assets. For example, Starr, Prescott and Bishara (2021[119]) find, using a large sample of US workers, that more than 40% of the workers bound by an NCA neither worked directly with clients nor had access to client information or other trade secrets.

Existing evidence from the United States suggests that facilitating the enforceability of NCAs unambiguously reduces job mobility and often depresses wages – see e.g. Marx, Strumsky and Fleming (2009[120]); Starr (2019[121]); McAdams (2019[122]); Lipsitz and Starr (2022[123]), except in firms and occupations where employers can credibly commit to share with employees the returns from enhanced investments in intangibles – see e.g. Lavetti, Simon and White (2019[124]).The negative effect of NCAs on job mobility and wages tends to be stronger for women, likely due to stronger preference for shorter commuting (Johnson, Lavetti and Lipsitz, 2021[125]).48 Research from Austria confirms that, even in other countries, reducing the enforceability of NCAs enhances job to job transitions to better paid jobs (Young, 2021[113]).

Most jurisdictions impose that, to be enforceable by courts, NCAs need to respect a number of reasonableness conditions, which are designed with the purpose of limiting abuse – see e.g. Meritas (2017[126]) for an overview of existing rules in OECD countries. However, courts typically assess the reasonableness of non‑compete agreements on a case‑by‑case basis, and costly litigation by workers often results in simply waving unenforceable covenants, with no additional gain for plaintiff workers (Krueger and Posner, 2018[127]). For this reason, governments could consider banning NCAs, or establishing a rebuttable presumption of abusive use.49 This would be particularly important in the case of certain type of positions, pay levels or skill requirements, for which a clear justification, such as the protection of trade secrets, seems implausible.50

However, even when NCAs are unenforceable, they may still be included in employment contracts as a way to put pressure on uninformed employees. For example, 19% of employees in California and North Dakota report having signed an NCA, despite the fact that these clauses are legally not enforceable in these states (Starr, Prescott and Bishara, 2020[128]), suggesting that employers still use them to deter mobility despite their formal lack of enforceability. This strengthens the case for a proactive role of enforcement agencies, including labour inspectorates, to curb abuses, but in order to be effective, they should have the possibility of imposing sanctions or taking the case to courts, which should be empowered to impose them. Governments could also consider imposing minimum compensation schedules during the period after separation in which the NCA binds, as done for example in Denmark, France, Norway and Sweden (Vuorenkoski, 2019[129]; Berjot, 2021[130]) and enacted recently in Finland (Autio, 2021[131]).

Occupational licenses

Over 20% of jobs in many OECD countries require some form of occupational license (Koumenta and Pagliero, 2018[132]; Hermansen, 2019[133]). By imposing minimum standards of competence to practice for pay, occupational licensing limits entry into the occupation to those practitioners whose skills have been recognised to be at or above the minimum requirements. Consequently, it reduces the pool of practitioners, thereby potentially giving them (or their employers) significant market power in the service market (Pagliero, 2011[134]), which tends to yield higher service prices (Wing and Marier, 2014[135]; Kleiner, 2017[136]). Conversely, licensing may improve service quality and customers’ protection, as well as wages and working conditions of workers in that occupation. Available research indeed suggests that licensing generates a wage premium in licensed occupations (Kleiner and Krueger, 2013[137]; Gittleman, Klee and Kleiner, 2017[138]; Zhang, 2018[139]; Koumenta and Pagliero, 2018[132]).

Recent research in the United States, however, has pointed out a more subtle effect of licensing in the labour market. Certain occupations are closer substitutes and similar tasks can be performed by workers in more than one occupation. Moreover, workers with similar competencies can work in multiple occupations. Imposing licensing requirements in one of them has a negative effect on wages in other, closely-related occupations, as in Kleiner et al. (2016[140]) and Dodini (2020[141]). This is consistent with a monopsony model, in which licensing, by reducing outside options for workers in closely related, non-licensed occupations, increase monopsony power of employers in these occupations (Kleiner and Park, 2010[142]). Dodini (2020[141]) further shows that his results cannot be explained by the labour supply shock induced by licensing in one occupation in closely related occupations.

Although more research is needed to confirm this evidence, policy makers may want to consider these cross-market effects when evaluating costs and benefits of occupational licensing. A valid substitute for licensing could be certification, which offers practitioners the option to join a scheme that verifies and guarantees their skills but without imposing any legal restriction (Koumenta and Pagliero, 2018[132]).

Labour market collusion

In most jurisdictions, competition law forbids collusion among buyers of intermediate goods or services, including labour services – see e.g. Blair and Wang (2017[143]).51 General statistics on collusion are difficult to collect, since figures on those illicit behaviours that escape investigation are typically not available. Statistics on non‑poaching covenants exist for franchising agreements, where these covenants are not necessarily unlawful – see OECD (2019[15]) for a discussion. Krueger and Ashenfelter (2022[144]) estimate that more than 50% of major franchise companies in the United States use no‑poaching clauses in their franchising agreements. Theory and empirical evidence also suggests that collusion is more likely to occur in concentrated markets, since co‑ordination among few actors is typically easier to sustain – see e.g. Asker and Nocke (2021[145]). Moreover, as it can more easily cover all or most of the actors, collusion is likely to be more damaging in concentrated markets: in the only academic study evaluating the impact on wages of major no-poaching agreements investigated by US antitrust authorities, Gibson (2021[56]) finds that each agreement among any two firms operating in the mid‑2000s, in the highly concentrated Silicon Valley high-tech sector, suppressed about 2.5% of annual wages.

Providing explicit guidance about labour market collusion is crucial to guide and set priorities for enforcement agencies. For example, US antitrust authorities have issued guidelines that explicitly refer to collusion in the labour market, present clear examples of illicit behaviours, and underline the importance of fighting them for their effects in the labour market (US Department of Justice; Federal Trade Commission, 2016[106]). Whistleblower protection and adequate leniency programmes, which offer immunity to the first cartel member that blows the whistle, are also important for effective enforcement since collusion is often discovered based on information provided by insiders (Dyck, Morse and Zingales, 2010[146]; Yeoh, 2014[147]; Luz and Spagnolo, 2017[148]). Last, but not least, public enforcement action by antitrust authorities has a key role to play in this context, as they are usually able to impose sanctions for collusive behaviours (OECD, 2020[149]). Private enforcement actions may also be brought by individual employees. However, individual employees often do not have the resources or incentives to sue employers for these types of antitrust violations since an antitrust standalone suit is usually much more costly than the individual damage compensation that may be awarded by the court (OECD, 2019[15]).

Mergers

Potentially colluding companies, however, could avoid unlawful labour market collusion by simply merging. If the merger substantially lessens or significantly impedes effective competition in a specific labour market, including by leading to the creation of a dominant employer, the merged entity would likely use its market power to reduce employment and wages in that market, similarly to what non‑merging colluding companies would do – see e.g. Hovenkamp and Marinescu (2019[41]). Antitrust authorities and courts have, however, usually paid relatively little attention to the effects of mergers in the labour market.52 Typically, the attention of antitrust authorities has been triggered only when the merger increased concentration above the threshold for high concentration. However, evidence suggests that horizontal mergers do not need to create dominant employers to have a significant effect in the labour market – see e.g. (Arnold, 2021[74]; Prager and Schmitt, 2021[77]) – and that undesirable effects of a merger may be induced also by very small increases in concentration (Nocke and Whinston, 2022[42]; Affeldt et al., 2021[43]).

In recent years, however, antitrust authorities, particularly in the United States, are reflecting on how to better incorporate the labour market in the analysis of mergers and on how to take a more proactive role (US Department of Justice; Federal Trade Commission, 2022[107]). Yet, one difficulty in assessing the impact of mergers on buyer power in the labour market has to do with the shortage of specific tools to analyse labour competition and, in particular, the difficulty of identifying the relevant market. Another difficulty is the evaluation of merger effects when merging firms are not direct competitors in downstream product markets – see OECD (2019[15]). This is an area in which more research is needed and more investment in developing adequate tools by governments and enforcement authorities would be welcome.

Collective bargaining and social dialogue

There is an older and sizeable literature on trade unions as counterbalancing the excess bargaining power with monopsony. Bilateral monopolies can typically yield efficient bargaining results (see for example Blair and Wang, (2015[150]) for a discussion.53 With the countervailing power of strong trade unions, concentration per se is unlikely to be sufficient to impose a situation of labour market monopsony – see e.g. MaCurdy and Pencavel (1986[151]) and Espinosa and Rhee (1989[152]). Sectoral collective agreements may also introduce a sectoral minimum wage which, if not too high, would make the relationship between marginal labour costs and employment less steep, thereby increasing both employment and wages (Ashenfelter, Farber and Ransom, 2010[7]) – see also the discussion on the anti-monopsonistic effects of the minimum wage in Section 3.4.2 below.

To the extent that the bargaining power of unions and collective bargaining coverage has faded over time in certain countries (OECD, 2019[153]), however, organised workers may not be as effective as before in exerting a sufficient counterbalancing force through social dialogue. Indeed, as mentioned above, OECD (2021[12]) finds that the elasticity of wages to concentration has increased over time. Benmelech et al. (2022[29]) also find the same pattern for the United States. The latter study also finds that this elasticity is higher where unionisation is lower. Similar results on the interaction between concentration and unionisation are found by Marinescu, Ouss and Pape (2021[34]) for France, and Abel, Tenreyro and Thwaites (2018[33]) for the United Kingdom. This literature tends to suggest that direct interventions to facilitate collective bargaining and social dialogue could have a strong impact on monopsony power – see OECD (2019[153]) for a discussion of policies to enhance collective bargaining and Chapter 5 for an application to working time issues.

Employers may, however, put in place organisational strategies to reduce the countervailing power of organised labour. For example, domestic or international outsourcing and franchising could be used as a way to split up one firm’s workforce and reduce co‑ordination among workers doing different jobs (OECD, 2021[65]). This could be particularly relevant in the case of firm-level bargaining, or in countries where firm-level employment thresholds trigger the possibility of collective action for employees.54 For example, franchising allocates workers into many separate legal entities (the franchisees), thereby preventing workers from co‑ordinating and bargaining together, even if these entities are de facto vertically integrated in the production and distribution structure of the franchisor (Callaci, 2018[154]).

3.4.2. Other policies

Minimum wages may counterbalance the adverse impact of monopsony

Minimum wages are often justified as a policy to reduce wage inequality and raise incomes at the bottom of the distribution – see e.g. Dube (2019[155]). Yet, in a standard model with competitive labour markets, the impact of the minimum wage on employment is unambiguously negative – see e.g. Brown (1999[156]) – which would make it difficult to attain its primary objective. Despite this theoretical prediction, the empirical evidence is much less conclusive and many studies have found no or small disemployment effects of minimum wage increases when it is maintained at moderate levels – see e.g. OECD (2015[157]) and the very complete recent survey of Dube (2019[158]).

Monopsony models provide a simple explanation for the lack of negative impact of moderate minimum wage hikes on employment (Manning, 2003[5]).55 The existing evidence therefore suggests the potential of the minimum wage to limit the negative effects of the employers’ market power on employment and wages.56 This conclusion is supported by three other pieces of evidence: Azar et al. (2019[159]) look at the impact of changes in minimum wages in the US retail sector using granular data on labour market concentration. They find that increases in the minimum wage significantly decrease employment of workers in low concentration markets while minimum wage‑induced employment changes become less negative as labour concentration increases, and are even estimated to be positive in the most highly concentrated markets. Popp (2021[83]) finds similar results as regards the impact of universally-binding but collectively negotiated sectoral minimum wages in Germany. Moreover, Johnson and Lipsitz (2022[160]) find that minimum wages have a less negative (more positive) impact on employment in low-wage occupations where non-compete agreements are more strictly enforced, thereby reducing more extensively the potential outside options for workers (see Section 3.4.1). These three results together confirm that where labour markets are more concentrated or outside options are artificially curbed, employers have greater market power and inefficiently reduce employment. By the same token, these results suggest that the objective of contrasting the negative effects of monopsony on labour market performance provide another justification for raising the minimum wage where it is too low, or introducing one where it does not exist, in particular when workers are not already covered by effective collective bargaining. Yet, as within the same country, the level of monopsony and concentration is heterogeneous (see Section 3.2.1), policy makers need to take into account that raising the minimum wage could hurt employment in some local labour markets while improving it in others.

Policies to promote remote work may make labour markets more competitive

Reducing barriers to geographic mobility beyond the worker’s travel-to-work area can expand the worker’s set of outside options. First-order tools in this sense are housing policies such as rental regulation, land-use and planning reforms, taxation on housing purchases, or investments in social housing (OECD, 2021[161]). Active labour market policies can also provide incentives for geographical mobility, by presenting jobseekers with opportunities that are not limited to their region of residence (OECD, 2005[162]). Caliendo, Mahlstedt, and Künn (2017[163]) exploit a natural experiment in Germany and find that a relocation subsidy for the unemployed also increases subsequent wages and job stability.57 Geographic mobility is further enhanced when workers’ qualifications and skills are recognised across regions, which relies on the existence of a national qualification framework, and of mechanisms for the recognition, validation or certification of workers’ prior learning (OECD, 2021[164]).

Policies to promote remote work can contribute to the same goal. There is increasing evidence that the upward trend in remote work made possible by telework – see e.g. OECD (2021[165]) is making labour markets more competitive. In the United States, the share of job applicants for local vacancies, who also apply to jobs in other labour markets, has risen steadily the past two years (Zhao, 2021[166]). Employers are also starting to report that remote work is increasing competition for workers who normally would not attract offers outside of their local labour market (Federal Reserve Board, 2021[167]).

If certain jobs could be practised fully remotely, workers could accept positions in a given occupation in the national (and even international) labour market, regardless of where the company is located. To assess the potential of teleworking to reduce labour market concentration, Figure 3.14 simulates average HHI if remote work were fully available for all vacancies in occupations for which telework is possible. The simulation assumes that for those occupations amenable to telework, all vacancies allow remote work, and the relevant geographical market for that occupation is therefore the entire country, rather than the region. As such, one should regard this as an upper bound on the efficacy of remote work to make labour markets more competitive. Occupations for which telework is possible are based on Basso et al. (2022[67]), who follow Dingel and Neiman (2020[69]), as in Section 3.2.3.

Figure 3.14. Teleworking may make labour markets less concentrated

On average across OECD countries in the sample, HHI would decline by a little over 20% if all vacancies allowed remote work, and workers could then search nationally within their occupation. The largest reductions were in Germany and Canada. In contrast, Singapore and Luxembourg see no effect of telework on HHI as both countries contain only one TL3 region.

Overall, these results suggests that incentivising full-time telework can make labour markets more competitive by enlarging outside options for workers. However, telework cannot be seen as a panacea – see OECD (2021[65]). In many countries the potential fall in concentration following a large swing towards teleworking would remain somewhat limited, and it may reinforce existing labour market inequalities. In addition, full-time telework may have other consequences on productivity and worker well-being that must be carefully assessed (see Chapter 5).

Skills may contribute to enlarge outside options policies

Training and skill policies can play an important role in enlarging outside options for workers. Workers who have retrained for a different occupation can search for jobs in an enlarged market represented by their origin occupation and the new occupation for which they have trained. This section presents a second simulation exercise, where workers are allowed to search for work not in a different geographical area, but in a different occupation from the one in which they are currently employed.

In the simulation, workers search for a new job in their occupation of origin or in one alternative occupation, for which they have trained. For a given occupation of origin, a destination occupation is only available after applying a strict set of criteria: a destination occupation should not be associated with significantly lower average wage or significantly higher educational attainment than in the origin occupation to ensure that the potential transition does not entail welfare losses. For the purpose of this simulation, a destination occupation is defined as that which minimises the retraining effort to do the transition, as long as this does not require more than 6 months of retraining.58 More details on the construction of the neighbourhoods of occupations are reported in Annex 3.C. A new HHI is then calculated for each TL3 region and jointly considering the postings of the occupation of origin and its associated destination. The analysis is performed for 2019 at the 3‑digit ISCO‑08 occupation level for Australia, Switzerland, the United Kingdom, the United States and all the European Union countries included in the main analysis.59

Figure 3.15 shows that allowing workers to search for employment not only in the occupation of origin but also in the occupation with the most similar skill bundle (within the limit of 6 months of retraining) has the potential to reduce aggregate labour market concentration by 18% on average across the OECD countries considered. For some occupations, the gain can be much larger than average, as shown in Annex Figure 3.A.6. For the occupations with a valid transition away from the occupation of origin, the average worker should retrain for 2.9 months to make the transition.60

These results suggest that reskilling and retraining policies can play a role in improving labour market conditions when markets are monopsonistic, but that their impact is limited by the extent of the retraining effort workers are willing or able to sustain in exchange for a gain in salary and other work amenities if moving to a less concentrated labour market. An important role is potentially played by career guidance counsellors. By providing individuals with information on skills requirements and on available retraining opportunities, they can help workers target training towards the most suitable alternative occupation. Crucially, career guidance for workers that seek to change jobs should be designed differently from that for the unemployed (OECD, 2021[168]).

Figure 3.15. Retraining can make labour markets somewhat less concentrated

The gains from training in terms higher salaries and better job quality should be assessed against the cost of retraining workers to switch occupations, which can be non-negligible (OECD, 2019[169]; Andrieu et al., 2019[170]). A review of successful adult learning reforms in Europe estimated the direct costs of delivering education and training to adults to reach EUR 200 to 2 500 per participant, while reforms that covered indirect costs of training (i.e. the salary of workers in training) were more expensive (OECD, 2020[171]). This emphasises the importance of deciding how the cost of training is shared between workers, employers and governments, which can vary a lot with the policy design (OECD, 2017[172]; OECD, 2019[173]).

Curbing employers’ monopsony power may reduce the provision of training by employers. Theoretical contributions by Acemoglu and Pischke (1998[174]; 1999[175]) show that firms are more likely to invest in general training when labour markets are monopsonistic, as workers cannot easily switch employer after training. Similarly, enhancing the enforceability of non-competes has increased firm-sponsored training in the United States (Starr, 2019[121]), even though, because of monopsony, it has simultaneously resulted in lower wages. Governments can intervene by designing programmes that strengthen the incentives of firms to invest in skills, from financial support tools (training subsidies and tax credits) to instruments that foster the co‑ordination of retraining programmes among firms with similar skill needs. For an extensive discussion of best policy practices to support firms’ supply of training – which falls beyond the scope of this chapter – see OECD, (2021[176]). More broadly, national skills policies should ensure that policy efforts to reduce monopsony do not impinge on the provision of firm-sponsored training.

3.5. Concluding remarks

This chapter reviews the measurement, consequences and policies concerning monopsonistic labour markets – markets where employers have power to set wages unilaterally. Whether due to job search frictions, differentiated workplace amenities or employer concentration, when employers find themselves with discretion in wage setting, theory predicts they will depress labour demand and wages in order to reap higher profits. In short, monopsony is likely to result in inefficiently low levels of employment and wages.

Using harmonised data, this chapter finds that a sizeable share of workers in OECD countries work in concentrated labour markets. On average, 16% of workers are in markets that are at least moderately concentrated and 10% are in highly concentrated markets, according to the conservative definition of concentrated markets used by US antitrust authorities. The distribution of workers across levels of market concentration is not even: workers in rural regions and those who worked on the front line during the COVID‑19 pandemic are more exposed to concentrated labour markets. Finally, the COVID‑19 crisis appears to have raised labour market concentration in many OECD labour markets, but the effect appears to be diminishing in the recovery. This is probably the result of fewer vacancies being posted during the peak of the pandemic by less resilient firms or firms affected by mandatory foreclosures. These evolutions in concentration may not have translated into actual changes in the wage‑setting power of employers, however.

Labour market concentration has a negative effect on job quality. Using matched employer-employee data, this chapter finds that, all other things equal, workers in more concentrated labour markets have reduced wages and job security. In addition, employers in concentrated labour markets can become choosier and turn to more skilled workers. Combined with a growing empirical literature on these issues, the results of this chapter suggest that monopsony is pervasive.

Monopsonistic labour markets can be considered a widespread market failure, and like most market failures, policy has an essential role to play in addressing them. Non-compete agreements between workers and firms, labour market collusion between employers, and often horizontal mergers, directly reduce workers’ outside options. Occupational licensing may also reduce outside options for workers in related, but non-licensed, occupations. By contrast, collective agreements and social dialogue can increase employment and raise wages by offsetting monopsony power by firms. In monopsonistic markets, minimum wages can also raise both employment and wages for workers at the bottom of the wage distribution, and could likewise be considered as a way of reducing the inefficiencies that result from monopsony.

These are just a few of the tried and tested policies for dealing with monopsony, but there are a wider suite of policies that require more consideration. For example, many OECD countries choose, to varying degrees, to give firms tax incentives to provide and pay for social spending such as health insurance, childcare and pensions. If firms are not strictly required to provide these benefits, and the benefits are not portable, this can make it difficult for employees to find suitable employers offering the same advantages. As a result, compared to a government programme that provides the social benefit directly to all employees, such social spending through employers may have the unintended consequence of making firms more differentiated and, therefore, making labour markets less competitive.

Similarly, all employment protection provisions that are increasing with job tenure – such as severance pay in most countries, see OECD (2020[99]) – can potentially tie the employee to his/her employer and yield ambiguous effects on market power.61 This argument also applies to all mandatory benefits and protections which increase with job tenure and are lost upon separation.

More generally, any programme that ties workers to firms will limit their outside options and likely reduce the competitiveness of labour markets. For example, some OECD countries have training programmes where a firm provides training for a worker and, in exchange, the worker is required to continue working for the firm for a particular length of time in order to “pay back” the cost of training. Leaving the firm before the end of this period is only possible at significant monetary cost to the worker. While better than a generic non-compete clause, such constraints on the worker’s mobility still enhance employer monopsony power. At the same time, there is an understanding that firms are more likely to invest in training when labour markets become monopsonistic, although the evidence is not clear-cut. More empirical research is, therefore, needed before drawing any policy conclusions.

In addition, it is not uncommon to find countries that require a mandatory notice period in the case of voluntary quits. There may be good reasons for such notice periods but, at the margin, they may make employers more reluctant to hire if they lead to long delays before a new recruit can actually start work. Hence, extended notice periods may also reduce workers’ outside options.

In addition to policy research, there are still some basic economic questions surrounding monopsony that require further attention. Theoretically, monopsonistic labour markets should result in inefficiently low employment, and wages marked down from the level which would have prevailed in competitive markets. The evidence in this chapter and the previous empirical literature confirms the theory that monopsonistic labour markets (at least as measured by concentration) result in lower wages. However, direct empirical proof for lower employment resulting from monopsony generally, and concentration specifically, is less abundant, often due to data limitations. More research is therefore needed on this issue. More generally, national statistical offices could invest to make available to researchers and policy makers exhaustive linked employer-employee data that could be used to, among others, track the evolution of concentration and its effects over time.

Lastly, this chapter did not examine the fact that employer power can also manifest itself in other input markets, as well as in output markets. Increasing levels of product market power have also been associated with lower demand for labour, lower labour force participation and lower wages – see e.g. De Loecker, Eeckhout and Unger (2020[177]). Future research could therefore explore how product and labour market power can co‑exist and interact to affect labour market outcomes.

References

[33] Abel, W., S. Tenreyro and G. Thwaites (2018), “Monopsony in the UK”, CEPR Discussion Paper, No. 13265, CEPR, London.

[175] Acemoglu, D. and J. Pischke (1999), “Beyond Becker: Training in Imperfect Labour Markets”, The Economic Journal, Vol. 109/453, pp. 112-142, https://doi.org/10.1111/1468-0297.00405 (accessed on 6 October 2021).

[174] Acemoglu, D. and J. Pischke (1998), “Why Do Firms Train? Theory and Evidence”, The Quarterly Journal of Economics, Vol. 113/1, pp. 79-119, https://www.jstor.org/stable/2586986 (accessed on 6 October 2021).

[43] Affeldt, P. et al. (2021), “Assessing EU Merger Control through Compensating Efficiencies”, CEPR Discussion Paper, No. 16705, CEPR, London, https://cepr.org/active/publications/discussion_papers/dp.php?dpno=16705.

[98] Albanese, A. and G. Gallo (2020), “Buy flexible, pay more: The role of temporary contracts on wage inequality”, Labour Economics, Vol. 64, p. 101814, https://doi.org/10.1016/j.labeco.2020.101814.

[61] Andersson, F., S. Burgess and J. Lane (2007), “Cities, matching and the productivity gains of agglomeration”, Journal of Urban Economics, Vol. 61/1, pp. 112-128, https://doi.org/10.1016/j.jue.2006.06.005.

[170] Andrieu, E. et al. (2019), “Occupational transitions: The cost of moving to a “safe haven””, OECD Science, Technology and Industry Policy Papers, No. 61, OECD Publishing, Paris, https://doi.org/10.1787/6d3f9bff-en.

[74] Arnold, D. (2021), Mergers and Acquisitions, Local Labor Market Concentration, and Worker Outcomes, mimeo, https://darnold199.github.io/jmp.pdf (accessed on 7 October 2021).

[50] Ascheri, A. et al. (2021), “Competition in urban hiring markets: evidence from online job advertisements”, Eurostat Statistical Working Papers.

[7] Ashenfelter, O., H. Farber and M. Ransom (2010), “Labor Market Monopsony”, Journal of Labor Economics, Vol. 28/2, pp. 203-210, https://doi.org/10.1086/653654.

[145] Asker, J. and V. Nocke (2021), “Collusion, mergers, and related antitrust issues”, in Ho, K., A. Hortaçsu and A. Lizzeri (eds.), Handbook of Industrial Organization, Elsevier, https://doi.org/10.1016/bs.hesind.2021.11.012.

[131] Autio, R. (2021), Finland: Change to the Employment Contracts Act concerning non-competition agreements from 1 January 2022, DLA Piper Monitor, https://knowledge.dlapiper.com/dlapiperknowledge/globalemploymentlatestdevelopments/finland-change-to-the-employment-contracts-act-concerning-non-competition-agreements-from-1-january-2022#page=1 (accessed on 1 February 2022).

[86] Autor, D., D. Dorn and G. Hanson (2013), “The China Syndrome: Local Labor Market Effects of Import Competition in the United States”, American Economic Review, Vol. 103/6, pp. 2121-2168, https://doi.org/10.1257/aer.103.6.2121.

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

[52] Azar, J., I. Marinescu and M. Steinbaum (2022), “Labor Market Concentration”, Journal of Human Resources, Vol. 57, pp. S167-S199, https://doi.org/10.3368/jhr.monopsony.1218-9914r1.

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

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

[57] Bassanini, A., C. Batut and E. Caroli (2021), “Labor Market Concentration and Stayers’ Wages: Evidence from France”, IZA Discussion Papers, No. 14912, IZA, Bonn, https://docs.iza.org/dp14912.pdf.

[37] Bassanini, A. et al. (2022), “Labour Market Concentration, Wages and Job Security in Europe”, IZA Discussion Papers, No. 15231, IZA, Bonn, https://docs.iza.org/dp15231.pdf.

[97] Bassanini, A. et al. (2013), “Working in Family Firms: Paid Less but More Secure? Evidence from French Matched Employer-Employee Data”, Industrial and Labor Relations Review, Vol. 66/2, pp. 433-466, https://doi.org/10.1177/001979391306600206.

[67] Basso, G. et al. (2022), “Unsafe Jobs, Labour Market Risk and Social Protection”, Economic Policy, https://doi.org/10.1093/epolic/eiac004.

[184] Bechichi, N. et al. (2018), “Moving between jobs: An analysis of occupation distances and skill needs”, OECD Science, Technology and Industry Policy Papers, No. 52, OECD Publishing, Paris, https://doi.org/10.1787/d35017ee-en.

[185] Bechichi, N. et al. (2019), “Occupational mobility, skills and training needs”, OECD Science, Technology and Industry Policy Papers, No. 70, OECD Publishing, Paris, https://doi.org/10.1787/30a12738-en.

[29] Benmelech, E., N. Bergman and H. Kim (2022), “Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?”, Journal of Human Resources, Vol. 57, pp. S200-S250, https://doi.org/10.3368/jhr.monopsony.0119-10007r1.

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

[130] Berjot, X. (2021), Indemnité de non-concurrence: Pas de réduction par le juge, https://www.village-justice.com/articles/indemnite-non-concurrence-pas-reduction-par-juge,40606.html (accessed on 31 January 2022).

[8] Blair, R. and J. Harrison (2010), Monopsony in Law and Economics, Cambridge University Press, Cambridge, https://doi.org/10.1017/cbo9780511778766.

[143] Blair, R. and W. Wang (2017), “Buyer cartels and private enforcement of antitrust policy”, Managerial and Decision Economics, Vol. 38/8, pp. 1185-1193, https://doi.org/10.1002/mde.2857.

[150] Blair, R. and W. Wang (2015), “Bilateral Monopoly, Two-Sided Markets, and the E-Books Conspiracy”, University of Miami Law Review, Vol. 69/Caveat 7, https://lawreview.law.miami.edu/bilateral-monopoly-two-sided-markets-e-books-conspiracy/ (accessed on 8 February 2022).

[62] Bleakley, H. and J. Lin (2012), “Thick-market effects and churning in the labor market: Evidence from US cities”, Journal of Urban Economics, Vol. 72/2-3, pp. 87-103, https://doi.org/10.1016/j.jue.2012.04.003.

[6] Boal, W. and M. Ransom (1997), “Monopsony in the Labor Market”, Journal of Economic Literature, Vol. 35/1, pp. 86-112, https://www.jstor.org/stable/2729694.

[188] Bowley, A. (1928), “Bilateral Monopoly”, The Economic Journal, Vol. 38/152, pp. 651-659, https://doi.org/10.2307/2224123.

[156] Brown, C. (1999), “Minimum wages, employment, and the distribution of income”, in Handbook of Labor Economics, Elsevier, https://doi.org/10.1016/s1573-4463(99)30018-3.

[9] Burdett, K. and D. Mortensen (1998), “Wage Differentials, Employer Size, and Unemployment”, International Economic Review, Vol. 39/2, pp. 257-273, https://www.jstor.org/stable/2527292.

[24] Cahuc, P., S. Carcillo and A. Zylberberg (2014), Labor Economics, MIT Press, Cambridge, MA.

[16] Caldwell, S. and E. Oehlsen (2018), Monopsony and the Gender Wage Gap: Experimental Evidence from the Gig Economy.

[163] Caliendo, M., S. Künn and R. Mahlstedt (2017), “The return to labor market mobility: An evaluation of relocation assistance for the unemployed”, Journal of Public Economics, Vol. 148, pp. 136-151, https://doi.org/10.1016/J.JPUBECO.2017.02.008.

[154] Callaci, B. (2018), “Control without responsibility: The legal creation of franchising 1960-1980”, Working Paper Series, No. December 2018, Washington Center for Equitable Growth, https://equitablegrowth.org/working-papers/control-without-responsibility-the-legal-creation-of-franchising-1960-1980/ (accessed on 25 June 2019).

[180] Cammeraat, E. and M. Squicciarini (2021), “Burning Glass Technologies’ data use in policy-relevant analysis: An occupation-level assessment”, OECD Science, Technology and Industry Working Papers, No. 2021/05, OECD Publishing, Paris, https://doi.org/10.1787/cd75c3e7-en.

[10] Card, D. et al. (2018), “Firms and Labor Market Inequality: Evidence and Some Theory”, Journal of Labor Economics, Vol. 36/S1, pp. S13-S70, https://doi.org/10.1086/694153.

[110] Colvin, A. and H. Shierholz (2019), Noncompete agreements: Ubiquitous, harmful to wages and to competition, and part of a growing trend of employers requiring workers to sign away their rights, Economic Policy Institute, https://www.epi.org/publication/noncompete-agreements/ (accessed on 31 January 2022).

[71] Conyon, M. et al. (2001), “Do hostile mergers destroy jobs?”, Journal of Economic Behavior & Organization, Vol. 45/4, pp. 427-440, https://doi.org/10.1016/s0167-2681(01)00155-x.

[76] Currie, J., M. Farsi and W. Macleod (2005), “Cut to the Bone? Hospital Takeovers and Nurse Employment Contracts”, Industrial and Labor Relations Review, Vol. 58/3, pp. 471-493, https://doi.org/10.1177/001979390505800309.

[63] Dauth, W. et al. (2018), Matching in Cities, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w25227.

[177] De Loecker, J., J. Eeckhout and G. Unger (2020), “The Rise of Market Power and the Macroeconomic Implications”, The Quarterly Journal of Economics, Vol. 135/2, pp. 561-644, https://doi.org/10.1093/qje/qjz041.

[104] Deming, D. (2017), “The Growing Importance of Social Skills in the Labor Market”, The Quarterly Journal of Economics, Vol. 132/4, pp. 1593-1640, https://doi.org/10.1093/QJE/QJX022.

[103] Deming, D. and L. Kahn (2018), “Skill requirements across firms and labor markets: Evidence from job postings for professionals”, Journal of Labor Economics, Vol. 36/S1, pp. S337-S369, https://doi.org/10.1086/694106/SUPPL_FILE/16202DATA.ZIP.

[49] Dijkstra, L., H. Poelman and P. Veneri (2019), The EU-OECD definition of a functional urban area, https://doi.org/10.1787/d58cb34d-en.

[69] Dingel, J. and B. Neiman (2020), “How many jobs can be done at home?”, Journal of Public Economics, Vol. 189, p. 104235, https://doi.org/10.1016/j.jpubeco.2020.104235.

[141] Dodini, S. (2020), “Occupational Licensing, Skills, and Labor Market Spillovers”, Mimeo, Cornell University, https://samueldodini.com/files/Dodini_occupation_licensing_9_21_2020.pdf (accessed on 3 December 2021).

[36] Dodini, S. et al. (2020), “Monopsony, Skills, and Labor Market Concentration”, CEPR Discussion Paper, No. DP15412, CEPR, London.

[68] DOL (2022), Bearing the Cost: How Overrepresentation in Undervalued Jobs Disadvantaged Women during the Pandemic, The United States Department of Labor , Washington, D.C., https://www.dol.gov/sites/dolgov/files/WB/media/BearingTheCostReport.pdf (accessed on 27 April 2022).

[158] Dube, A. (2019), Impacts of minimum wages : review of the international evidence, HM Treasury and Department for Business, Energy & Industrial Strategy, https://www.gov.uk/government/publications/impacts-of-minimum-wages-review-of-the-international-evidence (accessed on 7 January 2022).

[155] Dube, A. (2019), “Minimum Wages and the Distribution of Family Incomes”, American Economic Journal: Applied Economics, Vol. 11/4, pp. 268-304, https://doi.org/10.1257/app.20170085.

[14] Dube, A. et al. (2020), “Monopsony in Online Labor Markets”, American Economic Review: Insights, Vol. 2/1, pp. 33-46, https://doi.org/10.1257/aeri.20180150.

[146] Dyck, A., A. Morse and L. Zingales (2010), “Who Blows the Whistle on Corporate Fraud?”, The Journal of Finance, Vol. 65/6, pp. 2213-2253, https://doi.org/10.1111/j.1540-6261.2010.01614.x.

[152] Espinosa, M. and C. Rhee (1989), “Efficient Wage Bargaining as a Repeated Game”, The Quarterly Journal of Economics, Vol. 104/3, p. 565, https://doi.org/10.2307/2937811.

[189] European Commission (2003), Council Regulation on the control of concentrations between undertakings (2004/C 31/03), European Commission, Brussels.

[64] 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://doi.org/10.1787/b902cc00-en.

[167] Federal Reserve Board (2021), The Beige Book: Summary of Commentary on Current Economic Conditions, The Federal Reserve Bank of Chicago, https://www.federalreserve.gov/monetarypolicy/files/BeigeBook_20211201.pdf (accessed on 6 February 2022).

[38] Foote, A., M. Kutzbach and L. Vilhuber (2021), “Recalculating ... : How Uncertainty in Local Labour Market Definitions Affects Empirical Findings”, Applied Economics, Vol. 53/14, pp. 1598-1612, https://doi.org/10.1080/00036846.2020.1841083.

[54] Gathmann, C. and U. Schönberg (2010), “How General Is Human Capital? A Task‐Based Approach”, Journal of Labor Economics, Vol. 28/1, pp. 1-49, https://doi.org/10.1086/649786.

[22] Gerard, F. et al. (2021), “Assortative Matching or Exclusionary Hiring? The Impact of Employment and Pay Policies on Racial Wage Differences in Brazil”, American Economic Review, Vol. 111/10, pp. 3418-57, https://doi.org/10.1257/AER.20181596.

[56] Gibson, M. (2021), “Employer Market Power in Silicon Valley”, IZA Discussion Papers, No. 14843, IZA, Bonn, https://www.iza.org/de/publications/dp/14843/employer-market-power-in-silicon-valley.

[138] Gittleman, M., M. Klee and M. Kleiner (2017), “Analyzing the Labor Market Outcomes of Occupational Licensing”, Industrial Relations: A Journal of Economy and Society, Vol. 57/1, pp. 57-100, https://doi.org/10.1111/irel.12200.

[21] Goldin, C. (2014), “A Grand Gender Convergence: Its Last Chapter”, American Economic Review, Vol. 104/4, pp. 1091-1119, https://doi.org/10.1257/aer.104.4.1091.

[178] Goos, M., A. Manning and A. Salomons (2014), “Explaining Job Polarization: Routine-Biased Technological Change and Offshoring”, American Economic Review, Vol. 104/8, pp. 2509-2526, https://doi.org/10.1257/aer.104.8.2509.

[92] Grigsby, J., E. Hurst and A. Yildirmaz (2021), “Aggregate Nominal Wage Adjustments: New Evidence from Administrative Payroll Data”, American Economic Review, Vol. 111/2, pp. 428-471, https://doi.org/10.1257/aer.20190318.

[75] Gugler, K. and B. Yurtoglu (2004), “The effects of mergers on company employment in the USA and Europe”, International Journal of Industrial Organization, Vol. 22/4, pp. 481-502, https://doi.org/10.1016/j.ijindorg.2003.12.003.

[90] Haefke, C., M. Sonntag and T. van Rens (2013), “Wage rigidity and job creation”, Journal of Monetary Economics, Vol. 60/8, pp. 887-899, https://doi.org/10.1016/j.jmoneco.2013.09.003.

[79] Hafner, F. (2021), “Labor Market Competition, Wages and Worker Mobility”, Mimeo, Aalto University, https://papers.flaviohafner.com/cwm.pdf (accessed on 5 December 2021).

[93] Hahn, J., H. Hyatt and H. Janicki (2021), “Job ladders and growth in earnings, hours, and wages”, European Economic Review, Vol. 133, p. 103654, https://doi.org/10.1016/j.euroecorev.2021.103654.

[2] Haltiwanger, J., H. Hyatt and E. McEntarfer (2018), “Who Moves Up the Job Ladder?”, Journal of Labor Economics, Vol. 36/S1, pp. S301-S336, https://doi.org/10.1086/694417.

[84] Hausman, Leonard and Zona (1994), “Competitive Analysis with Differenciated Products”, Annales d’Économie et de Statistique 34, p. 159, https://doi.org/10.2307/20075951.

[118] Hausman, N. and K. Lavetti (2021), “Physician Practice Organization and Negotiated Prices: Evidence from State Law Changes”, American Economic Journal: Applied Economics, Vol. 13/2, pp. 258-296, https://doi.org/10.1257/app.20180078.

[78] Hensvik, L. (2012), “Competition, Wages and Teacher Sorting: Lessons Learned from a Voucher Reform”, The Economic Journal, Vol. 122/561, pp. 799-824, https://doi.org/10.1111/j.1468-0297.2012.02514.x.

[133] Hermansen, M. (2019), “Occupational licensing and job mobility in the United States”, OECD Economics Department Working Papers, No. 1585, OECD Publishing, Paris, https://doi.org/10.1787/4cc19056-en.

[44] Hershbein, B. and L. Kahn (2018), “Do Recessions Accelerate Routine-Biased Technological Change? Evidence from Vacancy Postings”, American Economic Review, Vol. 108/7, pp. 1737-1772, https://doi.org/10.1257/aer.20161570.

[94] Hijzen, A., W. Zwysen and M. Lillehagen (2021), “Job mobility, reallocation and wage growth: A tale of two countries”, OECD Social, Employment and Migration Working Papers, No. 254, OECD Publishing, Paris, https://doi.org/10.1787/807becdf-en.

[87] Hirsch, B., T. Schank and C. Schnabel (2010), “Differences in Labor Supply to Monopsonistic Firms and the Gender Pay Gap: An Empirical Analysis Using Linked Employer‐Employee Data from Germany”, Journal of Labor Economics, Vol. 28/2, pp. 291-330, https://doi.org/10.1086/651208.

[41] Hovenkamp, H. and I. Marinescu (2019), “Anticompetitive Mergers in Labor Markets”, Indiana Law Journal, Vol. 94/3, pp. 1031-1063, https://doi.org/10.2139/ssrn.3124483.

[20] Jacob, N. et al. (2019), “The disutility of commuting? The effect of gender and local labor markets”, Regional Science and Urban Economics, Vol. 77, pp. 264-275, https://doi.org/10.1016/j.regsciurbeco.2019.06.001.

[26] Jarosch, G., J. Nimczik and I. Sorkin (2019), Granular Search, Market Structure, and Wages, National Bureau of Economic Research, Cambridge, MA, R&R Review of Economic Studies, https://doi.org/10.3386/w26239.

[117] Jeffers, J. (2021), “The Impact of Restricting Labor Mobility on Corporate Investment and Entrepreneurship”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3040393.

[125] Johnson, M., K. Lavetti and M. Lipsitz (2021), “The Labor Market Effects of Legal Restrictions on Worker Mobility”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3455381.

[160] Johnson, M. and M. Lipsitz (2022), “Why are Low-Wage Workers Signing Noncompete Agreements?”, Journal of Human Resources, Vol. 57/3, pp. 689-724, https://doi.org/10.3368/jhr.57.3.0619-10274r2.

[53] Kambourov, G. and I. Manovskii (2009), “Occupational Specificity of Human Capital”, International Economic Review, Vol. 50/1, pp. 63-115, https://doi.org/10.1111/j.1468-2354.2008.00524.x.

[96] Kesternich, I. et al. (2021), “Money or meaning? Labor supply responses to work meaning of employed and unemployed individuals”, European Economic Review, Vol. 137, p. 103786, https://doi.org/10.1016/j.euroecorev.2021.103786.

[136] Kleiner, M. (2017), “The influence of occupational licensing and regulation”, IZA World of Labor, https://doi.org/10.15185/izawol.392.

[137] Kleiner, M. and A. Krueger (2013), “Analyzing the Extent and Influence of Occupational Licensing on the Labor Market”, Journal of Labor Economics, Vol. 31/S1, pp. S173-S202, https://doi.org/10.1086/669060.

[140] Kleiner, M. et al. (2016), “Relaxing Occupational Licensing Requirements: Analyzing Wages and Prices for a Medical Service”, The Journal of Law and Economics, Vol. 59/2, pp. 261-291, https://doi.org/10.1086/688093.

[142] Kleiner, M. and K. Park (2010), Battles Among Licensed Occupations: Analyzing Government Regulations on Labor Market Outcomes for Dentists and Hygienists, National Bureau of Economic Research, Cambridge, MA, https://doi.org/10.3386/w16560.

[40] Koh, L. (2013), In re High-Tech Employee Antitrust Litigation, 985 F. Supp. 2d 1167, United States District Court, N.D. California, San Jose Division, https://casetext.com/case/in-re-high-tech-emp-antitrust-litig-9 (accessed on 10 February 2022).

[182] Korbel, P. (2018), “Internet Job Postings: Preliminary Skills Analysis”, NCVER Technical Paper, National Centre for Vocational Education Research, https://www.ncver.edu.au/__data/assets/pdf_file/0023/2931440/Internet-job-postings-preliminary-skills-analysis-technical-paper.pdf (accessed on 14 April 2022).

[132] Koumenta, M. and M. Pagliero (2018), “Occupational Regulation in the European Union: Coverage and Wage Effects”, British Journal of Industrial Relations, Vol. 57/4, pp. 818-849, https://doi.org/10.1111/bjir.12441.

[144] Krueger, A. and O. Ashenfelter (2022), “Theory and Evidence on Employer Collusion in the Franchise Sector”, Journal of Human Resources, Vol. 57 (S), pp. S324-S348, https://doi.org/10.3368/jhr.monopsony.1019-10483.

[127] Krueger, A. and E. Posner (2018), A Proposal for Protecting Low-Income Workers from Monopsony and Collusion, Brookings, Washington, D.C., http://www.hamiltonproject.org/assets/files/protecting_low_income_workers_from_monopsony_collusion_krueger_posner_pp.pdf (accessed on 25 October 2018).

[91] Kudlyak, M. (2014), “The cyclicality of the user cost of labor”, Journal of Monetary Economics, Vol. 68, pp. 53-67, https://doi.org/10.1016/j.jmoneco.2014.07.007.

[23] Lang, K. and J. Lehmann (2012), “Racial Discrimination in the Labor Market: Theory and Empirics”, Journal of Economic Literature, Vol. 50/4, pp. 959-1006, https://doi.org/10.1257/jel.50.4.959.

[105] Lassébie, J. et al. (2021), “Speaking the same language: A machine learning approach to classify skills in Burning Glass Technologies data”, OECD Social, Employment and Migration Working Papers, No. 263, OECD Publishing, Paris, https://doi.org/10.1787/adb03746-en.

[116] Lavetti, K. (2021), “Noncompete agreements in employment contract”, IZA World of Labor, https://doi.org/10.15185/izawol.486.

[124] Lavetti, K., C. Simon and W. White (2019), “The Impacts of Restricting Mobility of Skilled Service Workers”, Journal of Human Resources, Vol. 55/3, pp. 1025-1067, https://doi.org/10.3368/jhr.55.3.0617-8840r5.

[19] Le Barbanchon, T., R. Rathelot and A. Roulet (2020), “Gender Differences in Job Search: Trading off Commute Against Wage”, The Quarterly Journal of Economics, Vol. 136/1, pp. 381-426, https://doi.org/10.1093/qje/qjaa033.

[72] Lehto, E. and P. Böckerman (2008), “Analysing the employment effects of mergers and acquisitions”, Journal of Economic Behavior & Organization, Vol. 68/1, pp. 112-124, https://doi.org/10.1016/j.jebo.2008.03.002.

[186] Lewi, T. et al. (2021), “Recent Federal and State Laws Restrict Use of Employee Non-Competition Agreements by Government Contractors and Other Employers | Inside Government Contracts”, Inside Government Contracts, Covington, https://www.insidegovernmentcontracts.com/2021/08/recent-federal-and-state-laws-restrict-use-of-employee-non-competition-agreements-by-government-contractors-and-other-employers/ (accessed on 1 February 2022).

[70] Lichtenberg, F. and D. Siegel (1990), “The Effect of Ownership Changes on the Employment and Wages of Central Office and Other Personnel”, The Journal of Law & Economics, Vol. 33/2, pp. 383-408, http://www.jstor.org/stable/725369.

[123] Lipsitz, M. and E. Starr (2022), “Low-Wage Workers and the Enforceability of Noncompete Agreements”, Management Science, Vol. 68/1, pp. 143-170, https://doi.org/10.1287/mnsc.2020.3918.

[148] Luz, R. and G. Spagnolo (2017), “Leniency, collusion, corruption, and whistleblowing”, Journal of Competition Law & Economics, Vol. 13/4, pp. 729-766, https://doi.org/10.1093/joclec/nhx025.

[101] Macaluso, C., B. Hershbein and C. Yeh (2019), “Concentration in U.S. local labor markets: evidence from vacancy and employment data”, Meeting Papers, No. 1336, Society for Economic Dynamics.

[151] MaCurdy, T. and J. Pencavel (1986), “Testing between Competing Models of Wage and Employment Determination in Unionized Markets”, Journal of Political Economy, Vol. 94/3, Part 2, pp. S3-S39, https://doi.org/10.1086/261398.

[25] Manning, A. (2020), “Monopsony in Labor Markets: A Review”, Industrial and Labor Relations Review, Vol. 74/1, pp. 3-26, https://doi.org/10.1177/0019793920922499.

[5] Manning, A. (2003), Monopsony in Motion: Imperfect Competition in Labor Markets, Princeton University Press.

[34] Marinescu, I., I. Ouss and L. Pape (2021), “Wages, hires, and labor market concentration”, Journal of Economic Behavior & Organization, Vol. 184, pp. 506-605, https://doi.org/10.1016/j.jebo.2021.01.033.

[35] Martins, P. (2018), “Making their own weather? Estimating employer labour-market power and its wage effects”, CGR Working Papers, No. 95, Queen Mary University, London.

[120] Marx, M., D. Strumsky and L. Fleming (2009), “Mobility, Skills, and the Michigan Non-Compete Experiment”, Management Science, Vol. 55/6, pp. 875-889, https://doi.org/10.1287/mnsc.1080.0985.

[17] Mas, A. and A. Pallais (2017), “Valuing Alternative Work Arrangements”, American Economic Review, Vol. 107/12, pp. 3722-3759, https://doi.org/10.1257/aer.20161500.

[122] McAdams, J. (2019), “Non-Compete Agreements: A Review of the Literature”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3513639.

[126] Meritas (2017), Employee Non-Compete Agreements in Europe, Middle East and Africa, Meritas, Minneapolis, Minn., http://www.meritas.org.

[102] Modestino, A., D. Shoag and J. Balance (2020), “Upskilling: Do employers demand greater skill when workers are plentiful?”, Review of Economics and Statistics, Vol. 102/4, pp. 793-805, https://doi.org/10.1162/REST_A_00835.

[48] Modestino, A., D. Shoag and J. Ballance (2016), “Downskilling: changes in employer skill requirements over the business cycle”, Labour Economics, Vol. 41, pp. 333-347, https://doi.org/10.1016/J.LABECO.2016.05.010.

[46] Naidu, S., E. Posner and G. Weyl (2018), “Antitrust Remedies for Labour Market Power”, Harvard Law Review, Vol. 132/2, pp. 536-601, https://doi.org/10.2139/ssrn.3129221.

[85] Nevo, A. (2001), “Measuring Market Power in the Ready-to-Eat Cereal Industry”, Econometrica, Vol. 69/2, pp. 307-342, https://doi.org/10.1111/1468-0262.00194.

[193] Nicandri, A. (2011), “The Growing Disfavor of Non-Compete Agreements in the New Economy and Alternative Approaches for Protecting Empoyers’ Proprietary Information and Trade Secrets”, University of Pennsylvania Journal of Business Law, Vol. 13/4, https://scholarship.law.upenn.edu/jbl/vol13/iss4/6.

[42] Nocke, V. and M. Whinston (2022), “Concentration Thresholds for Horizontal Mergers”, American Economic Review, Vol. 112/6, pp. 1915-1948, https://doi.org/10.1257/aer.20201038.

[161] OECD (2021), Brick by Brick: Building Better Housing Policies, OECD Publishing, Paris, https://doi.org/10.1787/b453b043-en.

[168] OECD (2021), Career Guidance for Adults in a Changing World of Work, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/9a94bfad-en.

[192] OECD (2021), FDI Qualities Policy Toolkit: Consultation paper for 6th FDI Qualities Policy Network Meeting, 16 November 2021, OECD Publishing, Paris, https://www.oecd.org/daf/inv/investment-policy/FDI-Qualities-Policy-Toolkit-Consultation-Paper-2021.pdf.

[176] OECD (2021), Incentives for SMEs to Invest in Skills: Lessons from European Good Practices, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/1eb16dc7-en.

[65] OECD (2021), “Labour market developments: The unfolding COVID‑19 crisis”, in OECD Employment Outlook 2021: Navigating the COVID-19 Crisis and Recovery, OECD Publishing, Paris, https://doi.org/10.1787/7e1e1ad3-en.

[45] OECD (2021), “Monopoly’s neglected twin? The effect of labour market concentration on wages and inequality”, in The Role of Firms in Wage Inequality: Policy Lessons from a Large Scale Cross-Country Study, OECD Publishing, Paris, https://doi.org/10.1787/fd80057f-en.

[165] OECD (2021), OECD Skills Outlook 2021: Learning for Life, OECD Publishing, Paris, https://doi.org/10.1787/0ae365b4-en.

[51] OECD (2021), “OECD Territorial Grids”, OECD, Paris, https://www.oecd.org/regional/regional-statistics/territorial-grid.pdf.

[100] OECD (2021), The New OECD Employment Protection Legislation Indicators for Temporary Contracts, OECD, Paris, https://www.oecd.org/els/emp/OECD-EPLIndicators-TemporaryContracts.pdf.

[164] OECD (2021), “The recognition of prior learning in adult basic education”, OECD, Paris, https://www.oecd.org/els/emp/skills-and-work/adult-learning/Prior_learning.pdf.

[12] OECD (2021), The Role of Firms in Wage Inequality: Policy Lessons from a Large Scale Cross-Country Study, OECD Publishing, Paris, https://doi.org/10.1787/7d9b2208-en.

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

[66] OECD (2020), “COVID‑19: From a health to a jobs crisis”, in OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis, OECD Publishing, Paris, https://doi.org/10.1787/cea3b4f4-en.

[171] OECD (2020), Increasing Adult Learning Participation: Learning from Successful Reforms, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/cf5d9c21-en.

[99] OECD (2020), OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis, OECD Publishing, Paris, https://doi.org/10.1787/1686c758-en.

[59] OECD (2020), “What is happening to middle-skill workers?”, in OECD Employment Outlook 2020: Worker Security and the COVID-19 Crisis, OECD Publishing, Paris, https://doi.org/10.1787/c9d28c24-en.

[173] OECD (2019), Getting Skills Right: Future-Ready Adult Learning Systems, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/9789264311756-en.

[153] OECD (2019), Negotiating Our Way Up: Collective Bargaining in a Changing World of Work, OECD Publishing, Paris, https://doi.org/10.1787/1fd2da34-en.

[15] OECD (2019), OECD Employment Outlook 2019: The Future of Work, OECD Publishing, Paris, https://doi.org/10.1787/9ee00155-en.

[169] OECD (2019), OECD Skills Outlook 2019 : Thriving in a Digital World, OECD Publishing, Paris, https://doi.org/10.1787/df80bc12-en.

[81] 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.

[172] OECD (2017), Financial Incentives for Steering Education and Training, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/9789264272415-en.

[58] OECD (2017), “How technology and globalisation are transforming the labour market”, in OECD Employment Outlook 2017, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2017-7-en.

[39] OECD (2016), OECD Regions at a Glance 2016, OECD Publishing, Paris, https://doi.org/10.1787/reg_glance-2016-en.

[157] OECD (2015), OECD Employment Outlook 2015, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2015-en.

[191] OECD (2014), OECD Employment Outlook 2014, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2014-en.

[183] OECD (2013), OECD Employment Outlook 2013, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2013-en.

[82] OECD (2012), Redefining “Urban”: A New Way to Measure Metropolitan Areas, OECD Publishing, Paris, https://doi.org/10.1787/9789264174108-en.

[162] OECD (2005), OECD Employment Outlook 2005, OECD Publishing, Paris, https://doi.org/10.1787/empl_outlook-2005-en.

[134] Pagliero, M. (2011), “What is the objective of professional licensing? Evidence from the US market for lawyers”, International Journal of Industrial Organization, Vol. 29/4, pp. 473-483, https://doi.org/10.1016/j.ijindorg.2010.09.002.

[60] Petrongolo, B. and C. Pissarides (2006), “Scale Effects in Markets with Search”, The Economic Journal, Vol. 116/508, pp. 21-44, https://doi.org/10.1111/j.1468-0297.2006.01046.x.

[89] Pissarides, C. (2009), “The Unemployment Volatility Puzzle: Is Wage Stickiness the Answer?”, Econometrica, Vol. 77/5, pp. 1339-1369, https://doi.org/10.3982/ecta7562.

[83] Popp, M. (2021), “Minimum Wages in Concentrated Labor Markets”, BGPE Discussion Paper, http://www.bgpe.de/texte/DP/214_Popp.pdf (accessed on 18 January 2022).

[77] Prager, E. and M. Schmitt (2021), “Employer Consolidation and Wages: Evidence from Hospitals”, American Economic Review, Vol. 111/2, pp. 397-427, https://doi.org/10.1257/aer.20190690.

[31] Qiu, Y. and A. Sojourner (2019), “Labor-Market Concentration and Labor Compensation”, IZA Discussion Papers, IZA.

[55] Redding, S. and E. Rossi-Hansberg (2017), “Quantitative Spatial Economics”, Annual Review of Economics, Vol. 9, pp. 21-58.

[32] Rinz, K. (2022), “Labor Market Concentration, Earnings, and Inequality”, Journal of Human Resources, Vol. 57, pp. S251-S283, https://doi.org/10.3368/jhr.monopsony.0219-10025R1.

[187] Robinson, C. (2018), “Occupational mobility, occupation distance, and specific human capital”, Journal of Human Resources, Vol. 53/2, pp. 513-551, https://doi.org/10.3368/JHR.53.2.0814-6556R2.

[190] Schmutz, B. and M. Sidibé (2019), “Frictional Labour Mobility”, The Review of Economic Studies, Vol. 86/4, pp. 1779-1826, https://doi.org/10.1093/RESTUD/RDY056.

[179] Schubert, G., A. Stansbury and B. Taska (2021), Employer Concentration and Outside Options, mimeo, https://doi.org/10.2139/SSRN.3599454.

[115] Shi, L. (2020), “The Macro Impact of Noncompete Contracts”, in Mimeo, Einaudi Institute for Economic and Finance, R&R Econometrica, https://www.liyanshi.com/s/noncompete_shi.pdf (accessed on 31 January 2022).

[73] Siegel, D. and K. Simons (2010), “Assessing the effects of mergers and acquisitions on firm performance, plant productivity, and workers: new evidence from matched employer-employee data”, Strategic Management Journal, pp. n/a-n/a, https://doi.org/10.1002/smj.843.

[11] Sokolova, A. and T. Sorensen (2021), “Monopsony in Labor Markets: A Meta-Analysis”, Industrial and Labor Relations Review, Vol. 74/1, pp. 27-55, https://doi.org/10.1177/0019793920965562.

[121] Starr, E. (2019), “Consider This: Training, Wages, and the Enforceability of Covenants Not to Compete”, ILR Review, Vol. 72/4, pp. 783-817, https://doi.org/10.1177/0019793919826060.

[114] Starr, E., N. Balasubramanian and M. Sakakibara (2018), “Screening Spinouts? How Noncompete Enforceability Affects the Creation, Growth, and Survival of New Firms”, Management Science, Vol. 64/2, pp. 552-572, https://doi.org/10.1287/mnsc.2016.2614.

[119] Starr, E., J. Prescott and N. Bishara (2021), “Noncompete Agreements in the US Labor Force”, The Journal of Law and Economics, Vol. 64/1, pp. 53-84, https://doi.org/10.1086/712206.

[128] Starr, E., J. Prescott and N. Bishara (2020), “The Behavioral Effects of (Unenforceable) Contracts†”, The Journal of Law, Economics, and Organization, Vol. 36/3, pp. 633-687, https://doi.org/10.1093/jleo/ewaa018.

[112] Streefkerk, M., S. Elshout and B. Cuelenaere (2015), Concurrentiebeding: Dataverzameling bij het LISS panel, CentERdata, https://docplayer.nl/52466944-Concurrentiebeding-dataverzameling-bij-het-liss-panel-in-opdracht-van-het-ministerie-van-sociale-zaken-en-werkgelegenheid.html.

[95] Taber, C. and R. Vejlin (2020), “Estimation of a Roy/Search/Compensating Differential Model of the Labor Market”, Econometrica, Vol. 88/3, pp. 1031-1069, https://doi.org/10.3982/ecta14441.

[80] Thoresson, A. (2021), “Employer concentration and wages for specialized workers”, IFAU Working Papers, No. 2021:6, IFAU, https://www.ifau.se/Forskning/Publikationer/Working-papers/2021/employer-concentration-and-wages-for-specialized-workers/.

[1] Topel, R. and M. Ward (1992), “Job Mobility and the Careers of Young Men”, The Quarterly Journal of Economics, Vol. 107/2, pp. 439-479, https://doi.org/10.2307/2118478.

[107] US Department of Justice; Federal Trade Commission (2022), Request for Information on Merger Enforcement, US Department of Justice and Federal Trade Commission , Washington, D.C., https://downloads.regulations.gov/FTC-2022-0003-0001/content.pdf (accessed on 26 January 2022).

[106] US Department of Justice; Federal Trade Commission (2016), Antitrust Guidance For Human Resource Professionals (October 2016), U.S. Department of Justice and the Federal Trade Commission, Washington, D.C., https://www.ftc.gov/public-statements/2016/10/antitrust-guidance-human-resource-professionals-department-justice (accessed on 15 November 2018).

[4] US Department of Justice; Federal Trade Commission (2010), Horizontal Merger Guidelines (08/19/2010), U.S. Department of Justice and the Federal Trade Commission, Washington, D.C., https://www.justice.gov/atr/horizontal-merger-guidelines-08192010 (accessed on 14 November 2018).

[108] US Department of the Treasury (2022), “The State of Labor Market Competition”, https://home.treasury.gov/system/files/136/State-of-Labor-Market-Competition-2022.pdf.

[109] Vestager, M. (2021), “A new era of cartel enforcement”, Italian Antitrust Association Annual Conference, https://ec.europa.eu/commission/commissioners/2019-2024/vestager/announcements/speech-evp-m-vestager-italian-antitrust-association-annual-conference-new-era-cartel-enforcement_en (accessed on 28 November 2021).

[88] Vick, B. (2017), “Measuring links between labor monopsony and the gender pay gap in Brazil”, IZA Journal of Development and Migration, Vol. 7/1, https://doi.org/10.1186/s40176-017-0099-x.

[129] Vuorenkoski, V. (2019), Kilpailukieltosopimukset Pohjoismaissa, AkavaWorks Blogit, https://akavaworks.fi/blogit/kilpailukieltosopimukset-pohjoismaissa/ (accessed on 1 February 2022).

[111] Vuorenkoski, V. (2018), “Kilpailukielto: ja salassapitoehtojen ongelmat ja ratkaisumallit”, AkavaWorks Artikkeli, No. 12/2018, AkavaWorks, https://akavaworks.fi/wp-content/uploads/sites/2/2019/10/Kilpailukielto-_ja_salassapitoehtojen_ongelmat_ja_ratkaisumallit_Artikkeli_1220181.pdf.

[181] Vu, V., C. Lamb and R. Willoughby (2019), I, Human: Digital and soft skills in a new economy, https://brookfieldinstitute.ca/wp-content/uploads/I-Human-ONLINE-FA.pdf.

[3] Wang, X. (2021), “US Permanent Residency, Job Mobility, and Earnings”, Journal of Labor Economics, Vol. 39/3, pp. 639-671, https://doi.org/10.1086/709689.

[13] Webber, D. (2016), “Firm‐Level Monopsony and the Gender Pay Gap”, Industrial Relations: A Journal of Economy and Society, Vol. 55/2, pp. 323-345, https://doi.org/10.1111/irel.12142.

[135] Wing, C. and A. Marier (2014), “Effects of occupational regulations on the cost of dental services: Evidence from dental insurance claims”, Journal of Health Economics, Vol. 34, pp. 131-143, https://doi.org/10.1016/j.jhealeco.2013.12.001.

[18] Wiswall, M. and B. Zafar (2017), “Preference for the Workplace, Investment in Human Capital, and Gender”, The Quarterly Journal of Economics, Vol. 133/1, pp. 457-507, https://doi.org/10.1093/qje/qjx035.

[30] Yeh, C., B. Hershbein and C. Macaluso (forthcoming), “Monopsony in the U.S. Labor Market”, American Economic Review, https://doi.org/10.1257/aer.20200025.

[147] Yeoh, P. (2014), “Whistleblowing: motivations, corporate self-regulation, and the law”, International Journal of Law and Management, Vol. 56/6, pp. 459-474, https://doi.org/10.1108/ijlma-06-2013-0027.

[113] Young, S. (2021), “Noncompete Clauses, Job Mobility, and Job Quality: Evidence from a Low-Earning Noncompete Ban in Austria”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3811459.

[139] Zhang, T. (2018), “Effects of Occupational Licensing and Unions on Labour Market Earnings in Canada”, British Journal of Industrial Relations, Vol. 57/4, pp. 791-817, https://doi.org/10.1111/bjir.12442.

[166] Zhao, D. (2021), Glassdoor Workplace Trends for 2022 - Glassdoor, https://www.glassdoor.com/research/2022-workplace-trends/ (accessed on 6 February 2022).

Annex 3.A. Additional figures

Annex Figure 3.A.1. Average concentration in OECD labour markets for the business sector, 2019

Annex Figure 3.A.2. Share of employment in moderately concentrated labour markets in the business sector by skill level, 2019

Annex Figure 3.A.3. Share of employment in moderately concentrated labour markets in the business sector by education, 2019

Annex Figure 3.A.4. Year-on-year quarterly change in Herfindahl-Hirschman Index (HHI), 2019‑21

Annex Figure 3.A.5. Ratio of the 90th percentile to the median of labour market concentration

Annex Figure 3.A.6. Some occupations benefit of large decreases in concentration when workers can retrain towards close‑by occupations, 2019

Annex 3.B. Country selection, standardisation and aggregation

In this chapter, the Herfindahl-Hirschman Index (HHI) is constructed from online job postings data web-scraped by Emsi Burning Glass (EBG). Hence, the validity of the estimates in the cross-country context relies upon the exhaustiveness of the EBG data. This annex illustrates the validation process of the EBG data, including the selection of the country sample, with the United States set as a benchmark. It also lists steps taken to standardise the data and make aggregated statistics comparable across countries.

EBG data have been previously used in the US literature on labour market concentration – see Azar et al. (2020[28]), Schubert, Stansbury and Taska (2021[179]) and Yeh, Hershbein and Macaluso (forthcoming[30]). Hershbein and Khan (2018[44]) show that the occupational distribution in EBG data is close to that emerging from Job Openings and Labour Turnover Survey (JOLTS), despite the fact that the former only contains online job postings. More importantly, Azar et al. (2020[28]) report that EBG data capture approximately 85% of the job openings estimated through JOLTS.

The data validation process used for this chapter is based on five steps:

1. Calculation of the new hire to vacancy ratio for each occupation in each country;

2. Calculation of the benchmark ratio for representativeness from the US data;

3. Selection of countries for the HHI calculation based on the benchmark ratio;

4. Calculation of the raw HHI at the local labour market (region by occupation) and quarter level for selected countries; and

5. Comparison of the raw HHI and a predicted HHI constructed by extrapolating information only drawn from local labour markets whose new-hire‑to-vacancy ratio is below the benchmark ratio.

Ultimately, out of 29 countries for which data are available, 16 countries (12 European OECD countries, three non-European OECD countries and one non-OECD country) have been retained.

In the first step, quarterly ratios of new hires to job postings for each occupation in 2‑digit ISCO (hereinafter referred to as “ratios”) are computed, and annual values are obtained averaging quarterly ratios using information from LFS and EBG data, for new hires and job postings respectively. Observations with either missing employer names or missing regions (TL3 regions or more granular units) are not taken into account. New hires are defined here as those whose job tenure is shorter than 3 months. Both new hires and vacancies in Armed Forces Occupations (ISCO 0) and Skilled Agricultural, Forestry and Fishery Workers (ISCO 6) are excluded. ISCO 0 is not recorded in many LFS. ISCO 6 is severely underrepresented in EBG data, according to the finding of Cammeraat and Squicciarini (2021[180]). The treatment of English-speaking countries required the use of a crosswalk to convert SOC‑6 to ISCO‑08 3‑digit occupations.

Some country-specific treatments were applied. The reference period is 2018‑19 for Australia, Canada, Singapore and the United Kingdom and 2019 for 24 European countries. Data for 2018 are not used for European countries as available data only contain postings from July to December. For European countries, job tenure with the current employer is not directly recorded in the EULFS microdata and thus it is inferred from the difference between the timing of the interview and the start year/month of employment. For the United States, job tenure is reported only on a two‑year basis by the January Current Population Survey (CPS) supplement and so job tenure refers only to the situation in January of even years. Accordingly, the EBG sample in a given year is constructed to contain observations from the beginning of the previous year to the January CPS supplement’s reference week. The quarterly average of vacancies is then obtained by dividing the number of these observations by four. The province of Quebec in Canada is excluded, as online job postings in French are under-scraped – see Lameb et al. (2019[181]). As for Germany, some TL3 regions corresponding to single‑municipality enclaves within larger regions are merged.62 Lastly, oversea territories are excluded to ensure comparability.

In the second step, a benchmark ratio is obtained by calculating the unweighted mean of quarterly ratios at the 2‑digit ISCO level in the EBG data of the United States. The unweighted mean of the 37 ratios (excluding ISCO 0 and ISCO 6) in the United States is 6.78, which is then used as benchmark threshold to test the representativeness of other countries’ EBG data.

In the third step, the percentage of those employed in occupations whose mean of quarterly ratios is lower than the US benchmark threshold is computed in each country. In the United States, for instance, 80.5% of the employed persons work in such 2‑digit occupations. Countries where less than 50% of the employed work in those occupations whose mean ratio is below the US benchmark thresholds are dropped. This leaves 20 countries in the sample. For Australia and Singapore, data regarding the number of new hires in 2‑digit occupations are not available. While their country-level ratio (i.e. the ratio of the total number of new hires to the total number of vacancies) is below 6.78, in practice the validation test is obviously much weaker for these two countries. For this reason, HHI statistics on these two countries must be taken with more caution. For Australia, Korbel (2018[182]) notes that the distribution of EBG job postings across occupations is different than the employment distribution as obtained from the labour force survey (LFS). While this may suggest further caution in using data from Australia, it must be noted that employment stock data represent a less rigorous benchmark for validating vacancy data than new hires, as done here.

In the fourth step, the HHI is calculated for the countries selected in the previous step.63 Six-digit SOC occupational categories are used for Australia, Canada, the United Kingdom, the United States and Singapore and 4‑digit ISCO occupational categories are used for the other countries.

When postings report the name of a job board64 instead of the true employer, the employer is considered missing. Missing employers are treated as individual, unique employers with one single posting. This is in line with what done by Azar et al. (2020[28]) and likely leads to underestimating concentration, but can be considered a conservative choice. In a robustness check, in order to reduce errors due to misreporting of employer names and imperfect cleaning, truncated employer names composed of the first word of the cleaned employer names are used with, nonetheless, similar results.

As a last step, the computed HHI aggregated at the 3‑digit ISCO‑2008 level is regressed on country and occupation dummies including only occupations whose ratio is below the benchmark threshold. The regression is estimated by employment-weighted OLS, excluding Australia and Singapore. Then, a HHI is predicted for excluded occupations. A country is selected for the analysis if the discrepancy between the mean of the actual HHIs and that of the predicted HHIs for a given country is lower than 10%. As a result, the number of retained countries goes down from 20 to 16: i.e. Australia, Austria, Belgium, Canada, the Czech Republic, Estonia, France, Germany, Latvia, Luxembourg, the Netherlands, Singapore, Sweden, Switzerland, the United Kingdom and the United States.

Raw HHIs are then computed at the most disaggregate level for the 16 countries that have been retained. Aggregation at the country or subnational level is based on a two‑step weighting scheme. First, based on available LFS data, the most disaggregated occupational level for which employment data is available is determined (usually 3 digits in the ISCO classification). Raw HHIs are then aggregated up to this level using the number of online job postings of each local labour market as weight. HHIs are then further aggregated at the national or subnational level using total employment of each occupation as weight.

To make comparable concentration statistics across countries, despite the cross-country heterogeneity in terms of average size of TL3 regions, the logarithm of aggregate measures of HHI is regressed on the logarithm of the country average population of TL3 regions, and the predicted value of the HHI for an average population of 200 000 individuals is obtained. Then the ratio of the predicted to the actual value is applied to adjust all concentration statistics obtained from online job posting used in this chapter.

Annex 3.C. Defining occupational transitions based on skill distances

The simulation exercise in Section 3.4.2 (“Skills policies”) requires identifying a set of potential alternative occupations a worker can work in (or destination occupations), given her own current occupation of employment (or occupation of origin), and some indicator of the skill distance between the two occupations.

There is no single, internationally agreed methodology to define such set of potential neighbouring occupations, but previous efforts in this direction were proposed at the OECD in OECD (2013[183]; 2019[169]), and Bechichi et al. (2018[184]; 2019[185]). As in those contributions, the current exercise imposes that a viable occupation of destination should not offer an average wage that is more than 10% lower than the average wage in the occupation of origin, and should not require an education attainment that is more than one year longer than that of the occupation of origin. Among the remaining possible occupations of destination, the worker chooses the occupation that minimises the skill distance between the two occupations and therefore the retraining effort, as long as the latter does not exceed 6 months in length.

The skill distance between origin and destination occupation, ${d i s t}^{o_d}$ is a function of the weighted average of the distance for each skills characterising the occupations. More precisely:

${d i s t}^{o_d} = \sqrt{\sum_{i = 1}^{35} {w_{i}^{d} (x_{i}^{o} - x_{i}^{d})}^{2}}$

where $o$ identifies the occupation of origin, $d$ that of destination, and $i = {1, 2, \dots 35}$ stands for one of the 35 skills that are used to describe occupations by the United States Occupational Information Network (O*NET). $x_{i}^{o}$ is the value of skill i in the origin occupation, corresponding to O*NET’s level of proficiency for that skill in the occupation, and $w_{i}^{d}$ is a weight measured as the importance of skill $i$ in occupation $d$ according to O*NET, relative to the skill’s importance in all possible occupations. Negative terms for $(x_{i}^{o} - x_{i}^{d})$ are set to zero to introduce an asymmetry in retraining between moving from $o$ to $d$ and the viceversa.65

A larger distance corresponds to a longer retraining effort if the worker chooses to move away from her occupation of current employment. The correspondence between distance and retraining time is obtained by regressing all skill distances on the difference in education requirements between origin and destination occupations. The coefficient on the education term estimates the number of points of skill distance that can be bridged with one extra year of education. This estimate is not used to define the one occupation of destination the worker may decide to retrain towards (for that, the skill distance is sufficient), but to limit the choice set of possible occupations of destination to those that require at most 6 months of training.

The information for the average educational requirement and the skill requirements of occupations is sourced from the United States O*NET, that for wages is from the United States Occupational Employment Survey. Patterns of transitions across occupations are therefore in common across all countries considered, which is equivalent to assuming – for the scope of this exercise – that patterns of potential transitions in the US labour market apply to other countries as well.

Notes

← 1. This chapter will refer to the situation describing wage‑setting power of employers using the terms “monopsonistic competition”, “monopsony” and “monopsonistic labour market” interchangeably. Similarly, the terms “wage‑setting power”, “monopsony power” and “employer market power” are used synonymously.

← 2. The estimates of the effect of labour market concentration on wages and job security presented in this chapter are the result of a collaboration with the Bank of Italy and the Institute for Employment Research (IAB) and are based on work by Giulia Bovini, Eve Caroli, Federico Cingano, Jorge Casanova Fernando, Paolo Falco, Florentino Felgueroso, Marcel Jansen, António Melo, Pedro Martins, Michael Oberfichtner and Martin Popp. The OECD Secretariat remains, however, the sole responsible of the views expressed in this chapter.

← 3. At the same time, concentration needs not imply monopsony power, where there exists countervailing market power on the part of workers. For example, in the mid‑20^th century, small towns in the United States typically had only one newspaper, so the local labour markets of typographers, who physically set-up and printed each edition, would have been considered to be very concentrated. And yet, at least until the 1970s, typographers were organised in a single, powerful union, and enjoyed significant wage premia with respect to workers in other manufacturing industries (MaCurdy and Pencavel, 1986[151]).

← 4. Certain policies also affect the degree of monopsony power that persists in the economy, as well as the way such employer power can reduce employment or wages. Non-compete covenants, for instance, respond to a specific policy objective, but limit the number of employers a worker can look to in order to find alternative employment. Regulatory provisions for social dialogue, conversely, are likely to limit the unilateral wage setting power of employers. As a consequence, policies are an extra source of monopsony power.

← 5. Models that depend on workers’ preferences over heterogeneous firms, and “dynamic” models of monopsony are not mutually exclusive. Manning (2020[25]), for instance, shows in a simple model that combining both approaches yields stronger monopsony power for firms.

← 6. Countries include Austria, Costa Rica, Estonia, Finland, France, Hungary, Italy, Portugal, the Slovak Republic and Spain.

← 7. Potentially one could rely on direct estimations of own-firm labour supply elasticities. However, this would require an instrument for wage changes, which would then complicate the estimation of the impact of labour supply elasticities on labour market performance (and in particular on wages).

← 8. A large number of European countries is covered in Ascheri et al. (2021[50]), who however restrict the statistical coverage to urban markets only.

← 9. Included countries are Australia, Austria, Belgium, Canada, the Czech Republic, Estonia, France, Germany, Latvia, Luxembourg, the Netherlands, Singapore, Sweden, Switzerland, the United Kingdom and the United States. Country selection is due to data availability.

← 10. Azar et al., (2020[28]) argue that the most disaggregated classification of occupation used in this chapter is still too broad, and that job titles may actually be the correct measure. The results in this chapter may therefore understate the extent of concentration.

← 11. Labour market concentration is measured using the Herfindahl-Hirschman index (HHI) computed on the basis of hiring, that is ${H H I}_{l, t} = \sum_{f = 1}^{F} s_{f, l, t}^{2}$ , where ${H H I}_{l, t}$ is the HHI in local labour market $l$ at time $t$ ; $F$ is the total number of firms on local labour market $l$ ; $t$ denotes time and $s_{f, l, t}^{}$ is the share of firm $f$ in employment, hiring or vacancies in local labour market $l$ at time $t$ . The index’s lower limit of 0 is reached only in the limit theoretical case in which there is an infinite number of firms. In a market with a finite number of firms n, the index is bounded from below by 1/n (the case of equal shares for each firm in the market).

← 12. Thresholds used by the European Commission are lower, however: 2 000 for high concentration and 1 000 for moderate concentration (European Commission, 2003[189]).

← 13. The countries considered suitable for the present analysis are: Australia, Austria, Belgium, Canada, the Czech Republic, Estonia, France, Germany, Latvia, Luxembourg, the Netherlands, Singapore, Sweden, Switzerland, the United Kingdom and the United States. Countries were selected based on data availability and a data validation exercise presented in Annex 3.B. Due to data limitations, no such validation exercise could be performed for Australia and Singapore, and results for these countries should be taken with more caution.

← 14. These are ISIC Rev. 4 sections O, Public administration and defence; P, Education; and Q, Human health and social work activities. The omission of industries where public employers often play a large role is motivated by the lack of robust evidence on whether public-sector employers use their wage setting power in the same way as private businesses.

← 15. In general, the results in this chapter accord well with the literature in most respects, especially when one considers that the definition of a labour market usually differs across studies in at least one dimension. The study using the closest definition of local labour market and HHI as in this section (See Box 3.1), reports an average HHI of 1 361 for the United States (Azar et al., 2020[28]), slightly higher but close to the US average found here (1 033). Remaining differences are likely due to data cleaning procedures (see Annex 3.B).

← 16. That time series is drawn from data on Austria, Costa Rica, Denmark, Finland, France, Portugal, and Spain.

← 17. These results are robust across individual countries. In particular, each occupation appearing in the top five most and least concentrated also appears in the top five at the country level in a majority of the 15 countries in the sample.

← 18. The number of workers in health-related occupations that are outside ISIC section Q (Human health and social work activities) is limited. Yet, health professionals and health associate professionals remain among the most concentrated occupations even when omitted industries are re‑included in the sample.

← 19. The chosen average encompasses all four quarters to absorb any seasonal variation separate from the effects of the pandemic. The average omits 2020Q1 due to the ambiguity over whether that quarter reflects dynamics before or after the onset of the crisis in all OECD countries.

← 20. Notable exceptions are Latvia and the United Kingdom, which both saw the largest increases in COVID‑19 case counts up until that point in each country, respectively.

← 21. The weights used to aggregate market-level concentration to the national level are kept constant and based on 2019 and do not therefore reflect e.g. changes in mandatory closures between 2020 and 2021.

← 22. Other studies looking at specific markets find more mixed results. For example Currie, Farsi and Macleod (2005[76]) and Prager and Schmitt (2021[77]) find no impact of mergers on employment in the US hospital industry.

← 23. In perhaps the only other study of this type, Popp (2021[83]) finds somewhat smaller but still large effects (1.5%), without controlling for product market competition and productivity.

← 24. The greater the number of new hires in a market, the greater the maximum number of firms that can hire in that market and the lower the theoretical minimum of the Herfindahl-Hirschman Index used to measure concentration.

← 25. Azar, Marinescu and Steinbaum (2022[52]), Rinz (2022[32]), Benmelech, Bergman and Kim (2022[29]), Arnold (2021[74]) and Schubert, Stansbury and Taska (2021[179]).

← 26. Martins (2018[35]), Abel, Tenreyro and Thwaites (2018[33]), Jarosch, Nimczik and Sorkin (2019[26]) Dodini et al. (2020[36]) Marinescu, Ouss and Pape (2021[34]), Bassanini, Batut and Caroli (2021[57]), Popp (2021[83]) and OECD (2021[12]).

← 27. Large estimates emerge only in studies not controlling for firm and individual fixed effects (Qiu and Sojourner, 2019[31]; Arnold, 2021[74]; Azar, Marinescu and Steinbaum, 2022[52]).

← 28. More precisely, OECD (2021[12]) estimates wage elasticities for several countries. However, as acknowledged in that study, the large confidence intervals for each country-specific estimate prevent country-by-country comparisons, and individual country estimates are used only to derive an average cross-country elasticity.

← 29. While the data on online job postings used in Section 3.2 allow for a greater country coverage, they do not contain information on individual wages or individual trajectories and characteristics. For this reason they are not used here. The analysis on skill demand in Section 3.3.3, focusing directly on the content of posted vacancies, will resort again to the online job posting data used above.

← 30. In the case of full-time workers, daily and hourly wages are likely to yield similar elasticities, which turns out to hold true also in the data used here. By contrast, the impact on daily wages beyond full-time workers is difficult to interpret as it is confounded by the effect of concentration on hours worked and the incidence of very short part time.

← 31. Larger standard errors for Denmark and Portugal are due to a “small country effect”. As in these countries the number of geographical areas is small, the instrument is somewhat weaker.

← 32. These values are obtained by multiplying the estimated wage elasticity by the logarithm of the ratio of the 9^th decile to the median of the distribution of concentration (see Annex Figure 3.A.5). It must be noted, however, that these estimates are more reliable at the sample average. Even more striking, in a few countries, the wage elasticity is even higher when estimated only on local labour markets with concentration below the average. Overall this implies that the negative effect of concentration materialises even in markets that are far less concentrated than the thresholds commonly used by antitrust authorities.

← 33. A few US studies (Qiu and Sojourner, 2019[31]; Arnold, 2021[74]; Azar, Marinescu and Steinbaum, 2022[52]) find elasticities lower than ‑0.1 in absolute terms, but other US studies, with closer specifications to those adopted here, find much smaller elasticities, comprised between ‑0.01 and ‑0.05 (Schubert, Stansbury and Taska, 2021[179]; Benmelech, Bergman and Kim, 2022[29]; Rinz, 2022[32]), which compare well to what found in European studies: Marinescu, Ouss and Pape (2021[34]) and Bassanini, Batut and Caroli (2021[57]) find an elasticity of ‑0.020 and ‑0.024, respectively, for France, while Martins (2018[35]), Dodini et al. (2020[36]) and Popp (2021[83]) obtain point estimates of ‑0.028, ‑0.010 and ‑0.043 for Portugal, Norway and Germany, respectively. Finally, OECD (2021[12]) finds an elasticity of ‑0.028, by pooling data for Austria, Denmark, France, Finland and Spain and Costa Rica.

← 34. This is consistent with results reported by Arnold (2021[74]), Bassanini, Batut and Caroli (2021[57]) and Thoresson (2021[80]) who find significant effect on incumbents’ wages in the United States, France and Sweden, respectively.

← 35. Interestingly, cross-country differences in terms of wage elasticities are small for incumbents while they are large for new hires: in fact the elasticity for Denmark (‑0.037) is more than twice as large as that of Germany (‑0.016), and the difference is significant, which suggests different wage adjustment patterns across countries.

← 36. While in specific cases, temporary contracts are sometimes associated to stable, good-quality jobs, the evidence suggests that on average they are associated with lower job security – see e.g. OECD (2014[191]) – and the incidence of the former represents therefore a good proxy for the latter.

← 37. Due to data limitations, the analysis is restricted to new hires as, in general, information on the type of contract for incumbents is missing in the available data. More precisely, in Germany and Spain it is always unavailable, while in Italy it is available only for workers hired after the beginning of the sample window (2012 for Italy).

← 38. 85% and 70% of new hires are on temporary contracts in the data in Spain and Italy, respectively.

← 39. While these effects are large in percentage terms, they are nonetheless small in percentage‑point terms, given the very low rate of conversions in these countries.

← 40. In these countries, contract type information is not regularly updated over the employment calendar.

← 41. This cannot be done with Portuguese data, however.

← 42. This definition of youth is slightly different than that offered earlier in the chapter reflecting the different data used in this section.

← 43. Removing apprentices from the sample would indeed reduce the point estimate of the effect of labour market concentration on the probability of being hired on a permanent contract by about 25%.

← 44. Modestino et al. (2016[48]; 2020[102]) explain the evidence with changes in recruitment intensity, i.e. the strategic behaviour of employers that invest greater resources in recruitment procedures when the supply of talent on the market is larger. Hershbein and Kahn (2018[44]) show that employers’ demand for skills increases permanently after demand shocks related to changes in technology or capital increases.

← 45. This subsection draws from and updates the discussions in OECD (2019[15]) and OECD (2020[149]).

← 46. Sometimes the literature distinguishes between “non‑compete” and “garden leave” clauses, the difference being that in the latter the worker is compensated after separating from the employer for the period of validity of the covenant, while in the former she is not – see e.g. Nicandri (2011[193]). For the purpose of this chapter, the term “non‑compete agreement” refers to both type of clauses, since there is an increasing number of countries and states in which a clause without worker compensation is not enforceable.

← 47. Mexico and certain few US states, including California, North Dakota and Oklahoma, are long-standing exceptions (OECD, 2019[15]). In 2020, the District of Columbia also enacted legislation banning non-compete agreements for employees (D.C. Law 23‑209: Ban on Non-Compete Agreements Amendment Act of 2020).

← 48. In almost all jurisdictions, NCAs must be limited in geographical scope to be enforceable. As a consequence, and particularly in low-skilled jobs, commuting to another city is often enough to overcome the constraint imposed by the clause.

← 49. A rebuttable presumption of abusive use means that the burden of proving that the use is not abusive is on the employer. If courts do not consider the alleged proofs convincing, the standard would be to consider the clause abusive.

← 50. A number of US states have introduced reforms in this direction in recent years, notably exonerating workers below a specified (and sometimes high) earnings threshold (Lewi et al., 2021[186]). In Europe, similar partial bans exist in Austria, Belgium and Luxembourg – see OECD (2019[15]).

← 51. Illicit collusion occurs, for example, when companies competing for the same type of workers agree on refraining from hiring those employed by the others (so‑called “non‑poaching agreements”) or when firms competing in the same labour market agree to apply a common compensation policy to employees (wage collusion), except when this occurs in the framework of sectoral collective bargaining.

← 52. The debate among regulators remains open on how to weigh the effects in labour and downstream product markets, in the cases where they are of opposite sign – see OECD (2019[15]) – although, in certain jurisdictions, there are clear guidelines suggesting that a merger in an upstream market should not be evaluated with reference to its consequences in the downstream market (US Department of Justice; Federal Trade Commission, 2010[4]).

← 53. The original formulation of the model of bilateral monopoly dates back to 1928 (Bowley, 1928[188]).

← 54. When international outsourcing is feasible, multinational corporations may threaten to relocate part of their production chains abroad with the objective to weaken the power of organised labour in their country of origin (OECD, 2021[192]).

← 55. In practice, in a monopsony model, in the unconstrained equilibrium, employment is lower than in the competitive equilibrium because the curve representing the marginal cost of labour is above (and steeper than) the supply curve. Moderate levels of the minimum wage shift down the marginal cost curve and make it flatter. As a result, employment is higher than in the unconstrained equilibrium and more reactive to changes in labour demand.

← 56. The fact that estimated effects of the minimum wage in the United States tend to become less negative (or even positive) when more recent sample windows are used (Dube, 2019[158]) may suggest that monopsony has become more pervasive over time.

← 57. Subsidies, however, are not necessarily cost-effective, in particular if targeting the employed, and may generate competition effects for workers in the destination region (Schmutz and Sidibé, 2019[190]).

← 58. An alternative unreported exercise assumes instead that the occupation of destination is the one that maximises the worker’s wage gain, conditional to at most 6 months of retraining.

← 59. Canada and Singapore are excluded because of missing data on employment at the 3‑digit ISCO 2008 level. The calculation of transitions at the 3‑digit ISCO‑08 level is standard (Bechichi et al., 2018[184]; Bechichi et al., 2019[185]) and a requirement in the present context, so as to associate the same transitions to the Australian, United Kingdom and the United States data (originally in SOC‑2010) as to the European data (originally in ISCO‑08). SOC occupations are converted in ISCO categories before calculating the HHI. Standard HHIs are also recalculated at the 3‑digit ISCO level for the purpose of this exercise.

← 60. Almost two‑thirds of all 3‑digit ISCO occupations do not find a valid transition to another occupation within the limit of 6 months of retraining.

← 61. On the one hand, as job tenure increases, greater dismissal costs reduce the employer’s bargaining power. On the other hand, however, the greater unwillingness to quit of more senior employees (because they would lose tenure‑related protections upon quitting) increases their employer’s monopsony power.

← 62. Merged TL3 regions are: Heilbronn (Stadtkreis) and Heilbronn (Landkreis); Baden-Baden (Stadtkreis) and Rastatt; Rosenheim (Kreisfreie Stadt) and Rosenheim (Landkreis); Landshut (Kreisfreie Stadt) and Landshut (Landkreis); Passau (Landkreis) and Passau (Kreisfreie Stadt); Straubing (Kreisfreie Stadt) and Straubing-Bogen; Regensburg (Kreisfreie Stadt) and Regensburg (Landkreis); Bamberg (Kreisfreie Stadt) and Bamberg (Landkreis); Bayreuth (Landkreis) and Bayreuth (Kreisfreie Stadt); Coburg (Kreisfreie Stadt) and Coburg (Landkreis); Ansbach (Kreisfreie Stadt) and Ansbach (Landkreis); Schweinfurt (Kreisfreie Stadt) and Schweinfurt (Landkreis); Würzburg (Landkreis) and Würzburg (Kreisfreie Stadt); Kaufbeuren (Kreisfreie Stadt) and Ostallgäu; Kempten (Allgäu – Kreisfreie Stadt) and Oberallgäu; Cottbus (Kreisfreie Stadt) and Spree‑Neiße; Bremerhaven (Kreisfreie Stadt) and Cuxhaven; Wilhelmshaven (Kreisfreie Stadt) and Friesland (DE); Bonn (Kreisfreie Stadt) and Rhein-Erft-Kreis; Trier (Kreisfreie Stadt) and Trier-Saarburg; Flensburg (Kreisfreie Stadt) and Schleswig-Flensburg; Weimar (Kreisfreie Stadt) and Weimarer Land.

← 63. The quarterly HHI of job postings can be written as follows:

$H H I_{t, i j} = \sum_{k = 1}^{K} {(% s h a r e o f v a c a n c i e s t, i j k)}^{2}$

where t is denoted for a quarter, i for a TL3 region, j for an occupation in 6‑digit SOC or 4‑digit ISCO and k for a firm. In other words, the HHI in the sample can be uniquely defined by a quarter, region and occupation (which is the definition of local labour market here – see Box 3.1).

← 64. A significant proportion of online job postings are by job boards and true employers are not observable, as EMG data do not indicate whether an employer is a job board or a true employer. Hence, the process of cleaning employer names adopted for this chapter requires the identification of job boards. In each country, the top 50 employer names are selected based on their share of vacancies. Reported employer names are then checked to assess whether they are job boards or not by identifying globally active recruiters and recruiting websites (e.g. Robert Walters, Michael Page, Adecco, Völker, Grafton, Hays, CV-Online, Page Personnel), systematically flagging words related to human resources (e.g. “career”, “headhunt”, “HR”, “job”, “manpower”, “personal”, “personnel”, “recruit”) and manually verifying their business activities on the Internet.

← 65. The functional form for the skill dissimilarity indicator follows Robinson (2018[187]) and is one of a range of options that are used in the literature. This metrics combines simplicity (in the use of Euclidean geometry) and the possibility to estimate bidirectional skill distances.

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment

Governance

Health

Industry, business and entrepreneurship

Regional, rural and urban development

Science, technology and innovation

Society

Taxation

Trade

Energy

Nuclear energy

Transport

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Countries

Countries A - C

Countries D - I

Countries J - M

Countries N - R

Countries S - T

Countries U - Z

Regional and global engagement

Publications

Publications

Featured publications

Data

Data

Featured data

News & Events

News & Events

Featured Events

About OECD

About

Engage with us

Work with us

Featured topics

Agriculture and fisheries

Climate change

Development

Digital

Economy

Education and skills

Employment

Environment

Finance and investment