Recent studies have documented rising product market concentration and analysed its effect on the labour share (Barkai 2016, Loecker and Eeckhout 2017). Another growing literature estimates employer concentration in the labour market and its impact on wages. One paper found that from 1977 to 2009, the concentration of employment in county- and industry-level labour markets increased by 5.8%. The authors also found a negative relationship between employer concentration and wages (Benmelech et al.2018).
The effect of increasing market concentration has often been overlooked by economists working under the assumption that the labour market is competitive. Under perfectly competitive markets, workers are paid their marginal product of labour – the additional value they bring to their employer. If a worker were to earn less, she would leave for another employer willing to pay them their worth. Assuming that wage differences reflect productivity differences, labour economists have typically turned to worker characteristics to explain wage variation between workers. However, recent findings on labour market concentration question this assumption, as does the growing literature on interfirm inequality (Card et al. 2016, Song et al. 2016). Labour market monopsony – employer power to set wages below workers’ marginal productivity – is an alternative cause of variation in labour market outcomes between workers.
In a recent paper, we used data from Burning Glass Technologies (BGT), a company that gathers job vacancy posting data from over 40,000 websites, corresponding to approximately 80% of the total job postings in the economy (Azar et al. 2018). The dataset contains many characteristics of each vacancy, including those related to occupation. The SOC code, the standardised job title, and the BGT occupation are all included. We restricted the data to 2016 to obtain the best possible data quality, dropped internships and data with missing SOC or commuting zones, and were left with over 24 million observations. For the analysis we kept the most frequently occurring 200 occupations, which represented over 90% of the postings. Focusing on the most frequent occupations allows us to discount the influence of rare and atypical occupations. The total number of markets (6-digit SOC occupation by commuting zone by quarter) in the main analysis is 117,707.
We used the Herfindahl-Hirschman Index (HHI) to measure labour market concentration. The HHI is defined as the sum of the squares of each firm’s share in vacancy postings in a given labour market and then multiplied by 10,000. In our baseline market definition, the average HHI is 3,953 and 54% of markets are highly concentrated. The HHI is widely used in industrial organization literature and in antitrust practice to measure market concentration. According to the Horizontal Merger Guidelines, a market with an HHI above 2500 is considered highly concentrated. It states that “mergers resulting in highly concentrated markets that involve an increase in the HHI of more than 200 points will be presumed to be likely to enhance market power” (DOJ 2010).
A crucial contribution of our paper is its discussion of market definition in labour markets. Defining a labour market is necessary in order to compute the HHI because we need to market shares for a particular market. The baseline market definition of the commuting zone by 6-digit SOC by quarter fulfils the requirement in the merger guidelines to define an antitrust market along two dimensions – a relevant product market and a relevant geographic market. HHI was calculated at the quarterly level, since the median unemployment duration is about 10 weeks. The commuting zones that define the geographical labour markets are based on clusters of counties that were developed by the USDA, using data from the 2000 Census on commuting patterns across counties. Given past research estimating the probability of job applications to a given vacancy as a function of distance (Marinescu and Rathelot 2017), we conclude that commuting zone is the closest empirical analogue to a geographic labour market.
We believe our proposed labour market definition is conservative because job vacancies vary within occupational codes. One paper that studied wage inequality amongst SOC codes and job titles by analysing job adverts on CareerBuilder.com found that unlike occupational codes, job titles explained nearly all of the wage variation amongst job openings (Marinescu and Wolthoff 2016). Contrary to standard assumptions in labour market search-and-matching, the authors found that within occupations, vacancies that posted higher wages received fewer job applications – suggestive evidence that job postings within occupations are too heterogeneous to be considered a single labour market. The authors point to the fact that, within an occupation, senior job title postings offer higher wages than junior job title postings and that senior positions attract more experienced applicants, but fewer in total. Therefore, the market concentration estimated in their paper may be too low, since a more properly defined market is plausibly job titles within commuting zones. The implication of this is that following a wage reduction by a potential monopsonist, a worker would be more likely to look for other jobs posted with their job title, not at all the jobs posted in their occupation.
The hypothetical monopolist test is used in antitrust policy to define a product market. The test defines the relevant antitrust market as the smallest market for which a hypothetical monopolist would find it profitable to implement a “small significant non-transitory increase in price (SSNIP)”. This idea is applied to labour monopsony with the monopsonist instead using a reduction in wages (SSNRW). Similar to the standard 5% increase in price used in antitrust policy, we consider a 5% reduction in wages. If a hypothetical monopsonist in a defined labour market would find it profitable to reduce wages by 5%, then the labour market definition is too narrow.
We show how a SSNRW test could be used to define labour markets. The hypothetical monopsonist profits from the productivity in excess of wages from each of their employees. The change in profits from the hypothesized 5% wage reduction equals the change in employed labour multiplied by the change in their excess productivity – the markdown. The monopsonist is profitable if the saving in wages is greater than the loss of labour to other firms, which in turn is determined by the firm-specific labour supply elasticity.
After rearranging the terms, we get the critical elasticity equal to the inverse of the markdown minus the change in wages. In our paper, we consider a markdown of 45% and use the wage reduction of 5% to provide an illustrative example. Using these numbers would give a critical elasticity of 2.
A market definition is too broad if the market-level elasticity of labour supply is less than the critical elasticity. Assuming a markdown of wages at 45% implies that the market is too broad if the defined labour market has an elasticity less than 2, and is too narrow if more than 2. Estimates of the firm-specific labour supply elasticity typically ranged between 0.1 to 0.4, with most estimates below 2. Unless one were to believe that the mark-up is well above 45%, the markets in the paper are too broadly defined.
Dube et al. (2018), who analysed the trade-off between the probability of task acceptance given a set reward on Amazon’s Mechanical Turk, calculated the labour supply elasticities to always be below 0.12.
In some ways, this paper and others estimating employer concentration in labour markets take research on monopsony back to where it began. The ‘new monopsony’ literature focuses on firm-level wage-setting power arising from sources other than out-and-out concentration of employers in the market, such as search frictions and asymmetric information (Manning 2011). One example from that literature is Dube et al. (2016), which analysed how quits responded to arbitrary differences between a worker’s wage and their peers at a large US retailer and found that “quits do not appear to be very sensitive to wages – consistent with the presence of search frictions”. By focusing on market concentration in well-defined markets, our paper revisits the old monopsony literature – employers have wage-setting power because workers can’t leave for another job, since there are few other employers. In fact, both mechanisms can operate together and reinforce each other to limit the mobility of individual workers and thus increase the wage-setting power on the part of employers.
Furthermore, the lack of competition amongst employers can potentially explain interfirm inequality among otherwise-similar workers (Song et al. 2016), with one paper finding that firm wage effects contribute to 20% of the overall variance in wages (Card et al. 2016). Those findings imply that outside offers are infrequent enough that they fail to equilibrate wage offers between firms.
Our paper contributes to the growing debate about whether market concentration might be one cause of stagnant wages and other labour trends. The results of this and other work suggests that the anti-competitive effects of labour market concentration could be important. Marinescu and Hovenkamp (2018) suggest that merger policy should consider harmful effects in the labour market. Although a potential merger may not reduce competition in the product market, this does not mean that the labour market will be unaffected. The analysis of labour market definition in our paper could be used to incorporate labour market concentration concerns as a factor in the legal review of mergers.
References
Azar, J, I Marinescu, M Steinbaum, and B Taska (2018), “Concentration in US Labor Markets: Evidence From Online Vacancy Data”, NBER Working Paper 24395.
Barkai, S (2016), “Declining Labor and Capital Shares”, University of Chicago Working Paper.
Benmelech, E, N Bergman, and H Kim. 2018. “Strong Employers and Weak Employees: How Does Employer Concentration Affect Wages?”
Card, D, A Rute Cardoso, J Heining, and P Kline (2016), “Firms and Labor Market Inequality: Evidence and Some Theory”, NBER Working Paper 22850.
De Loecker, J, and J Eeckhout (2017), “The Rise of Market Power and the Macroeconomic Implications”, NBER Working Paper 23687.
Dube, A, J Jacobs, S Naidu, and S Suri (2018), “Monopsony in Online Labor Markets”, NBER Working Paper 24416. See also on VoxEU.org.
Dube, A, L Giuliano, and J Leonard (2016), “Fairness and frictions: The Impact of Unequal Raises on Quit Behavior”, IZA DP 9149.
Manning, A (2011), “Imperfect Competition in the Labor Market”, Handbook of Labor Economics Vol. 4, Part B, 973-1041.
Marinescu, I, and H Hovenkamp (2018), “Anticompetitive Mergers in Labor Markets”, SSRN Scholarly Paper ID 3124483, Rochester, NY.
Marinescu, I, and R Rathelot (2017), “Mismatch Unemployment and the Geography of Job Search”, American Economic Journal: Macroeconomics, forthcoming.
Marinescu, I, and R Wolthoff (2016), “Opening the Black Box of the Matching Function: the Power of Words”, NBER Working Paper 22508.
Song, J, D J Price, F Guvenen, N Bloom, and T von Wachter (2016), “Firming Up Inequality”, UCLA Working Paper.
