An antitrust case in a labor market must answer one central question: where can these workers provide their labor? New Jersey’s August 4 complaint against Amazon, billed as the first monopsony conduct case brought by a state, defines the relevant market in a way no labor economist would recognize.
New Jersey’s complaint asserts a “Delivery Service Partner (DSP) Drivers Market” consisting of the labor of drivers employed by Amazon’s DSPs, in New Jersey and in the New York-Newark-Jersey City metropolitan area. Amazon’s alleged buyer power then follows almost automatically from the market definition.
The complaint offers little economic analysis to support its implausible central claims:
- Workers whose main prerequisite qualification is an ordinary driver’s license, and who receive no significant specialized training or credentials, are in some way:
- locked in to Amazon or its DSPs as employers, and
- unable to credibly consider alternate employment options such as other package delivery companies, other driving or delivery jobs, or any other job not requiring significant credentials or qualifications;
- This implausible worker lock-in results in Amazon achieving monopsony buyer power in the relevant market for its labor.
If the court accepts this dubious market definition, overly narrow market definitions could be expected to replace serious economic analysis in future antitrust complaints alleging monopsony conduct.
Almost Any Large Company Can Be Declared a Monopsonist If You Define the Relevant Market as New Jersey Does
“DSP driver” is a category defined by the identity of the company whose packages get delivered to customers: Amazon. While you can easily describe a job this way (“the job is to drive and deliver Amazon packages to customers”), you cannot credibly describe a labor market this way. Most jobs delivering packages for a company that contracts with Amazon require no commercial driver’s license, credential, certification, or firm-specific training a worker carries into the role. If you have a regular driver’s license, you can apply. What makes someone a DSP driver, as opposed to any other worker whose job consists mostly of driving and making deliveries, is that Amazon pays their employer for their services.
Market definition asks a distinct question: Which employers can a worker credibly threaten to leave for? The boundary is drawn by the substitution possibilities available to the worker. By defining the market around who is permitted to deliver Amazon packages, rather than what other employment options are available to workers, New Jersey has answered a question about Amazon’s supply chain and presented it as an answer about New Jersey’s labor market.
The test New Jersey needed to run is a wage-side analogue of the hypothetical monopolist test: could a single firm controlling every DSP driver job in New Jersey profitably impose a small but significant non-transitory worsening of employment terms, such as a reduction in compensation? Answering that question requires evaluating evidence about where workers go when pay falls or working conditions deteriorate.
Ordinary Search Frictions Are Not Labor Market Boundaries
New Jersey acknowledges that UPS, FedEx, and the Postal Service employ delivery drivers, then sets them aside as meaningful alternatives for DSP drivers on the grounds that a driver “cannot simply decide to work for an alternative employer; the employer must be hiring when the driver is looking and must select the DSP driver.”
That describes just about every labor market. Vacancies commonly open at particular moments. Applications get screened. Employers choose among candidates. These are search and matching frictions, the same widespread phenomena for which the 2010 Nobel Prize in economics was awarded. These are universal features of labor markets, not evidence that two jobs sit in different markets. If the inability to obtain a job unilaterally established non-substitutability, then no two employers would ever compete for the same worker, and every job in America would be its own antitrust market.
New Jersey attempts to add a bit more meat to this thin gruel of reasoning by discussing wages, but the State makes key errors in the process. The State infers the market boundary from the alleged wage gap itself, reasoning that because DSP drivers earn less than UPS drivers, those jobs must not be substitutes. Of course, wage differences across employers have many sources, including but not limited to capital intensity, route density, tenure structure, benefit composition, and unionization. Treating any gap as proof of the market’s boundaries, and then treating the boundaries as proof that the gap reflects monopsony harm, is circular reasoning.
Contrast this with how market definition works in product-market monopolization cases, where a plaintiff cannot define the market by pointing to the fact that the defendant charges more than other sellers. The hypothetical monopolist test asks the opposite question: whether enough buyers would switch to rival products after a small but significant price increase to make that increase unprofitable. The persistence of price differences across brands, retailers, or formats within a single market is treated as ordinary, not as evidence of separate markets. Applied to the labor side, the question New Jersey needed to answer is whether a small but significant wage cut by DSPs would cause enough drivers to move to other jobs to make the cut unprofitable.
The Missing Universe of Competing Jobs
The complaint’s discussion of substitutes assumes a DSP driver would only seriously consider other package delivery jobs. A whole universe of jobs outside package delivery is available to a light-delivery driver tomorrow.
The Bureau of Labor Statistics classifies light truck driving as an occupation requiring a high school diploma or equivalent, a driver’s license, and a month or less of on-the-job training. Occupation-specific human capital is close to zero. Those characteristics make a broad range of jobs plausible substitutes: any other job that can be performed by a worker with a high-school diploma or GED, a driver’s license, and a month or less of training. For example, warehouse and fulfillment work, food and grocery delivery, rideshare, moving and local freight, retail stocking, and non-parcel courier work are just some plausible options to test against the data. The complaint briefly mentions a few options like rideshare and food delivery but dismisses them as side gigs rather than substitutes for DSP work. Analysis of actual job substitutability is nearly absent.
The economic literature pushes against the complaint’s assumptions and reasoning. Dodini, Lovenheim, Salvanes, and Willén examine the universe of Norwegian workers and find that measuring concentration by occupation and industry, similar to what the complaint does, likely overstates employer power because worker skills are substitutable across firms, occupations, and industries. When they measure the same labor markets by skill rather than by job category, the estimated degree of concentration falls. In other words, a market definition built on job titles is biased toward overstating buyer power.
That bias is largest for workers like DSP drivers. In the Review of Economic Studies, Caldwell and Danieli build a measure of a worker’s actual option set that does not depend on occupation, industry, or geographic boundaries, and validate it by showing that workers with more options recover faster from mass layoffs. Holding willingness to commute constant, workers with less education have more relevant job options than workers with more, because credentialed workers cluster in occupations whose skills transfer poorly. Doctors and pilots have narrow option sets despite high pay. A worker whose primary qualification is a driver’s license has a broad set of options. This runs counter to the New Jersey complaint’s logic. If the package delivery business contracts, DSP drivers can typically switch to other jobs and even job types without abandoning any valuable credentials or employer-specific skills.
In the American Economic Review, Prager and Schmitt studied a decade of hospital mergers and found slower wage growth only where two conditions held together: the merger sharply increased concentration, and the affected workers’ skills were specific to the industry. For everyone else, including workers whose tasks were not hospital-specific, they could not detect a wage effect. Relevantly, DSP drivers largely lack industry-specific skills that constrain exit options.
Why This Matters Beyond Amazon
Labor market monopsony is a serious subject, and buyer power in labor markets deserves serious economic analysis and potential enforcement attention when procompetitive outcomes are lacking. That is precisely why market definition matters. A framework in which the relevant market is “workers who perform this task for this company or its contractors, inside this state” will find monopsony power everywhere, because it is built to. It would make any large employer an antitrust defendant whose job titles are distinctive and whose wages differ from a unionized competitor’s.
Courts should insist on real economic analysis: evidence of where workers actually go, drawn from mobility and employment data. New Jersey’s complaint substitutes a job description for a market definition, and forgoes serious economic analysis as a result.
