Where a Job Search Actually Happens
This chapter defines the local labor market in which unemployment-insurance time can become—or fail to become—bargaining power. The focus is not on memorizing a concentration index, but on understanding which employers are counted as alternatives, how their shares are measured, and what that measurement leaves out. That discipline matters because a claim that a market is “concentrated” has no stable meaning until its geography, type of work, and unit of analysis are clear.
Geography is part of choice
A worker does not usually compare every employer in a country. They compare jobs they can plausibly reach and perform. Commuting time, relocation costs, family obligations, licensing, training, and the portability of prior experience all narrow the practical set. A national company may advertise everywhere, but its local worksites may offer only a small number of positions in a particular occupation or industry. The relevant question is therefore not simply how many employers exist, but how many are realistic substitutes for this worker in this place.
Labor economists represent that idea by defining a market along two dimensions. The first is a section of the economy, usually a geography. The second is a line of commerce, which, for labor, is a type of work or group of jobs. Combining the two produces a market such as manufacturing in a metropolitan area or health care in a commuting zone. Workers supply labor in that market; employers demand it.
The geography cannot be chosen casually. The Bureau of Labor Statistics article uses metropolitan statistical areas as its baseline geography because they are designed to capture relationships among places connected by commuting. The Census working paper uses commuting zones for its worker-level analysis. These choices do not claim that every worker follows the same map. They are attempts to approximate the area in which a worker can search without undertaking a costly move.
This is also why a local labor market can matter even when firms operate nationally. A large employer may be one competitor among many across the country but a dominant buyer of labor in a particular town. If the next plausible employer is far away, the national number of firms does little to improve the worker’s immediate outside option. Conversely, a firm that enters a local market can increase local competition even if it makes the firm’s national position larger. Competition is experienced at the level where a worker can actually change jobs.
The second dimension is harder. The BLS study uses industry groups as a practical line of commerce, while noting that occupations would often be preferable because workers are less substitutable across occupations than across industries. An accountant may work in health care, banking, or construction, but that does not make an accountant interchangeable with a nurse, banker, or builder. If a market definition groups all of those jobs together, it may count employers as competitors even when their hiring needs do not overlap much.
The reverse problem is possible too. A narrow industry category may exclude employers that could hire the same worker. An experienced accountant may have options across several industries even if the study measures only one. A narrow line of commerce can therefore make the market look more concentrated than the worker’s actual search environment. The important lesson is not that one definition is always right. It is that every concentration result is conditional on a claim about substitutability.
How HHI turns employer shares into a market signal
Once a market is defined, the Herfindahl–Hirschman Index, or HHI, summarizes how its employer shares are distributed. The basic calculation squares each employer’s market share and adds the squared shares together. In symbols, if employer i has share sᵢ, then the index is the sum of sᵢ² across employers.
Squaring makes the distribution matter. A market with many employers of roughly equal size has a low concentration measure. A market in which one employer holds a large share has a high one, because the large share receives disproportionate weight after squaring. This is useful for the search-time question: a worker is not equally protected by the existence of ten employers if one controls most of the relevant jobs and the other nine are too small, too distant, or too different in their hiring needs to be credible alternatives.
The HHI is a market-structure measure, not a worker’s wage and not a direct measure of bargaining power. It tells us how employment or payroll is distributed among firms. It does not tell us whether a particular employer offers a safe workplace, predictable hours, good benefits, a manageable schedule, or a job that uses a worker’s skills well. It also does not prove that a large employer is exploiting workers. It identifies a condition under which employer power may be possible and makes markets comparable under a stated definition.
The BLS article expresses employment-share HHI on a scale from zero to 10,000. Under the 2010 Department of Justice and Federal Trade Commission taxonomy used there, values below 1,500 are classified as unconcentrated, values from 1,500 to 2,500 as moderately concentrated, and values above 2,500 as highly concentrated. These categories are screening conventions, not natural laws. The BLS article notes that the agencies’ 2023 guidelines use lower thresholds, so applying the older taxonomy understates the number of workers who would fall into a highly concentrated category under current guidance.
The Census paper uses a different construction. Its local HHI is based on firms’ shares of payroll in a commuting-zone-by-industry market and is expressed on a zero-to-one scale. A payroll share asks how much of the market’s pay bill is associated with each firm; the BLS employment measure asks how many jobs are associated with each firm. Both can describe employer concentration, but they answer different questions. Their values cannot be compared by moving the decimal point or importing a threshold from one study into the other.
This distinction matters for the series’ central theme. A worker’s ability to find a better job may depend on the number of jobs, the payroll attached to those jobs, the relevant skills, and the quality of the offers. Employment HHI and payroll HHI provide different windows into that environment. Neither is the whole window.
There is a further distinction between the number of markets and the number of workers in those markets. Imagine a country with many small towns, each dominated by one employer, alongside a few large metropolitan areas with many competing firms. Counting markets gives the small towns substantial weight because each town is one observation. Counting workers gives the large metropolitan areas more weight because more people work there. Both summaries can be accurate while answering different questions: how common is a market structure, and how many workers are exposed to it?
The BLS results later in this series use both perspectives. The average and median market can look highly concentrated even when the employment-weighted average is much lower, because concentrated markets tend to have smaller workforces. That is not a contradiction to be averaged away. It is information about the distribution of labor-market arrangements. A worker in a small, employer-dominated market may face a serious constraint even if most workers nationally are in larger, more competitive markets.
A second comparison concerns the denominator behind a firm’s share. Employment-based concentration treats each covered job as a unit. Payroll-based concentration weights firms by the pay attached to their workers. A firm with many lower-paid jobs and a firm with fewer high-paid jobs can have different positions under the two measures. The choice is therefore connected to the outcome under study. If the question is whether a worker has another employer to apply to, job counts may be intuitive. If the question concerns the distribution of pay or the bargaining environment reflected in a payroll bill, payroll shares may capture a different part of the relationship.
A third comparison concerns the level at which data are observed. Administrative records often arrive as establishment reports: a store, plant, hospital, or office reports its employment and wages. The worker, however, encounters an employer that may operate several establishments. Treating every worksite as a separate competitor can make a multi-site firm appear less dominant. Combining every location too aggressively can create the opposite problem if legally related establishments do not compete for the same workers in the relevant market. Firm identification is not a clerical step after the economic analysis; it helps determine what the economic participants are.
These comparisons suggest a practical reading rule. Whenever a study reports concentration, ask four questions before interpreting the number: what is the geographic boundary, what work is included, what is the denominator, and what counts as one firm? Then ask a fifth question: is the result weighted by markets or by workers? The answer will not resolve every ambiguity, but it will reveal which ambiguity the number is designed to manage.
The diagram shows why the index is downstream from several judgments. A researcher first decides which places belong together and which kinds of work compete. They then aggregate worksite reports to firms and choose whether the share should be based on employment or payroll. The final HHI inherits all of those choices. It is not a direct observation of “how much power” a firm has in every negotiation.
The boundary between a useful proxy and a misleading label
The aggregation step creates another important distinction: an establishment is a worksite, while a firm may own multiple establishments. To study employer concentration, the relevant participant is generally the firm, not each individual location. The BLS article uses employer identifiers to move from establishment-level reports toward firm-level analysis, but it also notes that the available employer identification number is not a perfect common key. Large, multistate firms may have more than one identifier, which can make a market appear more diffuse than it really is.
The Census analysis faces a related but different data problem. It uses the Longitudinal Business Database to construct a consistent firm-level payroll measure across state borders, then links that market measure to worker records in the Longitudinal Employer-Household Dynamics program. The result is well suited to its question about how UI extensions interact with local concentration, but it is still a constructed market measure. A commuting zone may cross state borders; an industry code may combine jobs with different skill requirements; and the records do not turn every unobserved alternative into an observed one.
These limitations do not make HHI useless. They tell us how to use it. A result that survives several reasonable geography and industry definitions is more informative than one that appears only under a single narrow partition. A result about employment concentration should be described as such. A result about payroll concentration should retain that denominator. A market classified as highly concentrated under the 2010 BLS thresholds should not be presented as if it were a universal diagnosis of every worker’s bargaining position.
The same restraint applies to remote work. Remote arrangements may enlarge the set of jobs a worker can consider, but the supplied evidence does not establish that remote work dissolves local labor markets. Job-seeking behavior, relocation costs, worksite requirements, and the specific occupation still determine how many employers are realistically available. The proper conclusion is that geography may be changing for some workers, not that geography has stopped mattering.
The HHI therefore performs a precise but limited job. It asks whether the measured employment or payroll in a defined market is spread across many firms or concentrated in a few. That answer helps explain why extra UI time may produce different results in different places. It does not by itself tell us whether workers actually wait longer, switch employers, obtain higher pay, or find better non-wage conditions.
The next question is empirical: once these markets are constructed, how common is high concentration, how much employment does it cover, and what relationship does it have with wages? Only then can the discussion move from the architecture of the measure to the scale of the problem it reveals.