The floor of a wage can come from several places. Subsistence locates one in the cost of maintaining a worker; the reservation wage locates another in the alternatives a worker can bear to refuse. The minimum wage adds a public condition to the private bargain: covered work must command at least the payment the law specifies. The worker may have accepted less and still be entitled to the floor; she may require more and decline an offer that satisfies it. This is the floor the law imposes, entering a market whose own terms may have been quite different. The question is what changes when it enters. The competitive model supplies a clean answer: a binding minimum raises the price of labor and reduces the quantity employers buy. The evidence that has troubled this answer asks how far that model describes the market in which the law operates. This plate follows the prediction through its quantities, then through the comparison needed to test it, to the opening that the next plate gives employer power.
In the competitive account, each employer takes the wage as given and hires labor until the contribution of another unit just covers its cost. The demand curve slopes downward because, with the other conditions of production held fixed, successively more labor contributes successively less at the margin. The supply curve records how much work people offer at each wage. Their meeting determines a clearing wage and a quantity traded. A minimum below that wage changes nothing in this simple model. A minimum above it separates the quantities: firms want less labor at the higher price, while workers offer more. Actual employment is limited by demand, and some of the labor offered goes unhired. The conclusion follows from the structure, including the absence of employer wage-setting power and the treatment of the curves as the appropriate descriptions of adjustment. An employer cannot restore the old quantity merely by wishing the new price away. The model's argument against the floor is an argument about the work that ceases to cover its cost.
An invented market puts exact numbers on the distinction between lost work and unsatisfied offers. Let the hourly wage on the demand curve be twenty minus one-tenth of the number of workers, and on the supply curve four plus one-tenth of that number. Each worker supplies one hour, and the only difference between the two situations will be the permitted wage. The curves meet at eighty workers and a wage of twelve. Set the floor at fourteen: firms now demand sixty workers, while a hundred offer their labor. Employment has fallen by twenty, but excess supply at the new rate is forty. Calling all forty lost jobs would count the additional workers drawn in by the better wage as though they had previously been employed. Calling that forty an official unemployment count would add another unwarranted step; the statistical category has its own requirements concerning availability and search. The diagram measures offers and purchases at stipulated prices. Its precision is useful because it forces the argument to name what has actually fallen.
The familiar claim that the floor must exclude the least-skilled adds an assumption about which workers and which contributions the curve represents. In a model of identical units of labor, nothing identifies a particular worker as the one who loses the job. With different skills and tasks, a floor can make some matches unprofitable, and firms may prefer applicants who contribute more at the stipulated wage. But the ordering depends on what each person can do in the jobs available, on substitution between workers and on how firms organize production. A person with little formal schooling need not have the smallest contribution in every task. Even the number of workers is an interpretive choice: the quantity on a labor-demand curve may be hours, with employment headcount unchanged. The competitive warning remains substantive. It says that increasing the required payment can remove marginal work. To identify the people affected, and how their circumstances change, the argument must move beyond a homogeneous curve to the opportunities those people actually face.
The gain to those who keep the work and the loss to those who lose it also prevent a higher hourly rate from settling the earnings question. In the invented market, the hourly payroll falls from twelve times eighty, or 960, to fourteen times sixty, or 840. The rise in the wage does not compensate the group for the reduction in work. Another demand schedule could give a smaller quantity response and a larger total payroll; elasticity matters. Neither total would tell us how the proceeds were distributed between the worker retained and the worker excluded. The same distinction operates within a job. In a separate invented example, forty hours at twelve yield 480 a week, while thirty-four hours at fourteen yield 476. The person remains on the payroll, with a higher hourly wage and slightly lower weekly earnings. The arithmetic establishes why a headcount result, a rate result and an earnings result are three findings, each answering a different question about the same policy.
Testing the competitive prediction requires a credible account of what employment would have done without the higher floor. A fall after an increase might be the effect of a recession; a rise might conceal work that would have grown faster without the policy. A comparison group offers a way to subtract movements that would have occurred anyway. In a deliberately invented comparison, treated establishments rise from a hundred jobs to a hundred and five while comparable untreated establishments rise from a hundred to a hundred and ten. Employment grew in both places; the difference between their changes is minus five. Interpreting that difference as the policy's effect requires the untreated change to represent the path the treated establishments would otherwise have followed. Different demand shocks, anticipatory adjustment or spillovers between the groups can unsettle the comparison. The method therefore moves the dispute to a sharper place. The issue becomes whether the selected comparison supplies the missing counterfactual, rather than whether somebody can point to jobs rising or falling after the law.
Card and Krueger made that comparison concrete when New Jersey raised its minimum from $4.25 to $5.05 on April 1, 1992, while Pennsylvania's remained at $4.25. They surveyed 410 fast-food restaurants before the increase and followed the establishments afterward, comparing employment changes across the boundary and between differently affected stores within New Jersey. In their 1994 paper, the increase produced no indication of reduced employment in the sampled restaurants. This was a challenge in a setting where wages did rise, rather than a case in which the floor sat harmlessly below what firms already paid. The study had a particular measure of employment, a particular industry, a particular interval and a particular comparison. Those features define its evidentiary force. A robust result there can make a general prediction less credible without establishing that every wage floor in every industry will behave the same way. The empirical question was now lodged inside the market that theory had been asked to describe.
The wage distribution offers another way to distinguish work that disappears from work whose price changes. If a covered, compliant job previously paid less than a new minimum, it must disappear from that part of the distribution. It may reappear at a higher wage with its work intact. Doruk Cengiz, Arindrajit Dube, Attila Lindner and Ben Zipperer used this distinction in their 2019 study of 138 prominent state minimum-wage increases between 1979 and 2016. Against their estimated counterfactual, jobs missing below the new minimum were closely matched by additional jobs at or somewhat above it, leaving overall low-wage employment essentially unchanged over the following five years. The comparison is about numbers of jobs in wage ranges. It does not establish that each particular worker kept the same job. Its advance is to expose where an employment estimate comes from: whether the disappearance of low-paid positions reflects destruction of work, a movement up the wage scale, or both.
A broad average still leaves room for differences across markets and for uncertainty about floors outside the range studied. Cengiz and his coauthors found negative but imprecise employment estimates in tradeable sectors and manufacturing, alongside their near-zero overall result. Their discussion expressly left higher, unstudied minimum wages an open question. Even within the observed range, a statistically insignificant estimate is accompanied by an interval of plausible effects; it is not a certificate that the true effect is exactly zero. Conversely, insisting that a loss must exist somewhere does not establish a loss of the magnitude originally predicted. The argument has to carry a level, a comparison, a time horizon and a measure. A modest effect on total low-wage employment can coexist with important effects on particular workers, and a very large national total can dilute the signal from the small part of the workforce directly exposed. The question becomes more exact as the evidence becomes more local.
A small employment response leaves the question of who bore the higher wage bill, and through what adjustment. Prices may rise, profits may fall, hours or other terms may change, and altered recruitment and turnover may change the cost of maintaining a workforce. These are possible margins to investigate, each requiring evidence of its own. Card and Krueger found meal prices rising in New Jersey relative to Pennsylvania, while prices did not rise more in the New Jersey stores most affected by the new floor. The combination complicated simple accounts of both competition and monopsony. That is an instructive difficulty. A story that fits the employment sign should also confront prices, wages and the organization of the firm. If the wage rises while recorded jobs hold steady, the event establishes room for that observed adjustment under those conditions. It does not reveal, by subtraction alone, how much profit was surrendered, how much efficiency was gained or precisely how far pay had been below the value of the work.
The next plate derives one reason that a higher minimum can increase employment: the employer may have been restricting hiring to hold the wage down. A firm facing an upward-sloping labor supply must offer more to recruit more, and under a uniform wage the higher offer also raises the pay of its existing workers. The marginal cost of another hire therefore exceeds that worker's own wage. A suitable floor changes this calculation. For the range over which workers are available at the legal rate, another hire need not raise everyone else's wage, so the firm can find it profitable to employ more. The range and the quantity willing to work at that rate matter; the supply constraint does not disappear merely because a floor exists. A sufficiently high minimum can again reduce labor demanded. This is a coherent alternative structure of the market, with its own conditions and comparative statics. The absence of a large measured job loss makes room to investigate it. Identification of the mechanism still requires evidence beyond that absence.
The choice of a floor therefore joins an assessment of the market to a judgment about the gains and losses that matter. Raising earnings for a worker who keeps her hours is a benefit; reducing access to work is a cost for the person excluded. A price increase can distribute part of the burden among customers, and a reduction in profit among owners. These effects cannot all be read from an employment total, nor can the people in one category be assumed to have the same circumstances as those in another. The instrument also reaches only the work to which it applies and the employers who comply. It cannot, by itself, supply earnings to everyone without a job. These distinctions make the policy's purpose more demanding and more intelligible: determine how much better a specified floor makes the worker's position, which other margins move, and who receives or bears the resulting change. The wage is the entrance to that inquiry, rather than its final measure.
The floor tests the wage by changing it. What moves with the wage tells us something about the bargain beneath it.
The minimum wage brings the volume's argument about worth and power into an unusually visible experiment. The law changes a term of the bargain, and the consequences disclose something about the market that had settled it. Under the competitive benchmark, a binding floor removes marginal work; in the historical settings examined here, the predicted large employment loss did not appear. That finding warrants revisiting the benchmark and examining the other adjustments, with employer power among the mechanisms capable of changing the answer. It gives no universal number below which a floor is harmless or above which a worker becomes dispensable. The question that survives is the useful one: what does this floor do, in this market, to pay and to the opportunity to earn it? The next plate returns to the wage before the law intervenes and changes one premise. The employer has to offer more to hire more. From that apparently small departure comes a different account of why the wage was low in the first place. The argument remains open.