What follows is wonkier than is typical for Cafe Hayek posts. The principal take-away is in the italicized paragraphs below.
A standard premise of any monopsony argument that is used to justify minimum-wage legislation is that hiring an additional low-skilled worker requires that an employer with monopsony power, to get that additional worker, not only pay that additional worker a wage higher than that employer had, up to that point, paid to all of its other low-skilled workers, but also that this employer raise the wages of all of its other, previously hired low-skilled workers up to the level of the wage paid to the newly hired worker.
Put differently, a premise of this model is that there is no ‘wage discrimination’ in the market for low-skilled workers (analogous to price discrimination in an output market). If, however, wage discrimination were possible, even employers with monopsony power would hire a number of workers up to, or closer to, the optimal number of workers, and each such employer would pay its last-hired worker an hourly wage at, or closer to, the ‘correct’ market wage of W’ (in the graph below). But, again, a standard premise of this model is that no such wage discrimination occurs. It is this premise that ensures that monopsony power results in too few low-skilled workers being hired and that each and every low-skilled worker who is hired is paid a wage that is below the value of his or her marginal product.
The result of this premise is shown in the nearby graph, which is familiar to every student who has taken a respectably decent course in intermediate microeconomics. [A summary explanation of the graph is below* for those who wish to read it. But good economics students will immediately know the correct meaning and interpretation of the graph.] The monopsony employer depicted by this graph employs only L workers (which is fewer workers than would be employed were this labor market competitive) and pays each an hourly wage of only W (which is lower than the value of these workers’ marginal product and, hence, lower than would be paid if the labor market were competitive).
This premise sits uneasily with any monopsony model used to justify minimum-wage legislation. If workers have little ability to quit their current employers in search of new jobs – or if, from the other perspective, employers face little risk of their current underpaid workers quitting – then there is no reason why employers with such monopsony power must raise the wages of their existing workers in order to hire new workers. After all, existing workers (according to the tenets of the monopsony model) being too poor, too poorly informed, too desperate, and without practically available employment alternatives, can’t or won’t quit merely because the new hire is paid a wage higher than theirs. So monopsony-power-laden employers can hire new workers without raising the pay of their existing workers, and it would be profitable for these employers to do so up to L’. The supply curve becomes the employer’s marginal-cost curve. Workers will be hired up to the optimal level (L’), with the last-hired worker paid an hourly wage of W’. It follows that a minimum wage that is set at a level designed to raise the wages of all low-skilled workers will put some such workers – the ‘marginal’ workers – out of jobs.
The last sentence of the preceding paragraph states what is perhaps the most basic conclusion from this analysis. This conclusion, however, does imply that a minimum wage set precisely at W’ (but no higher) will raise the wages of all inframarginal low-skilled workers without causing any loss (or gain) of jobs for low-skilled workers. (Note 1: See this related EconLog post by Scott Sumner  in which he quotes Matt Rognlie. Note 2: Here I assume, contrary to fact and logic but for sake of argument, that monopsony power is not only a necessary but also a sufficient condition for a minimum wage not to destroy jobs for low-skilled workers. In other posts  I’ve challenged this careless if common assumption – and I’ll say more about it in forthcoming posts.)
But the above analysis points also to a deeper conclusion. It is this: the existence of monopsony power strongly suggests, even if it does not logically imply, that the range of wages that each employer of low-skilled workers pays its low-skilled workers will be wide. Therefore, if in reality all low-skilled workers for each particular employer are paid roughly the same wage by that employer, then that employer is unlikely to possess monopsony power. There is simply no reason for an employer with monopsony power to raise the wages of previously hired workers 1 through 9 when that employer offers newly hired worker 10 a higher wage than that firm pays to workers 1 through 9.
So what do we observe in reality? Do we observe that each firm that employs low-skilled workers pays its low-skilled workers a wide range of different wages? If so, then the case that monopsony power really exists among such employers is stronger than it would be upon a contrary finding, but the argument that a minimum wage will increase the number of jobs for low-skilled workers (a’ la the monopsony model) is very much weakened. [And, do not forget, a necessary (although not sufficient) condition for the minimum wage not to cause some low-skilled workers to lose their jobs is that government officials are expert wage-target shooters: they must not set the minimum wage higher than W’.] Or do we observe that each firm that employs low-skilled workers pays each of its low-skilled workers roughly the same wage that it pays to all of its other low-skilled workers? If so, then the case that monopsony power really exists among such employers is weak – and, hence, the economic case for any legislated minimum wage revealed as being quite flimsy indeed.
(Another relevant question along these lines is: what is the degree of dispersion of the wages of low-skilled workers across different firms? The lesser the degree of dispersion of such wages, the less plausible is the claim that employers of low-skilled workers possess monopsony power. If Bob’s & Sally’s Diner in Anytown, USA, pays its low-skilled workers wages that are quite close to the wages paid to low-skilled workers at Anytown’s McDonald’s, Taco Bells, and Erma’s Maid Service, then this fact is evidence against the existence in Anytown of monopsony power.)
A final note: empirical tests along the lines suggested above are made more challenging not only by the real-world existence of different degrees of difficulty and different marginal values of tasks performed by low-skilled workers, but also by the existence of minimum-wage legislation. The reason is that such legislation artificially imposes more wage uniformity than would otherwise exist among low-skilled workers, and especially among those who are employed by employers that possess monopsony power.
The main point of this post is not to suggest empirical studies (although carefully done studies would be welcome). Instead, the main point is to indicate the tension, in the familiar monopsony model, between the divergent marginal-cost and supply curve and the ways that employers with actual monopsony power are likely to behave in reality. Employers with monopsony power in reality are, contrary to the textbook monopsony model, likely to practice wage discrimination resulting in the MC curve being pretty much identical to the S curve.
* MC = marginal cost, here the marginal cost to the firm of hiring an additional worker; this cost includes (in the monopsony model) not only the amount the firm must pay to the newly hired (“marginal”) low-skilled worker but also the extra amount that it presumably must pay to its other, previously hired (“inframarginal”) low-skilled workers.
MRP = marginal revenue product, here the extra revenue the firm earns by employing the additional (“marginal”) worker.
S = supply-of-labor curve, here the supply of low-skilled workers; this supply curve is upward-sloping showing that in order to increase the number of low-skilled workers it employs, the firm must raise the wage it pays to these workers.
The hourly wage rate is shown on the vertical axis while the quantity of low-skilled labor (which can be interpreted for convenience as the number of low-skilled workers) is shown on the horizontal axis.
In a competitive labor market the firm employs L’ workers and pays each an hourly wage of W’. This outcome is optimal, according to economic theory, because the hourly wage paid to each low-skilled worker is here exactly equal to the amount of extra revenue per hour that each such worker, reckoned as the marginal worker, makes it possible for the employer to earn.
But if this employer enjoys monopsony power, it employs only L workers and pays each an hourly wage of only W. The reason is that this employer’s marginal cost of hiring additional workers is not just the wages it pays to each additional worker who is hired; this cost includes also the hike in wages that this employer is presumed to give to all of its previously hired (“inframarginal”) low-skilled workers. So when contemplating whether or not it’s profitable to hire an additional worker, this employer compares the amount that that worker, if hired, would add to the employer’s hourly stream of revenue (MRP) to the cost that this employer would incur by hiring this additional worker (MC). Because the marginal cost of hiring this additional low-skilled worker includes both the wage that it must pay to this additional worker plus the hike in the wages that this employer must presumably pay to its already-hired low-skilled workers, this firm’s marginal cost of hiring an additional worker is above the wage it must pay to that additional worker – meaning that the firm’s marginal-cost curve lies above, and rises more steeply than, the supply curve of low-skilled workers that this firm faces.
The classic conclusion drawn from this static model is that a government-imposed minimum wage of W’ will cause this firm to increase the number of low-skilled workers it hires up to L’. The reason is as follows: now that all of the firm’s low-skilled workers must be – and, hence, are – paid at least the minimum wage of W’, adding an additional worker at that wage does not require the firm also to hike the wages of its previously hired workers. (Minimum-wage legislation establishing a minimum wage of W’ causes the supply curve to becomes horizontal and, hence, identical to the marginal-cost curve at wage W’ up to point C in the graph.)