… is from page 233 of Milton & Rose Friedman’s great 1980 book, Free To Choose  (first emphasis original; second emphasis added):
Moreover, the ability of unions to raise the wages of some workers does not mean that universal unionism could raise the wages of all workers. On the contrary, and this is a fundamental source of misunderstanding, the gains that strong unions win for their members are primarily at the expense of other workers.
The key to understanding the situation is the most elementary principle of economics: the law of demand – the higher the price of anything, the less of it people will be willing to buy. Make labor of any kind more expensive and the number of jobs of that kind will be fewer.
For example, while in practice a legislated minimum wage will reduce over time the number of jobs for low-skilled workers, in principle the negative effects of the minimum wage can be manifested almost exclusively in changes in the kinds of jobs filled by low-skilled workers. A minimum wage could, in principle, cause nearly all of the jobs filled by low-skilled workers to become so much more demanding for workers, more dangerous to workers, or otherwise less desirable for workers to hold (yet more productive from the standpoint of employers) that the number of such jobs doesn’t decline. This ‘change-in-the-nature-of-the-jobs’ effect is one among many reasons why empirical studies that purport to find no negative consequences of minimum-wage legislation for low-skilled workers should be approached with enormous helpings of skepticism: no such quantitative studies can possibly measure accurately the adjustments that occur on the many margins along which employment relations are defined in reality.