From Marginal Revolution:
Amazon’s widely touted increase in its minimum wage was accompanied by an ending of their monthly bonus plan, which often added 8% to a worker’s salary (16% during holiday season), and its stock share program which recently gave workers shares worth $3,725 at two years of employment. I’m reasonably confident that most workers will still benefit on net, simply because the labor market is tight, but it’s clear that the increase in the minimum wage was not as generous as it first appeared.
What lessons does this episode hold for minimum wage research? Amazon increased its wages voluntarily but suppose that the minimum wage had been increased by law. What would have happened? Clearly, Amazon would have, at the very least, eliminated their bonus plan and their stock share plan! In this situation, researchers examining employment data would discover that the increase in the minimum wage did not much lower employment. Such researchers might conclude that minimum wages don’t reduce employment much because the demand for labor is inelastic. The conclusion is correct but the reasoning is false. The correct conclusion and reasoning would be that the minimum wage didn’t reduce employment much because the minimum wage didn’t increase net wages much.
Amazon is a big and newsworthy employer so its actions have been closely monitored but in most cases we never know the myriad ways in which firms respond to a law. Even using administrative data it would be difficult to pick up changes in a stock share plan or a pension plan, as this compensation doesn’t show up in earnings until years after the work is completed. Even a simple employment contract is a complicated bargain with many margins. During the holiday season, for example, Amazon hires a CamperForce of workers who live in RVs and it pays their campsite fees–no big deal, but that is a form of compensation that is hard to find on a W-2. More generally, firms can respond to a minimum wage by changing compensation on non-wage margins, adjusting working conditions, reducing benefits, changing wage growth patterns, and adjusting the type of workers they hire, to give just a few examples–and notice that all of these changes are difficult to measure and none of them have a first-order effect on employment.