What follows will be a mere data gathering proposal. I'm not intrinsically arguing for or against a minimum wage with this post, just arguing for the data collection that will allow us to study the effects of the minimum wage. In a sentence: Every state should start tracking every worker's total hours. (Every state already tracks total earnings for the sake of workers compensation eligibility, among other things, so this is not much of a change.)
The Seattle minimum wage studies by the Jardim et. al. group were particularly well done and revealing, because they actually had the data to accurately measure wages and hours worked for basically everybody in the state of Washington. The only states that currently capture this data on "hours worked" are Washington, Oregon, Rhode Island, and Minnesota (according to Jacob Vigdor, who was kind enough to correspond with me by e-mail on this and other questions). Most states collect this data for the sake of tracking eligibility for workers comp, and most states just have a "total earnings" trigger for eligibility. Washington (and possibly these other states) also have a "total hours worked" trigger, so they have to capture this data to administer workers comp in compliance with state law. If every state were collecting this data, the empirical work on minimum wage would be a lot more accurate.
Knowing everyone's hours worked in addition to total earnings allows us to 1) figure out what each worker's average wage was (by simply dividing the one by the other) and 2) figure out if minimum wage increases cause workers' hours to be scaled back. Item 1) is important because most studies have used crude proxies for "minimum wage earners", effectively assuming that restaurant workers or teen employees are minimum wage workers. With "hours worked" we could accurately identify minimum wage (or near-minimum-wage) workers before and after the wage hike. Item 2) is important because employers may be more prone to scale back hours than to do outright layoffs. So the "disemployment" effect is due to loss of hours worked, not job losses. Indeed, this is what was found in Seattle, and low wage workers' hours were scaled back so much they actually lost out in terms of total wages.
My hesitation about writing this post came from reading Coyote Blog. The blog's author, is an employer and often writes about what a pain in the butt it is to gather data for various government reporting agencies. So I specifically asked him this question via e-mail, and he was kind enough to respond. He told me that almost all employers use a payroll provider anyway, who are almost certainly tracking the number of hours worked so as to accurately calculate employees' paychecks. It's possible that this data is being collected somewhere already, it's just not being reported to any central government for aggregating and doing studies. Collecting this might be as simple as flipping a switch already built into each payroll provider's code, or at the very most changing some code to make it similar to what's used in Washington. (He clarified that what's painful is when governments start asking for data that was never collected before, which the employer might not even know. Say, a demographic category, or something like "disability status", which an employer might not be legally allowed to even ask about. He's written extensively about this on his blog, so I'm not revealing any secret details from the e-mail.)
This proposal is agnostic about whether the minimum wage is a good idea or a bad idea. Minimum wage proponents should be saying, "Yes, that will give us the firepower we need to refute Seattle and put this to bed once and for all!" Minimum wage opponents should be saying, "Yes, that will allow us to replicate the findings of Seattle as more states and cities roll out minimum wage hikes!" Neutral parties can take a wait-and-see attitude. Economists should be celebrating the job security of having tons of extra data to analyze. "Yay! Now we can figure out new statistical tricks to make the effect get bigger/go away!"
To expand the proposal just slightly, we could also start tracking each employee's municipality, in case different cities have different minimum wages. The Seattle study started with all Washington workers, but it had to infer who did and didn't work in Seattle based on the employer's address. This didn't work for, say, McDonald's, because McDonald's has locations all over the state. Only businesses that were unique to Seattle could be located for the sake of the study. I think the Jardim et. al. papers handled this limitation adequately and their results are still valid, but some critics have latched on to this as a reason to dismiss the paper's results. So let's fix the problem going forward.