The debate over the effects of the minimum wage (and whether it should be increased and indexed to inflation) flares up every few years, especially when good-hearted politicians  promise to promote the public welfare by raising the price floor on wages.
The arguments marshaled by the two sides tend to be fairly similar every time this happens. And in many of the articles that cover the opinion and editorial pages of newspapers we get a repeat of some inadequate arguments that ought to be laid to rest. My goal in this post is to articulate some rules for the two sides that will make the debate a lot more productive.
For the political “left”:
1) Appeals to social justice are fun and all, but devoid of economics - It’s the consequences of a government policy that matter, not its intended consequences. Even if we accept as 100% true the fact that companies are evil, greedy, and dictatorial, if the economics of the minimum wage points to outcomes that are contrary to the stated intended outcomes, then all the social justice arguments are irrelevant – and useless.
Example: if raising the wages of 1 million workers by 20% requires 1 million workers to lose their jobs entirely, is this a good policy? Probably not. Of course, the estimated disemployment effects of minimum wages are far below the numbers I just gave. But the fact remains that appeals to social justice are pointless if we find out that economics tells us good intentions will ultimately be thwarted.
What should you do? If you find that your argument only claims that we need a “living”, “fair”, or “reasonable” wage and that corporations are exploiting us, don’t publish it anywhere! Include evidence from economics that a law in fact can achieve your goals without having large unintended consequences. If you don’t, you will not (and should not) be taken seriously. Of course, the evidence you present must avoid the other pitfalls set out below.
2) Pointing to Card and Krueger (1994) isn’t enough (and never was) - If we accept empiricism in economics, we know that it takes a whole body of research to establish the existence of an effect with confidence. Before Card and Krueger (1994), the minimum wage literature was strongly weighted in favor of minimum wages having significant disemployment effects. It is possible for an article to singlehandedly crack apart a whole body of literature if it has an impeccable argument for why the entire literature previously employed a methodology that systematically biased the results in one direction . Card and Krueger (1994) does not meet the bill. They do not have fantastic controls that correct for all previous confounding variables (or any other such groundbreaking improvement). As such, this study is not enough to overthrow the results found before their study.
What they found in the New Jersey/Pennsylvania study was that there were some positive employment effects in New Jersey after the minimum wage hike, even when compared to neighboring Pennsylvania, which did not raise its minimum wage. What they failed to do was provide a compelling theoretical argument for why this might have happened, especially when they did not find an increase in output (which would have been a slight bit of evidence in favor of the monopsony explanation). The 1994 study has been roundly criticized, and I feel fairly comfortable saying that it’s doubtful they did a very good job controlling all the factors that could have had an impact on employment, thereby identifying a causal effect of minimum wage on employment (or lack thereof).
Card and Krueger (1994) opened the door in the debate, but they did not get very far past the threshold. Simply throwing out the name is not enough to win a discussion legitimately. Don’t do it.
What can you do? You can point to other studies that have found minimal or insignificant disemployment effects due to minimum wage increases, many of which employ methodology that is leaps and bounds ahead of that in the 1994 article. An example is Dube’s recent work, which uses policy discontinuities in counties across state lines as counterfactuals. In English, they study lots of data from counties that are on state lines, where one state increases its minimum wage and the other does not (which, the argument goes, creates a situation where we can be more sure that the changes in employment are due to the change in minimum wage, since the bordering counties are [allegedly] so similar) .
3) Convincingly argue that non-wage factors are not relevant to the debate - While the simplest model of markets (perfect competition) predicts that a minimum wage will cause some people to be laid off, a more nuanced view states that it’s not necessary for workers to lose their jobs. Their total compensation package (which is what matters in the labor market) could remain the same while the minimum wage increases if non-wage benefits fall. That is, even if the wages of employees go up, it’s possible that the total compensation they get stays the same. What could that mean? It could be that firms will provide less job training, it could be that working conditions will be improved less quickly (or even fall into partial disrepair), it could be that workers are made to work harder than they’d otherwise like in order to make up for the increased wages, or it could be that other benefits (health benefits or others) could be curtailed or frozen. These are all called the margins of adjustment of the firm in response to a minimum wage hike. Even if studies show employment to be completely unaffected, there could be a significant reduction in employee welfare if these other factors take a hit. It’s your job to argue that either these effects are statistically insignificant (either through theoretical or solid empirical arguments) or to state that workers don’t really care about these benefits as much as wages (which seems like a tough task, given that employers choose to provide some of these benefits instead of additional direct wage compensation). You should also address the possibility that employment isn’t hit because low-skilled workers are replaced with high-skilled workers, leaving total employment levels unchanged but low-skilled workers worse off.
4) Back up the claim that current increases are similar to the ones studied before – and that inflation indexing will not have significant negative effects - Even if you believe the data shows that small increases in the minimum wage will not have a bad effect on employment, this does not mean that all the current proposals can use that data as evidence for their case. Some of the proposed increases are very substantial – far beyond the minor adjustments about which economists can make reasonable claims. We also don’t have very good data on indexing the minimum wage to inflation. What evidence would you present to argue that this won’t have a negative effect down the road? This requires some explanation, since there is no prima facie economic reason for why the wage of the lowest-paid (and lowest-skilled) workers ought to increase with the average price level of consumer goods in the economy (CPI) (or your price inflation index of choice). We don’t expect the price of potatoes, or of burgers, or of anything else to mirror increases in CPI. Why should the minimum wage (without negative side effects)?
For the political “right”:
1) Demolishing Card and Krueger (1994) isn’t enough - I am sure that people who are against the minimum wage can find plenty of reasons to discard Card and Krueger’s 1994 study as evidence against the minimum wage. Many of these reasons are likely very good. However, the fact is that the field has moved beyond that study, producing other empirical arguments against large disemployment effects, the most notable ones being found in the work of Dube.
Ultimately, it may be that most studies finding evidence against the existence of disemployment effects are wrong. The point is that if you think so, you have to critique the best of these arguments and not merely content yourself with showing Card and Krueger (1994) is inadequate. We know that that study is inadequate as a refutation of the preexisting literature. Move on. Critique the latest studies (and use Neumark and Wascher’s work to back you up in your critique, for example), or stick to theoretical arguments (and be prepared to explain why the empirical studies are not good enough).
2) Presenting simple regressions that show disemployment effects without controls is bad science - I covered this point in my earlier post on some sloppy economics flying around earlier this year. Here’s the gist of my argument: to be able to claim you are doing science, you have to prove the effect you’re seeing in the data is really due to the policy in question. If you simply look at employment before and after minimum wage increases, you are not doing this. As such, the data is useless to either prove or disprove employment effects. Why? Because it could be that there is a third factor that causes both the changes in minimum wage and the changes in employment. Example: leftist politicians both increase the minimum wage and enact other labor regulations. Even if the minimum wage has exactly 0 effect on employment, looking at the data without controlling for other variables will result in estimating negative employment effects. Another example is recessions: if the minimum wage tends to be increased in the middle of a recession (say, due to political reasons), the effects in the data could be due to the recession, not the minimum wage (Dube mentions this here).
A particularly egregious example of this is the following image, which has been circulating on Facebook:
Not only does this argument present a whopping 4 data points, but one of them is followed by a slight uptick in labor force participation, while two are fairly close to the beginning of recessions. Furthermore, there could be an underlying downward long-term trend in labor force participation that could in part explain the decreases in labor force participation. One idea: perhaps we’re becoming wealthier as a society, which means teenagers no longer want/need to work. Is this true? I don’t know. But the fact remains that without controlling for additional factors, we can’t know (from empirical observation) whether the declines are 1) bad 2) due to the minimum wage.
What should you do instead? Either stick to theory that counters pro-minimum wage arguments or present better empirical studies that do control for at least some confounds. Whether the confounds controlled can ever be enough is open to debate. But whether confound control is needed is not.
For both sides of the debate:
1) Both sides claim the weight of the evidence is on their side – you’re not special - If you make the statement that most studies agree with your position, be prepared to back this statement up, because everyone is pulling this card nowadays. It’s your job to link to a review of literature that takes stock of all relevant studies, weeds out the ones with bad methodology, and tallies up the score for both sides. This review of literature should also explicitly explain why the methodology of the rejected studies was inadequate (and make sure that the studies that it chose to retain do not make the same mistakes). Not only this, but you should present the main arguments of this review of literature within your own article; don’t expect readers to do it themselves. Readers too often take claims at face value; be responsible and at least give them the meat of the argument without overwhelming them.
So if you say “the evidence clearly shows,” be prepared to either throw your article in the garbage or provide serious theoretical backup for why the other side’s evidence should be ignored.
2) Producing a letter signed by hundreds of economists backing your side also doesn’t make you special - First and most importantly, neither science nor economics is settled by vote. 51% doesn’t win. Beyond that, as Greg Mankiw points out, there are literally hundreds of economists signing letters for either side of the debate. If you’re going to parade your letter around, make sure to also parade that of your opponents.
1) Give compelling justification for your model and a compelling critique of the opposition model! - For the left, this means critiquing the perfect competition model of the economy and backing up your preferred model. If this is the dynamic monopsony model, for example, explain why you think dynamic monopsony provides a convincing description of how the low-skill labor market functions and provide evidence that the dynamic monopsony model is good enough to explain the effect seen in the data. If, on the other hand, you think the minimum wage does not reduce employment, explain why you think the dynamic monopsony model is inadequate and defend your own model (whether it’s perfect competition or something else).
If your entire article consists of making social justice arguments, citing Card and Krueger’s 1994 study, claiming the evidence for yourself in passing, and pointing to a letter signed by economists, save it.
If you critique the C&K 1994 study, show some superficial data, and state that the evidence agrees with you, as do some economists who signed a letter, shelf it and read it back to yourself when you feel down.
These two types of articles neither add something new to the debate nor make rigorous arguments. Perhaps they could have earned a passing grade in middle school social studies; not so in the real world.
Seriously. Move beyond Card and Krueger, and if you’re going to use empirical studies, use ones that control for as many relevant factors as possible. If you don’t, then don’t expect any response other than a link to this article.
Notes and References:
 I know, I laughed too.
 I have an example of this in mind, although I am not sure how good it is: it’s the Miron and Tetelbaum (2007) paper “Does the Minimum Legal Drinking Age Save Lives?” I am not well-acquainted enough with the literature to be able to tell whether Miron and Tetelbaum’s review of previous literature is correct, but if we take their words at face value, it appears that they have identified a serious flaw in previous research that resulted in an incorrect estimation of the effect of minimum legal drinking age laws. While I am still hesitant to take their results as true (as I am with any econometric study), they present convincing evidence that, at the very least, the previous research was way off the mark.
 I’m open to the suggestion that these controls aren’t enough.
Views expressed are not endorsed or put forward by Turning Point USA.