Rudimentary Foray into the Minimum Wage & African-American Unemployment

A commonly cited cause of relatively high African-American unemployment rates is the minimum wage. This argument is often made by libertarians. To detractors: it’s not just white, middle-class males who argue this. Walter Williams is well known for making the case — see his book, Race and Economics. Thomas Sowell, as well. The relationship is intuitive, if you assume that African-Americans, on average, are less productive than whites (and other ethnicities/races with lower unemployment rates) and perfect competition is the best model to apply to low-skill labor markets. My prior is that binding minimum wages do reduce unemployment, and so I’ve too repeated the argument, most recently today.

Knowing that intuition is deceiving, I want to develop a model to test — nothing too fancy, but not necessarily primitive, either. I don’t have that model yet. I’m looking at what data is available, what data is the best to use, and how I should manipulate the data, based on the model (e.g. should I compare rates of change or levels? what should be the functional form?).

In the meantime, here is the quick and dirty result, from the data I do have. All data is for California, including the African-American unemployment rate (BLS) and the inflation (CPI, less food and energy) adjusted minimum wage. The two compared,

Percent Change Minimum Wage v Percent Change African-American Unemployment (California)

The black trendline includes all data points (the blue diamonds); the red trendline excludes the highest blue diamond (-0.01942, 0.1963). The R2 are .0365 and .0311, respectively. Very low. Indeed, percent change in the minimum wage is not statistically significant. Also not statistically significant is the effect on the inflation adjusted minimum wage on the unemployment rate, and neither is the effect of ln(adjusted minimum wage) on ln(unemployment rate) — for those who may not know what the difference is, this last one measures elasticity; i.e. a one percent change in the real minimum wage causes a percent change in unemployment equal to the coefficient estimated for the former.

A “quick look” suggests that maybe the relationship between the minimum wage and African-American unemployment isn’t so obvious (or intuitive), after all.

2 thoughts on “Rudimentary Foray into the Minimum Wage & African-American Unemployment


    Hey Jonathan,

    I think the percent changes you are analyzing are insignificant and are therefore too overly influenced by other factors to see such a correlation. I like the parallel that Sowell makes between Switzerland and USA more, where Switzerland has no minimum wage and 3% unemployment, very similar the USA when they did not have a minimum wage. In order to really see a significant change you would need to conduct an experiment that involves looking at an economy with 0% minimum wage vs. 10% minimum wage and you would see a very distinct difference. Better yet, make it 50 or 60%. Of course, lack of a time machine makes this experiment very difficult.

    The anti-libertarians concede on the logic for a 50 or 60% change but do not concede on the same logic for a 1 or 2% change, which is more telling IMO. Economics tells us that any minimum wage will make us poorer on average no matter what the percentage.

  2. M.H.

    Even a low r such as r=0.10 can be meaningful. You should probably try to perform multiple regressions and examine the unstandardized beta, not the standardized. For example, imagine your dependent var is wage where each unit is 1000 dollars earned per year. In your set of independent predictor variables, say, you control for a bunch of SES and demographic variables, age, year of experience at work, and some other things. You want to see if being tall correlates with earnings, net of all other variables. Your “height in inches” variable is measured in inches so it is coded as 1 unit equals 1 inches. In other words, for example, if your unstandardized B for height is, say, 0.40, that means a gain in one inch will earn you 400 dollars per year. Or, that 10 inches gain will earn you 4000 dollars per year. I believe the effect is large, here.

    Why I’m telling this, is also because i have read lot of regression analyses where the standardized coefficient is very low, somewhat, 0.05, and even such a low correlation can be meaningful when you look at the unstandardized regression coefficient.

    However, in your plot, I don’t think there is a relationship here, not because of your low correlation, for reasons mentioned above, but because you have a lot of “outliers”. This is largely sufficient to reject the plausible correlation between the two variables.


Leave a Reply to Cancel reply

Your email address will not be published. Required fields are marked *