I have not a shred of doubt that the minimum wage increases unemployment, especially among the most vulnerable group of workers: males aged 16 to 19.
Anyone who claims that the minimum wage does not affect unemployment among that vulnerable group is guilty of (a) ingesting a controlled substance, (b) wishing upon a star, or — most likely — (c) indulging in a mindless display of vicarious “compassion.”
Economists have waged a spirited mini-war over the minimum-wage issue, to no conclusive end. But anyone who tells you that a wage increase that is forced on businesses by government will not lead to a rise in unemployment is one of three things: an economist with an agenda, a politician with an agenda, a person who has never run a business. There is considerable overlap among the three categories.
I have run a business, and I have worked for the minimum wage (and less). On behalf of business owners and young male workers, I am here to protest further increases in the minimum wage. My protest is entirely evidence-based — no marching, shouting, or singing for me. Facts are my friends, even if they are inimical to Left-wing economists, politicians, and other members of the reality-challenged camp.
I begin with time series on unemployment among males — ages 16 to 19 and 20 and older — for the period January 1948 through June 2009. (These time series are available via this page on the BLS website.) If it is true that the minimum wage targets younger males, the unemployment rate for 16 to 19 year-old males (16-19 YO) will rise faster or decrease less quickly than the unemployment rate for 20+ year-old males (20+ YO) whenever the minimum wage is increased. The precise change will depend on such factors as the propensity of young males to attend college — which has risen over time — and the value of the minimum wage in relation to prevailing wage rates for the industries which typically employ low-skilled workers. But those factors should have little influence on observed month-to-month changes in unemployment rates.
I use two methods to estimate the effects of minimum wage on the unemployment rate of 16-19 YO: graphical analysis and linear regression.
I begin by finding the long-term relationship between the unemployment rates for 16-19 YO and 20+ YO. As it turns out, there is a statistical artifact in the unemployment data, an artifact that is unexplained by this BLS document, which outlines changes in methods of data collection and analysis over the years. The relationship between the two time series is stable through March 1959, when it shifts abruptly. The markedness of the shift can be seen in the contrast between figure 1, which covers the entire period, and figures 2 and 3, which subdivide the entire period into two sub-periods.
For the graphical analysis, I use the equations shown in figures 2 and 3 to determine a baseline relationship between the unemployment rate for 20+ YO (“x”) and the unemployment rate for 16-19 YO (“y”). The equation in figure 2 yields a baseline unemployment rate for 16-19 YO for each month from January 1948 through March 1959; the equation in figure 3, a baseline unemployment rate for 16-19 YO for each month from April 1959 through June 2009. Combining the results, I obtain a baseline estimate for the entire period, January 1948 through June 2009.
I then find, for each month, a residual value for unemployment among 16-19 YO. The residual (actual value minus baseline estimate) is positive when unemployment among 16-19 YO is higher than expected, and negative when 16-19 YO unemployment is lower than expected. Again, this is unemployment of 16-19 YO relative to 20+ YO. Given the stable baseline relationships between the two unemployment rates (when the time series are subdivided as described above), the values of the residuals (month-to-month deviations from the baseline) can reasonably be attributed to changes in the minimum wage.
For purposes of my analysis, I adopt the following conventions:
- A change in the minimum wage begins to affect unemployment among 16-19 YO in the month it becomes law, when the legally effective date falls near the start of the month. A change becomes effective in the month following its legally effective date when that date falls near the end of the month. (All of the effective dates have thus far been on the 1st, 3rd, 24th, and 25th of a month.)
- In either event, the change in the minimum wage affects unemployment among 16-19 YO for 6 months, including the month in which it becomes effective, as reckoned above.
In other words, I assume that employers (by and large) do not anticipate the minimum wage and begin to fire employees before the effective date of an increase. I assume, rather, that employers (by and large) respond to the minimum wage by failing to hire 16-19 YO who are new to the labor force. Finally, I assume that the non-hiring effect lasts about 6 months — in which time prevailing wage rates for 16-19 YO move toward toward (and perhaps exceed) the minimum wage, thus eventually blunting the effect of the minimum wage on unemployment.
I relax the 6-month rule during eras when the minimum wage rises annually, or nearly so. I assume that during such eras employers anticipate scheduled increases in the minimum wage by continuously suppressing their demand for 16-19 YO labor. (There are four such eras: the first runs from September 1963 through July 1971; the second, from May 1974 through June 1981; the third, from May 1996 through February 1998; the fourth, from July 2007 to the present, and presumably beyond.)
With that prelude, I present the following graph of the relationship between residual unemployment among 16-19 YO and the effective periods of minimum wage increases.
The jagged, green and red line represents the residual unemployment rate for 16-19 YO. The green portions of the line denote periods in which the minimum wage is ineffective; the red portions of the line denote periods in which the minimum wage is effective. The horizontal gray bands at +1 and -1 denote the normal range of the residuals, one standard deviation above and below the mean, which is zero.
It is obvious that higher residuals (greater unemployment) are generally associated with periods in which the minimum wage is effective; that is, most portions of the line that lie above the normal range are red. Conversely, lower residuals (less unemployment) are generally associated with periods in which the minimum wage is ineffective; that is, most portions of the line that lie below the normal range are green. (Similar results obtain for variations in which employers anticipate the minimum wage increase, for example, by firing or reduced hiring in the preceding 3 months, while the increase affects employment for only 3 months after it becomes law.)
Having shown that there is an obvious relationship between 16-19 YO unemployment and the minimum wage, I now quantify it. Because of the distinctly different relationships between 16-19 YO unemployment and 20+ YO unemployment in the two sub-periods (January 1948 – March 1959, April 1959 – June 2009), I estimate a separate regression equation for each sub-period.
For the first sub-period, I find the following relationship:
Unemployment rate for 16-19 YO (in percentage points) = 3.913 + 1.828 x unemployment rate for 20+ YO + 0.501 x dummy variable for minimum wage (1 if in effect, 0 if not)
Adjusted R-squared: 0.858; standard error of the estimate: 9 percent of the mean value of 16-19 YO unemployment rate; t-statistics on the intercept and coefficients: 14.663, 28.222, 1.635.
Here is the result for the second sub-period:
Unemployment rate for 16-19 YO (in percentage points) = 8.940 + 1.528 x unemployment rate for 20+ YO + 0.610 x dummy variable for minimum wage (1 if in effect, 0 if not)
Adjusted R-squared: 0.855; standard error of the estimate: 6 percent of the mean value of 16-19 YO unemployment rate; t-statistics on the intercept and coefficients: 62.592, 59.289, 7.495.
On the basis of the robust results for the second sub-period, which is much longer and current, I draw the following conclusions:
- The baseline unemployment rate for 16-19 YO is about 9 percent.
- Unemployment around the baseline changes by about 1.5 percentage points for every percentage-point change in the unemployment rate for 20+ YO.
- The minimum wage, when effective, raises the unemployment rate for 16-19 YO by 0.6 percentage points.
Therefore, given the current number of 16 to 19 year old males in the labor force (about 3.3 million), some 20,000 will lose or fail to find jobs because of yesterday’s boost in the minimum wage. Yes, 20,000 is a small fraction of 3.3 million (0.6 percent), but it is a real, heartbreaking number — 20,000 young men for whom almost any hourly wage would be a blessing.
But the “bleeding hearts” who insist on setting a minimum wage, and raising it periodically, don’t care about those 20,000 young men — they only care about their cheaply won reputation for “compassion.”
A relevant post by Don Boudreaux:
Here’s a second letter that I sent today to the New York Times:
Gary Chaison misses the real, if unintended, lesson of the Russell Sage Foundation study that finds that low-skilled workers routinely keep working for employers who violate statutory employment regulations such as the minimum-wage (Letters, September 8). This real lesson is that economists’ conventional wisdom about the negative consequences of the minimum-wage likely is true after all.
Fifteen years ago, David Card and Alan Krueger made headlines by purporting to show that a higher minimum-wage, contrary to economists’ conventional wisdom, doesn’t reduce employment of low-skilled workers. The RSF study casts significant doubt on Card-Krueger. First, because the minimum-wage itself is circumvented in practice, its negative effect on employment is muted, perhaps to the point of becoming statistically imperceptible. Second, employers’ and employees’ success at evading other employment regulations – such as mandatory overtime pay – counteracts the minimum-wage’s effect of pricing many low-skilled workers out of the job market.
Donald J. Boudreaux