Min wg panel model

September 26th, 2018 at 10:56 pm

I’ve got a piece in the WaPo today that presents findings from the model described in more detail here. What follows is for those who want the technical details of the data and model. As is unlike my wont, I make no effort to explain in layperson’s terms.

The model is a panel of the 50 states plus DC from 1979 to 2017 of wage, labor market, and minimum wage data. The results derive from fixed effects, panel regressions of the change in the log of the state real, 10th percentile wage on state unemployment and the change in the minimum wage variable. For this variable, I used the ratio of the state minimum wage in state i, year t, divided by the federal minimum wage in year t. Standard errors are clustered at the state level (results below).

The article mostly focuses on the first two rows of results below: outcomes for the log change in the real 10th percentile wage. But I also include the same regression for the 90th percentile wage, as a confirmation that the minimum wage variable is picking up variation in low wages. In fact, it is insignificant for the 90th percentile wage.

The WaPo article describes the results from a few simulations of the model over the past few years. To capture the impact of falling unemployment, all else equal, I held the unemployment rate at its 2015 level for each state in 2016-17, while the other variables took their actual values. I take simulated real wage levels and compared them to the path of actual wages.

The same procedure was followed for the change in the minimum wage ratio (state over federal) i.e., I did not allow it to go up in states that raised their minimum wages in 2015 or 2016. I then compared these simulated wages for states where the ratio rose to the actual wage trends for those states (which included any wage effects for the higher state minimum wage).

The full model for the wage results below had 1,938 observations and R-sq of 0.43. Restaurant employment data were only available for 21 states since 1990 (518 observations).

Wage data were derived by the Economic Policy Institute from CPS ORG files; state unemployment is from the BLS; state employment data were from the BLS CES survey; minimum wage data from Ben Zipperer (thanks to Ben for helpful comments, though any mistakes are mine).

Some key results follow (***=significant and <1% level):


Print Friendly, PDF & Email

3 comments in reply to "Min wg panel model"

  1. Kevin Rica says:

    This Wall Street Journal piece, “Labor Shortage Lifts Wages on Europe’s Eastern Flank” https://www.wsj.com/articles/labor-shortage-lifts-wages-on-europes-eastern-flank-1538049601?mod=hp_listc_pos1 makes clear how to raise wages.

    You create a real LABOR SHORTAGE. If you actually want wages to increase, as the East Europeans do, you don’t open the door to more immigrants so that you can prevent a labor shortage from raising wages.

    You can do one thing or the other. You can either 1. maintain the labor shortage to raise wages (if you really want to raise wages) or 2. open the door to let immigrants in to kill the labor shortage in order to prevent wages from rising (as the business lobby wants to because they want low wages). In case #2, you have to let rising wages decide which jobs you keep because they are productive enough to pay hig wages and which jobs are lost because no one wants to fill them at low wages.

    But you have to make up your mind whether you want higher wages or not. But if you let immigrants in, you can’t sit there and wonder why wages aren’t going up.

    Implicit with that is that you have to stand up and denounce WaPo and NYT journalists who insist that labor shortages are a bad thing that requires more immigrants.

  2. Frank Stricker says:

    Have no idea what this post is all about. Only for statisticians? I am interested in wage changes and write about the issue sometimes. But needed some help here.

    • Jared Bernstein says:

      Sorry–this was a rare, technical post for those wanted statistical background for the WaPo piece.