Be careful what you wish for with minimum wage hikes

Eighteen states began 2018 with higher minimum wages. Thirty states (including the District of Columbia) now have minimum wages higher than the federal minimum, ranging as high as $11.50 in Washington State and $12.50 in the District of Columbia.

By 2020, Arizona, California, Maine, Oregon and New York will all have minimum wages at or above $12.00.

{mosads}Minimum wage increases have a significant impact on the wage bill of firms reliant on minimum-wage workers in industries like services and transport, where margins are already tight. Notably, the cost of robots have been decreasing and robotics have been gaining capabilities.

There is also an increasing supply of skills to create, troubleshoot and maintain robotics as part of a production process among millennial workers. Right now, there are many occupations that low-skill workers rely upon that are on track to be automated.

These include delivery drivers (Waymo continue to strive in driverless technologies despite Uber and Tesla’s setbacks), cashiers (see Amazon GO) and bricklayers (Hadrian X is a robot that lays 1,000 standard bricks in 60 minutes).

My recent paper with David Neumark in labor economics is therefore timely given that we investigate whether minimum-wage changes in the past have caused an acceleration of automation that changed the type of work available for low-skilled workers.

In the past, jobs that can be replaced by technology have largely been those that involve a repeated sequence of actions and are easily codifiable.

Drawing on U.S. population data from 1980-2015 and a classification for automatable work introduced by Autor and Dorn (2013), we find that a 10-percent increase in the minimum wage leads to a 0.31 percentage point decrease in the share of automatable jobs done by low-skilled workers overall.

The effects are largest in manufacturing: A 10-percent increase in the minimum wage leads to a 0.73 percentage point decrease in the share of automatable jobs done by low-skilled workers.

Notably, we also find the largest effects for older, black and female workers. Overall, we find strong evidence that minimum wage increases are changing the type of work available for low-skilled workers by accelerating automation.

We also consider a more recent time period, 1995-2016, as we expect that effects will be stronger. This arises as technology has been getting cheaper and labor relatively more expensive over time. As expected, the estimates are stronger.

For example, we find that a $1 increase in the minimum wage leads to a 0.57-percentage-point decrease in the share of automatable jobs done by low-skilled workers (as compared to a 0.43-percentage-point decrease above).

We next modeled directly whether a higher minimum wage actually increases disemployment among low-skilled workers who were in jobs with routine tasks compared to those in non-automatable work.

Here, we find strong evidence indicating that the negative effects on employment shares in automatable jobs are in fact also associated with job loss and transitions to dis-employment among workers doing automatable jobs.

Once again we find differential effects by demographic groups, with the youngest and oldest workers most adversely affected and females more adversely affected than males.

In separate analyses, we show that low-skilled individuals working in automatable jobs are also more likely to work less hours and/or switch to non-automatable work following minimum wage increases. Switching jobs is costly to both individuals and to firms.

Overall, our work suggests that low-skilled individuals are vulnerable to their job disappearing quicker because of minimum wage increases. It is intuitive that the jobs lost to automation will be met with the creation of new jobs that require very different skills.

For instance, employers may substitute away from labor with technological innovations, such as supermarkets substituting self-service checkout for cashiers and assembly lines in manufacturing plants substituting robotic arms for workers.

At the same time, these firms may hire individuals to service, troubleshoot and maintain the new technology. It is very likely that older, low-skilled workers may be the biggest losers in such transitions, as they may not get so many opportunities to retrain and learn new skills.

It is also possible that we will end up with significantly fewer jobs as automation continues. Just because all jobs lost have been replaced in the past does not mean that this will continue to occur in the future.

Replacement becomes less likely as machines continue to learn and minimum wages put pressure on companies already operating in highly competitive environments to substitute workers for technologies.

So, there is a role for policy in the ongoing monitoring of trends and to consider how the rents earned by machines should be redistributed within society if technology adoption accelerates to a point where too many jobs have disappeared.

Grace Lordan is an associate professor of behavioral science at the London School of Economics and Political Science.