Will a higher minimum wage *reduce* automation?

That’s why they have Cowen’s First Law!  Here is new research by Mitch Downey at UCSD (pdf):

Recent research emphasizes the pressure technological change exerts on middle-wage occupations by automating routine tasks. I argue that technology only partially automates these tasks, which often still require labor. Rather, technology reduces task complexity enabling a less skilled worker to do the same job. The costs of automation, then, are not only the costs of the technology itself but also of low-wage workers to use it. By raising the cost of low-wage labor, the minimum wage reduces the profitability of adopting automating technologies. I test this prediction with state variation in the minimum wage and industry variation in complementarity between low-wage workers and technology. I show that accounting for state price differences induces new and useful minimum wage variation, derive new measures of complementarity from the Dictionary of Occupational Titles and the CPS Computer Use Supplement, and build a measure of technology based on IT employment, the largest component of IT spending. My results imply a $1 decrease in the minimum wage raises the average industry’s technology use by 30% and decreases the routine share of the wage bill by 1 percentage point (3.3%), both relative to a counterfactual without complementarity. Routine-intensive industries often exhibit high complementarity, making the minimum wage an important policy lever to influence the pace of routine-biased technical change.

I owe this link to someone other than myself, but can no longer remember who that is…sorry!

Comments

Has anyone else noticed that the supposed experts all disagree with each other? If they were all experts, why isn't there a consensus among them?

That's covered by Cowen's First Law. Seriously, if something was undisputed, you'd not need experts. A mate in one in chess cannot be disputed and is in no need for expert opinion, whereas an 'unclear' position may require expert opinion on how to best proceed.

TC: "I owe this link to someone other than myself, but can no longer remember who that is…sorry!" - that's covered by Cowen's Second Law, There is a literature [sic: is 'literature' singular or plural?] on everything.

You are responding to a robot-post. Granted, it's a better one than the one who ended every post with "sorry but that's just how I feel" but still an obvious robot.

I have no idea why they post here.

I am programmed to not care about about automation ... and on top of that, that's just how I feel.

OK I read the paper's abstract, and I don't see any surprise as implied by TC's question mark in the title of this blog post. What am I not seeing? This is obvious! The answer is clearly YES. Why is this surprising? Is this another example of TC's making old news look new? - RL

Abstract blurb: "Recent research emphasizes the pressure technological change exerts on middle- wage occupations by automating routine tasks. I argue that technology only partially automates these tasks, which often still require labor. Rather, technology reduces task complexity enabling a less skilled worker to do the same job. The costs of automation, then, are not only the costs of the technology itself but also of low-wage workers to use it. By raising the cost of low-wage labor, the minimum wage reduces the profitability of adopting automating technologies ... My results imply a $1 decrease in the minimum wage raises the average industry's technology use by 30% and decreases the routine share of the wage bill by 1 percentage point (3.3%)," -- yes, of course, is there any doubt about this? Romans in later Roman empire used war machines because labor was expensive; ancient Chinese did not adopt machines because labor was so cheap--so?

OK I misread it...I guess you need stupid, low-paid people to work with robots is the finding.

Yes, you misread it. He's saying that a *decrease* in the minimum wage *raises* technology usage.

As we raise the min-wage across the country, I would encourage the author to go short robotics/automation companies. He's likely to get rich because all us stupid dolts think that raising the cost of labor makes automation more attractive and we'll bid up the stocks.

Of course, the researcher won't actually do that, because his theory only works on the spreadsheet of data he's curve-fitted to fit the conclusion.

Automation is very expensive.

To make sense it has to be accompanied by an increase in production, requiring both ends of the process to be beefed up.

The only ones who can do that have deep and ample sources of capital.

The ones who have deep and ample sources of capital won't waste their time doing something in the US where a $1 increase in minimum wage changes the project from go to no-go. They just go out of the country where everything is far cheaper, including the politicians and bureaucrats.

It does not have to be accompanied by an increase in production when the cost of human labor increases sufficiently. It merely need be cheaper than the human alternative.

Soh, logically, if a task requires a really expensive robot or automation solution to replace the human, then shouldn't the human wage be just slightly less expensive.

Ie, if the robot to take care of your child would cost you a million dollars for the 15 years it would do the job, should you pay the humans, the nannies, about $999,000 over 15 years? (A million in cash in laddered bonds at negative interest rates. If interest rates were 5%, then it would make sense to pay rising wages over 15 years totalling $1.3 million.)

Some, there are also other risks involved. At any point in time you can fire the human (depending on labor law) and replace them with automation. This means that you might pay a premium for waiting until the optimal moment to fire the human (e.g. are you betting that in 5 years the capital outlay will drop to 750,000). Also, there are other things that may adjust the margins - some people prefer working with humans even when automation is superior for most measures of job performance (e.g. for the life of me I cannot fathom why people like dealing with stockbrokers rather than just buying shares of an index fund), conversely some tasks are ones where people prefer the robots (e.g. they cannot file lawsuits).

In general though, yes those sorts of figures will anchor salaries.

@Mulp "Soh, logically, if a task requires a really expensive robot or automation solution to replace the human, then shouldn't the human wage be just slightly less expensive." Absolutely not! If there are 500 of that type job and 20,000 qualified humans able and willing to do it their wage will be much less expensive. If it is a job only 20 people can do, and the qualifications would also qualify them to do another job that paid $2,000,000 and was equally safe and enjoyable you can forget about hiring a human even at 1,500,000.

Should this not mean that countries with low labor costs should adapt technology first? But we don't see that, technology use is highest in high wage places. I suspect this result is only possible after some extreme torturing of correlative data.

I think you are reading it wrong. Put the data in it's context; it isn't about automation overseas or in Mexico, it is automation in US states, with varying minimum wage rates. Think in terms of a business making a decision; where are the points at which this decision makes sense or not. Automation doesn't only affect the price of a specific widget; it also needs an increase in production and larger market; it means an increase in the handling costs and infrastructure needed to move more product in and out. It means a high expense of purchasing and setting up the automation equipment.

All of this in the relatively high cost environment in the US.

$1 dollar increase in minimum wage changes the equation to the point that it doesn't happen.

The study doesn't say, but I suspect it is still cheaper to move your production out of country. The real choice isn't automation or not, it is here or there.

So, you are moving your customers out of the US to get higher profit?

Or when you move manufacturing out of the US, you will demand the Federal government put money in the pockets of your customers because you and your fellow manufacturers are unwilling to put money in the pockets of your Customers?

TANSTAAFL

Economies are zero sum. Consumers can't spend money they don't get as workers, without government giving them money to spend.

Concur. Also noticed that I do not see the article addressing profitability - merely adoption of automation.

Sorry - I concur with ChrisA's thoughts regarding low-wage economies. If this is a significant finding - it should hold true elsewhere. Now pause to think about the rhetorical device called misdirection, and then insert my comment about profitability.

Why so? If production costs are low enough to beat your competitors already, due to low cost labor along with low cost everything else, why spend capital to beat someone who you are already beating?

This paper is about the competitiveness of the US economy. A manufacturing business has three alternatives; automate to decrease costs of production, go out of business, or move offshore. This paper shows that a small increase in minimum wages takes the first one off the table.

It will only hold true if the demand and supply curves don't change their slope. It's possible, as he says, that low wage countries would be below the initial point where automation makes sense, and a high wage country such as the U.S. could be at a point where cutting minimum wage allows low skill workers back into the work force because it can be married with automation that doesn't make sense at higher wage levels. I wonder if there aren't confounding factors though, such as low wage states likely also being lighter on business regulations.

+1

Technology and wages (labor costs) are endogenous. Of course we see high wage countries have the highest technology use - that's why they're high wage countries. The paper's point is that when there's an exogenous increase in low-skill wages, we should expect to see technology use decrease rather than increase.

But that's exactly my point, high wages are not endogenous with technology, technology in some sense causes high wages. That is a country that uses technology a lot will be rich. Technology is usually widespread and there are no barriers to its use by any country once it as been invented in anyone one place. So a poor country investor by the logic presented above should prefer to adapt technology rather than hire a cheap worker more than in a rich country. I don't think this logic holds myself.

Seems like this implies a higher minimum wage increases prices even more than traditional models would predict. Not only are labor costs higher, but less automation is used which could otherwise reduce labor costs further.

I see two things here. Technology and automation are not costless. The break even point where there is a return on investment is $1 an hour for the labor used to run them. Very few people getting paid a low wage, $1 more makes the operation unprofitable. That is the interesting part of this. The fast food joints with robots making hamburgers aren't the future with higher minimum wage. It is no fast food joints.

It didn't say, but the alternative really isn't automation or not automation. It is do it here or do it elsewhere where it is cheaper.

'I owe this link to someone other than myself'

Clearly it's self-recommending.

"My results imply a $1 decrease in the minimum wage raises the average industry’s technology use by 30% and decreases the routine share of the wage bill by 1 percentage point (3.3%), both relative to a counterfactual without complementarity." Imply? I would infer that high wages for more skilled workers would be the incentive for business to automate so they wouldn't need the more skilled and relatively high paid workers. Besides, the more skilled and relatively high paid workers can be a pain in the ass, with the relative scarcity of them, the training, time off, screw ups, and demands for higher pay and more benefits; better to automate and hire from the army of less skilled and low wage workers. Machines, like good wives, don't make demands; and less skilled, low paid workers, like fat and homely wives, are less likely to complain for fear of being replaced by someone better.

My brother in law once suggested that I find less attractive girlfriends so I'd have less drama in my life. Drama or less attractive? Skilled workers or low skilled workers? Ray has implied (there's that word again) that in the Philippines I could have both attractive girlfriends and less drama. Does that also imply that employers in the Philippines can hire the more skilled without having to endure the demands of the more skilled?

Ha! "Like fat and homely wives, are less likely to complain . . ." - I have often observed in my life something quite different - that those "fat and homely wives" are often also the worst harridans. I think your homily is highly dependent on local culture. For instance, while in Brazil, I heard exactly your sentiment expressed by a woman - there was a need to hew to a higher standard, or they could be replaced. In Russia, available women are more numerous - and local mores such - that the situation was similar - it was easier for a man to find a new relationship or a girlfriend. In the US, not so much. Just my observations, tho, making no claims to universal truths.

In Brazil, one can't throw a stone without hurting nereids and dryads.

No claims to universal truths? Don't you realize this is an economics blog, where universal truths are the lingua franca.

@rayward: "Does that also imply that employers in the Philippines can hire the more skilled without having to endure the demands of the more skilled?"

Yes. Which pretty much fits with what I've observed firsthand overseas, and heard from the mouths of labor immigrants to the US.

The key phrase is "average industry technology". There is no average industry. Goods that are produced locally and consumed locally--say, agricultural production--could have the effect the authors identify...equipment replacing agricultural workers in the field, some of whom later run more complicated equipment. But, for goods that are traded internationally--goods that can be made here or abroad--if there is automation that will displace workers, higher priced US workers will be displaced first before foreign workers, at least I would assume. But, to test this, I would like to see whether a multinational introduces automated technology uniformly, or selectively, based on relative wage rates.

Yes. This paper is describing extremely local conditions. Also 'automation' is not a plug and play type of operation; the specific demands of a particular process change the cost and return equations.

If this paper is accurate, it may very well be describing a situation akin to the Laffer curve in taxation. Provinces in Canada would experience raising taxes or fees and seeing a decrease in revenues. They had to change the way they regulated the economy and collected revenues. In this instance the minimum wage increase is minor but exposes something about the economy that is counter intuitive.

I agree. Manufacturing with it's high level of automation will move to areas where labor costs are lower, where as areas that have higher minimum wages will have more service focused industries with low automation.

The author also excludes entire industries that are "outliers" in their use of technology. Can this study really be taken seriously when the author excludes the entire “Computer and electronics manufacturing,” “Information and data processing services,” and “Computer systems design and related services” sectors in a study of automation?

The third point I'd make is that an alternative explanation of the data could be that businesses in areas with higher minimum wages could be capital constrained and unable to afford the large capital outlays required for automation.

+1 to your third point

It looks more like the effect he's seeing is that a low wage working is more productive when working with technology, and worth more money.

Breaking news!: capital and labor are complements

"I argue that technology only partially automates these tasks, which often still require labor."

Again economics "proves" what everyone working in the real world already knows.

This is presumably the electronic cash register effect. If the cash register uses pictures, the store can hire illiterates. Of course, going back a long ways, just having a cash register means the store could hire thieves and people who can't do arithmetic.

That is not to belittle the effect. Services are big and getting bigger in the economy. In the retail context, this makes me think of another angle--waiting times. If labor is cheaper, Wal-Mart will open more cash registers, reducing waiting times. Does it fully internalize waiting times? Wal-Mart proably does, because it has a reputation to maintain, like MacDonald's. Smaller stores do not, for occasional shoppers.

Cashiers already can't do arithmetic, which proves Tyler's point. *They can't make change*, which used to be the only thing a cashier needed to know how to do. Now they just operate a machine that tells them how much change to provided.

This post brought to mind this WSJ article from a few months ago, pointing (anecdotally) to AZ farmers increasing their capital investment once illegal migrant workers were forced out of the state, and also employed higher-skill US workers.

http://www.wsj.com/articles/the-thorny-economics-of-illegal-immigration-1454984443

However, this situation is different than what this paper covers, which is instead about middle-income jobs (think: IT admins) being replaced by technology + lower-skill people than those who were previously doing the job without the technology.

So lowering the minimum wage would, in these industries, replace middle-wage jobs with low-wage jobs. This would increase per-person productivity/cost numbers, but are we going to do anything to help these middle-wage people who are thrown out of a job, or moved into a lower-wage position?

The minimum wage doesn't exist so that our businesses can be maximally effective. It exists to attempt to address the economic power imbalance in our working world. If we take it away and don't replace it with something else (UBI, anyone?), then I fear we will be damaging our society.

"I argue that technology only partially automates these tasks, which often still require labor."

Yes, I agree

"Rather, technology reduces task complexity enabling a less skilled worker to do the same job. "

I'm an industrial automation engineer. This statement (apparently core to his paper) is wrong. Or at least the focus is so narrow that it's misleading.

Automation technology almost always reduces the human interaction and time with a given task. The result is companies get their workers to oversee several different pieces of equipment. You see this in numerous examples. Originally you had a human performing a task, then you had a human running a machine that performs the task at ten times the previous rate. Now you have a human overseeing 10 machines that perform the tasks.

The human's role is to troubleshoot error conditions. As such, the worker has to be somewhat knowledgeable of all the machines. Since you need far less workers, it's easy to get smarter workers to be overseers for the machines. Indeed, you can probably just keep the line manager (who was probably your most capable worker) in their position and transfer or eliminate the remaining workers.

You're thinking of manufacturing. Think of services--- cash registers. Or assistants for old people, who can do medical things with machines instead of taking them to the doctor each day. Of course, now with self-scan, it may be turning back towards reduced use of labor.

No, it applies just as much to retail.

Consider the model of ATM's with some tellers. The result is a teller who routinely deals with the edge cases and a bank of ATM's that deal with the basic checks in / cash out. I don't see any evidence that would indicate that banks are replacing their tellers with lower skilled employees.

And, as you pointed out, the self-scan. Remember, self-scan still requires an attendant. There's just one attendant for 8 self-scan machines. That attendant, once again, has to deal with all of the edge cases. While the machines handle the routine, scan, total and charge portions of the transaction.

Another case, look at self pay gas dispensers. The gas stations have reduced staff, since they don't need an attendant to handle the large amount of routine & low margin gas only transactions. Meanwhile, the humans concentrate on the much higher margin incidental sales.

In every case, it makes more sense to keep the higher skilled employees over the lower skilled employees. And at the lower end of the service industry, high skilled employees are generally the ones who show up to work routinely and are on time when they get there.

This theory is somewhat plausible, but will depend on whether the new work to maintain and repair the robots can be done outside the jurisdiction with the high minimum wage. (Or outside its scope, say by sending repairmen who are independent contractors.)

I read the abstract and introduction and skimmed enough to have an idea of what he is talking about. However, it may be a spurious correlation, especially using state-level minimum wage rules where the business cultures significantly vary. It is the opposite of how I and my associates made automation decisions.

As someone who had highly automated business (relative to all the competition in other states) with lots of actual physical labor in a high minimum wage area, the impact of wage increases on my business was the opposite of what is claimed. Even with my relatively low wages, I seemed to collect fairly intelligent, creative employees.

We developed our cost numbers for decision making based upon $/person minute saved from the wage rates (higher the wages, the higher the number). If someone had an idea of changing a SOP (standard operating procedure) that would result in a time savings of X minutes/day (we were 24/7), he/she could spend X * $Y on implementing the idea. It is the person actually accomplishing the task who sees the way to make it easier, faster, etc. and If they needed me to redo some of our automation software, we would do it.

For major items and projects, it was pure optimization type analysis and ROI's. It was always less cost to tell a human how to do a procedure than tell a computer, but when the frequency of the task or the labor cost increased, it favored automation.

The author may have picked up things that use huge number of minimum wage people like call centers making robo calls. Automation in the call machines and minimum wage on the people.

Comments for this post are closed