Economists Do It With Models

Warning: “graphic” content…

Bookmark and Share
Goldbricking In A Machine Shop, And At Micro Center…

June 24th, 2009 · 8 Comments
Behavioral Econ · Incentives

You, as my dear readers who have been putting up with me taking iPhone pictures of my hand-drawn graphs, will be pleased to learn that I finally went and bought a scanner. It was a surprisingly painless process, and not nearly as fraught with the perils of overchoice as I would have guessed. I was apparently in a good mood when I reached the checkout, since when the (very nice) cashier asked me for my email address, it didn’t even occur to me to not give it to her. This is probably why I was very confused when, as I started spelling the address, the cashier whispered to me “decline.” It didn’t occur to me what she meant, since I’ve been in grad school long enough for it to not be outside the realm of possibility for the word decline to be related to my credit card. Luckily for me, she eventually explained that I should decline giving her my email address. Still feeling agreeable, I said “uh, ok, decline,” and she literally typed “DECLINED” into the email field. I must have still looked confused, because she then explained to me that she had been asking people for their email addresses all day and no one had declined yet, despite the fact that she had had over 100 customers, so her manager was going to think she was doing something wrong.

I had to try really hard not to laugh, since I *was* at Micro Center and thus am pretty sure that the only thing she was guilty of was being a cute girl in a land of nerdy dudes. I then got some mental image of the poor girl’s manager accusing her of holding a gun to her customers’ heads until they gave up their email addresses. What was going on here? Clearly the store had given cashiers the directive of “ask people for their email addresses,” and perhaps even given an incentive for them to do so, so why was this cashier worried about doing her job too well? (In fairness, I am pretty sure that she phrased the question as “What is your email address?” rather than “Can I have you email address?”, but there was no holdup involved.)

At the risk of seriously disappointing you, I have to admit that I do not have an answer to this question. But this story reminded me of a sociology paper I read for one of my grad school classes. The paper, entitled “Quota Restriction and Goldbricking in a Machine Shop” (Donald Roy, American Journal of Sociology, 1952), is a first-hand account of the behavioral response to monetary incentives. From an article by Roger Martin on Monetary Incentives:

“Donald Roy’s machine shop was a classic piece-work shop of its day. Its machine tool operators fabricated a wide variety of items at their stations, with each assigned a point value based on management’s assessment of the level of difficulty required to produce the item, as assessed by engineers from the ‘timing department’. If an operator turned in 125 points in an hour, he (all were male) would earn $1.25 for that hour of work.The company provided the operators with base pay of $0.85 per hour, such that if the points they turned in were lower than 85, they would still earn the minimum amount. For anything over 85 points per hour, the direct monetary incentive was that one more point generated one more cent in hourly compensation.

Roy observed that the incentives embedded in the compensation structure generated behaviours that at first seemed quite odd. The largest fraction of work turned in (47 per cent of total hours) was for point totals between 115 and 134.The second largest fraction (24 per cent) was between 35 and 54 points. Every other range had tiny proportions of the total hours (see Exhibit One).

What caused these two high-frequency ranges at such odd levels? It turns out that amongst the workers, jobs associated with these ranges were referred to as ‘gravy’ and ‘stinker’, respectively.‘Stinker jobs’ were so difficult – i.e. their point totals were too low in relation to the work required – that the operator declared defeat immediately and slacked off – or ‘goldbricked’ to use the vernacular. But why not goldbrick entirely, not turn in 35 to 54 points? As Roy found out, any operator who turned in less than 35 to 54 points in an hour risked being fired for incompetence. So the extreme of goldbricking was in the range of 35-54 points per hour, which netted the basic $0.85 compensation.

In contrast, ‘gravy jobs’ were so easy in relation to their point values that the minimum could be reached with only a modicum of effort. But why stop at accumulating 134 points per hour? Roy found
out – by himself hitting 150 on a gravy job and being accosted by angry coworkers – that whenever a job yielded more than about 134 points per hour, the timing department would descend on the shop and lower – often dramatically – the point value for the job. So while the monetary incentive to keep working until accumulating 134 points was high, the monetary disincentive to go even a point past 134 was even higher. As a result, the operators engaged in ‘quota restriction’, by which they would work at an artificially slow pace or loaf for hours at a time to restrict themselves from turning in a point total over the informal maximum.

Exhibit One (page 7) overlays the true compensation structure – including threat of termination below 35 points per hour and threatened re-timing above 134 points per hour – on the distribution of work to demonstrate the power of monetary incentives to drive behaviour patterns that appear inexplicable at first blush. Roy estimated that goldbricking and quota restriction – which, within a month of joining the shop, he engaged in as dutifully as his co-workers – caused the machine shop to work at approximately half of its potential productivity, all a direct function of the structure of the monetary incentives.”

You will notice that the quote refers to an Exhibit One that is not included. To celebrate the arrival of my new scanner, I figured I would add one of my own:

How could the managers do better? (I assume that removing the guaranteed payout is not an option due to minimum wage laws.) A lot of the problem comes from the fact that the managers don’t commit to the piece rates. Committing to piece rates for the gravy jobs would solve the incentive problem on the upper end, even though it might result in overly generous compensation, but what about the lower end? One option would be to commit to only revise piece rates upwards, not downwards. (Note that you would effectively get a real piece rate decrease by not changing the point value if there is inflation.) But even that is not really ideal, since the managers don’t know what the “correct” piece rate is in the first place.

That last statement is actually the crux of the problem. How on earth are managers supposed to get an objective measure of how long things take or how hard they are? This is where thinking about the incentives of others becomes so important. You don’t want to ask current employees to be the guinea pigs for timing the tasks, since they have an incentive to make the tasks seem like they take longer than they actually do. It’s hard (read, expensive) to bring in random people off of the street to time the tasks, since you would need to train them to do the job all the while having them know that they would not be employed by the firm in the long term.

Economists refer to incentives where people can’t game the system as “strategy-proof.” Is there a strategy-proof incentive that could be offered in this case?

Tags: Behavioral Econ · Incentives

8 responses so far ↓

  • 1 Tripp Little // Jun 24, 2009 at 6:56 pm

    If the quota for each job is the expected norm, and if not meeting the quota carries the threat of job loss, why not only reward those who exceed the identified quota and not fire those employees who merely meet expectations? In other terms:
    quota : bonus

  • 2 Tripp Little // Jun 24, 2009 at 7:00 pm

    The above didn’t post correctly. It should read:

    Less than quota -> “you’re fired”

    Meet the quota -> job security

    Exceed the quota -> bonus $

  • 3 Erin Monahan // Jun 24, 2009 at 7:14 pm

    Micro Center is still in business?!

  • 4 SteveO // Jun 24, 2009 at 9:44 pm

    And I still have friends who refuse to believe people respond to incentives.

    I was discussing the feasibility of capital punishment as a deterrent for rape. My friend argued that rapists don’t respond to rational analysis.

    Not a minute later she offered on her own that perhaps the rapist would kill the woman for fear of being caught and punished.

    (When you throw everything at an argument, some of your rhetoric is bound to conflict with the rest of your arsenal.)

    I’m not so sure there is such a thing as game proof. Making someone a Residual Claimant is about the best you can do, but there are difficulties in employing that all the way through the system.

  • 5 Thomas // Jun 25, 2009 at 10:47 am

    If you could keep the individuals results private, you could easily pair random employees each pay period and set up a prisoner’s dilemma to maximize output and increase efficiency as well. However, if you can’t keep the results private, the workers will self regulate to generate a low output, low efficiency result.

  • 6 Sam Kinney // Jun 25, 2009 at 5:39 pm

    Auction off the jobs.

    It sounds to me like the employees are engaged in a kind of arbitrage against a set of artificial and fixed prices, bounded on the low end by termination and the upper end by a step change in the price. By auctioning them, you allow prices to fluctuate based on their difficulty and other factors like how busy the stop is.

    To reduce collusion (enforced at knife point in old machine shops), offer jobs in random bundles so the exact “point value” of a given job remains obscured, apparent only to the machinist good at running regressions.

    A rank-order auction that obscures clearing prices from non-winners can also help reduce collusion, although the knife point effect can still be real.

    In the 2009 environment, I’d consistently outsource some portion of these jobs to reduce the knife-point enforcement of the collusion. It’s harder to point your knife at the shop down the street or across the ocean.

    And do a Jack Welch evaluation system, where every 6 months, the lowest-earning 5% of workers gets let go. This ties termination to a longer pattern of productivity over multiple jobs, rather than tying it to a specific job.

    Don’t tell the union steward my home address, please.

  • 7 Pfloyd // Jun 28, 2009 at 6:45 pm

    I like the cut of Sam’s jib.

  • 8 jual mesin // Apr 29, 2014 at 9:13 am

    I have read so many articles about the blogger lovers however this paragraph
    is really a pleasant paragraph, keep it up.

Leave a Comment