Incentive theory

[From Bruce Abbott (971130.1600 EST)]

Bill Powers (971130.1240 MST) --

Fred Nickols (971130.1341 ET)

PCT predicts x + 10 after the new lathe. So does Nickols. So would almost
anyone I know who is familiar with incentives. I can't say why PCT predicts
that except to guess that the potential of the new lathe will not be
realized because the lathe operator is controlling for a rate of x + 10 and
simply introducing a new lathe doesn't alter that.

That's correct. If you solve the control equations, you find that doubling
the output amplification simply halves the error signal, having no effect
on the reference signal. The lathe operator would just work half as hard.
To make any change, as you suggest further on, you would have to persuade
the operator (and presumably the union) to raise his reference level for
parts produced. In other words, you have to deal with a higher level
control system in the operator.

Why is this _the_ correct answer? Perhaps the lathe operator adjusts his
efforts so as to maintain a certain cost/benefit ratio. He can now make
twice as many parts at the same cost in effort, but this results in a higher
ratio of benefit to cost. He can restore the same cost/benefit ratio by
increasing his effort, thus raising output to greater than 2(x + 10) parts
per hour.

In reality, employees would be more likely to work no harder than required
for a given pay, thus maximizing the benefit to cost ratio. The incentive
for raised productivity builds in a bonus in pay to offset the required
increase in effort -- an offer which the operator may or may not find
attractive. If pay does not depend on rate of output, the operator's best
deal in terms of cost/benefit ratio is to make no parts at all, regardless
of what lathe he uses. In that case the employer is left with the option of
convincing the operator to work harder (appealing to his sense of fairness,
for example), changing the costs associated with not working hard enough
(docking pay, threatening to fire the operator), or changing the way the
operator is paid (e.g., piece rate instead of salary). Incentive theory
takes account of the individual's subjective costs and benefits to predict
what the individual will do. I believe it assumes that these inputs yield
an equilibrium value, which depends on the values of those inputs as the
individual perceives them.

Regards,

Bruce

[From Bruce Gregory (971130.1620 EST)]

Bruce Abbott (971130.1600 EST)

In reality, employees would be more likely to work no harder than required
for a given pay, thus maximizing the benefit to cost ratio.

Any evidence to support this conjecture?

Bruce

[From Bruce Abbott (971130.2155 EST)]

Bruce Gregory (971130.1620 EST) --

Bruce Abbott (971130.1600 EST)

In reality, employees would be more likely to work no harder than required
for a given pay, thus maximizing the benefit to cost ratio.

Any evidence to support this conjecture?

I phrased this in a way that probably could be misunderstood. I'm
conjecturing that employees would be more likely to work no harder than
required for a given pay than to restore a cost-benefit ratio in which
benefits are too high by increasing their costs. Certainly there are other
motivations for doing one's job besides the employer's offering of pay. For
example, one may be willing to work hard for someone who seems to appreciate
one's work, one may like the work itself, and so on. Evidence? I don't
know -- I'm not "up" on this area of research. However, I might guess that
much of the former Soviet economy may provide an example: supposedly, a lot
of people did no more than the minimum required because working any harder
than that provided no incremental benefits to the individuals doing the labor.

Regards,

Bruce

[From Bill Powers (971201.1259 MST)]

Bruce Abbott (971130.1600 EST)--

Fred Nickols (971130.1341 ET)

PCT predicts x + 10 after the new lathe. So does Nickols. So would almost
anyone I know who is familiar with incentives. I can't say why PCT

predicts

that except to guess that the potential of the new lathe will not be
realized because the lathe operator is controlling for a rate of x + 10 and
simply introducing a new lathe doesn't alter that.

That's correct. If you solve the control equations, you find that doubling
the output amplification simply halves the error signal, having no effect
on the reference signal. The lathe operator would just work half as hard.
To make any change, as you suggest further on, you would have to persuade
the operator (and presumably the union) to raise his reference level for
parts produced. In other words, you have to deal with a higher level
control system in the operator.

Why is this _the_ correct answer?

Fred is correct in saying that this is kind of answer that PCT would
suggest, not that the answer is objectively correct. You propose a
different controlled variable, cost-benefit ratio, which is also a
possibility, along with cost-benefit _difference_, which would be more
convenient as it is linear. However, cost-benefit models have a great
problem in general, because there's no a-priori way to predict just what
cost-benefit ratio, or difference, would be the goal. Most people are not
willing to work in a break-even situation, where the cost equals the
benefit with nothing left over. So there is no neat answer to whether the
lathe operator would maintain benefits in some particular relation to
costs. Furthermore, my proposal, aside from its truth or falsity, at least
assumes that the immediate reference condition has to do with something
observable, the achieved rate of production. To test yours you would have
to find a way of observing both costs and benefits, which might be
difficult. Of course you could be right and I could be wrong. But I think
we would find out that I am wrong a lot quicker than we could find out
whether your proposal is either right or wrong.

Perhaps the lathe operator adjusts his
efforts so as to maintain a certain cost/benefit ratio. He can now make
twice as many parts at the same cost in effort, but this results in a higher
ratio of benefit to cost. He can restore the same cost/benefit ratio by
increasing his effort, thus raising output to greater than 2(x + 10) parts
per hour.

Actually, this would involve a reduction in the benefit-to-cost ratio for
the worker. If you assume that the lathe operator was working at break-even
when producing x+10 parts per hour, then when the more efficient lathe is
brought in he can now produce the same number of parts per hour, and earn
the same pay, at half the cost to himself in effort. If he were to produce
twice as much, his costs and his pay would be the same as before, and there
would be no improvement in benefit-to-cost ratio, or difference.

In reality, employees would be more likely to work no harder than required
for a given pay, thus maximizing the benefit to cost ratio.

It would bring the benefit-to-cost ratio to whatever it is when the
employee is working just hard enough to get the amount of pay that is
desired (if that is a permissible variable in his job -- i.e., piecework).
There's no reason to think the ratio would be maximized under those
conditions. I don't like maximizing models anyway; if the variable to be
maximized is less than maximum, should you increase or decrease your
effort? The error signal contains no sign information, so a rather
elaborate mechanism would be required. A control system is always simpler
than a maximizing system, if error information with sign is available.

And anyway, how can you ever find out whether a cost-benefit maximizing
model is correct? As I said, there's no way to observe the cost-benefit ratio.

The incentive
for raised productivity builds in a bonus in pay to offset the required
increase in effort -- an offer which the operator may or may not find
attractive. If pay does not depend on rate of output, the operator's best
deal in terms of cost/benefit ratio is to make no parts at all, regardless
of what lathe he uses. In that case the employer is left with the option of
convincing the operator to work harder (appealing to his sense of fairness,
for example), changing the costs associated with not working hard enough
(docking pay, threatening to fire the operator), or changing the way the
operator is paid (e.g., piece rate instead of salary). Incentive theory
takes account of the individual's subjective costs and benefits to predict
what the individual will do. I believe it assumes that these inputs yield
an equilibrium value, which depends on the values of those inputs as the
individual perceives them.

OK, that's one proposal. Can you model that in a way that can be tested?

I think my reply to you is really the same as my reply to Fred Nickols. You
seem to want to propose a different model instead of trying to use the PCT
model. You're trying to defend this cost-benefit concept and the idea of
incentives, and if that's what you want to do that's what you will do. I
think that a more plausible, not to mention more testable, model can be
constructed from the principles of PCT -- a model that would say nothing
about maximizing anything. But if you think your proposal is more plausible
and testable, then that's the way it is.

Best,

Bill P.