Incentives as ideology

“Incentives matter” says Econ 101. Reality, however, says otherwise. Here’s a new study:

Agents do not provide high effort when prize spreads are larger. When we take into account agents’ risk aversion, we find that while the effect of large spreads is not significant for risk-prone individuals, it is strongly detrimental for the performance of risk-averse individuals.

This is based on the performance of Italian students in exams. It is yet more in a long line of evidence that high-powered incentives can and do backfire, for example:

 - They can encourage outright dishonesty such as fiddling performance figures or Libor rates.

 - They encourage workers to do jobs which are easily measured and incentivized rather than ones which are less observable but which have longer-term payoffs, such as preserving the company’s reputation or solvency. Incentives, for example, gave us PPI mis-selling and banks’ issuing “liars’ loans.” As Benabou and Tirole show (pdf), bonus cultures can cause “significant efficiency losses, particularly in the long run.”

 - The urge to meet targets can crowd out innovation and creativity. As John Seddon has said:

“Best practice” in the service of big ideas promulgated from the centre and inspected for compliance is, in fact, worst practice, a nail in the coffin of innovation. For innovation to flourish the locus of control must shift to the frontline where people deliver public services. Innovation requires freedom to learn and experiment.

 - Big incentives can signal that a job is very difficult. This might encourage some to simply give up (pdf) and others to take big risks to meet their targets. Or they might have an adverse selection effect: only the irrationally overconfident would want to take on such jobs.

 - The Yerkes-Dodson law tells us that that big incentives can and do lead to choking – as we see when golfers miss easy putts and footballers important penalties. As Dan Ariely and colleagues say, “high reward levels can have detrimental effects on performance.”

All this poses the question: if high-powered incentives can so often fail or backfire, why do they still exist in some places? I suspect the main reason is that whilst the idea that big rewards are necessary to elicit good performance might be often wrong, it serves as a useful justification for inequality. What we have here is ideology masquerading as economics.

Comments

[Martinned]

Yes, incentives matter. All of your examples show exactly that. But they also show that there is such a thing as having incentives that are too high-powered, which is a familiar result from institutional economics and principal/agent theory going back decades.

The reason why we still have high-powered incentives - and remember, we don't have them nearly as much as we could in theory, that's what the Theory of the Firm is all about - is that in a lot of circumstances overshooting in the direction of having incentives that are too strong is less costly for the decision maker (i.e. the person putting those incentives in place in the first place) than undershooting the optimum. The decision maker/principal is either less likely to get caught overshooting, or he is less likely to get blamed for it if he does get caught.

[David]

50 years ago when I started work I discovered that bonus and performance related pay schemes had only short tem effects on output and in reality were a way to hide poor leadership and management.

During my working career i saw nothing to change this point of view. In fact events unfortunately proved my very minority opinion to be correct.

Along the way I discovered that eminent management gurus such as Drucker had the same point of view.

But still it is is the panacea for poor managememnt and leadership.

It is nice though to know one has always been right.

[Chris]

Have come across comical target chasing stories (such as a manufacturing site that 'shipped' everything in stock on the last day of the month, and put it back in as 'returns' on the first, to reduce 'inventory and increase 'sales', as 'returns weren't a target), but the one that personally wound me up was a company that I worked for.

Every month they would get in extra staff working overtime for the last weekend of the month to ship goods due in the first week of the next month. Effectively they were paying overtime every month for a one time increase in sales the first time they did it, and thereafter just to keep sales at the 'real' level.

[botogol]

all of your examples show that incentives DO matter!

the lesson is : because they DO matter, very much, it's very important to give people the right incentives.

[Neil]

Hi Chris, I've been reading your work for years, and in general I very much enjoy it. However, you DO have the habit of confusing "it's not as simple as that" with "that's wrong". This right here is an excellent example. Incentives matter. Of course they do. In general, rewarding certain behaviour results in more of that behaviour. This is obvious.

Sometimes, an incentive may create an unintended consequence which makes the incentive counter-productive. This is true, and is an important thing to bear in mind if you're trying to encourage certain behaviour via incentives.

But to go from that to the idea that belief in incentives is "ideology masquerading as economics" is a bit foolish.

"Incentives matter", says Econ 101, and it is completely true. But the incentive you INTEND to set and the one that you ACTUALLY set are not necessarily the same thing.

[An Alien Visitor]

"All this poses the question: if high-powered incentives can so often fail or backfire, why do they still exist in some places?"

One reason is management remuneration. A fundamental part of management and leadership is inspiring others to perform better. The academic evidence suggests all these assumptions are incorrect. One of the main ideological justifications for class exploitation is that workers need incentives from an elite group. Again the actual evidence proves otherwise.

Of course only a fool would believe anything is completely true but ideology creates fools. This is Econ 101.

[Endrew]

This is another great example of the difficulties of doing good social science.

The study tells us nothing more than that. For a start there is no estimates of the statistical power of the study design to answer its questions, so reporting the reams of "no significant relationships" in the results hardly tells you whether there actually isn't an important relationship.

There is hardly any statistical significance even in the data-mined relationships picked out of the mess that are shown (even the high-spread group). The medium spread group has "no significant" difference in performance relative to the low spread group. That DOES not mean that this incentive doesn't increase performance. Or indeed decrease it.

Worst of all, the incentive structure is very badly designed. In the ten question tournament part of the exam, regardless of your score (above a 20% threshold) you either get a win/loss payoff of 8/2, 7,3 or 6,4 points, depending on if you beat a single ability-matched individual. So the expected payoff per additional question you get right is only = 0.5 Because you are matched against a student of supposed similar performance, you have a roughly 50% chance of winning REGARDLESS OF HOW MUCH EFFORT YOU MAKE. You could perfectly well reason that if you decide to work harder for that part, so will your adversary. And vice versa. You don't know so 50% for all that uncertainty is a reasonable prior. That part of the exam is 25% of the whole.

The rest of the exam is 25 points and makes up the remaining 75% of the assessment. In this part, if you answer a question right, the expected payoff for that question is 1 full point - you do not have to beat some other guy overall, and in addition you are rewarded fully.

Now it gets worse. Each question in the tournament is worth less that each question in the conventional part of the exam. 25% / 10 = 2.5% per correct answer cf. 75% / 25 = 3 % per correct answer. So multiplying that by the payoff above, for each right answer, you get an expected percentage payoff per question of:

3 for the conventional test
1.25 for the tournament.

So which would you study harder for given the only REAL WORLD reward is a pass AND THAT ITALIAN STUDENTS CAN TAKE THE EXAM UP TO 5 TIMES!

And remember, these expected payoffs are INDEPENDENT OF INCENTIVE SPREAD LEVEL EVEN TAKING INTO A POSSIBLE CHOICE TO WORK HARDER.

So although the subject is interesting and I agree with a lot of what Chris says in general on incentives, this study does not enlighten such a discussion.

[sam]

if you wish to discard incentives from economics, you have to throw that Tirole paper you cite (which you give a too-simplistic account of IIRC) out the window with it!

what you show here is that there are some plausible countervailing tendencies which can cause extreme incentives to backfire. not quite as damning as 'ideology masquerading as economics'. not to mention, the list of examples of examples of incentives working is enormous and could fill a thousand blog posts.

also, you don't even have to be a capitalist to agree that some degree of inequality is justified for the good incentives it brings. consider a socialist utopia where the profit incentive is banned, means of production socialized, all that good stuff. the socialist government needs to build bridges, and they need those bridges not to fall down. studying civil engineering is very dull, and young people would rather study poetry or film or pure mathematics, or get drunk and sing songs of revolution. i think all socialists would be happier to live in a world where rich engineers build bridges that don't collapse above their heads than one of perfect equality where people have to row canoes across rivers, no?

[Kaleberg]

The problem is that incentives are hard to get right. I took a course from Evsey Domar on Soviet Economics and was surprised to discover that the commies were trying all sorts of incentive plans to improve productivity. The problem was that just about all of them could be gamed and they wound up being counterproductive. Meanwhile, I knew that all sorts of institutions in the capitalist US also had incentives and, back then at least, they were working fairly well.

I've pondered this for years. I think a lot of it was the low level of trust in the USSR, inherited from the low level of trust under the Tsars. There was a step on the next guy culture that dominated and a quick win was always worth it as the loser wasn't likely to get a second chance. In the US, again, back then, there was a higher level of trust. If sales went up, there was a good chance wages would go up. If you helped someone out, there was a good chance they'd help you out.

Unfortunately, most of us in the US have learned that working harder and producing more is unlikely to be rewarded. There is no sense of job security or investment in the joint venture. Corporations are just government chartered collectives, not all that different from those in the Soviet Union. There was just a different culture, though there is much less difference than there was.

[Keith]

This is also a lack of imagination. As Kaleberg said above incentives are the device used by authoritarian systems to try to push the plebs into doing what the boss wants on the assumption that the plebs cannot be trusted or secretly hate you so must be bullied into conformity. Carrot and stick. The class division of society is what communism is supposed to abolish but the attempts to do so have never worked. And Capitalism does not even try. If the men and women at the bottom cannot be trusted you get a lot of contortions trying to control them, from the KGB to million pound bonus pots. While the latter are better for the people who can work the system, as Banks found out recently the boss may find they do not work so well if they cause the Bank to collapse!

[reason]

I'm puzzled that nobody really put their finger on the real problem. The problem in business is that results are achieved by team effort, and individual incentives can destroy teamwork.

[Neil]

"The problem in business is that results are achieved by team effort, and individual incentives can destroy teamwork."

And you think no one has ever noticed that before?

That seems to contradict the available evidence: almost everyone assigns the biggest incentives to the people who manage teams - on the theory that if the manager gets a big bonus if the team performs well, he will take active steps to make sure that his team works together effectively.

Also, have you never worked at a company that did performance reviews? "Teamwork" is generally *one of the main criteria* by which they judge your performance (and therefore whether you get a bonus).

It doesn't always work, of course, but to assume that incentives destroy teamwork is to ignore how incentives are normally applied.

[Deviation From The Mean]

"studying civil engineering is very dull, and young people would rather study poetry or film or pure mathematics, or get drunk and sing songs of revolution"

Actually under capitalism writing poetry or music or film can earn you much more than civil engineering! Civil engineers tend not to be rich!

We live in a society where Abby Titmuss believed it was a career progression to stop being a nurse and becoming a slut (I have nothing against sluts btw).

So bring on the socialist utopia please. Though any utopia should have a place for sluts!

[Deviation From The Mean]

"Also, have you never worked at a company that did performance reviews? "Teamwork" is generally *one of the main criteria* by which they judge your performance"

This is not my experience. Performance reviews tend to be more task specific, albeit with some general marks against certain categories. Mt=y manager has rarely said to me, come and see me, your teamwork is not up to scratch. More often than not, it is, where is that fucking report I asked for. This is why managers and incentives destroy teamwork.

But we should also destroy a few myths:

Managers do not take on more responsibility, in fact when the News of the world went tits up Rebecca Brookes denied knowing anything!

Managers are not normal employees but with more ability, in my experience working in the business solutions (IT) sector, managers don't know where the enter key is on the keyboard! But they know a good strap-line when they see one!

[joe]

If you actually want to incentivise people to work harder then reduce the unit pay level and put people on piece work. They have to work harder to meet their basic payment needs. Once employee basic needs are met, then incentives bonuses waste more and more money for minimal increase in output. CEOs paid for failure are the worst example.
I worked on the Thames Barrier as a Civil Engineer -(the training was more exciting). Due to the urgency of the project we wanted 24 hour working and no strikes. No strikes incentive was a 50% uplift in basic pay and 24 hour working was further enhanced pay for unsocial 12 hour shifts. The workforce decided the uplift for 8 hours met all their normal spending needs and they went on strike refusing to work 12 hour shifts which eat into their leisure time.

Trying to incentivise people who do not do piecework type of jobs is a waste of time. Pay them to do what they are paid for. If they do not do that, then zero bonus or out the door.