A group of banking executives were, apparently, not convinced that Ariely's "incentive plans cannot work" conclusion would hold in the real world. Ariely offered to perform experiments on the employees of these banking firms to settle this "real world" question. But it seems this kind offer was declined.
Fortunately, however, a large number of executives have allowed economists to examine this question. Researchers in the expanding field of Personnel Economics have performed dozens of real experiments in real firms, on real workers, changing real pay plans, with real money at stake. These are actual controlled experiments where researchers randomly assign workers to "treatment" and "control" groups, just like you would do to test a new drug against a placebo. Accounts of these experiments are published in the leading peer-reviewed economics journals.
Here are some samples: Stanford economist Ed Lazear --- currently on leave from teaching while serving as Chairman of the President's Council of Economic Advisors --- worked with auto-glass installer Safelite Glass to structure a pay-for-performance plan for windshield installers. Randomly selected groups of workers were shifted from hourly pay plans to piece-rate plans in which installers were paid based on how many windshields they installed. The result? Productivity rose. About 44 percent.
Oriana Bandiera, Iwan Barankay and Imran Rasul tested a similar hypothesis on managers at an English fruit farm. At the farm, fruit-picking workers were paid piece rates, but managers received hourly wages that were independent of overall output. When managers' pay was shifted to being performance-based, overall fruit picking productivity rose. How could this be, given that managers themselves picked no fruit? Bandiera and colleagues show when managers were paid a simple hourly wage, they spent their time helping the employees they liked. When managerial pay depended on overall output, managers showed less favoritism, and instead seemed to allocate their "help" activities more productively. As with the Lazear study, the economic magnitudes of the impact of incentives are substantial.
There are many such studies. The broad conclusion is that, in real firms, incentives change behavior. Ariely's argument --- that tying pay to measured performance won't lead to improvements in measured performance --- is therefore not supported by the available real world evidence.
So, incentives work, right? Well, perhaps. If we take the question of "Do incentives work?" to mean "Do incentives improve performance on measured dimensions?", then I think the consensus of personnel economists is that it does. But if we instead interpret this question as "Does the use of incentives lead to higher profits or better organizational performance?", the answer is less clear.
Here's why: Incentives motivate employees to shift effort toward improving performance on _measured_ dimensions. Performance on dimensions of the job that are harder to measure is likely to slip as a result. If, for example, a banking executive is paid based on his firm's earnings for this year, then he may work to raise this year's earnings even if doing so worsens the firm's long-term prospects. If the performance measure fails to capture everything that a firm wants an employee to do, then it's not necessarily the case that using strong incentives will increase profits.
Measuring performance for bankers is hard specifically because these individuals make many short-term decisions that have long-term economic consequences. And it is problems with devising appropriate performance measures that contributed to our recent credit crisis, not Ariely's claim that using incentives does not improve measured performance.
The broader danger of Ariely's point, however, is that it gives academic credibility to those who would want to limit the ability of firms to use pay-for-performance incentives. Providing incentives for bankers is hard, and it may well be appropriate for government to play a role in determining how such incentives are structures. But the case for such regulation rests on problems with performance measurement and externalities from the financial sector on the rest of the economy, not on the unsupported-by-real-world-evidence claim that incentives just don't work.