The research question they tested was if the principal can set a minimum level of x, should they?  An economist would say, of course.  The agent has an incentive to not work at all.  Setting a minimum level of work would guarantee a return for the principal.  To test this, they set up the following experiment.  

• An agent (worker) must decide on a level of x, which is seen as a proxy for work.  The cost function is simply c(x)=x.  They have an initial salary of 120.  Thus, their payoff is 120-x.
• The principal (boss) wants the agent to choose a large x (aka to work as much as possible.  Their payoff is 2x.
  • If the agent decides x=20, then the worker gets 100 and the principal gets 40.

However, it turns out that setting a minimum level of x is counterproductive.  By setting the minimum level of x, this shows that the principal does not trust the agent.  This leads the agent to choose a lower level of x chosen by the worker.  For instance, if the principal chooses a minimum x of 5, the agents chose x=12.2 on average.  However, if the principal does not choose a minimum x, then the agent chose an average x of 25.1.Further, principals knew this would happen.  About two-thirds of principals decided not to dictate a minimum level of x.  

The best interpretation of these results can be shown by an excerpt from the memoirs of David Packard, the founder of Hewlett-Packard (HP):

“In the late 1930s, when I was working for General Electric…, the company was making a big thing of plant security. … GE was especially zealous about guarding its tool and parts bins to make sure employees didn’t steal anything. Faced with this obvious display of distrust, many employees set out to prove it justified, walking off with tools and parts whenever they could. … When HP got under way, the GE memories were still strong and I determined that our parts bins and storerooms should always be open. … Keeping storerooms and parts bins open was advantageous to HP in two important ways.  From a practical standpoint, the easy access to parts and tools helped product designers and others who wanted to work out new ideas at home or on weekends. A second reason, less tangible but important, is that the open bins and store- rooms were a symbol of trust, a trust that is central to the way HP does business” (David Packard, 1995, p. 135). 

• Falk, Armin and  Kosfeld, Michael (2006) “The Hidden Costs of Control” American Economic Review, v96(5):1611:-1630.

An article by Ryan, Watson, and Entwistle (Health Econ 2009) probes whether or not people really are irrational using discrete choice experiments (i.e., pick option A or option B).  What they found is that some of the “irrational” behavior is often due to additional assumptions by the participants.  To understand which idiosyncratic assumptions individual carry with them, the authors ask people to think aloud while they are answering the questions.  This way, the authors are able to evaluate if the participants are having the same assumptions the designer of the economic experiment.  

For instance, the paper found that some individual choose high cost over a low cost options.  The reason for the was that some people believe that high cost can be a proxy for quality.  

“You see I feel if you pay more you are obviously going to get more, they are going to do more for you, because that’s just a way of life isn’t it. I don’t mean to say that you have always got to pay a big price but sometimes, the saying is ‘pay cheap, pay dear’.”

Other people could not afford to pay for either option and thus they ignored the cost of each option.

“I thought with the 10 questions there, you were looking at it and thinking ‘what’s more important’? Before the cost was important to me, at the start. I was saying to myself ‘well I can’t afford £48, I know I can’t afford £48, I’m unemployed’. But the more I am answering the questions, the cost was coming out of the equation…”

This “think aloud” technique is a great way for experimental economists to verify that their experiments are actually testing what they claim to test.

• Ryan, Watson, and Entwistle (2009) ”Rationalising the irrational: a think aloud study of discrete choice experiment responses” Health Economics, v18(3):321-336.

The Wilson Quarterly reviews research by Benedetto De Martino and colleagues using an experimental framework to see if individual with autism spectrum disorder (i.e., ASD or autism) were less influenced by how gambles were framed.

“More than the contrl group, [individuals with autism] seemed to recognize the ‘lose’ and ‘keep’ options as the same.  Defiiciencies in emotional processing apparently have an upside, safeguarding people with ASD from at least some forms of emotion-driven irrationality.”

• “The ASD Difference” The Wilson Quarterly, Winter 2009, p. 16.
• De Martino B, Harrison NA, Knafo S, Bird G, Dolan RJ. 2008. “Explaining Enhanced Logical Consistency during Decision Making in Autism,” The Journal of Neuroscience, October 15, 2008, 28(42):10746-10750

• Hat-tip: Marginal Revolution

In “Risk Aversion in the Laboratory,” Harrison and Rutström review some of the techniques used to elicit risk aversion preferences. We will review 5 of these techniques: multiple price list (MPL), random lottery pairs (RLP), ordered lottery selection (OLS), Becker-DeGrot-Marschak (BDM) and trade-off (TO).

• Multiple price list (MPL). In this type of lottery, subjects are given a list of binary lottery choices to make all at once. The most famous example of MPL is the Holt and Laury (2002) study. MPL is probably the most widely used method to elicit risk preferences, but do suffer from the problem of framing. In the Holt & Laury study, subjects may have tended to choose a switching point in the middle of lottery list even if their actions in the real world would not have reflected this choice.
• Random lottery pairs (RLP). Under RLP, subjects face binary lottery choices in a sequence and must choose the preferred lottery. Hey and Orme (1994) used this methodology to test expected utility predictions. The experimenters’ elicited the subjects preferences over 100 pairs of lotteries, where the outcome values were fixed (£0, £10, £20, £30) but the probabilities for each outcome changed among the 100 lottery pairs.
• Ordered lottery selection (OLS). In this methodology, the subject chooses one lottery from an ordered set. For instance, Barr (2003) allowed villagers in Zimbabwe to choose from the following 50/50 lotteries: (100; 100); (90, 190); (80, 240); (60, 300); (20, 380); (0, 400). The OLS structure can help to answer questions about risk preferences, but since all lotteries are 50-50, they can not answer questions regarding higher order risk preferences (e.g., prudence, temperance). Further, this method does not allow for the analysis of any Kahneman and Tversky-style probability weighting.
• Becker-DeGrot-Marschak (BDM). In the words of Blavatsky & Köhler (2007), “Under the BDM procedure, individuals are asked to state their minimum selling price for a risky lottery. The experimenter then draws a random number between the lowest and the highest outcome of the lottery. If the price that the individual states is lower than or equal to the drawn number, she receives the drawn number as her payoff. Otherwise she has to play the risky lottery.” The benefit of BDM is that if preferences satisfy the independence axiom, then the bid will be the individuals exact certainty equivalent. However, it assumes that individuals do not make errors and understand the fairly complex nature of the game.
• Trade-off (TO). The trade off design gives subjects choices over lotteries and these lotteries are endogenously defined in real-time by prior responses of the same subject. This can lead to a more precise measure of the certainty equivalent, but does the lotteries played will vary by subject.

With any of these experiments it is important to pay real money for the subjects answers. Otherwise, many of the results will suffer from hypothetical bias (see Camerer and Hogarth, 1999).

• Cox and Harrison (2008), “Risk Aversion in the Laboratory,” Risk Aversion in Experiments, Research in Experimental Economics, volume 12.

I have compiled a document (“Lottery Presentation Styles“) to give you some suggestions.

However, another feature of individuals preferences can also influence how individuals evaluate risky situation. This concept is prudence. Let us go back to the coin flip game ($100 heads; $0 tails). Imagine you make $50,000 this year and you are going to get a raise next year so you will earn $60,000. Would you rather play the coin flip game this year or next year? A prudent person would want to take the risk when they are in the better financial situation.

Measuring prudence, however, is not a simple task. A working paper by Deck and Schlesinger tries to estimate prudence preferences in an experimental setting. The experimental question is posed as follows:

• You will receive $10.50 + (1|-1) if the con lands on Heads or Tails and $9.00 if the coin lands on the Same or Different outcome.

I think the phrasing of the question is unnecessarily complicated, but the question is fairly straight-forward. Everyone gets $10.50. Individuals must choose between Heads or Tails; and Same or Different. To simplify, let us assume that everyone chooses Heads, which means you earn $1 if the coin lands on heads and lose $1 if the coin lands.

Now we must decide between Same or Different. If you choose Same, that means that you get $9 if the first coin toss lands on heads and you also flip the coin a second time to see if you win or lose $1; if the first coin toss lands on tails then you get $0, but do not have to play the win/lose $1 game. If you choose Different, then if the coin lands on heads you get $9, and do not play the second coin toss; if the coin lands on tails you get $0 and do not play the second coin toss.

Individuals who choose Same are prudent because they take the financial risk (win/lose $1) when they are richer ($9 extra). Those who choose Different, are imprudent because they take the financial risk (win/lose $1) when they are poorer ($0 extra)

The authors asked 6 of these prudence of questions. They found that 61% of subjects responded to the questions in a prudent manner, but only 14% of individuals responded to all six questions prudently. A Kolmogorov-Smirnov statistic of 0.2225 indicates that people are making prudent choice more than would be the case if they were choosing randomly.

• Cary Deck and Harris Schlesinger (2008) “Prudence and Temperance: Exploring Higher Order Risk Effects in the Laboratory” Working Paper.

APPENDIX

The other 6 prudence questions were:

1. You will receive $30 + (25|-25) if the con lands on Heads or Tails and $25.00 if the coin lands on the Same or Different outcome.
2. You will receive $12.50 + $9.00 if the coin lands on Heads or Tails and (5|-5) if the coin lands on the Same or Different outcome.
3. You will receive $12.50 + (5|-5) if the con lands on Heads or Tails and $1.00 if the coin lands on the Same or Different outcome.
4. You will receive $10.50 + $9.00 if the coin lands on Heads or Tails and (1|-1) if the coin lands on the Same or Different outcome.
5. You will receive $12.50 + $5.00 if the coin lands on Heads or Tails and (5|-5) if the coin lands on the Same or Different outcome.
6. You will receive $14.50 + (9|-9) if the con lands on Heads or Tails and $1.00 if the coin lands on the Same or Different outcome.

However, an NBER working paper by Beshears Choi, Laibson and Madrian (2008) claims that sometimes, people’s revealed preferences may not coincide with their true preferences:

1. Passive choice. If individuals do not actively make a choice, such as would be the case for a default 401(k) enrollment at work, people may not be choosing optimally (at least initially).
2. Complexity. It is often difficult for individuals to act rationally in complex situations. For instance, when one is faced with a large number of choices or faced significant uncertainty regarding future prospects, individuals may act suboptimally.
3. Limited Personal Experience. “Human learning is often generated by feedback.” Thus, the more experience an individual has, the more rational they will act. You probably get a better deal buying groceries than buying a used car, because you have much more experience with the former.
4. Third Party Marketing. “Tom Sawyer tricked his friends into paying him for the privilege of painting his family’s fence. A great deal of real behavior is also influenced by marketing.”
5. Intertemporal Choice. How should people discount future utility? The authors claim that “Only discounting due to mortality risk is easily defended philosophically.”

How can we correct for these “incorrect” revealed preferences? Beshears and co-authors give some suggestions:

Structural estimation specifies a positive model with a precise set of economic and psychological motives (perhaps including non-Bayesian thinking and other decision-making errors). This model is then estimated using data, and the resulting positive preferences are mapped into normative preferences using normative axioms.

Active decisions …[require individuals] to explicitly state their preference without beinginfluenced by (or being able to rely on) a default option.

In most stationary economic environments, initial choices are likely to be further from normative optimality than choices made after many periods of experience. One should therefore give more weight to asymptotic choices [preferences revealed after having experience with the choices] when attempting to infer normative preferences.

When homogeneous individuals make noisy, error-prone decisions, their individualdecisions do not reflect normative preferences, but their aggregate behavior can …

Self-reported preferences reveal something about an agent’s goals and values. Normative economics should allow self-reports to have some standing.

• Beshears J, Choi JJ, Laibson D, Madrian BC (2008) “How are preferences revealed” NBER WP #13976.

A rational person would believe that buying insurance against an event will not alter the probability that it will occur–ignoring issues of moral hazard.  For instance, the act of buying health insurance should not make us less likely to be sick.  Using more preventive care which is cheaper due to insurance can prevent illness, but the act of buying health insurance should not effect the probability one gets sick holding constant the medical care levels.

A better example may be travel insurance.  “Last year, tens of millions of people bought life insurance for scheduled flights of airlines in the United States. Not one of those insured passengers died in a crash.”  Is this a waste of money?  Not if you are superstitious and believe that the act of buying life insurance affects the probability your plane will crash.

“So when we think about passing up flight insurance, we conjure up disaster just as easily as ancient Greeks imagined a thunderbolt from Olympus, and we too figure we can avert it through the equivalent of a bull sacrifice. Intuitively, we haven’t made great strides since Homer’s day. But at least our gods take credit cards.”

• Hat tip to Arnold Kling at EconLog.