Tag Archives: experiments

Behavioural economics 101 links

A new addition to the blog is the ‘Behavioural economics 101′ widget in the right hand sidebar. It is a short list of links to web pages that contain accessible, well-written or particularly captivating descriptions of the experiments that (behavioural) economists have designed over the last 30-odd years since the mid-20th century. If you are keen to learn more about behavioural economics, then I think it is much more engaging – and informative – to learn about the actual work than reading some boring, abstract description.

I encourage anyone interested in behavioural economics to check them out – and I welcome suggestions for newer or better links.

Buying a company: the winner’s curse experiment

Whenever somebody asks me about behavioural economics, a good way of preventing their eyes from glazing over is to start by telling them about running experiments with real people. And for a good reason, because it’s by far the most fun part of the work! So when we were asked to run some experiments for visitors of the University of Nottingham’s open day, my colleagues and I jumped at the chance. We ran two experiments: a common pool resource game – I will blog about that later – and a ‘buying a company’ game set up to bring the winner’s curse upon our unsuspecting participants.

The winner’s curse was first discovered during research on oil field auctions in the 1970s, which often lead to low rates of return for the oil company with the winning bid. The intuition behind this phenomenon starts with the assumption that the winning bidder is the one with the most optimistic prediction of the amount of oil in the oil field. If the average prediction of all oil companies in the auction is accurate (you may or may not believe this, it’s like Galton’s famous “guessing the weight of an ox” but then with lots of money at stake), then the most optimistic prediction must be an over-estimation of the actual amount of oil available. The winner of the auction thus makes a lot less money than predicted, and is thus ‘cursed’ by winning the auction.

The experiment we set up – which is based on Samuelson & Bazerman (1984) – has a nice and simple way of representing the uncertain value of the asset for sale. After a prospective buyer (a participant) submits an offer, we draw a ball from a bingo cage that might have any number between 0 and 100 on it (with equal probability). Next, the value of the asset (a fictitious company) is multiplied by 1.5 to give the value of the company to the prospective buyer. If the prospective buyer’s offer is higher than the number on the bingo ball, he will buy the company at his offered price. The profit or loss of a buyer is the difference between his offered price and the value of the company to him (to recap, this was 1.5 times the bingo ball number).

My friend and colleague Antonio entertains some prospective buyers of our fictitious company...

Sounds like good value? Here’s the catch: if the price at which you offer to buy the company is high enough to exceed the number of the bingo ball, the average value of the bingo ball number is halfway between zero and your offer. So if you offer a price of 60, the average value of the bingo ball for a winning bid is 30. Even though this number is multiplied by 1.5, you should still expect to make a loss! Most people make the mistake of disregarding the condition of winning (“if my offer is high enough”) and simply calculate the average bingo ball number in a random draw (this is 50) and base their bid on that. Our experiment on the open day showed exactly that – a lot of people submitted positive bids and most of them (2 out of 3, as a ‘fair’ bingo cage would predict) lost money. Amazingly, the only one to submit an offer consistent with Nash equilibrium (zero) was a 10 year old boy!

So you might say, “What does a simple experiment have to do with oil field auctions with big companies?” Well, think about what these experiments achieve: we are able to explain the outcome in the field (losing money) with a behavioural regularity (a failure to use conditional reasoning) that occurs in a particular environment (uncertain asset value, asset is worth more to buyer than seller). If economists can come up with a behavioural intervention that succeeds in ‘fixing’ behaviour in a laboratory experiment, then this intervention might be scaled up to work in the real world. (So far, laboratory experiments indicate that the winner’s curse in the ‘buying a company’ game is hard to get rid of completely.) Another way of thinking about this is that economists can use experiments to figure out which institutions (market rules) lead to the winner’s curse, and promote the use of alternative market structures (for example, more markets for information) such that people don’t bring themselves to financial ruin by making over-optimistic predictions.

Beyond Discovery has more information on winner’s curse research in economics.

Pay day loans experiment

Loans with a typical APR exceeding 500% – surely that’s impossible? Or even if they do exist, no one in their right mind would sign up for these terms, right? Well, ask the people walking into one of these:

Pay day loan store shop front

Because of the risk involved in making pay day loans, the cost to the borrower of obtaining such ‘alternative finance’ is higher than regular credit. The problem is that pay day borrowers usually don’t have access to regular credit. As the Economist eloquently puts it: “For someone who is truly hard-up, the only thing worse than borrowing $200 at 600% APR may be not borrowing $200 at all.”

So pay day loans are expensive. But are they harmful, and should they be regulated? The arguments for and against pay day lending are familiar and well-rehearsed economic justice arguments:

  • AGAINST: “Offering pay day loans is taking advantage of vulnerable people.”
  • FOR: “But no one is forcing the borrowers – it’s their own choice.”
  • AGAINST: “But poor people make poor choices when tempted by shrewd lenders – somebody do something!”
  • FOR: “Who? The government? Who says they know what’s best for people?”

Obviously, this discussion is going nowhere – it’s unavoidably political. But don’t despair, there is a third way: the behavioural intervention. A behavioural intervention is Nudge speak for giving people better structured information so that they can make better choices. This is what a pair of economists from the University of Chicago investigated in a well designed field experiment. The researchers manipulated the print on the cash envelopes used in 77 branches of a US pay day lending firm over a two-week period. They tried three print versions as experimental treatments:

  1. A comparison of pay day loan APR to other forms of credit
  2. A list of the accumulated cost of pay day loans in dollars, compared to credit card borrowing fees
  3. Information on how quickly people repay pay day loans

Compared to a control group who receive non-manipulated, company branded envelopes, the second envelope manipulation reduced the likelihood of the recipient re-applying for a pay day loan in the next pay cycle by 11%. Furthermore, individuals in all three envelope treatments reduce their average borrowing over the next pay cycle by more than the control group.

So, assuming that at least some pay day borrowers would rather not use pay day loans to make ends meet, the behavioural intervention is a success. Whilst I’m not sure that the conclusion “information disclosure that is inspired by, and responds to, cognitive biases or limitations that surround the payday borrowing decision has a significant effect” is warranted (the treatment is actually less effective for those who spend the loan money on eating out and holidays), the reduction in loans across the whole sample is encouraging. Perhaps this is one for the Behavioural Insights Team?

Experimental life

Laboratory experiments only get so interesting; if you want go a little further, you need to test your theories in the field. And when it comes to social science, what could be more engaging that using yourself as an experimental subject? Journalist and author AJ Jacobs has done exactly this for his book ‘The Guinea Pig Diaries’ (just released as a paperback titled ‘My Experimental Life‘). There’s a little taster on the Guardian site in the form of lengthy extract. The extract describes Jacobs’ attempts at uni-tasking (as opposed to multi-tasking) his way through life. He takes the principles of uni-tasking pretty far, which makes for a more entertaining read, but he makes some valuable observations on productivity along the way.

Jacobs reports that uni-tasking is a productivity booster; this rings true for me. Every time you stop yourself drifting towards a ‘quick’ browse of your favourite web sites, you win back at least 5-10 minutes of productive time. I’ve always thought that many of today’s ‘time savers’ (consider all the things you can do from your web browser that would have required an in-person visit in the past) force the mind to go into multi-tasking mode, thereby making for a double-edged sword in terms of net productivity. This is why I signed up for Tim Ferriss’ low information diet and haven’t looked back since.

 

The discontinuity effect = rationality polarisation?

Just found an interesting summary of the discontinuity effect. The name makes me wonder, do these people get paid on the basis of the number of ‘biases’ and ‘effects’ they discover? I quote:

(the researchers) named this the discontinuity effect because behavior in groups seemed discontinuous with the characteristics of individuals…

Wow. Presumably, if groups and individuals did behave similarly, they would’ve found a ‘continuity effect’? Or if the data were inconclusive, there would be an ‘continuity inconsistency effect’? OK I’ll stop now…

What’s interesting, though, is whether this particular individual-group discrepancy (groups are less cooperative in a repeated prisoner’s dilemma game)  could be defined as a kind of group polarisation. This kind of polarisation would then be called rationality polarisation. Now, imagine that one day, scientists will know all the behavioural dimensions in which polarisation occurs. Does that imply they would be able to predict group decisions based on the behavioural properties of and individual choices on a decision task?