Persistence by Hogarth

regimeshiftfigure1The last post Persistence looked at persistence as being partially determined by the distribution of the waiting times for the reward. A fat tailed distribution might rationally steer one toward giving up after a short waiting period. Robin Hogarth (Educating Intuition) has recently published a paper: “Ambiguous Incentives and the Persistence of Effort: Experimental Evidence” in the Journal of Economic Behavior & Organization, Volume 100, April 2014, page 1-19, with Marie Claire Villeval that looks at economic activities where the reward is mundane–money.  It is more aimed at looking at what determines our persistence from the employers point of view, but I believe it could be more broadly applicable.

Hogarth and Villeval explore ambiguous situations where economic agents reap the benefits of engaging in an activity across time until – unknown to them – there is a shift (the regime change) in the underlying process and pursuing the activity is no longer profitable. The term regime shift was new to me in the context.  For an old city planner, regime shift meant a new mayor or change in the form of government. Apparently the ecological term more or less runs with the old definition and means abrupt long lasting non-linear change.  Hogarth has helped me understand that humans have made an evolutionary career out of understanding linear change or functions that are linear over the relevant range, while we tend to be weak at non-linear functions. How long will the investor continue to place new orders and does this depend on the regularity of his previous outcomes? How long will an employee keep working in the same firm if she no longer receives a bonus? How is the decision affected by preferences regarding risk and ambiguity and/or the regularity with which bonuses have been paid in the past?

Hogarth and Villeval set up experiments to investigate the persistence in effort according to the individuals’ ability level and degree of risk and ambiguity aversion, and the differential effects of continuous and intermittent (random and fixed) incentive schemes before and after the introduction of a regime shift. I will give only a bare bones description of the experiments. The task itself involved counting specific letters in written paragraphs where words had been combined randomly to form meaningless sentences. They compared three treatments in a between-subjects design. In the Baseline treatment, participants were paid a piece-rate and each period gave rise to actual payment. This corresponds to a continuous incentive scheme. This is comparable to small regular monthly bonuses. In the Random Intermittent Incentive treatment, for each participant there was a one out of three chance that any particular period would be paid and the researchers determined individually randomized sequences of payment for all these participants. This is comparable to a bigger bonus on an unknown schedule. In the Fixed Intermittent Incentive treatment, one period out of three was paid according to a fixed schedule. This is comparable to a bigger bonus every 6 months. In all three treatments, payment was stopped after the 20th period.


Figure 1 simulates belief curves  over the 35 periods. (Belief is believing that you will get paid.) The model predicts different patterns of beliefs across periods for the three treatments. In the Baseline, belief increases in the early periods but drops rapidly after payment stops and stays below the levels of the other treatments through period 35. In the Fixed intermittent treatment, beliefs oscillate with the pattern of payment until after period 22 when there is a decline until the end of the game. As to the Random intermittent treatment, averaging random draws, there is a constant pattern through period 20 after which there is a decreasing trend toward the end of the game. However, beliefs are higher than those in the other treatments after the regime shift.

Hogarth and Villeval found that under ambiguity, more risk averse and less able participants are more likely to quit the task as soon as possible, i.e. before the regime shift, regardless of the treatment probably because they dislike having to bear all the risk. Participants with a higher marginal cost of effort have lower payoffs from undertaking the task. Higher risk aversion requires a higher risk premium to participate longer. The researchers cite Lazear in  “Speeding, Terrorism, and Teaching to the Test” who suggests that predictable tests provide the best incentives for less talented children to learn; less predictable tests are better for high ability students but result in no learning for less talented ones because their cost is too high.

In the Baseline treatment where it takes less time to identify the regime shift individuals exit soon after payment stops. Individuals who are used to receiving rewards in every period react quickly to payment stopping by quitting the task. On the other hand, participants in the Random Intermittent Incentive treatment exit either before or long after the regime shift, with the Fixed Intermittent Incentive treatment in an intermediate position. One reason for this is likely that it takes more time for individuals to determine that a regime shift has occurred. Therefore, under ambiguity, both the frequency and randomness of incentives matter in the decision to not persist, with randomness having more extreme effects in both directions. The gap between the individuals’ income expectations and their updated beliefs about the likelihood of payment is smaller than in the Baseline. Moreover, intermittent incentives lead to more persistent and higher effort levels.  This selection effect in terms of ability and risk attitudes combined with the impact of intermittent rewards on persistence lead to an increase in mean performance after the regime shift when incentives are intermittent.

So writing a novel that is a financial success probably has a fat tailed distribution of waiting times. Ceteris paribus it is probably reasonable to persist for awhile and then abandon the effort. Now Hogarth looks at other factors. Now if you have a family that you have to support, you are likely more risk averse than someone with a tidy trust fund. This was also mentioned by McGuire and Kable with the personal discount rate. If you discount the future steeply, you are likely to be risk averse and this will reduce your persistence on that novel. Ability is another factor that Hogarth unearths. If your writing education is top notch and you are an excellent writer, you will also tend to be more persistent. Finally, there are the incentives and their structure.  In the initial post on persistence, intermediate incentives/rewards were not really considered. For an activity like calling a cab that is appropriate, but with going to work or even longer term personal goals there are most often intermediate rewards that act as incentives.  Having a short story or novella receive an award every once in a while will help you persist especially if you have high ability and your personal discount rate is low.  Regime shift has probably already happened as far as writing a successful novel, but there can always be another regime shift. When it happens, those who are less risk averse, have more ability, and have been subject to random intermittent rewards, will persist more successfully.




Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.