The first behavioral concept I will outline is inter-temporal choice. Traditional economics predicts consumers should make inter-temporal trade-offs such that their marginal rate of time preference equals the interest rate. Further that consumers should be consistent in their inter-temporal choices: The discount rate used should be constant across situations and over time. However, research shows that depending on the context examined, the implied discount rates of observed behavior can vary from negative to several hundred percent per year (Lowenstein 1989). For instance an experiment by Thaler in 1981 in which “subjects were asked to imagine that they had won some money in a lottery conducted by their bank displays irregularities from time consistent discounting. Subjects could take the money now or wait until later. They were asked how much they would need to be paid to make waiting as attractive as immediate payment. The experiment manipulated three variables of interest: the length of time to be waited; the magnitude of the outcome; whether the outcome is a gain or loss. Three strong patterns emerged from the subjects' responses: discount rates declined sharply with the length of time to be waited, discount rates declined with the size of the reward, discount rates for gains were much higher than for losses” (Lowenstein 1989). “The magnitude effect, evident in both the Thaler and Benzion et al. studies using hypothetical questions illustrates implicit discount rates decline sharply with the size of the purchase. Discount rates for small amounts (under $100) were very high, while those for larger amounts were more reasonable. Named the sign effect, the third strong empirical regularity in the discounting surveys, is that the discount rate for gains is much greater than for losses. Subjects also display loss aversion with inter-temporal choice. The mean delay premium is at least twice the mean speed-up cost, subjects demand more to wait past the expected arrival date than they are willing to pay to speed up its expected arrival. The result is compatible with Kahneman and Tversky's notion of loss aversion, the idea that the disutility of losing a given amount of money is significantly greater in absolute value than the utility of gaining the same amount” (Lowenstein 1989). The ODonoghue 2000 paper describes the finding that when agents have a preference for immediate gratification this leads to under indulgence on tasks that involve immediate costs and delayed rewards. (Over indulge when immediate rewards and delayed costs). Inter-temporal preferences that are time consistent will maintain a relative preference for well being at earlier date over later date is same no matter when asked. In contrast preference for immediate gratification implies time inconsistent preferences: long run perspective for one set of preferences (prefer not indulge at period t) but when come period t, has different set of preferences (now wants to indulge). Long run- time consistent-impatience can be expressed by parameter delta whereas Beta models a preference for immediate gratification.
The next concept I will outline is loss aversion. To start Kahneman and Tversky proposed that the value function is defined on deviations from the reference point; generally concave for gains and commonly convex for losses and steeper for losses than for gains. (Kahneman 1979). Reflection is that an agent with the same utility function in the gain and in loss domain, simply by changing sign, will become risk seeking in loss domain and risk averse in gain domain. An example of reflection is the deposition effect. Traders hold onto stocks that have lost value for too long and sell stocks that have gained value too quick. (Odean – Genesove and Mayer). In contrast, loss aversion states that changes that make things worse (losses) loom larger than improvements or gains. Endowment effect and status quo bias can both be explained by loss aversion. The endowment effect is that people often demand much more to give up an object than they would be willing to pay to acquire it. This occurs because subjects perceive giving up the endowed object to be a loss. A status quo bias is a preference for the current state. Under loss aversion individuals have a strong tendency to remain at the status quo, because the disadvantages of leaving it loom larger than advantages (Kahneman 1991). An example of the Endowment effect is from an experiment presented in the Kahneman 1991 paper, in which 1/2 students in a classroom where given a mug/pen. Then were asked to trade or keep the item; since mugs/pens were assigned at random, half should want mugs, half should not. However, the 50 percent predicted volume of trade did not occur. “There were 22 mugs and pens distributed so the predicted number of trades was 11. In the four mug markets the number of trades was 4, 1, 2, and 2 respectively. In the pen markets the number of trades was either 4 or 5. In neither market was there any evidence of a trend over the four trials. The reason for the low volume of trade is revealed by the reservation prices of buyers and sellers. For mugs, the median owner was unwilling to sell for less than $5.25, while the median buyer was unwilling to pay more than $2.25-$2.75. The market price varied between $4.25 and $4.75. In the market for pens the ratio of selling to buying prices was also about 2. The experiment was replicated several times, always with similar results: median selling prices are about twice median buying prices and volume is less than half of that expected.” (Kahneman 1991). An example of the status quo bias, or default effect is organ donation rates correlation to the default choice. In countries where organ donation occurs under consent by default, participation rates are 25%–30% higher than in countries where donation occurs where no consent is assumed. (Abadie and Gay 2006). Another important aspect to loss aversion is given a stream of gains/losses how the agent evaluates them (how choices are bracketed). Segregating gains gives greater utility than a one time gain, whereas aggregating losses provides more utility than segregated losses. Further combining a loss and a gain to yield a smaller gain always leads to higher utility than a small loss and larger gain. (Variation occurs with small and gain and larger loss). A demonstration of bracketing is in a paper by Gneezy and Potters (1997) and Thaler (1997) ran experiments in which subjects make choices between gambles. The manipulation being the frequency which subjects receive feedback. In the Thaler study subjects made investment decisions between stocks and bonds at frequencies over 8 times per year, 1 time per year or 1 time every 5 years. Subjects in the long-term conditions invested roughly 2/3 of funds in stocks which in the frequent condition 59% in bonds. Stocks are more risky than bonds, thus demonstrating through aggregating potential losses (5 year period vs 8 times per year) subjects have a greater tolerance for risk. (A potential one time loss looms less than segregated losses summing to the same amount).
There have been various policy attempts to change real world behavior that draw on behavioral economic findings such as loss aversion and time inconsistent discounting. One aspect of loss aversion is the default effect: individuals have a strong tendency to remain at the status quo, because the disadvantages of leaving it seem larger than advantages. An example in the real world was a project aimed to save paper that tested the effect of printer paper used after changing the default setting from single sided to double sided. “In a project focused directly on the costs of government opera- ons, SBST collaborated with USDA’s Economic Research Service (ERS) to promote the use of double-sided printing among ERS employees. During a pilot period, if employees initiated a single-sided print job, a dialog box appeared that prompted individuals to change their default printer setting to double-sided, but still allowed them to continue their single-sided print job if that was preferred. Individuals who received this prompt were more likely to print double-sided. This prompt increased the likelihood of double- sided printing on a given job by 5.8 percentage points, from a baseline of 46.0 percent. Based on this result, ERS is changing the default settings of all of its printers to double-sided.” (Social and Behavioral Sciences Team). However using defaults to implement behavior change are not always as effective. The printing paper example may have worked because subjects didn’t have a strong preference for single or double sided printed pages. Thus in that case the default setting helped to establish a preference for the population. However if the population one is attempting to influence already has a clear preference, a reference point from which losses are measured will already be set, and modifying the default rule will cause a violation of these preferences. For instance, a study was done on Organization for Economic Co-operation and Development (“OECD”) employees, analyzing energy use after modifying default thermostat settings. During the winter, a one-degree Celsius decrease in the default led to energy savings as employees did not bother to change the setting. However, when the default setting was reduced by two degrees Celsius, employees thought that it was too cold and raised the temperature higher than the control setting (Testing the effect of defaults on the thermostat settings of OECD employees). In summary, when there are strong preferences, inertia will be overcome and the default won’t stick.Another example of loss aversion and framing in real life was a study done on teacher incentives. “The authors gave teachers money in advance and told them that if students did not show real improvements, the teachers would have to give the money back. The result was a significant increase in math scores A second treatment identical to the loss aversion treatment but implemented in the standard fashion, yields smaller and statistically insignificant results. This suggests it is loss aversion, rather than other features of the design or population sampled, that leads to the stark differences between our findings and past research.” (Roland G. Fryer, Jr, Steven D. Levitt, John List, Sally Sadoff) Giving teachers money in advance changes their reference point to a higher initial wealth state and having to give the money back is a loss from this original reference point. Giving the teacher’s the money if they boost students grades at the end of the program would in contrast be a gain.
Next, I’ll outline examples of attempts to change behavior using insights derived from time inconsistent discounting. One study provided financial incentives for subjects to lose weight. A major problem with dieting/weight loss is that people procrastinate, they promise themselves they will start a program the next day but when the next day comes they can’t do it. A meta‐analysis study of research on weight loss and financial incentives showed no significant effect of use of financial incentives on weight loss or maintenance at 12 months and 18 months. (Systematic review of the use of financial incentives in treatments for obesity and overweight by Paul-Ebhohimhen V1, Avenell A.) These programs may have been unsuccessful because weight loss becomes increasingly difficult throughout the course of a program. Subjects may lose the most weight in the first few months but then plateau of the next few and financial incentives are not a significant enough motivator to push them through this tougher period. However financial incentives for smoking cessation appear to be effective. Giles examined seventeen papers reporting on 16 studies on smoking cessation. He found “in meta-analyses, the average effect of incentive interventions was greater than control for short-term (≤ six months) smoking cessation (relative risk (95% confidence intervals): 2.48 (1.77 to 3.46); long-term (>six months) smoking cessation (1.50 (1.05 to 2.14)); attendance for vaccination or screening (1.92 (1.46 to 2.53)); and for all behaviors combined (1.62 (1.38 to 1.91)). The available evidence suggests that financial incentive interventions are more effective than usual care or no intervention for encouraging healthy behavior change.” (The effectiveness of financial incentives for health behavior change: systematic review and meta-analysis by Giles et al 2014) Providing financial incentives to quit smoking may have been effective because as a nicotine addict quits smoking for a longer and longer period of time the urge to smoke decreases. Thus financial incentives can effectively motivate the individual in the first few tough weeks to months and after the urge decreases. In summary, financial incentives have been effectively used to motivate behavior change in instances where subjects normally have time inconsistent preferences which prevents them from doing so alone. However financial incentives for behavior that becomes increasingly difficult to maintain over passage of time is likely to be less effective than for behavior that becomes easier to maintain over time.