Project Final Draft May 20, 2011
Economics 465 Gordon Atkins and Charlie Romero
Introduction:
Individuals below the poverty line have been shown to be more likely to engage in activities that reduce their future earnings. There exists a large body of prior literature which provides evidence of correlations between the poor and economically inefficient behavior, most notably summarized by Karelis (2007) to include the poor’s greater propensity to not work, not stay in school, not drink alcohol in moderation, not abide by laws, and not save money. While these observed correlations do not necessarily imply causation, data linking individuals below the poverty line to behavior that perpetuates their poverty is very clear (see i.e. U.S. bureau of the Census, 2000 data). Conventional explanations of this phenomenon are derived from one of two important presumptions: the poor are either “rational” and suffer from “restricted opportunity,” “flawed character” (Schiller 1976) or disincentives created by ineffective public policies, or the poor are simply acting contrary to “homo-economicus.” Karelis (2007) persuasively exposes each of the above “rational” explanations as flawed by citing surveys in which the poor themselves do not consider a lack of opportunities to work as the reason for their lack of work, the vast majority of poor and nonpoor consider the poor to have the same “moral values” as the nonpoor, and instances of poverty-perpetuating behavior in societies that do not suffer fromcreating perverse incentives for the poor to work. While there is no shortage of theories for the poor’s risk-seeking behavior being the consequence of acting contrary to “homo-economicus,” we seek evidence supporting a new theory explaining the cause of poverty: the poor’s risk-seeking behavior is in fact “homo-economicus,” instead it is our neo-classical [marginal] utility curve that needs to be changed for consumption of relief goods.
This above hypothesis challenges the widely accepted belief that marginal utility attained from goods—income, for instance—decreases as more of the good is consumed. This notion implies that goods are valued most by those who have the least. Correspondingly, the poor should stand to gain the most from an increase in income because their consumption of an additional dollar will result, ceteris paribus, in the attainment of greater marginal utility than if a nonpoor individual consumes the same dollar. However, this supposes that any “good” provides an increase in pleasure and does not account for the same “good” instead making a negative situation less bad. While we maintain that decreasing marginal utility will occur in the former scenario, we hypothesize that marginal utility is instead increasing when consumption of a good relieves pain (see Figure B). Consequently, the marginal utility of a given good will depend not only on how much it has been consumed already, but also on whether the individuals consuming it are above or below a certain level of sufficiency (see Figure A). Of most relevance is whether the subject is in a negative state of absolute utility[1] (in which misery would need to be relieved) or a positive state of absolute utility[2] (in which pleasure would be increased). This theory that goods provide increasing marginal utility below a level of sufficiency would be akin to a hungry individual[3] forgoing one bite of food for an unfair chance of consuming nine bites of food—it would seem that nearly eliminating one’s hunger would be more than nine times as valuable as just marginally reducing it. Of course, the plural of “anecdote” is not data, and thus our project seeks to find evidence for the following hypothesis: below a certain threshold of sufficiency, individuals are risk-seeking in pursuit of relief goods.[4]
The results of the above hypothesis are important not only because they could redefine the parameters of the utility curve,[5] but also because this information would be a plausible alternative explanation of extended poverty and the decisions of the poor that otherwise seem irrational.[6] Presuming that those below a given income of sufficiency (for argument’s sake, the poverty line) are in a state of misery, their marginal utility would be increasing up to a neutral level point of sufficiency. However, if our hypothesis proves correctand the marginal utility of the poor is increasing, we would be able to shed light on the motives behind the behavior of impoverished individuals: while the poor’s statistical tendencies to stay out of the workforce, drop out of school, and engage in similar behavior reduces their expected future income, our revision of the marginal utility curve explains how these individuals—who are presumably below a level of sufficiency—are actually acting rationally as they pursue such risk-seeking behavior. With a better understanding of why the poor behave the way they do, society will be able to implement policies that more effectively reduce poverty. One example would be that transfers increasing the incomes of the poor would theoretically increase the marginal utility of an added dollar, thus increasing poor individuals’ incentive to work and further supporting a positive cycle towards sufficiency.
Up to this point, not a lot of academic or experimental research has been done surrounding the concept of increasing marginal utility. It was first hypothesized by Friedman and Savage (1948) in an attempt to explain why individuals consume both insurance and lottery tickets. Essentially, the utility function provided by income is concave until a certain point at which it shifts to be convex; after a certain point, it then shifts back to concave for any point thereafter. Friedman and Savage provided this explanation without ruling out other potential explanations of this phenomenon,[7] and without relating inflection points to levels of consumption.
Karelis (2007) has most notably made the argument that marginal utility is increasing below a neutral level of sufficiency, although there has not yet been any experimental data supporting this idea. Karelis’ treatment of goods as either relievers (helping those below a level of sufficiency) or pleasers (above level of sufficiency) will prove instrumental in our explanations for how the nature of a good (and its consequent utility curve) can change depending on each individual’s status quo.
Kahneman and Tversky (1979) find experimental evidence which suggests that individuals are risk averse in the positive domain and risk seeking in the negative domain and propose “prospect theory” as an explanation. The positive and negative “domains” are defined as above and below a neutral asset level, but this level is merely defined as one that individuals have adapted to. As suggested by experimental data from Ericson and Fuster (2009), individuals tend to constantly and quickly adapt, meaning that attaining goods take individuals into the positive domain (where they will be risk averse) and losing goods take individuals into the negative domain (where they will be risk seeking).
While Kahneman and Tversky similarly acknowledge that a neutral level of wealth is the inflection point for utility curves, their analysis differs from ours given their assumption that individuals are risk averse when evaluating consumption of all goods. If our hypothesis is correct, a level of sufficiency is not defined by continuous adaptation but is instead an absolute level below which all individuals are uncomfortable.Franken, Georgieva and Muris (2006) provide additional experimental evidence that subjects who “have-not” are more likely that those “have” to exhibit risk-seeking behavior. However, their explanation for this is related to the effects of prior gains and losses in a gambling task create a moving reference point, rather than an absolute level of sufficiency. This idea of a critical level of income is similarly explored by Kwang (1965) in his explanation of why individuals buy lottery tickets. Kwang argues that the “indivisibility” of certain expenditures that a subject could otherwise not afford (i.e. a college education, a home for shelter, etc.)[8] can render gambling a rational choice given constraints of utility maximization. Although Kwang maintains that his conclusion is consistent with decreasing marginal utility, it would be similarly valid given our hypothesis of increasing marginal utility of relief goods.[9]
The concept of the increasing utility curve is further illustrated by Scholer, Zou, Fujita, Stroessner, and Higgins (2010) who reinforce the idea that risk-seeking is more likely when the individual is in a “state of loss.” Although this “state of loss” is also in reference to a malleable reference point, the results point to evidence consistent with our absolute level of sufficiency. The reasoning for this is that individuals are considered risk-averse in a positive domain because their motivation to maximize their returns is less than their motivation to avoid failure. However, if failure has already occurred (i.e. poverty has become their reality), then a fear of failure is reduced. The poor, in a sense, have much less to lose. It should also be noted that Scholer, Zou, Fujita, Stroessner, and Higgins (2010) also observed that if risky options were the only way to return to neutral level, individuals in a negative state would be risk-seeking; however, if both risky and conservative options presented opportunities to return to a neutral level, individuals would be risk neutral—in other words, they would be indifferent to choices with the same expected value. We expect the same to hold true within our experiments, largely because we maintain that the utility curve’s point of inflection (where it shifts back to decreasing marginal utility) is at the neutral level of sufficiency.
The concept that individuals tend to be risk-averse is illustrated by Holt and Laury (2002). Their study shows that individuals become even more risk-averse as the stakes increase. However, we do not feel obliged to control for individuals’ natural tendency towards risk-aversion in our experiment because if subjects in our experiment nonetheless make risk-seeking decisions, we can be even more confident that their utility curve is sloped differently for choices involving pain relief. Furthermore, Holt and Laury’s finding that subjects given merely hypothetical choices cannot accurately predict their behavior reinforces the importance of having subjects make decisions based off of choices with real consequences.
Should our hypothesis be correct, the utility curve would be redefined as follows:
3. Experimental Design:
3.1 Pre-Trial Organization
The goal of the experiment is to test whether individuals consuming goods below a level of sufficiency are risk-seeking in pursuit of relief. In order to adequately test this hypothesis, a sample of approximately two hundred Williams College students is necessary. All subjects will be assigned a specific ten minute time slot to report to the experiment, and will be notified via email. Subjects are to be brought from a waiting room into the experiment room and asked to not discuss the results with other participants upon the conclusion of the trial. The experiment room is concealed from the waiting room, preventing those waiting from observing the experiment. The procedure is expected to last about ten minutes, and the subjects are to be compensated following the completion study. One out of every twenty subjects will be compensated twenty dollars at random, under the condition that the subject completes the entire experiment. The winning subjects will be individually notified by email and asked to pick up their earnings at a predetermined location.
Once the subjects arrive, they are advised that they will undergo short, painful experiences, but that there are no major health risks. Subjects are informed of the compensation format and told that they can leave the experiment at any time, but reminded that only those who complete the entire experiment will be eligible for payment.
3.2 Pain Familiarization Stage
Before each subject enters, both water containers will be emptied and refilled with clean water to ensure cleanliness. In addition, a new layer of ice is added to the cold water to remain a temperature of approximately fourteen degrees centigrade for all trials. Fourteen degrees is consistent with Kahneman et al. (1993) and the duration of hand immersion falls within the range from this experiment. At the start of the experiment, all subjects will immerse their right hand to the wrist in the container of very cold waterfor one minute. Once the subjects place their hand in the container, they will be presented with two options for how they wish to spend the remaining five minutes. After the one minute period is over, the subjects will remove their hand from the container and will be given a one minute period to dry their hand in order to reduce the effects of numbing. The pain familiarization stage is integral to the decision-making process as it gives the subjects an opportunity to better calculate the value of each good. If the subjects did not undergo this initial submersion, uncertainty would largely factor into subject choices. Further, since the immersion is the exact same length as additional intervals, subjects experience less cognitive overload when attempting to compare the options.
3.3 Subject Decision
By the end of this one minute rest period, the subjects will be asked to choose one of the following options:
Option I (2 Minutes of Guaranteed Pain) Upon the completion of the minute, the subject puts theirleft hand to wrist in very cold water for one minute. Immediately following this interval, the subject places their right hand to wrist in very cold water for one minute. Immediately following this interval, the participant’s left hand to wrist will be placed in room temperature water. For the final minute, the participant’s right hand to wrist will be placed in room temperature water.
Option II (Opportunity for Relief) If they subject chooses Option II, the subject flips a coin, where heads yields Option A and tails yields Option B.
Option A requires the participant to put the left handto wrist in very cold water for one minute. Immediately after, the participant’s right hand to wrist will be placed in very cold water. Following this interval, the participant’s left hand to wrist will be placed in very cold water for one minute. For the final minute, the participant’s right hand to wrist will be placed in room temperature water.
Table AOption / Time / Hand/Temperature
Option I / Minute 1 / Left in Cold
Minute 2 / Right in Cold
Minute 3 / Left in Warm
Minute 4 / Right in Warm
Option II A (Heads) / Minute 1 / Left in Cold
Minute 2 / Right in Cold
Minute 3 / Left in Cold
Minute 4 / Right in Warm
Option II B (Tails) / Minute 1 / Left in Cold
Minute 2 / Right in Warm
Minute 3 / Left in Warm
Minute 4 / Right in Warm
Option B requires the participant to put their left hand to wrist in very cold water for one minute. Immediately after, the participant will then put their right hand to wrist inroom temperature water for an additional minute. Following this interval, the participant’s left hand to wrist will be placed in room temperature water for one minute. For the final minute, the participant’s right hand to wrist will be placed in room temperature water. In order to avoid the peak/end rule, both Option I and Option II conclude with the same condition (right hand in warm water). Figure A outlines the options presented to the participants.
It should be noted that during the fourth minute all options have the subject in lukewarm water in order to control for the “Peak-End” rule (Kahneman et al., 1993). In other words, we did not want to give subjects the option of ending under different conditions because their choices would have presumably been biased by a preference for ending the experiment on a high note. This phenomenon could even affect participants that are ex ante unaware of the “Peak-End” bias and influence their decisions. Additionally, regardless of the option the subject chooses, participants will all spend the exact same amount of time (six minutes) completing the study. One potential flaw that could arise would be that relief from the task provides the subject with free time. This leisure time could be used in a productive manner, and would incentivize the subject to choose one option over the other. With this design, Option I and II are structured in such a way that they both rule out the possibility of utilizing free time with regards to productivity.
Option I is the “homo-economicus” prediction because subjects that have decreasing marginal utility curves will value the certainty of an amount of relief more than a riskier alternative with equivalent expected relief. Option II is the “homo-Karelis” prediction because subjects that have increasing marginal utility curves will value the opportunity for increased relief greater than the certainty of a smaller amount of relief.