Toward A Common Behavioral Economics Perspective
With so many researchers, government agencies and consultants claiming to use behavioral science techniques, interest is booming! The business media appear to be responding with new and varied applications, methodologies and the promise of quick behavioral change, often without fully understanding much about BE. For example, we at the Behavioral Science Lab are repeatedly asked by marketers to “just apply behavioral economics” with little or no clarification of the motivations of those whose behavior we are asked to modify.
This excited readiness to accept the application of BE but with little understanding may be the result of there being very few results in marketing as attractive as those in BE. The strict adherence to the scientific method, clear definition of terms and focus on decision making all make BE results attractive to practitioners, especially marketers. For example, where in marketing research can you find a glossary of “effects” or a literature of decision biases or Nobel Prize winners?
Nevertheless, some of the confusion with which BE is perceived today must be attributable to the practitioners themselves. For example:
Do respondents express similar BE effects, or are there individual differences?
How important are individual differences?
How is BE best applied in the real world, how can it be improved?
By not answering these questions, are we investigators contributing to the confusion and misrepresentation of BE to potential users? If so, it behoves us to arrive at a common understanding of how BE works so that it reaches its full potential to enlighten and instruct.
To that end, this short discourse will attempt to offer a common perspective on BE and test it against selected findings. Most of the content will consist of “thought experiments,” intended not to minimize the value of past findings, but to place those findings in a context within which a more complete understanding of BE might be possible.
Basic Experimental Design
Many, but not all, BE studies use an experimental design that includes a choice decision with risk, payoff and often a reward. Results are interpreted as being reflective of how decisions are made and what decision commonalities or heuristics might be sufficiently consistent to be treated as new knowledge. In its simplest form, this design appears as the following:
Since any decision, even those in a laboratory setting, is not made in a vacuum, a more complete description of the decision process would be to include the presumed result of the Decision, which we will call the “Obtained Utility” as shown below.
Controlling Mental Process
Again, simply stated, we infer that selected options have greater utility than options not selected. From this behavior we infer the characteristics of a mental process or state guiding the observed choice behavior. Let’s add a component, Controlling Mental Process, to the model above about which we know neither its components nor method of operation.
Let’s try this Controlling Mental Process concept on some common BE findings.
One of the simplest BE effects to understand is Loss Aversion, the higher likelihood of selecting a choice option that avoids a loss of the same magnitude as an alternative that promises a gain. Often called the “losses loom larger than gains” phenomenon first reported by Kahneman and Tversky (1979), it is often used to explain the Endowment Effect. So applying our Controlling Mental Process concept, is there a state or process that guides loss aversion behaviour? If so, then there should be individual differences which affect decision making. In 2007, in a paper entitled “Individual-Level Loss Aversion in Riskless and Risky Choices,” Gächter, Johnson and Herrmann found that “ … in both choice tasks loss aversion increases in age, income and wealth and decreases in education.” Our conclusion from this and other studies is that some process or state affects loss aversion, exists prior to the choice task, and is related to these demographics. Operating in this way, this process or state would fit the definition of Controlling Mental Process.
This well-known BE effect is often described as the perceived higher likelihood of occurrence of events with positive, higher utility outcomes — in other words, “optimism” about the future. Unfortunately, not everyone experiences Optimism Bias equally. Strunk, Lopez and DeRubeis in “Depressive Symptoms Are Associated with Unrealistic Negative Predictions of Future Life Events” (2006) found that “A nonsignificant optimistic bias characterized participants with low depressive symptoms, whereas a significant pessimistic bias characterized participants with high depressive symptoms.” So we would conclude that differences in observed optimism appear related to the manner in which a yet unspecified Controlling Mental Process operates. A direct link between “unrealistic” optimism and physiology has been posited by Sharot, Korn and Dolan (2011), concluding that “… optimism is tied to a selective update failure and diminished neural coding of undesirable information regarding the future.” Our conclusions from these studies and others is that individual differences in Optimism Bias occur, i.e., not everyone behaves similarly as a result of it, and some pre-existing process or state “directs” it.
Intertemporal Choice (IC)
This is the well-known and researched higher likelihood of selecting a positive future outcome in the sooner it occurs. This effect does not follow the classic Discounted Utility model, but rather variable rate discounting. Intertemporal Choice suggests the higher selection likelihood of immediate versus delayed gratification, or Present Bias. So is there a uniform Present Bias effect across respondents or is there some process or state responsible for individual differences? In their 2007 paper, “Intertemporal Choice — Toward an Integrative Framework,” Berns, Laibson and Loewenstein suggest several control mechanisms for IC and cite neurological evidence for two decision states. Each of these states (control mechanisms) allow for individual differences in IC and may, themselves, conflict with each other according to the authors. So the concept of some Controlling Mental Process dealing with conflicting interactions and impacting the likelihood and degree of Present Bias appears not just feasible, but likely together with co-occurring differences in neural activity.
Finally, “internal” psychological variables such as self-control, and degree of innate risk aversion or acceptance appear to impact IC as do the “external” factors of time, stress, risk and payoff magnitude, and number of decisions.
So our premise that a Controlling Mental Process, as shown again below, regulates the liklihood of choice behavior (and perhaps the magnitude) associated with known BE effects (Loss Aversion, Optimism Bias and Intertemporal Choice) appears supportable.
What do we know about this concept? Daniel Kahneman in Thinking, Fast and Slow (2011) suggested that such a concept actually directs decision making in two ways, “fast” and “slow,” i.e., “System 1” and “System 2,” respectively. These “useful fictions,” as Kahneman refers to them, are aides in bundling the characteristics and mechanisms of such a controlling process. “Fast” decisions involve processes that are more intuitive, more impulsive, perhaps more related to perception than cognition and less likely to involve high-cost/-risk outcomes. Fast decisions are more likely to occur under time pressure, stress or an “overload” of the decision maker. One decision mechanism proposed for slow decisions is the Representativeness heuristic. This is the apparent comparison of current situations (decision options) to the utility of past situations so that an appropriate course of action can be selected with presumed greater “facility.” “Slow” decisions use mechanisms characterized by more deliberate “logical” processes with less impulsivity and more complete assessment of outcomes. Social impact and self-image also appear to impact these decisions. Anticipation (of outcomes) has been proposed as a slow decision mechanism in which an expected outcome (utility) is “forecasted.”
There is precedence for such an “expectational” basis of decision making. The Expectancy Theory of workplace performance proposed by Victor H. Vroom of Yale suggested that decisions were determined by a comparison of the effort required to achieve the expected increased salary and social prestige of advancement. Although the calculus for such decisions was not specified, Vroom was clear that the decision to assume advancement was dictated by the expectation of the sum of all rewards associated with it. This is a clear example of BE since there were both social and economic payoffs associated with the acceptance of advancement. Vroom also believed that the best strategy for management was to understand those nonfinancial “motivators” that were unique to each employee and contributed to individual higher performance.
Finally, Game Theory teaches us that feedback from the consequences of a decision must “feedback” to impact future decisions.
Utility Expectation Model
Since we are inferring something about a mental process that existed prior to the presentation of choice options, its function must be to direct actions that occur in the future together with their consequences. Its existence must precede the presentation of choice options even if its impact cannot be assessed until after a decision scenario is presented and behavior observed. So let’s relabel Controlling Mental Process the “Utility Expectation Model.” Such a model would have the following components:
Fast and slow decision processes
External factors impacting decisions (time, number of decisions, stressors, etc.)
Internal factors specific to the decision maker (risk aversion, self-control, wealth, etc.)
BE effects themselves (Loss Aversion, Optimism Bias, Intertemporal Choice, Endowment and others)
Feedback of experienced utility based either on the immediate result of a decision, the recollection of the results of prior decisions or the results of similar prior decisions.
Some educated guesses about how such a Utility Expectation Model would “manage” these components follow:
Affective states (delight, dread, anxiety, social approbation, etc.) and environmental decision conditions (time, number of decisions, others) could play significant roles.
“Conflict” (Approach-Avoidance) resolution suggests a systems structure with gates, paths and recursion.
Psychological variables (self control, risk aversion, locus of control) may act as nodes in the system structure.
BE effects may act as “moderators” or “accellerators.”
Heuristics might be just be the “short form” of the full model.
Mental constructs would have to have invariant meaning across respondents, so a new way of constructing and naming these constructs is probably necessary.
To avoid “pollution” with experimenters’ beliefs and biases, most of the “work” to define such mental constructs would probably be done by respondents themselves using a formal process.
Large scale sample sizes would be necsssary to capture the breadth of individual differences.
Individual respondent models would be necessary to fully explain the role of individual differences.
Respondents with similar individual models would be “aggregate-able” into common decision “types.”
Model components may impact the model quite differently dependent on decision type.
Validation on a respondent level and within decision type would be needed to confirm the results of multiple models.
Different decision types may map to different game playing strategies.
Common Behavioral Economics Perspective
Our Common Behavioral Economics Perspective now appears as follows with a feedback loop:
When complete, this Common Behavioral Economics Perspective could be used to answer some interesting questions as follow:
Are fast and slow process decisions really supported by the same underlying mechanism once the internal and external factors impacting decisions are known, or are they really two different mechanisms?
Are decisions that appear strictly “irrational” actually “rational” when the complete set of decision elements are known, such as the internal and external factors above? Are there really any “irrational” decisions, or is “irrational” just an explanation of not knowing the true drivers of the utility expectation?
What is the Model impact of known BE biases on the expectation of utility?
How many types of similar decision types for a purchase in a given product class are in use by respondents?
To what extent can respondents be disuaded of their own biases by “fixes” generated by understanding the bases of expected utility?
What would intervention “fixes” look like if a decision process was fully understood? Would there be only one optimal intervention per process, or are optima possible?
For decision processes involving investment, healthcare, education, risk protection, etc., could optimal decision set(s) be constructed for individuals?
Is there a metric for how well a decision option satisfies the “requirements” of the Utility Expectation Model?
How does culture impact the expectation of utility? What are the model’s components?
Let’s build on the content and scientific discipline of BE and expand it beyond its current limitations to form the basis of a broader understanding of human behavior. After all, isn’t our ultimate objective to better understand, ourselves, and our decisions?
Berns, G. S., Laibson, D. and Loewenstein, G. (2007). Intertemporal Choice–Toward an Integrative Framework. Trends in Cognitive Sciences 11(11), 482-488.
Gächter, S., Johnson, E. J. and Herrmann, A. (2007). Individual-Level Loss Aversion in Riskless and Risky Choices, IZA Discussion Papers 2961, Institute for the Study of Labor (IZA).
Kahneman, D. (2011). Thinking, Fast and Slow, Farrar, Straus & Giroux.
Kahneman, D. and Tversky, A. (1979). Prospect Theory: An Analysis of Decision under Risk, Econometrica, 47, 263-291.
Sharot, T., Korn, C.W. and Dolan, R.J. (2011). How Unrealistic Optimism Is Maintained in the Face of Reality. Natural Neuroscience, October 9, 14(11), 1475-1479.
Strunk, D.R., Lopez, H. and DeRubeis, R.J. (2006). Depressive Symptoms Are Associated with Unrealistic Negative Predictions of Future Life Events. Behavioral Research Therapy, June, 44(6), 861-82.
Vroom, V. H. (1964). Work and Motivation. McGraw Hill.
Special thanks for contributions to this paper to Christian Goy, Co-founder & Managing Director, Behavioral Science Lab and Isabelle Zdatny, Temkin Group.