“Attribute Substitution in Early Enrollment Decisions into Medicare PDPs,” Frakt, Pizer (2007)
A few years ago my colleague Steve Pizer and I were funded to study new plan options that became available to Medicare beneficiaries in 2006, chief among them stand-alone prescription drug plans (PDPs). PDPs quickly became the most popular means within Medicare for beneficiaries to obtain drug coverage (the other option being Medicare Advantage plans that offer drug benefits). However, PDPs were not uniformly more popular. There was (and is) some geographic variation in their popularity.
We were interested in understanding what factors were associated with geographic variation in PDP enrollment. That is, why were PDPs more popular in some regions than in others? Standard economics models to investigate such a question would include measures of price (premium), competition, and demand and supply factors. We developed such a model and then did something a little unusual.
On a lark, we threw in the percent of a county’s electorate that voted for Bush in 2004. Interestingly, this turned out to be very strongly and positively correlated with proportion of beneficiaries in a county who enrolled in a PDP. Why would this be? For an answer (or a hypothesis really) we turned to behavioral economics and wrote up the results in a 2007 Health Economics paper titled “Attribute Substitution in Early Enrollment Decisions into Medicare Prescription Drug Plans.”
The key notion from behavioral economics upon which our hypothesis hangs is that of “attribute substitution.” Attribute substitution is a form of intuitive thinking in which readily accessible attributes of an object are used as proxies for the less accessible attributes relevant to a rational decision. In the case of PDPs, we hypothesized that beneficiaries might have substituted the recommendations of respected political leaders for the less accessible calculations of expected financial values of Medicare plans.
To put it bluntly, perhaps some beneficiaries heard Bush and others in his Administration touting the benefits of the new drug plans. Finding it otherwise difficult to make their own independent assessment of the relative merits of various coverage options, beneficiaries may have substituted officials’ enthusiasm for PDPs for their own prediction of its benefits. That’s a type of shortcut many of us make: we rely on the “expert” advice of others we trust rather than do our own analysis. In this case, there is geographic variation of degree of trust in the Bush Administration, which we operationalized as proportion of 2004 Bush vote.
We found that elasticity of PDP enrollment with respect to the Bush vote to be 0.14 (a 10% change in Bush vote is associated with a 1.4% change in PDP market share). To obtain a sense of the relative importance of this effect, we calculated the change in PDP enrollment due to a one-standard deviation change in each of the independent variables in our model separately. We found that the effect of the Bush vote is larger than the effect of other variables that are generally accepted to be important and relevant factors associated with enrollment decisions: premium, level of beneficiary educational attainment, county urban/rural status, provider density, income, and diagnosis based risk score.
So, an administration’s enthusiasm and popularity can have a significant impact on the early response to a new program. That’s a fairly intuitive result, and it is nice to see it is supported by the data. This paper was an interesting walk through a small tract of behavior economics. It is something I’d like to pursue further but not something for which I’m funded. So it will likely be a long time before I try anything like this again.
Income Inequality and Behavioral Economics
This post is co-authored by Julian Jamison, economist with the Federal Reserve Bank of Boston (*), and Austin Frakt.
Bruce Bartlet reacted to last week’s The American Prospect article on income inequality by Dalton Conley thus:
I have never understood how I am worse off if the top 1% of households increase their share of national wealth or income as long as the absolute level of wealth and income of the other 99% is unchanged. It may be aesthetically displeasing, but it doesn’t impose any actual costs on anyone as long as the pie is not fixed.
Kevin Drum, whose post brought this latest blogosphere skirmish over the issue to my attention, thinks the real problem is stagnation of middle class wages:
Rising inequality, then, is just a symptom of the real problem: sluggish middle class wages in a country that’s been growing energetically for decades. That’s the core problem. Get median wages growing at the same rate as the country itself and inequality will take care of itself because there will automatically be less money left over for the rich.
As these quotes suggest, much of the focus in debates over income inequality is on its extent and causes, its socio-economic consequences, and what if anything can and should be done about it. But I think there’s something missing: why does income inequality bother some of us at all and others not in the least? One is tempted to answer, “It doesn’t bother the ‘haves’ because they’re on top. It bothers the ‘have nots’ (or the ‘have less’) because they, well, have less and want more. Don’t we all?” But I think the “have less” can get more (and some argue they have) and still be dissatisfied, and justifiably so. More doesn’t necessarily make us happy, even if neo-classical notions of rationality suggests it ought to.
For the moment let’s ignore income inequality entirely and ask whether more money (or GDP/capita) makes people happier. The neo-classical answer is yes, always. The recent stereotypical answer is: yes, very much up to some minimal level, but very little after that. Maybe it helps in the short-run, but people adapt very quickly to higher income. Furthermore, people don’t correctly predict this adaptation, so again they overestimate the effects of increased wealth. So this is a real problem with normal comparisons of who’s doing well.
Separately, behavioral economics studies also show that relative income seems to matter, and there are good evolutionary reasons for why that might be the case. It’s important for us social animals to be respected and held in high regard by our peers and potential competitors. Also, there’s the sense of schadenfreude when someone else does poorly, but this is probably not something we should promote in policy.
So again, this suggests that an unevenly rising tide might not be a great thing overall–not for ethical reasons (which are fine as well) but for purely utilitarian ones. Of course it also depends how one defines one’s reference group: community or country? people in your church or your workplace or who you grew up with? Can government change your framing so as to induce you to compare yourself to a different group and thus be happier? Would that be a reasonable policy goal? Would it need to be secret to work?
These questions are suggested by appeals to behavioral economics but are rarely considered in discussions of policy (as far as we know). Why not? Shouldn’t policy debates concerning income inequality be sensitive to the teachings of behavioral economics? After all, neo-classical notions of welfare, utility, and rationality do not have unique claims to legitimacy. They’re tradition, and in many instances they’ve been shown to be at odds with how we really feel, think, and behave. A sufficiently large degree of income inequality rubs (some of) us the wrong way because that’s the way we are. Of course it is hard to go from that statement of reality to policy prescriptions. But if we were to try, what would we come up with? What’s the behavioral economist’s solution to the problem of income inequality? By now at least that should be viewed as a legitimate question.
(*) The views and opinions expressed in this post do not necessarily represent those of the Federal Reserve Bank of Boston or the Federal Reserve System.
Behavioral Economics in the Supreme Court
The clash between a priori and experimental economics was joined in the Supreme Court last month in the case of Jones v. Harris Associates. The narrow question is the standard under which compensation to mutual fund managers should be judged excessive. But how the Court answers this question may have broad implications for future judicial recognition and remediation of market failure in diverse areas of law.
The question comes to the Court by way of a law-and-economics nerd’s equivalent of an Ali-Frazier bout. In the efficient-market corner, United States Seventh Circuit Court of Appeals Chief Judge and University of Chicago professor Frank Easterbrook dismissed concerns that most investors are too unsophisticated to compare prices on the ground that it generally takes just a few sophisticated investors to create sufficient competitive pressure to protect the rest. Against him, fellow Chicago professor and past Chief Judge Richard Posner marshaled empirical evidence that markets have in fact failed to curb excessive pay to fund managers.
While Posner couches his claim of market failure largely in terms of agency problems and moral hazard, a friend of the court brief by scholars Robert Litan, Joseph Mason, and Ian Ayres advances the argument on grounds of behavioral economics and informational asymmetries. The “cognitive anomalies” that they claim “render it nearly impossible for the vast majority of investors to assess the price-adjusted quality of mutual funds” include:
- “Misperceptions of chance” – e.g., the widespread and incorrect impression that “if a roulette wheel has repeatedly hit black, then red is somehow due.”
- “Sample-size neglect” which “means that investors will be far too likely to consider a few years of above-average mutual fund performance evidence of managerial skill when in reality it has been driven by the random fluctuations inherent to the stock market.”
- “Loss aversion” exacerbated by “mental accounting,” which causes investors to “segregate their portfolios based on the principle of limiting their disutility from losses” rather than “considering their financial investments in terms of an aggregate portfolio of investments as traditional financial economics assumes.”
The resulting “propensity to sell winners too early and hold losers too long,” coupled with information costs associated with identifying which funds are winners and losers net of fees, make it “difficult for investors to recognize and reward high-quality, low-cost mutual funds.”
The remainder of the brief goes on to show that high-cost, low-quality mutual funds are not being disciplined by the market, and concludes that the “few sophisticated investors” hypothesis advanced by Judge Easterbrook lacks empirical support. Nevertheless, Chief Justice Roberts and Justice Scalia were reportedly receptive to this theory during oral argument. Where the remaining Justices will come down remains to be seen. But if behavioral-economic arguments can attract a majority of the Court in this case, who can say what neo-classical economic orthodoxy will be the next to fall?
Does Behavioral Economics Matter?
Sendhil Mullainathan and Richard Thaler (M & T) address some interesting questions in their paper Behavioral Economics (September 2000). In doing so they justify the import and relevance of the concepts of bounded rationality, bounded willpower, bounded self-interest. Their application is finance and savings.
1. Don’t market incentives lead to rational choices? While it is true that incentives matter, in that they influence behavior, they are not decisive. The market may reward choice A over B so there is an incentive to choose A, but they do not force the choice of A.
2. Aren’t irrational choices arbitraged away? No arbitrage opportunities exist for all sets of choices. M & T use the example of the fictitious economist Sam who goes into the field of behavioral economics even though he could earn more money in finance. Remuneratively speaking he made an irrational choice. But there is no arbitrage opportunity for him or anyone else, not for this choice or his choice of how much to save for retirement, wife, car, and so forth.
3. Didn’t evolution select for rationality? Not at all. Evolutionary arguments can explain irrational behavior as well. It may, in fact, have been an advantage to be or appear overconfident. Appearing so would have provided an incentive for one’s foes to back down. Is the “irrational exuberance” of the modern age a vestige of evolutionary selection? It is at least plausible.
4. Don’t we learn from our irrational choices and correct them? Opportunity cost can prevent the switching from a sub-optimal choice to an optimal one. Even if one is not stuck in a non-optimal equilibrium, learning can take longer than the time scales of a changing environment allow. That is, by the time you’d have learned the optimal way to decide something the world has changed: there are new choices, you have a different income, different needs, and so forth. Finally, sometimes we only get to decide once. There are few, if any, chances to learn from our retirement saving decisions and the cost of experimentation is high.
Next, M & T focus on three unrealistic assumptions of standard economics: unbounded rationality, unbounded willpower, and unbounded selfishness.
5. What’s the problem with unbounded rationality? Well, it is just wrong. People don’t have unlimited brainpower. Assuming they do is bad economics, “the equivalent of presuming the existence of a free lunch.” There are loads of empirical studies that demonstrate various ways in which people are not rational; they solve problems with heuristics that lead to sub-optimal results.
6. Do people choose the optimum (rational) option even when they identify it? No, not always, due to lack of self-control. It is common for humans to overeat, over drink, over spend, under exercise, under save, under work, and so on. People procrastinate. In short, people have bounded willpower.
7. Finally, people are not as selfish as rational actors would be.
What are the implications for finance and savings?
8. First, M & T dispense with the efficient markets hypothesis (EMH) by describing some well-known ways in which it has clearly been violated.
9. If the EMH doesn’t hold, and moreover if markets sometimes over- and sometimes under-react, there doesn’t seem to be a unifying framework to explain market behavior. What next? M & T only make a few passing remarks about emerging research on this question. It isn’t clear they’ve made any real contribution in their paper so I’m not sure why they even raise the issue.
10. Turning to savings, M & T describe how bounded willpower explain lack of sufficient saving for retirement. The phenomenon of mental accounts explain the relative increase in savings that occurred when personal IRAs became available. When funds were mentally designated for retirement and placed in an IRA they were less likely to be used for other purposes.
Well, that was a rather weak conclusion to a paper with a very interesting start. No doubt behavioral economics can and has made important contributions to personal finance. They’re not to be found in this paper. I believe behavioral economics matters more than M & T let on. If you read their paper, do so for points 1-7 above. Those are well made.
Biennial Mammogram Screening and Behavioral Factors
The U.S. Preventive Services Task Force (USPSTF) new mammogram recommendations include biennial screening for women between the ages of 50 and 74. Prior recommendations were for an interval as short as one year and up to two. My experience suggests that two-year intervals are tricky, and I speculate that behavioral factors could lead to actual intervals that are much longer.
I’m young(ish) and healthy. Therefore, my primary care doctor recommends I visit him only ever other year, as opposed to the more customary annual visits most older and/or sicker individuals experience. But his office’s scheduling software does not permit visits to be booked more than one year out (he works at a large, sophisticated hospital). Therefore, instead of scheduling my next visit while I’m paying my co-pay, I have to remember to do so one year later (for the next year). By the way, my childrens’ pediatrician’s office cannot schedule beyond one year either. So this seems to be a common problem.
That’s not so good. How many people can remember to make a phone call in one year to schedule something that won’t occur for two years? The task is easier with modern personal electronic calendars, but I doubt a very large fraction of the population can manage this. The only way to make it work is for the physician’s office to handle it and, right now, I’m skeptical many can (extrapolating from my own experience).
Fundamentally there are behavioral considerations here. One year intervals are somewhat easy to keep track of. Anything longer isn’t. Will implementation of biennial mamogram screenings actually lead to much longer intervals? Is it possible that annual screenings are preferable just for behavioral reasons? I think it is at least plausible that the answers are “yes” to both of these. This would seem to be a good topic for behavioral economists. (I know a few…stay tuned.)
Notes on a Nobelist, Part III: Kahneman on Framing
This is the third post in a three-post series that summarizes Daniel Kahneman’s Nobel lecture as printed in The American Economic Review in December 2003 (full free version). The first post was on heuristics of judgement and the second was on choice with risk. This post is about framing effects. Most of what follows is a paraphrasing of Kahneman’s words. Comments that are more fully my own are [in brackets].
The framing effect is observed when different but informationally (factually) equivalent presentations lead to different outcomes. The effect is achieved by alteration of the salience of different characteristics. [This idea has become well-known through intense media focus on political and issue polls and how their outcomes differ by phrasing.]
A nice example is suggested by Thomas Schelling in his book Choice and Consequence: Are you morally offended by tax deductions? I suspect not. What do you think of a tax policy that allows a larger reduction in taxes for the rich than the poor? Abhorrent, no? In fact, that’s exactly the result of tax deductions. This is framing at work.
Kahnmen explains the principle of framing as “the passive acceptance of the formulation given.” This leads to the apparent preference of the default option. For instance, enrollment of organ donation programs is substantially higher when individuals are assigned to it and need to opt out as opposed to excluded unless they opt in. Clearly framing is inconsistent with assumptions of rationality.
Notes on a Nobelist, Part II: Kahneman on Choice with Risk
This post is short. Use the balance of your time to read something else. Can’t find anything? Try one of my favorite sources.
This is the second post in a three-post series that summarizes Daniel Kahneman’s Nobel lecture as printed in The American Economic Review in December 2003 (full free version). The first post was on heuristics of judgement and a subsequent post will be on framing effects. This post is about models of choice in the presence of risk. As before, my comments are [in brackets] (actually, there is only one comment).
We only need to consult our own experience to verify that perception is reference-dependent. A bath of 80°F will feel warm relative to a 60°F one and cool relative to a 100°F one. The same is true about our satisfaction with respect to different states of wealth. That is, we can be risk averse with respect to gains while loss averse with respect to loss. It matters a great deal what the starting point (the reference point) is.
Kahneman goes to great length in his article to stress that prior work that assumed reference-independence is a poor model of individual behavior. A main idea of his “prospect theory” is that the degree to which we value a unit of wealth depends on our initial endowment. Studies have found that the selling price an individual would assign to a good is often a factor of two higher than the buying price he would assign. That is, the value assigned to the good depends on whether or not you own it (the “endowment effect”).
Such reference-dependent decision making helps promote the maintenance of the status quo as the disadvantages of alternatives loom larger than advantages. [Political implications...]
Notes on a Nobelist, Part I: Kahneman on Heuristics of Judgement
I’ve been reading some behavior economics literature. I started with the excellent overviews by Camerer and Camerer and Lowenstein and then moved on to Daniel Kahneman’s Nobel lecture. For his work in behavioral economics Kahneman shared the Nobel prize in economics in 2002 with Vernon Smith. A revised version of Kahneman’s Nobel lecture was printed in The American Economic Review in December 2003 (full free version). Kahneman’s lecture (paper) reviews the three lines of his research: (1) heuristics of judgement, (2) models of choice under risk, and (3) framing effects. In this and two subsequent posts I provide my notes from my reading of the paper.
This first post is on (1) heuristics of judgement. Naturally, the second post will be on (2) models of choice under risk, and the third post will be on (3) framing effects. Most of what follows is a paraphrasing of Kahneman’s words. Comments that are more fully my own are [in brackets].
There are two modes of thought: reasoning and intuition. Reasoning may lead to more rational decisions while intuition is more likely to lead to faulty judgements. This is illustrated by a simple puzzle credited to Shane Frederick. “A bat and a ball cost $1.10 in total. The bat costs $1 more than the ball. How much does the ball cost?” Likely more than half of individuals can solve this puzzle with some moments of thought. Yet most provide an incorrect answer of “10 cents” relying on intuition alone.
A useful distinction between reasoning and intuition is the degree of effort. Reasoning requires effort and diverts the mind from other activities, forcing it to take shortcuts that would not be taken if not distracted by mental effort. Intuition requires little effort. One can operate a car intuitively (without much mental effort) and therefore conduct a conversation requiring substantial mental effort at the same time. [Though in such a case one may operate the car perfectly but navigate it poorly (missing turns, etc.).]
Intuition is akin to perception: it is immediately accessible and requires no effort. It generates impressions, many implicit, non-voluntary, and not expressed, of the attributes of objects and ideas. More reasoned judgement may be built upon impressions and are explicit and intentional. This hypothesis of intuition–that it is perception-like–motivates a focus on analogies from perception. Tricks of perception are at least metaphors of intuitive errors. The immediate accessibility (or “intuitive grasp”) of the properties of an object upon seeing it bear a cognitive resemblance to our readily intuited thoughts about an idea upon encountering it. The representation of a prototype of an object [e.g. an apple] is highly accessible, [as is our intuition about an idea, choice, puzzle, or decision (the source of bias?)]. The term natural assessments applies to the attributes immediately brought to mind without intention or effort. The natural assessment of “good” or “bad” plays a key role in our judgements and is one for which special brain circuitry exists.
Accessibility of properties about an object, situation, or thought is a continuum, varies by individual, and can be altered with training. A chess master has immediate access to the attributes of a configuration of chess pieces that a novice does not perceive even with intense effort. Salience plays a role in relative accessibility of attributes. Context, grammar, and various forms of presentation influence salience. The degree of accessibility of various thoughts differ upon hearing each of the sentences “Team A beat team B” and “Team B lost to team A” even though they are logically equivalent. Sex sells even when it has nothing to do with the product with which it is associated.
Heuristics of judgement are achieved by an operation of attribute substitution, a form of intuitive thinking in which readily accessible attributes (so called heuristic attributes) of an object are used as proxies for the less accessible attributes (so called target attributes) relevant to a rational decision. Optical illusions take advantage of perceptual attribute substitution. Attribute substitution is a pervasive shortcut and the list of heuristic attributes identified is long, reviewed by Kahneman in his paper, but beyond the scope of this post.
Subscribe
