Members Suffer Consequences Of Overconfidence
By: Richard Deaves
A recent survey shows Defined Contribution pension plan members have a low level of knowledge and, yet, are overconfident about their knowledge. Richard Deaves, of the Michael G. DeGroote School of Business at McMaster University, explores the findings of this survey and takes a look at what is best for employers that offer DC pensions plans.
In the winter of 2004, a survey of Canadian DC pension plan members was commissioned by SEI Investments to obtain information on the satisfaction of members with their current pension plans and to elicit desires for changes. More than 2,000 members from 17 different pension plans were surveyed. Participants were from a broad spectrum of industries, geographical areas, and socioeconomic groups.
In addition to a set of questions exploring satisfaction and desires, a number of additional questions were included in order to investigate the level of knowledge members had on the nature of their pensions and on investments in general. While it turned out that their state of knowledge was quite low, somewhat surprisingly, they were aware of this. It was also apparent that members were overconfident in the sense that they thought their knowledge was greater than it actually was. Let us turn to some of the survey evidence on this issue.
When respondents were asked to agree or disagree with the statement ‘You would consider yourself very knowledgeable about retirement planning,’ only 15 per cent strongly agreed. This lack of comfort level in their knowledge came through in a series of questions where respondents were asked to characterize their pensions. Unfortunately, the perceptions they had of the nature of their pension plan design were frequently at variance with reality. For example, 75 per cent thought they could personally make contributions to the plan, when in actual fact the true percentage was 51 per cent. And 47 per cent thought the plan was compulsory, when in fact that state of affairs was true for 71 per cent of respondents.
Additionally, several questions were included to investigate general investment knowledge. Two in particular dealt with (admittedly difficult) Canadian capital market history (see Question 1 and Question 2, in graphic below). The correct answers are the third alternative and the fourth alternative respectively.
On each of the questions, since there are five choices, a guess would have a 20 per cent probability of being correct. So if only 20 per cent of the entire sample get a question right, the logical inference is that there is virtually no knowledge. In actual fact, the correct response rates were 32 per cent on the first question and eight per cent on the second. On average, over the two questions, the percentage of correct answers was 20 per cent (which is equivalent to guessing). For the first question, there did seem to be some knowledge, since 32 per cent is different from 20 per cent in a statistically significant sense. For the second question, the percentage was eight per cent (which is worse than mere guessing). The likely explanation is that younger people did not realize that interest rates were much higher at the beginning of the 20-year period than at the end.
While lack of knowledge is undesirable, thinking you know more than you in fact do know, or being overconfident, is dangerous. The reason is while the former leads to indecision and the desire for education and advice, the latter can lead to misinformed decision-making.
Before investigating whether the survey respondents were overconfident, some background is appropriate.
Confidence And Overconfidence
What do researchers know about overconfidence and its presence in financial markets? Let me begin with a definition: overconfidence is the tendency for people to overestimate their knowledge, abilities, and the precision of their information. Let me stress that confidence (which can also be called certainty) and overconfidence are two very different things. The former is a positive attribute. It is a sign of high selfesteem and social skills that are likely to enhance job performance. But, for it to be a clear plus, it must be matched by knowledge. Overconfidence arises when the knowledge side of things falls short.
Indeed this seems to be the norm. That most people are overconfident is well-documented by researchers in both the psychology and financial economics literatures. Overconfidence has been observed and measured in several ways. When people are asked to rate themselves in terms of skills and abilities, generally more than 50 per cent say they are better than average. This is the so-called betterthan- average effect.
Another approach is to look for excessive optimism (or illusion of control). Those so afflicted think they have more control over events than objectively can be true. For example, when asked how likely it is that certain bad events (such as their small business failing) will occur, people tend to discount the likelihood of such outcomes. Another popular way to detect a discrepancy between knowledge and knowledge perception is to conduct what can be called a ‘calibration’ test. The idea is to look for a measurable imbalance between actual knowledge and perceived knowledge. Avariant of a calibration test was performed using the SEI survey data. To get a sense of individuals’ overconfidence levels, both of the investment knowledge questions outlined above had adjunct portions, which read as follows:
To obtain the average level of certainty on a question, we take a weighted average of the five responses. For example, suppose there are 100 respondents. Ten (10 per cent of them) say ‘absolutely sure,’ which obviously means 100 per cent sure; 20 (20 per cent of them) say ‘80 per cent sure;’ 30 (30 per cent of them) say ‘60% sure;’ 20 (20 per cent of them) say ‘40% sure;’ and the remaining 10 (10% of them) select ‘not at all sure,’ which we take to mean 20 per cent sure. The average level of certainty would be calculated as:
.1 * 100% + .2 * 80% + .3 * 60% + .2 * 40% and .1 * 20% = 60%
If less than 60 per cent of respondents get the question right, the group as a whole is judged to be overconfident; while if more than 60 per cent of respondents get it right, the group as a whole is underconfident.
It should be noted that two questions provide insufficient information to say if an individual is overconfident. The sample is simply too small. In fact, quite a few more questions would be needed. That said, given the sample size, we should be able to say something about knowledge versus knowledge perception for the group as a whole.
Knowledge And Actual Knowledge
What do we find? The average levels of certainty were 41 per cent and 43 per cent respectively, averaging in at 42 per cent. Defining overconfidence as the difference between knowledge perception (the level of certainty) and actual knowledge (the percentage of correct answers), simple arithmetic tells us that the average level of overconfidence in this sample was 22 per cent (42 per cent minus 20 per cent). This was highly significantly different from zero in a statistical sense. In short, the sample of Canadian DC pension plan members was quite overconfident.
What are the implications of these findings? There is evidence that overconfidence can lead to excessive trading. In experimental work that I recently completed with Erik Lüders, of Laval University, and Rosemary Luo, of McMaster University, it is shown that people’s relative level of overconfidence is correlated with their proclivity to trade. Those who were most overconfident traded too much. The result of their excess trading was inferior performance.
In the context of self-directed retirement accounts, excessive trading is probably less of a worry, as a number of research studies have shown that many members are subject to inertia and never alter their original allocation. The bigger danger is likely taking injudicious action at the outset and never altering course – all the time confidently assuming you are doing the right thing.
Other questions in the survey indicated a number of investment errors that were driven by psychologically-based (or behavioural) biases. To take one example, DC plan members misunderstood the nature of asset allocation. This was apparent because, when presented with a hypothetical menu of fund choices with more equity funds to choose from and asked to express preferences, they tilted towards equity. Conversely, when presented with another menu, this time with more bond choices, they tilted towards bonds. Their inability to maintain consistency revealed their lack of understanding in this regard. In the latter case, as many as 38 per cent thought that a portfolio equity share of 30 per cent (or even much lower) was appropriate. As is well known, such low equity exposure is likely to leave DC pension plan members with a shortfall when they retire.
Likely A Pipe Dream
So what is best? While an investmentsavvy DC plan membership is ideal, at present this is likely a pipe dream. Asecond best alternative is a membership which realizes that it has a lot to learn and is thus intent on making use of educational offerings or third-party advice. But this is where overconfidence gets in the way – “I know enough” or “I’ll be fine.” The future reality unfortunately might prove otherwise.
Sponsors are aware of the risk that when the future comes, they will be blamed. In the fall of 2003, SEI Investments commissioned another survey – this one of plan sponsors. There were 120 in the sample. Several findings were salient. There were fears of litigation, with 24 per cent of respondents feeling that their organization would face litigation within the next two years. As well, 53 per cent of plan sponsors believed that educating their plan members was the most important and effective way to protect against litigation, yet only 11 per cent thought they were performing well in this area.
Richard Deaves is chair of finance at the Michael G. DeGroote School of Business at McMaster University.
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -