February 2012

Review of Can Annuity Purchases be Influenced?

By G. A. (Sandy) Mackenzie

Both economists and actuaries are well aware of the contribution to well-being in retirement that a permanent stream of income can bring. Economists have long stressed the great utility of life annuities in providing insurance against unexpected longevity. Nonetheless, the market for these instruments has remained stubbornly narrow. Both objective and subjective reasons have been adduced to explain the annuity puzzle, as it has been called. Recently, the focus of researchers has shifted from objective to subjective explanations for the lack of popularity of annuities. Studies have suggested that the way the annuity investment decision is "framed" can have a material impact on the decision to buy. For example, the investment aspect of an annuity can be emphasized, which directs the prospective annuitant's attention to the risk of loss entailed by premature death. Alternatively, their consumption or insurance aspect can be emphasized, which directs attention to their basic role in avoiding the exhaustion of the annuitant's assets late in life.

In light of the importance of the annuity framing issue, the Society of Actuaries commissioned a major study to see how different kinds of information might influence the decision to purchase an annuity in an experimental setting. The study was carried out by a team from Behavioral Research Associates headed by Jodi di Cenzo. The BRA report, Can Annuity Purchases be Influenced? begins with an overview of recent developments that might affect the demand for annuities. It argues convincingly that a significant decline in the share of assets annuitized at retirement has taken place in the past few decades. This decline brings into sharp relief the need to increase the demand for annuities. The carefully constructed experiment described in this important report investigates the impact on demand for annuities of both the provision of information on annuities and retirement and the way that information is packaged and presented. In what follows, we shall briefly describe the design of the experiment and then summarize its most important results, making a few remarks on the significance of its findings as we go.

The experiment had 1009 participants aged 45 to 75. Participants were split about equally between men and women, and between retired and pre-retired. The average age of retired participants was 63.6 years, and of the pre-retirees was 55.9. The retired and pre-retired were each allocated to either a control group, or one of four treatment groups. Participants in the four treatment groups were each given a short summary of the circumstances of a hypothetical working man (for male participants) or woman (female participants), which was accompanied by a photo. They were then asked three questions intended to engage them in the issues of retirement finance: (1) At what age do you expect John/Jane to retire from his/her full time job, assuming he/she is a typical American employee; (2) Until what age do you expect John/Jane to live, assuming he/she is a typical American man/woman; (3) How likely do you think John/Jane would be to buy a life annuity?

Feedback, which differed across treatment groups was given after each question was answered. The first group was given general information on retirement ages, life expectancy and annuity suitability; the second was given an account of the behavioral biases that are thought to affect the annuity decision; the third was presented with a combination of the information presented to the first and second groups; and the fourth was given an anecdote that stressed the undesirable consequences of running out of money.

The treatment groups and the control group were then asked to answer two questions: the first gauged their intentions regarding annuity purchase and the second gauged the likelihood of a purchase on a seven-point scale. The first question was intended to capture the impact of information on "behavioral intentions," the second the effect on attitudes. The second question could also be interpreted as a way of assessing the desire underlying the stated intention to purchase.

The most striking conclusion was that the provision of information could undoubtedly sway the decision or expressed intent to buy an annuity for pre-retirees, if not for retirees. Taking no account of the influence of demographic or socio-economic factors on the choice, the share of pre-retirees opting for an annuity increased from nine percent for the control group to 30 percent for treatment group one, who were given factual information. Smaller but still large increases were observed for the other treatment groups. In the case of retirees, however, where the share of participants choosing an annuity in the control group was 18 percent, only the anecdotal treatment group, in which 23 percent opted for an annuity, recorded a higher share.

These findings were not overturned by probit analysis, which tested a large number of socio- economic and demographic variables in addition to the effect of the four treatments. The provision of factual information to pre-retirees was estimated to increase the probability of annuity purchase from six percent to 21 percent, and the provision of anecdotal information to increase it to 20 percent. These are large effects, and illustrate the potential influence of well-designed information campaigns. The finding that the provision of information alone is more effective than the joint provision of information and an account of behavioral biases seems puzzling, but could reflect the workings of information overload, particularly if the concept of behavioral bias is hard to convey. The effectiveness of anecdotal information also suggests that information in narrative form, although it may not be as comprehensive as a series of factoids or explanations of behavioral bias may be more memorable.

The study also found that the degree of self-assessed familiarity with annuities had a significantly positive effect on the annuity decision for both groups, and that pre-retirees who expressed concerns about their retirement or who expected to live longer than average were more likely to buy an annuity. As the authors point out, the relationship between degree of familiarity and purchase is not surprising if some degree of familiarity with annuities is a precondition for buying one. As for negative influences, participants expecting to invest in retirement assets and draw down their balances were less likely to purchase an annuity. Again, this disposition would tend to imply a lack of interest in annuities.

The apparent ineffectiveness of the experimental interventions with retirees is not totally surprising. A retiree is more likely to have addressed the issue of the distribution of his or her nest egg than a pre-retiree is, and is more likely to have actually bought an annuity. The information presented in the various treatments may in consequence be less novel to him or her than it would be to someone who was still working. In any event, the authors are right to conclude that waiting until retirement to provide information on annuities and their merits is likely to be a less effective strategy than providing it at an earlier stage, when there is less chance that decisions have been made or attitudes hardened.

How much annuitization (of wealth at retirement) is a good thing? The experiment provides us with insight into the effect of information on the propensity of future retirees to acquire an annuity; it cannot shed light on what determines the share of annuitized wealth. Even without behavioral biases and other impediments to demand for annuities, it is probable that far less than 100 percent of the population would wish to buy an annuity. A large share of the population would already have a very large share of their wealth at retirement in annuitized form because of Social Security. If the data permitted, it might be possible to uncover the relationship between interest in annuities and the share of wealth in annuitized form.

The findings of this important experiment could be very useful to both the insurance companies that sell annuities and to those government agencies that promote financial literacy. It would be relatively easy for an annuity provider to design brochures and other promotional material in the spirit of treatment one and three, if not treatment two. Insurance companies already do something like this. As for the promotion of financial literacy as a public good, the experiment suggests that a relatively simple or informal approach based on treatments one and four might be superior to a didactic approach as a means of encouraging interest in annuities. More generally, the study suggests that the demand for annuities could be materially increased by the provision of the right information at the right time.

G.A. (Sandy) Mackenzie is senior advisor in the Public Policy Institute at AARP. He can be reached at gmackenzie@aarp.org.