«FIVE STEPS TO PLANNING SUCCESS. EXPERIMENTAL EVIDENCE FROM U.S. HOUSEHOLDS Aileen Heinberg, Angela Hung, Arie Kapteyn, Annamaria Lusardi, Anya ...»
Individuals should not put all of their eggs in one basket, but rather choose welldiversified portfolios and avoid investing in only one asset, particularly if that asset is their employer’s company stock. There is very little knowledge about risk diversification (Lusardi and Mitchell, 2014). Understanding of risk diversification seems also influenced by affect and heuristics. For example, people often rate company stock, the stock of their employer, as a safer investment than a diversified fund (Benartzi and Thaler, 2007). Even when spreading assets among several investments, 401(k) investors often choose naïve diversification, with equity exposure tracking the relative number of equity funds in the menu of available funds (Benartzi and Thaler, 2001).
Tax treatment of retirement savings vehicles:
Retirement assets invested in tax advantaged vehicles such as 401(k)s and IRAs benefit from tax exemptions on contributions, capital gains, or withdrawals, allowing for more rapid potential growth. In fact, people possess limited attention and often do not deliberatively consider and appropriately weight all features of complex decisions. The impact of taxes on decision making therefore does not depend solely on their economic consequences, but also on the salience of these taxes (Congdon, Kling, and Mullainathan 2009).
Employer matches of defined contribution savings plans:
Many employers match (in different proportion, often one-to-one) the contributions employees make to retirement accounts, resulting in a much higher return on retirement savings. Failure to contribute up to the employer’s matching threshold is often the equivalent of leaving money on the table. However, a large portion of individuals do not take advantage of their employer’s full 401(k) matching contributions; evidence suggests this cannot be fully attributed to rational strategies. Choi, Laibson, and Madrian (2005) investigated a special case in which it was difficult to provide a normative explanation for failure to contribute up to the employer’s matching threshold. They examined a group of individuals who could withdraw assets from their 401(k)s at any time without tax penalties and found that, even among this group, half contributed below the match threshold. Notifying employees about the existence of this matching opportunity in the context of three brief written survey questions did not significantly impact the contribution rate.
II.B. Lowering Behavioral Barriers
To ensure that the program materials were delivered with low technical and time burden on the user, we designed them to be accessible, engaging, and relatively brief. We also drew on several well-established principles of psychology and marketing to increase motivation.
Guided by Bandura’s social cognitive theory (1989), we considered the role of self-efficacy in uptake of our educational program. Self-efficacy refers to the subjective belief that one (rightly or wrongly) has the ability to complete a particular task in a way that will lead to a successful outcome. “Learning self-efficacy,” one’s belief that one has the capacity to successfully learn from an educational program, increases motivation to utilize such a program. Since individuals are often intimidated by financial information, we took care to make our program accessible in order to increase self-efficacy and, in turn, motivation to utilize the program.
The five financial concepts were embedded in five simple short stories that describe the concepts verbally and present the benefits of taking action. Each story explains elements of one of the financial principles, consists of dialogue between two people, and is written to convey information in a relatively light-hearted, positive manner. Each story focuses on a few simple take-away points related to the concept and we minimize the use of complex jargon.
We adopted a narrative strategy. In commercial advertising, adult education, and public health, narratives are an established means of creating cognitive involvement and emotional immersion (Bruner, 1987; Green and Brock, 2000; Norris et al., 2005; McDaniel, Waddill, and Shakesby, 1996; Davidhizar and Lonser, 2003) and have been shown to outperform argument-based advertisements in improved comprehension for poor readers (Michielutte et al., 1992).
Bandura’s account of self-efficacy also informed our use of a narrative strategy. While a main focus of the program was on improving financial knowledge, we also aimed to provide that knowledge in a way that could impact financial behavior. According to social cognitive theory, bolstering self-efficacy beliefs (i.e., one’s belief that they can successfully perform a behavior) fosters behavior change. One way self-efficacy beliefs can be strengthened is through observing others successfully perform (or “model”) desired tasks and achieve desired outcomes (Bandura, 1989). We created narratives in which characters did just that: they accomplished desired tasks and achieved desired goals, even in the face of challenges. We included triumph over obstacles in the narratives because observing this type of persistence fosters a more resilient sense of selfefficacy. It has been noted that observers are more strongly influenced by others when the observed and observer are similar. Therefore, our stories incorporate a focus on relatable situations and characters. As the content focus of Five Steps is on basic retirement planning rather than catch-up strategies appropriate for those nearing retirement, we designed these features to be most relevant to young adults (going out, shopping, newlyweds, workplace experiences) and our videos employed actors between the ages of 20 and 40. By having the models appear relaxed and in an everyday setting and explicitly emphasizing lack of stress in taking action in the scripted dialogue, we tried to make the intervention accessible and unintimidating while at the same time imparting and reinforcing new information. It’s worth noting that the use of behavioral modeling itself has been found to decrease anxiety in stressful situations more than purely informational materials alone (Gagliano, 1988).
Our program design also applied other behavioral insights. The literature on present-bias and time-inconsistency documents that people often lack the self-control to take action that will results in future benefits (for review, see Frederick et al., 2002). This can manifest itself as a tendency to procrastinate when it comes to retirement savings or taking up financial education itself, which has largely long-term benefits (Benartzi and Thaler, 2007; Meier and Sprenger, 2009). To counter this, we designed the intervention to emphasize short-term benefits, whether tangible or not (e.g., increased peace of mind), in order to make near-term benefits more salient.
Finally, wherever possible, as related to the individual stories, we employed insights from various aspects of the existing literature. For instance, as research indicates that people perceive free items as especially attractive (Shampanier, Mazar, and Ariely, 2007), when describing 401(k) contributions, employer matches was deliberately framed as “free money.” II.C. Delivery Models: Videos versus Narratives As part of this study, we developed and produced our narrative-based materials using two alternative delivery methods: written narratives and videos.
While written materials are usually considered easy to disseminate, the online video increases the potential for efficient and scalable interventions. Online videos are rising in popularity; a 2010 report by the Pew Research Center (Purcell, 2010) found that 69% of internet users watched or downloaded videos online. The fixed cost of creating videos may be close to the cost of creating written/printed materials. However, with the rise of free Internet video hosting services, the marginal cost of dissemination is rapidly becoming negligible. While many interventions currently rely on written pamphlets and materials, studies of health education have found higher user satisfaction with educational videos as compared to written materials (Jeste et al., 2008;
Armstrong et al., 2011). Most importantly, while written narratives provide some degree of behavioral modeling, the direct visual experience provided by videos may have greater potential to provide the type of observational learning experience needed to have an impact on selfefficacy (Gist, 1989). Additionally, education research indicates that videos have the potential to create fertile opportunity for cognitive engagement (Kozma, 1991). As the popularity of online videos is an important potential new development in adult education, we sought to explore the relative impact of providing Five Steps through video versus written formats.
Videos and narratives were carefully matched on both informational content and stories, with the way concepts were explained closely mirrored in the two formats. Each of the five videos was approximately three minutes long. The narratives are reproduced in Appendix B, while the video links and titles (made available via YouTube.com, a leading free online video hosting service) are listed in Appendix C.
III. Field Experiment Approach The primary goal of the study was to develop and evaluate an innovative, engaging, and low-cost financial education program. Our empirical methodology employs qualitative and quantitative methods to design and test the effectiveness of the financial education program.
III.A. Qualitative Study Design
In order to qualitatively test the intervention materials and get feedback, as well as to gain insight on savings attitudes and behavior, we first conducted two focus groups in Washington, DC, with young workers between the ages of 25 and 40. We were particularly interested in young adults’ response to our intervention, as starting to apply knowledge of the financial concepts earlier in life could produce the greatest benefit, even though younger adults might consider retirement saving less relevant. The first focus group consisted of ten “savers,” people who were already saving for retirement. The second group consisted of eight “non-savers,” who were not currently saving for retirement. This stratification allowed participants with similar experiences to feel more comfortable and engage in a useful discussion. Focus group participants were presented with each of the five stories explaining the financial concepts. They were shown two stories in video format only, two stories in written narrative format only, and one story in both video and written narrative formats. After each story, comments and feedback on the interventions were solicited and the groups concluded with a general discussion of the different presentation formats as well as how the information might motivate any behavioral changes.
While the focus group provided us with a deeper understanding of the types of responses we may receive from a small group of savers and non-savers, it did not give us a great deal of understanding of program effectiveness. Thus, we turn to the main empirical strategy in our paper, a field experiment in which we rigorously test the causal effects of the program.
III.B. Field Experiment Design
To quantitatively test the Five Steps program, we designed a field experiment using the RAND American Life Panel (ALP), fielded from May through November 2010. At the time of the intervention, the ALP consisted of a sample of approximately 3,000 households who are regularly interviewed over the Internet. An advantage relative to most other Internet panels is that the ALP is a probability sample of the US population.1 Data routinely collected via the ALP ALP respondents participating in our experiments have been recruited in one of three ways. Most were recruited from among individuals age 18+ who were respondents to the Monthly Survey (MS) of the University of Michigan’s Survey Research Center (SRC). A subset of respondents (approximately 500) were recruited through a snowball sample; here respondents were given the opportunity to suggest friends or acquaintances who might also want to participate. Respondents without Internet (both in the Michigan sample and the snowball respondents) were provided with so-called WebTVs (http://www.webtv.com/pc/), which allow them to access the Internet using their television and a telephone line. The technology allows respondents who did not have previous Internet access to participate in the panel and furthermore use the WebTVs for browsing the Internet or using email. A new group of respondents (approximately 500) has been recruited after participating in the National Survey Project, created at include a wide array of variables about household and individual demographics. In our sample, about 30% of participants are below the age of 40, 55% are 41–64 years old, and 15% are 65+.
The education attainment is rather high—more than 40% have a college degree—and 59% are female. About 28% have income below $35,000, 40% have income between $35,000 and $75,000, and 32% have income above $75,000.
In May 2010, all members of the ALP (regardless of age) were administered a baseline survey with a series of questions on topics related to our five concepts: (1) compound interest; (2) inflation; (3) risk diversification; (4) tax treatment of retirement savings vehicles; and (5) employer matches of defined contribution savings plans. The questions are reproduced in Appendix A. While the survey questions are multiple choice, the correct answers to the quizzes are all binary (the answer is correct or incorrect).
Starting in August 2010, with the exception of a control group, respondents were exposed to six successive treatments, each presenting material for one topic in either video or written narrative form. Treatments were delivered over three successive waves of the survey, which were spaced two weeks apart. In order to maintain a reasonable length of each survey session, two treatments were given in each wave.
The randomization was implemented as follows:
The order of the five topics was randomized with equal probability for all respondents.
Within each treatment, format was randomized so that respondents would experience either (only) video or (only) written narrative format with equal probability, with one exception.
In one wave of the survey (randomly selected with equal probability), a respondent would experience two treatments for the same topic, one in video and one in written narrative format (in random order).