Introduction - University of Southern California



Two Dimensions of Uncertainty Predict Investment Decisions and ForecastsDaniel J. WaltersGülden ?lkümenCarsten ErnerDavid TannenbaumCraig R. FoxAbstractEstimating the future returns of any stock investment is an exercise fraught with uncertainty. This uncertainty can be attributed to failures in knowledge (i.e., being unfamiliar with an industry or company—epistemic uncertainty) or to random processes (i.e., asset volatility—aleatory uncertainty). In this paper we find attributing stock market uncertainty to failures in knowledge rather than random processes is associated with a variety of poor investing decisions including greater asset concentration, higher portfolio turnover, greater willingness to purchase expensive mutual funds, greater willingness to pay for financial advice, and more overconfident predictions of asset earnings. Perceptions of aleatory uncertainty can be primed to improve financial decisions, as reflected in lower portfolio concentration, but only among the financially unsophisticated, providing an intervention that could improve financial decision making for those who need it most. Further, perceptions of uncertainty affect underlying beliefs about asset characteristics— investors make more extreme earnings estimates for companies perceived to be operating in an environment of epistemic uncertainty and wider ranging confidence interval earnings estimates for companies perceived to be operating in an environment of aleatory uncertainty. Examination of actual earnings reveals differences in uncertainty are perceptual—there is no difference in the extremity or variance of realized earnings by perceived variant of uncertainty. As a result, perception of greater epistemic uncertainty leads to greater overconfidence.IntroductionHow an individual chooses to invest their assets is one of the most important financial decisions that a person can make, and people approach this decision in dramatically different ways. To illustrate, imagine two investors, Jack and Jill. Jack believes that the market rewards talent and expertise: with the right investment strategy and knowledge, he feels confident that he can consistently identify winning and losing assets and outperform the market. As a result, Jack spends considerable amounts of time and money researching and investing in individual assets. By contrast, Jill views the market as profoundly stochastic, and places little confidence in his (or anyone else’s) ability to reliably pick winners and losers. Instead, Jill focuses his energy on maintaining many different and diversified assets that tend to follow aggregate market trends, which she views as reducing noise (and therefore her exposure to risk) in a fundamentally random environment.Jill and Jack’s investment strategies represent two common yet fundamentally different viewpoints of the investment landscape. Jack views the market according to a logic of skill, while Jill views the market according to a logic of chance. In this paper we suggest that such differences arise from a fundamental psychological distinction in how people view the nature of uncertainty. This difference pertains not only to investment decisions, but to all facets of life that entail subjective assessments of uncertainty. In particular, people can view uncertainty as resulting from fundamentally random or chance processes (aleatory uncertainty) or as resulting from the awareness of deficiencies in one's knowledge, information, or skills to correctly assess an event that is, in principle, knowable (epistemic uncertainty). The distinction between epistemic and aleatory conceptions of uncertainty has been noted and debated since the very beginning of modern probability theory (Hacking, 1975), but has received scant empirical attention. We assert that this distinction reflects dual intuitions that investors hold when making judgments and decisions in a market laden with uncertainty. In this paper we examine how perceptions of epistemic and aleatory uncertainty can affect predictions that inform investment decisions (e.g., estimates of company earnings), and most importantly, subsequent investment behaviors and strategies.Recent research has suggested that the epistemic-aleatory distinction is psychologically rich. Across a wide range of age groups and contexts, people appear to intuitively distinguish between uncertainty due to randomness from uncertainty due to lack of knowledge. For instance, 4-6 year old children tend to behave differently when facing chance events yet to occur (in which aleatory uncertainty is presumably salient) versus chance events that have already been resolved but not yet revealed to them (in which epistemic uncertainty is presumably salient; Robinson et al. 2006). Meanwhile, brain imaging studies (Volz et al. 2005, 2004) have found distinct activation patterns when participants learn about events whose outcomes were determined in a rule-based (presumably epistemic-salient) manner compared to a stochastic (presumably aleatory-salient) manner. Furthermore, people rely on distinct linguistic expressions to communicate their degree of epistemic and aleatory uncertainty (?lkümen et al. 2015). Table XX outlines two characteristics of epistemic and aleatory uncertainty that are particularly relevant to investment behavior: representation and attribution of uncertainty.Table 1Investment PredictionsCharacteristicEpistemic (Knowable)UncertaintyAleatory (Random)UncertaintyEpistemic (Knowable)UncertaintyAleatory (Random)UncertaintyRepresentation of uncertain outcomesSingle caseClass of possible outcomesMore extreme point estimates Narrower confidence intervalsMore regressive point estimatesWider confidence intervalsAttribution of UncertaintyInadequate knowledge/skillStochastic behaviorGreater value placed on financial expertiseMore frequent tradingMore concentrated portfoliosMore diversified portfoliosRepresentation of Uncertainty. The focus of attention under pure epistemic uncertainty is a singular case that may occur (or a single statement that may be true), whereas the focus of attention under pure aleatory uncertainty is on classes of possible outcomes. Investors often make decisions based on their judgments of market fundamentals, such as estimates of companies’ future earnings. We expect that such point estimates, as well as the confidence intervals given around those point estimates should be influenced by investors’ perception of stock market uncertainty. First, an investor who perceives a company’s revenues and costs to depend primarily on epistemic factors (i.e. are in principle knowable in advance) and has good reason to believe those factors will change in the coming year (which also requires some knowledge of the company) will represent uncertain future earnings as a singular scenario, and therefore will predict more extreme changes in earnings from one year to the next. In contrast, an investor who perceives a companies revenues and costs to depend primarily in aleatory factors (i.e., mostly random), will consider multiple alternative outcomes (e.g., company earnings may increase, decrease or stay the same), and therefore his/her judgments should be more regressive than judgments based on singular representations. Second, an investor may predict a broader range of possible earnings changes (i.e. a wider confidence interval) to the extent she perceives revenues and costs to depend on primarily aleatory factors (i.e. be susceptible to chance). We test these predictions among investors in Study 1 and Study 2.Attribution of Uncertainty. It is generally accepted that while investors seek to maximize returns, they simultaneously try to minimize the uncertainty associated with those returns. We propose that the uncertainty around any given financial outcome can be perceived as primarily epistemic or as primarily aleatory. Importantly, we suggest that whether an investor perceives uncertainty around stock market outcomes to be primarily epistemic versus aleatory suggests different approaches they must take to reduce this uncertainty; and thus, predicts their overall investment philosophy. Since epistemic uncertainty is generally attributed to failures of knowledge, we expect that strategies to improve knowledge — such as seeking information or consulting experts — will be used to reduce epistemic uncertainty. In contrast, stock market uncertainty attributed to random processes under aleatory uncertainty should lead to strategies that minimize risk exposure. Diversification has been widely disseminated as a strategy to minimize random uncertainty (i.e., volatility) relative to returns in efficient portfolio theory (Jensen, 1969; Markowitz, 1991). We predict that investors perceiving more aleatory and less epistemic uncertainty in the stock market will be more likely to diversify and hold less concentrated portfolios. We test these predictions through experimental manipulation in Study 2 and real investment behavior in Study 3. Furthermore, attributing uncertainty to epistemic factors suggests that outcomes become more predictable with additional information. Thus, changes in information allow investors to draw new conclusions when epistemic uncertainty is high. We expect this will result in increased trading frequency since investors are constantly learning new information. We test this prediction in Study 3.Study 1: Perceptions of Uncertainty Predict Earnings EstimatesIn this study we examine how the perception of uncertainty impacts a critical component of stock investing: forecasts of company earnings. To make investment decisions, investors often study information about a company’s earnings in a previous period and try to predict the extent to which earnings will increase or decrease in the future. Following from table XX, if an investor perceives the environment in which a company is operating to be reasonably predictable (i.e., involving primarily epistemic uncertainty), this investor may tend to make predictions that stray further from prior year earnings, resulting in more extremity in their predictions. However, we expect this relationship between perceived epistemic uncertainty and extremity of earnings forecasts to hold more strongly for investors who perceive themselves to be familiar with a company—investors who are not familiar with a company should give earnings estimates that regress toward previous period’s earnings. Corroborating our predictions, past research with participants who were knowledgeable in NCAA basketball found that participants’ predictions were more regressive when they perceived the uncertainty governing the outcome of basketball games to be relatively more aleatory in nature, and more extreme in their predictions when they perceived the uncertainty to be relatively more aleatory in nature (Tannenbaum, Fox, and ?lkümen, 2015). In addition, we predict that investors will provide wider confidence intervals around their estimates, when they perceive greater aleatory uncertainty. The intuition for this prediction follows from table XX: when a person perceives greater randomness or variance in earnings she/he will be less confident in predicting next year’s earnings and will provide wider confidence intervals to capture a wider range of possible outcomes.MethodsWe invited participants (n=697) to complete our study through Amazon Mechanical Turk in exchange for $2. Potential participants first answered a question that asked them to characterize their knowledge about investing in the stock market (1=“I know nothing about investing in the stock market”, 7=“I consider myself an expert on stock market investing”). Participants then completed an attention screen (Appendix XX). Of the 2,402 participants who applied to take the survey, only those who both rated their knowledge as 4 or above, and passed the screen were allowed to participate, yielding 697 participants.We randomly selected 250 companies from the S&P 500 index. We assigned each participant to read 10 company profiles randomly selected from these 250 companies. Each company profile was downloaded directly from Reuters with no changes. After participants read each profile they were asked to complete the shorter, 6-item EARS to rate the uncertainty associated with the “approximate future yearly earnings amount of the company.” After completing the EARS for all 10 companies each participant rated their prior knowledge of the companies on a seven-point scale from very poor to very high. In total, we recorded 6,970 ratings for an average of 28 ratings per company. Next, we provided the participants once again with a one-paragraph description of the company as well as last year’s earnings, and asked the participants to give estimates of the next year’s earnings. Participants first provided a point estimate for the company’s earnings for next year, then they gave a 90% confidence interval, such that they believe there was a 90% chance that actual earnings next year would fall within the range. Participants repeated this task for each of the 10 companies.Finally, participants completed an 11-item financial literacy test (CITE) and provided demographic information. Participants were also asked to complete an attention check , where they described the business in which one of the 10 companies operates. Participants were then debriefed.ResultsWe first examined the time participants spent reading the company profiles. Given the complexity of information presented in the profiles, we excluded observations where the participants read and evaluated the profile on EARS at a rate faster than 300 words per minute. This resulted in us dropping 2,729 of the 6,970 observations, leaving us with 4,241 observations. Epistemic and aleatory ratings were negatively correlated, r = -.28, p < .001. Our theorization predicts that the items measuring perceptions of epistemic and aleatory uncertainty will load on two distinct factors. To test this prediction, we ran a confirmatory factor analysis with Varimax rotation. The results were consistent with our prediction. The three epistemic items load highly on factor 1, and the three aleatory items load highly on factor 2. The rotated component matrix is shown in table XX.Table XX: Factors Loads for the Reduced EARSWe calculated the earnings extremity as the absolute value of the predicted percentage change from previous years earnings. We also calculated the confidence interval around earnings as the upper prediction minus the lower prediction divided by the prior-year earnings. We further excluded 15 observations where the confidence interval was wider than five times the past year earnings, leaving us with 4,226 observations.We first analyzed confidence intervals with respect to perceptions of uncertainty. We examined the earnings confidence interval as the dependent variable with aleatory rating and epistemic rating as the independent variables while clustering the regression by participant. Confirming our prediction, we found that participants who perceived a company as involving higher aleatory uncertainty provided wider confidence intervals around their estimates of the company’s future earnings, b = .03, CI = [.01, .04], p < .001. In contrast, epistemic ratings were not related to the width of confidence intervals, b = .01, CI = [-.01, .02], p > .41. We next tested our hypotheses regarding earnings extremity. We examined the earnings extremity as the dependent variable and (1) epistemic rating (2) the interaction between epistemic rating and knowledge, and (3) aleatory rating as the independent variables while clustering the regression by participant. Confirming our prediction, we found that participants who perceived the company as involving more epistemic uncertainty and perceived themselves as knowledgeable made more extreme forecasts about the company’s future earnings, b = .002, CI = [.000, .003], p = .02. Participants who perceived a company as involving more aleatory uncertainty also made more extreme predictions about its future earnings, b = .014, CI = [.005, .023], p < .01. We next examined whether the differences in expected earnings growth were justified, i.e., did they predict differences in actual earnings growth. One year after these data were collected we recorded the real earnings data to calculate the real absolute growth for the pertinent year. We then examined the real absolute growth as the dependent variable and epistemic and aleatory rating as the independent variables and clustered by company. In contrast to the estimates, we find real growth does not vary by epistemic, p > .7 or aleatory rating p > .8, suggesting the difference is largely perceptual. Finally, we examined the accuracy of the estimates. If actual growth does not vary across variants of uncertainty, then the observed more extreme estimates under epistemic versus aleatory perceptions of uncertainty should reflect less accuracy. To test this prediction, we examined accuracy as the absolute difference between the real earnings growth and the predicted earnings growth. We regressed accuracy as the dependent variable and epistemic and aleatory rating as the independent variables while clustering by participant. Confirming our prediction, participants where less accurate in their estimates of companies that they rated high in epistemic uncertainty, b = 136.7, CI = [41.26; 232.13], p < .01. Aleatory ratings, in contrast, were not related to accuracy, b = -25.86, CI = [-117.24; 65.53], p > .5.DiscussionAs expected, investors forecasted more extreme future earnings, if they perceived themselves as knowledgeable about a company, and if they perceived the company to operate under primarily epistemic uncertainty. On the other hand, perceived aleatory uncertainty was associated with wider confidence intervals around earnings estimates. Moreover, perceived variants of uncertainty appear to influence predicted, but not actual earnings growth, leading to lower accuracy under epistemic uncertainty. Study 2: Priming Perceptions of UncertaintyIn Study 1 we showed that perception of uncertainty influenced judgments of company earnings. In Study 2 we extend this finding in an experimental design to demonstrate causality and to show perceptions of uncertainty influence investment behaviors. We sought to experimentally manipulate perceptions of stock market uncertainty by asking participants to represent forecasts of company operations as a single instance (to prime epistemic uncertainty) or as a class of possible outcomes (to prime aleatory uncertainty). Prompting individuals to engage in such “singular” versus “distributional” reasoning has been found to reliably influence perceptions of epistemic and aleatory uncertainty (Tannenbaum, Fox & ?lkümen, 2015). We expected greater portfolio concentration and greater departures in forecasts from the previous year’s earnings (i.e., forecast extremity) when primed with singular reasoning, and greater portfolio diversification and wider confidence intervals around forecasted earnings when primed with distributional reasoning.MethodParticipants were recruited from a large, diverse qualtrics panel and were each compensated $8 for their participation. The qualtrics panel is comprised of over 525,413 members ranging in age from 18-50 with a broad range of professional experience. Before completing the questionnaire, participants were screened for adequate financial experience. To be eligible for the study, participants had to possess more than $1,000 in stock market investments, be between the ages of 18 to 65, report making their own investment decisions, rated their knowledge of the stock market as a three or more on a five point scale (either, I consider myself an expert on stock market investing-5, I know what investing in the stock market is and have a moderate level of knowledge in the subject-4, or I know what investing in the stock market is but do not consider myself very knowledgeable in the subject-3), and were able to correctly answer two basic financial literacy questions. Of the 1,551 individuals who responded to the initial screening questions, 201 qualified for participation. All participants read the 2014 Q3 earnings release for two companies, and Ford (Appendix XX). These two companies were selected because of their familiarity to most subjects, and because their press releases were relatively non-technical and easy to read. Participants were then randomly assigned to one of two conditions. To prompt singular reasoning, participants were asked to construct a single scenario (12-15 sentences long) for each company about how much revenue would be generated in Q3 2015. To prompt distributional reasoning, participants were asked to construct a class of three possible scenarios (each 4-5 sentences long) for each company about how much revenue would be generated in Q3 2015. In order to ensure thoughtful responding, we removed 29 participants who failed to adequately engage in the writing task.After completing the writing task, participants forecasted each company’s revenue for the following year, and were also asked to place 90% confidence intervals around each estimate. As a manipulation check, participants also assessed the underlying uncertainty associated with the forecasting task on a shortened 6-item version of the EARS. Next, participants were asked to imagine investing $100 dollars between the two companies. As an incentive to respond truthfully, participants were notified that some subjects would be randomly selected to receive the actual payout of their investment decisions after six months. Finally, participants completed a basic financial literacy test (Appendix XX-Cite) and were debriefed.ResultsAs expected, participants primed with distributional reasoning reported lower epistemic ratings than participants primed with singular reasoning (Ms = 4.34 vs 4.65; t(170) = 2.21, p < .05, d = 0.34). Participants in the distributional prime condition also reported greater aleatory market uncertainty than participants in the singular prime condition, although this difference failed to reach statistical significance (Ms = 4.79 vs 4.62; t(170) = 1.05, p = .30, d = 0.16).Figure XX: Ratings of epistemic and aleatory uncertainty for next year revenues by condition. Figure XX: Confidence interval width of next year revenues by prime across both companies. We next examined confidence interval widths. Confidence widths were standardized for each participant by taking the difference between the upper and lower bounds of their estimates and dividing this number by their estimate on expected earnings. This was done separately for each company, and for each subject we took the average of their two interval widths. Confirming our prediction, participants prompted to think distributionally provided wider confidence intervals than participants prompted to think singularly (Ms = 0.27 vs 0.20; t(170) = 2.60, p = .01, d = 0.40). However, and contrary to our prediction, we find no reliable difference between the two priming conditions in the extremity of participants' forecasts (i.e., absolute divergence from the previous year’s earnings); if anything, participants in the distributional prime condition made slightly more extreme forecasts than participants in the singular prime condition (Ms = 0.23 vs 0.13; t(170) = 1.62, p = .11, d = 0.25).We next examined portfolio concentration, which was calculated by taking the absolute difference in percent invested between the two stocks (0 = equally concentrated between the two companies, 100 = complete concentration in one of the two companies). Confirming our prediction, participants primed with epistemic market uncertainty were more likely to selectively concentrate their investments compared to participants primed with aleatory market uncertainty (Ms = 60.5% vs 51.2%; t(170) = 2.15, p = .03, d = 0.33).For exploratory purposes, we also examined whether our findings interacted with participants' degree of financial literacy, which was operationalized as the percent correct across our 11 financial literacy questions. Using a fractional response model (Papke & Wooldridge, 1996), we regressed portfolio concentration onto experimental prime (0 = distributional, 1 = singular), financial literacy scores, and the interaction term between the two predictors. As illustrated in Figure XX, our experimental prime was most pronounced for participants low in financial literacy (b = –4.48, SE = 1.24, p < .001 for the interaction term). Based on our regression coefficients, participants relatively low in financial literacy (1 SD below the mean) would be expected to show a 24 percentage-point decrease in portfolio concentration when primed to think distributionally rather than singularly (predicted probabilities were 66.7 vs 42.5; b = 0.24, SE = 0.5, p < .001). In contrast, participants relatively high in financial literacy (1 SD above the mean) would not be expected to show a reliable difference in their portfolio behavior across the two priming conditions (predicted probabilities were 60.0 vs 54.4l; b = –.06, SE = 0.06, p = .36). Financial literacy scores did not interact with the experimental prime on any of the other dependent variables.Figure XX: The prime only influenced the asset allocation of those low in financial literacy. Asset concentration is shown on the vertical axis while financial literacy is shown on the horizontal axis. Financial literacy scores ranged from 27.7% to 82%. DiscussionStudy 2 demonstrates that perceptions of market uncertainty can be reliably influenced by prompting individuals to engage in either singular or distributional reasoning. In particular, participants were less likely to diversify their portfolios when prompted to think about singular cases, which reliably increased perceptions of epistemic market uncertainty. We also find that our experimental prime had a larger influence on those with the lowest levels of financial literacy — presumably those who were least informed about how to appropriately allocate their investments. Thus, shifting beliefs about epistemic and aleatory market uncertainty may prove to be an useful intervention technique for improving financial decision making among those who need it the most.Study 3: Individual InvestorsThe findings in the first two studies have been in the laboratory. In Study 3 we extend our research to real world investment decisions. A sample of investors was recruited and perceptions of stock market uncertainty were assessed. We predicted that viewing the stock market as entailing primarily epistemic uncertainty would be associated with lower asset diversification, higher portfolio turnover, greater willingness to purchase expensive mutual funds, and a greater willingness to pay for financial advice. MethodWe recruited experienced investors (n = 164) from the same subject pool reported in Study 2. Prospective participants (N = 1,146) were screened for basic financial experience, yielding 164 qualified investors who were each paid $10 for their participation. Participants completed all of the following tasks in random order.EARS rating. Participants evaluated stock market uncertainty using a 10-item scale that we have developed and validated elsewhere (Fox et al., 2015). The scale prompted participants to rate their agreement with a set of statements that measured both feelings of epistemic uncertainty (e.g., “The approximate return of an individual stock over 1 day is something that becomes more predictable with additional knowledge or skills”) and aleatory uncertainty (e.g., “The approximate return of an individual stock over 1 day is something that has an element of randomness”). In random order, participants rated (1) the approximate return of an individual stock over the course of one day, (2) the approximate return of an individual stock over the course of 10 years, (3) the approximate return of a diversified mutual fund over the course of one day, and (4) the approximate return of a diversified mutual fund over the course of 10 years.Real Investment Behaviors. Participants reported the amount of money they invest in each of the following categories: mutual funds, index funds, individual stocks, fixed income, cash, real estate, commodities, other (open response). We then operationalized diversification as the percentage of assets in index funds and mutual fund (i,e, diversified assets), compared to individual stocks, fixed income, cash and real estate (i,e, non-diversified assets). Participants then reported the number of individual stock transactions, the number of mutual fund, and the number of index fund transactions that they made over the past year. They also reported how frequently they checked their investments, how often they made changes to their investments, and the average holding duration of both their stocks and mutual funds. Finally, participants reported whether they had a financial advisor and, if so, the annual fee they paid to their financial advisor.Hypothethical Investment Decisions. Participants were asked to imagine having US $100,000 that could be invested in individual stocks, mutual funds, or index funds. Participants reported how they would divide the $100,000 among these three assets. Participants again reported how often they would both look over (access) and make changes to their investments on a 7-point scale (scale endpoints are listed in table XX). Participants also reported the likelihood they would use the money to invest in a foreign mutual fund, an expensive mutual fund (one that charges higher fees), an unfamiliar mutual fund, or a familiar mutual fund (choices made on a 7 point scale). Finally, participants reported how likely they would be to hire a financial advisor (on a 7-point scale), the amount of money they would allow the financial advisor to actively manage (free response in percentage assets), and the fee they would be willing to pay the financial advisor (choices in 1% increments from 0% to 7%). ResultsEARS rating. As expected, participants carved up the financial landscape into two distinct forms of uncertainty. An iterated principal factors analysis revealed two factors that correspond to epistemic and aleatory uncertainty (see Appendix for details). The two factor structures were modestly correlated, r = .22, suggesting that the two dimensions of uncertainty were largely independent from one another. Furthermore, alpha for each of the four domains returned adequate reliability (alphas ranged from XX to XX). For purposes of simplicity, we standardize each rating and then used the average of those ratings for both epistemic and aleatory uncertainty. Real Investment Behavior. Perceptions of epistemic and aleatory uncertainty were related to both scenario-based and real investment behaviors. Table XX lists each of our measured variables and how they relate to epistemic uncertainty and aleatory uncertainty. Investors who perceived the stock market as more epistemic engaged in a number of behaviors that suggested they viewed the market through a logic of skill: they held stocks and mutual funds for a shorter period (figure XX), were more likely to purchase a high priced mutual fund, were less likely to be diversified, were more likely to seek expertise and hire a financial advisor, reported a higher percentage of their assets were actively managed by a financial advisor, and were more willing to pay a higher fee to a financial advisor (as a percentage of assets under management). By contrast, perceptions of greater aleatory uncertainty were related to a lower likelihood of having a financial advisor, shorter stock holding period, more frequent changes to the portfolio, higher likelihood of purchasing a foreign or familiar mutual fund and more frequent checking of a hypothetical portfolio. It is perhaps not surprising that perceptions of aleatory market uncertainty were associated with lower willingness to hire a financial advisor and higher likelihood of purchasing a mutual fund (since such uncertainty is minimized through diversification). However, we did not predict that perceptions of aleatory uncertainty would be associated with more frequent trading. One possibility is that some investors consider the stock market to be both random and predictable. These investors might see some components of a stock to be predictable, for example the underlying company performance might be predictable, whereas other aspects are random, for example oil price fluctuations. As a result, this investor might see greater randomness as a trading opportunity as prices go up and down around a target price. To investigate this possibility we examined the interaction effects of perception of epistemic and aleatory uncertainty on changes in the portfolio in a regression. We find a significant interaction, b= –.15, CI= [-.25, -.04], p < .01. Figure XX shows the interplay of aleatory and epistemic uncertainty when making changes to a portfolio in a contour plot: As predicted, when uncertainty is perceived to be a combination of high epistemic and high aleatory (upper right quadrant), then investors make daily changes to their portfolios. When investments are perceived as low in epistemic uncertainty (bottom two quadrants) then investors make monthly changes. However, when investments are perceived to be low in aleatory uncertainty and high in epistemic uncertainty people make changes only yearly, presumably because the outcome is perceived to be so predictable that no unexpected events occur to necessitate a change.Table XX: Investment Behaviors and Perceived Uncertainty Figure XX: Real portfolio changes and perceived uncertainty. The vertical axis shows perceived epistemic uncertainty while the horizontal axis shows perceived aleatory uncertainty. People make the most frequent changes to their investments when both epistemic and aleatory uncertainty are perceived to be high and the least frequent changes when epistemic is perceived to be high and aleatory is perceived to be low.Figure XX: Average holding period of stocks by participants in either the upper quartile of epistemic rating or lower quartile of epistemic rating. Participants that rated stock uncertainty as more epistemic held stocks for a much shorter period on average compared to those that rated stock uncertainty as less epistemic.DiscussionThese findings show that perceptions of uncertainty are related to investment behaviors. Confirming our predictions, investors who perceive the stock market to be more epistemic are less likely to diversify, more likely to trade, more likely to hire a financial advisor, and more likely to pay high fees for a mutual fund. These result held after controlling for perceptions of aleatory uncertainty, demonstrating that these are two distinct constructs. Further, higher perceptions of aleatory uncertainty are associated with being less likely to hire a financial advisor, and being more likely to purchase a diversified asset, such as a mutual fund. We also find that making changes to an investment portfolio depends on the interplay of both variants of uncertainty. People appear to vary the frequency of making investment changes on the basis of two factors: First, the degree to which a person believed an action will predictably result in a greater investment return and second the frequency with which a person believes unexpected opportunities occur. Perceptions of epistemic uncertainty are associated with the belief that an outcome is predictable, so greater perception of epistemic uncertainty should drive the belief that investment actions can predictably drive higher returns. Perceptions of aleatory uncertainty are associated with the belief that random and unexpected events can occur, so greater perceptions of aleatory uncertainty should drive the belief that unexpected opportunities occur.We also find that greater perceptions of epistemic uncertainty were associated with lower financial literacy (b - .25, p <.001, Table XX). To further examine this result we compared perceptions of epistemic uncertainty in this study to those in a study of 37 practicing financial advisors who also evaluated stock market investing using the four EARS scales. In a separate sample t-test we find that financial advisors perceived lower epistemic uncertainty in stock investing (M = 2.90), compared to individual investors (M = 3.99; t = 4.49, p < .0001; Figure 3). In contrast, perceptions of aleatory uncertainty did not reliably differ between financial advisors (M = 4.47) and individual investors (M = 4.70; t = 1.09, p = .28).ReferencesBarber, B., Lehavy, R., McNichols, M., & Trueman, B. (2001). Can investors profit from the prophets? Security analyst recommendations and stock returns.The Journal of Finance,?56(2), 531-563.Chang, H. L., Chen, Y. S., Su, C. W., & Chang, Y. W. (2008). The relationship between stock price and EPS: Evidence based on Taiwan panel data.Economics Bulletin,?3(30), 1-12.Jegadeesh, N., & Kim, W. (2006). Value of analyst recommendations: International evidence.?Journal of Financial Markets,?9(3), 274-309.Jensen, M. C. (1969). Risk, the pricing of capital assets, and the evaluation of investment portfolios.?Journal of business, 167-247.Moshirian, F., Ng, D., & Wu, E. (2009). The value of stock analysts' recommendations: Evidence from emerging markets.?International Review of Financial Analysis,?18(1), 74-83.Patell, J. M. (1976). Corporate forecasts of earnings per share and stock price behavior: Empirical test.?Journal of accounting research, 246-276.Cutts, M. (2013). Oxford guide to plain English. Oxford university press. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download