Do Animated Line Graphs Increase Risk Inferences?

Graphical Display of Time-Varying Data

A time-varying data series conveys information that changes because of temporal updates in data values. As data values change from one point in a series to another, the transitions between individual data points also provide insight about the nature of the series (e.g., its trend, its pattern, the degree of noise surrounding such transitions).

Previous literature has suggested that graphical display forms a particularly effective way to represent time-varying data (e.g., Henry 1995; Wallgren et al. 1996). For example, when a stock’s price history is visually presented in a line graph, summary information on temporal variations such as trend (e.g., increasing or decreasing), pattern (e.g., cyclical), and noise around the trend (e.g., volatility) are easily recognized by the user. Thus, graphical display offers a compact and easily discernible representation that helps us derive insights from the data (Hutchinson, Alba, and Eisenstein 2010).

Because the human visual system has highly developed skills of perceptual sense making (Kosslyn 1994), the effectiveness of graphical data display is often self-evident. In that regard, research in psychology has largely focused on basic phenomena related to perception and comprehension of graphs, such as pattern recognition, mapping, and interpretation (e.g., Carpenter and Shah 1998; Zacks and Tversky 1999). Further, decision and marketing researchers have focused on substantive impacts of graphical display on outcomes, such as risk assessment, stock investment, and budget allocation (e.g., Chua, Yates, and Shah 2006; Hutchinson, Alba, and Eisenstein 2010; Raghubir and Das 2010).

Although previous studies have examined various aspects of data visualization, the primary focus has been on static forms of graphical display. Notably, a few initial practitioner-driven investigations on animated data visualization have been conducted in computer science and information technology (e.g., Heer and Robertson 2007; Robertson et al. 2008). However, this line of research has left significant gaps. First, these studies primarily explore the effectiveness of specific visualization tools available at that time (e.g., DynaVis, Trendalyzer), but are largely silent on potential underlying theory. Second, they focus on basic graph comprehension outcomes (e.g., recall or recognition of data values or patterns), rather than process and consequences such as judgments and decision making. Figure 1 depicts some of the findings and gaps in the data visualization literature, and how the current research fills the gaps in the literature.

OPEN IN VIEWER

Figure 1.

Relevant Findings and Gaps in the Data Visualization Literature.

These gaps are noteworthy because emerging literature has shown unique consequences of dynamic display (e.g., Goldstein et al. 2014; Kim and Lakshmanan 2015; Roggeveen et al. 2015). For example, dynamic (vs. static) display modes (e.g., video presentation of product images, animated online ads) enhance consumers’ involvement with the product (Roggeveen et al. 2015), shape impressions of ad design (Kim and Lakshmanan 2015), or, in some contexts, are seen as distracting and annoying (Goldstein et al. 2014). However, we are not aware of any prior studies that have systematically examined how animated graphical data display affects comprehension, judgments, and decision making. For instance, does animated data display simply grab more attention or distract viewers? Does it have an impact on basic graph comprehension? Does it change viewers’ affective state? Does it have a systematic and/or unique effect on downstream judgments and decisions? The current research sheds light on these questions by integrating insights from the salience literature and cognitive inference making to lay out a process outlining when and why animated display affects data-based judgments and decision making.

Animated Display and Transition Salience

Salience as a stimulus characteristic refers to the availability of specific information (Lurie and Mason 2007; Taylor and Fiske 1978). The salience literature in different streams has identified various factors that could become salient, such as gender, race, and identity (Forehand, Deshpandé, and Reed 2002; Taylor and Fiske 1978). In the visual perception domain, the intensity of sensory features, such as brightness of a light or loudness of a tone, affects stimulus salience (Tversky 1977). With respect to data visualization, prior work explores how the use of visual elements such as color, size, and line thickness enhances the salience of spatial dimensions in geographical data (Hegarty, Canham, and Fabrikant 2010; Ozimec, Natter, and Reutterer 2010). Extending prior work, we identify animation as a salience-inducing mode of displaying time-varying data.

In its basic form, animation depicts a series of scenes that differ from each other temporally. For instance, a flip-book represents a basic form of animation that shows how a series of static images combine sequentially to form a dynamic visual: successive transitions between each page create a visual impression of moving images. As such, animation is a visual medium that dynamically depicts how a narrative evolves over time. We suggest that animated display of data values accomplishes something similar.

Imagine a line graph depicting how a stock’s price has changed over the last month. In the animated version of this graph, stock price points are presented sequentially—with physical movement of the line—as the price trajectory unfolds. In this case, viewers visually trace transitions by seeing the prices physically appear by period. This unique aspect of animated display is especially well-suited to enhance the salience of the temporal dimension of data change. To elaborate, we propose that animated display of a data string draws viewers’ attention in real time to the moment-to-moment unfolding of the data string. In doing so, it is uniquely able to enhance the salience of transitions between data points. Other, more static, ways of drawing attention to the graph (e.g., using visual elements such as color or arrows) are less likely to highlight the temporal variation of the data, particularly since the viewer sees the entire data string at once. Just as a three-dimensional (as opposed to strictly two-dimensional) representation of a scene uniquely enables identification of depth, animated display uniquely enables the detection of variability embedded in the stream of changing data values by making transitions salient.

Utilizing Temporal Transitions for Making Inferences About Risk

Prior literature proposes that heightening the salience of a particular stimulus aspect directs viewers’ attention to it, and as a result, it exerts a prominent impact on subsequent judgments (Taylor and Fiske 1978). In our case, when animation unfolds the trajectory by period, transitions become more traceable and thus draw visual attention. Notably, the drawing of focused attention may not alone fully explain the downstream consequences of salience. Once a particular portion of a stimulus receives focused attention (say, due to its salience), the information contained in that portion is more likely to be disproportionately utilized in subsequent judgments (Taylor and Thompson 1982). For example, Raghubir and Das (2010) find that when stock prices are presented graphically, local maxima and minima appear more salient compared with the prices surrounding them, and this relative salience makes them more likely to be used to infer risk. Thus, the literature alludes to a sequential process in which salience shapes data-based judgments. Although stimulus salience occurs earlier in the overall process because of the focal features drawing differential attention, there is reason to expect an inferential process to follow, such that these visually salient features of the data are more heavily weighted in drawing meaning from the data.

Prior work has, to the best of our knowledge, not explicated the process role of utilizing salient features to form inferences. Indeed, recent research documents evidence that low-level visual attention metrics alone may not fully account for the effect of salience on subsequent judgments. For example, Duclos (2015) finds that graphically displaying a series of stock prices leads consumers to pay more attention to the recent price fluctuations to forecast a future price. However, the gaze duration on the recent price fluctuation does not mediate financial judgments, because such judgments involve “a variety of mental processes” (see Duclos 2015, p. 323). Similarly, McArthur and Ginsberg (1981) find that an individual whose visual appearance is more salient than others in a group draws longer fixations and is perceived as a causal agent in the social interaction. Yet, increased duration of attention to the salient individual does not directly mediate downstream causal attributions. To bridge the gap between salience-induced attention and downstream data-based inferences, we aim to theoretically disambiguate the utilization construct in the process.

We propose that enhanced salience of transitions increases the extent to which transitions are used as a diagnostic cue. Individuals often make judgments based on the given data instead of seeking all the information relevant to a judgment task (e.g., technical analysis in stock trading). In such cases, their judgments vary depending on which data features (e.g., overall trend, beginning/end, noise around a data string) they attend to and use to make judgments (Raghubir and Das 1999). In our case, when temporal transitions become more salient—because of animated unfolding (as opposed to all-at-once static display)—individuals are more likely to rely on these transitions to make risk judgments. Thus, we operationalize the utilization of transitions as the extent to which individuals base their risk judgments on changes in data values from one point in time to the next.

We further argue that greater utilization of transitions inflates downstream risk judgments. In financial investment, risk indicates the probability of losing capital (Duxbury and Summers 2018). Moreover, prior literature has documented that investors believe stock price volatility (i.e., the degree of change in stock prices over time) to be associated with risk (Raghubir and Das 2010). Volatility conceptually refers to a tendency to fluctuate from a stable outcome (Botner, Mishra, and Mishra 2020). Although volatility can signal both gain and loss in capital, the irregular nature of stock price sequences (i.e., random walks) makes it more strongly associated with downside risk (Pincus and Kalman 2004). Behavioral finance literature finds that when random sequences of stock prices are graphically presented, trajectories with greater price volatility correlate with greater risk judgments (Duxbury and Summers 2018). Work in consumer psychology also documents that retail investors infer greater risk from stocks with larger differences between the local maxima and minima—a proxy for the noise around a data trajectory (Raghubir and Das 2010).

Combining the preceding discussion on animated display, salience, and the utilization of transitions in judgment making, we formally hypothesize:

We examine these primary hypotheses regarding the animated display effect and underlying process in Studies 1 and 2 (see Figure 2 for our theoretical framework with associated studies).

OPEN IN VIEWER

Figure 2.

Theoretical framework with associated studies.

Moderating Transition Salience

Our explication of the underlying process mechanism begins with the salience-inducing effect of animated display. For transitions to be disproportionately used as a meaningful inferential cue, it is necessary for animated display to make successive transitions visually stand out. Thus, if animated display does not enhance transition salience earlier in the perceptual process, the core effect should attenuate. We test this expectation by directly manipulating transition salience.

In graph comprehension, the type of representation (e.g., line vs. bar graphs) determines the extent to which specific insights can be discerned from a graph. For example, transitions in time-series information are easier to discern in line graphs, whereas the absolute values of discrete data points are easier to perceive in bar graphs (Hutchinson, Alba, and Eisenstein 2010; Zacks and Tversky 1999). This is because line graphs visually represent the connection between data points, whereas bar graphs visually separate each data point (Duclos 2015; Zacks and Tversky 1999).

When they are animated, line graphs—because they better represent continuity along the trajectory—should make transitions more salient, leading to greater utilization in inferring higher or lower levels of risk. Thus, the core animated display effect and its underlying process should manifest with line graphs. In contrast, though animated display physically unfolds each data point by period, in bar graphs each bar is not visibly and directly linked to the next. Each data point is represented as an independent bar visually rooted in the x-axis. Thus, because of the relative lack of continuity, successive data points in animated bar graphs are less likely to lead viewers to trace transitions, compared with line graphs. Consequent downstream process effects should therefore also attenuate. Stated formally,

We test H₂ by directly manipulating transition salience via the type of graphical representation in Study 3.

Contextualizing the Role of Transitions

Our theorizing also implies that transitions should inflate risk judgments when they serve as signals of the probability of loss. Thus, if the data string contains features that frame the context as one where the prospect of capital loss is low, the risk-inflating role of transitions should attenuate. We explore this moderation argument by recruiting global trends as such a “contextualizing” data feature.

A string of time-varying data may have multiple features, such as trend and the noise around the trend. While transitions happen moment to moment along the time-varying trajectory, global trends are represented as a holistic pattern (i.e., upward/downward patterns) across the entire trajectory. Such global trends are perceptually salient in and of themselves and are easily recognizable, as people’s visual experience typically comes from a high-level recognition of the overall stimulus (principle of global precedence; Navon 1977). For these reasons, although animated display primarily increases transition salience, trends should remain salient irrespective of display mode (animated or static).

As a generally salient data feature, global trends should come into play in risk inferences by framing the context of financial judgments. When a data string has a clear downward or upward trend, it provides a reference point (Raghubir and Das 1999). That is, the starting data point can serve as a reference point with which the current stock price is compared. Thus, when trend is downward (upward), the current price is lower (higher) than the reference point, leading investors to anticipate losses (gains) and so infer greater (lesser) risk.

Given that global trends asymmetrically set investors’ expectations of outcomes, we expect these trends to contextualize the role of transitions in risk judgments. With a downward trend, losses are more likely, and so investors should remain sensitive to risk-diagnostic cues such as transitions in the data string (loss aversion; Kahneman and Tversky 1979). Therefore, for the downward trending stock, we expect animated display (vs. static display) to heighten risk judgments, as transitions are still a risk-relevant data feature. However, an upward trend establishes the context of a rising market, and in this environment, the prospect of losing capital is significantly reduced. Investors should therefore be generally less sensitive to risk-diagnostic cues such as transitions. Thus, for the upward trending stock, the impact of animated (vs. static) display on risk judgments would be mitigated. Formally,

H₃ outlines a condition in which salient transitions may be less likely to be utilized to heighten risk judgments because of the context established by global trends. In other words, the global trend data feature helps dissociate transition salience from downstream risk inferences. In doing so, H₃ helps indirectly explicate how transitions are utilized in inference making.

Moderating How Transitions Are Utilized

Further down the causal chain, theory suggests that transitions that have been made salient (on account of animated display) are more likely to be utilized to infer risk. Therefore, varying how transitions are utilized in the inference process should moderate the core effect of display. To minimize potential confounding with alternative data features, we test this expectation via contextual factors exogenous to the data, such as investment goals (H₄).

Our theory is that animated display inflates risk judgments because of beliefs that transitions are associated with risk. In other words, investors’ belief that frequent changes in stock price are indicative of risk to investments drives the animated display effect. It follows therefore that decision contexts that highlight an opportunity for profits should moderate how transitions are utilized to form inferences about higher or lower levels of risk.

Prior work suggests that investors’ investment goal can affect the meaning ascribed to daily changes in stock prices. For example, investors sometimes infer both risk and opportunity from investment instruments (Zhou and Pham 2004). Drawing on this insight, we propose that the meaning ascribed to price changes would differ depending on the investment goal (long- or short-term investing). Specifically, when forming their portfolio, long-term investors tend to seek instruments that are relatively more stable, that is, with less fluctuation over time (e.g., dividend stocks). This is because they generally wish to continue their investment in these instruments over longer durations. Therefore, multiple transitions in a stock’s price should be associated with risk in long-term investors’ beliefs. In contrast, short-term investors (e.g., day traders) tend to seek volatile instruments to maximize profits by taking advantage of price variations in a short period. Thus, they should relate short-term transitions not to risk, but to an opportunity to make higher returns.

Drawing on this notion, we expect the investment goal to moderate the effect of animated display by varying the meaning inferred from salient temporal transitions. Long-term investors should infer greater risk (H_1a). However, for short-term investors, because transitions represent more of an opportunity to make money, animated display should not discernibly inflate risk judgments.

We test H₃ and H₄ by introducing global trends and investment goal as moderators in Study 4 and Study 5, respectively. Table 1 summarizes our studies and main findings.

OPEN IN VIEWER

Table 1.

Summary of Studies and Main Findings.

Pilot Field Study: Evidence from Oil Trading Experts (N = 21; Mage = 38.14 years; 90% male; oil trading managers)
	Animated Display (n = 10)		Static Display (n = 11)
Risk judgments	5.90 (.77)		5.05 (.93)
Belief about speculation	4.60 (1.65)		4.45 (1.51)
Belief about hedging	5.80 (1.81)		5.82 (1.17)
Trading experience (in years)	10.70 (7.36)		10.18 (8.38)
Main findings: Presenting crude oil price history in an animated (vs. static) graph leads to greater risk judgments.
Study 1: Underlying Process—Eye-Tracking Evidence (N = 85; Mage = 21.46 years; 40% male; students)
	Animated Display (n = 43)		Static Display (n = 42)
Risk judgments	5.23 (.84)		4.80 (.96)
Number of fixation sequences	5.19 (1.75)		2.12 (.92)
Fixation duration (in seconds)	11.50 (5.97)		9.62 (7.49)
Pupil diameters (in millimeters)	4.46 (.81)		4.55 (.96)
Main findings: Animated (vs. static) display enhances the salience of transitions in stock prices—evidenced by the number of fixation sequences following the animated trajectory. Presenting monthly stock prices in an animated (vs. static) graph leads to greater risk judgments.
Study 2a: Serial Mediation (N = 190; Mage = 35.46 years; 67% male; MTurk)
	Animated Display (n = 93)		Static Display (n = 97)
Risk judgments	5.56 (1.08)		5.21 (1.24)
Transition salience	5.93 (.89)		5.62 (1.06)
Utilization of transitions	5.76 (1.08)		5.42 (1.13)
Frequency of price change	11.23 (7.36)		11.35 (11.21)
Range of price change	19.41 (8.27)		19.24 (8.89)
Arousal	3.84 (1.31)		3.82 (1.26)
Involvement with stock investment	4.73 (1.69)		4.58 (1.57)
Stock trading experience	2.71 (1.56)		2.77 (1.37)
Main findings: Animated (vs. static) display enhances the salience of transitions in stock prices, which in turn leads to greater utilization of the transitions and, consequently, increases risk judgments.
Study 2b: Arousal Effect of Animation Speed (N = 165; Mage = 21.51 years; 43% male; students; six participants who skipped the stimulus page were excluded)
	Three-Second Animated (n = 54)	Ten-Second Animated (n = 51)	Static (n = 54)
Risk judgments	5.27 (1.13)	5.17 (1.02)	4.74 (.99)
Frequency of price change	8.70 (4.49)	9.55 (7.25)	8.37 (4.84)
Range of price change	20.48 (6.00)	20.10 (6.30)	19.11 (5.49)
Arousal	3.39 (1.32)	3.32 (1.25)	3.75 (1.13)
Main findings: Animated (vs. static) display of stock prices increases risk judgments both in the three-second and ten-second animated modes without any differences in arousal.
Study 3: Manipulating Transition Salience—Graph Type (N = 335; Mage = 36.55 years; 52% male; MTurk)
	Animated Line Graph (n = 81)	Static Line Graph (n = 85)	Animated Bar Graph (n = 82)	Static Bar Graph (n = 87)
Risk judgments	5.28 (1.43)	4.78 (1.22)	4.41 (1.39)	4.70 (1.39)
Transition salience	5.84 (.93)	5.45 (1.01)	5.49 (.88)	5.52 (.94)
Utilization of transitions	5.88 (1.04)	5.52 (1.08)	5.38 (1.13)	5.59 (1.09)
Main finding Animated (vs. static) display of stock prices increases risk judgments through salience and utilization of transitions when the focal data are presented in line graphs. However, the core effect and its underlying process do not manifest when the same data are presented in bar graphs.
Study 4: Contextualizing the Role of Transitions in Risk Judgments—Global Trend (N = 208; Mage = 21.09 years; 34% male; students; five participants who failed an attention check and four participants who did not view or follow the experiment instruction were excluded)
	Animated Downward Trend (n = 49)	Static Downward Trend (n = 49)	Animated Upward Trend (n = 51)	Static Upward Trend (n = 50)
Risk judgments	5.87 (1.04)	5.41 (.99)	4.02 (1.15)	4.27 (1.17)
Range of price change	21.65 (5.64)	21.96 (5.94)	22.22 (5.90)	23.52 (5.59)
Main findings: Animated (vs. static) display of stock prices increases risk judgments when global trend is downward, but the effect is mitigated when global trend is upward.
Study 5: Moderating How Transitions Are Utilized—Investment Goal (N = 216; Mage = 35.05 years; 59% male; Prolific Academic; three participants who did not view the investment goal manipulation were excluded)
	Animated Long-Term Goal (n = 55)	Static; Long-Term Goal (n = 53)	Animated; Short-Term Goal(n = 51)	Static; Short-Term Goal (n = 54)
Risk judgments	5.51 (1.13)	4.73 (1.39)	5.24 (1.16)	5.25 (1.10)
Investment decision—amount willing to invest ($)	2,414.69 (2,686.82)	4,712.26 (4,307.70)	5,373.45 (5,387.58)	5,129.72 (4,917.32)
Stock trading experience	3.31 (1.17)	3.17 (1.16)	3.22 (1.14)	3.02 (.98)
Financial literacy	2.60 (.78)	2.68 (.58)	2.37 (.94)	2.67 (.61)
Main findings: Animated (vs. static) display of stock prices increases risk judgments for long-term investors, but not for short-term investors. The amount that long-term investors are willing to invest is lower in the animated (vs. static) display mode, whereas there is no difference between the two display modes in the amount that short-term investors are willing to invest.

Notes: Values are means with standard deviation in parentheses.

Stimuli, Design, and Procedure

Eighty-five undergraduate students (M_age = 21.46 years; 40% male) participated in a single-factor (display mode: animated/static) between-subjects laboratory experiment in exchange for course credit. Participants were told that they would make financial judgments and were given a brief overview of a fictitious company. They were then randomly assigned to review one of two (animated vs. static) stock price graphs (see the Appendix). The focal graph depicted 21-day stock prices (average monthly trading days in the United States) generated randomly around a predetermined mean and variance (for details, see Web Appendix D). The display mode was manipulated as outlined in the previous section.

While participants reviewed the graph, a screen-based eye tracker (Tobii Pro X2-60) captured their eye movement (i.e., they did not need to wear any additional apparatus). After viewing the graph, participants reported risk judgments relating to buying the focal stock (1 = “Not at all risky,” and 7 = “Very risky”; 1 = “Safe,” and 7 = “Unsafe”; r = .57; adapted from Raghubir and Das 2010). Manipulation check measures (r = .83) and demographics were collected thereafter.

Results

We created seven rectangular areas of interest (AOIs) to test our theorizing (see Web Appendix E). These AOIs were positioned contiguously every three time periods (with no space in between and no overlap) so as to capture the 20 paths linking each of the 21 stock price points to one another. For analysis, we specifically focus on identifying successive patterns in fixation sequences to test whether participants’ gaze points fixate systematically on the data trajectory that unfolds following the order of the AOIs (i.e., test whether the animated display indeed enhances the salience of temporal transitions). We examined gaze points fixated on the AOIs during the first five seconds for both static and animated display modes because it took five seconds from the first moment of exposure for the entire animated trajectory to unfold.

We also measured the total duration of fixations within AOIs to capture overall attention. Further, to test for arousal, we measured dilations of participants’ right and left pupils (Bradley et al. 2008). To enable proper comparison, we captured the pupil dilations for the first five seconds of graph exposure for both static and animated display modes.

Discussion

In Study 1, eye-tracking data lends physiological support that animated display enhances the salience of temporal transitions (evidenced by the number of fixation sequences following the animated trajectory). Study 1 also replicates the core animated display effect: risk judgments are greater when the focal data trajectory is animated (vs. static). Moreover, results show that animated display does not affect the amount of attention devoted or arousal, helping rule them out as potential process constructs.

Importantly, in Study 1 we found that the effect of display mode on risk judgments was not directly mediated by the number of fixation sequences (b = .1064, SE = .2090, 95% confidence interval [CI] = [−.2887, .5371]). This result supports prior research as well as our expectation that an inference-making process may more completely explain the effect of animated display on downstream risk judgments. To further explore the full causal chain, which involves cognitive inferences, we test a serial process in Study 2a where we examine how enhanced salience caused by animated display leads to greater utilization of temporal transitions and, consequently, the heightening of risk judgments. Study 2b further investigates the role of arousal as an alternative explanation (via manipulation of animation speed).

Method

One hundred ninety individuals from Amazon Mechanical Turk (MTurk; M_age = 35.46 years; 67% male) participated in a single-factor (display mode: animated/static) between-subjects experiment. Participants were randomly assigned to review one of two (animated vs. static) monthly stock price graphs of a fictitious NASDAQ company, called Pegatrans (abbreviated PEGA). The stock price data used in Study 1 were presented in line form. In the animated mode, all data values were presented in a shaded trajectory for one second and then became more clearly visible as the trajectory was sequentially traced with animation. Thus, in both display modes participants were able to recognize the overall trajectory at the time they were exposed to the graph. This method formed an experimental control for the potential effect of uncertainty-induced arousal.

After reviewing the graph, participants reported risk judgments (r = .71) followed by salience and utilization of transitions, our two mediating variables. We measured the salience of transitions using five items (“daily changes in the PEGA stock price were:” “salient,” “vivid,” noticeable,” “visible,” and “clear”) measured on seven-point scales (1 = “Not at all,” and 7 = “Very” or “Highly”; α = .85). These scale items were adapted from prior salience literature for our visual context (see Web Appendix C).

Next, in keeping with the overarching cognitive-inference based conceptualization (Raghubir and Das 1999) we aimed to directly capture the utilization-of-transitions construct via participants’ self-report. In a separate pretest (N = 99), individuals were shown the stock price graph used in Study 2a and were asked to describe particular aspects of the data that they had considered in their decision to invest. Unaided open-response protocols revealed that approximately 60% utilized temporal transitions as a basis for judgment. We reviewed the protocols and selected phrases reflecting the definition of temporal transitions (i.e., changes in data values from one point in time to the next). Using these phrases, we developed a four-item scale measuring utilization of transitions (“I rated the risk of the PEGA stock based on:” “daily changes in the stock price,” “how the stock price varied one day to the next,” “how often the stock price changed on a daily basis,” and “how much the stock price rose and fell from day to day”; 1 = “Strongly disagree,” and 7 = “Strongly agree”; α = .87). Items were worded to capture the degree to which participants relied on transitions in forming their risk judgments. Convergent and discriminant validity was verified using factor analyses (for details, see Web Appendix G).

Subsequently, participants reported the frequency of price changes (“How many times do you think the stock price rose and fell in the graph?”) along with the highest and lowest prices they saw. The difference between these prices formed the recalled price range. These two measures (i.e., frequency and range of price changes) captured comprehension of overall variability of the focal data. Next, participants reported their arousal at the time of viewing the graph (α = .85), manipulation check measures for the display mode (r = .82), involvement with stock investment (α = .95), and prior stock trading experience (on a frequency response scale). Finally, they provided demographic information.

Results

Discussion

Study 2a yields support for the proposed full causal chain (H_1d): enhanced salience of transitions caused by animated display leads to greater utilization of transitions in forming risk inferences and, consequently, increases risk judgments. Moreover, the data do not support competing accounts of variability comprehension and arousal.

Method

One hundred sixty-five undergraduate students (M_age = 21.51 years; 43% male) participated in a single-factor (display mode: three-second animated, ten-second animated, or static) between-subjects laboratory experiment in exchange for course credit. Six participants who skipped the stimulus page were excluded from the analysis. Participants were given the same cover story as in Study 2a and were then randomly assigned to one of the three display mode conditions. We used the stock price data used in previous studies. The display mode was manipulated in the same way as in previous studies, but in the ten-second animated condition, the entire trajectory took ten seconds to unfold (for stimuli, see the Appendix). After reviewing the graph, participants reported risk judgments (r = .66) and the frequency and range of stock price changes they saw in the focal graph. Next, participants reported their arousal at the time of viewing the graph (α = .89) and manipulation check measures for the display mode (r = .81), followed by demographic information.

Results

Discussion

Taken together, these findings indicate that animated display increases risk judgments and that potential differences in arousal do not explain this effect. In combination with the pupil dilation evidence from Study 1, the experimental results from Study 2b suggest that arousal may not be a primary driver of the animated display effect. In the next study, we further test the salience-based cognitive inference account by experimentally manipulating transition salience in time-varying data (operationalized via the type of graphical representation).

Stimuli, Design, and Procedure

Three hundred thirty-five individuals from MTurk (M_age = 36.55 years; 52% male) participated in a 2 (graph type: line/bar) × 2 (display mode: animated/static) between-subjects experiment. We followed the same procedure as in Study 2a. Stock price data used in previous studies were presented either in line (i.e., baseline condition) or bar graphs (see Appendix). We manipulated the display mode in the same manner as in previous studies. After reviewing the graph, participants reported risk judgments (r = .71) followed by salience (α = .83) and utilization (α = .86) of transitions, using the same measures as in Study 2a. Finally, participants responded to manipulation check measures for display mode (r = .85) and provided demographic information.

Results

Discussion

Study 3 provides converging evidence for our proposed process by visually manipulating transition salience. Specifically, when the focal data are presented in line graphs, animated display enhances the salience of transitions, leading to greater utilization and, consequently, increasing risk judgments. However, the core effect and its underlying process do not manifest when the same data are presented in bar graphs. This finding suggests that temporal transitions should stand out in the earlier, perception stage to be further utilized in the subsequent inferential process.

Stimuli and Pilot Study

We manipulated global trends by rearranging the stock price data used in Studies 1 through 3 in descending and ascending orders while maintaining the same price fluctuations (i.e., both the downward and upward stocks had the same run length [1.3]; for stimuli, see Appendix). In doing so, we made the stock price data used in Study 4 clearly have either a downward slope (−.7) or an upward slope (.7), compared with the baseline, which has a virtually flat slope (.1). A pretest also confirmed that the slopes of the downward and upward trending stock data were perceived as being more downward and upward than the baseline stock data, respectively (both p < .001; for details, see Web Appendix I).

Next, we conducted a pilot study with 174 MTurk participants (M_age = 35.72 years; 63% male) to explore how animated display affects the salience of transitions and global trends. Per our theorizing, we expect animated display to enhance transition salience for both downward and upward trending stocks. However, as a global feature, trend should be salient irrespective of display mode and serve to contextualize the role of transitions. Consistent with our expectations, a two-way ANOVA with global trends and display mode as independent factors revealed a significant main effect for display mode (M_animated = 5.65, SD = .82; M_static = 5.32, SD = 1.17; F(1, 170) = 4.46, p = .036) on transition salience. No other significant main or interaction effects emerged. On the other hand, trend salience was not significantly different across all conditions (p = .307; for details, see Web Appendix J).

Main Study

Two hundred eight undergraduate students (M_age = 21.09 years; 34% male) participated in an online experiment in exchange for course credit. They were randomly assigned to one of four conditions following a 2 (global trend: downward/upward) × 2 (display mode: animated/static) between-subjects design. Five participants who failed an attention check and four participants who did not view or follow the instructions were excluded from the analysis.

After participants reviewed the graph stimuli, we measured their risk judgments using the two items measured on seven-point scales as in previous studies (r = .72). Next, participants reported the range of stock price changes they saw, followed by manipulation check measures for display mode (r = .81) and global trend (r = .99) and finally demographic information.

Results

Discussion

Study 4 provides empirical support for H₃ and, in the process, shows that global trend contextualizes how transitions—made salient through animated display—are utilized for inference making. We find that although animation makes transitions salient, the risk-inflating effect of these salient transitions manifests only when risk-related inference-making is warranted, that is, when the overall trend is downward. When prices are generally upward trending, the prospect of capital loss is dramatically lowered, and in such an environment, salient transitions do not lead to inferences of higher risk. The dissociation of transition salience from risk judgments bolsters our theorizing that a higher-level (possibly cognitive) inference-making process is implicated in how stimulus factors cause systematic variations in downstream judgments.

This process evidence by moderation should be interpreted with caution because transition salience was not directly measured in the main study. In addition, our experimental manipulation does not directly and independently manipulate the utilization-of-transitions construct. Since the global trend manipulation we use is itself a data feature, it combines with animated display to jointly shape how transitions are used. In Study 5, we address this concern by employing contextual manipulations that are exogenous to the data string and that directly affect how transitions are used to form judgments.

Stimuli, Design, and Procedure

Two hundred sixteen (M_age = 35.05 years; 59% male) individual investors were recruited from Prolific Academic. On average, participants traded stocks on a quarterly basis and also currently held a retirement plan. They were randomly assigned to one of four conditions following a 2 (investment goal: long-term/short-term investing) × 2 (display mode: animated/static) between-subjects factorial design. Three participants who did not view the investment goal manipulation were excluded from the analysis.

Participants were told that they would be asked to make financial judgments based on market information. They were then asked to imagine themselves as long-term or short-term investors and were given an investment goal to keep in mind during the experiment. Adapting past work (Zhou and Pham 2004), for the long-term investing group, we informed participants that long-term investors intend to maintain long-term profits by avoiding short-term stock price fluctuations, whereas for the short-term investing group, we informed participants that short-term investors intend to make immediate profits by taking advantage of stock price fluctuations. Next, we provided participants with a brief overview of a fictitious NASDAQ company. Stock price data used in previous studies were presented in line graph form. We manipulated the display mode in the same manner as in previous studies.

After reviewing the graph, participants reported their judgments of stock risk (two items; r = .72). Then, participants were asked to imagine that they had $25,000 for their portfolio and to indicate the amount they wished to invest in the focal stock. Following this, they responded to manipulation check measures for display mode (r = .80) and investment goal. We also measured participants’ trading experience (on a frequency response scale) and financial literacy (Lusardi and Mitchell 2007). Finally, participants provided demographic information.

To assess the validity of our investment goal manipulation, we conducted a separate posttest with 123 MTurk participants (M_age = 34.85 years; 63% male). Participants were given the same investment goal manipulation and asked to describe their general thoughts on how they interpret daily changes in stock prices. Two independent coders classified thought protocols and reconciled disputes through discussion. Overall, 40% of the interpretations were related to risk, while 31% related to opportunity. Other interpretations were related to supply and demand (12%), normal market phenomenon (5%), market up and down (3%), firm performance (2%), trading volume (2%), and trend (2%). Lastly, 3% provided no clear interpretations.

We observed that risk-related interpretations (e.g., volatility, instability, uncertainty, unsteady flow of cash, massive loss) and opportunity-related interpretations (e.g., good deal, potential for earnings, making quick money, cashing out to make the most profit) differed significantly depending on the investment goal. Within the long-term group, 66% inferred meanings related to risk, whereas 3% inferred opportunity. Importantly, risk-related meanings were more prominent than all other meanings combined (χ² = 30.20, p < .001). In contrast, within the short-term group, 55% inferred opportunity, whereas 17% inferred risk. In this group, opportunity-related meanings were more prominent than all other meanings combined (χ² = 38.73, p < .001). Taken together, the posttest results confirm that our investment goal manipulation indeed varied the meaning ascribed to the daily changes observed in stock prices.

Results

Discussion

Using manipulation of the investment goal, Study 5 provides additional evidence for our theory by moderating how transitions are utilized to form risk judgments. Results show that for long-term investors who infer risk from the transitions in stock prices, animated display inflates risk judgments. Reflecting their risk judgments, it reduces the amount that they are willing to invest. However, for short-term investors, who might infer opportunity from the same transitions, animated display does not increase risk judgments. In fact, we find that short-term investors invest more in the given stock, regardless of display mode, compared with long-term investors; this pattern is consistent with our manipulation.

To further explore the effect of animated display on consequential decision making, we conducted a follow-up study that examines how an individual investor characteristic, elaboration on potential outcomes (EPO; Nenkov, Inman, and Hulland 2008), interacts with animated display to affect incentivized investment behavior.

Prior literature has proposed that high-EPO individuals have a stronger tendency than low-EPO individuals to generate a variety of potential consequences and evaluate the importance of these consequences before making decisions (see Nenkov, Inman, and Hulland 2008). Moreover, low-EPO (vs. high-EPO) investors are more susceptible to information presentation effects (Nenkov et al. 2009). Therefore, we expect that susceptibility to the animated display effect varies depending on individuals’ chronic tendency to engage in predecision elaboration on potential outcomes. In the follow-up study, we examined this with the actual number of dollars participants decided to invest based on the reward structure in the cover story (for full study details, including incentivized decision structure, see Web Appendix K). Results revealed a significant interaction between display mode and EPO (b_{display × EPO} = .18, SE = .08, t(184) = 2.19, p = .03) on the amounts invested by participants when they made the focal investment decision. Specifically, low-EPO individuals (the EPO scale of 3.72 and below) invested significantly lesser money in the animated (vs. static) display mode (b = −.36, SE = .18, t(184) = −1.97, p = .05). However, the amount that high-EPO individuals (the EPO scale above 3.72) invested was not significantly different in the two display modes (animated vs. static).

In general, high-EPO investors were less affected by the animated display, but, interestingly, animated display appears to have helped low-EPO investors make equally self-regulated decisions, compared with high-EPO investors. This finding has significant public policy implications in that animated display offers a way to effectively debias investors who may be predisposed to riskier decision making.