Research Lounge: Megan Crawford-Grime, “The Crisis of Suggestion: Propaganda and Risky Behaviour”

RL 13/14: Megan Crawford-Grime

Date: Wednesday, 22 Jan 2014
Location: PG Hub 1, Senate House
Topic: The Crisis of Suggestion: Propaganda and Risky Behaviour
Speaker: Megan Crawford-Grime

About the Speaker: Megan earned a B.A. in Psychology (Cognitive) and a B.A. in Philosophy (Logic) at Texas A&M University-Corpus Christi before attending the Behavioral & Economic Science, MSc program at the University of Warwick. Her main research focus for the last 3 years has been investigating the various aspects of risky behaviour and rational decision-making. She plans to expand her work into the employment and job-seeking arena while living in Coventry.

Overview: The United States (U.S.) housing market crash in late 2007 was considered by the U.S. Secretary of the Treasury as “the most significant current risk to our economy” (Andrews, 2007, para. 2). By 2008 it was difficult to find a media outlet not reporting shock, dismay, and fear over the quick economic downturn. My question is to what extent does the frequency and salience of these dire descriptions about our economic atmosphere prime our rational investment decisions?

One factor that may play a role is repetition priming. This tells us that frequent exposure to a message eases its retrieval and recognition, also known as the availability heuristic (Glanzer & Bowles, 1976; Kahneman, 2011; Leicht, 1968; Zielske, 1959). It is in the ease of retrieval and recognition which can bring about the illusion of a message’s truth (Tulving & Pearlstone, 1966; Tverksy & Kahneman, 1973). The illusion of truth resulting from frequent exposure is the basis of the propaganda effect (Begg, Anas, & Farinacci, 1992; Chomsky, 1997; Jowett & O’Donnell, 1992). In other words, it is not the truth of the message that will eventually permeate people’s mindset and influence rational investment decision, but rather the frequency. If message frequency can influence decisions, then repeated messages of economic stability should permeate the investors’ mindset, which then should influence investment decisions. But how does this frequency influence investment decisions?

We have a suggestion of how the link between the investor’s mindset and investor’s decisions occurs, from Kahneman & Tversky’s prospect theory (PT; 1979). One principle of PT suggests that consistent positive messages of economic growth & stability, like what the US experienced in the 1980’s, should lead to perception of an economic state that is more certain. Perceptions of stability should then lead to decisions that are more risk-seeking.

Conversely, PT suggests that consistent negative messages of economic recession & instability, like the one we began to experience in 2008, should lead to perceptions of an uncertain economic environment. Perceptions of uncertainty should then lead to decisions that are more risk-averse.

Experiment 1: My research began by priming participants (N = 128) with economic messages that either suggested positive growth, neutral stagnation, or a negative recession. Participants were randomly assigned to one of the conditions. My prediction was that through repetition priming of messages perhaps participants’ risk-taking decisions could be changed. Using prospect theory’s framework, I hypothesized that repeated positive messages of economic growth would lead to a more risk-seeking mindset and result in this group being more willing to invest their money. By contrast repeated negative messages of an economic recession would lead to a more risk-averse mindset and result in this group being less willing to invest their money. The stagnant economic messages were neither of growth nor recession and I hypothesized that repeated neutral messages of economic stagnation would prime a mindset somewhere between the extreme risky mindsets and result in investment willingness that falls between the extreme investment behaviours.

Task #1: Each participant read three articles modeled after real-world articles from popular publications. That is, a single participant read three articles that shared perspectives on a single economic state: either all positive economy articles, all stagnant economy, or all recession economy articles. The articles reported false events, with falsified evidence, about real cities in the state of Texas (e.g. Austin, Houston, Dallas).

The articles retained the same skeletal framework and differed only in strategically placed trigger words. Trigger words were matched between and within by frequency, letter & syllable count, part of speech (Wilson, 1988), and scores from the Rubin & Friendly goodness & emotional scale (1986). To help illustrate, trigger words from three cross-message articles, one from each message condition, will be compared. Article A, from the positive growth message condition, contains the trigger words OPPORTUNITY and PROSPERITY in the beginning of the article. Each word has a high score in both goodness (5.4 and 5.9, respectively) and emotionality (5.4 and 5.9, respectively), measured on a 7-point scale. In article A, from the neutral stagnant message condition, the trigger words INVESTIGATE and ECONOMIC hold the same word position as OPPORTUNITY and PROSPERITY, but have neutral goodness scores (3.0 and 3.2, respectively) and neutral emotionality scores (3.1 and 3.0, respectively). Finally, article A, from the negative recession message condition, contains the trigger words SICKNESS and POVERTY in the same word position, but have low goodness scores (both 1.6) and high emotionality scores (both 6.9). After reading three articles that reflected one of the economic conditions, participants performed a lexical decision task.

Task #2: To help determine whether the repetition priming of each condition successfully manipulated the mindsets of the participants, a lexical decision task was used as a manipulation check. In a lexical decision task, participants quickly decide whether a letter string is a word or not a word. The stimuli for the task were 45 words and 45 pronounceable non-words (e.g., BINT). Fifteen of the words consisted of bad words (HORRIFIC), 15 consisted of good words (HAPPY), and 15 consisted of neutral words (DOOR). The non-words were generated from the ARC Nonword database (Rastle, Harrington, & Coltheart, 2002). The words selected for the task were sampled from emotionally good, bad, and neutral lists (Ruben & Friendly, 1986), and matched by the same standards to the article trigger words as well as within (Wilson, 1988).

If priming was successful then participants in the positive growth condition should have faster reaction times (RTs) to the good words compared to the bad words. Conversely, participants in the negative recession condition should be faster at recognizing the bad words compared to the good words. To control for extraneous variables, RTs that were ±2.5 standard deviations from the overall mean were removed from the analysis. Participants primed with positive growth messages averaged a 17ms priming effect for recognition of good words relative to bad words. Participants primed with negative recession messages averaged a 21ms priming effect for recognition of bad words relative to good words. Finally, participants primed with intermediate stagnant messages had average RTs that fell between both groups on all word-types (see Figure 1). The results support the repetition priming predictions. In other words, the first task priming successfully manipulated the mindset of the reader with respect to the differing emotional messages of the economy. The next step is to determine whether this priming effect carries the type of influential weight required to alter risk-taking decisions of the primed participants.

Figure 1Figure 1

Experiment 2: Risk-taking decisions in past literature usually involved participants evaluating hypothetical scenarios. For example, participants would be asked, “If you had $10 and you encountered an investment scenario with a more risky choice and a less risky choice, what do you think you would do?” In an effort to increase the ecological validity of the results, a separate group of participants (N = 100) first received a real payment of $10 when they entered the lab, then performed the same priming task (Task #1). Primed and paid participants were then provided with an opportunity to invest their $10 for a chance to receive more money. As well, participants were offered a chance to invest another valued commodity, their personal time. That is, each participant was offered a chance to invest their future time for a chance to lessen their time in the lab.

Task #3: First the participants were required to decide whether to invest their money or time. “Investing” in this sense, meant they paid in their $10 for the chance at more money and risked a longer lab time for the chance at leaving early. If they chose to take one of the investment options, they were presented with a investment choices that offered different levels of riskiness as well as different payouts.

Half the participants were randomly assigned to a dual investment scenario. In the monetary domain, the less risky options offered a 63% chance to receive $5 more and the more risky option offered a 60% chance to double their money (+5, .63; -10, .37 or +10, .60; -10, .40). In the temporal domain, the less risky option offered a 65% chance to shorten their time in the lab by 3 minutes, and the riskier option offered a 60% chance to end their lab time immediately (-3, .65; +10, .35 or -10, .60; +10, .40).

The other half of the participants were randomly assigned to a three investment scenario. In the monetary domain, a higher probability option was added, 65% chance to receive 3 more dollars (+3, .65; -10, .35 or +5, .63; -10, .37 or +10, .60; -10, .40). In the temporal domain, a moderate option was added, 63% chance to shorten their participation time by 5 minutes (-3, .65; +10, .35 or -5, .63; +10, .37 or -10, .60; +10, .40). Adding a third option allows for a deeper exploration of the priming effect. Often, risk assessment research offers two gambling options, and conclusions are drawn from scenarios that require only one acceptance and one rejection. Perhaps by comparing investment behaviors, more subtle preference differences can be detected that are not possible to discover with only two options. Adding a third option also allows for a more in depth look at both the superordinate and subordinate levels. The superordinate level of risk-seeking behavior is initially choosing to invest their money and/or time. The subordinate level of risk- seeking behavior is the particular probability chosen in the investment.

Results: Overall, at the superordinate level, participants who read messages promoting a growing economy exhibited the most risk-seeking behavior by being the most willing to invest both their money and time (p̂ = .42, .42, respectively). Those that read messages conveying a neutral tone of a stagnant economy exhibited a slightly smaller proportion of willing investors in both domains (p̂ = .32, .35, respectively). Finally, participants who read economic messages which highlighted a failing economy, exhibited the most risk-averse behavior by having the smallest proportion of willing investors in both domains (p̂ = .26, .23, respectively; see Figure 2). Though there is a trend in both domains which supports the original hypothesis, there is no significant investment preference between the conditions in either the monetary domain, χ2 (2, n = 19) = .75, p > .05, or temporal domain χ2 (2, n = 26) = 1.48, p > .05. At the superordinate level of investment behavior, the trend of the results support the hypothesis that frequent exposure to a message can implicitly influence perceptions of truth. When this influence is based on the stability or volatility of their economy, people can quickly translate that state to their immediate valued commodities.

RL 13/14: Megan Crawford-Grime (2)Figure 2

Staying at the superordinate level, of the participants given two monetary investment options, the growth message condition yields the highest proportion of willing investors
(n = 17, p̂ = .24), the recession message condition yields a smaller proportion of willing investors (n = 15, p̂ = .07), and no one primed with a stagnant economic message choose to invest (n = 18, p̂ = .00)*. The trend reverses and is magnified when three monetary investment options are offered. The stagnant message condition yields the highest proportion of willing investors (n = 13, p̂ = .46), followed by the recession (n = 13, p̂ = .31), then growth message conditions (n = 14, p̂ = .29; see Figure 3). By increasing the number of monetary investment options, taking an initial investment risk appears to be most attractive to participants primed with neutral messages about a stagnant economy, χ2 (1, n = 31) = 9.07, p < .05, V = .54, but regarded as equivalently attractive to participants primed with either growth or recession economic messages, though the recession condition did approach significance, χ2 (1, n = 28) = 3.04, p > .05.

RL 13/14: Megan Crawford-Grime (3)Figure 3

Of the participants given two temporal investment options, the growth message condition yields the highest proportion of willing investors (p̂ = .24), than either the stagnant (p̂ = .17) or recession (p̂ = .13) message conditions. This investment behavior is also magnified when given three temporal investment options. The growth message condition yields the highest proportion of willing investors (p̂ = .50) than either the stagnant (p̂ = .46) or recession (p̂ = .31) message conditions (see Figure 4). The difference in proportions within the economic message conditions is significant for the growth message condition, but not the others, χ2 (1, n = 31) = 4.33, p < .05, V = .38. It appears that participants primed with positive messages about the economy are more effected by the temporal investment change than the participants primed with other economic messages. The three superordinate measurements so far show that participants’ investment decisions may be reflecting the manipulation of the different emotional priming messages of economic certainty.

RL 13/14: Megan Crawford-Grime (4)Figure 4

Curiously, there were more willing investors, overall, when an extra option was offered in either domain. The increased attractiveness of more investment options could be a reflection of an illusion of choice. The increased number of choices may have the effect of increasing the participant’s feelings of autonomy, or freedom of choice (Katz & Assor, 2007). This increased feeling of autonomy can increase the participant’s certainty that they can win, even though this belief has no rational base (Langer, 1975). This increased certainty of winning can, then, make a risk appear more attractive (Kahneman & Tversky, 1979).

At the subordinate level, within the monetary domain, all the investing participants given two options exclusively choose the higher probability. Of the investing participants given three investment options, the growth message condition evenly divides between the moderate and lower probabilities (p̂ = .50, .50, respectively) favoring riskier preferences; the stagnant message condition evenly divides between higher, moderate, and lower probabilities (p̂ = .33, .33, .33, respectively) exhibiting equal risky preference; and the recession message condition favors the lower over the moderate and higher probabilities (p̂ = .50, .25., .25, respectively) favoring riskier preferences overall, but not as extreme as the growth message condition (see Figure 5). Participants primed with a positive message about a growing economy are more likely to take riskier monetary investments. Furthermore, participants primed with stronger emotional messages about the economy are more likely to be riskier with their monetary investments than participants primed with neutral messages, once they have taken the first step to invest at all.

RL 13/14: Megan Crawford-Grime (5)Figure 5

At the subordinate level, within the temporal domain, all the investing participants given two options exclusively choose the higher probability, just as in the monetary domain. Of the investing participants given three investment choices, the growth message condition favors the lower over the higher probability (p̂ = .71, .29, respectively), and the stagnant message condition also favors the lower over the higher probability (p̂ = .83, .17, respectively). Both conditions exhibit stronger risk-seeking behaviour. The recession message condition evenly divides between the lower and the higher probabilities (p̂ = .50, .50, respectively) reflecting an equivalent preference for both the risky-seeking and risk-averse behaviour (see Figure 6).

RL 13/14: Megan Crawford-Grime (6)Figure 6

Concluding Thoughts: This experiment helps to expose the persuasive power of the media. Results show that repeatedly reading messages about a particular economic state can influence short-term investment behavior. Experiment 1 shows that message priming is effective with respect to emotionally charged words. Experiment 2 shows that messages of economic stability can affect the desire to invest across different domains. Participants who read messages promoting a growing economy exhibited the most risk-seeking behavior by being the most willing to invest both their money and time. Conversely, participants who read economic messages which highlighted a failing economy, exhibited the most risk-averse behavior by being the least willing to invest either their money or time. At the superordinate level of investment behavior, the results support the hypothesis that frequent exposure to a message will implicitly influence perceptions of truth. When this influence is based on the stability or volatility of their economy, people can quickly translate that state to their immediate valued commodities.

It is important to understand the impact messages like this have on the everyday consumer. Perhaps a shopper is less likely to wait in line to purchase a basket-full of groceries and housewares after reading the negative headlines on the newsstands than one who has read positive headlines. Further, how would gambling practices change if patrons had to sit next to newsstands at every slot machine?

Interestingly, adding an extra investment option showed that risky behavior did not translate equivalently within, nor between, the monetary and temporal domains. Investors in both domains showed an increase in willingness to invest. However, the investment trend reversed for money, but not for time. Personal values placed on money appear to be more susceptible to various influences than values of time. Perhaps this is symptomatic of the different inherent relationships people have with money verses time. Money can be gained, spent, lost, and found. These experiences can be planned or unplanned. Time, on the other hand, is a fixed linear experience. Time cannot be saved for later, or lost without knowledge in the same manner as money. The stability of risky investment behavior with personal time may be a reflection of the stability of our relationship with time. Conversely, the instability of risky investment behavior with personal money may also be a reflection of the instability of our relationship with money.

As much as investment behavior did not translate equivalently from one domain to another, neither did it from the superordinate to the subordinate level. However, investing participants primed with a positive message about the economy were consistently more risk- seeking than investing participants primed with a recession message. The inconsistency of subordinate investment behavior by neutrally primed participants exposes an interesting element of emotionally neutral information. Without an emotionally charged message to prime the reader, risk decisions can become influenced by other factors.

References:
Begg, I. M., Anas, A., & Farinacci, S. (1992). Dissociation of processes in belief: Source recollection, statement familiarity, and the illusion of truth. Journal of Experimental Psychology: General , 121, 446-458.

Chomsky, N. (1997). Media control: The spectacular achievements of propaganda. (G. Ruggiero, & S. Sahulka, Eds.) Seven Stories Press.

Glanzer, M., & Bowles, N. (1976). Analysis of the word-frequency effect in recognition memory. Journal of Experimental Psychology: Human Learning & Memory , 21-31.

Jowett, S., & O’Donnell, V. (1992). Propaganda and Persuasion (2nd ed.). Newbury, CA: Sage Publications.

Kahneman, D. (2011). Thinking Fast and Slow. NY: Farrar, Straus and Giroux.

Kahneman, D., & Tversky, A. (1979). Prospects theory: An analysis of decision under risk. Econometrica , 47, 263-291.

Katz, I., & Assor, A. (2007). When choice motivates and when it does not. Educational Psychology Review , 19.4, 429-442.

Langer, E. (1975). The illusion of control. Journal of Personality and Social Psychology , 32, 311-328.

Leicht, K. L. (1968). Recall and judge frequency of implicitly occurring words. Journal of Verbal Learning & Verbal Behavior , 7, 918-923.

Rastle, K., Harrington, J., & Coltheart, M. (2002). 358,534 nonwords: The ARC nonword database. Quarterly Journal of Experimental Psychology , 55A, 1339-1362.

Ruben, D. C., & Friendly, W. (1986). Predicting which words get recalled: Measures of free recall, availability, goodness, emotionality, and pronunciability for 925 nouns. Memory & Cognition , 14, 79-94.

Tulving, E., & Pearlstone, Z. (1966). Availability versus accessibility of information in memory for words. Journal of Verbal Learn- ing & Verbal Behavior , 5, 381-391.

Tverksy, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology , 5, 207-232.

Wilson, M. (1988). The MRC psycholinguistic database: Machine readable dictionary, version 2. Behavioural Research Methods, Instruments and Computers , 20, 6-11.

Zielske, H. A. (1959). The remembering and forgetting of advertising. Journal of Marketing , 23, 239-243.

* The number of participants in each condition (n) do not sum to the overall number (N) because some participant information had to be removed from the final analysis. This was due to extremely fast article reading times (< 10 seconds) or incomplete participation. However, no participant included in the final analysis had data from more than one article removed due to short reading time.

 

© The content is provided by the speaker Megan Crawford-Grime to present the opinions and findings completely and accurately. Reference and data source may not be fully displayed but can be requested from the author. The Full content is considered as intellectual or academic work and strictly protected by copyright law. All rights reserved.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s