The Relationship Between Shocks and Higher-Order Risk Attitudes

We use a controlled economic experiment to assess the relationship between shocks and higher-order risk attitudes. Particularly, we study if previous apparent discrepancies are generated by the impossibility of field data separating the effects of the income shocks on wealth from those with risk attitudes. To do so, we propose a between-subjects experiment in which we vary the amount of initial experimental endowments (wealth) that subjects receive and the occurrence or non-occurrence of the shock. To capture these relationships from a broader set of subjects than college students, the experiment is implemented via Forthright Access, an online platform with a representative sample of the U.S. population. To capture whether endowment shocks affect our subjects' decision processes, we record decision times and overinvestment attempts. We exogenously isolate shocks from wealth to assess their independent effect on higher-order risk preferences.
We implement an experiment with seven randomly assigned treatments between subjects, as described in Table 1.

Table 1. Treatments
Endowment\Shock Deterministic Shock
$20 Deterministic-$20 Negative Shock-$20 (Negative shock from a 50% chance of a ($20, $60) bundle)
$60 Deterministic-$60 Positive Shock-$60 (Positive shock from a 50% chance of a ($20, $60) bundle)
Negative Shock-$60 (Negative shock from a 50% chance of a ($60, $100) bundle)
$100 Deterministic-$100 Positive Shock-$100 (Positive shock from a 50% chance of a ($60, $100) bundle)

To further investigate within-subject changes in risk attitudes following economic shocks, we will conduct a follow-up experiment with new participants recruited through Forthright Access. In Task 1, all participants will begin with a $60 endowment and complete four rounds of the Investment Game (Gneezy & Potters, 1997) to establish baseline risk preferences. In Task 2, participants will be randomly assigned to one of three treatments that determine their post-shock endowment:
WIN: 50% probability of keeping $60, 50% probability of gaining $40 (final endowment = $100).
LOSS: 50% probability of keeping $60, 50% probability of losing $40 (final endowment = $20).
C$60: Guaranteed $60 endowment (control).
After the shock assignment, participants will complete four additional rounds of the Investment Game, allowing us to measure within-subject changes in the proportion of the endowment invested in the risky asset.

Finally, in Task 3, participants will complete 18 rounds of risk-elicitation tasks (Eckel & Grossman, 2008; Noussair et al., 2012) to measure risk aversion, prudence, and temperance, followed by a brief social preferences questionnaire.

In a third experiment, new participants will be randomly assigned to one of two treatment arm conditions: Negative Shock-$20 or Positive Shock-$100. This experiment mirrors the same structure as the follow-up experiment (Tasks 1–3), with participants completing the Investment Game before and after the shock, followed by the same set of risk-elicitation tasks and social preference questions. The key difference is that this study will be conducted in the lab, where participants will be connected to biometric devices (eye-tracking and galvanic skin response) to record continuous physiological data throughout the experiment.

In Stage I, subjects learn their endowment and the source of it.
In Stage II, subjects will perform an investment task for 4 rounds to measure their risk aversion (Gneezy & Potters 1997). This tool has previously been employed to assess risk tolerance for economic shocks (Cohn et al., 2015; Holden & Tilahun, 2021). In this simple risk elicitation task, participants will be asked how much of their initial endowment, based on the random treatment assignment in Stage I ($20, $60, $100), they wish to invest in a risky option and how much to keep. The invested amount yields a dividend of 2.5 or 5.0, depending on the round, with a 50% probability. The decision maker keeps the money not invested.
In Stage III, subjects will perform an additional lottery selection task to measure attitudes toward risk (Eckel and Grossman, 2008). More specifically, subjects select a lottery of their preference from six available options (Eckel and Grossman, 2008). Each lottery has the same outcome probabilities but varies the payments of the low and high outcomes. Subsequently, subjects complete 17 binary selection lotteries designed to capture higher-order risk attitudes following Noussair et al. (2014). Table 2 describes the 17 lottery decisions presented in are grouped into four parts: five choices to assess risk aversion by comparing a sure payoff with a risky lottery; five choices to measure prudence, defined as the preference to assign unavoidable risks to higher income states rather than lower ones; five choices to assess temperance, reflecting the preference to spread independent risks across states rather than concentrating them; and two additional choices to test relative risk aversion and prudence under expected utility theory. The choice of the left lottery indicates risk aversion, prudence, and temperance.
Left Lottery Right Lottery
Risk Aversion 1 20 [65_5]
Risk Aversion 2 25 [65_5]
Risk Aversion 3 30 [65_5]
Risk Aversion 4 35 [65_5]
Risk Aversion 5 40 [65_5]
Prudence 1 [(90+ [20_-20]) _60] [(90_ (60+ [20_-20])]
Prudence 2 [(90+ [10_-10]) _60] [(90_ (60+ [10_-10])]
Prudence 3 [(90+ [40_-40]) _60] [(90_ (60+ [40_-40])]
Prudence 4 [(135+ [30_-30]) _90] [(135_ (90+ [30_-30])]
Prudence 5 [(65+ [20_-20]) _35] [(65_ (35+ [20_-20])]
Temperance 1 [(90+ [30_-30]) _ (90+ [30_-30])] [(90_ (90+ [30_-30] + [30_-30])]
Temperance 2 [(90+ [30_-30]) _ (90+ [10_-10])] [(90_ (90+ [30_-30] + [10_-10])]
Temperance 3 [(90+ [30_-30])_(90+[50_-50])] [(90_ (90+ [30_-30] + [50_-50])]
Temperance 4 [(30+ [10_-10]) _ (30+ [10_-10])] [(30_ (30+ [10_-10] + [10_-10])]
Temperance 5 [(70+ [30_-30]) _ (70+ [30_-30])] [(70_ (70+ [30_-30] + [30_-30])]
Ra_EU1 [40_30] [50_24]
Prud_EU2 [(50+ [25_-25]) _30] [(50_ (30+ [15_-15])]

Finally, participants complete a short questionnaire to assess sociodemographic and personality dimensions relevant to risk and shock behavior. Table 3 briefly provides a chronology of the tasks with the number of choices or questions per task and source.



Table 3. Chronology of tasks
Task Observed choices per subject Source
Task 1 Risk Aversion 8 Gneezy & Potters (1997)
Task 2 RA Simple 1 Eckel and Grossman 2008
Task 2 Risk Aversion 5 Noussair et al., (2014)
Task 2 Prudence 5 Noussair et al., (2014)
Task 2 Temperance 5 Noussair et al., (2014)
Task 2 RA EU1 1 Noussair et al., (2014)
Task 2 PRudEU2 1 Noussair et al., (2014)
Questionnaire (Ten-Item Personal Information) 10 Gosling, S. D., Rentfrow, P. J., & Swann, W. B., Jr. (2003).

Further descriptions of our experimental tasks can be found in Eckel & Grossman (2008), Gneezy & Potters (1997), and Noussair et al. (2014). The Forthright Access panel provides additional socio-demographics, including age, nationality, residency, education, ethnicity, employment, gender, income, marital status, political ideology, religion, sexual orientation, etc.

The follow-up study uses a within-subject design to measure changes in individual risk attitudes before and after an exogenous shock. All participants begin with a $60 endowment and complete four rounds of the Investment Game (Gneezy & Potters, 1997) to establish baseline risk preferences.
Participants are then randomly assigned to one of three treatments:
WIN – 50% chance of keeping $60, 50% chance of gaining $40 (ending with $100)
LOSS – 50% chance of keeping $60, 50% chance of losing $40 (ending with $20)
C$60 – guaranteed $60 (control)
After treatment assignment, participants complete four additional rounds of the Investment Game to capture post-shock investment behavior.
Lastly, they perform the same higher-order risk preference tasks as our previous experiment with a between-subjects design (Eckel & Grossman, 2008; Noussair et al., 2014), allowing comparison of pre- and post-shock measures for risk aversion, prudence, and temperance.


Physiological Response Experiment: The experiment will be conducted in the lab. All participants taking part in the sessions will be uniformly and randomly assigned to one of the shock treatment arm conditions: Negative Shock-$20 and Positive Shock-$100. The experimental structure of this study is the same as the Follow-up study: participants begin with a $60 endowment and complete four rounds of the Investment Game (Gneezy & Potters, 1997), then experience the randomly assigned shock, complete four additional rounds of the Investment Game, and finally perform the 18 higher-order risk preference tasks and the sociodemographic and personality questionnaire. At the beginning of the session, participants will be fitted with biometric devices, including eye-tracking and galvanic skin response (GSR) sensors. Physiological data will be recorded continuously throughout the length of the sessions, with event markers at each decision screen and at each realization screen. This design allows us to link within-subject changes in investment behavior and higher-order risk attitudes to changes in physiological arousal before, around, and after the shock.

2025-04-21

2026-01-31

1. Risky Investment Choices
2. Lottery choice
3. Risk Aversion Measure
4. Prudence Measure
5. Temperance Measure
6. Change in amount invested before vs. after the shock treatment.
7. Mean Galvanic Skin Response (GSR) conductance

1. Risky Investment Task (Gneezy & Potters, 1997): The proportion of the participant’s endowment invested in the risky asset in each round. A higher proportion indicates greater risk tolerance; a lower proportion indicates greater risk aversion.
2. Lottery Choice (Eckel & Grossman, 2008): Choice of one out of six ordered lotteries, with higher-numbered lotteries corresponding to lower risk aversion.
3. Risk Aversion Measure (Noussair et al., 2014): Number of safe choices (0–5) in paired decisions between a sure payoff and a risky lottery. Higher counts indicate stronger risk aversion.
4. Prudence Measure (Noussair et al., 2014): Number of prudent choices (0–5) in decisions designed to detect a preference for smoothing consumption across uncertain states.
5. Temperance Measure (Noussair et al., 2014): Number of temperate choices (0–5) in decisions designed to detect a preference for reducing exposure to compound risks.
6. Change in Risky Investment Before vs. After Shock Treatment (Experiment 2): The within-subject change (ΔY) in the proportion of the actual endowment invested in the risky asset before and after experiencing the shock (endowment change). This captures how an exogenous gain or loss affects risk-taking behavior in the investment task.
Following prior literature (Deck & Schlesinger, 2010; Ebert & Wiesen, 2014; Noussair et al., 2014; Xu et al., 2024), the number of risk-averse, prudent, and temperate choices is interpreted as the strength of the corresponding higher-order risk attitude.
7. Mean Galvanic Skin Response (GSR) conductance: We will compute mean GSR conductance within a pre-shock window and within a post-shock window for each treatment arm and compare these mean trajectories between the Negative Shock-$20 and Positive Shock-$100 groups.

Correlation between risk aversion, prudence, and temperance
Time taken to complete each decision (DT)
Overinvestment (OI)
Quiz Answers in the first Attempt (QA1)
Number of blinks
Heart rate variability (HRV)
GSR peaks/minute

Decision time (DT): We will perform robustness checks either by aggregating this subject characteristic as a continuous control variable or a dummy for those who took significantly longer (shorter) decision periods.
Notice that if Decision Time is significantly affected by our treatment dimension (i.e., the endowment shocks), then an analysis of decision time as an outcome variable is properly due.
Decision time might be a proxy for cognitive processing or emotional reactivity. As robustness checks, we will analyze whether positive or negative shocks lead to faster, more intuitive, slower, or deliberate decisions.

Finally, we will register the subject’s attempt to invest amounts above their endowment (Overinvestment). This variable, also known as DT, could also be argued to be an intrinsic characteristic of the subject. Thus, to be plausible as a control variable for our analysis, or if it could be affected by our treatment dimension (i.e., the endowment shocks), an analysis of decision time as an outcome variable is properly due.

Quiz Answers in the first Attempt (QA1): We will perform robustness checks either by aggregating this subject characteristic as a control variable or by decomposing the analysis between two groups of subjects

The number of incorrect attempts to pass the comprehension checks serves as an indirect measure of attentiveness or understanding of the tasks. While we will include all participants in the main data analysis, we will conduct robustness checks based on the number of attempts to pass comprehension questions. We will use all participants in our sample for data analysis, but for robustness checks, we will categorize participants based on the number of attempts to correctly answer the attention/comprehension quiz to evaluate whether our results are driven by inattentive behavior or misunderstanding of the tasks.

Number of blinks: Using the eye-tracking data, we will record the number of blinks per minute and construct a measure of the change in blink rate around the shock.
Heart rate variability: HRV can be used as an indicator of negative stress. We will calculate the mean of HRV within each treatment arm in a pre- and post-shock window for each treatment arm condition using the standard deviation of the IBI numbers obtained from the GSR sensor. Then, we will compare this measure between the two treatment arms.
GSR peaks/minute: We will compute the number of GSR peaks per minute in a pre- and post-shock window and compute the change in peak rate associated with the shock. We will compare this measure between the Negative Shock-$20 and the Positive Shock-$100 treatment arms.

We use two controlled online economic experiments to assess the relationship between shocks and higher-order risk attitudes. The first is a between-subjects design, varying both the initial endowment ($20, $60, or $100) and the occurrence or non-occurrence of an economic shock. The second is a within-subjects design in which all participants start with the same initial endowment and are then randomly assigned to experience a positive shock, negative shock, or no shock. This allows us to measure within-subject changes in risky investment and compare across treatment arms to assess the effects of shocks on prudence and temperance.

In both experiments, participants have a 10% chance of receiving a payment. After completing all tasks, participants spin a digital roulette to determine if they are selected for payment, with one round randomly chosen to determine their payoff.

We conduct two controlled online economic experiments to assess the relationship between economic shocks and higher-order risk attitudes (risk aversion, prudence, temperance).
In the between-subjects design, participants are randomly assigned to one of several treatment arms that vary both the initial endowment amount ($20, $60, $100) and whether they experience an economic shock. In shock treatments, participants face a 50% probability of receiving a higher or lower endowment than their assigned starting amount.
Table 1. Treatments
Endowment\Shock Deterministic Shock
$20 Deterministic-$20 Negative Shock-$20 (Negative shock from a 50% chance of a ($20, $60) bundle)
$60 Deterministic-$60 Positive Shock-$60 (Positive shock from a 50% chance of a ($20, $60) bundle)
Negative Shock-$60 (Negative shock from a 50% chance of a ($60, $100) bundle)
$100 Deterministic-$100 Positive Shock-$100 (Positive shock from a 50% chance of a ($60, $100) bundle)

In Stage I, subjects learn their endowment and the source of it.
In Stage II, subjects will perform an investment task to measure their risk aversion (Gneezy & Potters 1997). This tool was previously employed to assess risk tolerance for economic shocks (Cohn et al., 2015; Holden & Tilahun, 2021). In this simple risk elicitation task, participants will be asked how much of their initial endowment based on the random treatment assignment in Stage I ($20, $60, $100) they wish to invest in a risky option and how much to keep. The invested amount yields a dividend of 2.5 or 5.0, depending on the round, with a 50% probability. The decision maker keeps the money not invested.
In Stage III, subjects will perform an additional lottery selection task to measure attitudes toward risk (Eckel and Grossman, 2008). More specifically, subjects select a lottery of their preference from six available options (Eckel and Grossman, 2008). Each lottery has the same outcome probabilities but varies the payments of the low and high outcomes. Subsequently, subjects complete 17 binary selection lotteries designed to capture higher-order risk attitudes following Noussair et al. (2014). Table 2 describes the 17 lottery decisions presented are grouped into four parts: five choices to assess risk aversion by comparing a sure payoff with a risky lottery; five choices to measure prudence, defined as the preference to assign unavoidable risks to higher income states rather than lower ones; five choices to assess temperance, reflecting the preference to spread independent risks across states rather than concentrating them; and two additional choices to test relative risk aversion and prudence under expected utility theory. The choice of the left lottery indicates risk aversion, prudence, and temperance.
Left Lottery Right Lottery
Risk Aversion 1 20 [65_5]
Risk Aversion 2 25 [65_5]
Risk Aversion 3 30 [65_5]
Risk Aversion 4 35 [65_5]
Risk Aversion 5 40 [65_5]
Prudence 1 [(90+ [20_-20]) _60] [(90_ (60+ [20_-20])]
Prudence 2 [(90+ [10_-10]) _60] [(90_ (60+ [10_-10])]
Prudence 3 [(90+ [40_-40]) _60] [(90_ (60+ [40_-40])]
Prudence 4 [(135+ [30_-30]) _90] [(135_ (90+ [30_-30])]
Prudence 5 [(65+ [20_-20]) _35] [(65_ (35+ [20_-20])]
Temperance 1 [(90+ [30_-30]) _ (90+ [30_-30])] [(90_ (90+ [30_-30] + [30_-30])]
Temperance 2 [(90+ [30_-30]) _ (90+ [10_-10])] [(90_ (90+ [30_-30] + [10_-10])]
Temperance 3 [(90+ [30_-30])_(90+[50_-50])] [(90_ (90+ [30_-30] + [50_-50])]
Temperance 4 [(30+ [10_-10]) _ (30+ [10_-10])] [(30_ (30+ [10_-10] + [10_-10])]
Temperance 5 [(70+ [30_-30]) _ (70+ [30_-30])] [(70_ (70+ [30_-30] + [30_-30])]
Ra_EU1 [40_30] [50_24]
Prud_EU2 [(50+ [25_-25]) _30] [(50_ (30+ [15_-15])]

Finally, participants complete a short questionnaire to assess sociodemographic and personality dimensions relevant to risk and shock behavior. Table 3 briefly provides a chronology of the tasks with the number of choices or questions per task and source.

Table 3. Chronology of tasks
Task Observed choices per subject Source
Task 1 Risk Aversion 8 Gneezy & Potters (1997)
Task 2 RA Simple 1 Eckel and Grossman 2008
Task 2 Risk Aversion 5 Noussair et al., (2014)
Task 2 Prudence 5 Noussair et al., (2014)
Task 2 Temperance 5 Noussair et al., (2014)
Task 2 RA EU1 1 Noussair et al., (2014)
Task 2 PRudEU2 1 Noussair et al., (2014)
Questionnaire (Ten-Item Personal Information) 10 Gosling, S. D., Rentfrow, P. J., & Swann, W. B., Jr. (2003).

Further descriptions of our experimental tasks can be found in Eckel & Grossman (2008), Gneezy & Potters (1997), and Noussair et al. (2014). The Forthright Access panel provides additional socio-demographics, including age, nationality, residency, education, ethnicity, employment, gender, income, marital status, political ideology, religion, sexual orientation, etc.

In the within-subjects design, all participants start with an initial endowment of $60 and complete 4 rounds of the same investment game proposed by Gneezy and Potters (1997). Next, participants are randomly assigned to one of three treatments to determine their endowment for Task 2:
WIN: 50% probability of keeping $60 and 50% probability of gaining $40 (endowment becomes $100).
LOSS: 50% probability of keeping $60 and 50% probability of losing $40 (endowment becomes $20).
C$60: Endowment remains $60.
Participants then complete another 4 rounds of the Investment Game with their new endowment. The change in the proportion of their endowment invested before and after the shock provides an isolated measure of the effect of experiencing an economic shock on risk preferences.
In Task 3, participants complete 18 rounds of risk-elicitation tasks to measure risk aversion, prudence, and temperance:
Round 1: Choose one of six gambles (Eckel & Grossman, 2008).
Rounds 2–18: Make 17 binary choices between lotteries designed to measure higher-order risk preferences (Noussair et al., 2012).
After the risk-elicitation tasks, participants complete a short questionnaire assessing social preferences.

For the physiological response experiment, we will add event markers to the shock screen to delineate pre- and post-shock windows. These markers allow us to measure the change in physiological arousal caused by experiencing the shock within each treatment arm and then compare these responses across the Negative Shock-$20 and Positive Shock-$100 conditions. The structure and payment scheme is the same as the one described in the follow-up study that uses a within-subjects design.

The computer will do randomization.
The randomization will be conducted within the Otree experimental platform; each individual will be randomly assigned to a different treatment with equal probability.

individual

No

as many clusters as individual observations

1,050 individuals for the between-subjects experiment design. 440 individuals for the within-subjects experiment design.

1,050 individuals, 150 subjects per treatment arm
440 individuals, 88 subjects per treatment arm
200 individuals, 100 per treatment arm, to account for possible dropouts and secure enough risk-averse individuals for heterogeneity analysis.

For the between-subjects design, based on power calculations and previous studies (Cohn et al., 2015), which found an effect size of d=0.42, we estimate we will need 100 participants per cell to detect a slightly smaller effect size (d=0.35), considering that our conditions and experimental design differ. Since online participants often fail standard attention checks, we will over-sample to account for dropouts and incomplete session studies. For the within-subjects experiment design, a power analysis by simulation indicates that detecting the same effect size as in our previous experiment (Cohen’s d = 0.35) with 80% power requires 66 individuals per treatment arm. In our previous work, approximately 90% of participants were classified as risk-averse; however, for robustness checks restricted to this group, we assume a more conservative proportion of 80%. Based on this assumption, and allowing for potential dropouts or incomplete sessions, we will recruit 88 participants per treatment arm. We use the same power analysis by simulation conducted for the online follow-up experiment, which indicated that detecting an effect size of Cohen’s d = 0.35 (as in our previous studies) with 80% power requires 66 individuals per treatment arm. In our previous work, approximately 90% of participants were classified as risk-averse; however, for robustness checks restricted to this group, we assume a more conservative proportion of 80%. Based on this assumption, and allowing for potential dropouts or incomplete sessions, we will recruit 100 participants per treatment arm.

Relationship between Shocks and HORA

Relationship between Shocks and HORA- Latest Version