Study 2: Behavioral Intentions and Expectations Data Analysis

Author

Jamie C. Lee

Published

July 9, 2026

Summary

This document presents descriptive analyses for Study 2, which examined how experimental manipulation — via easy vs. hard quiz conditions using either an estate planning quiz (EP) or Raven’s Progressive Matrices (RM) — affects behavioral intentions (BI) and behavioral expectations (BE) related to estate planning as well as information-seeking behavior.

Sample. 800 participants were recruited on Prolific for a 2x2 between-subjects design. Participants were screened by age (30-65 years old), number of previous submissions on Prolific (100+), and approval rate (100%). The sample is balanced by gender (50-50 split). Our sample has an average age of 43.5 (SD=9.3), with most participants falling between the ages 30 and 44. Most participants have not completed either a will or a trust, and 41.1% of the sample have also not named beneficiaries for any of their financial accounts. 20.5% of participants reported having a will, and 10.2% of participants reported having a trust.

Manipulation checks. The easy/hard difficulty manipulation appears to have worked reasonably well, as indicated by participants’ perceived difficulty ratings and their objective performance on the quizzes, particularly in the Raven’s Matrix conditions. However, the confidence variable did not vary across conditions as expected: hard conditions produced slightly higher rather than lower measured confidence. This pattern is likely driven by the way the confidence variable is constructed, which depends on the difference between expected and actual performance. A more detailed discussion of this issue is provided in Section 2 of the document.

Behavioral intentions (BI) and behavioral expectations (BE). BI and BE were moderately-to-strongly correlated across estate planning actions, supporting the view that they are related but distinct constructs. Accordingly, the main analyses focus on the BE variables. Overall, average BE variables were low across experimental conditions, falling below the midpoint of the scale. In other words, participants reported a low likelihood of taking estate planning-related actions in the near term (e.g., drafting a will, creating a trust, talking to others about estate planning, seeking out information, etc.).

Condition assignment does not appear to meaningfully affect BI or BE. In contrast, prior estate planning engagement is more strongly associated with BE values: participants who have previously taken estate planning-related actions generally report higher likelihoods of taking additional actions in the future. This pattern suggests that prior engagement may signal greater readiness to continue progressing through the estate-planning process, and that informational interventions may be more effective when targeted toward individuals who have already begun engaging. However, even among participants with prior engagement, average BE values remain relatively low and still fall below the midpoint of the scale.

Information seeking behavior. Hover time across the eight estate planning content areas did not appear to differ meaningfully across experimental conditions. Hover behavior did not appear to be strongly related to prior estate planning engagement; however, participants with higher confidence in estate planning processes tended to spend less time hovering over informational content.

Data Dictionary

This section documents every column in behavioral_intentions_03.06.csv (N = 800 participants). Variable families that repeat across the 8 task items (q1-q8), the 8 information-seeking topics (accordion/hover fields), or the BI/BE action set are documented once, with the repeating pattern noted, rather than listed item-by-item.

0.1 Session and Recruitment Metadata

Column Description
start_dt / end_dt Timestamp participant started / finished the survey.
duration Total time spent on the survey, in seconds.
ip_address Participant’s IP address (Qualtrics default; not used in analysis).
latitude / longitude Approximate geolocation inferred from IP address (Qualtrics default; not used in analysis).
PROLIFIC_PID Participant’s unique Prolific ID.
SESSION_ID / STUDY_ID Prolific session / study identifiers.

0.2 Demographics and Condition

Column Description
age Participant age in years (screened 30-65).
female Gender indicator (1 = female, 0 = male; sample recruited to a 50/50 split).
condition Experimental condition, 1-4: 1 = Easy EP Quiz, 2 = Hard EP Quiz, 3 = Easy Raven’s Matrix, 4 = Hard Raven’s Matrix.

0.3 Prior Estate Planning Engagement (Self-Reported)

Column Description
ep_completion Single-item self-reported stage of estate planning progress, 1-5 (1 = no steps taken … 5 = completed EP documents); 6 = preferred not to answer.
prior_will 1 = has a will; blank = does not.
prior_trust 1 = has a trust; blank = does not.
prior_beneficiaries 1 = has named beneficiaries; blank = has not.
prior_poa 1 = has power of attorney; blank = does not.
prior_none 1 = selected “none of the above” for prior EP actions.
prior_noresponse 1 = preferred not to answer the prior-actions question.
prior_total Count of prior actions completed (0-4): sum of prior_will + prior_trust + prior_beneficiaries + prior_poa.

0.4 Experimental Task: Quiz / Raven’s Matrix Responses

Column Description
q1-q8 Participant’s selected answer for each of the 8 task items (Estate Planning quiz questions for conditions 1-2; Raven’s Progressive Matrices items for conditions 3-4).
c1-c8 Correctness of each of the 8 responses (1 = correct, 0 = incorrect); actual_correct = sum of c1-c8.
actual_correct Total number of task items answered correctly (0-8).
t_q1_first_click … t_q8_first_click Time (seconds) from page load to first click, for each of the 8 task items.
t_q1_last_click … t_q8_last_click Time (seconds) from page load to last click, for each of the 8 task items.
t_q1_page_submit … t_q8_page_submit Time (seconds) spent on the page before submitting, for each of the 8 task items; summed into condition_time_sec in the analysis setup.
t_q1_click_count … t_q8_click_count Number of clicks recorded on each of the 8 task items.

0.5 Post-Task Self-Report and Manipulation Checks

Column Description
guess_correct Participant’s numeric guess of how many of the 8 task items they answered correctly (0-8); paired with actual_correct to compute overconfidence.
guessing Self-reported reliance on guessing when answering task items (0 = not at all to 7 = on every question).
overconfidence Pre-computed in the raw data (not derived in this qmd) as the self-reported guess minus actual_correct; positive = overconfident, negative = underconfident, 0 = perfectly calibrated.
motivation Self-reported motivation during the task (0 = Not at all to 7 = Extremely).
enjoyment Self-reported enjoyment of the task (0 = Not at all to 7 = Extremely).
difficulty Self-reported perceived difficulty of the task (0 = Very easy to 7 = Very difficul).
ai_yn Whether the participant used an AI tool (e.g., ChatGPT) while completing the task (yes/no).
ai_use Follow-up detail on how the participant used AI, if ai_yn = yes.

0.6 Behavioral Intentions (BI) and Behavioral Expectations (BE)

Each estate planning action has a BI item (hypothetical likelihood of taking the action) and a matching BE item (likelihood of actually taking the action in the near term), both on a 9-point scale (1 = extremely unlikely, 9 = extremely likely). The will/beneficiary/trust/POA items are split into two wordings depending on whether the participant has already started that action (e.g., “draft” vs. “update”); these are collapsed via coalesce() in the analysis setup (bi_will_collapsed, be_will_collapsed, etc.).

Column Description
bi_will_draft / bi_will_update BI: likelihood of drafting a new will / updating an existing will.
bi_bene_add / bi_bene_update BI: likelihood of naming new beneficiaries / updating existing beneficiary designations.
bi_trust_create / bi_trust_update BI: likelihood of creating a new trust / updating an existing trust.
bi_poa_create / bi_poa_review BI: likelihood of creating a new power of attorney / reviewing an existing one.
bi_talk BI: likelihood of talking to family about estate planning.
bi_lawyer BI: likelihood of consulting a lawyer about estate planning.
bi_info BI: likelihood of seeking out more information about estate planning.
be_will_draft / be_will_update … be_info BE equivalents of all the above items (same actions, expectation-framed instead of intention-framed).

0.7 Confidence Items

Global items use a general “how confident/comfortable are you overall” framing; specific items ask about confidence in a particular estate-planning topic. All are on the confidence items’ native 1-7 scale (see Tables 9-10).

Column Description
global_overall Global confidence: overall confidence in one’s estate planning knowledge.
global_comfort Global confidence: comfort with estate planning topics generally.
global_confident Global confidence: general self-rated confidence in understanding estate planning.
specific_nowill Specific confidence: understanding the consequences of not having a will.
specific_bene Specific confidence: understanding beneficiary designations.
specific_ros Specific confidence: understanding right of survivorship.
specific_poa Specific confidence: understanding power of attorney.
specific_willvstrust Specific confidence: understanding the difference between a will and a trust.

0.8 Dependent-Variable Page Timing

Column Description
t_dv_first_click / t_dv_last_click Time (seconds) from page load to first / last click on the page containing the BI/BE/confidence items.
t_dv_page_submit Time (seconds) spent on the BI/BE/confidence page before submitting.
t_dv_click_count Number of clicks recorded on the BI/BE/confidence page.

0.9 Information-Seeking Behavior (8 EP Topics)

Participants could hover over / expand an accordion panel to view information on 8 estate-planning-related topics: Will, Trust, Probate, Beneficiary, POA, Starting EP, Timing, and Strategy. Each topic has four parallel measures, named with the pattern {measure}_{topic} where {topic} is one of will, trust, probate, beneficiary, poa, start, timing, strategy.

Column pattern Description
accordion_clicks_{topic} Number of times the participant clicked to expand/collapse the accordion panel for that topic.
hover_count_{topic} Number of times the participant hovered over that topic’s content.
hover_time_ms_{topic} Total time (ms) spent hovering over that topic’s content; log-transformed into hover_log_{topic} in the analysis setup.
hover_order_{topic} The order in which the participant first hovered over that topic, relative to the other 7.
hover_time_total_ms Sum of hover_time_ms_{topic} across all 8 topics.

1 Demographic and Behavioral Variables

1.1 Age

As Carolyn anticipated, even though I screened for 30-65 year olds, without stratification, our sample ended up including more participants in the lower end of this range. This pattern is evident in Figure 1, which shows the continuous age distribution, and in Figure 2, which groups participants into five-year intervals.

1.1.1 Continuous Distribution

Our sample has a mean age of 43.5 (SD = 9.3).

1.1.2 Grouped Distribution

Participants were grouped into 5-year age bands. The largest group was 30-34 (n = 164, 20.5%), and the smallest was 60-65 (n = 49, 6.1%).

1.2 Prior Estate Planning Engagement

1.2.1 Estate Planning Completion Stage

Participants indicated their current stage of estate planning on a 6-point scale, ranging from having taken no steps (1) to having completed one or more documents (5); 6 = “I prefer not to answer.”

Table 1. Distribution of Estate Planning Completion Stage
Response Label N %
1 I have not taken any steps. 243 30.4
2 I have thought about it but haven't done anything yet. 232 29.0
3 I have started learning or talking about it. 128 16.0
4 I have begun some planning. 96 12.0
5 I have completed one or more estate planning documents. 100 12.5
6 I prefer not to answer. 1 0.1

The most commonly reported stages were Response 1 (“I have not taken any steps.”; n = 243, 30.4%) and Response 2 (“I have thought about it but haven’t done anything yet.”; n = 232, 29%). Combined, these two stages account for more than 50% of the sample, suggesting that most participants have yet to complete one of the primary estate planning documents (will or trust) and are still early in the estate planning process.

Table 1 frequencies are visualized in Figure 3 below.

1.2.2 Specific Estate Planning Actions

Participants indicated which specific estate planning actions (more broadly defined here) they had previously completed. A value of 1 indicates the action was taken; blank values indicate the action has not been taken.

Table 2. Prior Estate Planning Actions
Action N %
Has a Will 164 20.5
Has a Trust 82 10.2
Named Beneficiaries 421 52.6
Has Power of Attorney 125 15.6
No Prior Planning 323 40.4
Preferred Not to Respond 7 0.9

Of the various actions inquired about, the most common action was naming beneficiaries (52.6% of participants). Consistent with the completion stage frequencies reported above, a decent share of participants (40.4%) reported no prior estate planning involvement.

Among the four core estate planning actions (will, trust, beneficiary designations, and power of attorney), participants completed an average of 1 action (SD = 1.1).

Table 3. Number of Prior Estate Planning Actions Completed (out of 4)
Actions Completed N %
0 329 41.1
1 284 35.5
2 96 12.0
3 48 6.0
4 43 5.4

Table 3 frequencies are visualized in Figure 4 below.

2 Manipulation Checks: Difficulty, Actual Performance, and Confidence

The experimental conditions are coded as follows:

  • Condition 1 = Easy Estate Planning Quiz

  • Condition 2 = Hard Estate Planning Quiz

  • Condition 3 = Easy Raven’s Matrix

  • Condition 4 = Hard Raven’s Matrix.

If the easy/hard manipulation was successful, participants assigned to the hard conditions (2 and 4) should report higher perceived difficulty, score lower on the assessment, and show lower confidence than those assigned to the easy conditions (1 and 3). Generally, this is what we find. However, consistent with pre-testing results, the easy vs. hard estate planning quiz seems to be a milder manipulation compared to the Raven’s Matrix tests. Questions should be further refined to improve the strength of the manipulation.

2.1 Difficulty

Following completion of the quiz, participants were asked: To what extent did you find the questions easy or difficult to answer correctly? Difficulty was measured on a 7-point scale from 1: Very easy to 7: Very difficult.

Overall, hard conditions (2 and 4) show higher difficulty ratings than easy conditions (1 and 3), consistent with a successful manipulation. This difference is more pronounced in the Raven’s Matrix conditions than in the Estate Planning Quiz conditions, suggesting that the estate planning manipulation can be further strengthened.

Table 4. Difficulty by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 4.97 1.35 1 7
Cond 2 (Hard EP) 202 5.36 1.57 1 7
Cond 3 (Easy RM) 198 4.52 1.47 1 7
Cond 4 (Hard RM) 198 5.57 1.40 1 7

The distributions of difficulty ratings for each condition are visualized in Figure 5 below.

2.2 Actual Performance

In addition to differences in perceived difficulty ratings, the easy/hard manipulation should lead to differences in participants’ actual performance, measured by the number of quiz questions answered correctly (out of 8 questions total).

Overall, easy conditions (1 and 3) resulted in higher scores than hard conditions (2 and 4), consistent with the perceived difficulty differences observed above. Again, the score difference is more pronounced in the Raven’s Matrix conditions than in the Estate Planning Quiz conditions.

Table 5. Actual Correct by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 4.72 1.65 1 8
Cond 2 (Hard EP) 202 3.38 2.06 0 8
Cond 3 (Easy RM) 198 5.86 1.89 1 8
Cond 4 (Hard RM) 198 3.34 1.78 0 8

The distributions of correct responses for each condition are visualized in Figure 6 below.

2.3 Confidence

2.3.1 Initial Confidence Measure

Confidence was initially calculated by subtracting each participant’s actual number of correct responses from their speculated number of correct responses. Thus, positive values (> 0) indicate overconfidence (participants believed they answered more questions correctly than they actually did), while negative values (< 0) indicate underconfidence.

In theory, hard conditions (2 and 4) should produce lower confidence than easy conditions (1 and 3). However, participants in the hard conditions also achieved lower actual scores, and because this confidence measure is defined relative to these scores, the expected pattern does not clearly emerge in the observed data. In addition, a substantial share of participants are perfectly calibrated (confidence score = 0; 25.6% of the sample), which likely attenuates differences across conditions.

Table 6. Frequency Distribution of Confidence (Full Sample)
Score N %
-6 1 0.1
-5 2 0.2
-4 6 0.8
-3 36 4.5
-2 95 11.9
-1 173 21.6
0 205 25.6
1 147 18.4
2 78 9.8
3 34 4.2
4 14 1.8
5 4 0.5
6 3 0.4
7 1 0.1
8 1 0.1

Somewhat counterintuitively, the hard conditions appear to produce slightly higher measured confidence than the easy conditions (see Table 7). Although this pattern appears inconsistent with prior findings (e.g., Hadar et al., 2013), it can be explained by the construction of the confidence metric: because it is defined as the difference between expected and actual performance, lower actual scores in the hard conditions mechanically push the measure upward.

At the same time, participants’ expected scores (how many questions they think they got correct) should also be lower in the hard conditions. If both expected and actual scores shift downward together, the difference between them - our confidence measure - may remain relatively unchanged across conditions. This is consistent with what we observe in the data (see Table 7 and Figure 7). Table 8 shows that participants’ speculated number of correct responses declines in the hard conditions, which closely resembles the pattern in actual performance shown in Table 5.

Together, these observations suggest that this initial confidence metric may not exclusively capture the intended construct of interest. Therefore, we use a different measure of confidence described in the next section.

Table 7. Confidence by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 -0.33 1.32 -6 4
Cond 2 (Hard EP) 202 0.41 1.73 -4 5
Cond 3 (Easy RM) 198 -0.41 1.72 -4 7
Cond 4 (Hard RM) 198 0.35 1.91 -5 8

Table 8. Speculated Number Correct (guess_correct) by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 4.39 1.71 0 8
Cond 2 (Hard EP) 202 3.79 1.71 0 8
Cond 3 (Easy RM) 198 5.44 1.78 0 8
Cond 4 (Hard RM) 198 3.69 1.70 0 8

2.3.2 Alternative Confidence Measure

This section uses exploratory factor analysis (EFA) to identify latent structure across eight self-reported confidence items — three reflecting global confidence in estate planning generally, and five reflecting specific confidence about particular topics or decisions. Factor scores are then used as predictors of all the BI and BE outcomes in the subsequent sections.

2.3.2.1 Confidence Item Descriptions

All items were collected on 7-point scales.

Table 9. Confidence Item Descriptions
Group Item Name Item Wording Scale
Global Global: Overall I feel like I have a good overall understanding of estate planning. 1: Strongly disagree – 7: Strongly agree
Global Global: Comfort I would feel comfortable explaining the basics of estate planning to someone else. 1: Strongly disagree – 7: Strongly agree
Global Global: Confident I feel confident in my understanding of how estate planning works. 1: Strongly disagree – 7: Strongly agree
Specific Specific: No Will How well do you feel you understand how assets are distributed if someone dies without a will? 1: Not at all – 7: Very well
Specific Specific: Beneficiaries How well do you feel you understand how beneficiary designations determine who receives assets? 1: Not at all – 7: Very well
Specific Specific: Right of Survivorship How well do you feel you understand how joint ownership with right of survivorship affects who inherits property? 1: Not at all – 7: Very well
Specific Specific: POA How well do you feel you understand the role of powers of attorney when someone becomes incapacitated? 1: Not at all – 7: Very well
Specific Specific: Will vs. Trust How well do you feel you understand how wills and trusts differ in how assets are distributed? 1: Not at all – 7: Very well

2.3.2.2 EFA: Item Descriptives and Correlations

People, on average, do not seem very confident in their understanding of estate planning. However, across the 8 confidence items, there appears to be moderate-to-strong inter-item correlations, which is consistent with the presence of latent structure and supports factor analysis.

Table 10. Descriptives for Confidence Items
Item N Mean SD Min Max
Global: Overall 800 3.29 1.74 1 7
Global: Comfort 800 2.94 1.81 1 7
Global: Confident 800 3.22 1.72 1 7
Specific: No Will 800 3.35 1.82 1 7
Specific: Beneficiaries 800 3.93 1.88 1 7
Specific: Right of Survivorship 800 3.21 1.84 1 7
Specific: POA 800 4.16 1.81 1 7
Specific: Will vs. Trust 800 3.18 1.85 1 7

Items are moderately-to-strongly correlated overall, with global items tending to correlate more strongly with each other than with specific items, suggesting the presence of at least two latent factors.

Table 11. Inter-Item Correlation Matrix
Item Global: Overall Global: Comfort Global: Confident Specific: No Will Specific: Beneficiaries Specific: Right of Survivorship Specific: POA Specific: Will vs. Trust
Global: Overall 1.000 0.866 0.941 0.671 0.636 0.665 0.585 0.731
Global: Comfort 0.866 1.000 0.862 0.645 0.582 0.644 0.508 0.716
Global: Confident 0.941 0.862 1.000 0.672 0.640 0.678 0.591 0.737
Specific: No Will 0.671 0.645 0.672 1.000 0.655 0.699 0.618 0.717
Specific: Beneficiaries 0.636 0.582 0.640 0.655 1.000 0.673 0.711 0.673
Specific: Right of Survivorship 0.665 0.644 0.678 0.699 0.673 1.000 0.582 0.713
Specific: POA 0.585 0.508 0.591 0.618 0.711 0.582 1.000 0.624
Specific: Will vs. Trust 0.731 0.716 0.737 0.717 0.673 0.713 0.624 1.000

2.3.2.3 EFA: Parallel Analysis and Factor Solution

To determine the number of factors to retain, a parallel analysis was conducted. This method compares the eigenvalues from the observed data with eigenvalues generated from 500 randomly simulated datasets with the same sample size and number of variables. Factors are retained when the eigenvalue from the observed data exceeds the corresponding simulated eigenvalue. The scree plot below (Figure 9) displays both the observed and simulated eigenvalues and suggests retaining two factors.

Parallel analysis suggests retaining 2 factors.

From the factor loadings reported in Table 12 and Figure 10, Factor 1 appears to capture confidence in specific EP processes, whereas Factor 2 reflects more general or global confidence in one’s understanding of EP.

Table 12. EFA Factor Loadings (oblimin rotation, ML estimation)
Item Factor 1 Factor 2 Communality
Global: Overall -0.016 0.985 0.945
Global: Comfort 0.043 0.858 0.797
Global: Confident 0.017 0.954 0.937
Specific: No Will 0.756 0.080 0.676
Specific: Beneficiaries 0.892 -0.071 0.698
Specific: Right of Survivorship 0.742 0.091 0.669
Specific: POA 0.843 -0.086 0.600
Specific: Will vs. Trust 0.652 0.230 0.723
Loadings < |0.30| suppressed for clarity. Communality = h².
Table 13. Variance Explained by Factor Solution
Factor SS Loadings Prop. Variance Cum. Variance
ML2 Factor 1 3.202 0.400 0.400
ML1 Factor 2 2.843 0.355 0.756

2.3.2.4 Factor Scores by Condition

The experimental manipulation does not appear to have meaningfully affected participants’ self-reported confidence in their EP understanding.

Table 14. Factor Score Means (SD) by Condition
Easy EP
Hard EP
Easy RM
Hard RM
Factor Cond 1 Cond 2 Cond 3 Cond 4
Factor 1 -0.065 (0.918) -0.109 (0.957) 0.094 (1.011) 0.083 (0.935)
Factor 2 0.023 (0.967) -0.162 (0.954) 0.086 (1.038) 0.056 (0.976)

Pairwise t-tests confirm that the conditions did not meaningfully affect participants’ self-reported confidence.

Table 15. Pairwise t-Tests: Factor Scores by Condition (Holm-Corrected)
Factor Comparison t df p (raw) p (Holm) sig
Factor 1 Cond 1 vs. 2 (Easy vs. Hard EP) 0.469 401.3 0.639 1.000
Cond 3 vs. 4 (Easy vs. Hard RM) 0.105 391.6 0.916 1.000
Cond 1+2 vs. 3+4 (EP vs. RM) -2.593 795.4 0.010 0.058 †
Factor 2 Cond 1 vs. 2 (Easy vs. Hard EP) 1.940 401.9 0.053 0.221
Cond 3 vs. 4 (Easy vs. Hard RM) 0.302 392.5 0.763 1.000
Cond 1+2 vs. 3+4 (EP vs. RM) -2.015 794.8 0.044 0.221
Welch's t-tests. sig column based on Holm-corrected p. † p < .10, * p < .05, ** p < .01, *** p < .001

3 Behavioral Intentions and Behavioral Expectations

Following the manipulation, participants were asked a series of questions about actions they might consider taking related to estate planning in the future. Specifically, they reported both their behavioral intentions and behavioral expectations for 7 estate planning-related actions.

Behavioral intentions (BI, bi_ prefix) capture what participants say they intend to do. Participants indicated whether they presently intend to perform each behavior within the next week using a 9-point scale anchored at 1: No, definitely do not intend and 9: Yes, definitely do intend.

Behavioral expectations (BE, be_ prefix) capture what participants expect they will actually do. Participants reported how likely it is that they will perform each behavior within the next week using a 9-point scale anchored at 1: Extremely unlikely and 9: Extremely likely. Prior literature suggests these constructs are related but conceptually distinct (Warshaw & Davis, 1985; Tanner & Carlson, 2009).

3.1 BI and BE Descriptives by Condition

3.1.1 Aggregated across Full Sample

Overall, both behavioral intentions (BI) and behavioral expectations (BE) are low across all conditions, with average values falling below the midpoint of the scale. This suggests that participants generally report a low likelihood of taking estate planning-related actions in the near term. However, BI and BE values are slightly higher in the EP quiz conditions than in the RM conditions, indicating that participants who completed the estate-planning quiz report somewhat greater intentions and expectations to take related actions.

Table 16. Behavioral Intentions by Condition — Mean (SD)
Easy EP
Hard EP
Easy RM
Hard RM
Variable Cond 1 Cond 2 Cond 3 Cond 4
Will 3.16 (2.83) 3.07 (2.71) 2.81 (2.75) 2.61 (2.47)
Beneficiaries 3.71 (3.06) 3.65 (2.9) 2.98 (2.89) 2.64 (2.54)
Trust 2.77 (2.54) 2.68 (2.46) 2.48 (2.49) 2.11 (1.97)
POA 3.21 (2.89) 2.91 (2.62) 2.71 (2.59) 2.4 (2.41)
Talk 3.42 (2.78) 3.6 (2.76) 3.38 (2.87) 3.19 (2.64)
Lawyer 2.7 (2.44) 2.55 (2.34) 2.53 (2.55) 2.23 (2.17)
Info 4.23 (3.08) 4.33 (3.03) 3.73 (2.91) 3.6 (2.72)
Table 17. Behavioral Expectations by Condition — Mean (SD)
Easy EP
Hard EP
Easy RM
Hard RM
Variable Cond 1 Cond 2 Cond 3 Cond 4
Will 2.46 (2.43) 2.43 (2.35) 2.2 (2.21) 1.97 (1.99)
Beneficiaries 3.2 (2.93) 2.94 (2.63) 2.47 (2.48) 2.33 (2.3)
Trust 2.15 (2.14) 2.1 (1.98) 1.94 (1.92) 1.66 (1.53)
POA 2.52 (2.42) 2.22 (2.25) 2.07 (1.92) 1.93 (1.98)
Talk 3.03 (2.61) 3.23 (2.67) 2.92 (2.52) 2.79 (2.47)
Lawyer 2.23 (2.12) 2.09 (2) 2.02 (2.07) 1.75 (1.69)
Info 3.75 (3.04) 3.88 (2.97) 3.21 (2.7) 3.13 (2.55)

The box plots in Figure 12 below suggest that although the mean and median BI and BE values do not differ substantially across conditions, the distribution of scores appears somewhat higher in the EP quiz conditions. This pattern suggests that while the manipulation may not shift the overall distribution of responses, it may increase intentions or expectations for a subset of participants. This possibility motivates additional analyses of specific subgroups - for example, splitting participants based on their prior estate planning engagement, as explored in the following subsection and subsequent sections.

Because behavioral expectations (BE) have been shown to be more predictive of actual future behavior, all subsequent analyses focus on BE variables. BI variables are therefore omitted from the main analyses. Additional robustness checks examining the relationship between BI and BE variables are reported in Appendix C.

3.1.2 Split by Prior EP Completion Stage

Altogether, the tables below suggest that participants who report being in the stage of “I have started learning or talking about estate planning” tend to have higher BEs across the 7 estate planning-related behaviors. This pattern appears to hold regardless of condition assignment or manipulation. However, these average BEs still rarely exceed the scale midpoint (5).

3.1.2.1 BE: Will

BE: Will — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.41 (1.2) 1.41 (1.28) 1.64 (1.54) 1.4 (1.12) 1.21 (0.68)
Thought about it 2.28 (2.22) 2.62 (2.56) 2.51 (2.23) 2.06 (2.1) 1.95 (1.92)
Started learning/talking 3.34 (2.71) 3.9 (2.82) 3.46 (2.98) 3.22 (2.64) 2.72 (2.27)
Begun some planning 2.9 (2.5) 3.12 (2.76) 2.42 (2.18) 3.45 (2.86) 2.27 (1.53)
Completed EP documents 2.35 (2.64) 2.18 (2.38) 2.75 (2.85) 1.65 (1.87) 2.79 (3.19)

3.1.2.2 BE: Beneficiaries

BE: Beneficiaries — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.81 (1.85) 1.9 (2) 2.3 (2.28) 1.62 (1.62) 1.44 (1.32)
Thought about it 2.88 (2.61) 3.44 (3.05) 3.17 (2.62) 2.23 (2.15) 2.6 (2.39)
Started learning/talking 3.73 (2.91) 4.5 (3.05) 3.92 (2.99) 3.69 (2.82) 2.76 (2.6)
Begun some planning 3.3 (2.71) 4.08 (3.06) 2.27 (1.64) 3.76 (3.21) 2.87 (2.03)
Completed EP documents 2.86 (3.02) 3.27 (3.27) 3.33 (3.27) 1.96 (2.37) 2.96 (3.1)

3.1.2.3 BE: Trust

BE: Trust — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.34 (1.12) 1.38 (1.34) 1.52 (1.41) 1.31 (0.96) 1.14 (0.56)
Thought about it 1.96 (1.75) 2.32 (2.12) 2.32 (1.98) 1.64 (1.48) 1.57 (1.16)
Started learning/talking 2.88 (2.42) 3.33 (2.88) 2.59 (2.22) 3.41 (2.51) 2.17 (1.89)
Begun some planning 2.2 (2.07) 2 (1.5) 2.23 (2.2) 2.41 (2.47) 2.07 (1.98)
Completed EP documents 2.12 (2.35) 2.41 (2.67) 2.25 (2.4) 1.62 (1.88) 2.25 (2.49)

3.1.2.4 BE: POA

BE: POA — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.43 (1.28) 1.48 (1.32) 1.46 (1.27) 1.52 (1.38) 1.27 (1.14)
Thought about it 2.21 (2.13) 2.6 (2.43) 2.45 (2.31) 1.68 (1.45) 2.06 (2.08)
Started learning/talking 2.86 (2.36) 3.47 (2.7) 2.92 (2.75) 2.91 (1.8) 2.1 (1.86)
Begun some planning 2.85 (2.57) 3.35 (2.88) 2.08 (1.9) 3.28 (2.75) 2.53 (2.5)
Completed EP documents 2.49 (2.65) 2.91 (2.81) 2.75 (3.05) 1.73 (2.03) 2.64 (2.67)

3.1.2.5 BE: Talk

BE: Talk — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.8 (1.63) 1.74 (1.49) 2.11 (1.92) 1.59 (1.21) 1.76 (1.77)
Thought about it 3.08 (2.48) 3.25 (2.69) 3.4 (2.57) 2.96 (2.4) 2.75 (2.25)
Started learning/talking 4.38 (2.83) 4.3 (2.79) 4.46 (3.05) 4.28 (2.69) 4.48 (2.86)
Begun some planning 4.08 (2.76) 3.81 (2.91) 4.04 (2.72) 4.48 (2.92) 3.87 (2.39)
Completed EP documents 2.9 (2.77) 3.32 (2.95) 3.04 (2.85) 2.42 (2.56) 2.89 (2.82)

3.1.2.6 BE: Lawyer

BE: Lawyer — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.35 (1.11) 1.44 (1.43) 1.41 (1.26) 1.29 (0.75) 1.24 (0.86)
Thought about it 2 (1.93) 2.32 (2.18) 2.45 (2.19) 1.81 (1.97) 1.48 (1.09)
Started learning/talking 2.86 (2.32) 3.23 (2.39) 2.68 (2.32) 3.22 (2.51) 2.31 (1.97)
Begun some planning 2.5 (2.29) 2.62 (2.12) 1.88 (1.61) 2.9 (2.82) 2.6 (2.44)
Completed EP documents 2.19 (2.4) 2.32 (2.51) 2.38 (2.53) 1.62 (1.83) 2.46 (2.67)

3.1.2.7 BE: Info

BE: Info — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 2.21 (2.05) 2.3 (2.16) 2.84 (2.43) 2 (1.81) 1.73 (1.61)
Thought about it 3.87 (2.82) 4.29 (3.17) 4.42 (3.07) 3.28 (2.37) 3.48 (2.47)
Started learning/talking 5.14 (2.8) 5.2 (3) 5.08 (2.96) 4.94 (2.86) 5.38 (2.41)
Begun some planning 4.1 (2.87) 3.96 (2.88) 3.73 (2.86) 4.66 (3.22) 3.93 (2.19)
Completed EP documents 3.09 (3.11) 4 (3.57) 3.79 (3.41) 2.04 (2.36) 2.75 (2.88)

3.1.2.8 BE: Will

BE: Will — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.41 (1.2) 1.41 (1.28) 1.64 (1.54) 1.4 (1.12) 1.21 (0.68)
Thought about it 2.28 (2.22) 2.62 (2.56) 2.51 (2.23) 2.06 (2.1) 1.95 (1.92)
Started learning/talking 3.34 (2.71) 3.9 (2.82) 3.46 (2.98) 3.22 (2.64) 2.72 (2.27)
Begun some planning 2.9 (2.5) 3.12 (2.76) 2.42 (2.18) 3.45 (2.86) 2.27 (1.53)
Completed EP documents 2.35 (2.64) 2.18 (2.38) 2.75 (2.85) 1.65 (1.87) 2.79 (3.19)
BE: Will — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 0.247 113.8 0.806
Easy RM vs. Hard RM 53 63 0.277 106.7 0.782
Started learning/talking Easy EP vs. Hard EP 30 37 0.620 63.4 0.537
Easy RM vs. Hard RM 32 29 0.788 58.8 0.434
Begun some planning Easy EP vs. Hard EP 26 26 1.004 47.4 0.320
Easy RM vs. Hard RM 29 15 1.783 41.9 0.082 †
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.9 BE: Beneficiaries

BE: Beneficiaries — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.81 (1.85) 1.9 (2) 2.3 (2.28) 1.62 (1.62) 1.44 (1.32)
Thought about it 2.88 (2.61) 3.44 (3.05) 3.17 (2.62) 2.23 (2.15) 2.6 (2.39)
Started learning/talking 3.73 (2.91) 4.5 (3.05) 3.92 (2.99) 3.69 (2.82) 2.76 (2.6)
Begun some planning 3.3 (2.71) 4.08 (3.06) 2.27 (1.64) 3.76 (3.21) 2.87 (2.03)
Completed EP documents 2.86 (3.02) 3.27 (3.27) 3.33 (3.27) 1.96 (2.37) 2.96 (3.1)
BE: Beneficiaries — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 0.521 113.9 0.603
Easy RM vs. Hard RM 53 63 -0.895 113.5 0.373
Started learning/talking Easy EP vs. Hard EP 30 37 0.782 61.7 0.437
Easy RM vs. Hard RM 32 29 1.338 59.0 0.186
Begun some planning Easy EP vs. Hard EP 26 26 2.657 38.3 0.011
Easy RM vs. Hard RM 29 15 1.123 40.1 0.268
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.10 BE: Trust

BE: Trust — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.34 (1.12) 1.38 (1.34) 1.52 (1.41) 1.31 (0.96) 1.14 (0.56)
Thought about it 1.96 (1.75) 2.32 (2.12) 2.32 (1.98) 1.64 (1.48) 1.57 (1.16)
Started learning/talking 2.88 (2.42) 3.33 (2.88) 2.59 (2.22) 3.41 (2.51) 2.17 (1.89)
Begun some planning 2.2 (2.07) 2 (1.5) 2.23 (2.2) 2.41 (2.47) 2.07 (1.98)
Completed EP documents 2.12 (2.35) 2.41 (2.67) 2.25 (2.4) 1.62 (1.88) 2.25 (2.49)
BE: Trust — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 -0.009 112.8 0.993
Easy RM vs. Hard RM 53 63 0.280 97.7 0.780
Started learning/talking Easy EP vs. Hard EP 30 37 1.154 53.6 0.253
Easy RM vs. Hard RM 32 29 2.179 57.1 0.033
Begun some planning Easy EP vs. Hard EP 26 26 -0.443 44.1 0.660
Easy RM vs. Hard RM 29 15 0.505 34.5 0.617
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.11 BE: POA

BE: POA — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.43 (1.28) 1.48 (1.32) 1.46 (1.27) 1.52 (1.38) 1.27 (1.14)
Thought about it 2.21 (2.13) 2.6 (2.43) 2.45 (2.31) 1.68 (1.45) 2.06 (2.08)
Started learning/talking 2.86 (2.36) 3.47 (2.7) 2.92 (2.75) 2.91 (1.8) 2.1 (1.86)
Begun some planning 2.85 (2.57) 3.35 (2.88) 2.08 (1.9) 3.28 (2.75) 2.53 (2.5)
Completed EP documents 2.49 (2.65) 2.91 (2.81) 2.75 (3.05) 1.73 (2.03) 2.64 (2.67)
BE: POA — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 0.341 112.3 0.734
Easy RM vs. Hard RM 53 63 -1.168 110.5 0.245
Started learning/talking Easy EP vs. Hard EP 30 37 0.819 62.6 0.416
Easy RM vs. Hard RM 32 29 1.709 58.0 0.093 †
Begun some planning Easy EP vs. Hard EP 26 26 1.875 43.2 0.068 †
Easy RM vs. Hard RM 29 15 0.901 30.9 0.374
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.12 BE: Talk

BE: Talk — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.8 (1.63) 1.74 (1.49) 2.11 (1.92) 1.59 (1.21) 1.76 (1.77)
Thought about it 3.08 (2.48) 3.25 (2.69) 3.4 (2.57) 2.96 (2.4) 2.75 (2.25)
Started learning/talking 4.38 (2.83) 4.3 (2.79) 4.46 (3.05) 4.28 (2.69) 4.48 (2.86)
Begun some planning 4.08 (2.76) 3.81 (2.91) 4.04 (2.72) 4.48 (2.92) 3.87 (2.39)
Completed EP documents 2.9 (2.77) 3.32 (2.95) 3.04 (2.85) 2.42 (2.56) 2.89 (2.82)
BE: Talk — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 -0.291 112.1 0.772
Easy RM vs. Hard RM 53 63 0.497 107.8 0.620
Started learning/talking Easy EP vs. Hard EP 30 37 -0.223 64.0 0.824
Easy RM vs. Hard RM 32 29 -0.283 57.5 0.779
Begun some planning Easy EP vs. Hard EP 26 26 -0.295 49.8 0.769
Easy RM vs. Hard RM 29 15 0.750 33.9 0.458
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.13 BE: Lawyer

BE: Lawyer — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.35 (1.11) 1.44 (1.43) 1.41 (1.26) 1.29 (0.75) 1.24 (0.86)
Thought about it 2 (1.93) 2.32 (2.18) 2.45 (2.19) 1.81 (1.97) 1.48 (1.09)
Started learning/talking 2.86 (2.32) 3.23 (2.39) 2.68 (2.32) 3.22 (2.51) 2.31 (1.97)
Begun some planning 2.5 (2.29) 2.62 (2.12) 1.88 (1.61) 2.9 (2.82) 2.6 (2.44)
Completed EP documents 2.19 (2.4) 2.32 (2.51) 2.38 (2.53) 1.62 (1.83) 2.46 (2.67)
BE: Lawyer — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 -0.332 110.5 0.740
Easy RM vs. Hard RM 53 63 1.104 77.9 0.273
Started learning/talking Easy EP vs. Hard EP 30 37 0.962 61.4 0.340
Easy RM vs. Hard RM 32 29 1.581 57.8 0.119
Begun some planning Easy EP vs. Hard EP 26 26 1.401 46.6 0.168
Easy RM vs. Hard RM 29 15 0.362 32.3 0.720
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.14 BE: Info

BE: Info — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 2.21 (2.05) 2.3 (2.16) 2.84 (2.43) 2 (1.81) 1.73 (1.61)
Thought about it 3.87 (2.82) 4.29 (3.17) 4.42 (3.07) 3.28 (2.37) 3.48 (2.47)
Started learning/talking 5.14 (2.8) 5.2 (3) 5.08 (2.96) 4.94 (2.86) 5.38 (2.41)
Begun some planning 4.1 (2.87) 3.96 (2.88) 3.73 (2.86) 4.66 (3.22) 3.93 (2.19)
Completed EP documents 3.09 (3.11) 4 (3.57) 3.79 (3.41) 2.04 (2.36) 2.75 (2.88)
BE: Info — Welch’s t-Tests by EP Stage (Stages 2–4)
EP Stage Comparison N (g1) N (g2) t df p sig
Thought about it Easy EP vs. Hard EP 63 53 -0.223 111.7 0.824
Easy RM vs. Hard RM 53 63 -0.428 112.0 0.669
Started learning/talking Easy EP vs. Hard EP 30 37 0.162 61.8 0.872
Easy RM vs. Hard RM 32 29 -0.654 58.7 0.516
Begun some planning Easy EP vs. Hard EP 26 26 0.290 50.0 0.773
Easy RM vs. Hard RM 29 15 0.878 38.7 0.386
Welch’s two-sample t-test. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.15 BE: Will

BE: Will — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.41 (1.2) 1.41 (1.28) 1.64 (1.54) 1.4 (1.12) 1.21 (0.68)
Thought about it 2.28 (2.22) 2.62 (2.56) 2.51 (2.23) 2.06 (2.1) 1.95 (1.92)
Started learning/talking 3.34 (2.71) 3.9 (2.82) 3.46 (2.98) 3.22 (2.64) 2.72 (2.27)
Begun some planning 2.9 (2.5) 3.12 (2.76) 2.42 (2.18) 3.45 (2.86) 2.27 (1.53)
Completed EP documents 2.35 (2.64) 2.18 (2.38) 2.75 (2.85) 1.65 (1.87) 2.79 (3.19)
BE: Will — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 0.758 232.4 0.449
Easy RM vs. Hard RM 114 107 1.745 212.6 0.082 †
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.16 BE: Beneficiaries

BE: Beneficiaries — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.81 (1.85) 1.9 (2) 2.3 (2.28) 1.62 (1.62) 1.44 (1.32)
Thought about it 2.88 (2.61) 3.44 (3.05) 3.17 (2.62) 2.23 (2.15) 2.6 (2.39)
Started learning/talking 3.73 (2.91) 4.5 (3.05) 3.92 (2.99) 3.69 (2.82) 2.76 (2.6)
Begun some planning 3.3 (2.71) 4.08 (3.06) 2.27 (1.64) 3.76 (3.21) 2.87 (2.03)
Completed EP documents 2.86 (3.02) 3.27 (3.27) 3.33 (3.27) 1.96 (2.37) 2.96 (3.1)
BE: Beneficiaries — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 1.729 229.2 0.085 †
Easy RM vs. Hard RM 114 107 1.001 217.9 0.318
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.17 BE: Trust

BE: Trust — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.34 (1.12) 1.38 (1.34) 1.52 (1.41) 1.31 (0.96) 1.14 (0.56)
Thought about it 1.96 (1.75) 2.32 (2.12) 2.32 (1.98) 1.64 (1.48) 1.57 (1.16)
Started learning/talking 2.88 (2.42) 3.33 (2.88) 2.59 (2.22) 3.41 (2.51) 2.17 (1.89)
Begun some planning 2.2 (2.07) 2 (1.5) 2.23 (2.2) 2.41 (2.47) 2.07 (1.98)
Completed EP documents 2.12 (2.35) 2.41 (2.67) 2.25 (2.4) 1.62 (1.88) 2.25 (2.49)
BE: Trust — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 0.409 232.3 0.683
Easy RM vs. Hard RM 114 107 2.094 202.3 0.037
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.18 BE: POA

BE: POA — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.43 (1.28) 1.48 (1.32) 1.46 (1.27) 1.52 (1.38) 1.27 (1.14)
Thought about it 2.21 (2.13) 2.6 (2.43) 2.45 (2.31) 1.68 (1.45) 2.06 (2.08)
Started learning/talking 2.86 (2.36) 3.47 (2.7) 2.92 (2.75) 2.91 (1.8) 2.1 (1.86)
Begun some planning 2.85 (2.57) 3.35 (2.88) 2.08 (1.9) 3.28 (2.75) 2.53 (2.5)
Completed EP documents 2.49 (2.65) 2.91 (2.81) 2.75 (3.05) 1.73 (2.03) 2.64 (2.67)
BE: POA — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 1.430 231.9 0.154
Easy RM vs. Hard RM 114 107 1.041 218.0 0.299
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.19 BE: Talk

BE: Talk — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.8 (1.63) 1.74 (1.49) 2.11 (1.92) 1.59 (1.21) 1.76 (1.77)
Thought about it 3.08 (2.48) 3.25 (2.69) 3.4 (2.57) 2.96 (2.4) 2.75 (2.25)
Started learning/talking 4.38 (2.83) 4.3 (2.79) 4.46 (3.05) 4.28 (2.69) 4.48 (2.86)
Begun some planning 4.08 (2.76) 3.81 (2.91) 4.04 (2.72) 4.48 (2.92) 3.87 (2.39)
Completed EP documents 2.9 (2.77) 3.32 (2.95) 3.04 (2.85) 2.42 (2.56) 2.89 (2.82)
BE: Talk — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 -0.664 232.8 0.507
Easy RM vs. Hard RM 114 107 0.980 219.0 0.328
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.20 BE: Lawyer

BE: Lawyer — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 1.35 (1.11) 1.44 (1.43) 1.41 (1.26) 1.29 (0.75) 1.24 (0.86)
Thought about it 2 (1.93) 2.32 (2.18) 2.45 (2.19) 1.81 (1.97) 1.48 (1.09)
Started learning/talking 2.86 (2.32) 3.23 (2.39) 2.68 (2.32) 3.22 (2.51) 2.31 (1.97)
Begun some planning 2.5 (2.29) 2.62 (2.12) 1.88 (1.61) 2.9 (2.82) 2.6 (2.44)
Completed EP documents 2.19 (2.4) 2.32 (2.51) 2.38 (2.53) 1.62 (1.83) 2.46 (2.67)
BE: Lawyer — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 0.763 232.8 0.446
Easy RM vs. Hard RM 114 107 2.241 200.0 0.026
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.2.21 BE: Info

BE: Info — Mean (SD) by EP Completion Stage and Condition
By Condition
EP Completion Stage Overall Easy EP Hard EP Easy RM Hard RM
No steps taken 2.21 (2.05) 2.3 (2.16) 2.84 (2.43) 2 (1.81) 1.73 (1.61)
Thought about it 3.87 (2.82) 4.29 (3.17) 4.42 (3.07) 3.28 (2.37) 3.48 (2.47)
Started learning/talking 5.14 (2.8) 5.2 (3) 5.08 (2.96) 4.94 (2.86) 5.38 (2.41)
Begun some planning 4.1 (2.87) 3.96 (2.88) 3.73 (2.86) 4.66 (3.22) 3.93 (2.19)
Completed EP documents 3.09 (3.11) 4 (3.57) 3.79 (3.41) 2.04 (2.36) 2.75 (2.88)
BE: Info — Welch’s t-Tests (EP Completion Stages 2–4 Pooled)
Comparison N (g1) N (g2) t df p sig
Easy EP vs. Hard EP 119 116 -0.072 233.0 0.942
Easy RM vs. Hard RM 114 107 0.112 218.6 0.911
Welch’s two-sample t-test. EP completion stages 2–4 pooled. † p < .10, * p < .05, ** p < .01, *** p < .001

3.1.3 Split by Prior Specific EP Actions

There may be a small effect of prior engagement in specific EP behaviors on behavioral expectations (BE) for future engagement.

In each table below, participants are divided into two groups: those who have not previously taken the specified EP action (e.g., creating a will) and those who have. Within each of these groups, participants are further split by experimental condition. Each row reports the average BE value, participants’ reported likelihood of taking the action within the next week, measured on a 9-point scale (1: Extremely unlikely to 9: Extremely likely).

Overall, the results suggest that the EP quiz may slightly increase BE values relative to the Raven’s Matrix conditions, with the easy EP quiz appearing somewhat more effective. However, there may be a stronger pattern with respect to prior behavior: participants who have previously engaged in EP-related actions generally report higher BE values for future actions. This pattern suggests that prior engagement may be an important predictor of readiness to take additional steps, and that informational interventions may be more effective when targeted to individuals who have already begun engaging with EP.

That said, average BE values remain relatively low - below the midpoint of the scale - even among participants with prior EP engagement.

3.1.3.1 Has a Will

BE Outcomes by Has a Will and Condition — Mean (SD)
Has not done this
Has done this
Variable Easy EP Hard EP Easy RM Hard RM Easy EP Hard EP Easy RM Hard RM
Will 2.22 (2.23) 2.29 (2.26) 2.04 (2.01) 1.73 (1.6) 3.35 (2.93) 3 (2.61) 2.79 (2.76) 3 (2.96)
Beneficiaries 3.08 (2.89) 2.9 (2.61) 2.39 (2.34) 2.21 (2.15) 3.65 (3.05) 3.1 (2.73) 2.79 (2.93) 2.82 (2.82)
Trust 2.03 (2.07) 2.06 (1.92) 1.84 (1.69) 1.54 (1.29) 2.63 (2.32) 2.3 (2.21) 2.3 (2.58) 2.13 (2.26)
POA 2.29 (2.25) 2.11 (2.12) 1.9 (1.61) 1.77 (1.76) 3.37 (2.8) 2.65 (2.68) 2.7 (2.7) 2.61 (2.65)
Talk 2.77 (2.51) 3.1 (2.61) 2.8 (2.41) 2.72 (2.4) 4 (2.79) 3.75 (2.84) 3.37 (2.88) 3.11 (2.74)
Lawyer 2.03 (1.97) 2.02 (1.88) 1.9 (1.85) 1.56 (1.39) 2.98 (2.48) 2.35 (2.41) 2.47 (2.7) 2.53 (2.46)
Info 3.64 (3.05) 3.86 (2.91) 3.14 (2.56) 3.12 (2.5) 4.14 (3) 3.95 (3.27) 3.47 (3.19) 3.16 (2.8)

3.1.3.2 Has a Trust

BE Outcomes by Has a Trust and Condition — Mean (SD)
Has not done this
Has done this
Variable Easy EP Hard EP Easy RM Hard RM Easy EP Hard EP Easy RM Hard RM
Will 2.45 (2.43) 2.36 (2.3) 2.21 (2.2) 1.97 (1.97) 2.52 (2.54) 3.17 (2.75) 2.16 (2.3) 2 (2.28)
Beneficiaries 3.19 (2.93) 2.95 (2.65) 2.46 (2.44) 2.34 (2.3) 3.29 (2.99) 2.83 (2.43) 2.6 (2.83) 2.22 (2.39)
Trust 2.12 (2.13) 2.03 (1.89) 1.89 (1.84) 1.62 (1.45) 2.48 (2.18) 2.89 (2.65) 2.28 (2.41) 2 (2.2)
POA 2.47 (2.37) 2.21 (2.25) 2.02 (1.84) 1.92 (1.98) 2.95 (2.78) 2.28 (2.35) 2.44 (2.42) 2.06 (2.1)
Talk 2.94 (2.57) 3.24 (2.68) 2.9 (2.5) 2.84 (2.48) 3.76 (2.91) 3.17 (2.57) 3.08 (2.74) 2.28 (2.35)
Lawyer 2.17 (2.04) 2.04 (1.97) 2.01 (2.02) 1.73 (1.64) 2.71 (2.7) 2.61 (2.25) 2.12 (2.39) 1.94 (2.1)
Info 3.67 (2.98) 3.84 (2.94) 3.28 (2.7) 3.25 (2.59) 4.38 (3.53) 4.33 (3.4) 2.76 (2.76) 1.94 (1.83)

3.1.3.3 Named Beneficiaries

BE Outcomes by Named Beneficiaries and Condition — Mean (SD)
Has not done this
Has done this
Variable Easy EP Hard EP Easy RM Hard RM Easy EP Hard EP Easy RM Hard RM
Will 2.16 (2.19) 2.29 (2.17) 2.17 (2.21) 1.74 (1.67) 2.73 (2.6) 2.56 (2.5) 2.23 (2.21) 2.2 (2.25)
Beneficiaries 2.53 (2.58) 2.82 (2.56) 2.42 (2.44) 1.9 (1.76) 3.79 (3.1) 3.04 (2.69) 2.52 (2.52) 2.76 (2.67)
Trust 2.01 (2.11) 1.91 (1.56) 1.7 (1.56) 1.44 (1.14) 2.28 (2.16) 2.28 (2.28) 2.14 (2.16) 1.87 (1.82)
POA 2.12 (2.09) 2.05 (1.95) 2.03 (1.94) 1.73 (1.8) 2.88 (2.63) 2.37 (2.49) 2.1 (1.91) 2.14 (2.14)
Talk 2.33 (2.16) 2.93 (2.48) 2.89 (2.56) 2.43 (2.26) 3.65 (2.83) 3.51 (2.81) 2.95 (2.5) 3.15 (2.62)
Lawyer 2 (1.91) 1.92 (1.67) 1.93 (2.05) 1.39 (1.07) 2.43 (2.27) 2.25 (2.25) 2.09 (2.08) 2.1 (2.08)
Info 3.14 (2.79) 3.9 (2.84) 3.16 (2.62) 2.85 (2.5) 4.29 (3.15) 3.87 (3.1) 3.26 (2.78) 3.41 (2.59)

3.1.3.4 Has Power of Attorney

BE Outcomes by Has Power of Attorney and Condition — Mean (SD)
Has not done this
Has done this
Variable Easy EP Hard EP Easy RM Hard RM Easy EP Hard EP Easy RM Hard RM
Will 2.43 (2.37) 2.4 (2.3) 2.27 (2.22) 1.97 (1.96) 2.64 (2.76) 2.61 (2.63) 1.85 (2.12) 1.96 (2.2)
Beneficiaries 3.17 (2.91) 2.94 (2.59) 2.58 (2.55) 2.32 (2.28) 3.33 (3.08) 2.94 (2.85) 1.97 (2.1) 2.39 (2.48)
Trust 2.14 (2.09) 2.12 (1.9) 1.99 (1.96) 1.69 (1.58) 2.24 (2.37) 2.03 (2.4) 1.67 (1.73) 1.43 (1.23)
POA 2.45 (2.29) 2.19 (2.15) 2.07 (1.93) 1.86 (1.85) 2.88 (3.01) 2.35 (2.78) 2.09 (1.93) 2.39 (2.66)
Talk 2.98 (2.55) 3.13 (2.59) 2.91 (2.54) 2.73 (2.39) 3.3 (2.96) 3.77 (3.03) 3 (2.44) 3.18 (2.92)
Lawyer 2.15 (1.99) 2.05 (1.87) 2.15 (2.17) 1.73 (1.62) 2.64 (2.68) 2.29 (2.61) 1.39 (1.27) 1.86 (2.07)
Info 3.67 (2.98) 3.83 (2.88) 3.3 (2.71) 3.18 (2.52) 4.15 (3.32) 4.16 (3.47) 2.79 (2.64) 2.86 (2.77)

3.2 Relationship between Confidence Level and BE (using the Alternative Confidence Measure)

3.2.1 Specific EP Confidence

Higher confidence across the five specific aspects of estate planning is associated with greater expected likelihood of engaging in each of the 7 estate planning-related behaviors. However, this relationship appears to be driven largely by the fact that both factors (specific and global) of confidence are correlated with prior EP engagement (see Table 18 below). This suggests that the experimental manipulation may not have meaningfully affected confidence and/or that this particular measure of confidence may instead reflect a more stable characteristic shaped by prior behavior.

Table 18. Correlations: Confidence Factors × Prior EP Engagement
Factor Variable N r t df p sig
Factor 1 EP Completion Stage (1-5) 799 0.482 15.545 797 0 ***
Prior Number of Actions Completed (0-4) 799 0.430 13.455 797 0 ***
Factor 2 EP Completion Stage (1-5) 799 0.535 17.884 797 0 ***
Prior Number of Actions Completed (0-4) 799 0.482 15.546 797 0 ***
ep_completion values of 6 excluded. † p < .10, * p < .05, ** p < .01, *** p < .001

Table 18 looks at this relationship at the level of the aggregate specific/global confidence factor scores. As a finer-grained check, Table 18b tests each specific confidence item against the single prior-action item it most directly corresponds to (e.g., whether having already named beneficiaries predicts confidence specifically about beneficiary designations), with condition included as a covariate in one version and omitted in the other so the two can be compared directly. Specific: Right of Survivorship has no matching item in the prior-actions battery and is excluded here.

Table 18b. Specific Confidence Items × Matching Prior Action, With and Without Condition Covariate
Without Condition
With Condition (factor)
Item Prior Action N B SE p B SE p
Specific: No Will Has a Will 800 1.144 0.154 0 *** 1.148 0.154 0 ***
Specific: Beneficiaries Named Beneficiaries 800 1.243 0.125 0 *** 1.243 0.124 0 ***
Specific: POA Has Power of Attorney 800 1.236 0.171 0 *** 1.236 0.171 0 ***
Specific: Will vs. Trust Has a Trust 800 1.835 0.206 0 *** 1.828 0.206 0 ***
B = effect of having taken the matching prior action (0/1) on the specific confidence item (1-7 scale). 'With Condition' models add condition as a 4-level factor covariate. † p < .10, * p < .05, ** p < .01, *** p < .001

Coefficients are virtually identical with and without condition in the model (e.g., “Has a Will” → Specific: No Will confidence: B = 1.144 without condition vs. 1.148 with condition; all four pairs show similarly negligible shifts). This is consistent with the null manipulation-check result in Table 15 — since condition does not meaningfully move confidence, there is no confound for it to absorb, so partialling it out here does not change the substantive conclusion. Each specific item is strongly predicted by its own matching prior action (all p < .001), reinforcing the earlier point that these confidence items likely reflect real, stable prior experience rather than something manipulable in the moment.

As a robustness check, Table 18c reverses the direction — predicting the (binary) prior action from the (continuous) confidence item, via logistic regression, again with and without condition as a covariate. Because prior action was reported at baseline and confidence was measured after the randomized manipulation, Table 18b’s direction is the one that matches the study’s actual timeline; this reversed model is included only to confirm both directions of the association agree.

Table 18c. Logistic Regressions: Matching Prior Action ~ Specific Confidence Item, With and Without Condition Covariate
Without Condition
With Condition (factor)
Item Prior Action N B (log-odds) OR p B (log-odds) OR p
Specific: No Will Has a Will 800 0.353 1.423 0 *** 0.356 1.427 0 ***
Specific: Beneficiaries Named Beneficiaries 800 0.381 1.464 0 *** 0.390 1.478 0 ***
Specific: POA Has Power of Attorney 800 0.443 1.558 0 *** 0.448 1.565 0 ***
Specific: Will vs. Trust Has a Trust 800 0.554 1.740 0 *** 0.553 1.739 0 ***
DV = matching prior action (0/1). B = log-odds effect of the specific confidence item (1-7 scale); OR = corresponding odds ratio. 'With Condition' models add condition as a 4-level factor covariate. † p < .10, * p < .05, ** p < .01, *** p < .001

As expected, the two directions agree: each confidence item significantly predicts its matching prior action (all p < .001), and the odds ratios are essentially unchanged whether condition is included or not (e.g., “Specific: No Will” confidence → “Has a Will”: OR = 1.423 without condition vs. 1.427 with condition). This mirrors Table 18b and further supports treating condition as a non-factor in this relationship — the association between specific confidence and its matching prior action holds regardless of which variable is modeled as the outcome.

Table 18c’s logistic model assumes the relationship is linear in the log-odds — i.e., that a step from confidence 1→2 shifts the log-odds by the same amount as a step from 3→4. Figure 12b checks that assumption directly by plotting the observed proportion who took the matching prior action at each raw confidence score (points, sized by N) against the model’s fitted logistic curve.

The fitted curve tracks the observed points reasonably well overall, but there are visible bumps at a few scores with small N — e.g., Specific: POA dips slightly from confidence = 1 to confidence = 2, and Specific: Beneficiaries is essentially flat between confidence = 2 and 3 before rising. These wiggles line up with the thinnest bins (as few as N = 37–38 participants at the top of the “No Will” and “Will vs. Trust” scales), so they most likely reflect sampling noise rather than genuine non-monotonicity. The logistic curve looks like a reasonable, if imperfect, summary of the relationship.

On the error-term question: logistic regression doesn’t have a free residual-variance parameter the way lm() does. In OLS, Y = Xβ + ε, where ε ~ N(0, σ²) and σ² is estimated from the data — that residual variance is exactly what R² divides out. In logistic regression, once you know the predicted probability p, the variance of a 0/1 outcome is mechanically fixed at p(1 − p) — there’s no separate noise parameter to estimate, so there’s no direct equivalent of “residual standard error” or R².

What’s used instead are pseudo-R² measures, which approximate how much better the model does than one with no predictors. Table 18d reports two: McFadden’s R² (based on a ratio of log-likelihoods; conventionally read on a much more conservative scale than OLS R² — .2–.4 already counts as an excellent fit) and Tjur’s R² (the average predicted probability among those who took the action minus the average predicted probability among those who didn’t — this one actually is a proportion, in the same units as the outcome).

Table 18d. Pseudo-R2 for Logistic Models (Confidence Predicting Matching Prior Action, No Covariate)
Item Prior Action N McFadden's R2 Tjur's R2
Specific: No Will Has a Will 800 0.064 0.066
Specific: Beneficiaries Named Beneficiaries 800 0.082 0.109
Specific: POA Has Power of Attorney 800 0.077 0.070
Specific: Will vs. Trust Has a Trust 800 0.135 0.097
McFadden's R2 = 1 - logLik(model)/logLik(null-model). Tjur's R2 = mean predicted probability among those who took the action minus mean predicted probability among those who did not.

Both measures land in the .06–.14 range across the four items. That means confidence is a statistically reliable predictor of the matching prior action (consistent with the p < .001 results throughout), but it only explains a modest share of who has and hasn’t taken each action — most of the variation in prior behavior is coming from something other than current confidence. This is a common pattern with large samples: a real, precisely estimated effect can still leave most of the outcome’s variation unexplained.

3.2.1.1 A Note on Causal Direction

Everything above establishes that specific confidence and its matching prior action are correlated — it does not establish which one drives which. It’s entirely plausible that taking the action builds confidence (learning-by-doing), rather than confidence enabling the action. Both prior_will/prior_trust/prior_beneficiaries/prior_poa and the specific_* confidence items were reported within the same survey session, so there is no true “before/after” timepoint separating them here — the earlier appeal to timing (prior action as a baseline trait) reflects when the underlying behavior happened in the participant’s life, not when it was measured, and it does not by itself rule out reverse causation or a common third cause (e.g., general estate-planning engagement or conscientiousness driving both the behavior and the self-reported confidence).

Figure 12c looks at this from a different angle than Figure 12b: instead of the model-implied probability curve, it shows the full observed distribution of confidence responses separately for people who have and have not taken the matching action.

The two distributions are shifted relative to each other (as the earlier tables already implied) but they overlap substantially — the “have not done this” group is not just floored at 1–2. For example, among those who have not named beneficiaries, 24% still report top-two-box confidence (6 or 7); among those who have named beneficiaries, 15% report bottom-two-box confidence (1 or 2). That overlap is informative in its own right: if simply having done the action mechanically produced high confidence, the “have not done this” distribution should pile up near the bottom of the scale far more than it does. The substantial spread on both sides is more consistent with confidence and behavior being correlated-but-distinct — plausibly both downstream of a shared disposition — than with a clean one-directional causal story in either direction.

This distributional check doesn’t resolve the direction question, though, and honestly nothing in a cross-sectional, single-timepoint self-report survey like this one can fully rule out reverse causation or a common cause. One diagnostic that would add some (still imperfect) evidence is a specificity check: correlate each specific confidence item with all four prior-action variables (not just its matching one). If confidence about, say, powers of attorney tracks only prior_poa and not prior_will/prior_trust/prior_beneficiaries, that pattern is more consistent with confidence reflecting specific lived experience with that particular action (whichever direction it runs) than with a single generic “I’m an engaged planner” trait driving everything at once. If instead each specific item correlates with all four prior actions roughly equally, that would point toward a shared underlying disposition as the more likely explanation for the correlations in Tables 18b/18c, rather than a specific link between an item and its own matching action.

Table 18e. Specificity Check: Each Specific Confidence Item x Each Prior Action (Point-Biserial r)
Item Has a Will Named Beneficiaries Has Power of Attorney Has a Trust
Specific: No Will 0.255*** 0.217*** 0.182*** 0.272***
Specific: Beneficiaries 0.254*** 0.331*** 0.191*** 0.248***
Specific: POA 0.252*** 0.214*** 0.248*** 0.244***
Specific: Will vs. Trust 0.292*** 0.235*** 0.202*** 0.301***
Bold = matching item-action pair (same pairing as Tables 18b/18c). † p < .10, * p < .05, ** p < .01, *** p < .001

The pattern is mixed, and if anything leans toward a shared disposition rather than item-specific tracking. Only two of the four items show their matching action as the single strongest correlate: Specific: Beneficiaries (r = .331 with Named Beneficiaries, clearly its top correlate) and Specific: Will vs. Trust (r = .301 with Has a Trust, narrowly ahead of Has a Will at .292). The other two do not: Specific: POA’s strongest correlate is actually Has a Will (r = .252) rather than its own matching Has Power of Attorney (r = .248) — essentially a tie — and Specific: No Will’s strongest correlate is Has a Trust (r = .272), not Has a Will itself (r = .255). More broadly, every row shows meaningful, similarly-sized correlations across all four prior actions (roughly .18–.33 throughout) rather than a sharp diagonal standing out against near-zero off-diagonal values. That’s more consistent with a general “estate-planning engagement” disposition contributing to both prior behavior and self-reported confidence across the board than with confidence being built specifically by having done that particular action. Given this, I’d treat the specific confidence items primarily as markers of general planning engagement rather than as evidence of an item-specific causal link, and would flag the direction of causality as an open question in any write-up rather than assuming action → confidence.

The relationship between confidence and BE appears to be fairly noisy with poor model fit (see scatterplots in Figures 13 and 15 below). Thus, I would not treat it too seriously for now.

Table 19. Regressions: Factor 1 → BE Outcomes
Outcome N B SE R2 p sig
BE: Will 800 0.690 0.080 0.086 0 ***
BE: Beneficiaries 800 0.669 0.094 0.060 0 ***
BE: Trust 800 0.580 0.068 0.084 0 ***
BE: POA 800 0.670 0.076 0.088 0 ***
BE: Talk 800 0.745 0.091 0.077 0 ***
BE: Lawyer 800 0.635 0.070 0.095 0 ***
BE: Info 800 0.719 0.102 0.059 0 ***
† p < .10, * p < .05, ** p < .01, *** p < .001

3.2.2 Global EP Confidence

Like the specific EP results in the previous section, higher general confidence around estate planning is associated with greater expected likelihood of engaging in each of the 7 estate planning-related behaviors. However, again, this relationship appears to be driven largely by the fact that confidence is correlated with prior EP engagement.

Table 21. Regressions: Factor 2 → BE Outcomes
Outcome N B SE R2 p sig
BE: Will 800 0.603 0.078 0.070 0 ***
BE: Beneficiaries 800 0.544 0.092 0.042 0 ***
BE: Trust 800 0.533 0.066 0.076 0 ***
BE: POA 800 0.614 0.074 0.079 0 ***
BE: Talk 800 0.651 0.089 0.063 0 ***
BE: Lawyer 800 0.564 0.068 0.079 0 ***
BE: Info 800 0.572 0.100 0.040 0 ***
† p < .10, * p < .05, ** p < .01, *** p < .001

3.3 Information Seeking Behavior

Hover time across 8 informational content areas (will, trust, probate, beneficiary, POA, starting estate planning, timing, and strategy) was converted from milliseconds to seconds (÷ 1,000) and log-transformed using log(x+1) to address the right-skewness of the data. Overall, there does not appear to be any meaningful differences across conditions.

3.3.1 Distribution of Logged Hover Time (Aggregated across Conditions) by Topic

Unlike the results from Study 1, a substantial share of participants in this study (approximately 20–30%) did not hover to view the information for each topic. There is no clear evidence that participants who did not hover had more prior EP engagement than those who did. Moreover, across the full distribution of hover times, there does not appear to be a relationship between prior EP engagement and logged hover time (see Table 22 below).

Table 22. Correlations: Hover Time × EP Engagement (Holm-Corrected)
Topic Predictor N r t df p (raw) p (Holm) sig
Will EP Completion Stage 799 -0.081 -2.300 797 0.022 0.326
Prior Actions Completed 799 -0.080 -2.276 797 0.023 0.326
Trust EP Completion Stage 799 -0.021 -0.592 797 0.554 1.000
Prior Actions Completed 799 -0.043 -1.227 797 0.220 1.000
Probate EP Completion Stage 799 -0.012 -0.332 797 0.740 1.000
Prior Actions Completed 799 -0.024 -0.688 797 0.492 1.000
Beneficiary EP Completion Stage 799 -0.012 -0.330 797 0.742 1.000
Prior Actions Completed 799 -0.004 -0.103 797 0.918 1.000
POA EP Completion Stage 799 -0.038 -1.075 797 0.283 1.000
Prior Actions Completed 799 -0.032 -0.898 797 0.370 1.000
Starting EP EP Completion Stage 799 -0.031 -0.887 797 0.376 1.000
Prior Actions Completed 799 -0.079 -2.241 797 0.025 0.329
Timing EP Completion Stage 799 -0.060 -1.688 797 0.092 1.000
Prior Actions Completed 799 -0.101 -2.866 797 0.004 0.068 †
Strategy EP Completion Stage 799 0.030 0.856 797 0.392 1.000
Prior Actions Completed 799 0.001 0.025 797 0.980 1.000
Holm correction applied across all tests. ep_completion values of 6 excluded. † p < .10, * p < .05, ** p < .01, *** p < .001

3.3.2 Logged Hover Time Descriptives by Condition

Logged hover time across topics does not appear to differ meaningfully by condition. Pairwise t-tests indicate that these differences are not statistically significant. For brevity, these analyses are not included in the document.

Table 23. Logged Hover Time by Topic and Condition — Mean (SD)
Easy EP
Hard EP
Easy RM
Hard RM
Topic Easy EP Hard EP Easy RM Hard RM
Will 2.32 (1.22) 2.31 (1.21) 2.22 (1.17) 2 (1.25)
Trust 1.45 (1.05) 1.62 (1.02) 1.26 (1.04) 1.3 (1.02)
Probate 1.5 (1.18) 1.63 (1.17) 1.44 (1.17) 1.44 (1.19)
Beneficiary 1.56 (1.25) 1.56 (1.26) 1.26 (1.2) 1.29 (1.23)
POA 1.23 (1.06) 1.2 (1.03) 0.96 (1) 0.94 (0.99)
Starting EP 1.25 (1.05) 1.25 (1.02) 1.13 (1.02) 1.11 (1.05)
Timing 1.01 (0.9) 1.11 (0.93) 1.05 (0.91) 0.93 (0.93)
Strategy 0.83 (0.78) 0.83 (0.74) 0.72 (0.74) 0.73 (0.72)

3.3.3 Relationship between Confidence and Hover Time

Overall, across most EP topics, participants with higher confidence tended to spend less time hovering over the informational content.

Table 24. Correlations: Hover Time × Confidence Factors (Holm-Corrected)
Topic Predictor N r t df p (raw) p (Holm) sig
Will Confidence Factor 1 800 -0.148 -4.233 798 0.000 0.000 ***
Confidence Factor 2 800 -0.165 -4.712 798 0.000 0.000 ***
Trust Confidence Factor 1 800 -0.101 -2.878 798 0.004 0.029 *
Confidence Factor 2 800 -0.143 -4.084 798 0.000 0.001 **
Probate Confidence Factor 1 800 -0.046 -1.298 798 0.195 0.389
Confidence Factor 2 800 -0.098 -2.796 798 0.005 0.032 *
Beneficiary Confidence Factor 1 800 -0.083 -2.362 798 0.018 0.074 †
Confidence Factor 2 800 -0.116 -3.309 798 0.001 0.009 **
POA Confidence Factor 1 800 -0.095 -2.705 798 0.007 0.035 *
Confidence Factor 2 800 -0.113 -3.213 798 0.001 0.011 *
Starting EP Confidence Factor 1 800 -0.126 -3.601 798 0.000 0.003 **
Confidence Factor 2 800 -0.158 -4.522 798 0.000 0.000 ***
Timing Confidence Factor 1 800 -0.134 -3.830 798 0.000 0.002 **
Confidence Factor 2 800 -0.145 -4.126 798 0.000 0.001 **
Strategy Confidence Factor 1 800 -0.043 -1.216 798 0.224 0.389
Confidence Factor 2 800 -0.066 -1.859 798 0.063 0.190
Holm correction applied across all tests. † p < .10, * p < .05, ** p < .01, *** p < .001

4 Appendix

4.1 Appendix A: Randomization Check

Age and prior estate planning involvement are compared across the four experimental conditions to assess whether randomization was successful.

4.1.1 Age by Condition

Table A1. Age by Condition
Condition N Mean SD Min Max
Condition 1 202 44.3 9.7 30 65
Condition 2 202 43.8 9.3 30 64
Condition 3 198 42.8 9.2 30 63
Condition 4 198 43.0 9.0 30 65

Age appears broadly comparable across conditions.

4.1.2 Prior Estate Planning by Condition

Table A2. Prior Estate Planning Actions by Condition (N, %)
Action Condition 1 Condition 2 Condition 3 Condition 4
Has a Will 43 (21.3%) 40 (19.8%) 43 (21.7%) 38 (19.2%)
Has a Trust 21 (10.4%) 18 (8.9%) 25 (12.6%) 18 (9.1%)
Named Beneficiaries 107 (53%) 106 (52.5%) 109 (55.1%) 99 (50%)
Has Power of Attorney 33 (16.3%) 31 (15.3%) 33 (16.7%) 28 (14.1%)
Table A3. Mean Number of Prior Estate Planning Actions by Condition
Condition N Mean SD
Condition 1 202 1.01 1.12
Condition 2 202 0.97 1.05
Condition 3 198 1.06 1.19
Condition 4 198 0.92 1.12

The distribution of prior estate planning actions appears broadly similar across conditions, suggesting successful randomization with respect to prior engagement.

4.2 Appendix B: Additional Manipulation Checks

Motivation, enjoyment, guessing, and total time on task are examined across the four conditions. These variables should be relatively comparable across conditions, to confirm that the manipulation did not inadvertently affect other constructs.

4.2.1 Motivation

Table B1. Motivation by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 6.13 1.24 1 7
Cond 2 (Hard EP) 202 6.00 1.22 1 7
Cond 3 (Easy RM) 198 6.03 1.31 1 7
Cond 4 (Hard RM) 198 5.71 1.46 1 7

4.2.2 Enjoyment

Table B2. Enjoyment by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 4.28 1.77 1 7
Cond 2 (Hard EP) 202 4.29 1.74 1 7
Cond 3 (Easy RM) 198 4.46 1.91 1 7
Cond 4 (Hard RM) 198 3.63 1.85 1 7

4.2.3 Total Time on Task

Total time on task (condition_time_sec) is defined as the sum of time in seconds spent across all 8 quiz pages.

Table B3. Total Time on Task (seconds) by Condition
Condition N Mean SD Min Max
Cond 1 (Easy EP) 202 213.44 139.20 71.3 1184.1
Cond 2 (Hard EP) 202 255.24 163.45 68.8 1514.1
Cond 3 (Easy RM) 198 244.08 123.66 76.2 777.2
Cond 4 (Hard RM) 198 333.13 238.89 17.9 1831.3

Participants did not appear to spend substantially more time in the hard conditions compared to the easy ones, suggesting that any increased difficulty was not primarily driven by additional time on task.

4.3 Appendix C: Relationship between BI and BE

Based on prior literature, BI and BE are expected to be positively correlated, with BE values typically slightly lower than BI values. This section of the appendix examines whether the two measures show this expected pattern.

4.3.1 Correlations Between BI–BE Counterparts

Moderate-to-high but sub-1.0 correlations support treating BI and BE as related but distinct constructs, consistent with prior literature.

Table C1. Pearson Correlations Between BI and BE Counterparts
Action N r (BI–BE)
Will 800 0.731
Beneficiaries 800 0.786
Trust 800 0.716
Talk to Family 800 0.798
Consult Lawyer 800 0.760
Seek Information 800 0.823
POA 800 0.710

4.3.2 Signed Difference Scores (BI − BE)

Positive values indicate that participants report stronger intentions than expectations (BI > BE), while negative values indicate the opposite; a value of zero reflects perfect alignment between the two. As expected, BI values are, on average, slightly higher than their BE counterparts, which is consistent with patterns documented in prior literature.

Table C2. Signed Difference Scores (BI − BE): Descriptives
Action N Mean SD Min Max
Will 800 0.65 1.86 -8 8
Beneficiaries 800 0.51 1.82 -6 8
Trust 800 0.55 1.68 -7 8
POA 800 0.62 1.88 -8 8
Talk 800 0.40 1.70 -7 8
Lawyer 800 0.48 1.55 -6 8
Info 800 0.48 1.73 -6 8
Table C3. Signed Difference Scores (BI − BE) by EP Completion Stage — Mean (SD)
Action No steps Thought about it Started learning Begun planning Completed docs
Will 0.47 (1.45) 0.79 (1.89) 0.67 (2.1) 0.62 (1.92) 0.75 (2.28)
Beneficiaries 0.35 (1.37) 0.62 (1.98) 0.42 (1.99) 0.64 (1.75) 0.67 (2.21)
Trust 0.38 (1.22) 0.6 (1.88) 0.54 (1.78) 0.69 (1.59) 0.73 (2.05)
POA 0.45 (1.39) 0.74 (1.92) 0.74 (2.15) 0.72 (1.93) 0.54 (2.36)
Talk 0.26 (1.32) 0.58 (1.69) 0.36 (1.95) 0.34 (1.97) 0.46 (1.97)
Lawyer 0.3 (1.24) 0.52 (1.58) 0.59 (1.63) 0.62 (1.78) 0.55 (1.82)
Info 0.32 (1.34) 0.56 (1.72) 0.44 (2.01) 0.77 (2.11) 0.45 (1.78)
Positive values = BI > BE; negative values = BE > BI; zero = perfect alignment.