Entrepreneurial Financial Performance, Information Processing, and Decision-Making Study

Participants completed an online randomized vignette experiment concerning entrepreneurial financial-performance interpretation and intended business decisions. Each participant was assigned to a hypothetical service, retail, or manufacturing business and to one focal financial indicator: gross margin, operating margin, or return on assets. Within the assigned business and financial-indicator context, each participant reviewed three financial-performance conditions—healthy, mixed or ambiguous, and distressed—in randomized order.

The scenarios varied whether the focal financial indicator was around or above the benchmark for similar firms or below that benchmark and whether performance had improved or declined relative to the prior year. After each vignette, participants reported how they interpreted the financial signal and what business action they would be most likely to take.

The survey also collected nonrandomized measures concerning promotion and prevention focus, entrepreneurial experience and venture history, entrepreneurship training, accounting and finance training, financial-ratio familiarity and use, artificial-intelligence-assisted business analysis, performance-decision self-efficacy, and demographic characteristics. These participant characteristics were measured variables and were not randomized interventions.

2026-06-26

2026-09-05

The study has two primary outcome families corresponding to two analytically distinct components of the broader research project.

A. Randomized financial-performance vignette outcomes

1. Intended business action following each vignette: continue the current strategy, undertake strategic restructuring, seek or add funding, or exit the venture.

2. Reported likelihood of taking each of the four business actions, with each action rated separately on a seven-point scale.

3. Perceived concern regarding the financial-performance pattern, measured on a seven-category ordered scale.

4. Perceived clarity versus ambiguity of the financial signal, measured on a seven-point ordered scale.

B. Artificial-intelligence and standardized accounting-information outcomes

5. Reported use of gross margin, operating margin, and return on assets among respondents with prior business-starting or co-founding experience. Outcomes include use of each individual ratio, use of at least one of the three ratios, and the total number of the three ratios used.

6. Agreement that access to artificial-intelligence tools makes the respondent more likely to use financial ratios to understand business performance, measured on a seven-point agreement scale.

7. The respondent’s stated orientation toward how artificial intelligence affects the use of financial ratios. The response categories distinguish: artificial intelligence helping explain ratios; artificial intelligence helping calculate, compare, or interpret ratios; no change in ratio use; reduced ratio reliance because artificial intelligence provides broader business advice; reduced ratio reliance because the respondent may trust an artificial-intelligence-generated overall explanation; use of ratios to check or verify artificial-intelligence recommendations; and uncertainty regarding the effect.

The randomized vignette outcomes and the artificial-intelligence/accounting-information outcomes are analytically distinct.

For the randomized vignette component, each participant evaluates one healthy, one mixed or ambiguous, and one distressed financial-performance condition. Each participant therefore provides repeated observations for intended action, action likelihood, concern, and perceived ambiguity. Intended action is treated as a nominal four-category outcome. The four action-likelihood ratings are analyzed separately. Concern and ambiguity are treated as ordered seven-point outcomes after the exported Qualtrics values are recoded to preserve the response order displayed to participants.

For the artificial-intelligence and accounting-information component, actual ratio use is measured only for participants who report having started or co-founded at least one business. Respondents without prior venture experience were not asked the actual ratio-use question and will not be coded as nonusers.

Separate indicators will identify reported use of gross margin, operating margin, and return on assets. An any-ratio-use measure will equal one when the respondent reports using at least one of these three ratios and zero when an eligible respondent reports using none of the three. A ratio-count measure will range from zero to three.

Prior use of artificial intelligence to understand or evaluate business performance is a focal explanatory variable rather than the primary dependent variable in the initial artificial-intelligence paper. Business-specific artificial-intelligence experience may include use for a currently owned or managed business, a previously owned or managed business, or another actual business. Experience limited to coursework, training, consulting, or work-related analysis may be retained as a separate category in paper-specific analyses.

The seven-point artificial-intelligence/ratio item measures the extent to which access to artificial intelligence is associated with an increased stated likelihood of using financial ratios. The categorical artificial-intelligence/ratio item is treated as nominal because increased reliance, no change, reduced reliance, verification, and uncertainty are conceptually different configurations rather than points on a single ordered continuum.

For paper-specific analyses, the categories may also be grouped as follows: analytical complementarity, consisting of artificial intelligence explaining, calculating, comparing, or interpreting ratios; substitution, consisting of reduced ratio reliance because of broader artificial-intelligence advice or trust in an overall artificial-intelligence explanation; verification complementarity, consisting of using ratios to check artificial-intelligence recommendations; no change; and uncertainty.

Formal accounting or finance education and training may be examined as a moderator of the relationships between business-specific artificial-intelligence experience and both actual ratio use and artificial-intelligence/ratio orientation. Country, age, and gender may be included as control variables. Promotion focus, prevention focus, and paper-specific regulatory-fit measures may be examined as additional moderators in separately specified analyses.

The artificial-intelligence analyses are observational because prior artificial-intelligence experience, accounting training, regulatory focus, and ratio use were measured rather than randomly assigned. Results involving these variables will therefore be described as associations rather than randomized treatment effects.

Return on equity was not measured. The ratio outcomes in this study are gross margin, operating margin, and return on assets.

1. Preferred initial financing source when seek or add funding is selected: owner capital, debt financing, or outside equity or strategic-partner capital.

2. Financial-signal interpretation measures, including whether the business is classified as around or above versus below the relevant benchmark and whether the participant correctly identifies the vignette’s focal financial ratio.

3. Performance-decision self-efficacy.

4. Promotion-focus and prevention-focus measures and paper-specific regulatory-fit or regulatory-congruence measures.

5. Entrepreneurial experience and venture history, including number of ventures started or co-founded, founder role, venture type, operating status, closure experience and reason, venture duration, and revenue trend.

6. Entrepreneurship education or training and accounting or finance education or training.

7. Familiarity with gross margin, operating margin, and return on assets.

8. Context of prior artificial-intelligence use, including use for a current business, previous business, another business, coursework, training, consulting, or work-related analysis.

9. Demographic and background characteristics, including country of residence, age, gender, educational attainment, race or ethnicity, and employment status.

Financing preference is observed only when a participant selects seek or add funding as the primary intended action for a vignette.

Benchmark classification and focal-ratio identification may be used as interpretation measures, comprehension measures, or secondary outcomes in research examining the processing of accounting information.

Performance-decision self-efficacy is measured using five seven-point confidence items addressing the respondent’s ability to assess financial ratios, judge whether business performance is improving or declining, select an appropriate business response, identify a useful strategic change, and explain or justify a business decision. A paper using this construct will report the final scoring rule, missing-item rule, and reliability assessment.

Promotion focus and prevention focus will be scored as separate motivational-orientation constructs using the applicable item assignments and reverse-scoring rules. A regulatory-fit analysis may construct fit as the theoretically specified congruence between a participant’s regulatory orientation and an action, decision orientation, financial condition, or artificial-intelligence-related information-use configuration. The exact construction and corresponding hypotheses will be stated in the separate analysis plan and paper using that measure.

Regulatory focus and regulatory fit are measured or constructed participant characteristics; they are not randomized interventions. They may be examined as moderators of responses to randomized financial-performance conditions, as moderators of relationships involving artificial-intelligence experience and accounting-information use, or both.

Entrepreneurial experience, training, ratio familiarity, prior ratio use, artificial-intelligence experience, and demographic measures may serve as explanatory variables, moderators, controls, or outcomes depending on the research question addressed by an individual paper. Analyses using these measured characteristics will not be represented as randomized treatment effects.

This umbrella registration documents the full scope of the data collected. Individual papers will identify their own focal dependent variables, hypotheses, sample restrictions, variable constructions, and statistical specifications.

The broader study supports two types of analysis. The randomized component examines how assigned industry, focal financial indicator, financial-performance condition, and presentation order relate to financial judgments and intended business actions. Separate nonrandomized analyses examine relationships among prior artificial-intelligence experience, actual financial-ratio use, beliefs about whether artificial intelligence complements or substitutes for ratio analysis, accounting and finance training, regulatory focus or fit, entrepreneurial experience, and demographic characteristics. Only the assigned vignette conditions are interpreted as randomized treatments.Participants were recruited from multiple countries. Country of residence was not part of the treatment-assignment mechanism. Paper-specific analyses may examine whether experimental responses or measured relationships differ across countries.

The study also measures entrepreneurial and venture characteristics that may be used to define subgroups or entrepreneur profiles in later analyses. Such country-based and attribute-based analyses are observational heterogeneity analyses unless they examine interactions with the randomly assigned vignette conditions.

Not available

Computer-generated randomization was implemented through Qualtrics Survey Flow. An outer Qualtrics Block Randomizer set to “Evenly Present Elements” assigned each participant to one of three hypothetical industry contexts: service, retail, or manufacturing.

Within the assigned industry context, a nested Qualtrics Block Randomizer, also set to “Evenly Present Elements,” assigned each participant to one of 18 focal-ratio-by-presentation-order combinations. These combinations were formed by crossing three focal financial indicators—gross margin, operating margin, and return on assets—with six possible presentation orders of the healthy, ambiguous, and distressed financial-performance conditions.

The procedure generated 54 possible industry-by-ratio-by-order assignment sequences and was designed to maintain approximately balanced assignment across the sequences. Assignment was performed entirely by the Qualtrics computer system. No participant was manually assigned by the investigator.

Country of residence, entrepreneurial background, venture experience, regulatory orientation, accounting or finance training, artificial-intelligence experience, and other participant attributes were not used as treatment-assignment units. These measured characteristics may be used in paper-specific subgroup, moderation, or heterogeneity analyses.

The unit of randomization was the individual survey participant. Each participant was individually assigned to an industry context, a focal financial ratio, and one of six presentation orders for the financial-performance conditions.

Financial-performance signal was a repeated within-participant factor because every participant evaluated one healthy, one ambiguous, and one distressed vignette. The three financial-performance signals were therefore not assigned to separate groups of participants.

Country of residence and measured entrepreneurial attributes were not randomization units. Later papers may group participants by country or by measured characteristics for subgroup, moderation, profile, or heterogeneity analyses, but those groupings do not represent randomized treatment assignment.

No

Not applicable as a group-level cluster count because treatment was individually randomized. The planned sample consisted of 800 individual
randomization units—800 survey participants. There were no country-level, firm-level, session-level, or other group-level treatment-assignment clusters.

800 individual survey participants. Each participant was expected to provide responses to three financial-performance vignettes, producing up to 2,400 participant-vignette observations. The planned number of independently recruited participant-level observations was 800. The participant-vignette observations are repeated observations nested within participants. Country groups and entrepreneur-attribute groups are classifications of the same participants and do not increase the total number of independently recruited observations.

The planned sample of 800 participants was to be allocated approximately evenly across 54 randomized industry-by-ratio-by-presentation-order
sequences. The 54 sequences were formed from three industry contexts—service, retail, and manufacturing—three focal financial indicators—
gross margin, operating margin, and return on assets—and six presentation orders of the healthy, ambiguous, and distressed financial-performance conditions. Approximately 14 to 15 participants were expected in each complete industry-by-ratio-by-order sequence.

When presentation order is collapsed, the design contains nine industry-by-ratio arms, with approximately 88 to 89 participants planned for each arm:

1. Service × Gross Margin
2. Service × Operating Margin
3. Service × Return on Assets
4. Retail × Gross Margin
5. Retail × Operating Margin
6. Retail × Return on Assets
7. Manufacturing × Gross Margin
8. Manufacturing × Operating Margin
9. Manufacturing × Return on Assets

Every participant evaluated all three financial-performance signals. Healthy, ambiguous, and distressed were therefore repeated within-participant conditions rather than separate participant treatment arms. The study did not include an untreated control arm.

Country of residence was not a randomized treatment arm. Paper-specific analyses may compare participants across countries, include country indicators or country-level effects, or examine whether relationships vary across countries. These comparisons will be identified as country-level heterogeneity or observational subgroup analyses rather than randomized treatment comparisons.

Entrepreneurial background, venture experience, venture status, entrepreneurship training, accounting or finance training, regulatory orientation, artificial-intelligence experience, and related participant attributes were also not randomized treatment arms. These variables may define paper-specific subgroups, moderators, or empirically derived entrepreneur profiles. Results based on these classifications will be distinguished from the randomized vignette comparisons.

An a priori power analysis informed the planned sample of approximately 800 participants, with recruitment distributed across nine countries. Calculations assume a two-sided significance level of 0.05 and 90% statistical power. Treatment assignment occurs at the individual-participant level; therefore, no cluster-randomization design effect is applied. Repeated responses to the three financial-performance conditions are nested within participants. For pooled within-participant comparisons using approximately 800 participants, the design provides 90% power to detect a standardized mean difference of approximately 0.115 standard deviations, or 11.5% of one standard deviation. For pooled comparisons across three approximately equal groups of about 267 participants each, the minimum detectable omnibus effect is approximately Cohen's f = 0.126. A pairwise comparison between two groups of approximately 267 participants each provides 90% power to detect approximately Cohen's d = 0.281, or 28.1% of one standard deviation. For a binary outcome with a baseline probability near 50%, this corresponds to an approximate detectable difference of 13.8 percentage points between two groups. For the four-category intended-business-action outcome—continue the current strategy, undertake strategic restructuring, seek or add funding, or exit the venture—a three-group comparison provides approximately 90% power to detect an association of Cramér's V = 0.104. Country may be analyzed in theoretically defined groups, including two-, three-, or four-group classifications. For a roughly balanced two-group classification, the minimum detectable pairwise effect is approximately 0.23 standard deviations. For three approximately equal country groups of about 267 participants each, the minimum detectable pairwise effect is approximately 0.28 standard deviations. For four approximately equal groups of about 200 participants each, the minimum detectable pairwise effect is approximately 0.33 standard deviations. Where multiple pairwise country-group comparisons are treated as co-primary, multiplicity-adjusted inference will be used. These calculations support the principal pooled experimental analyses and broad country-group comparisons. Analyses involving artificial-intelligence experience, regulatory focus or fit, accounting or finance training, entrepreneurial characteristics, or outcomes restricted to participants with venture experience may have smaller effective samples. Paper-specific analyses of these relationships will therefore report sensitivity or minimum-detectable -effect calculations based on the final eligible sample, outcome distribution, and statistical model.