Auditing Knowledge, Professional Skepticism, Professional Ethics, and a Two-Item Auditor Career-Interest Proxy: A Cross-Sectional Study of Accounting Students in Serang City

Abstract

Prior audit-career research has emphasized professional familiarity, career perceptions, and general accounting-career choice, whereas professional skepticism and ethics have been studied mainly among practicing auditors. This study therefore examines whether auditing knowledge, professional skepticism, and professional ethics are associated with a narrowly operationalized auditor career-interest outcome among accounting students in Serang City. A quantitative cross-sectional survey was conducted with 55 students from three universities who had completed advanced auditing coursework and were recruited through incidental non-probability sampling. PLS-SEM indicated that auditing knowledge was positively associated with the two-item career-interest proxy (β = 0.445, t = 2.199, p = 0.028), whereas professional skepticism (β = 0.128, p = 0.563) and professional ethics (β = 0.177, p = 0.243) were not statistically significant; R² was 0.442. Supplementary diagnostics nevertheless require caution: the reconstructed SRMR was 0.1017, Harman's first component explained 51.6% of item variance, and the X2-X3 HTMT bootstrap interval [0.593, 1.024] included 1.00. A full- collinearity VIF check did not show severe collinearity, but it does not eliminate common-method concerns. Accordingly, the results are presented as provisional, sample-specific evidence rather than a comprehensive test of career intention. Practically, the findings support further evaluation of audit-knowledge exposure using stronger, validated multi-item career-intention measures and more rigorous sampling and method-bias controls.

Keywords: auditing knowledge; professional skepticism; professional ethics; auditor career-interest proxy; PLS-SEM.

Introduction

The audit profession is a crucial component of financial accountability because independent assurance supports confidence in financial reporting. Regeneration of the Indonesian public accounting profession has been a documented concern. The Ministry of Finance reported 1,368 active public accountants as of 21 January 2019; 26.1% were older than 60 years and 25.6% were aged 50-59, while new public-accountant licenses declined from 212 in 2017 to 86 in 2018 (Pusat Pembinaan Profesi Keuangan, 2019). A separate Ministry of Finance communication in 2020 similarly noted that more than half of registered public accountants were over 50. These official figures are historical rather than current headcounts, but they provide concrete evidence that professional regeneration has been a longstanding policy concern and make accounting students an important potential pipeline for Indonesian public accounting firms. International literature indicates that students' perceptions of auditing are shaped by professional familiarity, educational exposure, stereotypes, social identity, and access to career information. Thuy et al. (2022) found that proximity and knowledge were related to students' perceptions of auditors, while Pelzer & Nkansa (2022) emphasized social Research on accounting career choice has generally emphasized financial rewards, professional reputation, certification, family and peer influence, career prospects, and perceived job fit (Bhatt & Khan, 2023; Ilias et al., 2022; Kong et al., 2020; Marsintauli et al., 2022). Studies focused specifically on auditing also show that professional familiarity and perceived attractiveness matter (Graschitz et al., 2023; Thuy et al., 2022; Wulan et al., 2023). However, three gaps remain. Theoretically, TPB-based career studies rarely distinguish capability-related audit knowledge from professional-value constructs within one model. Empirically, Indonesian studies more often examine accounting or public- accounting careers broadly rather than the joint direct effects of auditing knowledge, skepticism, and ethics. Contextually, evidence from students in Serang City is scarce, despite differences in university exposure and access to public accounting firms compared with larger metropolitan centers. Auditing knowledge is central because students who understand standards, procedures, evidence, audit risk, materiality, and reporting may perceive the work as more manageable. Within TPB, this construct is most closely related to perceived behavioral control: knowledge can strengthen students' assessment that they possess, or can acquire, the capability required to perform audit tasks (Ajzen, 2020). This study does not directly measure self-efficacy or perceived behavioral control; rather, it treats auditing knowledge as an educational antecedent that is theoretically expected to support those perceptions (Rietz, 2018). Professional skepticism is a core audit mindset involving a questioning attitude, suspension of judgment, and critical evaluation of evidence. It may increase perceived person-profession fit among students who value analytical and investigative work. However, skepticism may also remain an abstract professional norm when students have limited engagement experience. Therefore, its relationship with career interest is conceptually plausible but not necessarily immediate or direct. Professional ethics provide the moral foundation of auditing through integrity, objectivity, confidentiality, competence, and responsibility to the public interest. In TPB terms, ethics should not be equated automatically with subjective norms, which refer to perceived social pressure. In this study, professional ethics are treated primarily as a moral- norm and value-based extension of TPB that may shape attitudes toward the profession. Ethical awareness may enhance the perceived dignity and social contribution of auditing, but it may also be broadly endorsed across occupations and therefore have limited power to differentiate career choices. Digital audit technologies provide relevant background because they are changing the competencies expected of new auditors, but digital competence is not a tested construct in the present model. Accordingly, technology is treated only as contextual motivation and as a potential extension for future research rather than as an explanation of the present structural paths (Felski, 2023; Zhang et al., 2024). Accordingly, this study examines the direct associations of auditing knowledge, professional skepticism, and professional ethics with auditor career interest among accounting students in three universities in Serang City. Its novelty is primarily theoretical and contextual: it integrates one capability-related educational construct and two audit- specific professional-value constructs in a TPB-informed model, while testing that model among prospective rather than practicing auditors. Practically, the study evaluates whether classroom-based knowledge and professional-value education are equally salient for students' career intention. The restricted local sample is used to generate context-specific evidence rather than broad population estimates. Theoretical Review and Hypothesis Development

The Theory of Planned Behavior and Auditor Career Interest

The Theory of Planned Behavior explains behavioral intention through attitudes toward the behavior, subjective norms, and perceived behavioral control (Ajzen, 2020). Applied to career choice, students are more likely to consider auditing when they view the profession positively, perceive relevant social or professional support, and believe they can perform the required work. This study uses TPB as an organizing framework rather than claiming that the three predictors are exact substitutes for the original TPB dimensions. Auditing knowledge is positioned as a capability-related antecedent of perceived behavioral control because technical understanding may make audit tasks appear learnable and manageable. Professional skepticism is positioned as an audit- specific mindset that may contribute to attitudes and perceived person-profession fit. Professional ethics are treated as a moral-norm extension that may shape the perceived social value and legitimacy of the profession, rather than as a direct measure of subjective norms. This more cautious mapping recognizes that perceived behavioral control, attitudes, subjective norms, and professional identity were not measured as separate mediators in the present model.

Auditing Knowledge and Interest in Becoming an Auditor

Auditing knowledge includes an understanding of auditing standards, audit objectives, risk assessment, materiality, audit procedures, audit evidence, and reporting. In accounting education, this knowledge is acquired through lectures, practicums, case studies, simulations, and internships. Students with greater audit knowledge tend to have a more realistic view of the auditor's work, thereby reducing the perception of career uncertainty. Knowledge and familiarity can reduce uncertainty about professional work and improve career attractiveness (Ali et al., 2024; Thuy et al., 2022). Indonesian evidence also suggests that professional knowledge, learning exposure, and career information are associated with interest in public accounting (Setiawan et al., 2025; Welly et al., 2022). Students who understand audit procedures and evidence may be better able to imagine themselves completing audit assignments and developing the required competence. Therefore, H1 is formulated as follows: H1: Auditing knowledge is positively associated with accounting students' interest in becoming auditors.

Professional Skepticism and Interest in Becoming an Auditor

Professional skepticism is a mental attitude that includes questioning, suspension of judgment, seeking knowledge, and critical evaluation of evidence. In audit practice, skepticism is necessary to assess the risk of material misstatement, respond to indications of fraud, and prevent auditors from relying too heavily on management representations. Previous research has shown that professional skepticism is relevant to audit judgment and audit quality (Fawziah et al., 2023; Savira et al., 2021; Sumartono et al., 2022). For students, skepticism may signal alignment with analytical and investigative work and thus strengthen perceived job fit. Conversely, because skepticism is usually developed and understood through ambiguous evidence, client interaction, and engagement experience, students may endorse it as an important professional norm without using it as a career-selection criterion. The direct association is

Professional Ethics and Interest in Becoming an Auditor

The professional ethics of an auditor encompass integrity, objectivity, professional competence, confidentiality, and professional conduct. Ethics are the foundation of public trust because auditors work to provide assurance to stakeholders. Research on audit quality shows that auditor ethics is related to audit quality, audit judgment, and the ability to maintain independence (Parluhutan et al., 2022; Rizky & Astuti, 2023; Surya et al., 2021). Ethical awareness may improve attitudes toward auditing by emphasizing integrity, public responsibility, and professional legitimacy. However, ethics may also be perceived as a general obligation across professions, operate indirectly through professional identity, or show limited variance because respondents tend to endorse socially desirable ethical statements. Thus, a positive direct relationship is plausible, but it may not emerge among students with limited professional exposure. Based on the expected value-fit mechanism, H3 is formulated as follows: H3: Professional ethics are positively associated with accounting students' interest in becoming auditors. Methods

Research Type

This study used a quantitative, cross-sectional survey design to estimate associations among auditing knowledge, professional skepticism, professional ethics, and a narrowly operationalized two-item proxy for auditor career interest. The three predictor constructs were modeled reflectively. The outcome was retained because only two career-related items were collected in the original survey; it is therefore not presented as a comprehensive or validated career-intention scale. Because all variables were measured at one point in time using self-reported perceptions, the design does not establish temporal order or causality. PLS-SEM was used to estimate the reported measurement and structural relationships within the available sample, but all structural findings are treated as provisional and sample-specific. The measurement model was evaluated using item loadings, average variance extracted (AVE), Cronbach's alpha, composite reliability, the Fornell-Larcker criterion, and the heterotrait-monotrait ratio (HTMT). The structural model was Table 2. Operational Definition of Variables Variable Operational definition Auditing Knowledge (X1) Understanding of audit standards, p materiality, and evidence. Critical mental attitude in evaluating delaying judgment until sufficient evide Professional Skepticism (X2) Moral and professional principles gu maintaining public trust. Professional Ethics (X3) Auditor Career Interest Proxy (Y) Internal tendency to choose auditor as perceived fit and career prospects. Source: Research instrument developed from the thesis data and evaluated using R², adjusted R², path coefficients, inner VIF, f² effect sizes, bootstrap inference, Q²_predict, and SRMR. A supplementary reanalysis used 5,000 percentile bootstrap resamples with a two-tailed 5% significance level. The reconstructed SRMR was calculated as the square root of the mean squared off-diagonal standardized residuals between the empirical indicator-correlation matrix and the model-implied correlation matrix generated from the retained outer loadings and construct-correlation matrix; the exact value was 0.1017. Because 0.1017 exceeds the commonly cited 0.08 benchmark, model-data discrepancy remains material and no claim of satisfactory global fit is made (Hair et al., 2020). Theory-driven results were therefore supplemented with reduced-model SRMR checks and a composite-score bootstrap regression as robustness analyses rather than post hoc model optimization.

Population and Sample

The operational sampling frame consisted of accounting students who were reachable through the three participating universities in Serang City during the data-collection period. Eligible students were approached through printed- questionnaire distribution and circulation of a Google Forms link via WhatsApp. Inclusion criteria were: (1) enrollment as an accounting student at Universitas Serang Raya, Universitas Pamulang Serang Campus, or Universitas Sultan Ageng Tirtayasa; (2) completion of an audit concentration or equivalent advanced audit coursework; and (3) voluntary submission of a usable questionnaire. Responses not meeting these criteria or incomplete for study variables were excluded. Incidental non-probability sampling was used because recruitment depended on eligible students who were accessible during the collection period. A total of 55 usable Table 1. Distribution of Respondents Category Group Frequency Percentage University Universitas Serang Raya 28 50.9% Universitas Pamulang Kampus Serang 21 38.2% University Universitas Sultan Ageng Tirtayasa 6 10.9% University Gender Female 48 87.3% Gender Male 7 12.7% Total All respondents 55 100% Source: Primary data processed, 2026. on Indicators Scale Likert 1-5 X1.1 standards; X1.2 risk-materiality- procedure; X1.3 practicum experience; X1.4 audit evidence analysis. , procedures, risks, Likert 1-5 X2.1 suspension of judgment; X2.2 search for knowledge; X2.3 self-esteem. ng information and dence exists. Likert 1-5 X3.1 public interest; X3.2 integrity; X3.3 objectivity; X3.4 confidentiality; X3.5 technical standards. guiding auditors in Likert 1-5 as a career based on Y1 challenging work; Y2 career path. and recent audit career literature.

Instrumentation or Tools

The instrument was a closed-ended questionnaire administered in Indonesian using a five-point Likert scale from strongly disagree to strongly agree. The items were developed for the original thesis-based study and were conceptually informed by audit education, professional skepticism, professional ethics, and accounting-career literature, but the available records do not provide verifiable documentation that the items were adopted or systematically adapted from established validated scales (Table 2). The records also do not document a formal pilot test or expert content-validity assessment before data collection. Therefore, pre-use content validity cannot be established retrospectively and is treated as a major measurement limitation, not merely a reporting omission. The questionnaire was originally developed and administered in Indonesian, so English-to- Indonesian translation or back-translation was not part of data collection. The outcome contains only two items concerning challenging work and career path. Measurement literature generally favors multi-item measures for broader latent constructs because they provide wider content coverage and more stable construct representation; shortened measures require strong conceptual justification (Diamantopoulos et al., 2012; Hair et al., 2020). Because these two items primarily reflect perceived job attractiveness and career structure rather than explicit plans, willingness, or likelihood of entering auditing, the outcome is retained only as a narrow auditor career-interest proxy. Complete item wording is provided in Appendix A.

Data Collection Procedure and Research Ethics

Data were collected from 29 April to 8 May 2026 through printed questionnaires and Google Forms distributed via WhatsApp. Before participation, respondents received information about the study purpose, voluntary participation, confidentiality, anonymity, and their right to decline or withdraw before submitting the questionnaire. No directly identifying information was included in the analytical dataset. The study was not submitted to an institutional ethics committee before data collection, and no formal ethics approval, waiver, or separate institutional permission letter was obtained. Accordingly, the manuscript does not claim institutional ethical approval or exemption. The study nevertheless applied basic participant-protection procedures through voluntary participation, prior study information, anonymity, confidentiality, and analysis of de-identified responses. No separate formal data-protection protocol was used beyond these de-identification and confidentiality procedures. This absence of prospective institutional ethics Table 3.

Descriptive Statistics

of Research Variables Variable N M Auditing Knowledge (X1) 55 15 Professional Skepticism (X2) 55 12 Professional Ethics (X3) 55 21 Auditor Career Interest Proxy (Y) 55 7.8 Source: SmartPLS 4 output, 2026. review or formal waiver is reported transparently as a research- governance limitation and remains subject to the journal's ethics requirements for publication.

Data Analysis

Analysis proceeded in two stages. Descriptive statistics summarized respondent characteristics and construct scores. PLS-SEM was then used to assess the reported measurement and structural relationships. Convergent validity was evaluated using outer loadings (>0.70) and AVE (>0.50), internal consistency using Cronbach's alpha and composite reliability (>0.70), and discriminant validity using Fornell-Larcker and HTMT. Structural assessment included R², adjusted R², inner VIF, f², path coefficients, bootstrap confidence intervals, Q²_predict, and SRMR. Because all variables were collected from the same respondents in one questionnaire, common method bias (CMB) was evaluated using two diagnostics: Harman's single-factor test and a full-collinearity VIF assessment. Following (Kock, 2015), full-collinearity VIF values below 3.3 were treated only as evidence against severe collinearity-related method bias, not as proof that CMB was absent. A marker-variable adjustment could not be implemented because no theoretically unrelated marker variable was included in the original questionnaire (Lindell & Whitney, 2001). Result and Discussion

Descriptive Statistics

Table 3 reports summed construct scores. To make constructs with different indicator counts comparable, mean scores per item were also calculated: auditing knowledge = 3.91, professional skepticism = 4.24, professional ethics = 4.31, and auditor career interest = 3.92. Using a descriptive interpretation in which 3.41-4.20 indicates agreement and 4.21-5.00 indicates strong agreement, auditing knowledge and career interest fall in the agreement range, while skepticism and ethics fall in the strong-agreement range. These statistics describe response tendencies only and are not used to infer structural relationships.

Measurement Model

The reported predictor measurement models met the stated convergent-validity and reliability thresholds: outer loadings exceeded 0.70, AVE values exceeded 0.50, and Cronbach's alpha and composite reliability exceeded 0.70. The two career-related items also loaded strongly in the original SmartPLS specification, but high loadings and reliability do not resolve construct underrepresentation. The items 'challenging work' and 'career path' capture only a narrow portion of what a comprehensive behavioral career-intention construct would normally include. Consistent with measurement guidance favoring adequate content coverage for broad constructs (Diamantopoulos et al., 2012; Hair et al., 2020), the outcome is not interpreted as a validated latent career-intention scale. Instead, all structural findings involving Y are framed as provisional evidence for a narrowly operationalized two-item career-interest proxy. Figure 1 presents the SmartPLS 4 outer model for the measurement model evaluation. Mean Std. Dev. Minimum Maximum 15.655 4.133 4 20 12.709 2.492 4 15 21.527 4.185 7 25 7.836 1.983 3 10 Figure 1 shows that the retained indicators had loadings above 0.70. This evidence supports indicator reliability but does not establish the substantive importance of any individual item for career interest. Accordingly, the loadings are interpreted only as measurement evidence; substantive conclusions are based on structural path estimates and additional effect-size diagnostics (Table 4). Discriminant validity requires particular caution. The HTMT point estimate between professional skepticism and professional ethics was 0.884, below 0.90 but above the more conservative 0.85 guideline (Henseler et al., 2015). A 5,000-resample percentile bootstrap reanalysis yielded a 95% interval of approximately [0.593, 1.024], so discriminant validity between X2 and X3 is not established. Indicator-level inspection also showed notable cross-construct correlations, especially X2.1 with X3.4 (r = 0.713), X3.3 (r = 0.698), and X3.5 (r = 0.681) (Table 5), suggesting that students may perceive delayed judgment, objectivity/fairness, confidentiality, and procedural conduct as overlapping aspects of desirable professional behavior. A post hoc deletion sensitivity check reduced HTMT to 0.776 when X2.1 was omitted and to 0.841 when X3.1 was omitted. Neither deletion was adopted: removing X2.1 would leave skepticism with only two indicators, while deleting X3.1 solely to improve a threshold lacks a priori conceptual justification. The original constructs are therefore retained for transparency, but their Figure 1. Outer Model Schematic Source: SmartPLS 4 Output, 2026. Table 4.

Measurement Model

Evaluation Construct Outer loading range AVE Auditing Knowledge (X1) 0.747-0.868 0.694 Professional Skepticism (X2) 0.773-0.914 0.708 Professional Ethics (X3) 0.754-0.906 0.706 Auditor Career Interest Proxy (Y) 0.905-0.927 0.839 Source: SmartPLS 4 output, 2026. separate structural paths are treated as empirically unresolved and interpreted only provisionally.

Structural Model and Hypothesis Testing

The R² value for the career-interest proxy was 0.442 (adjusted R² = 0.409), indicating moderate in-sample explanatory power. Inner VIF values were 1.880 for auditing knowledge, 3.045 for professional skepticism, and 2.272 for professional ethics. Effect sizes were f² = 0.186 for auditing knowledge, 0.010 for skepticism, and 0.024 for ethics. Repeated 10-fold cross-validation yielded Q²_predict values of 0.276 for Y1 and 0.184 for Y2. The exact reconstructed SRMR was 0.1017, above the 0.08 benchmark. Sensitivity checks did not show that simple theory-reducing respecification solved the discrepancy: SRMR was 0.0909 when skepticism was omitted, 0.1119 when ethics was omitted, and 0.1049 in a knowledge- only model. Because none reached 0.08, the original theory- driven specification was retained rather than selecting a model solely to improve fit. Structural conclusions are consequently treated as provisional. As an additional robustness check independent of the latent-score specification, standardized composite means were regressed on the two-item outcome proxy and evaluated with 5,000 nonparametric bootstrap resamples. The composite- score coefficients were 0.448 for auditing knowledge, 0.086 Composite E Cronbach reliability Decision alpha 94 0.852 0.900 Valid and reliable 08 0.795 0.879 Valid and reliable 06 0.895 0.923 Valid and reliable 39 0.809 0.912 Valid and reliable Figure 2 presents the path coefficients and significance testing results generated through the bootstrapping procedure in SmartPLS 4. The original SmartPLS bootstrapping results showed that auditing knowledge was positively associated with auditor career interest (β = 0.445, t = 2.199, p = 0.028). Professional skepticism had a positive but non-significant direct path (β = 0.128, t = 0.578, p = 0.563), as did professional ethics (β = 0.177, t = 1.168, p = 0.243) (Table 6). In the supplementary 5,000-resample percentile bootstrap reanalysis, the approximate 95% confidence intervals were [0.123, 0.920] for auditing knowledge, [-0.343, 0.562] for skepticism, and [- 0.151, 0.491] for ethics (Table 8). Thus, only H1 was supported. The non-significant results for H2 and H3 indicate that direct associations were not detected in this sample; they do not demonstrate that skepticism or ethics are unimportant to audit practice or career development. The full-item-level outer loadings for all indicators are presented in Table 7, with values ranging from 0.747 to 0.927, all exceeding the 0.70 threshold. Common method bias diagnostics yielded mixed evidence. Harman's single-factor analysis of all 14 questionnaire items showed that the first unrotated component accounted for 51.6% of total standardized item Figure 2. Path Coefficient and Significance Test (Bootstrapping) Source: SmartPLS 4 output, 2026. Table 6. Hypothesis Testing Results Hypothesis Path Original p- values t- statistics sample H1 X1 -> Y 0.445 2.199 0.028 S H2 X2 -> Y 0.128 0.578 0.563 N s H3 X3 -> Y 0.177 1.168 0.243 N s Source: SmartPLS 4 output, 2026. variance, slightly above the conventional 50% screening threshold; therefore, common method bias cannot be dismissed and may inflate or distort observed relationships. As an additional full-collinearity assessment, each construct score was regressed on the remaining construct scores. The resulting VIFs were 2.181 for auditing knowledge, 3.019 for professional skepticism, 2.311 for professional ethics, and 1.720 for the career-interest proxy, all below the 3.3 threshold proposed by (Kock, 2015). This second diagnostic does not show severe full- collinearity contamination, but it does not negate the Harman result or the single-source, single-time design. Because no marker variable or temporal/source separation was available, CMB remains a material threat to inference rather than a resolved issue.

Interpretation of Key Findings

H1 was supported in the original PLS specification: auditing knowledge was positively associated with the two-item career- interest proxy (β = 0.445, p = 0.028; f² = 0.186). The Table 5. HTMT Discriminant Validity Construct X1 X2 X3 Y X1 X2 0.822 X3 0.596 0.884 Y 0.737 0.681 0.588 Source: SmartPLS 4 output, 2026. g) Decision Interpretation Supported Positive direct association with the two-item proxy supported in this sample. Not supported Direct association not supported; null result interpreted cautiously. Not supported Direct association not supported; null result interpreted cautiously. The result has practical relevance because auditing has substantial technical and procedural demands. Classroom exposure to risk assessment, internal control, evidence evaluation, working papers, and reporting may allow students to form a more concrete view of professional work. The structural coefficient, together with the effect size to be reported, should be used to assess practical importance. Outer loadings are not used to infer which learning experience causes career interest, because they indicate measurement quality rather than substantive impact. H2 was not supported: no statistically significant direct association was detected between professional skepticism and the two-item career-interest proxy (β = 0.128, p = 0.563; f² = 0.010). This null result cannot establish absence of a relationship. The sample has limited power for small effects, the outcome is narrowly operationalized, CMB cannot be excluded, and X2 is not empirically distinct from X3 under the bootstrap HTMT criterion. The separate skepticism path should therefore be interpreted as indeterminate rather than evidence that skepticism is unimportant to career development. H3 was not supported: professional ethics did not show a statistically significant direct association with the two-item career-interest proxy (β = 0.177, p = 0.243; f² = 0.024). As with H2, the null result is inconclusive because content validity is limited, the sample may not detect small effects, common-method influences remain possible, and the X2-X3 distinction is empirically unresolved. Ethics may also operate through professional identity or value fit rather than through a direct path, but those mechanisms were not tested. Taken together, the analyses provide provisional evidence that auditing knowledge is positively related to the narrowly defined career-interest proxy in this sample, while separate direct associations for skepticism and ethics were not established. The pattern should not be generalized to a comprehensive model of auditor career intention because the outcome contains only two job-characteristic-based items, the sample is small and non-probabilistic, nonresponse bias cannot be assessed, SRMR remains elevated, common method bias remains possible, and skepticism and ethics are not empirically distinct under the HTMT bootstrap test.

Comparison with Previous Studies

The provisional knowledge finding is directionally consistent with international research linking audit-career perceptions to professional proximity, knowledge, and perceived attractiveness (Amondarain et al., 2023; Graschitz et al., 2023; Thuy et al., 2022). However, because the present dependent measure is only a two-item career-interest proxy, the result should not be interpreted as demonstrating the same construct as validated behavioral career-intention scales used elsewhere. The result is also compatible with accounting-career research showing associations with perceived benefits, career prospects, personal capability, and educational exposure (Kong et al., 2020; Marsintauli et al., 2022; Srirejeki et al., 2019). In the present data, auditing knowledge has the largest reported structural coefficient, but the model's elevated SRMR and measurement limitations prevent a firm conclusion that competence-based exposure is more influential than professional-value variables in the wider student population. The non-significant skepticism and ethics paths differ from studies of practicing auditors, where these constructs are often associated with audit judgment, quality, and professional behavior. That comparison should be made cautiously: the present X2-X3 discriminant-validity problem means that their separate coefficients may be unstable or difficult to interpret, while students also have less engagement experience than practicing auditors. The current study therefore treats these null paths as unresolved rather than as evidence that professional values do not matter. The digital audit literature also helps interpret these findings. Felski (2023), Damerji & Salimi (2021) and Eilifsen et al. (2020) show that audit education must keep pace with technology and analytics. If students associate audit with outdated manual checking, they may not see the profession as attractive. Conversely, if they encounter audit analytics, data visualization, and AI-supported audit tools, auditing knowledge may become more engaging and contemporary. Thus, integrating digital audit content may strengthen the effect of knowledge and gradually make skepticism and ethics more meaningful through data-driven cases.

Practical Implications

The findings support specific curriculum actions. Accounting programs can establish audit case laboratories in which students complete risk assessment, control evaluation, substantive testing, and working-paper documentation. Audit analytics modules can require students to use spreadsheets or specialized tools to identify anomalies and visualize evidence. Skepticism can be assessed through evidence-conflict cases, delayed-judgment exercises, and mock client interviews, while ethics can be assessed through dilemma analysis, Table 7. Full Item-Level Outer Loadings Indicator Outer loading X1.1 0.866 X1.2 0.844 X1.3 0.747 X1.4 0.868 X2.1 0.832 X2.2 0.914 X2.3 0.773 X3.1 0.792 X3.2 0.754 X3.3 0.875 X3.4 0.864 X3.5 0.906 Y1 0.927 Y2 0.905 Source: SmartPLS 4 output, 2026. Table 8. Supplementary Structural Diagnostics 95% bootstrap Predictor Inner VIF f² Effect-size interpretation CI for path Auditing knowledge 1.880 0.186 Medium [0.123, 0.920] Profession al skepticism 3.045 0.010 Negligible [-0.343, 0.562] Profession al ethics 2.272 0.024 Small [-0.151, 0.491] Source: SmartPLS 4 output, 2026. Collaboration with public accounting firms, professional associations, and alumni auditors can also improve professional proximity. Structured internships, mentoring, practitioner-led case reviews, and audit-technology demonstrations may clarify the nature of current audit work. Such interventions should be evaluated using pre-post measures of audit self-efficacy, professional identity, and career intention rather than relying only on satisfaction surveys.

Limitations and Cautions

This study has several substantial validity limitations. First, incidental sampling, the small sample, unequal representation across universities, and the predominance of female respondents limit external validity and may create selection/composition bias. Recruitment used printed questionnaires and WhatsApp-distributed Google Forms, but no complete invitation denominator or nonrespondent record was retained; therefore, the response rate and nonresponse bias cannot be assessed. The sensitivity analysis also indicates limited power for small effects. Second, the instrument lacks documented pre-use validation: the available records do not verify adoption or systematic adaptation from validated item-level scales, formal pilot testing, or expert content validation. This materially weakens content-validity claims and constrains interpretation of all structural paths. Third, all variables were collected from the same respondents at the same time. Harman's first component explained 51.6% of variance, whereas full-collinearity VIFs were below 3.3; these mixed diagnostics mean CMB cannot be ruled out. Fourth, the cross-sectional design cannot establish temporal order or causality. Fifth, the outcome consists of only two items on challenging work and career path and therefore represents a narrow career-interest proxy rather than a validated behavioral career-intention construct. Sixth, discriminant validity between professional skepticism and professional ethics is not established because the HTMT bootstrap interval includes 1.00; indicator-overlap sensitivity checks did not justify a defensible post hoc deletion strategy. Seventh, the exact reconstructed SRMR of 0.1017 exceeds 0.08, and reduced-model checks also remained above 0.08, indicating residual model-data discrepancy. Finally, the study was not submitted for institutional ethical review before data collection and did not receive a formal ethics waiver or separate institutional permission letter. Although voluntary participation, anonymity, confidentiality, and de-identified analysis were applied, the absence of prospective institutional ethics oversight is an important research- governance limitation that must be considered by the journal in determining compliance with its publication ethics requirements.

Recommendations for Future Research

Future research should use probability-based or stratified recruitment where feasible, document the number invited so a response rate can be reported, and determine sample size References Ajzen, I. (2020). The theory of planned behavior: Frequently asked questions. Human Behavior and Emerging Technologies, 2(4), 314–324. https://doi.org/10.1002/hbe2.195 Ali, N. A. M., Bustamam, K. S., Saidin, A., Adanan, S. A., & Sani, A. A. (2024). An exploration of accounting student’s perspective on factors influencing a priori through power analysis. Most importantly, auditor career intention should be measured with at least 3-5 validated intention items that directly assess willingness, plans, and likelihood of pursuing an audit career rather than relying on job- characteristic perceptions. The skepticism and ethics measures should also be refined and pretested to establish clearer discriminant validity. Future studies should incorporate procedural and statistical controls for common method bias, including temporal or source separation where feasible and a theoretically unrelated marker variable that permits marker- variable adjustment. Longitudinal and mixed-method designs could test whether auditing knowledge affects career intention through self-efficacy or perceived behavioral control, and whether professional skepticism and ethics operate indirectly through professional identity, value fit, or practical audit exposure. Conclusion This study examined associations between auditing knowledge, professional skepticism, professional ethics, and a narrowly operationalized two-item auditor career-interest proxy among 55 accounting students from three universities in Serang City. In the original PLS model, auditing knowledge was the only statistically significant direct predictor (β = 0.445, p = 0.028), and a supplementary composite-score bootstrap check showed the same qualitative pattern. These results are provisional and sample-specific rather than a comprehensive test of career intention. Inference is materially constrained by incidental sampling and unknown response rate, undocumented pre-use instrument validation, the two-item outcome proxy, possible common method bias, an SRMR of 0.1017, and unresolved discriminant validity between skepticism and ethics. Accordingly, the study's contribution is primarily diagnostic and exploratory: it suggests that audit knowledge deserves further investigation while also identifying measurement and design conditions that must be strengthened before broader claims are made. Future research should use validated multi- item career-intention scales, prevalidated skepticism and ethics measures, larger probability-based samples with documented recruitment denominators, prospective ethics oversight, stronger procedural and statistical common-method controls, and model specifications that demonstrate adequate discriminant validity and residual fit. Author contributions Denny Putri Hapsari contributed to conceptualization, data collection, data processing, and initial manuscript development. Denny Kurnia contributed to supervision, theoretical refinement, methodology review, interpretation of findings, and final manuscript editing. Both authors approved the final version of the manuscript. Acknowledgements The authors thank the participating accounting students and the academic units that facilitated non-identifying access to respondents. No institution influenced the design, analysis, interpretation, or reporting of the study. audit career choice. Information Management and Business Review, 16(3), 407–419. https://doi.org/10.22610/imbr.v16i3(I).3892 Amondarain, J., Aldazabal, M. E., & Espinosa-Pike, M. (2023). Gender differences in the auditing stereotype and their influence on the intention to enter the profession. Journal of Behavioral and Experimental Finance, 37,

Theoretical Review and Hypothesis Development

Methods

Research Type

This study used a quantitative, cross-sectional survey design to estimate associations among auditing knowledge, professional skepticism, professional ethics, and a narrowly operationalized two-item proxy for auditor career interest. The three predictor constructs were modeled reflectively. The outcome was retained because only two career-related items were collected in the original survey; it is therefore not presented as a comprehensive or validated career-intention scale. Because all variables were measured at one point in time using self-reported perceptions, the design does not establish temporal order or causality. PLS-SEM was used to estimate the reported measurement and structural relationships within the available sample, but all structural findings are treated as provisional and sample-specific. The measurement model was evaluated using item loadings, average variance extracted (AVE), Cronbach's alpha, composite reliability, the Fornell-Larcker criterion, and the heterotrait-monotrait ratio (HTMT). The structural model was Table 2. Operational Definition of Variables Variable Operational definition Auditing Knowledge (X1) Understanding of audit standards, p materiality, and evidence. Critical mental attitude in evaluating delaying judgment until sufficient evide Professional Skepticism (X2) Moral and professional principles gu maintaining public trust. Professional Ethics (X3) Auditor Career Interest Proxy (Y) Internal tendency to choose auditor as perceived fit and career prospects. Source: Research instrument developed from the thesis data and evaluated using R², adjusted R², path coefficients, inner VIF, f² effect sizes, bootstrap inference, Q²_predict, and SRMR. A supplementary reanalysis used 5,000 percentile bootstrap resamples with a two-tailed 5% significance level. The reconstructed SRMR was calculated as the square root of the mean squared off-diagonal standardized residuals between the empirical indicator-correlation matrix and the model-implied correlation matrix generated from the retained outer loadings and construct-correlation matrix; the exact value was 0.1017. Because 0.1017 exceeds the commonly cited 0.08 benchmark, model-data discrepancy remains material and no claim of satisfactory global fit is made (Hair et al., 2020). Theory-driven results were therefore supplemented with reduced-model SRMR checks and a composite-score bootstrap regression as robustness analyses rather than post hoc model optimization.

Population and Sample

The operational sampling frame consisted of accounting students who were reachable through the three participating universities in Serang City during the data-collection period. Eligible students were approached through printed- questionnaire distribution and circulation of a Google Forms link via WhatsApp. Inclusion criteria were: (1) enrollment as an accounting student at Universitas Serang Raya, Universitas Pamulang Serang Campus, or Universitas Sultan Ageng Tirtayasa; (2) completion of an audit concentration or equivalent advanced audit coursework; and (3) voluntary submission of a usable questionnaire. Responses not meeting these criteria or incomplete for study variables were excluded. Incidental non-probability sampling was used because recruitment depended on eligible students who were accessible during the collection period. A total of 55 usable on Indicators Scale Likert 1-5 X1.1 standards; X1.2 risk-materiality- procedure; X1.3 practicum experience; X1.4 audit evidence analysis. , procedures, risks, Likert 1-5 X2.1 suspension of judgment; X2.2 search for knowledge; X2.3 self-esteem. ng information and dence exists. Likert 1-5 X3.1 public interest; X3.2 integrity; X3.3 objectivity; X3.4 confidentiality; X3.5 technical standards. guiding auditors in Likert 1-5 as a career based on Y1 challenging work; Y2 career path. and recent audit career literature.

Instrumentation or Tools

The instrument was a closed-ended questionnaire administered in Indonesian using a five-point Likert scale from strongly disagree to strongly agree. The items were developed for the original thesis-based study and were conceptually informed by audit education, professional skepticism, professional ethics, and accounting-career literature, but the available records do not provide verifiable documentation that the items were adopted or systematically adapted from established validated scales (Table 2). The records also do not document a formal pilot test or expert content-validity assessment before data collection. Therefore, pre-use content validity cannot be established retrospectively and is treated as a major measurement limitation, not merely a reporting omission. The questionnaire was originally developed and administered in Indonesian, so English-to- Indonesian translation or back-translation was not part of data collection. The outcome contains only two items concerning challenging work and career path. Measurement literature generally favors multi-item measures for broader latent constructs because they provide wider content coverage and more stable construct representation; shortened measures require strong conceptual justification (Diamantopoulos et al., 2012; Hair et al., 2020). Because these two items primarily reflect perceived job attractiveness and career structure rather than explicit plans, willingness, or likelihood of entering auditing, the outcome is retained only as a narrow auditor career-interest proxy. Complete item wording is provided in Appendix A.

Data Collection Procedure and Research Ethics

Data were collected from 29 April to 8 May 2026 through printed questionnaires and Google Forms distributed via WhatsApp. Before participation, respondents received information about the study purpose, voluntary participation, confidentiality, anonymity, and their right to decline or withdraw before submitting the questionnaire. No directly identifying information was included in the analytical dataset. The study was not submitted to an institutional ethics committee before data collection, and no formal ethics approval, waiver, or separate institutional permission letter was obtained. Accordingly, the manuscript does not claim institutional ethical approval or exemption. The study nevertheless applied basic participant-protection procedures through voluntary participation, prior study information, anonymity, confidentiality, and analysis of de-identified responses. No separate formal data-protection protocol was used beyond these de-identification and confidentiality procedures. This absence of prospective institutional ethics review or formal waiver is reported transparently as a research- governance limitation and remains subject to the journal's ethics requirements for publication.

Data Analysis

Analysis proceeded in two stages. Descriptive statistics summarized respondent characteristics and construct scores. PLS-SEM was then used to assess the reported measurement and structural relationships. Convergent validity was evaluated using outer loadings (>0.70) and AVE (>0.50), internal consistency using Cronbach's alpha and composite reliability (>0.70), and discriminant validity using Fornell-Larcker and HTMT. Structural assessment included R², adjusted R², inner VIF, f², path coefficients, bootstrap confidence intervals, Q²_predict, and SRMR. Because all variables were collected from the same respondents in one questionnaire, common method bias (CMB) was evaluated using two diagnostics: Harman's single-factor test and a full-collinearity VIF assessment. Following (Kock, 2015), full-collinearity VIF values below 3.3 were treated only as evidence against severe collinearity-related method bias, not as proof that CMB was absent. A marker-variable adjustment could not be implemented because no theoretically unrelated marker variable was included in the original questionnaire (Lindell & Whitney, 2001).

Table 1. Distribution of Respondents

Source: Primary data processed, 2026.

Table 2. Operational Definition of Variables

Source: Research instrument developed from the thesis data and recent audit career literature.

Result and Discussion

Descriptive Statistics

Table 3 reports summed construct scores. To make constructs with different indicator counts comparable, mean scores per item were also calculated: auditing knowledge = 3.91, professional skepticism = 4.24, professional ethics = 4.31, and auditor career interest = 3.92. Using a descriptive interpretation in which 3.41-4.20 indicates agreement and 4.21-5.00 indicates strong agreement, auditing knowledge and career interest fall in the agreement range, while skepticism and ethics fall in the strong-agreement range. These statistics describe response tendencies only and are not used to infer structural relationships.

Measurement Model

The reported predictor measurement models met the stated convergent-validity and reliability thresholds: outer loadings exceeded 0.70, AVE values exceeded 0.50, and Cronbach's alpha and composite reliability exceeded 0.70. The two career-related items also loaded strongly in the original SmartPLS specification, but high loadings and reliability do not resolve construct underrepresentation. The items 'challenging work' and 'career path' capture only a narrow portion of what a comprehensive behavioral career-intention construct would normally include. Consistent with measurement guidance favoring adequate content coverage for broad constructs (Diamantopoulos et al., 2012; Hair et al., 2020), the outcome is not interpreted as a validated latent career-intention scale. Instead, all structural findings involving Y are framed as provisional evidence for a narrowly operationalized two-item career-interest proxy. Figure 1 presents the SmartPLS 4 outer model for the measurement model evaluation. Mean Std. Dev. Minimum Maximum 15.655 4.133 4 20 12.709 2.492 4 15 21.527 4.185 7 25 7.836 1.983 3 10 Figure 1 shows that the retained indicators had loadings above 0.70. This evidence supports indicator reliability but does not establish the substantive importance of any individual item for career interest. Accordingly, the loadings are interpreted only as measurement evidence; substantive conclusions are based on structural path estimates and additional effect-size diagnostics (Table 4). Discriminant validity requires particular caution. The HTMT point estimate between professional skepticism and professional ethics was 0.884, below 0.90 but above the more conservative 0.85 guideline (Henseler et al., 2015). A 5,000-resample percentile bootstrap reanalysis yielded a 95% interval of approximately [0.593, 1.024], so discriminant validity between X2 and X3 is not established. Indicator-level inspection also showed notable cross-construct correlations, especially X2.1 with X3.4 (r = 0.713), X3.3 (r = 0.698), and X3.5 (r = 0.681) (Table 5), suggesting that students may perceive delayed judgment, objectivity/fairness, confidentiality, and procedural conduct as overlapping aspects of desirable professional behavior. A post hoc deletion sensitivity check reduced HTMT to 0.776 when X2.1 was omitted and to 0.841 when X3.1 was omitted. Neither deletion was adopted: removing X2.1 would leave skepticism with only two indicators, while deleting X3.1 solely to improve a threshold lacks a priori conceptual justification. The original constructs are therefore retained for transparency, but their separate structural paths are treated as empirically unresolved and interpreted only provisionally.

Structural Model and Hypothesis Testing

The R² value for the career-interest proxy was 0.442 (adjusted R² = 0.409), indicating moderate in-sample explanatory power. Inner VIF values were 1.880 for auditing knowledge, 3.045 for professional skepticism, and 2.272 for professional ethics. Effect sizes were f² = 0.186 for auditing knowledge, 0.010 for skepticism, and 0.024 for ethics. Repeated 10-fold cross-validation yielded Q²_predict values of 0.276 for Y1 and 0.184 for Y2. The exact reconstructed SRMR was 0.1017, above the 0.08 benchmark. Sensitivity checks did not show that simple theory-reducing respecification solved the discrepancy: SRMR was 0.0909 when skepticism was omitted, 0.1119 when ethics was omitted, and 0.1049 in a knowledge- only model. Because none reached 0.08, the original theory- driven specification was retained rather than selecting a model solely to improve fit. Structural conclusions are consequently treated as provisional. As an additional robustness check independent of the latent-score specification, standardized composite means were regressed on the two-item outcome proxy and evaluated with 5,000 nonparametric bootstrap resamples. The composite- score coefficients were 0.448 for auditing knowledge, 0.086 Composite E Cronbach reliability Decision alpha 94 0.852 0.900 Valid and reliable 08 0.795 0.879 Valid and reliable 06 0.895 0.923 Valid and reliable 39 0.809 0.912 Valid and reliable Figure 2 presents the path coefficients and significance testing results generated through the bootstrapping procedure in SmartPLS 4. The original SmartPLS bootstrapping results showed that auditing knowledge was positively associated with auditor career interest (β = 0.445, t = 2.199, p = 0.028). Professional skepticism had a positive but non-significant direct path (β = 0.128, t = 0.578, p = 0.563), as did professional ethics (β = 0.177, t = 1.168, p = 0.243) (Table 6). In the supplementary 5,000-resample percentile bootstrap reanalysis, the approximate 95% confidence intervals were [0.123, 0.920] for auditing knowledge, [-0.343, 0.562] for skepticism, and [- 0.151, 0.491] for ethics (Table 8). Thus, only H1 was supported. The non-significant results for H2 and H3 indicate that direct associations were not detected in this sample; they do not demonstrate that skepticism or ethics are unimportant to audit practice or career development. The full-item-level outer loadings for all indicators are presented in Table 7, with values ranging from 0.747 to 0.927, all exceeding the 0.70 threshold. Common method bias diagnostics yielded mixed evidence. Harman's single-factor analysis of all 14 questionnaire items showed that the first unrotated component accounted for 51.6% of total standardized item variance, slightly above the conventional 50% screening threshold; therefore, common method bias cannot be dismissed and may inflate or distort observed relationships. As an additional full-collinearity assessment, each construct score was regressed on the remaining construct scores. The resulting VIFs were 2.181 for auditing knowledge, 3.019 for professional skepticism, 2.311 for professional ethics, and 1.720 for the career-interest proxy, all below the 3.3 threshold proposed by (Kock, 2015). This second diagnostic does not show severe full- collinearity contamination, but it does not negate the Harman result or the single-source, single-time design. Because no marker variable or temporal/source separation was available, CMB remains a material threat to inference rather than a resolved issue.

Interpretation of Key Findings

H1 was supported in the original PLS specification: auditing knowledge was positively associated with the two-item career- interest proxy (β = 0.445, p = 0.028; f² = 0.186). The g) Decision Interpretation Supported Positive direct association with the two-item proxy supported in this sample. Not supported Direct association not supported; null result interpreted cautiously. Not supported Direct association not supported; null result interpreted cautiously. The result has practical relevance because auditing has substantial technical and procedural demands. Classroom exposure to risk assessment, internal control, evidence evaluation, working papers, and reporting may allow students to form a more concrete view of professional work. The structural coefficient, together with the effect size to be reported, should be used to assess practical importance. Outer loadings are not used to infer which learning experience causes career interest, because they indicate measurement quality rather than substantive impact. H2 was not supported: no statistically significant direct association was detected between professional skepticism and the two-item career-interest proxy (β = 0.128, p = 0.563; f² = 0.010). This null result cannot establish absence of a relationship. The sample has limited power for small effects, the outcome is narrowly operationalized, CMB cannot be excluded, and X2 is not empirically distinct from X3 under the bootstrap HTMT criterion. The separate skepticism path should therefore be interpreted as indeterminate rather than evidence that skepticism is unimportant to career development. H3 was not supported: professional ethics did not show a statistically significant direct association with the two-item career-interest proxy (β = 0.177, p = 0.243; f² = 0.024). As with H2, the null result is inconclusive because content validity is limited, the sample may not detect small effects, common-method influences remain possible, and the X2-X3 distinction is empirically unresolved. Ethics may also operate through professional identity or value fit rather than through a direct path, but those mechanisms were not tested. Taken together, the analyses provide provisional evidence that auditing knowledge is positively related to the narrowly defined career-interest proxy in this sample, while separate direct associations for skepticism and ethics were not established. The pattern should not be generalized to a comprehensive model of auditor career intention because the outcome contains only two job-characteristic-based items, the sample is small and non-probabilistic, nonresponse bias cannot be assessed, SRMR remains elevated, common method bias remains possible, and skepticism and ethics are not empirically distinct under the HTMT bootstrap test.

Comparison with Previous Studies

The provisional knowledge finding is directionally consistent with international research linking audit-career perceptions to professional proximity, knowledge, and perceived attractiveness (Amondarain et al., 2023; Graschitz et al., 2023; Thuy et al., 2022). However, because the present dependent measure is only a two-item career-interest proxy, the result should not be interpreted as demonstrating the same construct as validated behavioral career-intention scales used elsewhere. The result is also compatible with accounting-career research showing associations with perceived benefits, career prospects, personal capability, and educational exposure (Kong et al., 2020; Marsintauli et al., 2022; Srirejeki et al., 2019). In the present data, auditing knowledge has the largest reported structural coefficient, but the model's elevated SRMR and measurement limitations prevent a firm conclusion that competence-based exposure is more influential than professional-value variables in the wider student population. The non-significant skepticism and ethics paths differ from studies of practicing auditors, where these constructs are often associated with audit judgment, quality, and professional behavior. That comparison should be made cautiously: the present X2-X3 discriminant-validity problem means that their separate coefficients may be unstable or difficult to interpret, while students also have less engagement experience than practicing auditors. The current study therefore treats these null paths as unresolved rather than as evidence that professional values do not matter. The digital audit literature also helps interpret these findings. Felski (2023), Damerji & Salimi (2021) and Eilifsen et al. (2020) show that audit education must keep pace with technology and analytics. If students associate audit with outdated manual checking, they may not see the profession as attractive. Conversely, if they encounter audit analytics, data visualization, and AI-supported audit tools, auditing knowledge may become more engaging and contemporary. Thus, integrating digital audit content may strengthen the effect of knowledge and gradually make skepticism and ethics more meaningful through data-driven cases.

Practical Implications

The findings support specific curriculum actions. Accounting programs can establish audit case laboratories in which students complete risk assessment, control evaluation, substantive testing, and working-paper documentation. Audit analytics modules can require students to use spreadsheets or specialized tools to identify anomalies and visualize evidence. Skepticism can be assessed through evidence-conflict cases, delayed-judgment exercises, and mock client interviews, while ethics can be assessed through dilemma analysis, Collaboration with public accounting firms, professional associations, and alumni auditors can also improve professional proximity. Structured internships, mentoring, practitioner-led case reviews, and audit-technology demonstrations may clarify the nature of current audit work. Such interventions should be evaluated using pre-post measures of audit self-efficacy, professional identity, and career intention rather than relying only on satisfaction surveys.

Limitations and Cautions

This study has several substantial validity limitations. First, incidental sampling, the small sample, unequal representation across universities, and the predominance of female respondents limit external validity and may create selection/composition bias. Recruitment used printed questionnaires and WhatsApp-distributed Google Forms, but no complete invitation denominator or nonrespondent record was retained; therefore, the response rate and nonresponse bias cannot be assessed. The sensitivity analysis also indicates limited power for small effects. Second, the instrument lacks documented pre-use validation: the available records do not verify adoption or systematic adaptation from validated item-level scales, formal pilot testing, or expert content validation. This materially weakens content-validity claims and constrains interpretation of all structural paths. Third, all variables were collected from the same respondents at the same time. Harman's first component explained 51.6% of variance, whereas full-collinearity VIFs were below 3.3; these mixed diagnostics mean CMB cannot be ruled out. Fourth, the cross-sectional design cannot establish temporal order or causality. Fifth, the outcome consists of only two items on challenging work and career path and therefore represents a narrow career-interest proxy rather than a validated behavioral career-intention construct. Sixth, discriminant validity between professional skepticism and professional ethics is not established because the HTMT bootstrap interval includes 1.00; indicator-overlap sensitivity checks did not justify a defensible post hoc deletion strategy. Seventh, the exact reconstructed SRMR of 0.1017 exceeds 0.08, and reduced-model checks also remained above 0.08, indicating residual model-data discrepancy. Finally, the study was not submitted for institutional ethical review before data collection and did not receive a formal ethics waiver or separate institutional permission letter. Although voluntary participation, anonymity, confidentiality, and de-identified analysis were applied, the absence of prospective institutional ethics oversight is an important research- governance limitation that must be considered by the journal in determining compliance with its publication ethics requirements.

Recommendations for Future Research

Future research should use probability-based or stratified recruitment where feasible, document the number invited so a response rate can be reported, and determine sample size References Ajzen, I. (2020). The theory of planned behavior: Frequently asked questions. Human Behavior and Emerging Technologies, 2(4), 314–324. https://doi.org/10.1002/hbe2.195 Ali, N. A. M., Bustamam, K. S., Saidin, A., Adanan, S. A., & Sani, A. A. (2024). An exploration of accounting student’s perspective on factors influencing a priori through power analysis. Most importantly, auditor career intention should be measured with at least 3-5 validated intention items that directly assess willingness, plans, and likelihood of pursuing an audit career rather than relying on job- characteristic perceptions. The skepticism and ethics measures should also be refined and pretested to establish clearer discriminant validity. Future studies should incorporate procedural and statistical controls for common method bias, including temporal or source separation where feasible and a theoretically unrelated marker variable that permits marker- variable adjustment. Longitudinal and mixed-method designs could test whether auditing knowledge affects career intention through self-efficacy or perceived behavioral control, and whether professional skepticism and ethics operate indirectly through professional identity, value fit, or practical audit exposure.

Table 3. Descriptive Statistics of Research Variables

Source: SmartPLS 4 output, 2026.

Table 4. Measurement Model Evaluation

Source: SmartPLS 4 output, 2026.

Table 5. HTMT Discriminant Validity

Source: SmartPLS 4 output, 2026.

Table 6. Hypothesis Testing Results

Source: SmartPLS 4 output, 2026.

Table 7. Full Item-Level Outer Loadings

Source: SmartPLS 4 output, 2026.

Table 8. Supplementary Structural Diagnostics

Source: SmartPLS 4 output, 2026.

Conclusion

This study examined associations between auditing knowledge, professional skepticism, professional ethics, and a narrowly operationalized two-item auditor career-interest proxy among 55 accounting students from three universities in Serang City. In the original PLS model, auditing knowledge was the only statistically significant direct predictor (β = 0.445, p = 0.028), and a supplementary composite-score bootstrap check showed the same qualitative pattern. These results are provisional and sample-specific rather than a comprehensive test of career intention. Inference is materially constrained by incidental sampling and unknown response rate, undocumented pre-use instrument validation, the two-item outcome proxy, possible common method bias, an SRMR of 0.1017, and unresolved discriminant validity between skepticism and ethics. Accordingly, the study's contribution is primarily diagnostic and exploratory: it suggests that audit knowledge deserves further investigation while also identifying measurement and design conditions that must be strengthened before broader claims are made. Future research should use validated multi- item career-intention scales, prevalidated skepticism and ethics measures, larger probability-based samples with documented recruitment denominators, prospective ethics oversight, stronger procedural and statistical common-method controls, and model specifications that demonstrate adequate discriminant validity and residual fit.

Author Contributions

Denny Putri Hapsari contributed to conceptualization, data collection, data processing, and initial manuscript development. Denny Kurnia contributed to supervision, theoretical refinement, methodology review, interpretation of findings, and final manuscript editing. Both authors approved the final version of the manuscript.

Acknowledgements

The authors thank the participating accounting students and the academic units that facilitated non-identifying access to respondents. No institution influenced the design, analysis, interpretation, or reporting of the study.

References

Ajzen, I. (2020). The theory of planned behavior: Frequently asked questions. Human Behavior and Emerging Technologies, 2(4), 314–324. https://doi.org/10.1002/hbe2.195

Ali, N. A. M., Bustamam, K. S., Saidin, A., Adanan, S. A., & Sani, A. A. (2024). An exploration of accounting student’s perspective on factors influencing audit career choice. Information Management and Business Review, 16(3), 407–419. https://doi.org/10.22610/imbr.v16i3(I).3892

Amondarain, J., Aldazabal, M. E., & Espinosa-Pike, M. (2023). Gender differences in the auditing stereotype and their influence on the intention to enter the profession. Journal of Behavioral and Experimental Finance, 37, Hapsari et al. 100784. https://doi.org/10.1016/j.jbef.2022.100784

Bhatt, M. A., & Khan, S. T. (2023). Determinants of accounting students’ decision to pursue a career as ACCA-certified accountants: A case study of Omani students. Management & Sustainability: An Arab Review, 2(3), 217–238. https://doi.org/10.1108/MSAR-09-2022-0043

Damerji, H., & Salimi, A. (2021). Mediating effect of use perceptions on technology readiness and adoption of artificial intelligence in accounting. Accounting Education, 30(2), 107–130. https://doi.org/10.1080/09639284.2021.1872035

Diamantopoulos, A., Sarstedt, M., Fuchs, C., Wilczynski, P., & Kaiser, S. (2012). Guidelines for choosing between multi-item and single-item scales for construct measurement: A predictive validity perspective. Journal of the Academy of Marketing Science, 40, 434–449. https://doi.org/10.1007/s11747-011-0300-3

Eilifsen, A., Kinserdal, F., Messier, W. F., & McKee, T. E. (2020). An exploratory study into the use of audit data analytics on audit engagements. Accounting Horizons, 34(4), 75–103. https://doi.org/10.2308/HORIZONS-19-121

Fawziah, I. N., Iqbal, S., & Trisno, S. (2023). The role of locus of control in moderating the effect of auditor’s independence and professional skepticism on audit quality. The Indonesian Accounting Review, 13(2), 233–247. https://doi.org/10.14414/tiar.v13i2.3791

Felski, E. (2023). Audit technologies used in practice and ways to implement these technologies into audit courses. Journal of Accounting Education, 62, 100827. https://doi.org/10.1016/j.jaccedu.2022.100827

Graschitz, S., Holzknecht, S., & Steller, M. (2023). Perceived attractiveness of tax consultancy and auditing professions: Insights from a German-speaking area. SAGE Open, 13(1), 1–25. https://doi.org/10.1177/21582440231153100

Hair, J. F., Howard, M. C., & Nitzl, C. (2020). Assessing measurement model quality in PLS-SEM using confirmatory composite analysis. Journal of Business Research, 109, 101–110. https://doi.org/10.1016/j.jbusres.2019.11.069

Henseler, J., Ringle, C. M., & Sarstedt, M. (2015). A New Criterion for Assessing Discriminant Validity in Variance-Based Structural Equation Modeling. Journal of the Academy of Marketing Science, 43(1), 115–135. https://doi.org/10.1007/s11747-014-0403-8

Ilias, A., Baidi, N., Ghani, E. K., & Rahman, R. (2022). Factors driving the intention to pursue internal auditing certification and career among future graduates in Malaysia. Universal Journal of Accounting and Finance, 10(2), 549–558. https://doi.org/10.13189/ujaf.2022.100219

Kock, N. (2015). Common Method Bias in PLS-SEM: A Full Collinearity Assessment Approach. International Journal of E-Collaboration, 11(4), 1–10. https://doi.org/10.4018/ijec.2015100101

Kong, Y., Ngapey, J. D. N., & Qalati, S. A. (2020). Effects of financial rewards, parents and peers, and benefits and costs on choosing accounting career: A global perspective. Journal of Asian Finance, Economics and Business, 7(11), 157–167. https://doi.org/10.13106/jafeb.2020.vol7.no11.157

Marsintauli, F., Situmorang, R. P., & Suminar, S. R. (2022). Understanding the drivers’ factors for choosing an accounting student’s career as a professional accountant. Business Economic, Communication, and Social Sciences Journal, 4(2), 133–144. 10.61194/ijtc.v7i4.2558 https://doi.org/10.21512/becossjournal.v4i2.8450

Parluhutan, C. A., Ermawati, E., & Widyastuti, S. (2022). The influence of auditor ethics, auditor motivation, locus of control on audit quality: Role of professional skepticism. Universal Journal of Accounting and Finance, 10(1), 267–275. https://doi.org/10.13189/ujaf.2022.100127

Pelzer, J. E., & Nkansa, P. (2022). Why aren’t more minority accounting students choosing auditing? An examination of career choice and perceptions. Accounting Education, 31(4), 347–369. https://doi.org/10.1080/09639284.2021.1991404

Pusat Pembinaan Profesi Keuangan. (2019). Statistik dan Direktori Akuntan Publik. https://pppk.kemenkeu.go.id/

Rietz, S. D. (2018). Information vs Knowledge. Accounting Auditing & Accountability Journal, 31(2), 586–607. https://doi.org/10.1108/aaaj-01-2013-1198

Rizky, M. A., & Astuti, C. D. (2023). The influence of auditor competence, workload, and time budget pressure on audit quality with audit ethics as a moderating variable. Jurnal Ekonomi Bisnis, Manajemen Dan Akuntansi, 3(2), 514–526. https://doi.org/10.47709/jebma.v3i2.2824

Savira, J. A., Rahmawati, R., & Ramadhan, A. (2021). The influence of professional competence and skepticism on audit quality. Jurnal Ilmiah Akuntansi Kesatuan, 9(1), 23–30. https://doi.org/10.37641/jiakes.v9i1.435

Setiawan, S., Tin, S. E., & Meythi, M. (2025). Accounting students’ interest in the auditing profession in the digitalization era. Jurnal Ilmiah Akuntansi Kesatuan, 13(2). https://doi.org/10.37641/jiakes.v13i2.3228

Srirejeki, K., Supeno, S., & Faturahman, A. (2019). Understanding the intentions of accounting students to pursue career as a professional accountant. Binus Business Review, 10(1), 11–19. https://doi.org/10.21512/bbr.v10i1.5232

Sumartono, S., Ermawati, Y., & Rorong, T. M. (2022). Influence of experience and professional skepticism on audit judgement of government external auditors. Jurnal Akuntansi Dan Auditing Indonesia, 26(2), 211–223. https://doi.org/10.20885/jaai.vol26.iss2.art10

Surya, F., Pramudita, A., Wijaya, D., Permadi, D., & Putra, E. (2021). Professional ethics of auditors and its impact on audit quality. Golden Ratio of Auditing Research, 1(2), 99–107. https://doi.org/10.52970/grar.v1i2.368

Thuy, N. T. T., Truong, D. T., & Ho, T. H. M. (2022). Students’ perception of the auditing profession in Vietnam: The effect of proximity and knowledge on auditors’ stereotype. Cogent Business & Management, 9(1), 2154058. https://doi.org/10.1080/23311975.2022.2154058

Welly, Y., Loist, C., Wijaya, A., Sisca, S., & Sianipar, M. Y. (2022). Factors influencing the interest of accounting students’ career as public accountant: Work environment as moderators. E-Jurnal Akuntansi, 32(2), 3684. https://doi.org/10.24843/EJA.2022.v32.i02.p07

Wulan, D. F., Oktavia, R., & Syaipudin, U. (2023). The reasons why accounting student (not) pursuing public accounting profession: A systematic literature review. Ilomata International Journal of Tax and Accounting, 4(4), 646–667. https://doi.org/10.52728/ijtc.v4i4.775

Zhang, L., Wang, Q., & Li, H. (2024). AI-powered audit analytics: Enhancing analytical capabilities and decision-making skills in audit education. Journal of Accounting Education, 70, 100792. https://doi.org/10.1016/j.jaccedu.2023.100792