Ethical Leadership, Work Engagement, and Employee Creativity in Mogadishu’s Telecommunications Sector: Evidence From a Cross-Sectional Survey

Introduction

Employee creativity - the generation of ideas that are both novel and useful - is increasingly important in technology-intensive service organizations. Creative employees can improve work practices, respond to operational problems and identify new ways of serving customers. Yet creative effort is inherently uncertain: employees must decide whether an unconventional suggestion is worth pursuing and whether it is safe to expose an unfinished idea to evaluation. This makes the surrounding work environment, not only individual ability, central to understanding creativity (Amabile & Pratt, 2016).

Creative performance is not simply a matter of effort. A technically feasible suggestion has to be new and useful, and an employee may decide not to articulate it if supervisors discourage disagreement. This distinction makes the interpersonal setting relevant alongside individual capability and job demands (Woodman et al., 1993). The analytic focus is therefore on associations, not a claim that managerial virtue automatically generates ideas.

Ethical leadership is a feature of the workplace environment rather than a proxy for employee creativity. In Brown et al.'s (2005) formulation, it combines normatively appropriate conduct with active reinforcement and communication of ethical standards. Supervisors who listen, distribute opportunities fairly and explain decisions may affect whether employees believe proposing a novel idea is socially acceptable. Previous studies often report positive relationships, but their designs, measures and proposed pathways differ (Chughtai, 2016; Esguerra, 2022; Wadei et al., 2023). These findings do not, by themselves, establish that ethical leadership produces creativity in every setting.

Work engagement offers a complementary motivational explanation. Employees who bring energy, dedication and concentration to their roles may be more willing to persist with difficult problems and invest effort in alternative solutions. Evidence from service organizations has linked work engagement with employee creativity but also suggests that this relationship depends on other conditions such as perceived support and intrinsic motivation (Aldabbas et al., 2023). Engagement is therefore relevant to creativity without being equivalent to it.

A related empirical difficulty is that favorable self-reports about leaders, personal involvement and one's own creativity can move together for reasons unrelated to actual creative output. This threat is particularly salient where all variables are surveyed at one moment. A scientifically cautious analysis must distinguish the substantive rationale for studying these concepts from the strength of evidence the survey can provide (Podsakoff et al., 2003).

Background of the study

Somalia's communications environment is relevant to the interpretation of workplace behavior but should not be treated as a measured explanatory variable. The National Communications Law of 2017 established the National Communications Authority and a framework for licensing and communications regulation (Federal Republic of Somalia, 2017). The regulator reported that six licensed operators completed voice interconnection in March 2023 (National Communications Authority, 2023). These publicly documented developments describe the sector, not conditions within the sampled employers or a causal explanation of survey responses.

The regulator also describes interconnection obligations under Articles 49–53 of the National Communications Law (National Communications Authority, n.d.). Such institutional information makes the setting more specific than a generic national label, but it does not identify the employers or work arrangements of the 86 respondents. Published telecom-sector developments cannot be treated as substitutes for missing recruitment records, nor do they justify claims about the respondents' exposure to particular regulatory changes.

Research problem, objectives and significance

A focused empirical problem remains: the extent to which established ethical-leadership and work-engagement associations with creativity are visible within the same cross-sectional dataset of telecommunications employees in Mogadishu is not well documented. Much of the cited research examines other industries and national settings and, in several cases, indirect mechanisms rather than two directly estimated paths. The present article therefore aims (1) to examine the recorded association of ethical leadership with employee creativity and (2) to examine the recorded association of work engagement with employee creativity, while explicitly qualifying what can be inferred from the original model and incomplete analysis record. The corresponding questions are whether each relationship is positive and how consistently its recorded statistical evidence supports that direction. This is a quantitative, hypothesis-guided observational study with preliminary evidence; 'exploratory' describes the limited evidential reach, not a mixed-method sequence or a hypothesis-generating qualitative stage.

The two objectives concern statistical association within the originally estimated model. They do not extend to testing leadership interventions, comparing operators, determining how many Somali telecom workers are creative or explaining regulatory effects. A modest contribution is still useful when clearly bounded: it identifies which relationships merit replication and exposes where record-keeping and survey construction need improvement for management research in this setting.

The distinction between statistical and substantive uncertainty is particularly important for the two paths. The estimands are conditional on the original model's other predictors, including a “Psychological” variable whose precise item content is missing from the questionnaire appendix. Moreover, the reported sample was not drawn probabilistically. Both issues limit claims even where a coefficient passes a conventional statistical threshold.

Theoretical anchorage, empirical literature and conceptual framework

Foundational perspectives on creativity and ethical leadership

The componential perspective treats creativity as the production of ideas that are novel and useful, arising from the interaction of expertise, creative thinking and motivation within an enabling social environment (Amabile & Pratt, 2016). Woodman et al. (1993) similarly situated organizational creativity in interactions among persons, groups and organizational contexts. These classic and subsequent accounts caution against attributing employee creativity to a single managerial practice. A workplace can provide encouragement without supplying technical capability, time or opportunities to use new ideas.

Social learning theory explains how individuals learn behavioral expectations by observing consequential models (Bandura, 1977). Brown et al. (2005) applied this perspective to ethical leadership, in which managers communicate expectations through conduct, decisions and reinforcement. Work engagement supplies a separate motivational lens: vigor, dedication and absorption describe the psychological investment employees bring to work (Schaufeli et al., 2006). This distinction motivates testing ethical leadership and engagement as separate correlates of creativity, while recognizing that their empirical estimates may overlap with omitted organizational and individual conditions.

Critical synthesis of related research

The empirical literature should not be read as a collection of uniformly positive confirmations. For example, studies that investigate psychological empowerment or voice specify mechanisms that cannot be assessed from a direct-path-only model. Work-engagement investigations frequently involve organizational resources or team factors absent here. Differences in measures, source ratings and sample settings also limit direct effect-size comparison (Aldabbas et al., 2023; Chen & Hou, 2016; Wadei et al., 2023). Accordingly, the present framework is intentionally modest: two recorded direct relationships are interpreted with explicit acknowledgment of unmeasured processes.

Ethical leadership and employee creativity

Creative behavior may expose employees to embarrassment, criticism or the cost of failed proposals. Fair and consistent managerial responses could reduce anticipated interpersonal costs of suggesting improvements. Yet this is a theoretical explanation, not an established psychological-safety mechanism: although the original PLS-SEM output includes a construct called “Psychological,” its P1–P4 item wording is absent from the thesis questionnaire, and no mediation analysis was documented. The observed association might also reflect employees' favorable general judgments about their managers or workplace.

Evidence is supportive but does not identify a universal causal pathway. Chughtai (2016) examined leader–member exchange and empowerment; Chen and Hou (2016) considered employee voice and innovative climate; Wadei et al. (2023) reported evidence from Ghana involving creative self-efficacy and job autonomy; and Li et al. (2024) explored resource-related mechanisms. Research involving the Ghanaian organizational sample offers a particularly useful African comparison, but it is not equivalent to the Mogadishu sample in sampling, industry or measurement. Together, these studies justify examining a direct association while acknowledging that leadership, motivation and organizational affordances may be interdependent.

H1. Ethical leadership is positively associated with employee creativity.

Work engagement and employee creativity

Work engagement refers to vigorous, dedicated and absorbed involvement in work rather than creativity itself (Schaufeli et al., 2006). Bakker and Demerouti (2008) link engagement to resources available at work and argue that energetic employees may invest additional effort. Creativity, however, requires that effort to take an exploratory direction. A highly engaged employee may also devote substantial effort to prescribed routines, especially where discretion is limited. Positive associations should therefore be interpreted as context-dependent rather than inevitable.

Aldabbas et al. (2023) connected support, motivation, engagement and creativity within service organizations, whereas Tu et al. (2023) reported engagement-related creativity in a multilevel, matched-response investigation. The different designs matter: evidence from multi-source or multilevel studies cannot remove the shared-measure concerns in the present single-survey study. These findings support the directional rationale for H2, but neither establishes that a small cross-sectional survey in Mogadishu can estimate the effect without omitted-variable bias.

H2. Work engagement is positively associated with employee creativity.

Research methodology

Research design, study population and sampling

The research used a quantitative, cross-sectional observational survey of individuals working in the telecommunications sector in Mogadishu. The original undergraduate thesis (Ahmed, 2024) reported a target population of 148 people and 86 analyzed questionnaires. It also recorded a sample-size calculation using a stated 7% margin, but it did not identify a complete employee sampling frame or show a probability selection procedure. The thesis mentions sector operators in its general background but does not identify the participating employers, define the 148-person sampling frame, establish recruitment dates or document how many eligible employees actually received the survey. Accordingly, the 148 figure is a thesis-reported target-population estimate, not evidence of 148 questionnaires distributed. A response rate cannot be calculated reliably as 86/148. Recruitment is best described as nonprobability convenience sampling; a sample-size calculation does not confer random selection or a measured margin of sampling error, and the phrase 'selected individuals' in the questionnaire invitation does not demonstrate random selection.

The instrument was self-administered and combined demographic questions with study measures using a five-point agreement response scale (1 = strongly disagree to 5 = strongly agree). Eighty-six observations are sufficient to describe this obtained sample but do not establish population representativeness. A formal statistical-power assessment cannot be reconstructed from the available materials; nevertheless, small samples produce imprecise estimates and limit the reliability of significance and model diagnostics. No longitudinal follow-up or independent supervisor ratings were documented. These features are limitations of the actual design and are not reconstructed as a sequential mixed-method study.

This research design can examine statistical patterns but does not allow the measured variables to be causally ordered. The study does not include a control group, random assignment or experimental manipulation. The inferential target is the observed 86-person sample and its original conditional model coefficients. Claims about 'effectiveness' or 'impact' must not imply a treatment effect that the method cannot identify.

Measures and instrument adaptation

The appended undergraduate-thesis questionnaire (Ahmed, 2024) contained six ethical-leadership statements closely resembling material from the 10-item Ethical Leadership Scale of Brown et al. (2005): listening, attention to employee interests, fair decisions, ethics discussion, role modeling and response to unethical conduct. The preserved measurement output retains indicators EL1–EL5. The available documentation does not explain why EL6 was not retained. The article therefore describes the observed five-indicator model rather than claiming that an a priori five-item validated version of the scale was administered.

The six work-engagement items resemble items from the nine-item Utrecht Work Engagement Scale (Schaufeli et al., 2006), covering vigor, dedication and absorption. The survey used agreement responses instead of the scale's original frequency format, limiting equivalence to prior UWES scores. Employee creativity was represented by eight statements: five resemble the creativity measure used by Ganesan and Weitz (1996), while three resemble items associated with Zhou and George's (2001) supervisor-rated instrument. These latter statements appear in third-person phrasing in a self-administered survey. The mismatch between respondent and item perspective is a genuine measurement concern; no post hoc claim of a validated composite is made.

Item adaptations also matter for construct comparability. Calling a set of statements 'based on' an established instrument is not equivalent to showing that the published factor structure and scoring rules remained intact. In the absence of respondent-level data, alternative specifications, item deletion sensitivity checks and a substantive validity assessment cannot be performed. Internal consistency coefficients can be described without equating them to full measurement validity.

The original thesis (Ahmed, 2024) explicitly labels a fourth latent variable “Psychological” and reports four indicators (P1–P4) in its measurement tables. However, the questionnaire appendix does not contain the P-item statements, and the thesis alternately uses psychological factors, safety and well-being without defining the model construct consistently. Its precise measurement and theoretical identity therefore cannot be verified. It has therefore been excluded from the two focal hypotheses and substantive interpretation of this article, but it was present in the original model from which the two reported path estimates were obtained. Importantly, removing the construct from the narrative does not mathematically remove it from the original estimation. Estimates for a genuinely reduced model remain unavailable without the individual-level dataset and a rerun of the analysis.

Analysis strategy and inferential scope

The original undergraduate thesis (Ahmed, 2024) reports analysis using SmartPLS 4.0.8.6 and partial least squares structural equation modeling (PLS-SEM). Its documented measurement outputs comprise selected outer loadings, Cronbach's alpha, rho_A, composite reliability (rho_C) and average variance extracted (AVE). The thesis’s Table 4.5, titled “discriminant validities,” is not clearly identified as HTMT, a Fornell–Larcker matrix or construct correlations, and cannot responsibly be relabeled. The preserved VIF values relate to indicators rather than predictors in the structural equation and cannot establish the absence of structural collinearity (Hair et al., 2019).

The structural estimates reproduced here are coefficients, bootstrap standard deviations, t values, originally reported p values and f² statistics for the two identifiable paths in the original, larger model. No new PLS-SEM run has been carried out. In particular, R² from the original larger model is not treated as R² for an unestimated two-predictor model. Neither bootstrap confidence intervals nor iteration settings were available for independent verification. The reported p values agree approximately with one-sided tests, but the original tail choice is undocumented. The conventional |t| > 1.96 rule is used only as a transparent, approximate two-sided 5% reference for interpreting the recorded t statistics; it is not represented as a newly run bootstrap test or a substitute for original confidence intervals.

Ethical considerations and thesis provenance

The original undergraduate thesis appendix (Ahmed, 2024) invites individuals to participate and promises confidentiality, but it does not preserve a distinct record of consent, withdrawal rights, ethical committee approval or documented exemption. The analysis uses the original survey record without asserting procedures that cannot be verified. Research records suitable for confirming institution-specific ethics requirements would be needed for a stronger ethics statement.

The manuscript contains no author byline or contact details, consistent with the journal’s anonymous review format. The undergraduate thesis is acknowledged through scholarly citation in the text and reference list, with thesis provenance also explained confidentially in the response to the editor. The provided records do not establish a documented research-ethics approval, formal exemption, individual written consent procedure, or participating organizations, so none is asserted.

Data analysis and results

Respondent characteristics and descriptive-statistics availability

The analytic sample comprised 86 respondents: 53 men (61.6%) and 33 women (38.4%). Education was reported as bachelor's degree for 76 (88.4%), high-school qualification for six (7.0%) and master's degree for four (4.7%). Fifteen respondents (17.4%) reported less than one year of work experience, 59 (68.6%) one to three years and 12 (14.0%) four to six years. These sample descriptions are supported by the thesis demographic frequency tables (Ahmed, 2024); they are not population estimates. Age is not tabulated because the reported categories in the questionnaire and results record do not agree. Means, standard deviations and construct correlations were not preserved in the available results. The unidentified inter-construct matrix cannot safely serve as a correlation table. Respondent-level data or a clearly labeled construct-level SmartPLS descriptive output are needed to calculate the table requested by the reviewer.

Note. Source: demographic counts reported in Ahmed (2024). Percentages recomputed from n = 86, rounded to one decimal; totals may differ from 100% because of rounding. Age is not summarized because reported categories do not match the original questionnaire.

Table 1 summarizes sample demographics directly recorded in the source thesis. Tables 2 and 3 show, respectively, the available measurement and structural results. Construct-level means and standard deviations cannot be calculated from the thesis tables without the respondent-level responses or a documented construct-score output. Likewise, the original matrix entitled “discriminant validities” does not specify the statistic and must not be relabeled a correlation matrix. The reviewer’s request for construct means, descriptive SDs and correlations consequently remains only partially addressed.

Measurement properties

The retained loadings for employee creativity, ethical leadership and work engagement were reported between .727 and .913. Across the three constructs, Cronbach's alpha was .906–.914, rho_A was .910–.915, composite reliability was .928–.930 and AVE was .624–.727. These reported values exceed frequently used minimum criteria for internal consistency and convergent validity (Hair et al., 2019). Such values do not establish that the adapted scales have the same meaning as the published originals or that common-source bias is absent. Table 1 contains the specific values recoverable from the documented output.

The relatively high reliability coefficients are also compatible with redundant or closely overlapping items, and do not resolve whether ethical leadership is empirically distinct from a favorable overall appraisal of the workplace. This matters because theoretical separateness and discriminant validity are not identical. Independent validity evidence is needed before strong interpretations of the separate path estimates are warranted.

Note. α = Cronbach’s alpha; rho_A = coefficient rho_A; rho_C = composite reliability; AVE = average variance extracted. Values are reproduced from the original four-construct model report, not a newly estimated reduced model.

The original thesis’s “discriminant validities” matrix contains off-diagonal values from .761 to .855. Without verification of the statistic and how the construct score or ratio was produced, that matrix cannot demonstrate either adequate discriminant validity or the presence of severe multicollinearity. If the values represented HTMT, a value of .855 would exceed a conservative .85 benchmark, but treating this hypothetical interpretation as a confirmed test would be inappropriate. Structural VIF and common-method diagnostics cannot be reconstructed. These uncertainties reduce confidence in the separation of ethical leadership, engagement and creativity.

Structural model and hypothesis interpretation

The original estimated model reported R² = .609 and adjusted R² = .595 for employee creativity. These statistics describe that original model, which included the thesis-labeled but insufficiently specified “Psychological” predictor, and are not evidence of the explanatory power of a model limited to ethical leadership and work engagement. Within the original model, the recorded standardized path coefficients were β = .363 for ethical leadership and β = .213 for engagement. Their reported f² values were .135 and .038, respectively; both are small using conventional effect-size descriptions, with .135 near but below the .15 medium benchmark (Hair et al., 2022). These are conditional coefficients from the original specification; their values could change after removing its additional predictor.

The preserved output reports t = 2.144, p = .016 for ethical leadership and t = 1.651, p = .049 for work engagement. The p values are consistent with directional testing, but the original bootstrap test specification is unverified. Under the explicitly labeled approximate two-sided 5% t-value reference, H1 is provisionally supported in the original model, whereas H2 is not supported: its positive coefficient does not cross |t| = 1.96. Neither finding is presented as causal, nationally representative or independently replicated. Because there is no re-estimated reduced model, support for H1 cannot be transferred automatically to a hypothetical two-predictor equation.

The phrase 'provisionally supported' for H1 means that its preserved original-model t statistic exceeds a conventional threshold; it does not signify confirmation by an independently reproduced analysis. H2's recorded one-sided p value should not be converted into a claim of robust two-sided significance without a documented directional test plan. The uncertainty associated with multiple analytic choices and the missing raw data should accompany any account of the hypothesis outcomes.

Note. β = standardized path coefficient; bootstrap SD = reported standard deviation of the bootstrapped path estimate (not a construct-level descriptive SD); f² = reported effect-size statistic. All figures originate in the original larger SmartPLS model, which included the insufficiently specified “Psychological” construct. Original p values are reproduced as documented; their compatibility with one-tailed inference does not establish the test settings. Approximate |t| > 1.96 is shown only as a conventional two-sided reference, not as a replacement for missing bootstrap confidence intervals.

Discussion of findings

The two relationships differ in their recorded inferential strength. Ethical leadership shows the larger positive path and passes the approximate two-sided reference in the original model; work engagement has a smaller positive estimate that does not. This contrast is substantively relevant because the study treats leadership conduct and employees' involvement as distinct constructs, but sampling error, overlap among predictors and the unidentified additional construct prevent ranking them as stable independent causes. A positive directional p value does not resolve this uncertainty when the original test configuration is unknown.

Social learning and creativity theories offer an interpretation, not a proven mechanism. Observing supervisors behave fairly and articulate ethical expectations may signal that responsible initiative is acceptable (Bandura, 1977; Brown et al., 2005). The componential perspective further emphasizes motivational and environmental conditions (Amabile & Pratt, 2016). Prior studies identify relational exchange, voice, self-efficacy and job discretion as plausible intermediate conditions (Chen & Hou, 2016; Chughtai, 2016; Wadei et al., 2023). None of these variables was tested as a mediator here; attributing the observed coefficient to them would exceed the evidence.

The positive but statistically uncertain engagement estimate is also consistent with a conditional motivational account. Work engagement may provide persistence and energy, yet without discretion or a receptive climate these resources need not yield original ideas. Research in broader service and team contexts has found links among support, engagement and creativity (Aldabbas et al., 2023; Tu et al., 2023). In the present analysis, H2 should not be described as confirmed at the conventional two-sided 5% reference level. A larger study with adequate precision and measures of job design may clarify whether the observed pattern reflects a real association or sampling variability.

Somalia's documented communications law and reported operator interconnection show that the sector has an identifiable regulatory environment (Federal Republic of Somalia, 2017; National Communications Authority, 2023). However, those public facts are not participant-level variables in this study, and no comparative dataset tests the effect of regulation, competition or post-conflict conditions on respondents. The defensible contextual claim is narrower: a positive ethical-leadership association was recorded within one convenience sample of employees working in Mogadishu's telecommunications services. The sample does not establish that the result is uniquely Somali or significantly different from results obtained elsewhere.

Rival explanations

A credible rival explanation is common method variance: respondents who prefer to present themselves positively may rate their supervisors, engagement and creativity favorably in the same questionnaire (Podsakoff et al., 2003). A second is predictor collinearity or insufficient discriminant validity, which could make individual paths unstable. The original “Psychological” predictor, whose item wording is not preserved in the thesis appendix, creates a further interpretive problem because it may account for variance shared with either focal construct. Restricted responses or workplace selection processes are also possible, but the data required to assess them are not available. These possibilities are not demonstrated biases; they are unresolved alternatives that should temper the article's claims.

Contribution and interpretation

The principal contribution is limited and descriptive: two established workplace correlates were considered in the same documented survey of telecommunications employees in Mogadishu, and their source-thesis estimates do not display equal evidential strength. The analysis is consistent with, rather than a challenge to or revision of, social learning and componential accounts. It cannot claim to discover a novel theoretical mechanism, identify a contextual moderator or establish population-level prevalence. A further contribution of the revised account is transparent separation of the original model's reported estimates from the conceptually simplified discussion; this is an improvement in reporting, not a new empirical finding.

Cross-setting comparison therefore remains an agenda for future evidence rather than an accomplished theory test. Evidence from Ghana and other regions suggests several mechanisms worth studying, but heterogeneity across organizations, jobs, measurement systems and national institutions could produce different associations (Wadei et al., 2023). A stronger contribution would require prospective design, validated measurement, documented recruitment, sufficient observations and an explicitly estimated model in which relationships can be compared with comparable studies.

The study's value should not be overstated as a 'contextual extension' automatically equivalent to theoretical contribution. It documents a limited association in an underrepresented setting while exposing boundaries on inference. Transparency about the incompletely documented “Psychological” predictor, missing descriptives and inconclusive discriminant validity is necessary for readers to evaluate how much weight to assign the recorded coefficients.

Implications for management and research practice

For managers, the observed ethical-leadership association provides a reason to examine whether everyday supervisory communication and fair treatment accompany employees' willingness to suggest improvements. It does not demonstrate that training programs, mandated listening sessions or any particular organizational intervention will raise creativity. Such interventions should be treated as testable management options, preferably evaluated with baseline and follow-up data, not prescriptions derived from this cross-sectional survey. No participating companies can be named from the available sampling record, so recommendations refer to the sector in general.

The engagement coefficient provides still weaker grounds for direct action. Organizations considering initiatives aimed at motivation or job involvement should also consider opportunities for voice, task discretion and feedback, but the present dataset cannot establish which design will be effective. Employers could measure both perceived ethical behavior and independently assessed creative contributions before deciding whether a local pilot merits expansion. These are proposals for evaluation, not statistically established consequences of the reported paths.

An employer contemplating a pilot should specify what observable change would count as creative performance, collect evaluations from more than one source and compare results over time. Without this additional evidence, the present survey supports inquiry into supervision and engagement, not a cost-benefit judgment or an assurance of improvement. This practical boundary aligns the implications with the measured cross-sectional associations.

Limitations and directions for future research

First, the cross-sectional design cannot establish time order or rule out reverse causation; employees who see themselves as creative might evaluate leaders more favorably. Second, convenience recruitment and the small analytic sample restrict precision and generalization. The sampling frame, organizational composition, recruitment period and response denominator are not documented, preventing nonresponse assessment and subgroup comparisons. No defensible population-based response rate can be calculated from 86 completed surveys and a reported accessible population of 148. A future study should document eligible employees, actual invitations, completion counts and participating firms, and perform a justified a priori power or precision analysis.

Third, all focal constructs were assessed through a single self-administered questionnaire. Shared response tendencies and social desirability may inflate or obscure associations (Podsakoff et al., 2003). Item wording was also inconsistent: several creativity items originated in a supervisor-oriented source but remained in third-person form, and the work-engagement response scale was changed from its established frequency format. Possible restriction of response range cannot be assessed without item-level observations. Repeating the study with consistent validated measures, supervisor-rated creativity, temporal separation of measures and transparent instrument translations would help distinguish conceptual relationships from method artifacts.

Fourth, the original estimated specification included a four-item predictor labeled “Psychological” in the thesis; because the P1–P4 questionnaire statements and construct definition are missing, its precise measurement and role cannot be established. Consequently, alternative predictor-overlap and model-specification explanations cannot be dismissed. Discriminant validity remains uncertain, the available VIF table is indicator level and no construct correlation matrix, construct means, standard deviations or bootstrap confidence intervals can be independently recovered. The recorded original R² and path estimates must not be interpreted as results of a new reduced model. Future work should retain raw anonymized data, full SmartPLS project files, item codes, consent and ethical review documentation, and a reproducible analysis plan. Where feasible, collect data longitudinally and test specific mechanisms such as psychological safety, empowerment, employee voice and creative self-efficacy using validated measures rather than assuming they operated here. Interview-based follow-up may clarify local workplace experiences; such work would be a new study, not part of the present design.

A more discriminating future model could specify psychological safety as a mediator between ethical leadership and creative voice, or assess whether job autonomy moderates the association between engagement and idea generation. Such proposals would require explicit measurement, adequate sample size, theoretical justification and suitable design. An appropriately planned multi-site study could test cross-organization variation without naming firms that have not been documented as participants in this survey.

Conclusion

This cross-sectional quantitative study reports a limited re-examination of originally documented associations among ethical leadership, work engagement and employee creativity in a convenience sample of 86 telecommunications employees in Mogadishu. Within the original larger PLS-SEM model, ethical leadership had a positive path that met an approximate two-sided 5% reference; the work-engagement path was positive but did not. This pattern is consistent with established perspectives linking supervisory behavior and employee motivation to the social conditions around creative work, without demonstrating a causal process.

The inference remains constrained by incomplete sampling information, shared-source measurement, uncertainty about the original “Psychological” measure and unavailable raw data. The original R² cannot be attributed to a two-predictor model, and several statistical diagnostics requested during review cannot be reconstructed responsibly. These limitations do not negate the recorded observations but place clear boundaries around their interpretation. A larger, preregistered, multi-source and well-documented replication is needed before making strong organizational recommendations or population claims.