What determines digital accounting systems’ continuance intention? An empirical investigation in SMEs

The study examines what drives organisations, specifically SMEs, to continue using digital accounting systems (DAS). In modern business, information technology underpins decision-making, and DAS provide reliable, timely financial information for managerial decisions. While adoption benefits of DAS have been explored, less is known about post-adoption continuance intention (ICU-DAS), especially in less developed countries (LDCs) like Yemen. Prior literature emphasizes that long-term success from IT investments depends more on continuance than initial adoption. Building on established theories (ISSM, TAM, ECM), this paper compares these models and a synthesised model to identify the determinants of ICU-DAS in Yemeni SMEs. The study addresses a critical gap by focusing on continuance during a challenging context (political instability, COVID-19) and aims to inform SMEs, policymakers, and technology providers about factors enhancing sustained DAS use.

The research background covers Yemen’s SME context (97% of private firms; key roles in employment and GDP, but facing instability and COVID-19 impacts). The concept of ICU-DAS is defined as users’ readiness to continue using DAS, influenced by system quality, information quality, ease of use, perceived usefulness, satisfaction, and confirmation. Three theoretical lenses are reviewed: (1) ISSM (DeLone & McLean) with quality (system quality, information quality), use (use, satisfaction), and impacts dimensions; flexible use of constructs allows integration with other models. (2) TAM (Davis), where perceived usefulness (PU) and perceived ease of use (PEU) shape behavioural intention; PEU influences PU and intention; in continuance contexts, satisfaction can proxy for attitude. (3) ECM (Bhattacherjee), positing that continuance intention depends on post-adoption PU, confirmation (CON), and satisfaction (S). The paper motivates integrating constructs from ISSM, TAM, and ECM to improve explanatory power in DAS continuance. Reviewed studies (Table 2) show diverse applications of these models across contexts (e.g., e-finance, cloud ERP, e-government, chatbots), but limited evidence in LDCs and DAS, justifying the current investigation. Eleven hypotheses are developed: SQ→S (H1), SQ→ICU-DAS (H2), IQ→S (H3), IQ→ICU-DAS (H4), PU→S (H5), PU→ICU-DAS (H6), PEU→PU (H7), PEU→ICU-DAS (H8), CON→PU (H9), CON→S (H10), S→ICU-DAS (H11).

Design: Quantitative, cross-sectional survey targeting Yemeni SMEs (≤50 employees). Population: Managers and owners using DAS with accounting backgrounds across manufacturing, services, and commerce sectors. Instrument: Seven constructs measured—System Quality (SQ), Information Quality (IQ), Perceived Usefulness (PU), Perceived Ease of Use (PEU), Confirmation (CON), Satisfaction (S), and Intention to Continue Using DAS (ICU-DAS). Items adapted from prior studies and contextualized to DAS; 5-point Likert scale. Example items: SQ (flexibility, accessibility, reliability), IQ (accuracy, relevance, timeliness), PU (improved performance, productivity, control), PEU (low mental effort, simplicity, clarity), CON (experience exceeds expectations), S (meets expectations, satisfaction with support and interaction), ICU-DAS (intention to continue and use regularly). Data collection: 440 questionnaires distributed; 318 responses (72.3% response rate); 308 valid for analysis. Bias checks: Non-response bias deemed less serious; common method bias assessed via full collinearity VIFs—values between 1.007 and 2.261 indicated no CMB concerns. Measurement quality: Factor loadings generally high; Cronbach’s alpha (CA) and composite reliability (CR) satisfactory for all constructs (e.g., CA ranging approximately 0.735–0.870+, CR approximately 0.796–0.921). Convergent validity supported by AVE values (e.g., SQ 0.745; IQ 0.758; PU 0.763; PEU 0.567; CON 0.635; S 0.797; ICU-DAS 0.742). Sample characteristics (n=308): Gender: Male 271 (88%), Female 37 (12%). Position: Owner 197 (64%), Manager 111 (36%). Age: <25 (13%), 25–35 (45%), >35 (42%). Experience: <5y (11%), 5–10y (32%), 11–15y (30%), >15y (27%). Education: School/Diploma (14%), Bachelor (64%), Postgraduate (22%). Analysis: Partial least squares structural equation modeling (PLS-SEM) using SmartPLS. Justification: Suitable for complex models with multiple constructs, non-normal data, bootstrapping, and moderate sample sizes.

Model fit and explanatory power: All four models fit satisfactorily. Variance explained (R²) in ICU-DAS: Synthesised model 63.6%, ISSM 57.9%, TAM 55.6%, ECM 50.1%. Synthesised model outperformed individual models. Additional R²: PU explained by CON alone in ECM 2.3%; by CON and PEU in synthesised model 38.1%; by PEU alone in TAM 36.3%. Satisfaction R²: Synthesised 59.7%, ISSM 53.5%, ECM 45.6%. Hypothesis testing (synthesised model): - H1 SQ→S: β=0.334, p<0.001 (accepted). - H2 SQ→ICU-DAS: β=0.115, p<0.05 (accepted). - H3 IQ→S: β=0.243, p<0.01 (accepted). - H4 IQ→ICU-DAS: β=0.213, p<0.001 (accepted). - H5 PU→S: β=0.343, p<0.001 (accepted). - H6 PU→ICU-DAS: β=0.250, p<0.001 (accepted). - H7 PEU→PU: β=0.599, p<0.001 (accepted). - H8 PEU→ICU-DAS: β=0.192, p<0.01 (accepted). - H9 CON→PU: β=0.130, p<0.05 (accepted). - H10 CON→S: β=−0.011, ns (rejected). - H11 S→ICU-DAS: β=0.194, p<0.01 (accepted). Baseline models: - ISSM significantly linked SQ, IQ, and S to ICU-DAS; SQ and IQ significantly influenced S; R²(ICU-DAS)=57.9%. - TAM showed PU and PEU as major determinants of ICU-DAS; R²(ICU-DAS)=55.6%. - ECM supported its role in ICU-DAS with all paths positive except CON→S; R²(ICU-DAS)=50.1%. Measurement model: Strong factor loadings; reliability (CA, CR) satisfactory; AVEs generally >0.5; VIFs 1.007–2.261; CMB not a concern.

The study confirms that multiple quality and belief constructs drive SMEs’ continuance intention to use DAS. System quality and information quality enhance satisfaction and directly increase ICU-DAS, reinforcing ISSM findings that technical robustness and high-quality information promote continued use. Perceived usefulness strongly affects both satisfaction and ICU-DAS, highlighting that performance benefits are central in contexts where use may be expected or mandated. Perceived ease of use substantially boosts perceived usefulness and also directly supports ICU-DAS, underscoring the importance of intuitive interfaces and low effort. Satisfaction positively influences ICU-DAS, aligning with ECM’s post-adoption logic that positive experiences drive continuance. Notably, confirmation did not significantly affect satisfaction, contrary to ECM expectations, although it did increase perceived usefulness. This suggests that even when experiences exceed expectations, satisfaction may be shaped more by ongoing perceptions of usefulness, system performance, and information quality than by expectation confirmation alone in this context. Overall, the synthesised model (integrating ISSM, TAM, and ECM) provides superior explanatory power, indicating that continuance intentions in DAS are multifaceted and best understood through a combined lens.

The study compared ISSM, TAM, ECM, and an integrated synthesised model to explain SMEs’ continuance intention to use digital accounting systems. The synthesised model achieved the highest explanatory power (R²(ICU-DAS)=63.6%), validating the integration of system and information quality, user beliefs (usefulness, ease of use), satisfaction, and confirmation. Key determinants of ICU-DAS are system quality, information quality, perceived usefulness, perceived ease of use, and satisfaction; satisfaction is driven by system quality, information quality, and perceived usefulness; confirmation and perceived ease of use enhance perceived usefulness. Contributions include extending and validating these models in an LDC (Yemen) and in the DAS context, enhancing cross-cultural applicability of continuance theories. Practical implications: SMEs and policymakers should invest in improving DAS quality and information outputs, design for ease of use, and focus on delivering tangible performance benefits to sustain use; technology vendors should align product features with these determinants. Future research directions include testing extended versions of the base models, examining demographic effects, conducting longitudinal studies to capture evolving behaviours, probing the unexpected non-significant CON→S link, and assessing generalisability across sectors and countries.

• The study employed the original ISSM, TAM, and ECM; extended versions (e.g., DeLone & McLean 2003; UTAUT; ECM extensions) were not tested. - Demographic moderators (e.g., age, role, experience, education) were not examined for their effects on ICU-DAS. - Cross-sectional design limits insights into temporal changes; longitudinal analyses are recommended. - The non-significant relationship between confirmation and satisfaction was unexpected and warrants deeper investigation. - The sample focused on Yemeni SMEs (an LDC context), which may limit generalisability to other countries or larger firms.