SMEs in Nairobi that report using data analytics also report better tax compliance, with larger firms gaining the most; results come from a 391-firm survey and are correlational rather than causal.

Main Finding

The study of 391 Nairobi CBD SMEs finds that investments in data analytics—both basic descriptive analytics and a composite “advanced analytics” index (diagnostic, predictive, prescriptive combined due to multicollinearity)—are positively associated with better tax compliance (i.e., reduced tax evasion). Advanced analytics show a larger effect than descriptive analytics. Firm size moderates these relationships: the compliance gains from analytics are significantly stronger for larger SMEs.

Key Points

• Sample and context: 391 SME owners/managers in Nairobi Central Business District, Kenya; cross-sectional survey (structured Likert questionnaire).
• Theoretical framing: Economic Deterrence Theory, Resource Dependency Theory, and Laffer Curve Theory.
• Main effect sizes and significance:
  • Descriptive analytics → tax compliance: β = 0.175, p = 0.033.
  • Diagnostic, predictive, prescriptive analytics showed severe multicollinearity (pairwise correlations 0.831–0.862), so they were combined into an Advanced Analytics Index; Advanced Analytics → tax compliance: β = 0.408, p = 0.003.
• Moderation by firm size:
  • Interaction descriptive_analytics × firm_size: β = −0.198, p < 0.001.
  • Interaction advanced_analytics × firm_size: β = −0.229, p < 0.001.
  • Authors interpret these results to mean larger firms derive greater compliance benefits from analytics-based strategies than smaller firms.
• Practical recommendations from the paper: Kenya Revenue Authority (KRA) should adopt integrated analytics platforms, promote digital recordkeeping among SMEs, and design firm-size–specific compliance strategies.
• Corroborating evidence cited: Positive effects of fiscal devices and e-invoicing on compliance in Rwanda, Tanzania, and Uganda; KRA’s stated move toward AI-driven enforcement.

Data & Methods

• Design: Cross-sectional explanatory quantitative study; positivist approach.
• Sampling: Stratified random sampling of SMEs across sectors in Nairobi CBD; simple random selection of respondents within strata.
• Data collection: Structured questionnaire (adapted from validated measures), 5‑point Likert items for descriptive, diagnostic, predictive, prescriptive analytics, firm size, and tax evasion/compliance; pilot-tested.
• Analysis: SPSS v27; descriptive statistics, Pearson correlations, multiple regression for main effects, and hierarchical moderated regression (interaction terms) to test moderation by firm size.
• Key data processing decision: High multicollinearity among diagnostic/predictive/prescriptive analytics led to creation of a composite Advanced Analytics Index.
• Limitations noted or implicit: cross-sectional and self-reported data (limits causal inference, potential common-method bias); multicollinearity constrained separate estimates for advanced analytics components.

Implications for AI Economics

• Revenue gains from AI/analytics: Empirical evidence that analytics—especially advanced analytics—can materially improve tax compliance among SMEs, implying positive returns to public investment in AI-driven tax administration.
• Heterogeneous returns to analytics: Firm size matters—larger SMEs capture more compliance benefit—suggesting uneven distribution of the gains from AI deployment. Policymakers should consider equity and efficiency trade-offs when scaling analytics programs.
• Implementation challenges that affect economic outcomes:
  • Data fragmentation, interoperability, and data quality are binding constraints; they affect the effectiveness of analytics and the realized fiscal gains.
  • Skills and model governance (validation, transparency, third-party data-sharing rules) are essential to sustain trust and accuracy; failures here can reduce the net benefit of AI.
• Policy design and cost externalities:
  • Analytics-based enforcement can raise detection probabilities and thus deter evasion, but may impose higher compliance/administrative costs on smaller firms; targeted support (digital recordkeeping subsidies, simplified interoperable tools) can mitigate adverse distributional effects.
  • Third-party data and fiscal devices (ETRs, e-invoicing) complement analytics; integrating these sources increases traceability and detection power.
• Research and evaluation needs for AI economics:
  • Longitudinal or quasi-experimental studies linking administrative tax records to firm-level analytics adoption would improve causal estimates of fiscal impact and cost-effectiveness.
  • Cost–benefit analyses comparing centralized (tax authority) analytics investments versus subsidizing SME adoption of interoperable digital record systems.
  • Examination of equilibrium behavior: how increased detection changes firm entry/exit, informality, pricing, and broader welfare.
• Takeaway for economists and policymakers: Data analytics/AI can strengthen tax capacity in developing-country settings, but maximizing social welfare requires attention to data infrastructure, governance, skill gaps, and firm-size–sensitive policies so that benefits are broad-based rather than concentrated among larger firms.