1. Data

This study integrates the World Development Indicators (WDI), World Governance Indicators (WGI), and Doing Business Project data, comprising 1,187 country-year observations from 2015-2019. The WDI dataset, established in 1997 and maintained by the World Bank’s Development Data Group, collects standardized statistics from national agencies and international organizations to support evidence-based policymaking for sustainable development goals.

The dependent variable is the Ease of Doing Business (EDB) score (0-100), measuring regulatory environment conduciveness for SMEs. Explanatory variables include economic indicators (GDP per capita, inflation, interest rate spread, current account balance, CPI), infrastructure metrics (electricity access, broadband and mobile subscriptions), governance indices (government effectiveness, rule of law, control of corruption), and business process measures (time and cost to start a business). Categorical variables include country, year, region (following World Bank classifications), and income groups (Low, Lower-Middle, Upper-Middle, and High Income).

Table 1: N = 937 for variables with complete observations, after droping the regional aggregated data. The sample comprises 1,187 country-year observations from the World Bank’s World Development Indicators, World Governance Indicators, and Doing Business Project databases for the period 2015-2019. The sample size varies by variable due to differential reporting across countries. Categorical variables (Country, Region, Year, and Income Group) are not presented. Government Effectiveness, Rule of Law, and Control of Corruption are standardized indices (mean = 0, SD = 1) with higher values indicating stronger institutional quality. Current Account Balance is measured as a percentage of GDP, with negative values indicating deficits. Interest Rate Spread represents the difference between lending and deposit rates. Ease of Doing Business scores range from 0 to 100, with higher values indicating more business-friendly regulatory environments.
	N	Mean	SD	Min	Max
Ease of Doing Business	937	61.19	13.91	19.98	87.17
Access to Electricity (%)	932	82.86	26.67	4.20	100.00
Current Account Balance (% of GDP)	867	-3.09	8.89	-52.52	40.01
Consumer Price Index	888	180.22	623.53	97.75	15749.19
Interest Rate Spread	581	6.87	6.10	-2.09	45.00
GDP per Capita (USD)	914	20553.57	20968.85	753.46	121403.82
Broadband Subscriptions (per 100 people)	897	13.26	13.66	0.00	62.21
Mobile Subscriptions (per 100 people)	919	106.28	37.31	14.22	286.22
Government Effectiveness	937	-0.09	0.98	-2.44	2.25
Rule of Law	937	-0.08	0.98	-2.38	2.05
Control of Corruption	937	-0.08	0.99	-1.80	2.24
Days to Start a Business	937	21.65	26.77	0.50	230.00
Cost to Start a Business (% of income)	937	25.41	43.89	0.00	422.40

The dataset has several limitations: national-level aggregation obscures subnational variation; governance indicators reflect expert perceptions rather than direct observations; data completeness varies by country income level; the discontinuation of the Doing Business Project in 2021 raises methodological concerns; and the pre-pandemic timeframe (2015-2019) limits insights into post-COVID regulatory adaptations. Despite these constraints, the dataset enables robust cross-national analysis of business environment determinants across diverse development contexts.

2. Questions

Building on this comprehensive dataset, my analysis addresses several interconnected research questions regarding the determinants of business-friendly regulatory environments. First, I examine the roles of infrastructure and economic stability indicators in predicting a country’s ease of doing business score, specifically investigating how factors such as access to electricity, current account balance, consumer price index, and interest rate spreads correlate with business environment quality. This inquiry aims to identify which economic indicators serve as the strongest predictors of business-friendly environments across diverse contexts. Second, I investigate whether regional patterns shape the relationship between economic indicators and the business environment, exploring systematic differences in how these factors affect business environments across geographic regions and whether certain regions exhibit distinct patterns in their business environment development trajectories.

Additionally, I analyze how governance factors interact with economic and infrastructure indicators, examining the relationship between governance metrics (Rule of Law, Government Effectiveness, Control of Corruption) and ease of doing business, while also investigating whether these governance factors moderate the relationship between economic indicators and business environment quality. Finally, I explore the temporal effects of economic stability on ease of doing business, assessing whether lagged economic indicators better predict current business environment quality and evaluating the stability of these relationships over the 2015–2019 period.

3. Visualization

Figure 1. Regional trends in Ease of Doing Business scores from 2015 to 2019. The figure displays country-level Ease of Doing Business (EDB) scores (points) and regional trends (dashed lines) across the five-year period preceding the COVID-19 pandemic. Each point represents an individual country-year observation (n = 937), with colors denoting World Bank regional classifications. Regional trend lines were estimated using locally weighted scatterplot smoothing (LOESS). North America consistently maintained the highest business environment quality (EDB ≈ 82), while Sub-Saharan Africa showed the lowest average scores (EDB ≈ 50), though with substantial intra-regional variation. All regions demonstrated modest improvement over the study period, with South Asia exhibiting the steepest positive trajectory (approximately 8-point increase). These patterns suggest persistent regional disparities in regulatory quality despite global convergence toward business environment improvement during this period.

Figure 2. Correlation heatmap of business environment determinants. The figure displays Pearson correlation coefficients between 13 key variables related to business environment quality. Correlation strength and direction are represented by color intensity (blue for positive, red for negative correlations) and numeric values. Hierarchical clustering was applied to organize variables by similarity pattern. The strongest positive relationships appear among governance indicators (government effectiveness, rule of law, control of corruption; r = 0.9), which also correlate strongly with ease of doing business (EDB) scores (r = 0.5-0.8). Infrastructure variables (electricity access, mobile and broadband penetration) show moderate positive associations with both governance indicators and EDB scores. Business entry barriers (days and cost to start a business) demonstrate negative correlations with governance and EDB measures (r = -0.3 to -0.5). Economic stability indicators (CPI, interest spread) exhibit weaker relationships with other variables, suggesting their more complex and contextual influence on business environments across diverse development settings.

Figure 3. Relationship between electricity access and ease of doing business across income groups. The scatter plot illustrates the association between population access to electricity (x-axis) and Ease of Doing Business scores (y-axis) across 937 country-year observations (2015-2019), with points colored by World Bank income classification. A significant positive linear relationship exists between electricity access and business environment quality (gray dashed regression line with 95% confidence interval, r = 0.5, p < 0.001). High-income economies (purple) cluster predominantly at maximum electricity access (100%) with superior business environment scores (60-90), while low-income countries (blue) demonstrate greater variability in both dimensions. The relationship exhibits notable heterogeneity across income groups, with several low-income countries achieving relatively high business environment scores despite electricity limitations, suggesting complex interaction between infrastructure development and regulatory quality. The non-uniform distribution at 100% electricity access indicates that while basic infrastructure represents a necessary condition for business-friendly environments, it is insufficient alone to explain regulatory quality differences, particularly among upper-middle income economies (yellow).

Figure 4. Regional variation in digital infrastructure and institutional quality. Box plots display the distribution of fixed broadband subscriptions per 100 people (left panel) and Rule of Law index scores (right panel) across seven World Bank regions (n = 937 country-year observations, 2015-2019). Regions are ordered by median values. Digital infrastructure shows pronounced stratification, with North America exhibiting the highest broadband penetration (median ≈ 35 subscriptions per 100 people), followed by Europe & Central Asia (median ≈ 28), while Sub-Saharan Africa demonstrates minimal connectivity (median < 1). Similarly, institutional quality reveals persistent regional disparities, with North America and Europe maintaining significantly stronger rule of law environments (medians ≈ 1.7 and 0.5, respectively) than other regions. East Asia & Pacific displays considerable heterogeneity in both dimensions, reflecting diverse development trajectories within the region. The parallel regional hierarchies across both panels suggest potential co-evolutionary relationships between digital infrastructure development and institutional quality, though with notable outliers (represented as individual points) indicating context-specific divergences from regional patterns.

Figure 5. Principal component analysis of business environment determinants. The biplot visualizes the contributions of 11 key variables to the first two principal components, which together explain 61% of total variance across 937 country-year observations. Component 1 (horizontal axis, 50.6% variance explained) represents overall development level, with governance indicators (rule_law, gov_effectiveness, control_corruption) and infrastructure metrics (electricity, broadband, mobile) loading strongly on the negative side. Component 2 (vertical axis, 10.4% variance explained) distinguishes institutional quality (positive loadings for governance indicators) from economic conditions (negative loadings for infrastructure and stability metrics). Vector colors indicate variable contribution strength (dark blue to yellow representing low to high). The clustering pattern reveals three distinct dimensions of business environment: governance quality (upper left), infrastructural development (lower left), and macroeconomic stability (right side, represented by interest_spread and cpi). Notably, the ease of doing business score (edb) aligns more closely with infrastructure variables than governance metrics, challenging conventional emphasis on institutional factors alone and suggesting a more holistic approach to business environment reforms that incorporates physical and digital infrastructure development alongside institutional improvements.

4. Models

The regression models offer substantive insights into the determinants of business-friendly regulatory environments, directly addressing my research questions regarding infrastructure, economic stability, and their interactions.

Regarding the first research question on infrastructure and economic stability indicators, Model 1 demonstrates that both dimensions significantly predict business environment quality, with infrastructure (electricity access) and financial market efficiency (interest rate spread) emerging as the strongest predictors. The standardized coefficients (not shown) would likely indicate that these factors outweigh other economic indicators, suggesting a hierarchy of importance in business environment determinants.

Table 2 presents coefficient estimates from four regression models predicting Ease of Doing Business scores (2015-2019) with robust standard errors in parentheses. Access to electricity demonstrates a consistently strong positive relationship with business environment quality across all specifications (β = 0.203-0.241, p < 0.001). Interest rate spread exhibits the largest negative effect (β = -0.409 to -0.439, p < 0.001), indicating substantial penalties to business environment quality from financial market inefficiencies. Current account balance shows significant positive associations in non-interaction models (β = 0.253-0.296, p < 0.001) but becomes non-significant when interacted with infrastructure indicators. The interaction model (column 4) reveals significant synergistic effects between electricity access and external balance (β = 0.006, p < 0.001), suggesting amplified infrastructure benefits in macroeconomically stable contexts. Model fit remains substantial across specifications (Adj. R² = 0.384-0.432), with the interaction specification demonstrating better explanatory power.
	All Years	All Years (w/ IV Lags)	All Years (w/ IV Lags and Fills)	Interaction Model
+ p < 0.1, * p < 0.05, p < 0.01, * p < 0.001
Access to Electricity	0.214***	0.203***	0.205***	0.241***
	(0.015)	(0.017)	(0.016)	(0.016)
Current Account Balance	0.296***	0.274***	0.253***	-0.203
	(0.045)	(0.053)	(0.052)	(0.139)
Consumer Price Index	-0.001	-0.002+	-0.002+	-0.006
	(0.000)	(0.001)	(0.001)	(0.004)
Interest Rate Spread	-0.409***	-0.418***	-0.430***	-0.439***
	(0.065)	(0.074)	(0.072)	(0.077)
Electricity × CAB				0.006***
				(0.002)
CPI × Interest Spread				0.000
				(0.000)
Num.Obs.	544	431	452	544
R2	0.420	0.390	0.391	0.439
R2 Adj.	0.415	0.384	0.386	0.432

The temporal dynamics addressed in my fourth research question are illuminated through comparison of contemporaneous (Model 1) and lagged specifications (Models 2-3). The stability of coefficient magnitudes across these models indicates that the relationships between economic conditions and business environments reflect enduring structural patterns rather than transient correlations. This temporal consistency enhances confidence in these relationships as reliable predictors of business environment quality.

The interaction model (Model 4) provides critical insights into the third research question by revealing conditional relationships between infrastructure and economic stability indicators. The significant positive interaction between electricity access and current account balance demonstrates that infrastructure improvements yield enhanced regulatory benefits when supported by macroeconomic stability. This synergistic relationship suggests that effective policy approaches should integrate infrastructure development with sound macroeconomic management rather than pursuing these objectives in isolation.

While the models do not directly incorporate governance indicators or regional fixed effects (addressing research questions two and three), the substantial unexplained variance (approximately 57% in the best-fitting model) suggests important roles for institutional quality and regional dynamics beyond the economic factors explicitly modeled. Future research should extend this analysis by incorporating both governance measures and regional heterogeneity to develop a more comprehensive understanding of business environment determinants across diverse development contexts.

5. Results, Analysis, and Discussion

Comparative modeling reveals significant complexity in the determinants of business environment quality across global economies. Gradient boosted trees demonstrate markedly superior predictive performance compared to linear regression models (R² = 0.88 versus 0.64; RMSE = 3.99 versus 6.99), indicating substantial non-linearities and interaction effects in the relationships between economic indicators and regulatory environments. This performance differential persists across validation approaches, suggesting robust underlying patterns rather than statistical artifacts.

The machine learning approach’s enhanced predictive capacity directly addresses my first research question concerning infrastructure and economic stability indicators. While these factors significantly influence business environment quality (as demonstrated in Table 2), their effects manifest through complex functional forms that traditional linear specifications inadequately capture. Gradient boosted trees likely identify critical threshold effects in these relationships—particularly evident in infrastructure indicators, where electricity access appears to yield marginal benefits that vary considerably across development contexts.

Figure 6. Comparative model performance for predicting Ease of Doing Business scores. The left panel displays predicted versus actual EDB scores using linear regression (RMSE = 6.99, R² = 0.64), while the right panel shows results from gradient boosted trees (RMSE = 3.99, R² = 0.88). Points aligned with the red dashed line (perfect prediction) indicate accurate model forecasts. Gradient boosted trees demonstrate superior predictive accuracy with reduced scatter, capturing approximately 88% of variance in business environment quality compared to 64% for linear regression. This substantial improvement suggests significant non-linear relationships and interaction effects among predictors that traditional linear models cannot adequately capture, particularly at extreme values of the EDB distribution.

Regional heterogeneity in predictive accuracy (Figure 7) offers compelling evidence regarding my second research question on geographic patterns. Both modeling approaches exhibit systematic variation in performance across regions, with particularly pronounced accuracy differentials in South Asia (MAE differential of 3.6 points). This regional variation suggests regionally distinct determinant structures, potentially reflecting historical institutional trajectories and cultural factors not explicitly incorporated in my economic variables.

The interaction effects identified in the regression analysis—most notably between electricity access and current account balance (β = 0.006, p < 0.001, Table 2)—receive further validation through the superior performance of tree-based models that inherently accommodate complex variable interactions. This addresses my third research question by confirming that governance factors and economic indicators function synergistically rather than independently in shaping regulatory environments. The gradient boosted tree’s flexible specification captures these interdependencies without requiring explicit pre-specification.

Figure 7. Regional variation in predictive accuracy across modeling approaches. The bar chart displays Mean Absolute Error (MAE) by region for both linear regression (teal) and gradient boosted trees (red). Lower values indicate better predictive performance. The machine learning approach consistently outperforms linear regression across all regions, with the performance gap most pronounced in South Asia (5.5 vs. 9.1 MAE). Both models demonstrate superior accuracy in Sub-Saharan Africa and East Asia & Pacific regions compared to others. The systematic regional variation in model accuracy suggests that determinants of business environment quality follow regionally distinct patterns, with some regions exhibiting more complex, non-linear relationships than others.

My findings yield three main conclusions. First, business environment quality emerges from complex, non-linear relationships among infrastructure, macroeconomic stability, and institutional factors, with important threshold effects and diminishing returns. Second, regional context substantially moderates these relationships, suggesting the limited efficacy of universalistic policy prescriptions. Third, the interaction between infrastructure development and macroeconomic stability indicates potential complementarities in reform implementation that merit further investigation.

Several limitations needs to be acknowledged. The models incorporate a constrained set of economic indicators, potentially omitting relevant variables including trade openness, human capital development, and political stability metrics. While gradient boosted trees effectively capture non-linearities, they sacrifice interpretability, complicating the derivation of specific policy recommendations. The regional analysis, though revealing important patterns, necessarily treats regions as relatively homogeneous units, potentially obscuring significant intra-regional variation in business environment determinants.

Future research would benefit from incorporating additional dimensions including political economy factors, regulatory implementation quality, and subnational business environment variation. Extended longitudinal data would facilitate more robust analysis of reform sequencing and path dependence effects. Complementary qualitative case studies of specific reform episodes would provide valuable contextual understanding of the mechanisms linking economic conditions, governance quality, and regulatory outcomes across diverse development contexts.

6. Impact

This analysis suggests several actionable policy priorities for improving business environments globally. First, infrastructure development—particularly expanding electricity access—emerges as a foundational prerequisite for business-friendly environments, with potentially amplified benefits in countries maintaining strong external balances. Second, financial sector reforms that reduce interest rate spreads appear particularly impactful, suggesting that banking competition and credit market efficiency merit special attention. Third, the better performance of non-linear models indicates that policymakers should focus on country-specific thresholds rather than universal benchmarks when designing reforms.

These findings align with recent shifts in development thinking that emphasize contextual factors over one-size-fits-all prescriptions. If implemented, such targeted policies would primarily benefit small and medium enterprises by reducing operational uncertainty and costs, potentially unlocking entrepreneurial activity in sectors previously constrained by infrastructure limitations. The primary beneficiaries would likely include businesses in underserved regions, though established firms might face increased competition.

If accurate, these models could inform more efficient allocation of development resources by identifying specific binding constraints in each country’s business environment. However, misspecification could misdirect reform efforts toward areas with limited impact. The most significant risk involves overemphasizing infrastructure and macroeconomic factors while neglecting corruption or bureaucratic inefficiency, which might entrench existing power imbalances. Given these models explain up to 88% of variation in business environment quality, they offer substantial but incomplete guidance, highlighting the need for complementary institutional analysis when designing comprehensive reform programs.