Machine learning reveals which economic policies really drive human development in the world’s richest nations

What makes a country truly developed? For decades, governments have chased gross domestic product figures as if they were the ultimate scoreboard, yet a growing body of research insists that the real measure of national success is people: how long they live, how much they learn, and how comfortably they live. Now, a team of researchers from Arden University, Manchester Metropolitan University, the University of Hafr Al Batin, MIT, and Taif University has combined classical statistics with machine learning to dissect exactly which fiscal and monetary policy levers move the Human Development Index (HDI) in the world’s most developed nations. Their findings, published in Frontiers in Human Dynamics, challenge several assumptions that have shaped economic policymaking for years.

The study, led by Andre Amaral of Arden University and Taysir E. Dyhoum of Manchester Metropolitan University, analyzed cross-sectional data from 49 countries with the highest HDI rankings, from Switzerland at 0.962 down to Kuwait at 0.831. Liechtenstein was excluded because key variables were unavailable. The researchers drew on data from the United Nations Development Programme and the World Health Organization, examining nine independent variables: gross domestic product, GDP per capita, inflation, the debt-to-GDP ratio, interest rates, unemployment, average income tax rates, life expectancy, and expected years of schooling. The dependent variable, the HDI, is a composite measure introduced in the 1990 UNDP Human Development Report by Pakistani economist Mahbub ul Haq, combining life expectancy at birth, expected years of schooling, and gross national income per capita on a scale from 0 to 1.

What sets this research apart is its dual methodology. The team deployed Ordinary Least Squares (OLS) regression alongside supervised machine learning techniques, splitting the dataset into a 70 percent training set and a 30 percent test set using the caTools package in R. This random split, a widely recognized approach in machine learning, helps limit selection bias and reduces researcher discretion. The OLS framework estimates each coefficient by minimizing the sum of squared residuals, providing an interpretable measure of each variable’s marginal contribution to HDI while holding all other factors constant. The machine learning component then tests whether the model generalizes to unseen countries, guarding against overfitting, the phenomenon where a model captures random noise rather than genuine signal.

The Pearson correlation matrix revealed striking patterns. HDI correlates most strongly with life expectancy at 0.754, followed by GDP per capita at 0.662 and expected years of schooling at 0.552. Aggregate GDP, by contrast, shows only a weak correlation of 0.162, suggesting that the sheer size of an economy matters far less than how its wealth is distributed per person. The debt-to-GDP ratio shows essentially no correlation with HDI at 0.038, a finding that seems to let governments off the hook regarding fiscal discipline. But the story does not end there.

When the researchers ran the full OLS regression, only three variables emerged as statistically significant predictors of HDI: GDP per capita, life expectancy, and expected years of schooling. GDP per capita, life expectancy, and education were significant at conventional levels, while inflation, the debt-to-GDP ratio, interest rates, unemployment, and average income tax rates were not. The model achieved an adjusted R-squared of 0.8826, meaning it explains nearly 88 percent of cross-country variation in HDI within this high-HDI sample. The residual standard error of 0.01322 on 25 degrees of freedom indicates remarkable precision, and the joint F-statistic of 29.41 decisively rejects the null hypothesis that all slope coefficients are zero.

Diagnostic testing confirmed the model’s validity. A quantile-quantile plot showed residuals approximating a normal distribution with only mild tail deviations. The Breusch-Pagan test for heteroskedasticity returned a p-value exceeding 0.05, meaning the researchers failed to reject the null hypothesis of constant error variance, a desirable outcome that strengthens confidence in the standard errors and associated inference. A residuals-versus-leverage plot confirmed that no observation exceeded Cook’s distance threshold, ruling out influential outliers that could distort coefficient estimates. The model then achieved outstanding precision on the test data, with predicted HDI values closely matching actual figures for countries such as Switzerland, Denmark, Germany, and Japan.

But the most intriguing finding lies in what statisticians call suppression effects. A suppressor variable has a weak or non-significant direct relationship with the dependent variable, yet when included in the regression model, it increases predictive power by removing unexplained variance. The team tested whether inflation, unemployment, interest rates, and the debt-to-GDP ratio act as suppressors by comparing the GDP per capita coefficient with and without each suspected variable. The results were remarkable: the interest rate and the debt-to-GDP ratio both act as suppressor variables, strengthening the GDP per capita coefficient when included, while inflation and unemployment do not. When the interest rate was included, the GDP per capita coefficient rose from 0.000000441 to 0.000000445. When the debt-to-GDP ratio was included, it rose from 0.000000439 to 0.000000445.

This suppression analysis reveals that fiscal and monetary policies influence human development through channels that simple bivariate analysis cannot detect. The debt-to-GDP ratio, though showing no direct correlation with HDI, operates indirectly through its relationship with GDP per capita. High debt can constrain fiscal space, increase financing costs and risk premia, and reduce private investment and income growth, thereby lowering the per-capita income that drives HDI. Similarly, interest rates and inflation, though not directly significant, negatively affect GDP per capita, which is in turn moderately correlated with HDI. The researchers also found evidence of indirect links between the debt-to-GDP ratio, inflation, unemployment, and HDI: these variables are negatively correlated with GDP per capita, meaning increases in them may be associated with lower per-capita income and, through that channel, lower human development.

The study’s practical implications are significant. The researchers recommend prudent fiscal policies aimed at reducing government debt to levels that limit inflation risk and support low, stable interest rates. Notably, they challenge the view that rising debt improves HDI, suggesting instead that sustainable investment in public services should be pursued alongside fiscal balance. Their data show that high-HDI countries maintain an average debt-to-GDP ratio of 57 percent, compared with 71 percent for lower-ranked countries, and they advise keeping the ratio below 60 percent. The team also demonstrated the model’s predictive utility by projecting the United Kingdom’s HDI under hypothetical scenarios: if GDP per capita increased by 3 percent and life expectancy and years of schooling each increased by 2 percent, the UK’s HDI would improve from its current 0.93, potentially raising its rank from 17th.

The research also highlights the importance of considering both direct and indirect relationships when evaluating macroeconomic policy impact. By integrating interpretable statistical approaches with machine learning techniques, the study provides a comprehensive assessment of human development determinants that earlier single-country or GDP-focused studies missed. The authors point to several directions for future work, including expanding the dataset to incorporate technological innovation, income distribution, and social capital, conducting longitudinal analyses to clarify causal links over time, and deploying advanced machine learning methods such as decision trees, random forests, and neural networks to detect non-linear patterns that traditional regression may overlook. For now, the message is clear: in the world’s most developed nations, the path to higher human development runs through per-capita prosperity, health, and education, with fiscal discipline and monetary stability operating as essential but hidden enablers behind the scenes.

Subject of Research: Statistical and machine learning analysis of fiscal and monetary policy determinants of human development in high-HDI countries

Article Title: Statistical and machine learning analysis of fiscal and monetary policy determinants of human development in high-HDI countries

Article References: Amaral, A., Dyhoum, T. E., Alqahtani, M., Alharthi, A. S., & Aljohani, H. M. (2026). Statistical and machine learning analysis of fiscal and monetary policy determinants of human development in high-HDI countries. Frontiers in Human Dynamics, 8, Article 1874322. https://doi.org/10.3389/fhumd.2026.1874322

Image Credits: AI Generated

DOI: 10.3389/fhumd.2026.1874322

Keywords: Human Development Index, GDP per capita, fiscal policy, monetary policy, machine learning, OLS regression, public debt, inflation, interest rates, life expectancy, education, suppressor variables