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

1 Introduction: research background & motivation

Economic growth alone does not necessarily translate into improvements in human well-being. Over recent decades, policymakers and economists have increasingly recognized that development should be assessed beyond traditional economic indicators such as gross domestic product (GDP), incorporating broader dimensions including health, education, and living standards. The Human Development Index (HDI) is one of the most widely used multidimensional measures of development, combining indicators of income, education, and life expectancy. Understanding the macroeconomic factors that influence HDI remains an important challenge for governments seeking to achieve sustainable and inclusive development.

Policymakers and economists have worked to balance economic growth with human development. A central debate in this effort concerns the roles of the private sector and government in achieving this balance, (Mardikaningsih and Hariani, 2021; Ranis et al., 2000; Tan, 2022; Ul Haq, 1995). Both private and public sectors are expected to improve societal living standards, but the question remains: at what cost, and how should resources be allocated to achieve these gains? In the private sector, this often involves a trade-off in which firms sacrifice part of their profits to pay wages above the minimum wage, fund employee education, or offer better health insurance. On the other hand, it is widely agreed that governments must invest in their societies, but resources are limited. Like any economic agent, governments face financial constraints. In many cases, revenue collected through taxes is insufficient to cover expenses, resulting in public debt. Public debt is the total liabilities of a country, typically financed through the issuance of government bonds.

Against this backdrop, this paper focuses on the government’s role in human development, seeking to answer two critical questions: To explore these questions, we analyzed cross-sectional data from 49 countries with the highest HDI rankings. The HDI is an economic metric, calculated and published annually by United Nations Development Programme (2026), that measures development by focusing on the most critical element of the economy: people. The HDI was developed by Ul Haq (1995) [with contributions from Anand and Sen (1994)] and introduced in the 1990 UNDP Human Development Report. Hence, by leveraging data, this study provides empirical evidence that specific economic variables (such as GDP per capita, the debt-to-GDP ratio, and interest rates) can influence the HDI. The HDI has three key components: (1) Life Expectancy at Birth: The average lifespan of a newborn; (2) Expected Years of Schooling: The average number of years a child entering school is expected to spend in education; and (3) Gross National Income (GNI) Per Capita: The average income per person, measured in USD. The HDI ranges from 0 to 1, where 1 represents the highest possible score and 0 the lowest.

• Should governments prioritize human development at all costs?
• What is the relationship between public debt and the HDI?

The role of fiscal policies in promoting human development has been examined in various studies [e.g., Barro (1991), Man and Vatavu (2022), Sachs et al. (1995), and Shettigar et al. (2024)], while the impact of monetary policies has been explored in the context of specific countries or economies (Ann Lantion et al., 2023; Arestis et al., 2021; Blanchard and Sheen, 2013; Kisel’áková et al., 2020; Taylor, 1993). However, the existing literature often lacks a comprehensive, cross-country analysis that combines these policies with key HDI components such as life expectancy, education, and income levels. Moreover, many studies use simple regression techniques or focus on theoretical models without testing their assumptions through real-world data from high-HDI countries. Therefore, this paper fills these gaps by employing a novel combination of Ordinary Least Squares (OLS) regression and supervised machine learning (ML) techniques to examine the impact of fiscal and monetary policies on HDI across a global sample of 49 high-HDI countries. Unlike prior studies, which often treat economic factors in isolation, this research examines their interactions and suppressive effects, shedding light on how certain variables may amplify or dampen the effects of others on human development. This study not only provides deeper insights into the economic foundations of human development but also lays the groundwork for future research on the role of policy interventions in improving societal well-being.

The intersection of fiscal and monetary policies with human development has been a significant focus of economic research. Advances in technological tools, such as ML algorithms, have further expanded our understanding of this relationship, enabling more nuanced analysis. The HDI remains a central metric in academic discussions of poverty reduction and socio-economic progress. Several studies have examined the relationship between macroeconomic factors and the HDI. For instance, Cashin et al. (2001) examined variations in HDI across countries from 1975 to 1998, revealing a moderate association between economic growth and well-being, though they noted the ambiguous role of individual macroeconomic policies. In a more focused context, (Ridha and Parwanto, 2020) analyzed the relationship between macroeconomic factors and human development in Indonesia using time-series data from 1985 to 2015. They employed the Error Correction Model (ECM) and found that Gross Fixed Capital Formation had a positive long- and short-term impact on GDP growth and human development. Their results also highlighted the ECM model’s robustness, particularly in the presence of multicollinearity. Dzihny et al. (2023) examines the effect of specific macroeconomic variables, including inflation, unemployment, trade openness, and foreign direct investment, on HDI in Organisation of Islamic Cooperation (OIC) countries. It finds that trade openness has a significant positive effect on HDI, whereas foreign direct investment is significantly and negatively associated with HDI; inflation and unemployment are not found to have a statistically significant direct association with HDI in their sample.

Monetary policy’s role in economic development has been examined through various models. Building on earlier VAR-based work on monetary policy shocks, Bernanke and Gertler (1995) discussed evidence on the effects of monetary policy shocks on output and prices, highlighting the role of interest rates in the transmission mechanism. Later, Amaral et al. (2022) found that expansionary monetary policy positively affects short-term GDP growth, but in the long term, it increases inflationary pressure and increases government debt. Yolanda (2017) focused on Indonesia, using time-series data from 1997 to 2016 to examine the relationships between macroeconomic variables, including the Indonesian Bank Rate, oil prices, and inflation, and their impacts on HDI and poverty. While the model demonstrated that inflation significantly impacts HDI (with a determinant coefficient of 18%), it also underscored the complex nature of inflation’s effects on human development. Further research by Yumashev et al. (2020) utilized a multiple linear regression model to explore the relationship between HDI and consumption quality, finding that these factors influence HDI ratings. Similarly, Khan et al. (2019) examined the role of information and communication technology (ICT) in human development in Pakistan. Their study, using ARDL and VECM models, highlighted that ICT positively influences economic and human development and recommends policies to further integrate ICT into economic strategies to foster growth. More recently, ML methods are increasingly used to estimate or predict HDI (Biaba et al., 2025; Georgieva, 2023; Raza and Ahmed, 2025). Still, relatively few studies use ML to examine the joint influence of fiscal and monetary policy indicators on HDI. Most studies have traditionally used GDP or GDP per capita as primary indicators of economic performance, leaving a limited understanding of the broader socio-economic outcomes that affect human development.

While these studies provide valuable insights, a significant gap remains in the literature on integrating fiscal and monetary policy analysis using advanced statistical techniques, particularly across multiple countries with high HDI rankings. Existing research has largely focused on traditional economic indicators, such as GDP or GDP per capita, which, while useful, fail to capture the multidimensional nature of human development. Moreover, existing studies have rarely combined machine learning techniques with interpretable statistical methods to examine the simultaneous influence of fiscal, monetary, and socioeconomic variables on human development. In particular, limited attention has been paid to identifying indirect relationships and suppressive effects, in which certain variables may alter the apparent contribution of other determinants after controlling for broader economic conditions.

This paper aims to address this gap by incorporating ML algorithms to assess the combined impact of fiscal and monetary policies on the key components of the HDI (life expectancy, education, and income). Specifically, as: This paper is structured as follows: the theoretical review and motivation in Section 1 explore prior research and existing knowledge on the topic. Methodology in Section 2 outlines the data collection process and statistical tools used. Data analysis Section 3 describes how the data were processed and interprets the findings. Finally, in Section 4, the study’s results are summarized and discussed in terms of their implications for current and future research.

• Quantify the relationships between key economic variables (e.g., GDP per capita, debt-to-GDP ratio, inflation, and interest rates) and HDI.
• Examine potential suppressive effects of selected variables to provide a more nuanced understanding of the associations between macroeconomic policy and HDI.
• Offer actionable insights for policymakers seeking to balance economic stability with human development.

2 Methodology

This study consists of secondary data from several data sources, mainly the United Nations Development Program (UNDP) https://hdr.undp.org/data-center/documentation-and-downloads and the World Health Organization at https://www.who.int/data/nutrition/nlis/info/human-development-index. The data is a cross-sectional series of the countries with the highest HDI (Table 1). The initial sample comprised 50 countries; however, we excluded Liechtenstein because some key variables required for the analysis were unavailable. The analysis uses HDI as the dependent variable, with the following independent variables: GDP, GDP per capita, inflation, the debt-to-GDP ratio, the interest rate, the unemployment rate, the average income tax rate, life expectancy, and expected years of schooling (education). At this stage, it is pivotal to clarify that all variables are endogenous. These variables were selected for their theoretical relevance and established connections to human development. For example, inflation serves as a proxy for monetary policy effectiveness, while the debt-to-GDP ratio and average income tax represent fiscal policy impacts. Life expectancy and education reflect the social dimensions of HDI.

Table 1. Top human development index (HDI) rankings by country for 2021, United Nations Development Programme (2026).

We use linear regression to assess whether high debt levels are positively correlated with HDI and whether a higher average income tax is more beneficial for achieving a higher HDI. The estimated model is also used for prediction, enabling us to examine the optimum levels of key variables associated with higher HDI within the sample. All analyses are implemented in R (see Supplementary Appendix for details), given its robustness and wide use for statistical computing.

2.1 The linear regression model

We first employ multiple linear regression to examine the statistical relationships between HDI and a set of macroeconomic and socioeconomic indicators. Linear regression provides an interpretable framework for estimating the contribution of individual explanatory variables while controlling for the influence of other factors (Gujarati, 2004). The general simple linear regression model can be expressed as:where Y is the response (dependent variable), x is a scalar predictor (independent variables), α is the intercept, β is the slope, and ε is a zero-mean disturbance term capturing other influences on Y (e.g., omitted factors, measurement error, and unobserved heterogeneity). Given observations {(xi,Yi)}i=1n, the ordinary least squares (OLS) estimators (α^,β^) minimizeThe first-order optimality conditions yield the closed-form estimators,andwhere x¯=1n∑i=1nxi, Y¯=1n∑i=1nYi, and ?>Equations 2, 3 show that OLS estimates the slope β^ from the centered sum of squares defined in (4), and the centered cross-product defined in (5) and then recovers the intercept α^ using the sample means.

Y=α+βx+ε,(1)

∑i=1n(Yi−α−βxi)2.

β^=SxySxx,(2)

α^=Y¯−β^x¯,(3)

Sxx=∑i=1n(xi−x¯)2=∑i=1nxi2−(∑i=1nxi)2n,(4)

Sxy=∑i=1n(xi−x¯)(Yi−Y¯)=∑i=1nxiYi−(∑i=1nxi)(∑i=1nYi)n.(5)

We extend (Equation 1) to a multivariate setting with p predictors,where μ is the error term. In this study, Y denotes the HDI, and the predictors are:

Y=α+∑j=1pβjxj+μ,(6)

x1 = Gross Domestic Product (GDP) [USD],

x2 = GDP per capita [USD],

x3 = Inflation [%],

x4 = Debt-to-GDP ratio [%],

x5 = Interest rate [%],

x6 = Unemployment rate [%],

x7 = Average income tax rate [%],

x8 = Life expectancy [years],

x9 = Expected years of schooling [years].

We include these covariates because they capture commonly studied links between macroeconomic conditions and human development. In particular, x1 and x2 capture the income-related dimension of development, with x2 providing the per-capita measure. We use x4 and x7 as fiscal indicators, x3 and x5 as monetary indicators. The variable x6 captures labor-market conditions, while x8 and x9 align with the health and education dimensions of HDI. All percentage variables are reported in percent, and HDI is unitless on a 0–1 scale

To enhance predictive accuracy and minimize overfitting, the dataset was split into a training set (70%) and a test set (30%), a widely recognized approach in ML (Joseph, 2022). The split was performed randomly using the caTools package in R, which helps limit selection bias and reduces the scope for researcher discretion (Williams et al., 2017). We implemented the analysis in R because of its strong support for regression, diagnostics, and visualization. The ML method for this analysis is the standard linear regression in RStudio. Linear regression is used to quantify the association between HDI and the explanatory variables, making it appropriate for examining correlations and exploring the levels of key predictors associated with higher HDI within the sample (Ali and Younas, 2021). Finally, relying on secondary data allows us to draw from widely used international sources (e.g., UNDP and the World Bank), supporting cross-country comparability and data consistency.

3 Data analysis and discussion

This part starts with an analysis of each variable used in the research, followed by applying the Pearson correlation and the linear regression tests. The top 49 nations with the highest HDI as of 2021 are listed in Table 1. In Figure 1, we visualize the largest HDI countries listed in Table 1 by a regional distribution map. We observe clustering of the countries with the highest HDI along the western side of Europe.

Using HDI data from the UNDP Reports, we first inspect pairwise relationships between HDI and key macroeconomic indicators (Figure 2). Panel A suggests a weak positive association between aggregate GDP and HDI: countries with larger GDP (larger markers) are mostly observed at HDI values above 0.9. The same panel also shows a clearer upward trend between GDP per capita and HDI, indicating that higher income per person is generally associated with higher human development in this sample. Panel B examines inflation, which can erode purchasing power and adversely affect households (Anyars and Adabor, 2023). Visually, lower inflation rates tend to coincide with higher HDI values, although this pattern should be validated quantitatively using correlation measures. Finally, Panel C plots the debt-to-GDP ratio against HDI and does not reveal a clear monotonic relationship in the data. As discussed in the introduction, one of the key questions this work addresses is whether fiscal policy affects HDI. While some governments are known for austerity measures and others for more expansionary policies, Panel C does not visually answer this question. From a theoretical perspective, the debt-to-GDP ratio can be linked to HDI through competing channels, so its net association need not be monotone. On the one hand, public borrowing may finance productive spending on health, education, and infrastructure, which can improve HDI components. On the other hand, high debt can constrain fiscal space, increase financing costs and risk premia, and through tighter macroeconomic conditions, might reduce private investment and income growth. In a cross-sectional setting, these offsetting mechanisms can yield a weak direct association between debt-to-GDP and HDI, while still allowing for indirect effects operating through income-related variables (e.g., GDP per capita) and other macro indicators. Figures 2, 3 provide a first-pass, descriptive view of how HDI relates to the main macroeconomic and policy indicators in our dataset. We next complement the patterns suggested by these scatter plots with formal correlation and regression results (Tables 2, 3 later).

Table 2. Pearson correlation matrix for macroeconomic and human development variables (lower triangle shown).

Positive values indicate that two variables tend to move together, while negative values indicate an inverse association. The HDI is most strongly associated with life expectancy (r=0.754), GDP per capita (r=0.662), and expected years of schooling (r=0.552), suggesting that better health/education outcomes coincide with higher overall development in this sample.

Table 3. OLS regression of HDI on macroeconomic and policy indicators. GDP per capita, life expectancy, and expected years of schooling are statistically significant predictors, while other covariates are not significant in this specification.

Significance codes: 0 “*” 0.001 “” 0.01 “*” 0.05 “.” 0.1 “ ” 1

Residual standard error = 0.01322 on 25 degrees of freedom.

Multiple R2=0.9137; adjusted R2=0.8826

F-statistic = 29.41 on 9 and 25 DF, p-value =3.942×10−11.

Figure 3 summarizes pairwise relationships between HDI and additional policy and social indicators. Panel A suggests a negative association between the interest rate and HDI, with higher HDI values occurring more frequently at lower interest rates. This pattern is descriptive and should not be interpreted causally. Panel B shows a weak-to-moderate positive association between the average income tax rate and HDI in this sample. Panels C and D display strong positive relationships between HDI and life expectancy and between HDI and expected years of schooling, respectively. These trends are expected, since both life expectancy and education enter directly into the construction of the HDI. For this reason, we treat Panels C and D primarily as consistency checks and focus our policy interpretation on variables that are not mechanically embedded in the HDI definition.

Every government and most central banks aim to support economic stability, and maintaining low unemployment is a central objective because labour-market conditions directly affect household income and welfare. Figure 4 plots HDI against the unemployment rate for the 49 high-HDI countries in our sample. It shows a cluster of dots around the 5% unemployment rate as the HDI approaches 1 (higher). The scatter suggests a weak negative association: higher unemployment rates are more often observed alongside lower HDI values, although the relationship is not strictly monotone and includes several outliers. As with the other bivariate plots, this visual pattern is descriptive and is evaluated more formally using the Pearson correlation matrix and the regression results reported later. For the fiscal policies mentioned earlier in this article, we also collected taxation data for 49 countries and calculated the average. The aim is to determine whether potential tax increases can promote human development or whether there is no significant relationship between them.

Table 4 summarizes the descriptive statistics for all variables in the dataset. These descriptive patterns motivate the correlation and regression analyses that follow. In Panel A, GDP shows a large gap between its mean and median (mean ≈$1.2 trillion vs. median ≈$397 billion), indicating a strongly right-skewed distribution in which a small number of very large economies pull the mean upward. In this dataset, the United States has the largest GDP and appears as an outlier: its GDP exceeds that of Japan (the second-largest economy in the sample) by approximately $18.4 trillion. To characterize dispersion around the mean, we report the variance, standard deviation, and coefficient of variation (CV). The large standard deviation and high CV for GDP indicate substantial variability, suggesting that GDP is unlikely to be normally distributed in this sample. This is further supported by the skewness statistic, which exceeds the ±2 threshold, and by the kurtosis value of 39.03, well above the ±7 threshold reported by Hair et al. (2010). In contrast, GDP per capita exhibits more stable behavior: the mean and median are relatively close, dispersion is lower (CV ≈0.57), and skewness and kurtosis remain within the stated bounds. For inflation, the most notable feature is the high kurtosis (32.89), indicating a heavy-tailed distribution with outliers. The remaining variables exhibit comparatively low variability and show no strong departures in skewness or kurtosis. Panel B reports labour and human-development variables. Compared with macroeconomic indicators, life expectancy, expected years of schooling, and HDI exhibit relatively low variability, as expected given that the sample is restricted to high-HDI countries. Overall, Table 4 provides context on scale, dispersion, and outliers, which helps interpret the correlation patterns reported in Table 2.

Table 4. Descriptive statistics for the variables in the dataset.

Table 5.

The Pearson correlation matrix reported in Table 2 summarizes the strength and direction of linear associations between pairs of variables. Correlations range from −1 to 1: values close to 1 indicate a strong positive linear relationship, values close to −1 indicate a strong negative linear relationship, and values near 0 suggest little to no linear association. In our sample, HDI is weakly positively correlated with aggregate GDP (r=0.162), implying that countries with larger GDP tend to have slightly higher HDI, although the association is modest. By contrast, HDI is strongly positively correlated with GDP per capita (r=0.662) and shows essentially no correlation with the debt-to-GDP ratio (r=−0.038). HDI is also moderately positively correlated with the average income tax rate (r=0.424) and expected years of schooling (r=0.552), and it is most strongly correlated with life expectancy (r=0.754). Conversely, HDI is moderately negatively correlated with inflation (r=−0.316), interest rates (r=−0.294), and unemployment (r=−0.320).

The OLS regression estimates for HDI are presented in Table 3. Each coefficient measures the marginal association between the corresponding predictor and HDI, holding the remaining variables constant. The intercept α in Equation 6 represents the predicted HDI when all covariates are set to zero. The standard errors quantify sampling uncertainty in the coefficient estimates; the reported t-statistics test the null hypothesis that a coefficient equals zero, and the corresponding p-values summarize statistical significance. The results indicate that the intercept is statistically significant at the 0.1% level (p=1.31×10−4), so we reject the null hypothesis H0:α=0. This indicates that the baseline predicted HDI value when all covariates are set to zero differs from zero. Similarly, GDP per capita is statistically significant (p<0.001), suggesting a positive association with HDI after controlling for the other covariates. Life expectancy is also significant (p=0.006), and expected years of schooling is highly significant (p<10−6), consistent with the fact that both variables are included in the construction of the HDI. Conversely, GDP, inflation, the debt-to-GDP ratio, interest rates, the unemployment rate, and the average income tax rate are not statistically significant at conventional levels; therefore, we cannot reject the null hypothesis that their coefficients are zero in this specification. Life expectancy and expected years of schooling are, as expected, linearly related to HDI, with significance levels below 0.1%. At this stage, we also find evidence of indirect links between the debt-to-GDP ratio, inflation, and unemployment and HDI: These variables are negatively correlated with GDP per capita, which is a statistically significant predictor of HDI. Consequently, increases in these variables may be associated with lower GDP per capita and, through that channel, lower HDI. Overall model fit is strong: the residual standard error is 0.01322, and the adjusted R2 is 0.8826, indicating that the model explains a large share of cross-country variation in HDI within this high-HDI sample. The joint significance test rejects the null that all slope coefficients are zero (F-statistic p=3.942×10−11).

To assess the adequacy of the linear regression model, we examine standard diagnostic assumptions about the error term. First, we evaluate the normality of residuals using a quantile–quantile (Q–Q) plot (Figure 5). In Panel A, the points lie close to the reference line, suggesting that the residual distribution is approximately normal, with only mild deviations in the tails. Panel B provides a complementary check by scale–location plot, displaying the spread of standardized residuals across fitted values; the dispersion is moderately variable but does not show a strong systematic pattern. Second, we test for homoscedasticity (constant error variance), which is important for reliable inference based on OLS standard errors. We use the standard Breusch–Pagan (BP) test (Breusch and Pagan, 1979), which evaluates the null hypothesis of constant variance against the alternative of heteroscedasticity. The test is implemented via an auxiliary regression of squared residuals on the regressors, with the resulting statistic evaluated under a χ2 distribution (Astivia and Zumbo, 2019). The BP test results are reported in Table 5. Since the p-value exceeds the 0.05 significance level, we fail to reject the null hypothesis of homoscedasticity. This is a desirable outcome, as it indicates no evidence of heteroscedasticity in the model and suggests that the assumption of constant error variance is reasonably satisfied. Consequently, this strengthens confidence in the validity of the corresponding standard errors and associated statistical inference.

Table 5. Breusch-Pagan (BP) test for heteroskedasticity.

The null hypothesis is that errors are homoskedastic (constant variance). The large p-value indicates that we do not reject the null at conventional significance levels, providing no evidence of heteroskedasticity in the fitted model.

Another important aspect to examine is the effect of outliers in the dataset. This can be assessed using the Residuals vs. Leverage plot (Figure 6). In this figure, any observation lying beyond Cook’s distance, marked by the dotted lines, may indicate an outlier with a substantial influence on the regression coefficients. However, this is not the case here, since none of the data points exceeds the threshold. As the model therefore appears stable and reliable, it can now be evaluated on the test data to generate predictions and assess possible overfitting. If the model is overly fitted to the training data, it may begin to capture random noise, thereby reducing its ability to accurately predict unseen observations. Accordingly, large discrepancies between predicted and actual values may indicate overfitting.

The model achieves high and outstanding precision on the test data, as shown in Table 6, supporting its reliability and aligning with the residual standard error reported in Table 3. Accordingly, if the model is accurate, the regression equation can be used to identify the corresponding levels of GDP per capita, life expectancy, and years of education. Taking the United Kingdom as an example and substituting into Equation 6 the statistically significant variables identified by the model, we obtain the projected HDI (Equation 7) under the following assumptions: GDP per capita increases by 3%, from 47,334.36 to 48,754.39; life expectancy increases by 2%, from 80.7422 to 82.357; and years of schooling also increase by 2%, from 17.3097 to 17.6559.Using the residual standard error of 0.01322 reported in Table 3, the approximate prediction interval corresponding to equation (7) is:Here, x*=(GDP per capita,life expectancy,years of schooling)= (48,754.39,82.357,17.6559) denotes the assumed future values of the explanatory variables. Since the United Kingdom is currently ranked 17th in Table 1 with an HDI of 0.93, this projection suggests a potential improvement in its HDI under the assumed scenario.

HDI^UK(x*)=4.389×10−1+(5.328×10−7)(48,754.39)+(3.729×10−3)(82.357)+(9.754×10−3)(17.6559)=0.9442.(7)

HDI^UK(x)±0.01322,⇒HDI^UK(x)∈[0.93098,0.95742].

Table 6. Comparison between predicted and actual HDI values for test-set countries, showing that the model’s predictions closely match the observed values.

Finally, given the stability and reliability of the model, it is possible to argue that even if there is no direct relationship between HDI and, for example, the debt-to-GDP ratio, an indirect relationship may still exist between the latter and GDP per capita, or perhaps between average income tax and GDP per capita, if both the debt-to-GDP ratio and average income tax per capita reflect an expansionary fiscal policy. Similarly, although interest rates and inflation do not show a direct relationship with GDP per capita, their indirect relationships should also be considered. Testing for indirect effects, also known as mediation, is therefore important, as noted by Loeys et al. (2015). Mediation analysis is commonly used for this purpose because it provides a more comprehensive understanding of the relationships among the variables in the model (Preacher and Hayes, 2004). Since, among the variables discussed above, only GDP per capita is statistically significant in the model, we conclude that mediation is absent. An alternative, therefore, is to test for suppression, which has long been studied, e.g., Ludlow and Klein (2014) and Mendershausen (1939). A suppressor variable has a weak or non-significant direct relationship with the dependent variable, but when included in the regression model, it increases predictive power. The test consists of estimating the regression with and without the suspected suppressor variable and comparing the coefficient β2 of GDP per capita (Tzelgov and Henik, 1991).

Table 7 demonstrates that inflation and the unemployment rate do not act as suppressor variables, whereas the interest rate and the debt-to-GDP ratio do. A suppressor effect occurs when adding the suspected suppressor strengthens the relationship between x2 (GDP per capita) and Y by removing unexplained variance in Y that is associated with the added variable but is not captured by x2 alone. Thus, Table 7 compares the GDP per capita coefficient when the suspected suppressor is included in the model and when it is excluded, using a logical test implemented in Excel with the “if” function. Hence, when included in the model, the interest rate and the debt-to-GDP ratio enhance the coefficient of GDP per capita. The code used in the paper is available in Supplementary Appendix.

Table 7. Logical suppressor test indicating that interest rate and debt-to-GDP ratio act as suppressor variables, whereas inflation and unemployment rate do not.

4 Conclusion

This study highlights the central role of people in national development and supports a balanced approach to economic growth and human well-being. Although GDP is often treated as a key measure of success, the analysis shows that it is only weakly associated with HDI components such as GDP per capita, life expectancy, and education. Instead, other macroeconomic variables, including inflation, the debt-to-GDP ratio, and interest rates, warrant greater attention given their more nuanced roles in human development.

The main findings show that inflation and interest rates negatively affect GDP per capita, which is, in turn, moderately correlated with HDI. Although central banks’ control of interest rates does not directly suppress HDI, keeping inflation low emerges as an important monetary policy objective. Fiscal policy, proxied by the debt-to-GDP ratio and average income tax, shows no direct statistical significance for HDI. However, the debt-to-GDP ratio displays a suppressor effect, highlighting the complex relationship between fiscal stability and economic growth.

Accordingly, the study recommends prudent fiscal policies aimed at reducing government debt to levels that limit inflation risk and support low, stable interest rates. The findings challenge the view that rising debt improves HDI and instead suggest that sustainable investment in public services should be pursued alongside fiscal balance. A debt-to-GDP ratio below 60% appears advisable, consistent with the data showing that high-HDI countries maintain an average ratio of 57%, compared with 71% for lower-ranked countries.

This study contributes to the literature on economic policy and human development by combining statistical and ML approaches. Linear regression and related statistical methods provide useful insights into the relationship between macroeconomic policies and human development outcomes. Unlike earlier studies, such as (Ridha and Parwanto, 2020; Sachs et al., 1995), which often focused on single-country cases or emphasized traditional indicators such as GDP, this study offers a broader cross-country perspective. It also identifies suppressor effects, particularly for the debt-to-GDP ratio, thereby providing a deeper understanding of the interaction between fiscal and monetary policies and HDI. By integrating statistical methods with ML, the study extends previous research by revealing complex relationships among macroeconomic factors and HDI components, including life expectancy, education, and income.

At the same time, this research points to several directions for future work. Expanding the dataset to include additional variables, such as technological innovation, income distribution, and social capital, could provide a more comprehensive view of human development. A longitudinal analysis of relationships over time could also help clarify causal links between fiscal and monetary policies and HDI components. In addition, advanced ML methods, such as decision trees, random forests, and neural networks, could detect non-linear patterns and interactions that traditional regression models may overlook. Such steps could further improve our understanding of how fiscal and monetary policies influence human development outcomes and offer more practical guidance for policymakers.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fhumd.2026.1874322/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Adejimi Adeniji, Tshwane University of Technology, South Africa

Reviewed by: Renea Shinta Aminda, Pakuan University, Indonesia

Haşmet Gökırmak, Istanbul Sabahattin Zaim University, Türkiye