Behind Uneven Income Growth: The Signals Hidden in Women’s Health and Education

The findings identify relationships worth investigating, without establishing that changing any particular indicator would directly increase economic growth.

Women's health and girls' education feature prominently in research exploring why income levels have followed different paths across developing countries. An analysis of 62 low- and lower-middle-income economies found that indicators describing women's causes of death carried some of the strongest information about long-term income trends. The findings identify relationships worth investigating, without establishing that changing any particular indicator would directly increase economic growth.

The research, titled "Predicting Long-Run Income Growth from Women's Indicators in Developing Countries," was published in Social Sciences and written by Manuel Chaves-Maza of Universidad Pablo de Olavide in Spain. The researcher examined 128 indicators from the World Bank Gender Data Portal, using available observations between 1960 and 2023 to describe changes in health, education, employment, population and empowerment. Its broad approach moves beyond a single gender-equality score to examine which aspects of women's lives contain the most useful information about national income patterns.

The strongest prediction model explained 38 percent of the differences in income trends across countries when tested on countries excluded from its training data. Health indicators occupied six of the ten leading positions, with changes in the injury-related share of deaths among women aged 15–59 ranking first. A surprising education pattern also emerged, as countries recording the fastest increases in girls' schooling often had flatter income trends, a relationship the paper connects to their disadvantaged starting positions.

Measuring Change Across 62 Economies

The analysis covered 39 health indicators, 22 education indicators, 26 labour and business measures, 33 population and demographic indicators, and eight empowerment measures. Each country received a trend estimate for every indicator, calculated from the observations available rather than by filling gaps with invented values. The World Economic Forum's Global Gender Gap Index provided background context and was excluded from the prediction models.

Income performance was measured as the average annual change in GDP per person in current US dollars, rather than as a percentage growth rate. Across the sample, this measure averaged about $31 per year, with Samoa and Angola recording the steepest increases and Guinea recording the only negative trend. Because current-dollar values reflect prices and exchange rates, the measure does not isolate real economic growth or reveal whether poorer households experienced improvements.

The researcher compared several machine-learning methods with a linear statistical model, repeatedly separating countries into training and testing groups. A random forest, which combines predictions from many decision trees, performed best, explaining 38 percent of variation against 29 percent for the linear benchmark. Results changed depending on how countries were divided, and roughly three-fifths of the variation remained unexplained, placing clear limits on the model's predictive reach.

Women's Mortality Provides the Strongest Signals

The leading predictor was the trend in the proportion of deaths among women aged 15–59 attributed to injuries, with faster increases associated with flatter income trends. This indicator was selected in every one of the 50 training rounds, strengthening confidence in its importance within this dataset. The next three predictors were the injury-related share of deaths among women of all ages, the share of working-age female deaths linked to communicable, maternal and nutritional causes, and girls' gross primary-school enrolment.

These mortality indicators describe the composition of deaths, rather than the number of women dying or their individual risk of death. An injury-related share can rise because deaths from infectious diseases decline, making it unsafe to interpret every increase as evidence that injuries have become more common. The underlying cause-of-death figures were modelled estimates available for only six years between 2000 and 2021, giving them a shorter observation period than the income data.

Two different methods agreed on the four leading predictors, giving those findings firmer support than the detailed ordering of other indicators. Rankings below those positions differed substantially, so the paper discourages treating the remaining list as a precise guide to importance. The results fit part of human-capital theory, which links health and education to productive capacity, without testing the mechanisms through which those relationships operate.

Why Faster Schooling Gains Accompanied Flatter Income Trends

All eleven education indicators that met the paper's initial correlation threshold were negatively associated with the income trend. Countries with the fastest improvements in female enrolment and educational gender parity generally recorded smaller annual dollar increases in GDP per person. The paper interprets this pattern mainly through starting conditions, since countries with low initial enrolment had greater room for rapid educational expansion.

Countries beginning with stronger female primary-school enrolment tended to record steeper income trends, making the pace of educational improvement different from the level already achieved. Child survival followed a related pattern, with female under-five mortality falling fastest in countries where it had initially been highest. These findings do not show that educating girls weakens economic performance, because the analysis compares overlapping national trends without establishing cause and effect.

An exploratory grouping separated 21 countries with slower schooling expansion and steeper income trends from 41 countries with faster schooling and child-survival gains but flatter income trends. The second group consisted largely of countries in sub-Saharan Africa and South Asia, where improvements frequently began from more disadvantaged conditions. The groups overlapped and changed under repeated checks, suggesting a continuum of development experiences rather than clearly separated national pathways.

Research Priorities Without Promises of Growth

The paper identifies preventable female deaths, female schooling and fertility-related constraints as priorities for research that can establish causal effects. Programmes addressing road and workplace injuries, violence against women and maternal complications could be evaluated for their economic returns alongside their health benefits. Whether such programmes increase growth, how much they cost and who benefits cannot be determined from these predictive rankings.

The analysis leaves open the possibility that rising incomes finance better health and schooling, or that political stability, institutions and conflict influence both social conditions and economic outcomes. Its small sample, uneven data coverage and overlapping measurement periods limit generalisation, and performance in other countries or future periods remains untested. Further work using inflation-adjusted income measures, indicators recorded before subsequent outcomes and changes within countries could assess whether the same patterns persist.

The processed data and analysis code are openly available, allowing other researchers to examine the findings and test alternative explanations. The central contribution is a clearer picture of which women's indicators carry information about income trajectories, with mortality patterns and primary schooling standing out most consistently. Any policy interpretation needs to recognise that higher average national income does not automatically reduce poverty or distribute opportunities more fairly.