Why Do Some Countries Win So Many Olympic Medals?
Wealth, Cold Climates, and the Political Economy of Sporting Success
The 2026 Winter Olympics, held in Milan and Cortina d'Ampezzo from February 6 to 22, drew significant global attention. The Games were the most-watched Winter Olympics in the United States since 2014, averaging around 23.5 million viewers on NBC and affiliated platforms, a 96 percent increase over Beijing 2022.
On the field, two outcomes stood out. Norway finished at the top of the Winter medal table once again. Italy, as host, recorded one of its strongest Winter performances on record, winning 10 gold, 6 silver, and 14 bronze medals, the most medals the nation has ever achieved in the Winter Olympics.
The results beg two questions that come up every Games: why do the same countries keep winning, and why does hosting help so much?
To answer that, I built statistical models of Olympic medal performance across modern Summer and Winter Games, adjusting for population size and team-event inflation, and incorporating economic, geographic, and climatic variables. I also examine whether national IQ predicts medal performance independently of long-run income.
Data and Sample
Olympic medal data are drawn from the Kaggle dataset, which covers Winter Games through 2014 and Summer Games through 2016. Team events are deduplicated to match official medal counts, and medals are weighted 3–2–1 (Gold–Silver–Bronze). The analysis is restricted to:
Summer Games: 1992–2016
Winter Games: 1994–2014
GDP per capita and population data come from the World Bank’s World Development Indicators. Climate variables are derived from WorldClim global temperature rasters, aggregated to country-level area-weighted means. Mountain share is calculated from global elevation data as the proportion of land area above 1,000 meters. National IQs were obtained from Parra & Kirkegaard, 2025.
All models are negative binomial regressions with log population offsets.
Structural Results
Because the models use a log link, coefficients are reported as incidence rate ratios (IRRs). An IRR expresses the multiplicative change in medal output associated with a one-unit increase in a predictor. An IRR of 2.00 implies medal output doubles; an IRR of 1.50 implies a 50% increase.
In the Winter Olympics, long-run GDP per capita has an IRR of approximately 3.97 (p < 0.001). That magnitude is large: moving from a lower-income to a substantially higher-income country corresponds to nearly quadruple medal output per capita. Winter temperature is also significant (p = 0.028), with warmer climates associated with lower performance. Once income and climate are controlled, mountainous terrain has no independent effect (p = 0.52).
The Summer Olympics show the same economic pattern without the climate effect. Long-run GDP per capita has an IRR of approximately 2.08 (p < 0.001), roughly doubling medal output. Average temperature is not statistically significant (p = 0.22).
Across both seasons, income is the dominant structural predictor. Climate contributes in Winter; hosting produces a temporary boost; mountains largely disappear once income is included. The scale of the GDP coefficient is substantially larger than any other variable in the model.
The figure below summarizes the core relationships: income and medal output in both seasons, the climate gradient in Winter performance, and the host-country boost.
Panel Results: Within vs Between Effects
The structural models pool performance across Games. The panel models instead separate long-run income differences from short-term economic fluctuations within countries.
In the Summer Games (1992–2016), only the long-run income component matters. Countries that are persistently wealthier win more medals. Year-to-year deviations from a country’s income trend have no detectable effect (p = 0.59). Olympic performance in Summer reflects structural capacity, not temporary economic cycles.
The Winter Games (1994–2014) differ. Long-run income remains significant (p = 0.001), but short-term income fluctuations also predict medal output (p = 0.0006). Winter performance appears more responsive to recent economic conditions. This is consistent with the capital-intensive nature of winter sports and the importance of current infrastructure and training investment.
The host-country effect remains strong in both seasons. In the Summer Games, hosting corresponds to an estimated 74% increase in medal output. In Winter, the increase is approximately 43%. Unlike income, however, this effect is temporary rather than structural.
Climate behaves differently across specifications. In the Summer panel model, warmer climates are associated with lower medal output even after controlling for income (p < 0.001). In the Winter panel, colder climates are associated with higher performance, though the effect is marginal in this specification and stronger in the structural aggregation.
The role of national IQ and its mediation through income requires more careful treatment. That analysis follows below.
National IQ and Mediation
National IQ also correlates strongly with medal performance when entered alone. In the Winter structural model without GDP, a one–standard deviation increase in IQ is associated with an IRR of 2.11 (p = 0.027). In the Summer model, the corresponding IRR is 1.99 (p < 0.001).
That apparent effect does not survive the inclusion of income.
Once long-run GDP per capita is added, the IQ coefficient falls close to zero in both seasons. In the Winter model, IQ’s coefficient drops from 0.75 (log scale) to 0.10 (p = 0.63). In the Summer model, it becomes slightly negative and statistically insignificant (p = 0.59). GDP remains highly significant in both cases (p < 0.001).
The indirect (mediated) effect of IQ through GDP is sizeable. The product-of-coefficients calculation corresponds to approximately a 76% increase in Winter medal output (exp(a×b) ≈ 1.76) and more than a doubling in Summer (≈ 2.15). Statistically, the IQ–medals association is fully mediated by income.
This does not establish a causal ordering between IQ and GDP; cross-national data cannot resolve that. What the models demonstrate is narrower: conditional on national income, IQ does not independently predict Olympic performance. Income is the proximate determinant.
Conclusion
Long-run national income is the dominant structural predictor of Olympic performance in both Summer and Winter Games. In the Winter structural model, the coefficient on log GDP per capita corresponds to an IRR of approximately 3.97 (p < 0.001). Because GDP is entered in logarithmic form, a one-unit increase represents roughly a 2.7-fold increase in income. Moving from a middle-income country to a country with roughly 2.7 times higher GDP per capita is associated with nearly quadruple medal output per capita.
In the Summer model, the IRR on log GDP per capita is approximately 2.08 (p < 0.001). The same income increase is associated with roughly double the medal output.
Sustained wealth finances training systems, coaching depth, sports science, athlete development pipelines, and institutional continuity. Olympic success is expensive and cumulative. Countries do not improvise elite performance; they build it over decades.
Climate plays a role in the Winter Games. Warmer winter temperatures are associated with significantly lower medal output, even after controlling for income. The likely mechanism is not temperature per se, but exposure and opportunity. Cold climates make winter sports culturally embedded and physically accessible. Children grow up skiing, skating, and competing outdoors. Facilities are integrated into daily life rather than constructed artificially. That early exposure compounds over time.
Mountains, by contrast, do not independently predict success once wealth and climate are accounted for. Terrain without infrastructure is insufficient.
The host-country effect is real but temporary. In the panel models, hosting increases medal output by roughly 40–70 percent depending on season. This is due to mechanisms suchg as automatic qualification slots, increased funding cycles leading up to the Games, home-field advantages, and judging effects in some events.
National IQ shows a strong association with medal output when examined alone. However, once long-run GDP per capita is introduced, the IQ coefficient disappears. The entire statistical association operates through income.
In short, the medal table reflects accumulated state capacity, long-term investment and, in the Winter Games, cultural immersion in cold-weather sport.