Declining population and GDP growth

The paper investigates whether population decline necessarily leads to economic stagnation or recession. Motivated by longstanding debates from Malthus to Hansen, Keynes and Schumpeter, and more recent concerns about aging and secular stagnation, the authors examine how GDP, GDP per capita, unemployment, and labour force participation evolve in countries experiencing actual declines in population. The context includes historical and contemporary perspectives on population dynamics and growth, alongside environmental considerations associated with population size. The study’s purpose is to provide empirical evidence from 19 countries with recent population declines to assess long-run and short-run relationships between population changes and key macroeconomic outcomes, clarifying whether fears of economic decline are warranted.

The introduction surveys classical and modern views on population and the economy. Historically, Plato, Aristotle, More, and Malthus addressed optimal or stable population sizes, while Hansen (1939) warned that slowing or negative population growth could reduce investment and cause secular stagnation; Schumpeter (1942) countered that aging and decline need not constrain output. Postwar debates shifted from high population growth’s impact on per capita output (Kuznets 1967; Ehrlich 1968) to contemporary concerns about slowing growth rates and aging. Modern growth-theory contributions (Elgin and Tumen 2012; Jones 2022; Strulik 2022; Sasaki and Hoshida 2017) often find that with declining population, per capita output can still rise, frequently via human capital and technology channels. Related literatures consider environmental limits and optimal population (Daly; Ehrlich and Holdren; Daily et al.; Meadows et al.; Bradshaw et al.; Wackernagel et al.), and the macroeconomic effects of demographic change and aging on productivity and growth (Lindh and Malmberg; Feyrer; Acemoglu and Restrepo; Maestas et al.). The paper also outlines plausible mechanisms by which population decline could affect economies: supply-side labor scarcity, demand-side composition shifts, labor-saving technologies, migration and remittances, institutional reforms, foreign investment, and increased productivity, particularly salient for post-Soviet and transition economies.

Data: Panel of 19 countries with recent population decline or near-stability, many being former USSR members or Soviet-bloc/transition economies; Japan, Italy, and Portugal are included due to recent population declines despite 1990–2019 increases. Population changes over 1990–2019 range from −28% (Latvia) and −25% (Bosnia-Herzegovina) to −6% (Hungary) and −2.4% (Russia); Japan/Italy/Portugal saw modest increases over 1990–2019 but declines in the 2010s. Rates of change in GDP (total and per capita), population, labour force participation, and unemployment for 2000–2020 are tabulated. Data sources include World Bank (GDP, GDP per capita PPP 2017$, total GDP constant 2015 US$, population, CPI, external balance), ILO via World Bank (unemployment, labour force participation), Gapminder (historical population), UNECE/OECD/Eurostat/CEID (wages), with indicator codes provided.

Empirical strategy: The study estimates three long-run relationships using pooled mean group (PMG) estimation within an ARDL error-correction framework on panel data to account for cointegration among non-stationary variables: (1) per capita GDP as a function of population (with asymmetric decomposition), (2) per capita GDP as a function of labour force participation rate, and (3) total GDP as a function of the unemployment rate. Variables: GDP is per capita GDP (PPP constant 2017 USD); POP is total population; LABOR is labour force participation rate (% ages 15+, ILO estimate); TOT_GDP is GDP in constant 2015 USD; U_ILO is unemployment rate (ILO). Asymmetries in population are modeled via POP+ and POP− constructed using a mean threshold following Shin et al. (2014) and Greenwood-Nimmo et al., to capture differing effects of above/below-average changes.

Econometric procedures: Stationarity is assessed with the Im–Pesaran–Shin panel unit root test. Results indicate U_ILO is I(0) while GDP (per capita), TOT_GDP, LABOR, and POP asymmetry components are I(1), enabling PMG estimation with mixed I(0)/I(1) regressors. The ARDL lag order is selected by AIC, with country-specific preliminary selection and adoption of the most common lag structure across countries. Final dynamic specifications reported include ARDL(4,5,5) for the population–per capita GDP model, ARDL(1,4) for the labour–per capita GDP model, and ARDL(3,4) for the unemployment–total GDP model. PMG estimates long-run coefficients constrained to be homogeneous across countries while allowing heterogeneity in short-run dynamics, intercepts, and error variances; maximum likelihood estimation is used.

Model diagnostics: Dynamic stability is verified via negative and significant error-correction terms ϕ (e.g., −0.293 for the population model; −0.076 for the labour model; −0.125 for the unemployment model). Cointegration is inferred from the significance and sign of the error-correction coefficient. The homogeneity of long-run slopes is tested by Hausman-type tests comparing PMG and MG estimators; p-values (0.508; 0.282; 0.240 for the three models) support the PMG homogeneity restriction. Panel covers 19 countries, annual data, 32 time periods (1990–2021), 608 observations per model.

• Descriptive context (2000–2020): Despite population declines, most countries experienced sizable GDP and per capita GDP growth. Examples: Armenia GDP +202% (per capita +241%) with population −11%; Georgia GDP +159% (per capita +184%) with population −8.7%; Belarus unemployment −67% and labour participation +7.3% amid population −6%. Italy is an exception with GDP −5.3% and per capita GDP −9.3% over 2000–2020.

• Population and per capita GDP (Table 5): Long-run population effects on per capita GDP are positive but very small and asymmetric. Coefficients: POP+ = 0.0177 (p = 0.001) and POP− = 0.0085 (p = 0.028). A Wald test rejects equality of the two coefficients (χ² = 76.57, p < 0.001), indicating that a population decrease raises per capita GDP less (in magnitude) than a comparable population increase, though both effects are small. Short-run population effects on per capita GDP are generally insignificant across lags. Error-correction term ϕ = −0.2934 (p < 0.0001) indicates about 29.3% annual adjustment toward long-run equilibrium. Hausman test supports long-run homogeneity (χ² = 1.35, p = 0.508).

• Labour force participation and per capita GDP (Table 6): Long-run coefficient on LABOR is positive and small (0.0168; p = 0.008), implying higher participation is associated with higher per capita GDP. Short-run coefficients are mostly insignificant; contemporaneous ALABOR has a modest positive effect (0.0173; p = 0.024). Error-correction coefficient ϕ = −0.0760 (p = 0.002) indicates 7.6% annual adjustment. Hausman test: χ² = 1.16, p = 0.282.

• Unemployment and total GDP (Table 7): Long-run coefficient on U_ILO is negative and small (−0.0187; p < 0.0001), indicating higher unemployment relates to lower total GDP. In the short run, the contemporaneous change in unemployment reduces total GDP (−0.0164; p < 0.0001); other lagged short-run terms are insignificant. Error-correction coefficient ϕ = −0.1248 (p < 0.0001) indicates 12.5% annual adjustment. Hausman test: χ² = 0.62, p = 0.240.

Overall: The evidence shows that in the sample of 19 countries, periods of population decline coexisted with growth in total and per capita GDP, rising labour force participation in several cases, and falling unemployment, with estimated long-run elasticities small in magnitude but of expected signs.

The analysis addresses the central question of whether population decline inevitably depresses economic performance. The cointegration results indicate that per capita GDP is only weakly related to population changes, and that higher labour force participation supports higher per capita GDP while higher unemployment is associated with lower total GDP. The weak long-run elasticities and generally insignificant short-run effects suggest population decline per se is not a dominant driver of per capita income dynamics in these settings. The findings align with modern growth-theory results indicating that per capita output can rise even as population falls, potentially via human capital accumulation, technological adaptation, and institutional changes. In the examined countries—many undergoing post-socialist transitions—mechanisms such as labour-saving technologies, market reforms, foreign investment, migration/remittances, and productivity improvements plausibly mitigated supply- and demand-side headwinds from declining populations. Thus, the empirical evidence challenges the presumption that population decline must lead to stagnation, instead highlighting the role of structural and productivity channels in shaping outcomes.

Using PMG cointegration on panel data for 19 countries with recent population declines, the study finds that: (i) long-run effects of population changes on per capita GDP are positive but very small and asymmetric; (ii) higher labour force participation is associated with higher per capita GDP; and (iii) higher unemployment is associated with lower total GDP. Short-run effects are generally weak. Overall, population decline can coincide with rising GDP and per capita GDP, higher participation, and falling unemployment. The results complement theoretical work showing that per capita income growth can persist under declining population, likely supported by technology adoption and productivity growth. Future research could examine whether similar relationships hold under a global population decline (with potentially reduced migration, trade, and resource demand), explore heterogeneous effects across institutional and development contexts, investigate sectoral and distributional impacts, and test alternative asymmetric specifications and thresholds.

• Sample composition: Sixteen of the 19 countries are post-Soviet/transition economies where population decline was driven largely by emigration and institutional transition, which may not generalize to countries with fertility-driven declines or different institutional settings.
• Time horizon and period heterogeneity: The analysis uses annual data over 1990–2021 with country-specific short panels and substantial structural changes; short-run and long-run dynamics may differ across subperiods (e.g., crises, EU accession), potentially affecting stability.
• PMG assumptions: Long-run slope homogeneity across countries is imposed; while supported by Hausman tests here, true heterogeneity may still exist and affect inference. Short-run parameters are allowed to vary but model specification choices (lag orders, threshold for asymmetry) may influence results.
• Variable measurement: Population asymmetry decomposition uses the mean threshold; alternative thresholds (median/zero) yield similar results per authors, but different approaches could change estimates. GDP per capita (PPP) and labour participation/unemployment measures rely on international datasets with potential comparability issues.
• Omitted channels: The models are bivariate per outcome (population, labour participation, unemployment). Other determinants (e.g., capital formation, TFP, policy reforms, trade, remittances) are not explicitly modeled, which may bias or obscure mechanisms.