Do Output Losses After Recessions Recover? Local Projections for 193 Economies, 1961–2025, Against Exact Random-Walk and Trend-Stationary Benchmarks

I estimate the path of real GDP after the start of a recession, defined as the first year of negative growth, in 193 economies with at least 30 years of World Bank data (1,146 recession starts, 1961 to 2025). With country and year effects and two lags of growth, log GDP is 8.1 points below its pre-recession level path in the start year, 12.9 points below after five years and 13.5 points (standard error 1.1) after ten years, with no sign of recovery. I prove two benchmarks for this design. If growth is independent over time, the estimated loss is the same at every horizon and equals the growth shortfall in the start year; a simulation with each country's own growth mean and variance gives about minus 8.7 at all horizons. If instead output is trend-stationary, the loss converges to minus the difference in the pre-recession cyclical position between start and non-start years, which is negative because recessions start from booms; a simulation with autoregressive coefficient 0.6 gives minus 5.5 at ten years against the analytical limit of minus 5.7. The observed loss deepens by 5.4 points after the start year, so it exceeds both benchmarks. Deep recessions (growth below minus 3%) leave a ten-year loss of 17.6 points and shallow ones 8.8 points. India's four recession episodes since 1965 show gaps below the pre-recession trend of 0.9 to 9.8 points five years later; the 2020 gap was 12.4 points in 2020 and 6.8 points in 2025.