Authoritative economic growth forecasts are often optimistically biased. One possible reason is that growth often has negatively skewed variation–negative shocks tend to have larger magnitudes than positive shocks, as the Great Recession and COVID-19 crisis illustrate. Negative skewness means that average growth over decades is smaller than typical-year (median or mode) growth, which would positively bias forecasts based on typical years. Here, we quantify this aspect of negative skewness in real per-capita GDP growth since the Industrial Revolution (1820-2016), by comparing medians and means across countries, regions, and time windows. Over decadal periods, we find mean growth rates to be <1%/y smaller than median growth rates in most countries and regions (median 0.23%/y across countries). Surprisingly, we find these differences are driven by negative skewness in both large- and medium-magnitude shocks, rather than only large shocks ('black swan' events). We also find our measure of negative skewness correlated with slow average per-capita GDP and population growth, high per-capita GDP growth volatility, and high per-capita GDP and population, building on previous studies. We find that recent over-projections of growth–by the International Monetary Fund (IMF), the U.S. Congressional Budget Office (CBO), and the Shared Socioeconomic Pathway (SSP) scenarios informing climate change research–have mostly been larger than can be explained solely by negative skewness, suggesting other sources of bias exist.