Forest soil carbon (C) storage is a significant component of the global C cycle, and is important for sustaining forest productivity. Although forest management may have substantial impacts on soil C storage, experimental data from forest harvesting studies have not been synthesized recently. To quantify the effects of harvesting on soil C, and to identify sources of variation in soil C responses to harvest, we used meta-analysis to test a database of 432 soil C response ratios drawn from temperate forest harvest studies around the world. Harvesting reduced soil C by an average of 8 ± 3% (95% CI), although numerous sources of variation mediated this significant, overall effect. In particular, we found that C concentrations and C pool sizes responded differently to harvesting, and forest floors were more likely to lose C than mineral soils. Harvesting caused forest floor C storage to decline by a remarkably consistent 30 ± 6%, but losses were significantly smaller in coniferous/mixed stands (-20%) than hardwoods (-36%). Mineral soils showed no significant, overall change in C storage due to harvest, and variation among mineral soils was best explained by soil taxonomy. Alfisols and Spodosols exhibited no significant changes, and Inceptisols and Ultisols lost mineral soil C (-13% and -7%, respectively). However, these C losses were neither permanent nor unavoidable. Controls on variation within orders were not consistent, but included species composition, time, and sampling depth. Temporal patterns and soil C budgets suggest that forest floor C losses probably have a lesser impact on total soil C storage on Alfisols, Inceptisols, and Ultisols than on Spodosols, which store proportionately large amounts of C in forest floors with long C recovery times (50–70 years). Mineral soil C losses on Inceptisols and Ultisols indicate that these orders are vulnerable to significant harvest-induced changes in total soil C storage, but alternative residue management and site preparation techniques, and the passage of time, may mitigate or negate these losses. Key findings of this analysis, including the dependence of forest floor and mineral soil C storage changes on species composition and soil taxonomic order, suggest that further primary research may make it possible to create predictive maps of forest harvesting effects on soil C storage.