Experts agree that reverse transcription-polymerase chain reaction (PCR) testing is critical in controlling coronavirus disease 2019 (COVID-19), but decision makers disagree on how much testing is optimal. Controlling for interventions and ecological factors, we used linear regression to quantify testing's impact on COVID-19's average reproduction number, which represents transmissibility, in 173 countries and territories (which account for 99 percent of the world's COVID-19 cases) during March-June 2020. Among interventions, PCR testing had the greatest influence: a tenfold increase in the ratio of tests to new cases reported reduced the average reproduction number by 9 percent across a range of testing levels. Our results imply that mobility reductions (for example, shelter-in-place orders) were less effective in developing countries than in developed countries. Our results help explain how some nations achieved near-elimination of COVID-19 and the failure of lockdowns to slow COVID-19 in others. Our findings suggest that the testing benchmarks used by the World Health Organization and other entities are insufficient for COVID-19 control. Increased testing and isolation may represent the most effective, least costly alternative in terms of money, economic growth, and human life for controlling COVID-19.