1AbstractDeaths are frequently under-estimated during emergencies, times when accurate mortality estimates are crucial for pandemic response and public adherence to non-pharmaceutical interventions. This study estimates excess all-cause, pneumonia, and influenza mortality during the COVID-19 health emergency using the June 12, 2020 release of weekly mortality data from the United States (U.S.) Mortality Surveillance Survey (MSS) from September 27, 2015 to May 9, 2020, using semiparametric and conventional time-series models in 9 states with high reported COVID-19 deaths and apparently complete mortality data: California, Colorado, Florida, Illinois, Massachusetts, Michigan, New Jersey, New York, and Washington. The May 9 endpoint was chosen due to apparently increased reporting lags in provisional mortality counts. We estimated greater excess mortality than official COVID-19 mortality in the U.S. (excess mortality 95% confidence interval (CI) (80862, 107284) vs. 78834 COVID-19 deaths) and 6 states: California (excess mortality 95% CI (2891, 5873) vs. 2849 COVID-19 deaths); Illinois (95% CI (4412, 5871) vs. 3525 COVID-19 deaths); Massachusetts (95% CI (5061, 6317) vs. 5050 COVID-19 deaths); New Jersey (95% CI (12497, 15307) vs. 10465 COVID-19 deaths); and New York (95% CI (30469, 37722) vs. 26584 COVID-19 deaths). Conventional model results were consistent with semiparametric results but less precise.Official COVID-19 mortality substantially understates actual mortality, suggesting greater case-fatality rates. Mortality reporting lags appeared to worsen during the pandemic, when timeliness in surveillance systems was most crucial for improving pandemic response.