Reopening of schools and workplaces during the ongoing coronavirus disease 2019 (COVID-19) pandemic requires affordable and convenient population-wide screening methods. Although upper respiratory swab is considered the preferable specimen for testing, saliva offers several advantages, such as easier collection and lower cost. In this study, we compared the performance of saliva with upper respiratory swab for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) detection. Paired saliva and anterior nares specimens were collected from a largely asymptomatic cohort of students, faculty, and staff from the University of Pennsylvania. Paired saliva and combined nasopharyngeal/oropharyngeal (NP/OP) specimens were also collected from hospitalized patients with symptomatic COVID-19 following confirmatory testing. All study samples were tested by real-time PCR in the Hospital of the University of Pennsylvania. In the university cohort, positivity rates were 37 of 2500 for saliva (sensitivity, 86.1%) and 36 of 2500 for anterior nares (sensitivity, 83.7%), with an overall agreement of 99.6%. In the hospital study cohort, positivity rates were 35 of 49 for saliva (sensitivity, 89.3%) and 28 of 49 for NP/OP (sensitivity, 75.8%), with an overall agreement of 75.6%. A larger proportion of saliva than NP/OP samples tested positive after 4 days of symptom onset in hospitalized patients. Our results show that saliva has an acceptable sensitivity and is comparable to upper respiratory swab, supporting the use of saliva for SARS-CoV-2 detection in both symptomatic and asymptomatic populations.