Assessing the Effect of the U.S. Vaccination Program on the Coronavirus Positivity Rate With a Multivariate Framework

Abstract

Abstract The factors influencing the incidence of COVID‐19, including the impact of the vaccination programs, have been studied in the literature. Most studies focus on one or two factors, without considering their interactions, which is not enough to assess a vaccination program in a statistically robust manner. We examine the impact of the U.S. vaccination program on the SARS‐CoV‐2 positivity rate while simultaneously considering a large number of factors involved in the spread of the virus and the feedbacks among them. We consider the effects of the following sets of factors: socioeconomic factors, public policy factors, environmental factors, and non‐observable factors. A time series Error Correction Model (ECM) was used to estimate the impact of the vaccination program at the national level on the positivity rate. Additionally, state‐level ECMs with panel data were combined with machine learning techniques to assess the impact of the program and identify relevant factors to build the best‐fitting models. We find that the vaccination program reduced the virus positivity rate. However, the program was partially undermined by a feedback loop in which increased vaccination led to increased mobility. Although some external factors reduced the positivity rate, the emergence of new variants increased the positivity rate. The positivity rate was associated with several forces acting simultaneously in opposite directions such as the number of vaccine doses administered and mobility. The existence of complex interactions, between the factors studied, implies that there is a need to combine different public policies to strengthen the impact of the vaccination program.

Keywords

• factors feedback and simultaneity
• U.S. vaccination program
• complex relationships
• error correction models
• machine learning
• time series and panel data