Background Social distancing measures to prevent the spread of COVID-19 included restrictions on retail services in many countries. In some countries, the governments also subsidized consumer spending on part of retail services to help struggling businesses. To evaluate the costs and benefits of government interventions in retail services, it is necessary to measure the infectiousness of each type of consumer activity. Methods This study regresses the log difference over seven days in the number of new confirmed cases of COVID-19 in Japan on lagged values of household expenditures per household on eating out, traveling, admissions to entertainment facilities, clothing and footwear, and the other items, as well as a measure of mobility in public transportation in the past 14 days. The sample period of the dependent variable is set from March 1, 2020, to February 1, 2021, in order to avoid a possible structural break due to the spread of mutant strains in 2021. The regression model is estimated by the Bayesian method with a non-informative (improper) prior. The estimated model is evaluated by out-of-sample forecast performance from February 2, 2021, onward. Results The out-of-sample forecasts of the regression by the posterior means of regression coefficients perform well before the spread of the Delta variant in Japan since June 2021. R2 for the out-of-sample forecasts from February 2, 2021, to June 30, 2021, is 0.60. The dependent variable of the regression overshot the out-of-sample forecasts from mid-June to August 2021. Then, the out-of-sample forecasts overpredicted the dependent variable for the rest of 2021. Conclusion The estimated model can be potentially useful in simulating changes in the number of new confirmed cases due to household spending on retail services, if it can be adjusted to realtime developments of mutant strains and vaccinations. Such simulations would help in designing cost-efficient government interventions.
ASJC Scopus subject areas