Japan's voluntary lockdown

Abstract

Japan's government has taken a number of measures, including declaring a state of emergency, to combat the spread COVID-19. We examine the mechanisms through which the government's policies have led to changes in people's behavior. Using smartphone location data, we construct a daily prefecture-level stay-at-home measure to identify the following two effects: (1) the effect that citizens refrained from going out in line with the government's request, and (2) the effect that government announcements reinforced awareness with regard to the seriousness of the pandemic and people voluntarily refrained from going out. Our main findings are as follows. First, the declaration of the state of emergency reduced the number of people leaving their homes by 8.5% through the first channel, which is of the same order of magnitude as the estimates obtained for lockdowns in the United States. Second, a 1% increase in new infections in a prefecture reduces people's outings in that prefecture by 0.027%. Third, the government's requests are responsible for about one quarter of the decrease in outings in Tokyo, while the remaining three quarters are the result of citizens obtaining new information through government announcements and the daily release of the number of infections. The findings suggest that what mattered for containing the spread of COVID-19 was not strong, legally binding measures but the provision of appropriate information that encouraged people to change their behavior.

Fig 1. Stay-at-home measure and number of new infections, Tokyo.

Fig 2. Stay-at-home measure and number of new infections, Ibaraki.

Fig 3. Our Measures and Google’s mobility measure.

The blue line shows the percentage change in the level of outings from the corresponding value at the baseline, where the level of outings is defined as the difference between the nighttime and the daytime population. The green line shows the percentage change in the extent to which people stayed at home, which is defined as the daytime population, from the corresponding value at the baseline. Both are constructed from the DoCoMo data. The red line shows the measure of “time spent in residential places” extracted from Google’s Community Mobility Reports.

Fig 4. Cross-prefecture comparison.

The horizontal axis shows the percentage change in the level of outings from the corresponding value at the baseline, where the level of outings is defined as the difference between the nighttime and the daytime population. The vertical axis shows the measure of “time spent in residential places” extracted from Google’s Community Mobility Reports. Both are the average of daily observations from April 22, 2020 to April 28, 2020.

Fig 5. Start and end of school closures.

Fig 6. Start and end of state of emergency.

Fig 7. Estimates of time fixed effects.

The blue line shows the coefficients on the time dummies. The dotted red line shows the fitted values obtained as follows: (1) we regress the estimated time fixed effects on the number of infections nationwide, the declaration of the state of emergency (April 7 and 16), the lifting of the state of emergency (May 14 and May 25), and Weekend/Holiday dummy; (2) we set the coefficient on the Weekend/Holiday dummy to zero in order to eliminate periodical fluctuations due to the weekend/holiday effect.

Fig 8. Decomposition of changes in the stay-at-home measure for Tokyo.

Changes in the stay-at-home measure for Tokyo are decomposed into the intervention and the information effects using the estimation results from specification (5) in Table 1. To eliminate seasonal fluctuations, only weekday observations are used for the stay-at-home measure.