In response to the Covid-19 outbreak in the spring of 2020, governments around the world agreed to lockdowns of different intensities to control the spread of the pandemic, alongside other pharmaceutical (PI) and non-pharmaceutical (NPI) interventions. After months of intense and critical emergency, the government then lifted the strictest measures, hoping the use of masks, social distancing and remote work could prevent a pandemic. However, in recent weeks, fears of a second wave have started to increase as the number of new infections continues to rise in many countries. Several countries, such as Israel, have implemented new lockdowns; others, particularly in Europe, are discussing applying tougher measures to economic and social mobility and activity. Since spring, we have learned some facts about the spread of the virus – for example, the use of masks is fundamental (Lyu and Wehby 2020) – that are changing the discussion of a second lockdown. However, we don’t know much about the effectiveness of severe pre-locks during the peak of the first wave of the epidemic.

Italy’s economy lockdown

In a recent study, we answered this question (Borri et al. 2020). We studied the Italian case to estimate how much the intensity of the lockdown reduced the number of deaths from Covid-19. The focus on Italy – one of the first countries to hit Covid-19 – is motivated by the lockdown design implemented by the government, which offers a reasonable source of variation in intensity at a detailed level. In response to the Covid-19 outbreak, the Italian government imposed the closure of all schools on March 5 and the first closure on March 11, closing many business activities open to the public including restaurants and fitness centers. It then imposed a second economic lockdown on March 22. For this second lockdown, the government established a list of essential economic activities that were allowed to continue operating, while others were suspended or only allowed to operate remotely. These policy provisions lead to heterogeneous geographical variations in the share of active workers across cities, predetermined by economic activity in a particular city and independent of the pandemic. We matched the list of important economic activities with data on the number of workers in those important jobs at the 3-digit city-NACE level. Important to our analysis, we included a granular sample of 7,089 cities, with an average population of 2,443 residents and a median area of ​​21 square kilometers, each belonging to one of Italy’s 110 provinces.

Impact on mobility

To measure the locking effect, we divided a sample of Italian cities into two groups. For each province in Italy, we calculated the median share of the reduction in active population after the economic shutdown on March 22, and then we grouped the cities below and above their provincial median. Cities above the median experienced an average decline in the percentage of active residents by 42.5 percentage points, while cities below the median experienced an average decline of 17 percentage points. Similar to Glaeser et al. (2020), we first estimate the decline in mobility caused by the economic lockdown by comparing cities above and below the median, before and after 22 March.

In Figure 1, we report the mobility patterns for the two city groups. We observed that reduced mobility occurred long before the first and second closures for all municipalities. However, the second lockdown on March 22 led to an additional drop in mobility in cities where the decline in the share of active workers was above the provincial median. In particular, we estimate that cities with a greater contraction in the share of active workers experience a decrease in daily mobility by about 53 kilometers per 1,000 population in relation to municipalities with a smaller contraction in the share of active workers.

Image 1 Evolution of the 5-day moving average in kilometers per 1,000 inhabitants (as residual due to fixed day of week effect)

Inactive workers and deaths from Covid-19

We represent Covid-19 deaths with excess mortality at the city level – measured as the difference between the daily number of deaths in 2020 and the average number of deaths on the same day and in the same city between 2015 and 2019. This addresses, at least in part, problems related to differences in the classification of deaths due to Covid-19, testing, and hospital capacity (Buonanno et al. 2020, Galeotti and Surico 2020). To capture the consequences of second locking in excess mortality, we followed the medical literature and adopted a conservative approach by insisting that the potential effect on mortality be manifested by a two-week lag from transmission (Wilson et al. 2020, Sun et al. 2020).

In Figure 2, we plot the evolution in five-day moving averages of excess mortality for two city groups. The figure shows that, despite schools being closed and other measures of containment from the first closure, both city groups experienced an excessive increase in the number of deaths in early March that lasted until the end of the month. Importantly, the municipalities above the median experienced a sharp increase in excess mortality in the first lockout period (March 11-22) while they experienced a sharper decrease in the second period (March 22-April 30) compared with those below the median. In line with the mechanisms linking active workers and deaths from Covid-19, this suggests that municipalities are characterized by a higher proportion of active workers and consequently mobility pays for a higher number of deaths prior to the economic lockdown, while they experience a greater reduction in Covid-19 deaths after locking is done.

At the same time, Figure 2 does not appear to support the assumption of standard parallel trends – the requirement that in the absence of treatment, the differences between the treatment and control groups are constant over time – which is necessary for differences in design differences. For this reason, we address the potential for violation of this assumption by controlling for the dynamics of excess mortality, by estimating our estimated limits (Angrist and Pischke 2008) and by showing that our results match our choices for sample cities with more similar or nearly identical pre-trends. , such as a city with less than 5,000 inhabitants.

Figure 2 Excess mortality in 2020 compared to 2015-2019

Our main findings suggest that the intensity of the economic lockdown is associated with a statistically significant reduction in mortality by Covid-19, particularly for the 40-64 age group and older, with greater and more significant effects for individuals over 50. -the-envelope indicates that in the 26 days between April 5 and April 30, a total of 4,793 deaths could have been avoided in 3,518 cities experiencing a more intense lockdown. The results are robust for including time-varying shocks for provinces and alternative specifications such as looking at the linear model (i.e. considering a linear decrease in the share of the active population in excess mortality) and weighting the share of active workers on proximity and occupation-from-home indices, among others.

Alternative mechanism

Interestingly, we found no significant effect of the lockdown in southern Italy. While these results are not surprising given the modest epidemic effects in this part of the country, they are at the same time a useful placebo exercise that excludes other confounding mechanisms. For example, our findings may be contaminated by the fact that in municipalities with a greater decline in active population there are also less lethal car or workplace accidents. However, the fact that we found no significant effect in the south, where the spread of the pandemic is negligible, would suggest that the effect of this alternative channel is, if any, small. In our paper, we discuss other possible confounding mechanisms that could explain our results, such as a return to the mean or the effect of the Italian first locking. We argue that the available evidence does not support this alternative mechanism. While we cannot precisely identify the channels in which lockdowns may have reduced transmission, the prime candidate is a reduction in the mobility of active workers caused by lockdowns.

Conclusion

Our results are the first step to evaluating the costs and benefits of severe lockdowns and can help to guide policymakers in their decisions regarding the safe relaxation of these measures after the current medical emergency, and to evaluate which policies are more effective to control. . future pandemics. Obviously, we can’t claim the same effect will apply in different settings – for example, when more masks become available (Lyu and Wehby 2020) or better contact tracing systems are implemented. More generally, our empirical strategy provides a simple methodology for future research aimed at assessing the impact of the economic lockdown on reducing Covid-19 deaths in other countries.

Reference

Angrist, JD and JS Pischke (2008), The least dangerous econometrics: The empiric’s colleague. Princeton: Princeton University Press.

Borri, N, F Drago, C Santantonio and F Sobbrio (2020), “The “Great Lockdown”: Workers inactivity and death from Covid-19”, CEPR Discussion Paper 15317.

Buonanno, P, S Galletta and M Puca (2020), “Estimating the severity of Covid-19: evidence from the Italian epicenter”, Plos One (will come).

Galeotti, A and P Surico (2020), “User’s Guide to Covid-19”, VOX.org, 27 March. https://voxeu.org/article/user-guide-covid-19

Glaeser, EL, CS Gorback and SJ Redding (2020), “How much will Covid-19 increase with mobility? Evidence from New York and four other US cities ”, NBER Working Paper 8345.

Lyu, W and GL Wehby (2020), “Community use of face masks and Covid-19: Evidence from state-mandated natural experiments in the US”, Health Affairs 39 (8).

Sun, P, X Lu, C Xu, W Sun and B Pan (2020), “Understanding Covid-19 based on current evidence”, Journal of Medical Virology 92 (6).

Wilson, N, A Kvalsvig, LT Barnard and MG Baker (2020), “Estimated case-fatality risk for Covid-19 calculated using the time lag for death”, Emerging Infectious Diseases 26 (6).