A growing and diverse literature is emerging on the COVID-19 pandemic’s socioeconomic effects and the efficacy of alternative mitigating policies. A series of empirical studies have provided preliminary findings on the impact of non-pharmaceutical interventions on health outcomes. Focused on the initial months of the pandemic, these studies have for the most part documented significant effects in reducing the spread of the virus (Askitas et al. 2020a,b, Wong et al. 2020, Deb et al. 2020a,b, Li et al. 2020, Caselli et al. 2020).
However, because they are based on data from the first semester of 2020, they fail to capture the incidence of lockdown fatigue, namely, non-linear effects due to the cumulative economic and psycho-sociological burden of the restrictions and the diminishing degree of compliance.1
In a recent paper (Goldstein et al. 2021), we evaluate the effectiveness of lockdowns and de facto declines in mobility been over time. To do this, we build a panel dataset of 152 countries with data from the onset of the pandemic until 31 December 2020, including two high-frequency cross-country metrics of restrictions to social interactions:
- The Stringency Index from Oxford´s COVID-19 Government Response Tracker, which measures the rigidity of school closures, workplace closures, restrictions on public events, among others containment policies
- Google’s measure of workplace mobility based on anonymised location history data from Google Maps.
We analyse the impact of both metrics on the effective reproduction number (Rt) and the evolution of COVID-19 deaths per million. To estimate the effect of non-pharmaceutical interventions over time, we follow Deb et al. (2020a,b) in their use of the local projections methodology, first introduced by Jordà (2005).
By estimating one-step-ahead ordinary regressions for each time period – instead of approximating the data globally through, for example, a vector autoregression – local projections provide impulse-response functions that are not only more suitable for non-linear and flexible relations but also less susceptible to misspecification as well as more amenable to statistical inference.
Figure 1 shows the estimated dynamic response of cumulative and daily growth rates of the reproduction number and the number of deaths per million (7-day averages to smooth out reporting and seasonal noise) to a change of one standard deviation in the Stringency Index over the 90-day period following the intensification of containment measures.
Stringency is associated with a gradual, significant, and negative variation of the spread of the virus and COVID-19-related deaths. The effect is fairly persistent: its cumulative effect on deaths peaks at about 60 days after the increase in intensity; the effect on the reproduction rate peaks at 20 days. Similar results are observed when analysing Google mobility data, confirming the findings of the previous literature.
Figure 1 Impact of OxCGRT Stringency Index on effective reproduction rate and COVID-19-related deaths
Notes: The graph represents the estimated impulse-response function for a change of one standard deviation in the OxCGRT Stringency Index. The shaded area represents the 90% confidence interval for the coefficient.
Source: Goldstein et al. (2021).
How long does this effect last? More precisely, how much is lost if we go from a one-month to a four-month lockdown – would lockdowns have the same effectiveness if resorted to in subsequent waves of the pandemic?
To do assess that, we include a length dummy that equals 1 after 120 days of strict (and possibly intermittent) lockdown and interact it with the policy variables: stringency and mobility. As can be seen in Figure 2, there are significant differences in the effect of distancing measures in reducing deaths from COVID-19 when comparing the onset of the pandemic with its re-imposition (proxied by the interaction with the length dummy).
Containment policies generate lower reductions in deaths from COVID-19 than in the first stage of the epidemic, and the effect tends to lose its statistical significance faster. By contrast, no significant differences are observed between the two phases of the pandemic for impact on the reproduction rate.
Figure 2 Impact of OxCGRT Stringency Index on effective reproduction rate and COVID-19-related deaths (coefficient and interaction term at 120 days)
Notes: The graph represents the estimated impulse-response function for a change of one standard deviation in the OxCGRT Stringency Index, including the effect of the duration-Stringency interaction valued at the specified period. The shaded area represents the 90% confidence interval for the linear combination.
Source: Goldstein et al. (2021).
A priori, it could be assumed that the fading impact of the lockdown may owe in part to the fact that compliance with mobility restrictions is hard to sustain economically and socially for long periods of time (Levy Yeyati and Sartorio 2020). If that were the case, one would expect that the estimated effect of the de jure stringency should decline by more than the effect of the de facto mobility variations, simply because de jure restrictions are increasingly ignored.
Indeed, when analysing the effect of more stringent containment policies in reducing workplace mobility (Figure 3), we can see that the impact of the Stringency Index on workplace mobility, while significant, tends to weaken after 120 days of strict lockdown.
Figure 3 Impact of OxCGRT Stringency Index on Google workplace mobility (coefficient and interaction term at 120 days)
Notes: The graph represents the estimated impulse-response function for a change of one standard deviation in the OxCGRT Stringency Index, including the effect of the duration-Stringency interaction valued at the specified period. The shaded area represents the 90% confidence interval for the linear combination. Source: Goldstein et al. (2021).
However, even if we take as given an increase in de facto non-compliance, the time pattern of the impact of mobility restrictions is still there: after 120 days of strict lockdown, a decrease in workplace mobility has a significantly more attenuated effect on the reduction of COVID-19 deaths and does not have a significant impact on Rt (Figure 4). In other words, non-pharmaceutical interventions lose their effectiveness beyond the effect of diminishing compliance, suggesting other channels may also be affected by lockdown fatigue.
Figure 4 Impact of Google Workplace Mobility on effective reproduction rate and COVID-19-related deaths (coefficient and interaction term at 120 days)
Notes: The graph represents the estimated impulse-response function for a change of one standard deviation in the Google Workplace Mobility index, including the effect of the duration-Mobility interaction valued at the specified period. The shaded area represents the 90% confidence interval for the linear combination.
Source: Goldstein et al. (2021).
These results suggest that restrictions applied for a long period or reintroduced late in the pandemic (for example, in the event of a resurgence of cases) would exert, at best, a weaker, attenuated effect on the circulation of the virus and the number of casualties. Combined with the results in Haug et al. (2020), they suggest that lockdowns should be strict and brief.
These findings are particularly relevant for emerging and developing countries that may face considerable delays in the start of massive vaccination campaigns and will likely face successive waves of infections without having achieved herd immunity levels. After a year of strong economic downturn and substantial health costs, the intensity of lockdowns continues to be an influential factor in the economic and social life of low- and middle-income countries. Even if restrictions played a role early on, they had a one-off effect that would be hard to replicate going forward. This suggests that the heavy reliance on lockdowns as in the early stages of the pandemic may not be advised.
Askitas, N, K Tatsiramos and B Verheyden (2020a), “Flattening the COVID-19 curve: What works", VoxEU.org, 5 June.
Askitas, N, K Tatsiramos and B Verheyden (2020b), “Lockdown strategies, mobility patterns and COVID-19”, IZA DP 13293.
Battisti, M, A Kourtellos and G Maggio (2021), “School openings affect local COVID-19 diffusion”, VoxEU.org, 2 February.
Caselli, F, F Grigoli, W Lian and D Sandri (2020), “Protecting lives and livelihoods with early and tight lockdowns”, VoxEU.org, 16 November.
Deb, P, D Furceri, J D Ostry and N Tawk (2020a), “The economic effects of COVID-19 containment measures”, VoxEU.org, 17 June.
Deb, P, D Furceri, J D Ostry and N Tawk (2020b), “The effect of containment measures on the Covid-19 pandemic”, IMF Working Paper 20/159.
Goldstein, P, E Levy Yeyati and L Sartorio (2021), “Lockdown fatigue: The diminishing effects of quarantines on the spread of COVID-19”, Covid Economics 67: 1–23.
Haug, N, L Geyrhofer, A Londei, E Dervic, A Desvars-Larrive, V Loreto, B Pinior, S Thurner and P Klimek (2020), “Ranking the effectiveness of worldwide COVID-19 government interventions”, Nature Human Behaviour 4(12): 1303–12.
Jordà, Ò (2005), “Estimation and inference of impulse responses by local projections”, American Economic Review 95(1).
Li, Y, H Campbell, D Kulkarni, A Harpur, M Nundy, X Wang, H Nair, et al. (2020), “The temporal association of introducing and lifting non-pharmaceutical interventions with the time-varying reproduction number (R) of SARS-CoV-2: A modelling study across 131 countries”, The Lancet Infectious Diseases 21(2): 193–202.
Wong, M C S, J Huang, J Teoh and S H Wong (2020), “Evaluation on different non-pharmaceutical interventions during COVID-19 pandemic: An analysis of 139 countries”, The Journal of Infection 81(3).
1 Haug et al. (2020) offer another proof of nonlinearity, reporting a more-than-proportional effect of stricter lockdowns. Battisti et al. (2021) have more recently provided estimates of the effect of school openings in Sicily in the second semester of 2020.