The Covid-19 containment effects of public health measures - A spatial difference-in-differences approach

Abstract

Since mid-March 2020 the Federal and state governments in Germany agreed on comprehensive public health measures to contain the spread of Covid-19. We study the effects of the policy actions on the progression of the pandemic in the first containment period in spring 2020 before easing of measures have started to work by the end of April. To exploit both the time and spatial dimension in the dissemination of the virus, we conduct a spatial data panel analysis for German NUTS-3 regions. Specifically, we employ a spatial difference-in-differences approach to identify the effects of individual public health measures. We find that contact restrictions and closure of schools substantially contributed to flattening the infection curve. Additionally, a strong treatment effect of wearing face masks is established in the few treated regions during the containment phase. No incremental effect is evidenced for closure of establishments and shutdown of shopping malls and other stores. These findings prove to be highly robust to changes in model specification. By contrast, the dampening effect of restaurant closure is sensitive to model variation.

Early assessment of epidemiological trends associated with SARS-CoV-2 variants of concern in Germany

“…We also include a spatially lagged variable covering the number of SARS-CoV-2 cases in region i ’s (direct) spatial neighbors during the last, the second last and the third last week. A spatial lag is important because infections can easily spread from one region to another region nearby, e.g., due to commuting or general mobility (Kosfeld et al, 2020). Spatial association between regions is measured through first-order contiguity, i.e.…”

The effects of non-pharmaceutical interventions on COVID-19-related mortality: A generalized synthetic control approach across 169 countries

“…We tackle this problem by applying a Generalized Synthetic Control (GSC) method [22]. The intuitive idea behind the conventional Synthetic Control method is to construct a synthetic control unit for a single treated unit by re-weighting observations from the pool of control-units [23][24][25].…”

“…The traditional Synthetic Control approach applies only to the case of one treated unit. Yet, the GSC method [22] provides a framework which generalizes the synthetic control method to multiple treated units by using factor augmented models. We start with the general model…”

“…L it subsumes unobserved factors approximating the outcome trend. The original GSC method based on interactive fixed effects models directly estimates L = ΛF , with Λ being an N × r matrix of unknown factor loadings (unitspecific intercepts), and F an r × T matrix of unobserved common factors (time-varying coefficients) [22]. This however requires the number of factors r to be specified correctly, for instance by using cross-validation methods.…”

“…Second, compared to other studies [17], we use data of more countries (N= 169) covering 98% of the world population. Third, to estimate the average treatment effect on the treated (ATT), we apply the Generalized Synthetic Control (GSC) method [22]. The GSC approach is a combination of the synthetic control approach [23][24][25] and traditional difference-in-differences (DiD) methods.…”