Subsidizing the spread of COVID-19: Evidence from the UK’s Eat-Out to-Help-Out scheme
The EOTHO scheme may have led to a 8% to 17% rise in infections according to a new CAGE working paper by a Warwick University Economist.
The Eat-Out to-Help-Out scheme was designed to revitalise the struggling hospitality sector by subsidising the cost of sit-in meals and non-alcoholic drinks. This incentivised risk-taking among the public which may, in the long-run, have inadvertently increased the overall economic harm of COVID-19 as new and extended ‘lockdown’ policies are employed to quell the 2nd wave.
Firstly, Thiemo Fetzer (the paper’s author) used regional data from UK’s official COVID-19 reporting dashboard and from HMRC’s restaurant-finder app to piece together the likely impact of the scheme on transmission rates throughout August and early September.
This showed significant increases in the rate of restaurant visits across the period of the scheme in August up-to >200% at peak. This likely shifted consumption (from the weeks before and after), but also appears to have increased it as the scheme explicitly intended.
The identification strategy firstly employed utilised the differential take-up rates of the scheme in a difference-in-differences design. This is where the change in infections in areas with high EOTHO take-up over the period are compared to the change in infections in an area with low EOTHO take-up.
This design demonstrated that in areas with notable adoption of EOTHO COVID-19 infections increased one week after the start of the scheme and decreased the week following. Note: 97% of non-asymptomatic patients developed symptoms within 8.2 to 15.6 days of infection and at least 50% of symptomatic patients report symptoms within 2 to 5 days of infection.
Using this design Fetzer suggests that per his ‘back of the envelope’ calculation, between 8% and 17% of all detected cases of COVID-19 infections over the relevant period were attributable to the EOTHO scheme.
While infection levels were low at this point, given that the virus spreads exponentially, the importance in stemming infection growth early is not easily overstated.
In order to ensure the robustness of this finding the paper exploits weather variation using and Google mobility variation. As expected, when there was rainfall during lunch and dinner hours the relative frequency of restaurant visits fell (hence theoretically reducing transmission rates).
When estimating the impact of rainfall on EOTHO days on the emergence of infection clusters later in the week, the paper found that infections were notably lower. This was measured against placebos e.g. the four weeks prior or post-scheme (below: Panel B and C, respectively).
Additionally, the impact of rainfall falling on the same days — but outside the core lunch and dinner hours had a null effect, therefore suggesting that, on average, rainfall falling outside the lunch or dinner hours has no notable impacts on mobility during the day.
The relationship survives robustness checks on other potential confounding variables e.g. population density, commuters, universities reopening and tenure types (e.g. renters vs owners), which are accounted for with additional non-parametric controls.
In conclusion, the EOTHO scheme appears to be successful in so much as it increased the level of restaurant usage. However, given that it appears to have precipitated the 2nd wave, the policy may have caused more harm than good to the hospitality sector, particularly when compared to the direct cash-transfer counterfactual.
Disclaimer: This is a working paper (not peer reviewed). Although the micro-econometrics used look appropriate to me the author is an economist not an epidemiologist and so please read skeptically!
Source: https://warwick.ac.uk/fac/soc/economics/research/centres/cage/manage/publications/wp.517.2020.pdf
