How did COVID-19 affect the likelihood of companies exporting their products? While this question might seem straightforward and simple to answer, the reality is far more complex and depends on various factors. For instance, a statistical measure of its impact may be tricky as no counterfactuals are available as I explain later on in this post.
It is out of any doubts that pandemic hit hard on international trade:
COVID-19 impacted our everyday life and continues to do so. The pandemic hit so hard that Econometrics itself was involved. Indeed, any counterfactual analysis (i.e. finding firms that were not exposed to COVID-19 so that we could measure the impact of the latter on export probability) seems impossible as COVID-19 had a "vertical" impact on every individual and firm. Thus, the construction of a "control group" is very difficult, not allowing for any causal inference...or maybe not?
Machine Learning can help uncovering apparently hidden patterns and constructing a counterfactual for firms in the pandemic!
The idea
The main idea is to create a Shock Unaware Machine (#SUM) based on 2018-2019 firm's export data and a Shock Aware Machine (#SAM), alias "COVID-19 aware". By comparing the difference of the two, we are able to detect the effect of COVID-19.
The results
Contrary to what you might expect, while a significant number of businesses have experienced a negative impact on their export probability due to COVID-19, a detailed analysis that considers the different characteristics of firms shows a diverse range of effects. Interestingly, this analysis uncovers that some companies have actually benefited in their exporting activities because of the pandemic.
As you can see, the answer is not as straightforward as it might first appear!
Where do I find out more?
The paper is under review in OBES!
Stay tuned for news as it we are almost ready with final refinements! If you cannot wait check out our #ASSA2022 poster below: