COVID-19 State Executive Orders from a Data Analytics Perspective

The effects of COVID-19 have presented a variety of health-related and economic challenges, and these challenges have been dealt with on levels ranging from decisions by individuals to approaches requiring worldwide coordination. In order to continue to develop and refine the strategies used to deal with this crisis, it is crucial to understand the actions and responses of the moving parts of the system being targeted by the coronavirus, such as a country’s government, businesses, and people.

The United States, which has been setting record numbers of daily new COVID cases, would greatly benefit from a more detailed understanding of precisely how people and businesses respond to official proclamations or directives from the government. While the general epidemiological, economic, and social effects of these actions are easy to see via the news or our daily lives, it is more difficult to make a quantitative prediction of such effects for an upcoming U.S. government order. For example, if a shelter in place order were to be given with an effective date of a week from now, how much of the population would already be sheltering in place more than a few days before that date? It is this disconnect between the true responses to an order and the order itself which requires closer analysis, as it is what makes the answers to valuable questions like “how many days did people stay at home for the majority of the day” more difficult to reach than by just looking at the effective dates of the order. Balancing economic considerations with the spread of the virus is a delicate task, one delicate enough to require specific information that might only be reached through these types of analyses.

However, the exact measurement of citizen and business responses to U.S. government emergency orders is not the topic of this blog. Topics such as lockdown compliance and the economic effects of COVID-19 have been analyzed many times, and deserve continued inspection as this pandemic rolls forward. Rather, this blog will take a look at something more simple, a single axis of the coronavirus response system—the content of the state executive orders themselves.

When I first decided to research the effects of statewide restrictions (and the lifting of those restrictions) on confirmed cases and the economy, one of the things I found most interesting was the source of those restrictions. These state-by-state directives seemed like the most natural source of information on statewide actions, since while you cannot guarantee that everyone followed an order, you can guarantee that the order itself was issued. But as I looked through each order to compare things like effective dates and which businesses were closed in each state, it became apparent to me how inconsistent the orders were from state to state. Some of these inconsistencies are more obvious and expected, like the fact that different states outright took different actions at different times, but there are smaller details which are harder to see without reading orders across multiple states.

For example, consider the language and phrasing used in the following excerpts of executive orders:

“All businesses or organizations are encouraged to provide delivery…” (Florida EO 20-91)

“In Stage 2, all individuals should stay at home or their place of residence as much as practicable…” (Indiana EO 20-26)

Strongly encourage all businesses and governmental entities to allow workers to telecommute whenever possible…” (Colorado D 2020 091)

Less forceful words like “should” and “encouraged” are ubiquitous across these kinds of directives that were issued after the expiration or rescission of the pre-reopening emergency orders. The language used in the above examples is clearly distinct from the language in an order requiring something. As opposed to decisive words such as “ordered”, “shall”, or “permitted”, the writing in the above statements makes it difficult to pinpoint the amount of businesses or individuals that will regularly adhere to those guidelines without other methods. There is even the question of whether seemingly similar words or phrases, like “are expected” or “should”, could have a subtle effect on the interpretation of an order. After all, while some people may learn about a new directive through the news or by word of mouth, there are still many who either must read through it or are just curious about the actual order itself.

I’d like to point out that the objective of this blog is not to analyze the effectiveness of an order based on its language, but to instead draw attention to the fact that it may be a detail relevant to data analysts looking through and comparing these government orders. COVID-19 has resulted in a massive influx of orders, advisories, and recommendations from different levels of the government in the United States, and because of that it is expected that there would be many differences between these unprecedented documents. Taking a look at small details such as language usage could be the subject of a smaller study or bolster larger studies focusing on lockdowns and closures in general.

Another inconsistency among the state executive orders were are businesses or services that are considered “non-essential”, and the way states went about defining this term. Some states explicitly listed all of the businesses that were not allowed to remain open, while others only gave a list of businesses that could do so. Some states chose to add previously “non-essential” items to the “essential” businesses and services lists as they began to reopen those businesses or services, while others kept the lists relatively constant. Alabama, for one, listed bookstores as “non-essential” and ordered the closure of them on March 28th, but later included them on the list of “essential businesses and operations” on their April 4th stay-at-home order.

As for the initial “essential” or “non-essential” lists themselves, some businesses, like marijuana dispensaries, that might be considered “essential” for one state, were considered “non-essential” for another. The differences between the businesses and services considered “essential” and “non-essential” creates a problem with comparing, for example, the effect of “non-essential” business closures on the state economies of two different states. While the words “essential” and “non-essential” do carry a general meaning across states, the imprecision of those terms could potentially prevent a study from being more accurate if it doesn’t consider its variance.

Although the inconsistencies across the state executive orders make it harder to quickly measure and analyze the results of the orders from a data analytics perspective, they may also prove to be an interesting subject of further study themselves. To fight back against this crisis, it would be ideal if we could understand which orders were most effective for curbing the virus and which were least damaging to the economy.  This would require a measurement scale that compare actions correctly. Given the inconsistencies mentioned and others such as the orders on different levels like county-wide orders, it may only be possible to create such a detailed scale by taking into account these many intricacies. Of course, with a complicated pandemic like COVID-19, it may seem as though there are always more details to take into account.  However, the importance of this situation implores us to look at small details like these to develop a precise, effective policy response.

Related Post

Leave A Reply