Evaluating Risk to the US Supply Chain From COVID-19
Through any crisis, it’s tough to know how much tinfoil in your hat is too much. Despite a brief stint in the Boy Scouts, I don’t have any natural inclination to survivalism — I think the most disaster preparation I’ve ever done was a week worth of groceries for Hurricane Sandy. The early data on the coronavirus felt different and by the first week of March the impulse to go full Doomsday prepper was too strong to resist.
“You loading up for the virus?” asked the butcher, chuckling to himself while he heaved another set of porkchops on his scale. Even at that point, some of the shelves were starting to look bare. It seems a lot of the grocery employees looked at the panic buying with some bemusement.
I must have blushed a bit in reaction — I felt a fair amount of self-consciousness as I tugged around the two shopping carts. He backtracked in consolation, “Ah, it won’t be all that bad. Just put it in the freezer and you won’t have to shop for a couple months.”
A week later, I’d be on quarantine, presumptive for COVID-19. The following week, New York shut down. Now a month that felt like a decade later, I hope that guy is doing alright. This can’t be an easy time to be doing what he is doing for a living.
Hopefully, his freezer is full too.
Your social media timelines are likely as full as mine of people complaining about lack of delivery slots for delivery services like Prime Now and Instacart. Bloomberg is reporting today that ordering delivery online today feels like Ticketmaster, documenting the tidal wave crushing companies like Amazon, Fresh Direct, Peapod, and Postmates with the kind of demand that is eclipsing even the holiday shopping system.
The scarce availability of home delivery is likely a harbinger of broader supply chain disruption the United States could experience over the next six weeks. If the last mile delivery from the store shelf to the home is getting crushed and we’re still only at the earliest stages of the pandemic, what is going to happen to the wholesale supply chains that keep those shelves stocked in the first place?
Much of the international supply chain disruption has been explored. Are we at risk of disruption to the supply chain within the United States? Could that affect our ability to get groceries, medication and other essentials?
Getting the Data
First, we need to understand where our stuff comes from. In the United States, the retail products you and I purchase are nearly entirely stored in warehouses and delivered on trucks. Automation remains a distant future for this industry — transportation and warehousing still runs on people. It is one of the most labor intensive sectors of our economy, accounting for a national workforce of 6.91 million people. Getting products on shelves is a people-powered endeavor uniquely susceptible to disruption from an infectious disease.
To understand the risk of infection to these nearly seven million souls, we need to identify where they are. Reonomy — a commercial real estate data provider — appears to have a sample of 30,000 warehouse locations out of the estimated 110,000 that exist. A quick comparison of these locations seems to match DataUSA’s distribution of the workforce.
There are a few problems with this sample:
- We don’t know what kind of warehouses these are. Are these long-term storage or last mile fulfillment?
- We don’t know how much square footage they represent. Are they big warehouses or tiny warehouses?
- We don’t know what they are storing. Is this food and medicine which we urgently need or stockpiles of Mandalorian merch for the holiday season?
More robust data is available from research firms for five to six figures, which is beyond the scope of this project. Those caveats firmly in hand, it does look like we can learn something at least directionally about the supply chain risk from this sample set of warehouses.
With this sample, next we need to understand the risk to the workers at these 30,000 locations. I’ve written about COVID-19 testing a few times. Looking at these predominantly rural warehouse locations, we know the testing is poor there. We also know the cases and deaths are starting to grow in rural America —we can now assume the virus is already there. The question then becomes, which places are likely to have more undetected spread than others?
The New York Times reported a powerful dataset looking at the change in travel for counties across the United States during the month of March. By looking at the lack of social distancing in the counties where America’s warehouses are located, we can start to evaluate the risk exposure to these warehouse workers at a low level.
Finally, we need to understand when the pandemic’s spread is going to be the worst. COVID-19 is a frustrating pandemic to model, even for some of the best quant shops in the world. Out of the many projections out there, there is a lot to like about those coming out of the Institute for Health Metrics and Evaluation (IHME). The model considers a comprehensive collection of data, gets updated regularly throughout the day and continues to get informed by the spread in states like Washington and New York. This can give us a sense of when each state is likely to hit its peak.
With a sample of warehouses, how they well their counties distanced socially and when their peak is likely to occur, we can get a better sense of the risk to America’s supply chain.
Half Of US Warehouses Aren’t Under A Lockdown Order
If we plot a choropleth of social distancing by county in the United States and overlay the warehouse locations from our sample, the risk becomes evident. Half the warehouses in our sample are located in counties that did not do enough social distancing in March.
The lack of lockdown orders for those 15,051 warehouses had a profound influence on the amount of travel reduction observed in their counties. Counties with lockdown orders reduced travel by an average of 82.1%. Counties without lockdown orders reduced their travel on average only 63% — nearly twenty points less.
Compounding the lack of social distancing is the lack of visibility into COVID-19’s spread in these parts of the country. Currently The Covid Tracking Project provides testing data on the state level. If we normalize rank the number of warehouses in our sample by state and then compare to COVID-19 testing, the pattern is disturbing. States with many warehouses are doing less testing than states with fewer warehouses.
When Could Supply Chain Disruption Occur?
Looking at the lack of social distancing and COVID-19 testing in counties with a high density of warehouses, we can understand some real risk is evident. If we think COVID-19 infection is likely for these counties, the next question is timing. If the infection of the workforces surrounding these warehouses is staggered over months, the disruption it would cause would be less significant.
Looking at what we’ve observed in New York, COVID-19’s community spread comes in three waves over a month. First, the emergency rooms are flooded with cases. 7–10 days later, the hospital beds are full with admissions. Finally, a similar interval later there is a peak of death, a wave we are still enduring.
According to the latest projections from IHME, 76% of counties with warehouses in our sample have not yet peaked. The majority of the counties with warehouses in our sample are projected to reach peak hospitalization over the next two weeks. Two-thirds of the warehouses in our sample will reach peak death by the conclusion of April.
These projections don’t suggest a staggered spread of infection with peaks reaching different parts of the supply chain at different times. These projects instead suggest three-quarters of our warehouses will be surrounded by peak COVID-19 during the same three week period.
Stories of supply chain disruption are starting to appear. Yesterday, The New York Times reported on meat processing facilities shutting down in Georgia, Pennsylvania and South Dakota. These three states have 2,454 warehouses in our sample. The counties they are located in only reduced travel 63%, placing it in the second quartile of social distancing nationally. The state reported in the lede of the story — Georgia — holds half of the total warehouses among the three states and performed the worst with social distancing, reducing travel by only 55%. It ranks 10th nationally among states with the lowest social distancing.
Nestlé CEO Mark Schneider warned yesterday that the supply chain challenges his business will face are expected to increase in coming weeks with logistics a key concern in the United States. The Wall Street Journal is reporting this week supply chain software companies are struggling to manage disruption effectively.
The Worst Is Yet To Come
Given the volume of warehouses in counties at high-risk for significant COVID-19 infection, we can imagine a few scenarios.
Best case, we have not yet seen the worst supply chain disruption and there is still some limited time to mitigate impact. With the data from Washington and New York, we can more refine regional public health responses to limit spread and hasten treatment. We can also observe these counties are mostly rural and won’t see the kind of infection spread observed in denser metropolitan areas. It’ll be bad, but the shelves won’t go empty.
Worst case, half the logistical backbone of the United States didn’t do enough to prevent spread and have no idea what is coming. Signals are starting to show infection is already rife in the production side of the supply chain. While infection rate will be lower due to lack of density, medical infrastructure will be weaker and quickly oversubscribed, leading to higher mortality than we’ve seen in the cities. The continued inability to scale testing will prevent workers from returning when they are no longer infectious or when they are likely immune. The curve could be longer and higher than we saw in New York, meaning 4–5 months of disruption instead of what will likely be our 3–4. Meaningfully long shortages of food, medication and other essentials will radiate through the country.
Hopefully, these counties experience what we’ve seen with modeling so far and the impact will land somewhere in the middle. While the South and Tornado Alley have a large number of warehouses in counties with low social distancing, there is a significant density of warehouses in the Rust Belt, the Pacific Northwest and California that sit in the top quartile of travel reduction. While much of Texas had close to normal travel and no statewide lockdown order, Harris County — home to the most warehouses nationally in our sample — did have a lockdown order early and had a travel reduction of 75%, almost ten percent higher than the statewide average. And these data are imperfect: the IHME projections could be too high, our warehouse sample size could be overindexed rurally and counties have had eight days to improve their social distancing.
All that taken into consideration, we’re in an unprecedented time that will produce unprecedented outcomes. Among the many we should be considering is a medium term disruption in the US supply chain. I don’t think the United States is about to devolve into Mad Max-esque bedlam descending into a barter economy denominated in beans. But in the near and medium terms, there will be surprising shortages across many of the things we previously associated with normal life. For my part, I’m squirreling away enough non-perishable food, stockpiling medicine and loading up on enough essentials to make it through the summer.
It’s hard not to feel self-conscious while doing so. I never thought I’d be a Doomsday prepper, but the data is impossible to ignore.
