Did Halloween Cause the COVID-19 Spike in Onondaga County?
There was a lot of concern around how Halloween impact COVID-19 cases locally. After an already exhausting spring and summer, when cases had generally remained low, and an early fall where Syracuse University students returned to Central New York and school-aged kids started going back to in-person classes, the prospect of Halloween gatherings ruining all the progress was something I’d heard talked about a lot.
Now that we’re in the midst of the worst COVID-19 outbreak in the county, it it natural to look back a couple of weeks and figure that it must have been Halloween that caused the spike, but I’m not so sure. On the face of it, I’m skeptical. We’re in the midst of a global rise in COVID-19 cases — one day of Halloween wouldn’t cause sustained increases in infections around the world.
I’ve been tracking COVID-19 data that New York State makes available periodically since the spring. Onondaga County has done a good job publishing dashboards, but the underlying data has not been easily available so digging into the information from local sources has been challenging. But, in the last couple of weeks, I’ve stubbornly figured out how to get access to the underlying data, and figured I’d use it to investigate this question about Halloween.
Why does investigating this question matter? It may not. But, blaming Halloween may make people feel more relaxed about other things they were doing that potentially are risky. A couple of notes, though: I don’t have access to all of the data, there are definitely weird things that happen in the data that might be misleading me, I’m not a contact tracer so I don’t hear the specific comments from people about activities they were doing that might have caused them to test positive (the County Executive said contact tracers report Halloween as a key factor in case increases).
First, let’s look at how cases have grown since the beginning of the pandemic in the spring:
The purple bars are the daily case counts and the orange line is a seven day rolling average. The vertical black dashed line is Halloween. You can clearly see that there is a spike of cases, and that spike has happened recently. But, does this prove Halloween is the cause? If anything, the cases already seem to be growing by the time Halloween comes around.
Another way to understand the rapid growth of cases locally is to see how quickly the next 1,000 confirmed cases happen.
The chart above shows both the number of days it has taken to reach the next 1,000th case, as well as the date on which that case was reported. Between 4,000 and 5,000 cases, it took 58 days to reach 1,000 cases, and that happened on October 11. From 5,000–6,000 cases, it dropped to just 22 days. The 6,000th case was reported on November 2, just after Halloween, but too soon to see a spike when considering the likely incubation period. Following the 6,000th case, it took 9 days to reach our 7,000th case, and just 5 days to reach the 8,000th case. Those rapid increases did happen after Halloween, but I would argue we had already seen a spike begin prior to Halloween beginning.
Some might argue that this is all the fault of students at SU. The University also has provided a dashboard, and I have been collecting data from there, too.
The red dots are case numbers each day, the red line is the 7 day average number of cases. In this case, there was a bump in the middle of October, and then it does like like there was a spike about 10 days after Halloween. SU has now shut down in person classes after the sudden rise, so maybe Halloween gatherings amongst SU students caused issues there. But, while some community members might have been exposed to COVID-19 at the same gatherings at SU students, most likely were not and would have only suffered secondary exposure, which would take even longer to show up in the data.
I thought it would be worthwhile to overlay the County case data over the SU data. Note in this visualization, the SU cases should be a part of the overall County cases as well:
In the above visualization, the black line is County cases, the red line is SU cases. While we can continue to see a bump after Halloween at SU, it is clear that there are a rise in cases across the county weeks earlier. In fact, the initial bump in cases at SU in mid-October seems to line up closely with the rise in cases across the County.
There are some other key dates during the late summer and fall. SU students returned to school, suburban schools reopened for in-person learning, city schools first opened for remote learning and then hybrid learning (2 days in person per week) for some students. How do those key dates line up with case rise?
Above, the lines are still black for County cases and red for SU cases. The vertical lines align with the key dates (note the red bar is there because schooling started on slightly different dates in early September). It becomes clear that Halloween happens very much in the middle of the rise in cases. If anything, looking 10–14 days after in-person schooling started seems to align more with the outbreak. I don’t know any of the specific reasons, but it is hard to see the case rise and think that Halloween was the cause. Potentially if an outbreak was already occurring, the risk of someone showing up to a Halloween gathering who had COVID-19 was higher, leading to even faster growth, but I would argue that while the growth may have been accelerated by the holiday, it was not caused by it.
All of the data and visualizations I used in this post are available here: https://github.com/samedelstein/Onondaga_COVID. I’m doing my best to keep all of the data updated each day (though it would be awesome if I didn’t have to collect it from these sites myself — dashboards are great but the raw data that informs those spreadsheets are also important).
What do you think? Did Halloween cause this outbreak?
