Forecasting COVID-19: Vaccines, Testing, Economic Impact

At the start of the COVID-19 pandemic earlier this year, Metaculus launched multiple efforts aiming to provide timely forecasts on various facets of the pandemic, with the ultimate goal of informing actions taken by decision makers and the general public. In this way, we hope to have been helpful in aiding the global response to the pandemic.

On our domain for COVID-related content, Pandemic, we are hosting a wide variety of questions proposed by Metaculus users as well those created by moderators and admins. We have additionally supported research collaborations with multiple external institutions, including with: Professor Thomas McAndrew of Lehigh University to predict the early course of the pandemic in the U.S. and to solicit expert forecasts on vaccine and therapeutic solutions to COVID-19, Epidemic Forecasting at the University of Oxford to assess the impact of lockdowns this past spring and the consequences of lifting of those lockdowns, the medical center at Texas Tech University El Paso to accurately predict their spring peak in hospitalizations and ICU bed usage, and Dr. Amanda Ortella to forecast the pandemic's social consequences (check out an upcoming event we have with Dr. Rotella on 20 November). Overall, our publicly accessible collaborative projects to date have gone well and the results have helped inform decision making at various levels. Metaculus also has ongoing tournaments with prizes that forecasters can participate in — two such examples are the Salk Tournament and the Li Wenliang prize series, both of which continue to be important and relevant.

Below we highlight some of the key take-aways from our report. You can find the full report here. This is the first of multiple such reports since, of course, our work is not yet done — a long winter awaits.

SARS-CoV-2 vaccines

A safe, efficacious, and widely accessible SARS-CoV-2 vaccine — or better yet, multiple vaccines — is likely the best long-term solution to the ongoing COVID-19 pandemic.

Safety: Experts assigned a median of 0.98 (80% CI: 0.20, 1.00) to the probability of at least ten serious adverse events being attributed within one year to the first SARS-CoV-2 vaccine approved in the U.S. through a normal approval process, and assigned a median prediction of 0.98 (80% CI: 0.37, 1.00) to the probability of the same for a SARS-CoV-2 vaccine approved in the U.S. through an emergency approval process. However, experts also importantly commented that they expect any approved vaccine to be safe for most people and expected the most common adverse events to be mild — local pain and fatigue. The expectation is that the benefits of vaccination against SARS-CoV-2 will far outweigh the risks.

Efficacy: Until recent positive news on the Pfizer/BioNTech SE vaccine candidate, there has been a lot of uncertainty as to whether vaccine candidates currently being tested would be sufficiently efficacious to end or at least slow the pandemic. As of 13 November 2020, after the positive vaccine news from Pfizer/BioNTech SE, Metaculus forecasters think that a safe SARS-CoV-2 vaccine will demonstrate an efficacy rate >75% in n>500 randomized controlled trial in early December 2020 (80% CI: November 2020, February 2021) and assign a median of May 2021 (80% CI: November 2020, April 2023) to when a safe SARS-CoV-2 vaccine will demonstrate an efficacy rate >95% in n>500 randomized controlled trial.

Timelines for approval: A vaccine will probably be approved through the emergency process much quicker, a median of 10 months sooner, than through a normal approval process. The Metaculus consensus as of 13 November 2020 on when normal approval of a vaccine candidate will occur, which includes more recent forecasts made on the basis of the promising news regarding the Pfizer/BioNTech SE vaccine candidate, assigns a median of January 2021 (80% CI: November 2021, June 2021).

Distribution: As of 13 November 2020, Metaculus forecasters predicted a median of February 2021 for when a SARS-CoV-2 vaccine candidate that has demonstrated an efficacy rate greater than 75% in a n>500 RCT will be administered to at least 10 million people (80% CI: December 2020, May 2021) and a median of October 2021 for when a SARS-CoV-2 vaccine candidate that has demonstrated an efficacy rate greater than 95% in a n>500 RCT will be administered to at least 10 million people (80% CI: January 2020, > April 2023).

Testing and Tracking

Until a vaccines are approved and widely distributed, testing and tracking will continue to play important roles in mitigating the spread of SARS-CoV-2.

Testing: Metaculus users have closely followed the ramp-up of testing, especially earlier on the pandemic when they accurately predicted the number of tests performed in the U.S. by June 1 and August 1. On the question of when an FDA-approved in-home diagnostic test for SARS-CoV-2 infection can first be purchased by a U.S. consumer from Amazon, the Metaculus prediction is a median of February 2021 (80% CI: November 2020, May 2022). Such a test would play an important role in mitigating the spread of SARS-CoV-2. Moreover, it would help confirm cases that would otherwise go undetected — as of 13 November 2020, the median prediction for how many cases will be confirmed globally before 2021 is 81.6 million, which is only a fraction of the estimated 645 million infections that will have occurred before 2021.

Tracking: In the U.S., the lack of a centralized data repository has meant that non-government efforts to track the pandemic, like that of the Covid Tracking Project (CTP) and Johns Hopkins’ Center for Systems Science and Engineering (JHU CESSE), have taken on great importance. Many Metaculus questions have and will resolve on the basis of these sources, and so many Metaculus users have been closely tracking them. Indeed, we have been tracking the probability of the continued maintenance of these sources themselves — for instance, the Metaculus prediction for whether the JHU CESSE COVID-19 dashboard is maintained through the end of 2020 is now 98%, though in early April it was only 69%. Metaculus users have also been addressing controversial issues, such as the probability that there will be credible claims that China deliberately misreported its number of COVID-19 infections — as of 13 November 2020, there is a 55% chance of this occurring.

Economic Impact

The use of non-pharmaceutical interventions, as well as the practicing of social distancing by individuals, has had an unprecedented adverse economic impact. Here, we focus on questions related to the U.S. economy in particular.

Unemployment: Metaculus forecasters think that that the height of U.S. unemployment this year has already peaked at 14.7% in April. Moreover, Metaculus forecasters expect continued improvement — the median prediction for U.S. unemployment in 2021 is 7.7% (80% CI: 5.0%, 10.9%). However, it will likely take years to get back to an unemployment rate of 4.0% — the median Metaculus prediction for when this will occur is November 2024 (80% CI: May 2022, > December 2024).

Return to in-person work and activity: There is significant uncertainty as to when, if ever, most of these workers will return to in-person work. It is also not known when work-related activities like air travel and metro area transit activity will recover to pre-pandemic levels. On the question of when most U.S. tech workers will return to the office, the Metaculus median is August 2021 (80% CI: December 2020, >May 2024).

Stock market: Perhaps the best-performing part of the economy since the March/April 2020 nadir is the stock market. The Dow Jones Industrial Average and SP500 recently reached intraday all-time highs on the news of the Pfizer/BioNTech SE vaccine being highly efficacious. And, according to pertinent Metaculus questions, the Dow Jones is very likely to close at all-time high record in the very near future. Of course, the stock market is only one part of the economy — in many respects, a very long recovery awaits.