Tyler Cowen told me to go viral on TikTok
Here's what happened next
At the end of 2021, Tyler Cowen assessed the state of economics on TikTok. He was not impressed.
TikTok is now the most popular website in the world, by one measure, and as such its influence on how young people see and think about themselves is attracting ever more attention. I am an economist, so I would like to focus on a considerably more narrow subject: what TikTok’s videos say about how young people see and think about economics. The news — for my profession, not necessarily for TikTok — is not good.
Since Market Power would not exist as it does today without Tyler’s support through Emergent Ventures, I read this article as a direct challenge. “Craig, make a viral TikTok.”
Today, I get to report that I succeeded. This last week, I made a TikTok video that got 4 million views.
Except, it wasn’t on TikTok. And my experience highlights why economics is struggling there and where it will succeed.
The Experiment
After reading Tyler’s article, I started an experiment. A year earlier I had already created a TikTok account, and of course, I’m already on YouTube. YouTube is trying to compete with TikTok on short-form, vertical content, so you can upload TikTok-style videos to YouTube (called Shorts). My experiment: post videos to both TikTok and YouTube to see which platform is better for economics.
You might think this is a biased experiment because of my established audience on YouTube. But TikTok is known for turning random videos into viral sensations. And this is not the first time I ran this experiment. In January 2021, I posted a short comparing the GameStop short squeeze to the Panic of 1907. On YouTube, with my established audience, the video has 4,800 views. On TikTok, with a new account, it got over 15,000 views. Seems like I have a chance to succeed on TikTok.
The key to this experiment is understanding that TikTok and YouTube have the same objective: get people to stay on the platform for as long as possible. The key to success, then, is to make videos that people want to watch all the way through. Both platforms help you achieve this by reporting statistics on viewer behavior. Here’s an example from YouTube:
This is a report from a video I made about Encanto. The video is 27 seconds long, and the average viewer watched 23 seconds. YouTube is great because it shows you exactly how many people are watching at each second in the video. About 66% of viewers are still watching at second 26. Amazingly, 156% of viewers are watching at the first second. That’s because the video has a narrative loop. TikTok and YouTube both automatically restart the video when viewers hit the end, and if you construct the video properly, you can trick the viewer into watching the beginning of the video again. Both platforms reward this behavior because it means people are watching longer.
One of the key results from this experiment was the huge difference in viewer retention across the two platforms. But more on that after the strategy.
Strategy
The strategy was straightforward. I was going to create two types of videos. For Strategy 1, I would find trending topics that I could connect to economics. Since the demographic on TikTok is mostly teenagers, I needed something that you could imagine seeing on the front page of reddit. That’s how my GameStop video got 15,000 views. This is easily the most common strategy on any social platform: make a video that appeals to the broadest cross-section of people.
For Strategy 2, I would try more niche videos that would appeal to a very specific audience. This is actually the strategy everyone should be following. Not only is it hard to make something that appeals to everyone, it’s incredibly competitive. But if you can find something that appeals to a niche, TikTok and YouTube will quickly promote it to everyone in that niche, and you can make more consistent hits.
You can see the difference in these two strategies by looking at the two most popular videos I published in 2021. At the beginning of the year I posted a video on NFTs, which fits Strategy 1. Then in July I posted an explanation of the assassination of Haiti’s president, perfect embodiment of Strategy 2. The NFT video got 20,000 views, which is a huge success for my channel. The Haiti video got 34,000 views. More niche, more views.
Under Strategy 1, I posted videos on trending movies (Don’t Look Up and Encanto) and under Strategy 2 I posted a video about economics books and a result from an economics paper.
Then, about two weeks ago, I posted a video about MrBeast, the biggest YouTuber in the world. This is a blend between the two strategies: MrBeast is popular with teens, but he’s also a niche on YouTube. That’s the one that went viral.
Results
This past week was the craziest in my channel’s history. Since the channel started in December 2018, I have made 217 videos. Those videos have attracted 5.2 million views. 75% of those views came in the last 7 days.
That’s because my MrBeast video went viral. MrBeast launched a chocolate bar with a sweepstakes, and I calculated the odds of winning it using the basic expected value framework. Super easy. I even slightly messed it up (but got really close to the actual answer anyway!). But nothing exciting. I posted it to YouTube and TikTok.
On YouTube, it got 4 million views.
On TikTok, it got 47. Not 47 million. 47. As in, 50 minus 3.
Why the discrepancy?
One factor is that MrBeast is much more popular on YouTube than TikTok. But the main factor is that people on YouTube watched it longer than people on TikTok.
After 4 million views, the average watch time on YouTube was 36 seconds. The average watch time on those 47 TikTok views was 20 seconds. That’s a large discrepancy on a video that is only 43 seconds long.
And that discrepancy persists across videos. In this experiment, the only video to do well on TikTok was a book recommendation (validating strategy two—target a niche). It got 636 views, and the average watch time was 5.7 seconds. That video has 4,000 views on YouTube, where the average watch time is 17.0 seconds.
For both videos, the average watch time on TikTok is much shorter than the average watch time for the same video on YouTube. This leads me to two hypotheses:
TikTok viewers have been trained to have much shorter attention spans
YouTube does a much better job finding people interested in your videos
It’s easy to sit on my front porch and complain about the first hypothesis, but I can actually use data to support the second hypothesis.
When people think of a viral video, they probably imagine a constant growth in views. But that’s not how YouTube’s algorithm makes videos go viral. It tests them in waves.
Here was the first evidence that the video was going viral.
Over a 48 hour period, there was a discrete jump in the number of views it was getting. The algorithm had been collecting data on the people who watched the video, then it sent it to a wider audience. As the wider audience watched the video, it collected more data and predicted who else would like it. 48 hours later, the views looked like this
You have to look at the right axis for scale, but for reference that big spike on the first graph is the left-most bar in this second graph. What you can see is YouTube keeps experimenting with a wider and wider audience, trying to predict who will like the video.
And it does a pretty good job. Here’s what average view duration looked like as it pushed the video to more people.
Basically flat. It kept finding people who liked the video. This isn’t that impressive until you compare it to my book recommendation video.
Maybe it’s a surprise to you, but people interested in MrBeast’s chocolate bar are not very interested in books about economics. But YouTube doesn’t know that at first. As this video went viral, it started promoting my book recommendation video to the people who watched the chocolate bar video. Below you can see exactly when that push came.
That sudden increase in views led to a decrease in the average watch time.
Before the push, the average viewer watched 21 seconds of the video. During the push, it was 12 seconds. After YouTube saw that these viewers did not like the video, it stopped pushing it. YouTube’s main goal is to get the right video to the right viewer at the right time. If your video isn’t the right one, then it stops promoting it.
So YouTube does a great job finding the people most interested in the video. But, supposedly, TikTok does too. If you watch one video on a drug cartel, suddenly your FYP is flooded with cartel videos.
This leads to a third hypothesis: the people who are interested in economics are on YouTube, not TikTok.
I don’t know how to verify this. I’ll be experimenting more, but I suspect that the type of economics content that succeeds on TikTok is mostly “armchair socialist screeds” and less “economic principles in real-life.” Have any ideas on what will do well? Feel free to comment.
Other Signals
I want to end on one last point that Tyler mentioned.
First, TikTok is one of the dominant modes of presenting and debating issues and ideas, including economics, yet it is hardly used or even discussed by professional economists. (University of Houston Professor Chris Clarke is a notable exception.) Economists are ignoring the market signals — to our own detriment.
I am willing to concede that as far as views are a market signal, then economists are probably underrating TikTok. But if we take prices as a market signal, economists are justifiably ignoring short-form content.
Off of 4 million views, my channel earned $200 from the MrBeast video. Before this video blew up, my most-viewed video was an 11 minute video about majoring in economics. It has about 150,000 views, yet it has earned about $600. And since it addresses an evergreen topic, I expect it to continue to earn money for years to come. If you want a closer comparison, across my whole channel I had about 1.5 million views before the viral video, and the channel had $5,500 in ad revenue. Even with one-third of the views, the content generated 25x more revenue. Long-form videos earn more money because the marginal revenue product is much higher. One long-form video can get you to watch 2-4 ads, while you can watch 20+ short-form videos without seeing a single ad.
So if you want views, try Shorts. If you want revenue, you need long-form content.