Blog
Who predicted the recession? YouTubers' calls, checked
By Callsight · Published Oct 9, 2026 · Updated Oct 12, 2026
Figures as of Nov 4, 2026. They change as new windows settle.
As of Nov 4, 2026, Callsight’s record holds 24 forecasts by four finance YouTubers saying, in the speaker’s own words, that a US recession would come. All 21 with a one-year verdict are Incorrect, against a base rate of 24.1% on the same calls: the share of past one-year windows that held a recession month.
Who predicted a recession, and did one come?
Callsight counts 24 forecasts, from four channels, that said in the speaker’s own words that a US recession would come. At one year, 21 have a verdict and none of them is Correct; 3 are still open. The National Bureau of Economic Research (NBER), whose dates decide these calls, has dated no recession since the one that ended in April 2020.
The table splits each channel’s counted calls on a US recession by what they said. Each window shows how many were Correct out of the calls decided there.
| Channel | Said | Calls | 30 days | 90 days | 180 days | 1 year | 3 years | 5 years |
|---|---|---|---|---|---|---|---|---|
| Coin Bureau | A recession would come | 9 | 0 of 9 | 0 of 9 | 0 of 9 | 0 of 9 | 0 of 4 | — |
| Coin Bureau | None would come | 2 | 1 of 1 | 1 of 1 | 1 of 1 | — | — | — |
| Andrei Jikh | A recession would come | 8 | 0 of 7 | 0 of 7 | 0 of 7 | 0 of 6 | 0 of 5 | — |
| Graham Stephan | A recession would come | 4 | 0 of 4 | 0 of 4 | 0 of 3 | 0 of 3 | 1 of 2 | 1 of 1 |
| Joseph Carlson | A recession would come | 3 | 0 of 3 | 0 of 3 | 0 of 3 | 0 of 3 | 0 of 2 | — |
| Base rate, calls that said one would come | 11.5% (23) | 13.9% (23) | 16.9% (22) | 24.1% (21) | 42.9% (13) | 59.1% (1) | ||
One of the calls that said a recession would come, Coin Bureau’s, is filed as an economic indicator, so its card says the recession will rise. The rule judged it as a call that one would come, and it is counted that way here. One more counted call is about another country’s economy, which the rule cannot check against US dates, so it is left out. Another 68 calls on a recession from these videos are listed and never counted: 28 forecasts read out from someone else, each shown with its source, and 40 of the channels’ own, held back because their direction could not be read with confidence or because a review questioned how they were recorded. Among them, 13 of Coin Bureau’s recession calls, 2 of Andrei Jikh’s and 1 of Graham Stephan’s are held out of every count because a review found that the way they were recorded may not match the words; each card gives the reason. Every figure here is of calls filed under a recession, and calls on a technical recession (two quarters of falling output) are outside this article.
Three of the forecasts that said a recession would come, one for each channel that had a call meeting the rule at the end of this article:
“it seems like a recession is pretty much inevitable at this point”No recession is dated between Feb 7, 2022 and Feb 7, 2023. More than 15 months have passed since the window closed, so the period is settled.
30 days
Incorrect
90 days
Incorrect
180 days
Incorrect
1 year
Incorrect
No deadline was stated; the headline is the one-year window, where it is Incorrect. At 3 years it is Incorrect; its window at 5 years is still open.
“from where I'm standing I think that it will be hard to Stave off a recession”No recession is dated between May 4, 2022 and May 4, 2023. More than 15 months have passed since the window closed, so the period is settled.
30 days
Incorrect
90 days
Incorrect
180 days
Incorrect
1 year
Incorrect
No deadline was stated; the headline is the one-year window, where it is Incorrect. At 3 years it is Incorrect; its window at 5 years is still open.
“it's all but confirmed that we're very likely to head towards a recession”No recession is dated between Jun 1, 2022 and Jun 1, 2023. More than 15 months have passed since the window closed, so the period is settled.
30 days
Incorrect
90 days
Incorrect
180 days
Incorrect
1 year
Incorrect
No deadline was stated; the headline is the one-year window, where it is Incorrect. At 3 years it is Incorrect; its window at 5 years is still open.
The forecasts that said a recession would come, by the year they were made, at one year:
| Year made | Calls | Correct | Incorrect | Still open |
|---|---|---|---|---|
| 2018 | 1 | 0 | 1 | 0 |
| 2022 | 15 | 0 | 15 | 0 |
| 2023 | 4 | 0 | 4 | 0 |
| 2024 | 2 | 0 | 1 | 1 |
| 2025 | 1 | 0 | 0 | 1 |
| 2026 | 1 | 0 | 0 | 1 |
At one year, the 3 forecasts made in 2024 and 2025 that a recession would come stand at 0 Correct, 1 Incorrect and 2 still open. The 2 counted calls that said no recession would come were made by Coin Bureau, and both are still open at one year. Every call on the subject, from every channel and with its reason where it is not counted, is on the recession page.
Why does an Incorrect recession call take so long to settle?
The NBER dates a recession after it has started, and sometimes long after it has ended. Its dating committee announced on Dec 1, 2008, that the economy had peaked in December 2007, the start of that recession, and on Jun 8, 2020, that it had peaked in February 2020. The end of the 2020 recession, in April of that year, was dated on Jul 19, 2021.
The rule waits for the dating. A month dated as recession inside a call’s window settles the call as soon as the date is published: Correct if it said one would come, Incorrect if it said none would. With no month dated, nothing is settled until 15 months after the window closes, about as long as the committee needed to date the end of each of the last two recessions (it dated a trough in June 2009 on Sep 20, 2010, and the April 2020 one on Jul 19, 2021). The call is then Incorrect if it said one would come and Correct if it said none would. A forecast checked over one year therefore waits 27 months from the day it was said before it can be Incorrect. The 3 forecasts still open at one year are waiting for the year to end or for the months after it to pass.
What is the base rate for a recession?
It is how often a recession was dated anyway. For each call, the method takes every past window of the same length that had closed by the day the words were said, back to 1980 in FRED’s recession indicator (USREC, which marks the months the NBER dates as recession), and counts the share that held at least one recession month. A longer window catches a recession more often, so the base rate grows with the window. Averaged over the forecasts that said one would come and have a verdict, it is 11.5% at 30 days (23 calls), 24.1% at one year (21) and 42.9% at 3 years (13). The table above gives every window.
At one year the count is 0 Correct out of 21, beside a base rate of 24.1%. If the 21 had been separate tries, each with that chance, about 5 would have come true. They were not separate tries: 19 of them were made in 2022 and 2023, about the same economy and over windows that overlap, so a single recession in those years would have turned many of them Correct together, and the NBER dated none. Professional forecasters saw the same risk at the time. In November 2022 the 38 forecasters in the Philadelphia Fed’s Survey of Professional Forecasters put the chance that real GDP would shrink in a given quarter of 2023 at between 43.5% and 49.4%, and the minutes of the Federal Reserve’s March 2023 meeting record that its staff projected a mild recession starting later that year.
At 3 years, one forecast that said a recession would come is Correct: Graham Stephan’s from October 2018, whose window takes in the recession of 2020.
What did the Sahm rule and GDP show meanwhile?
Two faster signals are watched while the dating committee works, and neither decides a verdict here. The first is two quarters of falling real GDP in a row, often called a technical recession. When the Bureau of Economic Analysis published its first estimate for the second quarter of 2022, on Jul 28, 2022, it showed output shrinking for a second quarter running. The second is the Sahm rule, which signals a recession when the three-month average of the unemployment rate rises 0.50 percentage points above its lowest three-month average of the previous 12 months. Its real-time version on FRED (SAHMREALTIME) crossed that line with the July 2024 figures, published on Aug 2, 2024. The NBER has dated a recession after neither.
A forecast here is judged on the NBER’s dates alone, so neither signal moved a verdict in this article. A call that leaned on one of them is judged like any other: Incorrect once the wait is over, if no month inside its window is dated as recession.
Market correction vs crash: what counts as which?
There is no official line between a correction and a crash. FINRA, which regulates US broker-dealers, describes a correction as a reversal of at least 10 percent, and the SEC’s Investor.gov says a bear market generally means a broad market index falling 20% or more over at least two months. Callsight’s rules use those two sizes for a correction and a crash, and add a third for a bubble bursting:
| The call | Correct at a fall of | Partly correct at a fall of | Measured on |
|---|---|---|---|
| Market correction | 10% | 5% | The S&P 500 (SPY), from its highest close since the call |
| Market crash | 20% | 10% | The S&P 500 (SPY), from its highest close since the call |
| Bubble burst | 30% | 15% | The S&P 500 (SPY), from its highest close since the call |
| Housing crash | 10% | 5% | The national Case-Shiller index (CSUSHPINSA), from its highest figure since the call |
The fall has to come inside the window. On the stock market it is measured from the first close after the video went public and from any higher close after that, so a market that rises first and then falls the full amount counts; on housing, from the index’s last figure at the time of the call and any higher figure after it. How fast the fall comes does not enter the rule. On the stock market, a larger size the speaker names is used in place of the table’s, provided it is in the speaker’s own words and is neither odds nor the top of a range. A call that the fall will not come is judged the other way round: Correct while the fall stays under half the size, Incorrect once it reaches the full size, and Partly correct in between. The rules in full are on the methodology page.
How did the crash and correction calls turn out?
The crash rule reads the US stock market as a whole, through SPY, so a call about crypto, one sector or another market is Not verifiable on it. Of the counted calls filed under these topics, 64 are of that kind, 55 of them from Coin Bureau. Each topic’s page lists them with the reason, and the table leaves them out. All 6 counted calls on a bubble bursting are among them. The calls the rule could read, at one year:
| Channel | Said | Calls | Correct | Partly correct | Incorrect | Still open |
|---|---|---|---|---|---|---|
| Market correction: a fall of 10% | ||||||
| Graham Stephan | A correction would come | 6 | 4 | 1 | 0 | 1 |
| Graham Stephan | None would come | 1 | 0 | 0 | 1 | 0 |
| Coin Bureau | A correction would come | 3 | 0 | 3 | 0 | 0 |
| Andrei Jikh | A correction would come | 2 | 2 | 0 | 0 | 0 |
| Joseph Carlson | A correction would come | 1 | 1 | 0 | 0 | 0 |
| Market crash: a fall of 20% | ||||||
| Graham Stephan | A crash would come | 6 | 1 | 3 | 2 | 0 |
| Graham Stephan | No crash would come | 3 | 1 | 0 | 0 | 2 |
| Coin Bureau | A crash would come | 4 | 0 | 2 | 2 | 0 |
| Joseph Carlson | A crash would come | 2 | 0 | 2 | 0 | 0 |
| Joseph Carlson | No crash would come | 1 | 1 | 0 | 0 | 0 |
Many are still open at the longer windows: at 3 years, 13 of these 29 calls have no verdict yet, and at 5 years, 22. Every call is on the market correction, market crash and bubble pages.
Who predicted a housing crash?
Mostly, the forecasts said the opposite. Of the 20 counted calls on a US housing crash, 18 said none would come and 2 said one would. At one year, those that said none would come stand at 12 Correct, 2 Partly correct, 0 Incorrect and 4 still open, and the 2 that said one would come at 0 Correct and 2 Incorrect. Most came from one channel: Graham Stephan made 17 of the 20. The most recent counted call that said a US housing crash would come was made in August 2022. Two more counted calls were about housing outside the US and are Not verifiable.
| Channel | Said | Calls | Correct | Partly correct | Incorrect | Still open |
|---|---|---|---|---|---|---|
| Graham Stephan | A crash would come | 1 | 0 | 0 | 1 | 0 |
| Graham Stephan | No crash would come | 16 | 11 | 1 | 0 | 4 |
| Coin Bureau | A crash would come | 1 | 0 | 0 | 1 | 0 |
| Coin Bureau | No crash would come | 1 | 0 | 1 | 0 | 0 |
| Andrei Jikh | No crash would come | 1 | 1 | 0 | 0 | 0 |
A housing crash call is judged on the S&P Cotality Case-Shiller U.S. National Home Price Index (CSUSHPINSA on FRED), as first published. A fall of at least 10% from its highest figure since the call, inside the window, is a crash, and a fall of at least 5% is Partly correct. A call that no crash will come is Correct while the fall stays under 5%. Two of the calls that said none would come, in the speaker’s words:
“even though that sounds severe, it's probably not going to lead to an all-out crash.”Reading when said
300.41
Sep 14, 2022
Reading at close
312.31
Sep 14, 2023
Change
+11.9
The house price index (CSUSHPINSA, first releases) stood at 300.4 in September 2022, the figure at the start of the window. Between then and September 2023, its deepest fall from a running peak was −2.6%, in January 2023. A fall of at least 10.0% counts as a housing crash, and at least 5.0% as partial. The call was that no housing crash would come.
30 days
Correct
90 days
Correct
180 days
Correct
1 year
Correct
No deadline was stated; the headline is the one-year window, where it is Correct. At 3 years it is Correct; its window at 5 years is still open.
“we're not going to be seeing a 10 to 20% real estate collapse anytime soon outside of a few localized markets.”Reading when said
331.51
Jun 9, 2025
Reading at close
336.66
Jun 9, 2026
Change
+5.15
The house price index (CSUSHPINSA, first releases) stood at 331.5 in June 2025, the figure at the start of the window. Between then and June 2026, its deepest fall from a running peak was −1.5%, in January 2026. A fall of at least 10.0% counts as a housing crash, and at least 5.0% as partial. The call was that no housing crash would come.
30 days
Correct
90 days
Correct
180 days
Correct
1 year
Correct
No deadline was stated; the headline is the one-year window, where it is Correct. Its windows at 3 years and 5 years are still open.
Every call on the subject, from every channel, is on the housing crash page. The article on Graham Stephan’s calls on housing goes through his housing-crash calls in more detail, with his calls on house prices.
How were these checked?
Each call is quoted word for word from the video and linked to the second it was said. It is checked at six windows (30 days, 90 days, 180 days, 1 year, 3 years and 5 years), and the one-year window is the headline. A recession call is judged on FRED’s recession indicator (USREC) as first published, which follows the NBER’s dates as they stood at the time. A fall in the stock market is measured on the closes of SPY, a fund that tracks the S&P 500, from the first close strictly after the video went public. A housing call is judged on the Case-Shiller index as first published, from its last figure at the time of the call. A forecast read out from someone else is listed with its source and never counted. The rules are the same for every channel, and the methodology page gives them in full. If you think a call was misread, the corrections page says how to report it. Two articles check single channels the same way: Andrei Jikh’s Bitcoin and economic calls and Graham Stephan’s calls on housing, rates and stocks.
How these examples were chosen
The cards were chosen on Oct 7, 2026 by a fixed rule. For a recession, the candidates were each channel’s counted calls that said a US recession would come and have a one-year verdict. A call was passed over if its reading is being checked again, if the deadline on its card does not match the words, or if its words could fairly be read as something other than such a forecast: a historical average, a remark on rising odds, whether the economy was already in one, a recession abroad or worldwide, somebody else’s argument, a recession at some unnamed point, or only a possibility. Of the calls left, each channel’s card comes from its most widely watched video, the first such call said in it. On that day none of Joseph Carlson’s calls was left. On housing, where Graham Stephan made 17 of the 20 counted calls, the cards are two of his calls that said no crash would come and have a one-year verdict: the one from his most widely watched video and his most recent one. Every call is listed on each channel’s page and on the recession and housing crash pages.
Figures are computed by code from the same record as the rest of the site. This article was drafted with the help of AI tools and reviewed by a person before publication. It is not investment advice.