When the Data Argues Back | Five Lessons from Joe Davis
Vanguard’s chief economist on trusting the structure when the story gets loud.
Joe Davis helped build a framework he describes as a “living, breathing system.” Inside it, the forces that shape the economy do not line up neatly in a forecast; they compete in real time, in what he calls a horse race. On one side is artificial intelligence and the growth it might unleash. On the other are the headwinds every developed economy carries: aging populations, rising debt, and the slow retreat of globalization. The race has no permanent winner, and the framework exists to tell Joe, quarter by quarter, which way the wind is blowing.
Joe is Vanguard’s global chief economist and the author of Coming Into View: How AI and Other Megatrends Will Shape Your Investments. What makes his work distinct is the discipline and long-term focus behind it. He built a machine designed to keep the story from outrunning the evidence and, as he put it, to keep him out of “the narrative business.”
Here are five things we took away from our conversation.
Lesson 1: Follow the Technology Out of the Tech Sector
When we asked where the second wave of AI winners might be, Joe did not start with a stock list. He started with the history of general purpose technologies and how their impact shows up far beyond the companies that first commercialize them.
Technology, in his framework, works along three channels: it automates work, it augments workers, and in some cases it becomes a platform for entirely new products and industries. Electricity, the internal combustion engine, and the personal computer all fit that pattern. Each began in a relatively narrow sector and ended as infrastructure other industries built on.
He drew the line from that history to AI directly. If AI follows the same path, its biggest economic impact will not come only from the firms that develop the models, but from the businesses that learn to use it to change how they operate. In his language, the question is whether AI becomes a “general purpose technology like the personal computer,” a platform that can “enable new products, new industries.”
Joe’s emphasis here was not on picking sectors in a vacuum, nor on abandoning the companies building AI. He was pointing to the places where adoption can lift productivity in a service-based economy that has lacked automation for decades. The investors who stop at the obvious technology layer risk missing the everyday businesses that quietly compound the gains.
Lesson 2: Growth Is Not Always a Demand Story
One of Joe’s findings genuinely surprised him, and he said so plainly. “I was shocked to find,” he admitted, what happened when his team integrated long-term megatrends into the standard business-cycle picture.
Macroeconomics and asset pricing often treat short-term swings in GDP as a demand story: if growth speeds up, consumers must be spending more, so the central bank should raise rates to cool them off. Joe’s work found that explanation was only half the story. “What we find is that’s true half the time,” he said. “But the other half of the time, it is the emergence of new ideas,” with investment accelerating because productivity and innovation are about to shift.
He also found that changes in those megatrends explain roughly half of the S&P 500’s quarterly movements, a result he did not know going in and was not looking for. Long-term forces, in other words, were showing up inside the short-term noise far more often than the usual models assume.
His analogy is the one any commodity trader would recognize. The price of oil depends not only on how much we drive but also on the supply of oil, and growth works the same way. There is a supply of ideas, and when it rises, the same GDP number can mean something very different. If growth and earnings are rising because innovation is increasing, inflation ultimately can ease, rates may not have to rise with debt, and higher valuations can be supported instead of punished. The investor who treats every strong number as overheating demand keeps bracing for a rate shock that, half the time, was never coming.
Lesson 3: Put Every Worry on One Ledger
Strip away the 130 years of data, Joe suggested, and the most useful thing he does is something any investor can copy. He keeps what he called a multi-factor scorecard and described it as “really simple.” Oil prices, geopolitical tensions, deficits, AI as a risk and AI as an opportunity all go onto a single page. The point, he explained, is to weigh “all those factors, good and bad, on one ledger” rather than reacting to each in isolation.
The reason is behavioral. A headline about higher oil prices, louder fiscal worries, or a specific geopolitical shock almost always argues for action if you read it alone. On a common ledger next to a trend in innovation or a shift in demographics pulling the other way, those same data points often shrink back to their actual size. Many days, they do not justify changing anything.
What Joe has found, doing this over and over, is that the full ledger rarely calls for drastic action. A single factor examined alone almost always argues for a bigger move than the complete picture justifies. It does not mean he never adjusts; it means the adjustments are smaller and less frantic than the headlines themselves.
The alternative is the trap that catches investors who watch the news closely but do not have a framework. Reacting to each shock as it arrives, he reasoned, “would lead me to really whipsaw a portfolio more often than not.” The scorecard predicts nothing. Its job is to keep you anchored to a policy when the data get noisy, so your responses reflect the whole environment rather than whichever story happened to break today.
Lesson 4: When the Data Argues With Your Priors, Lose the Argument
For most of his career, Joe told us, he was a 2 percent growth, 2 percent inflation guy. “I was very comfortable in my 2 percent growth, 2 percent inflation planning world,” he said. It always made sense: “We may get a little bit of technology lift, yeah, but we got all these negatives” from demographics, debt, and other headwinds on the other side. He filled out consensus surveys with those numbers and did not think twice.
His own model took that comfort away. When he and his team ran AI and the other megatrends through their framework, the work pointed to economic growth projections 50 percent above the consensus, with most of the gap coming from technology rather than anything visible in the old data. In a disappointing AI scenario, the same system assigned a non-trivial probability that the 10-year Treasury yield could reach above 9 percent over the next five to ten years. These are not the numbers of an economist hedging toward the crowd.
What he refused to do was dress any of it up as a hunch. The baseline that AI could be more transformative than the personal computer, and the tail risk that disappointing AI could leave deficits dominating, come from the analytics, not from his personal instincts. “We did not want to be in the narrative business or, ‘Hey, it’s Joe’s personal opinion that AI could be transformational,’” is how he put it. “It’s not my opinion,” he said, insisting on separating the finding from himself.
There is a discipline in that worth borrowing. The forecast that flatters your existing view is the one to interrogate hardest, and the conviction worth holding is the kind you arrive at reluctantly. Joe did not reach a bolder call on growth because it made a good story; the data dragged him there over his own objections, and he let it.
Lesson 5: Being Right About the Technology Won’t Spare You the Drawdown
Early in the conversation, Joe offered a sentence that should give every AI optimist pause. “There has never been a great technology that has not had a significant drawdown in stock prices,” he said, “which is code word for saying what some would say there’s a bubble forms.” In his data, a painful reset has shown up alongside every major technology, including the ones that went on to change the world.
He is reluctant to use the word bubble himself, and the reason is precise. To most listeners, he noted, the word evokes tulips, something with “no intrinsic value other than looking at a…picture of a flower.” AI is not that. These technologies can lift productivity, create industries, and raise living standards. But the market built around them still consolidates, sometimes sharply, while the transformation is underway. The economic story and the equity story separate for a time.
Joe’s concern is timing rather than authenticity. Either “the trend for growth is going materially higher because of the innovation of AI that overcomes the demographics and the debt levels we have,” which he described as “by far our most likely outcome,” or AI disappoints and deficits dominate. In both paths, stocks can move ahead of the economy, and what really runs on the way up tends to sow the seeds of a future drawdown.
For an investor, that gap is the whole point. Conviction about AI’s long-term impact does nothing to protect you from a consolidation that arrives while the rest of the market, and the economy, are still catching up. Joe is not calling AI tulips. He is reminding us that in his work, being right about the technology and avoiding a significant drawdown have never been the same thing.
The Bottom Line:
What Joe believes, underneath all the modeling, is that investors are hurt more by loyalty to a story than by being wrong about a single forecast. We commit to a narrative, the bubble that must pop or the boom that cannot, and then we defend it against the evidence instead of updating with it. His whole apparatus is a guard against that loyalty. It forces every force into the same arena, lets them compete, and reports the score without flinching when the answer is inconvenient.
Most of us will never run 130 years of data through a living, breathing system, or parse the business cycle into demand and the supply of ideas. But the posture is portable. It means holding your views in pencil, putting each frightening headline onto the same ledger as everything else you know, and being most suspicious of the forecast that feels most comfortable. The horse race does not care which horse you backed. Joe’s edge is only that he watches it as it actually runs, and he has built a structure willing to tell him when his own favorite story is losing.
Watch the full episode here: