The Big Macro Force That's Been Driving Stocks Higher for Years

ELI5/TLDR

The Shiller CAPE keeps screaming “overvalued” but a Minneapolis Fed economist found that if you swap the denominator from earnings to free cash flow, US stocks look much more reasonably priced — because firms have been keeping more of the pie (less to workers, less to investment) and handing it to shareholders. The catch: big tech is now flipping from cash machines to cash burners as they pour hundreds of billions into AI infrastructure, and nobody knows yet whether the investment will pay off or just crater the metric that made everything look fine.

The Full Story

The metric that cried wolf

Jonathan Heathcote, an economist at the Minneapolis Fed, co-authored a paper called “A Macroeconomic Perspective on Stock Market Valuation Ratios.” His starting point is familiar: the Shiller CAPE and traditional P/E ratios have been flashing red for years. They keep drifting up. Mean reversion never arrives. Everyone who bet on it got run over.

Heathcote’s team found something different when they swapped the denominator. Instead of earnings, they looked at price-to-free-cash-flow — the money actually left over after a firm pays all its bills, including capex.

“If you look at that ratio, the value of all the firms in the US relative to the total cash flow they’re generating — it bounces around a bunch over time. But it doesn’t have a long-term drift.”

The price-to-FCF ratio in 1980 was roughly the same as in Q2 2022. Both near the historical average. The last three years have pushed it above average, but not wildly outside the 70-year range.

Why free cash flow tells a different story

The gap between the two metrics comes down to two forces.

Declining labor share. Corporate wages and salaries have fallen about eight percentage points of GDP since 1980. A larger slice of the pie goes to firm owners. Earnings grew fast — but free cash flow grew even faster.

Low investment intensity. For most of the tech era, the biggest companies generated enormous returns with minimal capital expenditure. Software, SaaS, platform businesses — all relatively asset-light. When you don’t need to spend much to sustain earnings, almost all of it flows through to shareholders.

“These big tech firms have been generating mountains of cash, and their high values — they’re not built on sand, they’re not built on an expectation that we’re going to make big future profits. The profits are there now.”

Firm-level data from Compustat confirms this: about 50 firms account for most of the growth in total market value, and those same 50 firms had the fastest growth in cash flow. Value and cash flow moved in lockstep.

The AI capex problem

Here is where it gets interesting for the present moment. The big tech firms that were cash machines are now spending aggressively on data centers, chips, energy capacity — real, tangible, expensive stuff. Some have flipped to negative free cash flow.

Heathcote’s macro data (through Q3 2025) doesn’t yet show a decline in aggregate corporate FCF — AI capex is booming, but other investment (residential, for example) remains weak, so the total isn’t outsized. But the concentration matters. The handful of firms that drove the entire valuation story are the same ones now burning cash.

“The optimistic view is, well, this is 1 or 2 years of investment that’s going to generate a ton of free cash flow going forward.”

Joe Weisenthal makes a sharp observation: strategists will conveniently cite whichever ratio flatters the market at any given moment. When FCF was strong, everyone pointed to price-to-FCF. Now that FCF is weakening, watch them quietly switch back to P/E.

The inequality angle

If declining labor share is one engine of high valuations, there is an uncomfortable tension. Stock owners and wage earners are not the same people. The market going up is partly a story about workers getting a smaller cut.

AI could push this further — but with a twist. The old automation fear was about robots replacing factory workers. The new fear is about AI replacing knowledge workers. Heathcote notes this could actually compress inequality (nurses and construction workers stay, analysts and coders get displaced), though Tracy Alloway prefers the version where the bottom gets pulled up rather than the top gets pulled down.

The 1980 parallel

Heathcote cites a paper by Hobijn and Jovanovic (circa 2000) that tried to explain why stock prices were so low around 1980. Their story: investors could see the IT revolution coming but didn’t know which firms would win or lose. The uncertainty itself depressed valuations. The winners (Intel, Microsoft) weren’t even public yet.

The AI parallel is imperfect but real. A massive technological shift is underway. The investment is enormous. The winners are not yet obvious — or rather, the current winners (the model builders spending billions) may not be the ultimate beneficiaries. Tracy and Joe note that in 2026, US markets have underperformed the rest of the world. Maybe the real AI winners will be European chemical companies or drug discovery firms that adopt AI cheaply without building the infrastructure.

Claude’s Take

Solid, not spectacular. This is a well-structured Odd Lots episode that lands on a genuinely useful insight: price-to-free-cash-flow is a better valuation metric than P/E for understanding the long arc of US equities, and it tells a much less alarming story. That alone is worth knowing.

Heathcote is careful and measured — refreshingly so for a Fed economist on a podcast. He doesn’t oversell, doesn’t make predictions, and freely admits the limits of quarterly macro data. The intellectual origin story is nice: they stumbled into this from research on the US net foreign asset position, not from chasing headlines. That lends credibility.

The weakness is that the episode doesn’t quite close the loop on the most pressing question: what happens when the firms that drove the entire FCF story simultaneously flip to massive capex mode? Heathcote’s aggregate data doesn’t show it yet, and he’s honest about that. But the concentration risk — 50 firms driving most of the value — means the aggregate might be misleading precisely when it matters most.

The inequality discussion is interesting but underdeveloped. It surfaces the tension without resolving it, which is fine for a 35-minute podcast but leaves you wanting more.

Score: 7/10. A clear, well-argued piece of research translated competently for a general audience. Not groundbreaking if you already think in FCF terms, but the macro data backing it up and the historical perspective add real value.

The Big Macro Force That's Been Driving Stocks Higher for Years | Odd Lots