Is the "buffer" for US stocks about to disappear?

👆Click the blue text to follow us

JPMorgan Fund Flow Report

Now let’s look at this week’s Flow and Liquidity report. Last time, we mentioned that JPMorgan believes the revenue growth of AI companies is beginning to catch up with capital expenditures, making this round of AI construction more economically viable than it was six months ago. However, JPMorgan also set an important boundary: revenue growth only indicates that demand is materializing, but whether it can support stock prices depends on profit margins, return on capital, and valuation levels.

In this issue, JPMorgan pushes the discussion to the valuation level. According to JPMorgan, the equity risk premium (ERP) of the S&P 500 has dropped to 2.1%, the lowest level since 2002. What does this mean?

We know that investing in stocks comes with risks such as profit fluctuations, valuation reductions, and operational failures, so the expected return on stocks theoretically needs to compensate for these risks and be higher than the risk-free rate—meaning higher than the 10-year U.S. Treasury yield. The portion above this rate is known as the equity risk premium.

So, what impact does the equity risk premium falling to 2.1% have? JPMorgan summarized three points.

First, U.S. stocks will become more sensitive to Treasury yields.

When the equity risk premium is high, even if Treasury yields rise, it does not affect stocks much because stocks have a sufficiently large excess return as a buffer (high ERP).

However, now this buffer is only 2.1 percentage points. If real Treasury yields rise further, investors will demand higher returns from stocks. How does the stock market provide higher returns? That’s right—corporate profits and growth expectations must be revised up accordingly. Otherwise, the market can only restore attractiveness by compressing valuations. That’s the first impact.

Second, long-term funds may rebalance from stocks into bonds.

JPMorgan’s asset allocation indicators show that global non-bank investors, as well as pension and insurance funds in the US, UK, Europe, and Japan, have their largest overweight in stocks versus bonds since 2002. In the past, the main reason for long-term funds heavily favoring stocks was that the expected return on stocks was noticeably higher than on bonds. Now, as the return gap narrows quickly, the relative attractiveness of bonds in asset allocation is rising.

If the real yield on 10-year U.S. Treasuries continues to rise, pensions, insurance companies, and multi-asset funds may sell some stocks and shift into bonds. Due to the large size of these funds, such rebalancing doesn’t have to happen all at once; as long as it continues, it could form a medium- to long-term fund flow that suppresses stock valuations.

However, this is just an assessment. As of September 2, global equity funds averaged $11.2 billion of net inflows per week over the past four weeks, and bond funds had $12.4 billion, so both asset classes are still absorbing funds. The difference is that U.S. equity funds only saw $1.6 billion per week (below the 2025 average of $3.5 billion), while U.S. bond funds had $6.8 billion (above the 2025 average of $4.3 billion).

From a relative perspective, JPMorgan believes the pressure for stock-to-bond rebalancing is rising, but in the short term, investors are not showing a comprehensive retreat from equities.

Also, the current low ERP can still persist for a long time. If AI investments ultimately improve productivity and drive sustained profit growth, stocks may still handle current valuation pressures through stronger earnings. The main risk is that profit growth will be slower than Treasury yield increases. This is the second impact.

Third, stocks and bonds could become more likely to rise and fall together.

JPMorgan noted that the traditional 60% stocks, 40% bonds portfolio treats bonds as a hedge for stocks. When the economy weakens and stocks fall, U.S. Treasury yields usually drop, bond prices rise, and this 40% bond allocation offsets part of the stock losses.

But since the inflation shock of 2022, the correlation between stocks and bonds has turned positive again. When inflation rises, higher Treasury yields suppress both stock valuations and bond prices—the two assets can now fall together.

JPMorgan believes the current low ERP will further strengthen this positive correlation. On one hand, higher Treasury yields more easily suppress stock valuations via the valuation channel; on the other, if inflation remains above pre-pandemic levels for a long time, both stocks and bonds will remain driven by the same macro factors.

As a result, high stock-bond correlation weakens the hedging effect of bonds and may also force multi-asset investors to purchase more put options on stocks to manage downside risk using derivatives.

This brings up a key question. Is the recent rise in real Treasury yields a sign the market is optimistic about AI-driven economic growth, or does it reflect investors demanding higher term premiums? Both factors push yields higher, but the implications for stocks are very different.

If the rise in real yields comes from improved expectations for economic growth and productivity, then future corporate profits may also improve and provide support for equity valuations.

But if the rise mainly reflects fiscal deficits, bond supply, and term premiums, corporate profits might not benefit and stocks would face pure valuation pressure.

JPMorgan cited the Federal Reserve’s updated DKW model to break down the real reason behind rising real yields. This model splits the 10-year Treasury yield into two parts: one is the market-expected real yield (representing economic growth), and the other is the term premium (representing extra compensation investors demand for locking up funds long-term).

The model shows that, since 2022, the rise in real yields has primarily come from real rate expectations. But since late February this year, about half was due to higher real rate expectations and the other half to rising term premiums.

This suggests the recent rise in real Treasury yields includes both optimism about economic improvement and risk compensation from fiscal deficits, bond supply, and tighter monetary policy by other developed-economy central banks. Thus, this signal is mixed, and rising yields can't be simply interpreted as wholly bullish or bearish. Only half the move hints at economic growth, which would benefit stocks.

Finally, JPMorgan updated on how hedge funds recovered in August after July’s deleveraging.

According to preliminary data from Pivotal Path, all hedge funds rose 1.4% in August, recovering nearly all of July’s 1.36% drop. Both equity long-short and multi-strategy funds also recouped July’s losses. Macro hedge funds and CTAs gained 2.14% and 2.31% in August, not only recouping July’s drop but with macro and equity long-short funds’ beta rising significantly, indicating these funds re-increased equity exposure.

However, one sector lagged significantly. Preliminary figures show TMT sector funds fell 8.27% in July and rose only 2.66% in August—recovering about a third of the loss. While still up 19.54% for the first eight months of the year, TMT funds did not rebound as quickly as other hedge funds in August, indicating that risk budget constraints from July deleveraging still lingered.

This aligns with JPMorgan’s previous view. Overall market risk appetite has recovered, with macro hedge funds, CTAs, equity long-short, and multi-strategy funds all resuming risk-taking; but the more TMT-concentrated funds, particularly in semiconductors, AI, and tech, are recovering at a slower pace. In other words—PTSD (post-traumatic stress disorder).

Jason believes there is an interesting point in this report—the lowest equity risk premium since 2002.

First, as JPMorgan mentioned, ERP is back-calculated, not directly observed from market pricing, so results can vary widely based on model, method, and environment.

I checked U.S. ERP on Bloomberg, and the results are completely different from the 2.1% that JPMorgan mentions, or any historically low ERP. We don’t need to focus on the exact percentage—as mentioned, different models yield different results—but we can focus on the trend of the ERP.

From the chart, you can see that since 2000, the U.S. ERP structurally shifted higher, declining at the end of 2001 (when the dot-com bubble burst), then slowly rising up to the financial crisis, fluctuating afterward for quite some time, dropping a notch during the pandemic, then seeing a notable increase since late last year. Two preliminary conclusions can be drawn from these data.

First, ERP can remain at a certain level for a long time—for example, from 2014 to 2020, ERP barely fluctuated, yet during that period, the U.S. stock market enjoyed a prolonged bull run.

Second, since late 2025, the current ERP is actually at a historical high. This means the difference between U.S. earnings expectations and Treasury yield rises is growing, not narrowing as JPMorgan suggested.

I remember about half a year ago, Dalio said the U.S. ERP was negative; today, JPMorgan claims it’s 2.1%, while Bloomberg data puts it close to 10%. Clearly, ERP varies greatly by environment and model. So, we can use its trend as a reference but should always interpret the absolute value rationally.