Stock-Bond Correlation
Large language models could effectively filter the Fed's “noise” into tradable output. These models measure the Fed’s policy bias—hawkish or dovish. The output looks like an oscillator, such as the relative strength often used in technical analysis.
Running an RSI index on the SPY ETF and the Fed LLM model shows a notable oscillation in the short—and long-term momentum differences. The language model is a hawkish signal in the RSI’s “overbought" territory. Conversely, the SPY’s RSI is to the downside in "oversold.”
While it is not customary to think of Fed speakers as overbought or oversold, the RSI has a relationship with hawkish or dovish biases. The relationship is logical when the SPY moves to “oversold” and the Fed is hawkish.
That is what is happening now (Figure 1). Thus, the current equity and bond selloff should be reversed when the Fed returns to dovish. But in the context of a Fed in a "lame-duck position," that may not be as easy.
Figure 1: Fed and SPY: oversold/overbought vs. hawkish/dovish
Source: FedWatch, SPDR, Bloomberg
Last Friday, the Fed received bad news on inflation expectations. As the Michigan survey pointed out, people are getting anxious about tariffs.
The share of consumers worried about tariffs rose to 24% from less than 2% before the election, and these groups believe tariffs will be passed on in the form of higher prices. These same consumers have been “buying in advance” to avoid future price increases, the proliferation of which would generate further momentum for inflation.
This is having a peculiar effect on the economy. The buy-in advance has accelerated recent activity while, at the same time, expectations for the economy are bullish but for inflation hawkish. Markets price in a lift in growth above potential while the rise in inflation expectations also affects potential output, raising the neutral rate (Figure 2).
Figure 2: Michigan expectations and the estimated neutral rate (%)
Source: Michigan, CME, Bloomberg
The lock-step between Michigan 5-10Y and the market’s pricing of the neutral rate (as measured by the 5Y OIS 10 years from today) has added term premium to bonds, leveraged long positions on the dollar (Figure 4 below), and flipped the stock-bond correlation to positive, highest since the 2022 bear market (Figure 3).
Noteworthy, as the correlation is positive, credit spreads are unnaturally tight. Rising yields on an economy that may overheat imply that a future downturn gets into the cyclical horizon. For credit spreads to correct, tightening financial conditions through positive correlations must result in slowing activity.
Thus, credit spreads will still trade with a significant lag to the selloff in equities.
Figure 3: Stock-bond correlation and credit spreads
Source: Michigan, CME, Bloomberg
In June 2022, Michigan expectations shot up, and the Fed ‘signaled’ jumbo-size rate hikes to control a surge in inflation. Fast forward, and the Fed’s stance may not be hawkish enough to counter inflation expectations.
The lineup of Fed speakers ahead of the blackout (starts on 1/18) is considered centrist, dovish, or moderately hawkish (Williams, Goolsbee, Barkin, Khaskari). This may push the Fed's large language model signal further into the hawkish territory, driving the S&P deeper into oversold.
Regards,
Ben
Figure 4: leveraged bets have driven dollar strength
Source: CFTC