The Fed Removed Its Own Warning System

Warsh removed forward guidance from the July 29 statement and published no dots. The vote was 9-3, with Hammack, Kashkari and Logan dissenting for a 25bp hike. Musalem, Aug 6: financial conditions remain highly accommodative, many asset prices remain elevated, the central bank must act on its own judgment rather than be led by markets, and "there are moments when the central bank can surprise markets." Barkin, Aug 13: the Fed is "not currently in a forward guidance place." Also: whether rates must rise to reach 2% remains an open question; "AI investment seems impervious to the level of rates"; current federal debt levels are an inflationary wind the Fed must navigate against. Hammack, Aug 13, the same morning PPI printed soft: "My view is we need to act now." The labor market is stable, payroll data is noisy, the last two inflation reports are welcome but she has no confidence the progress continues, and "if it takes another 3-4 years to get to 2% inflation, can the public wait?" She also named her financial-stability watchlist: leverage being used to buy Treasuries, private credit, and whether AI is a bubble. Against that, prediction markets now price roughly 70% hold and 29% for a 25bp hike in September, down from a 50/50 split in interest rate swaps a week ago. VIX printed 14.84, a seven-month low. One-month SPX skew sits at the 0th percentile of the past year.

Two historical episodes frame this, and they are not the same risk. September 4, 1987. Greenspan had been chairman for three weeks. His first move raised the discount rate from 5.5% to 6%, a surprise into an extended, concentrated market with no communication track record for anyone to calibrate against. The market topped August 25. October 19 was down 22.6%. The parallel is not the hike, it is the configuration: a new chair with something to establish, no history for the market to read, and stretched positioning underneath. February 1994 is the better analog for the regime rather than the event. Before that year the Fed did not announce rate decisions at all; markets inferred policy from open market operations. Greenspan's first announced move was 25bp on February 4, 1994. What followed was roughly 300 basis points inside a year, including 50 in May, 50 in August and 75 in November, and it produced the bond market massacre, Orange County and the tequila crisis. The lesson is that a Fed without a communication apparatus does not deliver one surprise. It delivers a sequence, because every meeting becomes live and the market has no anchor for the terminal rate. Hammack has already described "some number of rate hikes" with an explicitly unspecified endpoint. The relevant modern calibration is smaller than either. The Bank of Japan raised by 15 basis points in July 2024 and triggered the August 5 carry unwind. The size of the move was irrelevant. What mattered was the shift in the distribution of expected policy, because levered positions are sized to that distribution rather than to the level. The current positioning is the reason this matters. CTAs have tripled their global bond underweight to a record, with roughly $300M of P&L impact per basis point on the 10-year per UBS. Margin debt is at $1.50T with the net credit balance at negative $1.06T. Households hold roughly 46% of financial assets in equities, the highest in an eighty-year series. Dealer gamma is long below spot and effectively absent above it, so a decline is cushioned only until it passes through the strikes where that gamma lives. Downside protection is at its cheapest in twelve months. On Jackson Hole, the risk is misidentified. The Fed does not announce policy there, but chairs use it to reframe: Bernanke pre-signalled QE2 in 2010 and QE3 in 2012, Powell introduced average inflation targeting in 2020 and delivered the "pain" speech in 2022. Timiraos has reported that Warsh's task forces appear aimed at building a framework to replace the current one, which remains in effect until then. The live risk at Jackson Hole is therefore a framework announcement rather than a rate signal, and that is the larger event, because it changes what every subsequent print means rather than what one meeting does. The asymmetry that makes every option unsatisfying is contained in Barkin's own sentence. If AI investment is impervious to the level of rates, then a hike does not slow the thing generating the inflation. It reaches housing, where mortgage demand is already softening at one-year-high rates. It reaches the consumer Barkin describes as "bound and determined to spend, and finding ways to finance it." It reaches the levered Treasury basis and private credit, both of which Hammack says she watches. And it reaches every operator running interest expense at a quarter of revenue. The Fed can only reach the parts of the economy that are not causing the problem.