3.3% became 3.0%: the inflation the Fed hiked on was revised away two weeks later — how reliable is the data your MPF bond bet rests on?

The core proposition: the hike was built on numbers that no longer exist

On 16 September, the Federal Open Market Committee voted 12–0 to raise the federal funds rate by 25 basis points to 3.75%–4.00% — the first Fed rate hike since July 2023. Chair Kevin Warsh’s justification at the press conference was blunt: “this summer’s inflation readings do not tell me that underlying inflation trends have meaningfully improved.”

The “summer readings” he was holding: July core PCE inflation at 3.3%, as first published on 26 August.

Then, on 30 September, the US Bureau of Economic Analysis (BEA) did something that sounds boring: it rewrote the methodology behind three components of the PCE price index — portfolio management and investment advisory services, legal services, and computer software and accessories — and applied the revisions retroactively to January 2021. Five years of inflation history, rewritten in one sweep. After the rewrite, the very same July reads: core PCE down from 3.3% to 3.0%, headline PCE down from 3.7% to 3.4%.

Not a single July price moved. What moved was the ruler.

That is the first truth MPF members need to face: “data-dependent” monetary policy stands on quicksand. The numbers behind the Fed’s 16 September hike no longer exist two weeks later.

Through the data: one July, four answers

The before-and-after, from the BEA release via Reuters on 30 September:

Goldman Sachs and JPMorgan had already run the numbers back in July: the methodology change alone would cut May core PCE from 3.4% to 3.2%–3.3%. The actual revision turned out larger than the banks estimated — this is not a tweak, it is a narrative-level rewrite. The entire slope of the post-pandemic inflation story has been pressed down.

The most telling detail hides in the like-for-like comparison. Restated on the revised basis, July headline PCE was actually 3.36% and August was 3.42% — August inflation did not cool, it ticked up slightly. The market’s celebrated “below expectations” was mostly an artefact of the new ruler, not genuinely falling prices. Forecasters built their expectations with the old ruler; the BEA published the result with a new one. That is how the “surprise” was manufactured.

The same annual revision rewrote two more numbers: second-quarter GDP was finalised at 2.2% against a 1.5% consensus, and July’s personal saving rate was lifted from 3.0% to 4.6%. The economy was stronger than thought, inflation weaker than thought, savings fatter than thought — three directions, three rewrites.

Now the dot-plot paradox. The officials’ median projection published on 16 September demands core inflation at 3.4% by year-end (revised up from 3.3% in June). August core inflation sits at 3.0%, so hitting the target requires an average monthly rise of 0.33% from September through December — August’s monthly print was just 0.2%. In other words, the market’s current bet on “one more hike this year” (16 of 18 officials expect at least one more, median year-end rate 4.1%) is not built on “inflation staying high” but on “inflation must re-accelerate.” The policy path is priced on a curve that has not happened yet.

One footnote: Chair Warsh, who voted for the hike, submitted no rate projection of his own. He does no forward guidance — he would not even put his forecast on paper. The Fed says it is data-dependent while its chair declines to write down what he expects.

The CME FedWatch probability rollercoaster says it all: in the week before the 30 September data, October 28 hike odds collapsed from 70% to 51.5%, then to 35% after the release. Numbers get revised, bets get revised — that is what “data-dependent” really looks like: dependence on data that is itself unstable.

The compounding toll: a triple squeeze on the bond sleeve

For MPF members, this is not a Wall Street academic debate — it is real net asset value. A triple squeeze:

First: duration. In a hiking regime, bond-fund price risk is mathematics, not opinion. The Fidelity Hong Kong Bond Fund runs a duration of 3.8 years — every further 100-basis-point rise in yields knocks roughly 3.8% off NAV. The 10-year Treasury sits at 5.25%–5.34% (highest since 2002), the 30-year above 5.64%, and T. Rowe Price’s David Clewell has publicly called a move toward 5.5%–6% on the 10-year “credible.” Put it in per-member terms (average balance HK$343,242, MPF Ratings 24 September): a 30% bond sleeve (about HK$102,973) loses roughly HK$3,913 on another 100bp — no equity crash required, one percentage point of yield is enough.

Second: the 2022 precedent. The same “stubborn inflation → hikes → bonds get hit” script played out four years ago: the Manulife Hong Kong Bond Fund fell 9.46% in 2022 (its worst year on record), the BCT Global Bond Fund fell 13.57%, the HSBC Age 65 Plus Fund fell 13.21%. A 13.57% drawdown on a 30% bond sleeve is about HK$13,973 per member. Bond funds were never the “safe leg” of an MPF portfolio — in a hiking cycle they are simply slower-falling risk assets.

Third: the statutory de-risking mismatch. Under the DIS mechanism, members are automatically shifted about 6.7 percentage points a year into the Age 65 Plus Fund from age 50 — a fund that is roughly 80% bonds. In the 2026 hiking regime, the statutory machinery is marching the members closest to retirement into the highest-duration-risk asset, year after year. This is not a member’s choice; it is the system’s design.

And anyone tempted to “trade the Fed meeting” faces the forward-pricing mechanism: MPF switches execute at unknown T+1/T+2 NAVs. Place an order around the 27–28 October FOMC and you buy the post-meeting price — the event itself can never be traded. The CME odds swinging from 70% to 35% in a week already tells you professional traders have no consensus on the next move; the expected value of timing on headlines is negative. Illustrative arithmetic: a ±2% timing error on the average balance is about HK$6,865 — roughly eighteen months of management fees at the market-average 1.36% charge. One timing attempt, eighteen months of fees gone.

The way out: allocate bonds to need, not to the Fed

If the Fed’s own data gets revised, the only rational response for MPF members is this: do not bet retirement money on the Fed’s next move.

First, size the bond sleeve to need, not to forecast. For members 20-plus years from retirement, bonds in the portfolio are a volatility buffer, not a return engine — hold a strategic weight, keep duration short, and do not load up on long bonds “playing the pivot.” The lesson of 2022: long duration in a hiking cycle is a loss amplifier, not a shelter.

Second, the Conservative Fund is a car park, not a destination. Parking there through a rate-decision week is understandable, but remember two sums: the Conservative Fund’s ten-year annualised return is about 1.2%–1.5% against August underlying inflation of 1.9% — the real return is negative. And the 30-year opportunity cost is brutal: HK$5,000 a month compounding at 1.3% for 30 years reaches about HK$2.24 million; at the DIS Core Accumulation Fund’s 7.3% it reaches about HK$6.47 million — a gap of HK$4.23 million, more than twice the HK$1.8 million of total contributions. Park too long and you park away the compounding.

Third, put discipline on the calendar, not bets on the meeting. The 27–28 October FOMC and the 8–9 December final meeting of the year — no tactical switching in the three trading days around either window. Instead of short-term switching around the Fed, use the once-a-year free ECA transfer where it counts: move balances out of high-fee schemes into the market’s lowest-fee plans and lock in a certain 0.5–1 percentage point a year — money that needs no forecast of whether inflation data gets revised again.

Remember this line: July’s prices never fell; the basis for the hike did. When policy stands on quicksand, a member’s only moat is discipline — set the strategic weights, harvest the certain fee savings, and let time do the work.