Friday’s US employment report did not beat expectations. It tripled them. Nonfarm payrolls printed +162,000 against a consensus near 56,000, unemployment held at 4.1%, and June and July were revised up by a combined 55,000 — June from 20K to 31K, July from a 23K contraction to a 21K gain. Average hourly earnings rose 0.3% on the month, 3.1% on the year.
The reaction was immediate and it was not about jobs. It was about whether the Federal Reserve raises rates on 16 September.
The number that moved was the odds, not the dollar
Before the release, market-implied probability of a September hike had drifted down toward 50%. Within hours it was back above 60%. J.P. Morgan Wealth Management now expects a 25bp increase, citing roughly a 65% implied chance. The current target range is 3.50–3.75%, held unchanged in July.
The reasoning is not really about the labour market. Two things are doing the work:
- Energy. Supply disruption from the Iran conflict has run longer than assumed. WTI near $90 keeps the inflation path unresolved in a way a soft labour market would otherwise have settled.
- Credibility. The July hold came with limited forward guidance. When a central bank leaves the market unsure whether it will respond to an inflation impulse, the incentive to make a visible, credibility-reinforcing move rises. That is an argument for hiking that has nothing to do with the data.
The disagreement is the tradeable fact
Look at what those numbers actually describe. Implied odds near 50% two weeks ago, above 60% now, roughly 65% by J.P. Morgan’s reading. That is not a market with a firm view. At 65%, better than one participant in three is positioned for no hike — nine days from the decision, with the deciding data still unreleased.
For a system trader that framing is more useful than the point estimate. A tight consensus produces a small reaction to an in-line print and a large one to a miss. A split consensus produces a large reaction either way, because a substantial share of positioning is wrong whatever the number does. The distribution of Friday’s move is fatter on both tails than a routine CPI, and it is fatter because of the disagreement, not because of the forecast.
It is also worth noting what moved the odds. The labour market did not deteriorate — it did the opposite, and a strong labour market is not by itself a hiking argument at 4.1% unemployment. The odds moved because a strong print removed the excuse not to respond to an energy-driven inflation impulse. That is a fragile basis for pricing, and fragile pricing revises fast.
Friday 11 September, in order. All times UTC.
- 06:00 — UK GDP m/m. A negative print takes pressure off Bank of England hike expectations; sterling risk, and GBP/JPY is already carrying yen beta.
- 12:30 — US August CPI. Consensus looks for headline around +0.3% to +0.4% m/m, against July’s +0.1%. Energy and food are pushing up; shelter is the offset worth watching in the core.
- Later the same day — a large USD/JPY option expiry at the 154.00 strike rolls off at the New York cut.
- 14:00 — Prelim University of Michigan consumer sentiment.
That third line is the one most calendars will not show you. A large expiry at 154.00 pins price toward the strike as the cut approaches — right up until the print arrives and the pin is irrelevant. You get the pinning behaviour and the event risk on the same day, at the same handle. A mean-reversion system that has spent the morning being rewarded for fading moves back to 154 will be holding exactly the wrong position at 12:30.
Earlier in the week, with the same UTC convention:
- Tue 8 Sep, 10:00 — US NFIB small business index (low impact)
- Wed 9 Sep, 01:30 — China CPI and PPI
- Wed 9 Sep, 12:15 — US ADP weekly employment change (high impact)
- Thu 10 Sep, 12:15 — ECB decision; 12:30 — US PPI and jobless claims
Note Thursday: the ECB decision and the US PPI print are fifteen minutes apart. That is not two events. It is one window with two ways to be wrong in it.
Building the window, not reacting to it
The mistake in most news filters is that they are built around the release time. The release is not where the damage happens.
- The first tick is not the move. On a split-consensus print, the initial spike is frequently reversed inside two minutes as the second interpretation lands. A filter that resumes trading at T+60s is trading the reversal, not avoiding the event.
- Widen before, not after. A stop widened at T+0 is a stop that was too tight at T−1. If your risk manager only reacts once volatility is measured, it is always one event late.
- The cluster is the unit. Thursday’s 12:15 and 12:30 belong in one blackout, not two. So do Friday’s 12:30 print and the option cut.
- Spread, not just volatility. Execution cost around a CPI print is where backtested edges quietly disappear. If your backtest filled at the level, live will not.
This is the case for a machine-readable event feed rather than a human reading a calendar — which is why our FX Events page is being built toward a unified schema an EA can consume directly. The EA needs when and how big. It does not need the number.
What this does not tell you
It does not tell you what CPI will print, and it does not tell you what the Fed will do on 16 September. A 65% implied probability means the market itself is not sure, and the honest position is that we are not either.
It also does not tell you the reaction will be large. Split positioning raises the variance of the outcome; it does not guarantee a big move. Markets have absorbed genuinely surprising prints with a shrug before, and will again.
Finally, none of this is a directional view. Everything above is about when to be exposed and how much, not about which way to face.
Related
- FX Events calendar — every release above, with confirmed times
- Signals — direction and strength on the majors, from a signal-only analyzer
- EA Track Record — verified performance, including through news windows
