← FINBORS ACADEMY

Working with data / WORKSHOP

Actual above forecast: why price direction is still unknown

Release time, retrieval time and reference period answer different questions. Read a release without false precision.

Educational illustration: risk and testing a market idea
Hypothetical example diagramHypothetical data: thousands of jobsNew month: 180 − 150 = +30Revisions: −40 − 30 = −70(80 + 70 + 180) − (120 + 100 + 150) = −40−40 ≠ job losses in the new monthPrice response is not determined
Hypothetical numbers from this lesson. A diagram, not market quotes or a trading signal.

How it works

An economic calendar connects the scheduled release time to a reference period and indicators. A previous value refers to an earlier period or an earlier estimate of the same period. Consensus represents the expectations of a stated analyst sample, not an official result or a probability of a market move.

BLS revises initial CES employment estimates in the next two months as additional survey responses arrive. Annual revisions may follow. The latest database series is not the information available at an earlier release: keep the archived release for historical analysis. “No data” does not mean zero.

What to check

  1. Check the original source, timezone and reference period. An October release may describe September.
  2. Check units: percent, percentage points, thousands of jobs, monthly change or annual change.
  3. Record the consensus source and date, actual result and previous-value revisions separately. Without consensus, surprise cannot be calculated.

A hypothetical example

All figures are invented, in thousands of jobs, with the same seasonal adjustment. Before release: growth in month A +120, B +100, and hypothetical consensus for C +150. After release: A revised to +80, B to +70, first actual for C +180. C surprise: 180 − 150 = +30. Revisions: (80 − 120) + (70 − 100) = −70. The earlier figures plus expected C totalled 370; the new sum is 330, a difference of −40. This compares two information sets; it does not mean 40 thousand jobs were lost in C, whose growth is +180. The calculation assumes the level before A is unchanged; if it is revised, reconcile the levels separately.

When to skip

Skip a trading decision when the period, units or expectation source is unclear, or when wider spreads violate your predefined risk limit.

Counterexample: “180 beats 150, so the whole report is stronger than expected” ignores −70 of revisions. Yet the calculated −40 does not predict a price fall either: revisions may have been anticipated, while wages and unemployment describe other aspects of the report. Never add thousands of jobs to percentages.

Try it yourself

Choose the next calendar release and verify the agency schedule. Record period, units, seasonal adjustment, time with timezone, and consensus source. For an employment report, preserve the previous estimates of two months, then record their revisions and the first actual for the new month separately. For another indicator, check its revision rules first; do not apply the CES rule to CPI. Never replace missing consensus with the previous actual.

In the exercise, identify the new-month surprise, combined revisions and difference between the two three-month sums. What can be calculated if consensus +150 is unavailable?

Show explanation

Respectively +30, −70 and −40 thousand. Without consensus, C actual +180, revisions −70 and the new three-month total +330 remain available. Surprise and comparison with the earlier total that included expected C are undefined. These calculations do not determine a trade direction.

Separate the new actual, historical revisions and expectations. Evaluate past decisions using information available then, not today’s version of the series.

Sources and definitions

BLS · release schedule ↗BLS · Public Data API ↗BLS · CES revisions and vintage estimates ↗

Scenarios and exercises are educational FINBORS examples, not a verified trading strategy.