Details
The releases that matter most
| Release | Agency | Usual time (US Eastern) | Frequency |
|---|---|---|---|
| Consumer Price Index (CPI) | BLS | 8:30 | Monthly |
| Employment Situation (payrolls, unemployment) | BLS | 8:30, usually the first Friday | Monthly |
| FOMC statement | Federal Reserve | 14:00, press conference 14:30 | 8 scheduled meetings a year |
At 8:30 ET US stock futures, Treasuries and FX are open, so the first reaction happens there. Cash equities only open an hour later at 9:30, often with a gap.
What moves price is the surprise: the actual number minus what the market expected. A CPI print of 0.3% month on month is bullish for bonds if the consensus was 0.4% and bearish if it was 0.2%.
What the data looks like
Our free economic indicators sample comes from FRED. The CPI series (CPIAUCSL, all urban consumers, index level) for the latest months:
| symbol | series_name | date | value |
|---|---|---|---|
CPIAUCSL |
CPI, all urban consumers | 2026-08-01 | 334.131 |
CPIAUCSL |
CPI, all urban consumers | 2026-09-01 | 336.298 |
CPIAUCSL |
CPI, all urban consumers | 2026-10-01 | 337.476 |
Source: FRED, Federal Reserve Bank of St. Louis
Month-on-month CPI was +0.65% for September and +0.35% for October. Payrolls (PAYEMS, in thousands) rose by 60 in September and 165 in October, so about 60,000 and 165,000 jobs. The unemployment rate (UNRATE) was 4.1% for October.
Note the date column: it is the month the data refers to, not the day it was released. October CPI comes out in November. That difference is the single biggest source of errors in backtests that use economic data.
Pricing
Measure changes and line them up with prices
"""Example: month-on-month CPI and payroll changes from the FRED sample."""
import pandas as pd
econ = pd.read_csv("us-economic-indicators.csv", parse_dates=["date"])
wide = econ.pivot(index="date", columns="symbol", values="value")
monthly = wide[["CPIAUCSL", "PAYEMS", "UNRATE"]].dropna(how="all")
monthly = monthly.asfreq("MS") # exposes any missing month as a NaN row
monthly["cpi_mom_pct"] = monthly["CPIAUCSL"].pct_change(fill_method=None) * 100
monthly["payrolls_change_k"] = monthly["PAYEMS"].diff()
print(monthly.tail(4).round(2))
print("missing months:", list(monthly.index[monthly["CPIAUCSL"].isna()].date))
Run it and the second print lists a gap: there is no October 2025 value for CPI or unemployment in the series. Without asfreq, pct_change would quietly compare November 2025 with September 2025 and call it one month. With the gap exposed, the November 2025 change comes out as NaN, which is what you want.
To study market reactions, you then need the release timestamp for each value, and intraday prices around it. A typical measurement is the move from one minute before to 15 or 30 minutes after the release, compared with the size of the surprise.
Watch out
Common mistakes
Using the reference date as the release date. If your backtest knows October's CPI on 1 October, it is trading on information that didn't exist for weeks. Use the release date, which our X-ECON schema carries as release_date.
Using revised values. Payrolls are revised in the following months. The number in today's database is often not what was published on the day. For backtests you need point-in-time (first release) values.
Trading the headline only. Markets react to core CPI, wage growth and revisions as much as to the headline. A good headline with a bad core print can reverse in minutes.
Holding tight stops through the release. Spreads widen and prices gap in the first seconds. A stop can fill far from its level. Many traders flatten or widen stops before 8:30.
Details
Get the data
Free US economic data sample. 880 rows: Treasury yields, the 10-year minus 2-year spread, fed funds, CPI, unemployment, payrolls and GDP.
- Download US economic indicators
- Source: FRED, Federal Reserve Bank of St. Louis
- Release times, actuals and previous values: economic calendar API, planned for the Free plan in early access. Point-in-time history: economic data API, planned for Starter ($29/mo).
FAQ
Frequently asked questions
What time is the CPI release?
The BLS usually publishes CPI at 8:30 a.m. US Eastern time, around the middle of the month, for the previous month. Check the agency's schedule, because dates move around holidays and occasional disruptions.
How do you trade a CPI release?
Most traders either stay flat into the number and trade the reaction after spreads normalise, or trade the surprise using futures that are open at 8:30. Either way, know the consensus beforehand, because the market reacts to the gap between actual and expected.
Why did the market fall on a good jobs number?
Because "good" depends on expectations and on what it means for rates. A strong report can push yields up, which can hurt stocks. The reaction follows the surprise and its rate implications, not the headline alone.
What is the difference between release date and reference date?
The reference date is the period the data describes, such as October. Its release date is when it was published, such as mid-November. Backtests must use the release date.
Keep reading