A chapter-by-chapter field manual for 49 U.S. government economic indicators, covering where each one comes from, how it is built, and how much to trust it. It never tells you what to do with a release, only how to read one correctly before you react.
The one idea
Frumkin, an economist who spent years inside and around federal statistical agencies, organizes the book as 49 near-identical chapters, one per indicator, from Average Weekly Earnings to Value of the Dollar. Every chapter runs the same six sections: Where and When Available, Content, Methodology, Accuracy, Relevance, and Recent Trends. That rigid format is the entire value proposition. You can open to any indicator and know exactly where to find the release schedule, the government agency that compiles it, the exact definitions it uses, and its published margin of error.
The book's real argument is that headline numbers hide a lot of construction detail that changes what the number means. The unemployment rate is not one number but six (U-1 through U-6), each counting a different slice of labor slack. The CPI is not one index but three (CPI-U, CPI-W, C-CPI-U), built with different weighting math. The PMI is a diffusion index, meaning it tells you the direction most firms are moving, not the size of the move. None of this is optional trivia if you plan to trade off a release: the definition determines what a surprise actually means.
It is explicitly a reference, not a narrative. There is no thesis about markets, no discussion of how bonds, equities, or currencies react to a beat or a miss, and no attempt to rank indicators by trading importance. The book's job ends at explaining the data. What you do with it afterward is entirely on you.
What it actually teaches
The book is not sequential, but its six-part chapter template amounts to a repeatable checklist for reading any government data release correctly.
CHECK THE SCHEDULE
Every chapter opens with Where and When Available: which agency compiles the number, how many weeks after the reference period it is released, and when it gets revised. Example: total unemployment is released the first or second Friday of the following month, then revised every January for the previous five years based on updated seasonal factors. A number you react to today may not be the number that survives revision.
READ THE DEFINITION BEFORE THE NUMBER
Content sections spell out exactly who and what is counted, and it is narrower than most people assume. Unemployed, for the BLS, means you actively sought work in the last four weeks through a specific list of actions (interviews, applications, agency contact); scanning job ads does not count, and 'discouraged workers' who gave up looking are excluded from the labor force entirely, not counted as unemployed. Frumkin gives six separate versions of the unemployment rate: U-3 (the headline number) stood at 5.4 percent in February 2005, while the broadest measure, U-6, which adds part-timers who want full-time work, stood at 9.3 percent, a gap of 3.9 percentage points on the same underlying labor market.
WEIGH THE STATED ACCURACY
Methodology and Accuracy sections give the book's most concrete numbers. The CPI's monthly percentage change carries a sampling error of plus or minus 0.06 of a percentage point at one standard error, so a reported 0.30 percent increase has a true value somewhere between 0.24 and 0.36 percent two times out of three. The PMI, by contrast, has no published sampling or revision error at all: it is a composite of five component diffusion indexes (new orders 30 percent, production 25 percent, employment 20 percent, delivery times 15 percent, inventories 10 percent) built from roughly 350 surveyed manufacturers, and the book states its own working threshold for overall GDP: a PMI above 43, not 50, has historically corresponded to an expanding economy, because non-manufacturing output tends to keep growing even when factories contract.
CROSS-CHECK WHAT THE INDICATOR CAN AND CANNOT PREDICT
Relevance sections state the honest limits. On the Conference Board's Leading, Coincident, and Lagging (LCLg) indexes, the book is blunt: the leading index is a forecasting tool but 'does not provide actual forecasts.' It cites the Conference Board's own rule of thumb for a pending recession: a decline of 1 to 2 percent in the leading index over six months, together with declines in at least half of its ten components. It also lists the system's specific failures: the lagging index ran counter to theory ahead of the 1990-91 and 2001 recessions by leading rather than lagging, and the leading index gave false recession signals in 1966, 1984, and 1995 that never materialized.
What it looks like Monday morning
A reader does not walk away with a trade idea. What changes is how the next CPI, jobs, or PMI print gets read: before reacting to a headline, check which version of the number moved (U-3 versus U-6, CPI-U versus core), whether the move is inside the published error band, and whether the release is preliminary or already revised. That habit alone prevents overreacting to noise that the compiling agency itself flags as within normal sampling error, like a 0.1 point CPI wiggle that sits inside the book's own plus-or-minus 0.06 band.
The other practical change is knowing where each number actually comes from and when it updates, so a trading calendar can be built around real release and revision dates (BLS Friday jobs report, mid-month CPI, ISM's first-business-day PMI) instead of vague assumptions about what 'this week's data' means.
| Component | Weight |
|---|---|
| New orders | 0.30 |
| Industrial production | 0.25 |
| Employment | 0.20 |
| Delivery schedules | 0.15 |
| Inventories | 0.10 |
A reading above 50 signals more manufacturers reporting improvement than decline, but the book states the historical breakeven for overall GDP growth was closer to 43, since non-manufacturing output tends to keep expanding even when factories contract.
The LCLg system is a forecasting tool, it does not provide actual forecasts.Norman Frumkin, Guide to Economic Indicators
Where it fails
- It disclaims trading use directly, and means it. The book states plainly that even its most forecasting-oriented tool, the Leading, Coincident, and Lagging index system, does not forecast quantitative growth rates or the timing of turning points, and that it gave false recession signals in 1966, 1984, and 1995. There is no section anywhere on market reaction, consensus-versus-actual surprise, or trading windows around a release. That is not an oversight; the book was never built to answer those questions.
- The specific numbers are frozen to 2004-2005 and are already the wrong vintage to trade on. The PMI table runs January 2003 to December 2004. The unemployment underutilization table (U-1 through U-6) is dated February 2005. The CPI accuracy figure and the PMI 43-point GDP threshold are both this edition's snapshot, not live constants; a reader who quotes '43' or '0.06 percentage point' as a current rule without checking BLS or ISM's current published figures has taken a 2006 textbook number for a live one, which is exactly the mistake the site's edition-tracking rule exists to catch.
- Several of the underlying series have since been redefined or rebased. The Conference Board has rebased and re-weighted its Leading, Coincident, and Lagging indexes more than once since this edition (the book cites 1996 = 100), and ISM has periodically revised how it maps the PMI to GDP growth, so the '43' breakeven the book gives is not the number ISM currently uses. Anyone treating this book as a current reference for either series needs to verify the live methodology on the source website, not this printing.
- It assumes a reader who already knows introductory statistics and business-cycle concepts. Accuracy sections lean on sampling-error language (standard errors, confidence intervals) with no primer beyond a short appendix, and Relevance sections assume familiarity with NBER business-cycle dating and diffusion-index mechanics. A reader without that background will get the numbers but miss why they matter.
- It is a reference to be looked up, not a book to be read straight through. Forty-nine chapters in an identical six-part template make for repetitive reading in sequence; the format that makes it useful as a lookup tool (find your indicator, read six sections, done) works against it as continuous reading. Padding is the wrong word since every section earns its place, but the structure is built for dipping in, not for finishing.
Who it is for
Buy it if
You trade or analyze around U.S. macro releases and want to know exactly what a headline number counts, how reliable it is, and when it gets revised, before you react to it. It also suits anyone who wants a single place to look up how CPI, GDP, PMI, or unemployment are actually constructed.
Skip it if
You are looking for a trading system, a rule for reacting to beats and misses, or any discussion of market reaction to a release; none of that is here. Also skip it if you just need today's numbers, since this edition's figures are a 2004-2005 snapshot and current data belongs on the BLS, BEA, or ISM sites directly.
As an Amazon Associate I earn from qualifying purchases. It costs you nothing extra, and it does not change what gets recommended or what the verdict says.