The Balance Sheet

The Balance Sheet: How a Business Story Gets Made

This is a standing guide to the other half of economic literacy: not what the numbers mean, but how they reach you. Nearly every recurring complaint about business coverage — that it is alarmist, that it contradicts itself, that it never explains anything — has a mechanical explanation in how the story was produced. Knowing the mechanics makes the coverage far easier to read and considerably less frustrating.

Where the story comes from

Economic stories have a small number of origins, and the origin shapes the story more than anything else.

• Scheduled data releases. Statistical agencies publish on a fixed calendar, often at a fixed minute. Reporters know the date months ahead and frequently have the release under embargo, which is why a substantial article can appear within minutes. The calendar, not the news, decides that today's story is about employment.
• Company disclosure. Earnings, filings and announcements arrive on their own schedule and are written by the company's own communications staff before any journalist sees them.
• Research and reports. Institutes, banks, trade bodies and universities publish studies with press releases attached. The release is written to be quotable and is often the only part read under deadline.
• Official statements. Central bank decisions, budgets and regulatory announcements are events with fixed times and prepared texts.

Notice what this list implies: the supply of economic news is largely determined by other people's publication schedules. A quiet week in the data calendar produces analysis pieces; a heavy one produces a run of short reactive stories. Neither reflects the underlying economy changing pace.

What deadline pressure does to a number

A statistical release can run to hundreds of pages of tables with a technical note explaining the methodology and the confidence intervals. A reporter may have forty minutes. In that time the sequence is almost forced: take the headline number, compare it with the consensus forecast, find the largest sub-component that moved, get one quotation from an economist and one from someone affected, and file.

Three predictable distortions follow, and none requires anyone to intend them.

The forecast becomes the baseline

Because the fastest available comparison is the analysts' consensus, a figure gets described as a "miss" or a "beat" against a forecast rather than in terms of its own history. A perfectly ordinary reading that landed below expectations is reported as a disappointment. The forecast, not the economy, is the reference point.

Sampling error disappears

Survey-based statistics have published confidence intervals, and for month-to-month changes those intervals are often wider than the change itself. A move that is statistically indistinguishable from zero can be reported as a decisive turn. The interval is in the release; it almost never survives into the article, because it makes the sentence longer and less definite.

Revisions never get the same billing

First estimates are provisional and get revised as more source data arrives. The initial figure lands on a news day and gets full coverage; the revision arrives quietly weeks later. Readers therefore hold a picture assembled from first estimates, which is systematically noisier than the final data — and occasionally points the other way.

Framing, and why it is mostly structural

Framing is the choice of what a story is about, and most of it is settled before any judgement is made. A change has to be described relative to something, and the choice of comparison point — last month, last year, the pre-downturn peak, the all-time high — can make the same series look like a recovery or a collapse. Levels and rates of change tell different stories: a rate can be falling while a level is at a record.

Then there is the asymmetry of the concrete. A factory closing is a place, a date and people who can be interviewed. The diffuse benefit spread across millions of consumers has no address and no spokesperson. Any journalism that prefers the specific to the abstract — which is to say, all of it — will over-represent concentrated effects relative to dispersed ones. That is a property of storytelling rather than of any newsroom's politics, and it applies to every subject on this site.

Sources and who is available

Under deadline, a reporter needs someone who will answer the phone, speak in complete sentences, and not require three days' notice. Organisations that invest in press operations are therefore quoted disproportionately relative to their share of expert opinion — a selection effect on availability, not on merit. When a story quotes two economists, they are usually the two who were reachable, not a sample of the profession. This is the single most useful thing to remember when a piece appears to establish a consensus.

A short checklist

Six questions handle most economic stories.

1. What is the source document, and is it linked? If a study is described but not linked, that is worth noticing.
2. Is this a level or a change? And if a change, compared with when?
3. Nominal or real? Any multi-year money comparison that has not been inflation-adjusted is telling you about prices, not quantities.
4. Is the change larger than the margin of error? For survey data on a one-month horizon, frequently not.
5. Is this an estimate, a projection, or an outcome? Three different kinds of claim, routinely written in the same tense.
6. Who benefits and who pays, and are both in the story? Concentrated costs are easy to report; dispersed benefits are not.

What this guide is not

This is not a critique of any news organisation and does not assess anyone's record. The pressures described here are structural, apply everywhere, and are largely acknowledged by working journalists. The point is not that coverage is untrustworthy but that it is compressed, and that a reader who knows what compression removes can put much of it back.

For worked examples, see the special report on recurring misreadings of economic statistics and the one on how the jobs numbers are built. For the underlying series, see data sources. The Bureau of Labor Statistics publishes a Handbook of Methods that is the single best free reference on how the statistics discussed here are actually collected.