Output and Growth: What the Number Measures

The identity, and what each term tells you

Output is measured as the sum of what is spent on it: consumption, private investment, government spending and net exports. Consumption is the largest share in a developed economy — roughly two-thirds in the United States — and it is the most stable, because households smooth spending against income. Investment is a much smaller share and far more volatile: it responds to interest rates, to expected demand and to cash flow, which is why a recession is usually visible in investment before it is visible in consumption. Government spending is a policy variable, so its contribution has to be read against what was intended rather than as a signal about private behaviour. Net exports capture the difference between what the country produces and what it spends, and they respond to relative growth and to the currency. Two derived figures do most of the analytical work. Final sales to private domestic purchasers strips out inventories, government spending and trade, leaving what households and businesses actually bought — the cleanest read on domestic private demand, and the one to compare with the headline when the two diverge. The other is the contribution of inventories, which is the difference between production and sales: a positive contribution in a slowing quarter means goods were made that nobody bought, and it is a subtraction from future growth, while a negative contribution in a weak quarter means demand was met from shelves and future production has to rise to replace them. Three practical warnings belong with all of it. The first estimate of a quarter is published weeks after it ends and is revised twice more, with the revisions often larger than the difference between the release and the consensus. Annualised quarterly growth, the convention in the United States, multiplies a single quarter by four, which turns a small quarterly figure into a large annual rate and makes the number noisier than it looks. And the deflator is itself an estimate, so the real number carries the uncertainty of two measurements rather than one — which is why the market reaction to a growth release is usually a reaction to the inflation detail inside it. One quarter, decomposed — Consumption, 1.2%, weight ≈ 68%: contributes about 0.8pp · Government spending, 4.1%, weight ≈ 18%: contributes about 0.7pp · Inventories, +1.4pp: produced but not sold — a subtraction from the next quarter ← · Final sales to private domestic purchasers, 0.4%: the number that describes private demand Nominal and real growth are not interchangeable. Revenue, earnings and debt are all nominal, so a company can report record profits in a recession if prices are rising fast enough — and real wages can fall while nominal wages rise. Always ask which one is being quoted.

Why growth drives asset prices

Equity earnings are a share of output, so growth sets the revenue line the whole valuation rests on. But the link is not one-for-one, and the timing is the trap: the market discounts expected growth, so a strong release that is weaker than expected moves prices down, and a weak release that is better than feared moves them up. What the market trades is the change in expectations, which is why the first reaction to a release says more about the consensus than about the economy. Growth also sets policy, and policy sets the discount rate, so a growth release reaches valuations through two channels that can oppose each other. Stronger growth raises expected earnings and also raises the expected policy rate, and the second effect dominates in the short run whenever policy is the binding constraint — which is what happens in the late stage of a cycle. That is the mechanism behind the counterintuitive reaction where good news is bad news for stocks, and it is not a quirk: it is the discount rate doing its work. For a portfolio, the practical implication is that growth data matters most where it changes the path of policy, and least where policy is already constrained. And it matters differently across assets: long-duration equities and long bonds are both hurt by a rise in the discount rate, credit spreads widen when growth is expected to fall, and the dollar generally strengthens with relative growth. Reading one release for a whole portfolio is a mistake; reading it for which of those four channels it changes is the job. • Earnings are a share of nominal output, so nominal growth sets the revenue line. • The market trades the change in expectations, not the level of the data. • Growth reaches valuations through earnings and through the discount rate, and they can oppose. • The same release means different things for equities, credit, the curve and the currency. A useful habit: before reading a release, write down what would have to be in it to change your view — on policy, on earnings, on the curve. The number then has something to be measured against beyond the consensus estimate of a statistician you will never meet.

The number you trade is not the number that exists

The growth figure markets react to is a **first estimate** produced from incomplete source data, and it is revised, often more than once. The statistical agency publishes an advance estimate weeks after the quarter ends, then a second and third as more complete data arrive, then updates the whole series in an annual revision — and every five years or so reworks the entire history in a comprehensive revision that can change the level of output by a noticeable amount. The revisions are not noise around a fixed truth that anyone can observe; the truth is only ever known through the revisions, and the early prints are the ones that move markets. That has two consequences worth carrying. The first is that a market reaction to a first print is a reaction to a partly-measured number, which is why a release that misses expectations can be followed by a revision that removes the miss and nobody re-prices. A trade built on the surprise in an advance estimate is a trade on the estimate, not on the economy. The second is that revision *direction* is informative in itself: if a run of quarters is being revised down, the underlying momentum is weaker than the releases suggested, and the market’s memory of the stronger initial prints is stale. There is also a structurally different way to measure the same economy, and its divergence is a useful warning light. Gross domestic income measures the same activity from the income side and is, in principle, identical to gross domestic product. In practice the two are estimated from different data and differ, and the difference is published as the **statistical discrepancy**. Big persistent gaps are not a misprint to be ignored — measurement is hardest at turning points, exactly when accurate readings matter most, so a widening discrepancy is a reason to lower your confidence in any smooth story about the quarter. The same quarter, published four times — Advance estimate: Incomplete source data; this is what the market trades · Second and third estimates: Better coverage, and revisions that are often material · Annual revision: The whole series is restated against fuller tax and survey data · Comprehensive revision: Levels change; long-run growth can look different afterwards ← Gross domestic income versus gross domestic product is the paired reading to watch. When the two measures disagree, both are telling you that the quarter is harder to measure than usual — and turning points are when that happens.

Where growth comes from over ten years

The identity in this lesson breaks a quarter’s growth into its spending components, which is what tells you where demand came from. A different decomposition answers a different question — where growth comes from over a decade — and it is the one that sets the ceiling on everything else: what the economy can produce, given how many people are working and how much each of them produces per hour. That decomposition has only two terms. **Labour input** is the number of workers times average hours, and **productivity** is output per hour. Multiplied, they give potential output. The arithmetic is trivial and the implications are not, because the two terms move on completely different timescales. Labour input responds to demographics and participation, both of which change over years and are largely predictable — a population pyramid is one of the few genuinely forecastable things in economics. Productivity responds to capital investment, technology, and the organisation of work, and it is far less predictable, which is why the range of long-run forecasts is dominated by the productivity assumption rather than by the demographic one. This is the source of the distinction that makes the framework useful: **cyclical** growth is output moving within the neighbourhood of potential, and **structural** growth is potential itself moving. Policy can affect the first powerfully and the second only slowly, so an economy whose potential growth has fallen cannot be returned to its old trend by demand management — the shortfall is not a gap that closes when demand recovers, it is a ceiling that moved. Reading a period of disappointing growth as cyclical when it was structural is how a central bank ends up easing into an inflation it did not expect. For an investor, the practical content is the decomposition of a long-run return. A market’s long-run earnings growth is tied, loosely and with plenty of drift, to nominal growth, which is real potential growth plus inflation. That is why demographic decline and weak productivity growth are treated as valuation-relevant facts rather than as background: they set the growth rate that a terminal value in a model is constrained by. A terminal growth assumption of three percent in a currency area whose potential growth is under one percent is not a modelling choice, it is an arithmetic error with a price attached. The honest caveat is that the estimate of potential output is itself unobservable and revised, often substantially. It is an estimate of a trend through noisy data, so it moves when the data moves, and revisions to it are not news about the economy but news about the estimate. Treat it as a framework for organising a forecast rather than as a number to be read off a release. • Potential growth is labour input times productivity, and the two move on different timescales. • Demographics are forecastable; productivity is where the uncertainty lives. • Cyclical shortfalls close, structural ones do not — easing cannot fix a lower ceiling. • Terminal growth assumptions are bounded by potential growth plus inflation, not chosen freely. Useful discipline for any long-run model: state the potential growth rate you are assuming and check the terminal growth against it. Most implausible forecasts are not implausible because of the near years; they are implausible because the terminal value quietly assumes a productivity boom.

Sources

• How GDP is defined, measured and revisedBureau of Economic Analysis, “Concepts and Methods of the U.S. National Income and Product Accounts”
• Real versus nominal measurement and the deflatorStandard national-accounts methodology; Federal Reserve education materials
• Revisions and the information content of the first estimateBEA revision studies; standard macroeconomic data-quality literature