US GDP and the quarterly national accounts

What gross domestic product measures 

Gross domestic product is the market value of the final goods and services produced within a country's borders in a given period. Two words in that sentence do most of the work. Final excludes intermediate inputs, so that a component sold to a manufacturer and the finished product it goes into are not both counted. Within a country's borders makes it a measure of production located in the country, regardless of who owns the producer, which is what distinguishes it from gross national product.

The same total can be reached three ways, and the United States national accounts publish two of them. The expenditure approach sums what is spent on final output: consumption, investment, government purchases, and exports less imports. The income approach sums what is earned in producing it: compensation, profits, rents, interest and taxes on production. Conceptually the two are identical. In practice they are estimated from different source data and differ by a residual the Bureau of Economic Analysis publishes explicitly as the statistical discrepancy, which is an unusual and commendable piece of honesty in official statistics.

Imports enter the expenditure identity with a negative sign, and this is the single most misunderstood feature of the accounts. It is not because imports reduce output. It is because consumption, investment and government purchases have already counted spending on imported goods, and subtracting imports removes what was produced abroad from a measure of domestic production. A rise in imports mechanically lowers the headline while telling nobody anything about domestic weakness.

Real, nominal and the deflator 

Nominal GDP is measured in the prices of the period. Real GDP removes price change so that the remainder is a change in volume, and it is real GDP that the headline reports. The price index used to do that removal is the GDP deflator, which is derived from the accounts themselves rather than imposed on them: it is the ratio of nominal to real output, and its basket is therefore everything the economy produced, reweighted every period, rather than a fixed consumer basket.

That makes the deflator a broader inflation measure than a consumer price index, and a different one. It includes investment goods, government purchases and exports, and it excludes imported consumer goods, which a consumer price index includes. The two therefore report different inflation rates for the same quarter as a matter of construction, not of error.

Why the headline is annualised 

The United States quotes quarterly GDP growth as a seasonally adjusted annual rate: the quarterly change compounded as if the same rate of change persisted for four quarters. Most other advanced economies quote the plain quarter on quarter change. The two conventions describe the same underlying movement and produce numbers that differ by roughly a factor of four, which is a recurring source of confusion in cross-country comparison.

The three estimates and the annual revision 

Each quarter is published three times. The advance estimate arrives roughly a month after the quarter ends and rests on incomplete source data, with the Bureau explicitly assuming values for the months not yet reported. The second estimate incorporates fuller trade, inventory and services data. The third estimate incorporates more still. An annual update then revises several years at once against benchmark sources, and a comprehensive revision every few years can change definitions, base years and the treatment of whole categories.

The practical consequence is that the advance estimate, which is the one that appears on a calendar as the GDP release, is the least reliable of the three, and that the version of history available today is not the version that was published at the time. Analysis of how the economy behaved around past turning points, conducted on current vintages, uses information nobody had.

The components practitioners separate out 

• Personal consumption expenditures, the largest component of the expenditure total by a wide margin, split into goods and services. The accompanying price index for this component is the one on which the Federal Reserve's inflation objective is formally defined.
• Private fixed investment, split into non-residential structures, equipment and intellectual property, and residential. Interest-sensitive and volatile, and conventionally read as the component most responsive to financing conditions.
• The change in private inventories. This enters as a change in a change, so inventories add to growth when they accumulate faster than in the previous quarter and subtract when accumulation merely slows. It is the component most likely to make a headline unrepresentative of underlying demand.
• Net exports, exports less imports, which carries the sign convention described above and can swing sharply when firms bring forward purchases ahead of an anticipated change in trade policy.
• Government consumption and gross investment, federal, state and local, which counts government purchases of goods and services but not transfer payments, since a transfer is not a purchase of output.

Because inventories and net exports are both volatile and both weakly related to domestic demand, analysts commonly quote final sales to private domestic purchasers, which strips both of them out along with government. It is a published aggregate, and the convention treats it as a cleaner reading of underlying demand than the headline. Like every such convention it is a choice about what to exclude, and the excluded components are part of output whether or not they are convenient.

Cadence and what it is not 

Publication is quarterly, in the morning Eastern Time, with the three estimates arriving in successive months so that a GDP release appears on the calendar every month even though a new quarter appears only every third one. The Bureau publishes the schedule in advance, along with the full component detail and the statistical discrepancy.

GDP is also a measure with well-known boundaries. It counts production that passes through a market, so unpaid household work and volunteering are outside it. It counts expenditure without regard to whether the expenditure repairs damage or creates value. It says nothing about distribution, and a rise in the total is consistent with any distribution of that rise. These are not defects in the estimate; they are properties of the definition, and they are the reason no statistical agency describes GDP as a measure of welfare.