GDP is intended to serve as a measure of all economic activity in an economy. But not every transaction is tracked, so one has to rely on estimates and models for parts of GDP. There are three ways one can measure GDP: adding up all expenditures, adding up all incomes, and adding up all value-added in the economy. All three should give you the same number. In practice, they don’t quite match up because of measurement issues. The difference between the expenditure and income approaches is called the statistical discrepancy. It is graphed above as a share of GDP, shown in red.
Another issue occurs when converting nominal values to real values: This is accomplished by applying a calculation to a basket of GDP components for a particular base year and keeping those prices constant for the other years. (The actual calculation is a bit more complex than this.) Adding up these GDP components does not exactly achieve GDP, and this difference is called the GDP residual. It is graphed above as a share of GDP, shown in blue. It is very small around the base year, but it deviates more substantially in earlier years, up to 6%.
How this graph was created: Search for GDP residual and add “Real Gross Domestic Product: Residual” (billions of chained 2009 dollars, quarterly, seasonally adjusted) to the graph. Then add real GDP (billions of chained 2009 dollars, quarterly, seasonally adjusted) to series 1 and apply transformation a/b. For the second line, search for GDP discrepancy and add “Gross Domestic Product (GDP); statistical discrepancy…” (quarterly, millions of dollars, seasonally adjusted) to the graph. Then add nominal GDP (quarterly, billions of dollars, not seasonally adjusted; not real GDP, as the discrepancy is nominal) to series 2 and apply transformation a/b/1000 . The division by 1000 here is because one series is in billions of dollars and the other is in millions of dollars.
Suggested by Christian Zimmermann
View on FRED, series used in this post: