Ask Dr. Econ

May 2005

Why is there such a time lapse getting the latest report on Gross Domestic Product GDP? In May 2005, we are just getting the preliminary GDP report for the quarter that ended in March 2005.

First, what is Gross Domestic Product?

The definition of Gross Domestic Product, or GDP, is best provided by the U.S. Department of Commerce’s Bureau of Economic Analysis (BEA), the agency that calculates GDP. The BEA defines GDP as follows:

The market value of goods and services produced by labor and property in the United States, regardless of nationality; GDP replaced gross national product (GNP) as the primary measure of U.S. production in 1991.

Economists examine trends in quarterly GDP figures and GDP growth rates to evaluate how well the economy performed during a given time period. GDP represents the broadest measure of output for the overall economy and its behavior is carefully analyzed by economists and widely reported in the press, especially changes in real (measured in inflation-adjusted dollars) GDP. For the first quarter of 2005, nominal (measured in current dollars) GDP was $12,198.8 billion at an annual rate, and real GDP was (measured in chained 2000 dollars) $10,999.3 billion at an annual rate.

Calculating GDP is complicated

Calculating GDP for a nation that has a population over 293 million (2004) and houses millions of businesses is complicated. It takes time to collect and estimate the roughly 1,500 economic data points that are used to generate GDP each quarter. The BEA describes the process:

The source data include a variety of economic measures, such as sales or receipts, wages and salaries, unit sales, housing stock, insurance premiums, expenses, interest rates, mortgage debt, and tax collections. For most components, the source data are “value data”; that is, they encompass both the quality and price data required for current-dollar estimates.

This means that the BEA must not only estimate physical quantities produced and the volume of services provided, but prices as well, to generate estimates of the value of goods and services produced during a time period. The process is further complicated by adjustments that must be made for quality and price changes, as well as for goods and services that are involved in international trade.

Time lag

Given the complexity of calculating GDP, it is impressive that it only takes about one month after the end of a quarter before the advance GDP estimate for that quarter is released to the public. Over the next two months, two additional GDP revisions for the same quarter are released: the preliminary estimate and the final estimate. Thus, the final estimate is available about three months after the end of a quarter. Table I shows the sequence of releases of GDP estimates for the first quarter of 2005.

Table I – Gross Domestic Product Releases for the First Quarter 2005

Why three estimates?

In calculating GDP, there is an important tradeoff between time and accuracy. The advance estimate gives economists their first indication of the strength and direction of GDP growth and is a useful tool for economists, analysts, and decisionmakers tracking changes in the performance of the economy. However, since the earliest estimates are based on incomplete information and partial data, they are not as accurate as later data—to get the advance GDP estimate, the BEA must extrapolate, or infer what trends are implied from available data, even though hundreds of data points aren’t yet available. 1 In the two months after the advance GDP figure is released, more complete data about the economy become available and are incorporated into the BEA’s GDP estimates. This additional information, including revisions in series used to estimate GDP, result in more accurate final GDP estimates than the earlier advance or preliminary estimates. As displayed in Table I, the estimates of first quarter 2005 Real GDP growth rates were revised with each new estimate as new information became available to the BEA.

How accurate are GDP estimates?

The BEA website provides a succinct answer to this important question:

Early vintages of the GDP estimates are based on partial and incomplete source data. Subsequent GDP estimates incorporate increasingly comprehensive and improved source data. Periodically, BEA conducts studies of reliability. In these studies, reliability is defined as whether the successive vintages of GDP estimates present a consistent, general picture of the economy. The most recent reliability study found (as have previous studies) that the early estimates of GDP present a useful picture of economic activity. They consistently indicate whether growth is positive or negative, whether growth is accelerating or decelerating, whether growth is high or low relative to the trend, and where the economy is in relation to the business cycle. In addition, the latest study found that the average revisions to GDP are small and positive, indicated a tendency toward upward revisions[.]

Providing three data releases for GDP gives users the best available data in a timely manner based on a flow basis, starting about one month after the end of a quarter. Approximately two months following the release of the advance estimate, the final GDP estimate is published. It has been revised to include additional information that makes the final data a more accurate indicator of GDP than the earlier advance or preliminary estimates.

Lots of sources and reasons for lags

Gross Domestic Product estimates of the output of the entire economy require the BEA to use information on many different sectors of the economy and from many different sources to estimate quarterly real (inflation adjusted) and nominal (current dollars) GDP, the GDP deflator (price level data) and national income statistics. Table II lists some of the agencies that provide data and information for key GDP components.

Table II – Sources of Data Used to Generate GDP

Some data that are used as inputs to calculate GDP are frequent and current—for example, interest rates are available daily and without a time lag. Other data may be available monthly with only a short lag of several weeks, such as payroll employment. Still other data might only be available less frequently, like manufacturing surveys, which often are only reported after a long lag.

Issuing three GDP estimates for each quarter means the BEA must rely on different sources for the advance, preliminary, and final GDP data. Over time, even the “final” estimate will be revised due to periodic revisions to base series; the benchmarks that help maintain the accuracy of the data and update seasonal factors. Revised and benchmarked historical data are used to generate updated future and revised historical seasonal factors. 2

Go to the BEA for GDP data and for state-level Gross State Product (GSP) data (links as of August 10, 2005)

Bureau of Economic Analysis. Homepage provides data on GDP for the nation as well as Gross State Product (GSP), an output estimate for each of the states. Note that the state data are only available after a much longer lag.

Bureau of Economic Analysis. National Economic Accounts, GDP, Current Estimates and Percent Change. This link provides access to the latest news release for GDP.

Technical Note. Gross Domestic Product. Second Quarter of 2005 (Advance). This BEA release describes some of the differences in data availability between the advance estimate and the later estimates of GDP.

“Updated Summary of NIPA Methodologies.” Bureau of Economic Analysis. November 2004.

Endnotes

1. J. Steven Landefeld, director of the BEA, estimated that one-third of the roughly 1,500 data points used to calculate GDP are estimates at the time of the advance GDP release. For additional information, see “ECONOMIC VIEW; A Number That's Meant to Be Second-Guessed,” by Anna Bernasek, New York Times, July 31, 2005.

2. For example, seasonal adjustments are used to correct data for seasonal effects, such as crop or weather cycles, or seasonal shopping patterns, that might cause regular seasonal movements in the data but that are essentially unrelated to the data itself. This process helps isolate the data’s underlying trend. Because a long-term data trend may change shape over time, a seasonal adjustment may eventually be slightly revised, causing the original data to be revised as well.