A Series Authored by Walter J. "John" Williams

"Gross Domestic Product"
(Part Five in a Series of Five)

October 6, 2004

Overstated GDP growth has meant that the 1990 and 2001
recessions were much more severe than recognized, and that lesser
downturns in 1986 and 1995 were more or less missed entirely.


The Gross Domestic Product (GDP) is one of the broader measures of economic activity and is the most widely followed business indicator reported by the U.S. government. Upward growth biases built into GDP modeling since the early 1980s, however, have rendered this important series nearly worthless as an indicator of economic activity. The analysis in this Installment will indicate that the recessions of 1990/1991 and 2001 were much longer and deeper than currently reported, and that lesser downturns in 1986 and 1995 were missed completely in the formal GDP reporting process. Furthermore, the current economic circumstance is suggestive of an early-1980s-style double-dip recession.

The distortions from bad GDP reporting have major impact within the financial system. For example, Alan Greenspan's heavy reliance on productivity gains to justify some of his policies is equally flawed, since the methods applied to GDP estimation influence the numerator in the productivity ratio. As with the CPI distortions discussed in Installment III, the Federal Reserve Chairman knows better.

With reported growth moving up and away from economic reality, the primary significance of GDP reporting now is as a political propaganda tool and as a cheerleading prop for Pollyannaish analysts on Wall Street.

Reporting Basics

The GDP is compiled and reported by the Bureau of Economic Analysis (BEA) of the Department of Commerce. Quarterly estimates are updated monthly, with the "advance" estimate published at the end of the first month following the close of a quarter. The first and second revisions are called the "preliminary" and "final" estimates. In turn the "final" estimate is revised in annual revisions (usually in July), and every five years or so a benchmark revision is published that revises all data back to 1929, the first year of formally estimated economic activity.[1]

The popularly followed number in each release is the seasonally adjusted, annualized quarterly growth rate of real (inflation-adjusted) GDP, where the current-dollar number is deflated by the BEA's estimates of appropriate price changes. It is important to keep in mind that the lower the inflation rate used in the deflation process, the higher will be the resulting inflation-adjusted GDP growth.

Due to a lack of good-quality hard data, the "advance" GDP report is little more than a guesstimate. The BEA comes up with three estimates of growth, a high, low, and most likely. The numbers then get re-massaged so that the reported growth rate is moved closer to whatever the economic consensus is expecting. There actually is a belief at the BEA that there is some value to economic consensus estimates.[2]

The estimation process does not improve much with the "preliminary" and "final" estimates. The BEA reports that 90% of ultimate revisions to the "final" estimate fall within a range of +3.1% to -2.6%. Where average growth has been about 3.5% over the years, that means that most reporting is not statistically significant. The upward bias shown in the revisions is due to what I call "Pollyanna Creep," where methodological changes regularly upgrade near-term economic growth patterns. These patterns will be explored shortly.

The GDP is a large component of the National Income and Product Accounts (NIPA), representing "the output of goods and services produced by labor and property in the United States."[3] The NIPA was the concept and development of the National Bureau of Economic Research, a private organization founded in 1920. The NBER work evolved into the BEA and the current NIPA accounting.

The NBER remains a consultant to the process and retains the position as official arbiter of U.S. recessions. At one time, the NBER did define a recession as two consecutive quarters of negative GNP/GDP growth that were not distorted by an event such as a truckers' strike.[4] The NBER used trends in indicators such as industrial production and payroll employment to time a recession's beginning and end, to the month. More recently, though, the NBER has abandoned the GDP as a recession indicator and has relied instead on those other economic series. My presumption is this change resulted from an unofficial recognition of the declining value of the GDP reports. In theory, the NBER is apolitical, although the timing of some of its recent calls on the ends of recession are suspect. Specifically, there is no such thing as a jobless recovery. If jobs are being lost, the economy still is in recession.

There Is a Problem in the Basic Structure

As part of the NIPA, the construct of the GDP is heavily reliant on economic theory for composition, unlike other data series such as retail sales or the trade deficit, which are relatively simple surveys that end up contributing to the GDP estimations.

The related Gross National Product (GNP) is the broadest U.S. economic measure and includes the GDP plus the balance of international flows of interest and dividend payments. For net debtor nations such as Guinea-Bissau and the United States, GDP usually will show the stronger growth than GNP, since the outflow of interest payments does not get charged against economic activity. For this reason, the United States switched its primary reporting from the GNP to the GDP in 1991. Put in perspective as of the "final" estimate of second-quarter 2004, annualized real GDP growth was 3.3%, down from 4.5% in the first quarter, while GNP growth for the same period was 1.9%, down from 3.9%.

I respect the intellect and creativity of those who have anchored their careers in academia. Frankly, though, most economic theories have little practical use in the real world. Concepts such as free trade being a boon to the world's economy [5], a weak currency helping turn a nation's trade deficit[6], or personal income including what the average homeowner would receive from himself in rental income if he charged himself to live in his own house, fall in to the "not in the real world" category.[7]

Varied academic theories, often with strong political biases, have been used to alter the GDP model over the years, resulting in Pollyanna Creep, where changes made to the series invariably have had the effect of upping near-term economic growth. Whether the change was to deflate GDP using "chain-weighted" instead of "fixed-weighted" inflation measures, to capitalize rather than expense computer software purchases, or to smooth away the economic impact of the September 11th terrorist attacks, upside growth biases have been built into reported GDP with increasing regularity since the mid-1980s.

The accompanying table shows the net impact of these changes over time. The GNP level for various years from 1929 through 1980 and GDP for 1980 and 1990 are shown in billions of current dollars. Once set, these GNP/GDP levels should not change. With redefinitions and methodological shifts, however, earlier periods have been restated so as to be on a consistent basis with the latest reporting. Accordingly, the GNP/GDP levels are shown as they were reported variously in 1950, 1984 and at present.[8]

What becomes evident when looking at these data is that the biggest reporting changes have taken place since 1984 and have accelerated coming forward in time. For example the 1980 GDP that had been reported as $2.708 trillion in 1992 had crept up by 2.9% to $2.786 trillion based on 2004 reporting. The 1990 GDP, however, had Pollyanna Creep of 5.3% over the same period.
---------------------------------------- POLLYANNA CREEP ---------------------------------------- Change in Reporting Year 1950 1984 2004 2004/1992 ---------------------------------------- GNP (Billions of Current Dollars) ---------------------------------------- 1929 103.8 103.4 104.4 +0.97% 1933 55.8 55.8 56.7 +1.61% 1940 101.4 100.0 101.7 +1.70% 1950 284.2 286.5 295.2 +3.04% 1960 -- 506.5 529.5 +4.54% 1970 -- 992.7 1044.9 +5.26% 1980 -- 2631.7 2823.7 +7.30% ---------------------------------------- GDP (Billions of Current Dollars) ---------------------------------------- Change in As Reported Reporting in 1992 2004 2004/1992 ---------------------------------------- 1980 2708.0 2785.5 +2.86% 1990 5513.8 5803.1 +5.25% ----------------------------------------
Double-Entry Bookkeeping

The NIPA effectively is a double-entry bookkeeping system, where an item on the consumption side of the ledger, in the GNP/GDP accounts, is offset on the income side of the ledger, in Gross Domestic Income (GDI) accounts. In theory, the GNP and the GDI should be identical. In practice they rarely are, with the latest "statistical discrepancy" showing GNP to be $67 billion, or 0.6% higher than the GDI. This is due to the BEA's inability to reconcile the two series.

Part of the problem is that source data often are estimated without regard to actual numbers otherwise available. As an example of how far from reality the GNP/GDP/GDI reporting has gone, consider data from a high quality and unbiased resource: the Internal Revenue Service (IRS).

Based on its analysis of income tax returns, the IRS reports that, "For the second consecutive year, Adjusted Gross Income (AGI) fell, decreasing by 2.3% to $6.0 trillion for 2002. This represents the first time since prior to 1950 that total AGI reported on individual tax returns has fallen for two successive years."[9]

While one might expect to see some parallel income reporting in the GDI, it only happens by coincidence. Although the BEA considers the IRS data, it never has been able to reconcile the differences between GDI assumptions and IRS reality. Of course, the BEA sticks with the GDI assumptions, which have income rising in 2001 and 2002. The following table shows some of the specifics of comparable income components. Where wages and salaries are the single largest component in the GDI, they grew by 6.8% in 2002, according to the BEA, but the IRS reports a 0.4% contraction.
----------------------------------- INCOME GROWTH 2002/2001 -- IRS VERSUS BEA (Not Adjusted for Inflation) ----------------------------------- Income Category IRS GDI ----------------------------------- Wages & Salaries -0.4% +6.8% Interest Income -20.9% -6.4% Dividend Income -14.9% +5.1% -----------------------------------
Part of the difference is in imputations, which gets back into the theoretical structure of the NIPA. Any benefit one receives, either living in one's own house, or receiving free checking from a bank has an imputed income component. Free checking, for example, is calculated as imputed interest income. Not only did imputed interest income account for 21% of all personal interest income in 2002, but also it grew at an annual rate of 8.3%! As an aside, renting the house you own from yourself gets imputed as 62% of total rental income.

Another issue is distortion in underlying series. The bias factors (now reported as net business birth/death modeling) inflate reported payroll employment, as discussed in this series' first installment. GDI estimates of wages and salaries are calculated off the payroll numbers and are inflated on a parallel basis.

Deflation Wonders

As emphasized earlier, the lower the inflation rate that is used to deflate the GDP, the higher will be the resulting inflation-adjusted growth.

One of the deflation stars is the computer. While computer prices have come down over time, the quadrupling and re-quadrupling of memories provided with a standard computer have, through hedonics and quality adjustments (see Installment III on the CPI), enhanced the decline in prices used in deflating computer consumption in the GDP. According BEA deflators, $1,000 computers bought in 1990, 1995 and 2000 would cost $48.63, $95.84 and $526.58, respectively, today. I bought computers in each of those time frames and could not replicate any one of them for the suggested proportionate price in deflated dollars, regardless of free memory enhancement.

One of the more significant changes to GDP inflation was made in 1996, when the deflator was shifted from fixed-weighted to a chain-weighted basis. The chain-weighted basis weights inflation for a two-year grouping of a related GDP component, rather than using the weighting of the benchmark year. One happy side effect of this change is that the components of inflation-adjusted GDP do not add up to the total, with the difference being allocated to the residual category. As of the "final" second-quarter 2004 real GDP, the residual was a negative $35.6 billion, or 0.33% of total GDP. The residual usually gets worse the more removed it is from the benchmark year, which is 2000 at present. As of the fourth-quarter 1990, for example, the residual is 13.4% of GDP. Before 1990, the BEA does not publish the detailed breakout of accounts, because of the large residual. For some reason, this bothers a number of well-reputed economists.

A Tempting Target for Manipulation

In the introduction to this series on government reporting, I mentioned political manipulation of the GNP/GDP in the Johnson and first Bush administrations that went beyond overly positive methodological changes. In both instances, my sources were consulting clients who had been involved directly in the process. In the latter instance, an individual at the BEA also confirmed the situation.

Few people argue with the GNP/GDP reports, so when Lyndon Johnson kept sending the initial GNP estimates back to the Commerce Department for correction, he eventually got what he wanted, and the media dutifully reported stronger than actual economic growth.

Near the end of the first Bush administration, an outside-the-system manipulation was worked. A senior member of the Executive Branch approached a senior officer of a large computer company and requested that reporting of computer sales to the BEA be inflated. This was done specifically to help with the reelection effort. The request was granted, and thanks to the heavy leverage of computer deflation, reported GDP growth enjoyed an artificial spike.

There are suggestions of other direct manipulations over time, specifically involving the Clinton administration and the current Bush administration. Most recently, a bizarre annual revision to the GDP data eliminated the 2001 recession, at least as traditionally defined with two consecutive quarters of real GDP contractions.

Where little public attention is paid to the GDI, however, it is interesting to note that the revisions did not follow the same pattern on the inflation-adjusted income side of GDP. Pre-revision numbers showed quarterly real GDP contractions in third-quarter 2000 and the first- through third-quarter 2001. In the 2004 annual revisions, second-quarter 2001 GDP growth turned positive (from -0.6% to +1.2%), breaking up any consecutive quarterly GDP declines. The patterns were repeated in revisions of the GNP. Following the latest annual revisions, however, the GDI-same as GNP in theory-showed contractions in fourth-quarter 2000, second- through fourth-quarter 2001 and third-quarter 2002.

Estimating Economic Reality

Based on my analysis of the GDP/GNP revisions and redefinitions over time, over-deflation and economic reporting as published before later political corrections, reporting of real GDP growth at present is overstated by roughly three percent per year against a more realistic, pre-Pollyanna Creep period.

Where the period of bloated GDP reporting began after the severe double-dip recession of 1980 and 1981/1982, it includes the last two recessions that were severe enough to generate reported GDP contractions. Both the 1990/1991 and 2001 recessions were deeper and longer than currently estimated. The recession from July 1990 to March 1991 (timing per the NBER) really began in late-1989 and persisted into 1992, perhaps even 1993. Such was evident in the underlying data of the time. Due to the NBER's early call of the recession's end, however, the first "jobless recovery" was seen.

Similarly, the recession that was timed from March to November 2001, began in late-2000 and persisted into 2003. Again, because of an early call to the recession's end, a "jobless recovery" was seen.

There also were economic downturns in 1986 and 1995 that were evident to most companies dealing in real world economic activity at he time. Although the contractions showed up in a number of measures, they were not severe enough to turn bloated GDP growth negative.

As the economy once again appears to be faltering, or losing traction, risk is high of renewed or a double-dip recession, of which the 2001 downturn eventually will be counted as the first leg.

I have only touched upon some of the highlights in problems with GDP reporting. Unfortunately, though widely followed, the series is probably the least meaningful of the major economic statistics followed by investors and the financial media.

Footnotes to Installment Four

[1]Full definitions and methodologies are available at the BEA's wbsite BEA.

[2]The information on the guesstimation process is based on my conversations with individuals at the BEA during the last 25 years. The economic consensus misses turning points in the economy about 100% of the time.


[4]Though the NBER now denies such a definition was ever used, the NBER supplied me with this definition in a conversation back in the 1980s.

[5]Free trade theory assumes all involved nations are at full employment. When that is not the case the wealthiest and highest salaried countries end up with a declining standard of living and redistributing their wealth to the other free-trade participants, as is the current circumstance for the United States.

[6]While currency values can have relatively quick impact on trade in pure commodities, products with quality differentiation combined with the financial and marketing creativeness of importers and exporters often bypass standard theory.

[7]This is an actual component of the income side of the GDP.

[8]BEA, various historical editions of the Statistical Abstract of the United States, Department of Commerce.

[9]Individual Income Tax Returns, Preliminary Data, 2002, IRS website IRS.