MANAGER’S BRIEFCASE: USING THE DATA FOR WAGE AND UNIT LABOR COSTS
3.10 Preliminary and revised data
Approximately one month after the end of each quarter, the BEA announces its
‘‘advance’’ estimates of real GDP and inflation. These figures are greeted with great anticipation, and can influence policy decisions. However, these advance estimates consist largely of estimated data. At the time the advance estimates are released, the BEA figures are based on fairly firm data for the first month of each quarter, preliminary data for the second month, and mostly guesses for the third month. As a rough order of magnitude, only about 50% of the data that are used to generate the advance estimates of GDP are actually known.
The BEA acknowledges that these numbers are guesses, and even circulates the estimates of missing data they use to calculate GDP and its components. The alter-native, which would be waiting another month or two, would simply encourage private sector forecasters to fill in the blanks. No criticism of the BEA is implied because they use estimates for some of the components of GDP.5
Nonetheless, users of these data should be aware these estimates change signif-icantly. According to the BEA, the average error for the percentage change in real GDP between the advance and final estimate is 1.4%. For example, if the advance estimate shows an increase of 3.0% in real GDP, there is a 1 in 3 chance that the final number will be less than 1.6% or more than 4.4%. Managers should keep that in mind when the BEA initially announces that the recession has ended because real GDP rose 0.5% last quarter.
The GDP numbers are a compendium of data collected from other sources: retail sales, inventory stocks, foreign trade data, manufacturers’ shipments, construction put in place, and so on. The advance estimates for retail sales are based on a small sample that is only about 10% of the full sample, and hence are subject to larger revisions than most other data. Also, more than 50% of the data used in the initial estimate of industrial production is based on guesswork.
In one well-documented case, the advance estimate of retail sales for May 1995 rose only 0.2%; revised data showed an increase of 0.9%. That caused Alan Blinder, then Vice-Chairman of the Federal Reserve Board, to state that he would not have voted to reduce the Federal funds rate in July 1995 if he had known retail sales were that strong. In this particular case, inflation did not rise, so the Fed ultimately turned out to have made the right decision. Nonetheless, serious errors in advance and preliminary data can lead to policy errors.
There are many other similar examples. In another important case, revised data showed that the 1990–91 recession was about twice as severe as the preliminary data suggested; knowledge of that point probably would have led to earlier Fed easing and a more robust recovery instead of the near-stagnation that occurred for the rest of that year. A similar error by the BEA occurred in underreporting the severity of the 2001 recession, although in that case the Fed eased more rapidly.
The NIPA data are usually rebenchmarked every five years, based on the com-prehensive economic censuses taken at that interval. These are not the same as the
decennial population censuses, but contain complete enumeration of all firms in manufacturing, agriculture, and services. These data represent the most compre-hensive benchmarks for US economic data. Such revisions usually do not alter the quarterly changes very much, but affect the underlying trends.
The change in methodology that introduced the chained deflators in 1995 resulted in a substantial downward revision of previous growth rates. Real growth in 1993 was reduced from 3.1% to 2.2%; growth in 1994 was lowered from 4.0% to 3.5%, and the increase in real GDP for the first half of 1995 was diminished from an annual rate of 1.5% to 0.5%. Hence this one-time change in methodology sliced almost 1% off the growth rate for this 212 year period. Admittedly, these were one-time changes due to a major switch in methodology that probably will not be repeated for many years. Nonetheless, the profile of how fast the economy was growing was seriously altered by this change in methodology. In October 1999, these growth rates were boosted back to 2.4%, 4.0%, and 1.1%. Further changes are expected in the future.
The main thrust of these comments for managers is that the preliminary govern-ment data must generally be greeted with some skepticism, since they are often revised substantially. That certainly does not mean the reports should be ignored, however. If your sales figures are moving in a direction that does not agree with preliminary government data, that might mean your company is gaining or losing market share – or it might mean the data will be revised. Particularly for retail sales and industrial production, it is best to wait at least one additional month before taking any definitive action based on preliminary government data. To track the economy on a monthly or quarterly basis, managers are well advised to focus on the index of coincident indicators rather than the preliminary GDP data.
KEY TERMS AND CONCEPTS
Chained Price Indexes Leading Indicators
Coincident Indicators Producer Price Index (PPI) Consumer Price Index (CPI) Productivity
Cyclical Unemployment Seasonal Unemployment
Fixed-Weight Price Indexes Structural Unemployment
Frictional Unemployment Unit Labor Costs
Full Employment Implicit Price Deflators
SUMMARY
• The key numbers describing the overall econ-omy are the growth rate in real GDP and the
rise in the price level, not the current dollar level of GDP.
• The government publishes three different types of price indexes: fixed-weight, implicit deflators, and chained price indexes. The fixed-weight index invariably overstates the true rate of inflation, while the implicit deflator understates it if some products have rapidly decreasing prices. The chained price index provides the closest approximation to the true rate of inflation, but has the draw-back that the chained-dollar components of GDP do not sum to total real GDP.
• All price indexes may be overstated if they fail to take into account the timely introduc-tion of new products, and the improvement in quality of existing products.
• The Boskin Commission estimated that through 1995, the consumer price index (CPI) overstated the true rate of inflation by an average of 1.1% per year. The Bureau of Labor Statistics (BLS) has implemented most of the Commission recommended changes. Cur-rently, the CPI probably understates the true rate of inflation by 0.5% to 1.0% per year.
• There are four theoretical types of unem-ployment: seasonal, frictional, cyclical, and structural. Even at full employment, there will always be some seasonal and frictional unemployment in a dynamic capitalist soci-ety. Seasonal unemployment is erased by adjusting the data and is not reported.
• The BLS publishes six different measures of unemployment. The ‘‘official’’ rate, which is the one most widely quoted, consists of four categories: job losers, job leavers, new entrants, and reentrants.
• Full employment is not zero unemployment.
It is often defined as the rate when the num-ber of people looking for work equals the number of job vacancies, or the rate at which everyone who wants to work can find a job.
• The full-employment rate is not a fixed number, but varies considerably with changes in demographic factors, government policies, and the long-run growth rate. In the US, the full-employment rate of unemploy-ment rose from less than 4% in the early post-WWII period to almost 6% by 1980, and then declined to 4% by the late 1990s.
• The BLS publishes three different estimates of labor costs. The most comprehensive measure, which includes fringe benefits and various types of bonuses, rose much more rapidly in 2000 than indicated by the more common measures of labor costs and inflation.
• Virtually all government data are seasonally adjusted. In some cases, the seasonal pat-terns will shift over time, so that these data will provide erroneous information about recent changes in underlying data.
• The advance and preliminary data published by various government agencies are subject to substantial revision, and the initial esti-mates are sometimes far off the mark. That is particularly true for retail sales, indus-trial production, and real GDP; errors for employment, unemployment, and inflation are usually much smaller.
QUESTIONS AND PROBLEMS
1. Calculate the various measures of inflation for the following data. Assume all price indexes are 100 in Year 1.
Year 1 2 3
College tuition 40 @ $20,000 40 @ $25,000 40 @ $30,000 Apples 300,000 @ $1.25 250,000 @ $1.65 280,000 @ $1.50 Computers 600 @ $1,200 750 @ $1,000 1,000 @ $800
(A) What factors cause the change in the CPI to overstate the actual rise in the cost of living?
(B) What are the differences between the CPI and the implicit deflator for GDP?
Which do you think is a better measure of inflation?
(C) What is a chained price deflator? What are the benefits and drawbacks to this measure of inflation?
(D) Calculate the CPI, implicit consumption deflator, and chained price con-sumption deflator in years (2) and (3) with these data.
(E) What do you think the ‘‘actual’’ rate of inflation was in years (2) and (3)?
Explain your answer.
2. (A) How is full employment defined, and why does the full-employment rate of unemployment keep changing?
(B) In December 1998, the BLS reported the following unemployment rates:
1. Unemployed over 15 weeks 1.1%
2. Job losers 2.0%
3. ‘‘Official’’ unemployment rate 4.3%
4. Plus discouraged workers 4.5%
5. Plus marginal workers 5.2%
6. Plus part-time workers 7.5%
Which do you think is the best measure of actual unemployment? Explain your choice.
(C) In 1990, most economists thought the full-employment rate of unemploy-ment was about 6%, but in 2000 they thought it was about 4%. Were they wrong in 1990, or was this change based on economic factors? If so, what factors caused them to change their opinion?
3. Classify the following job losses as seasonal, frictional, cyclical, or structural unemployment – or no change in the unemployment rate at all.
(A) No work for bricklayers in Illinois in February.
(B) GM lays off 50,000 workers after Fed restricts credit.
(C) GM closes plant in Flint, MI; builds new assembly plant in Mexico.
(D) IBM shuts down economic research department.
(E) North Dakota wheat farmers abandon farms after Asian recession reduces worldwide price of wheat.
(F) NBA basketball players locked out by owners.
(G) Firm relocates headquarters from New York to Dallas. Half of staff decides not to accept relocation.
(H) Unnamed fast food chain goes bankrupt after E-coli bacteria poison customers.
(I) High interest rates cut housing starts by 50%, reducing construction employ-ment by 650,000.
(J) Recent June graduates have difficulty finding employment.
4. In 1955, the last year when social security payments included only old-age payment (before disability), payments totaled $4.9 billion. For 2001, the figure was $433 billion (excluding Medicare, which was another $218 billion). During the intervening period, the CPI rose at an average annual rate of 4.2%, and the number of people over 65 rose an average of 2.0% per year. What would social security have been in 2001 if the program had not expanded after 1955 (i.e., if the only increases were due to inflation and population)? Now suppose the CPI had been overstated by 1.1% per year. What would social security payments have been in 2001 if the actual rate of inflation had been used?
5. What are the major causes of error in the preliminary NIPA data, and what could be done to reduce these errors?
6. If economists generally agree that fixed-weight price indexes overstate the actual rate of inflation, why is the CPI still the most popular and widely quoted measure of inflation?
7. What major factor causes the implicit price deflator and the chained price deflator to diverge over time? Under what circumstances would we expect to find very little difference in these two measures of prices?
8. As a manager, one of your responsibilities is to monitor recent changes in infla-tion. However, the recent statistics appear to be quite confusing. The CPI rose only 0.1%, but the PPI for finished goods increased 0.8%. A further examination of the core rates shows a gain of 0.2% in the CPI and 0.5% in the PPI.
(A) What further information would be required to determine the underlying rate of inflation?
(B) Suppose the core rate for the intermediate rate rose 0.1%. How would that affect your answer?
(C) What other components of the CPI and PPI would you examine closely to try and narrow the discrepancy?
9. In August 2002, preliminary data showed that payroll employment rose 39,000, household employment rose 429,000, and the unemployment rate fell from 5.9%
to 5.7%.
(A) Based on these data, what conclusion would you reach about the current employment situation?
(B) How would your answer change if you also noted that (a) the unemployment rate for men 20 years and over was unchanged, and (b) the number of initial unemployment claims rose from 386,000 to 398,000?
(C) How would your answer change if you also noted that the biggest declines in the unemployment rate occurred in (a) teenagers, and (b) experienced wage and salary workers?
(D) Based on this welter of confusing information – which is not unusual – what conclusions can be drawn about basing business plans on the latest monthly employment and unemployment data?
Notes
1. For a detailed discussion of these issues, see ‘‘BEA’s Chain Indexes, Time Series, and Measures of Long-Term Economic Growth’’ in the May 1997 issue of the Survey of Current Business.
2. Michael Boskin et al., Journal of Economic Perspectives, Winter, 1998.
3. While this seems counterintuitive, the reasoning is as follows. ‘‘Output’’ in the financial service sector is defined as wages plus profits. Productivity is output per employee-hour. Thus if ATM machines replace tellers, wages, and hence ‘‘output,’’ have fallen.
4. Because the Conference Board is a private organization, they have no obligation, as does the government, to supply these data free of charge. The reports can be purchased at
$500 per year, including historical data. The latest monthly report, however, is always available on their website free of charge.
5. In the late 1970s and early 1980s, the BEA used to prepare a ‘‘flash estimate’’ of GDP based on information available until the last week of any given quarter. Because of the high degree of estimation, this figure was designed only for internal use, but it was widely ‘‘leaked’’ and eventually was issued on a public basis. However, the procedure was flawed enough that it was soon discontinued. At the time it was claimed that private sector economists would fill the void, but that did not happen, and today no one prepares a flash estimate.