Revisions to BEA s Estimates of GDP and GDI

Transcription

1 Revisions to BEA s Estimates of GDP and GDI Dennis Fixler Presentation at Quarterly Meeting of Council of Professional Association on Federal Statistics (COPAFS) December 7, 2012

2 Outline Why are there revisions? Assessing accuracy Long-term patterns Short-term patterns Plans for the 2013 Comprehensive Revision GDP vs. GDI 2

3 Revisions are not errors Schedule is determined and is a commitment Error correction is a term that is reserved for true errors Main cause: availability of more complete and revised data 3

4 Why are there revisions? Revised estimates are made because the available source data becomes more complete and of higher quality with the passage of time BEA produces a number of vintages of estimates of GDP, each successive one incorporates the better and more complete source data 1. Three current quarterly vintages; advance, second, and third 2. Three annual revision vintages; first, second, and third annual 3. Comprehensive benchmark estimates: occur about every five years, and incorporate both a quinquennial input-output table that uses data from economic censuses and definitional revisions that keep the NIPAs abreast of a changing economy. 4. GDI also has a fourth current quarterly vintage that incorporates data from the Quarterly Census of Employment and Wages. 4

5 Why are there revisions? (continued) Source data flow mainly from Census, BLS and SOI The source data for the first four vintages of estimates of GDP change a lot. Advance Second 5

6 Where are there revisions? (continued) Third First Annual trend monthly plus trend monthly of quarterly revised annual 6

7 New source data vs. indirect indicators For some components of GDP and GDI, direct indicators aren t available State & local government spending Current data on employment and construction Data on other types of spending available from Census Bureau with a 2-year lag Corporate profits Current data on public corporations Comprehensive IRS data available (including private and S corporations) with a 2-year lag IRS data on proprietors and partnership income with a 2-year lag 7

8 Why are there revisions? (continued) For the advance estimates, nearly two-thirds of the estimates are either entirely trend projections, or incorporate at least some trend projections The trend projections in the second estimates are down to about one-quarter The trend projections are a bit more than one-fifth of the third estimates In the first annual estimates, the trend projections are just one-twentieth, and revised quarterly or monthly data and annual data are used about equally for the rest. The source data for GDI estimates undergoes a similar process of change and improvement with successive estimates. 8

9 Example: Benchmarking manufacturing Indicator: Monthly M3 ( manufacturers shipments, inventories, and orders ) Voluntary reporting 4,300 reporting units with more than $500 million in shipments (represents 60% of manufacturing) 89 industry categories No product detail Benchmark: Annual survey of manufactures Mandatory reporting 50,000 establishments (representing all 328,500 manufacturing establishments) 471 industries 1,384 product classes 9

10 Shipments - Nondefense capital goods excluding aircraft Levels (millions of dollars, seasonally adjusted) 70,000 68,000 66,000 64,000 62,000 60,000 58,000 56,000 54,000 52,000 50, m1 2008m1 2009m1 2010m1 2011m1 Previously published Revised Source: Census Bureau, July 3,

11 Business inventories 1,600,000 Levels (millions of dollars, seasonally adjusted) 1,550,000 1,500,000 1,450,000 1,400,000 1,350,000 1,300, m1 2009m1 2010m1 2011m1 Previously published Revised Source: Census Bureau, July 20,

12 Retail sales excluding motor vehicles Levels (millions of dollars, seasonally adjusted) 300, , , , , , , , m1 2008m1 2009m1 2010m1 2011m1 Previously published Revised Source: Census Bureau, March 30,

13 Why does BEA revise its estimates? Improve accuracy of the estimates by incorporating the most complete and reliable source data Benchmarking indicators to annual surveys Replacing trends or indirect indicators with comprehensive source data Provide a more detailed picture of the economy Make improvements to methods used for preparing the estimates 13

14 Assessing accuracy BEA judges the accuracy of its early estimates by whether they present the same picture of the economy as its latest estimates in terms of: 1. Long-term growth rates 2. Tends in saving, investment, government spending, corporate profits, and other key components of GDP and GDI 3. Broad features of business cycles, including the timing and depth of recessions, the strength of recoveries, and the major components contributing to growth and contractions 4. Patterns of quarterly growth, including whether it is high or low relative to trend, acceleration or deceleration, and whether it is positive or negative. 14

15 Long-run patterns Revisions to GDP growth average less than 0.1 percentage point over the comprehensive revisions in the period Revisions to the contributions of key components of GDP growth are small and do not substantially change the ordinal rankings of the contributions contributions. The overall pattern of change in GDP over business cycles is little changed by revisions. 15

16 Long-run patterns (continued) For the period , real GDP successfully indicated: 1. The direction of change in real GDP 97 percent of the time 2. The acceleration of growth about 72 percent of the time 3. Whether growth was above, near, or below trend four-fifths of the time 4. The cyclical peaks before 5 of the 6 recessions in The cyclical troughs of 4 of the recessions. Source: Fixler, Greenaway-McGrevy and Grimm, Revisions to GDP, GDI, and Their Major Components, July 2011 Survey of Current Business 16

17 Long-run patterns (continued) 17

18 Long-run patterns (continued) 18

19 Long-run patterns (continued) Accuracy for long-term trends Over the full sample period , the shares of GDP taken by consumption, fixed investment, inventory investment, exports, imports and government are little changed from the early to the latest estimates Consider the ratio to GDP of government expenditures, which since 1985 has seen no revisions more than.01; it exhibits the same general patterns over time in both the early and latest estimates In contrast, fixed investment experienced a major revision in the 1999 comprehensive revision and its ratio to GDP was raised by several hundredths The ratios of components to GDI likewise have also changed only modestly 19

20 Long-run patterns (continued) 20

21 Long-run patterns (continued) 21

22 Long-run patterns (continued) The ratio of national saving to GDI has been revised only modestly: Ratio of National Saving to GDI Q1 2000Q3 2001Q1 2001Q3 2002Q1 2002Q3 2003Q1 2003Q3 2004Q1 2004Q3 2005Q1 2005Q3 2006Q1 2006Q3 2007Q1 2007Q3 2008Q1 2008Q3 2009Q1 Third 2009 comp. Rev. Although there has been some quarter-to quarter smoothing, the largest revisions have been generally less than about

23 Long-run patterns (continued) Corporate profits have undergone some smoothing on a quarter-to-quarter basis Q1 2000Q3 2001Q1 2001Q3 2002Q1 2002Q3 Corporate Profits 2003Q1 2003Q3 2004Q1 2004Q3 2005Q1 2005Q3 2006Q1 2006Q3 2007Q1 2007Q3 Third July comp. Rev. 2008Q1 2008Q3 2009Q1 23

24 Long-run patterns (continued) The ratios for the sum of the other components of operating surplus (net interest, proprietors income, and rental income) have also been largely maintained, with no revisions in either series more than about Other Components of Operating Surplus 2000Q1 2000Q3 2001Q1 2001Q3 2002Q1 2002Q3 2003Q1 2003Q3 2004Q1 2004Q3 2005Q1 2005Q3 2006Q1 2006Q3 2007Q1 2007Q3 2008Q1 2008Q3 2009Q1 Third July comp. Rev. 24

25 Long-run patterns (c0ntinued) Similarly for the ratio of national saving to GDI; although there has been some quarter-to-quarter smoothing, the largest revisions have been generally less than about Ratio of National Saving to GDI Q1 2000Q3 2001Q1 2001Q3 2002Q1 2002Q3 2003Q1 2003Q3 2004Q1 2004Q3 2005Q1 2005Q3 2006Q1 2006Q3 2007Q1 2007Q3 2008Q1 2008Q3 2009Q1 Third 2009 comp. Rev. 25

26 Long-run patterns (c0ntinued) Average revisions of GDP and GDI are modest, and of similar sizes: Average Revisions to Anual Estimates of GDP and GDI; [Percentage points] Mean Revision Mean Absolute Revision Vintage GDP GDI GDP GDI First annual Second annual Third annual Nominal values; same price index would be used for real 26

27 Long-run patterns (continued) Revisions to Annual Estimates of GDP and Its Major Components in 1983 to 2009 [Percentage points] Mean Revision Mean Absolute Revision Current-dollar GDP Real GDP Current-dollar GDP Real GDP Gross domestic product First annual Second annual Third annual Personal consumption expenditures First annual Second annual Third annual Gross private domestic investment First annual Second annual Third annual Exports First annual Second annual Third annual Imports First annual Second annual Third annual Government consumption expenditures and gross investment First annual Second annual Third annual

28 Long-run patterns (continued) Generally, and for GDP, most of the revision is between the first and second annual estimates. The revisions between second and third annual estimates are typically smaller. Revisions to Annual Estimates of GDI and Selected Components in 1983 to 2009 [Percentage points] Mean Revision Mean Absolute Revision First annual Second annual Third annual First annual Second annual Third annual Gross domestic income Compensation of employees Proprietors' income with IVA and CCAdj Corporate profits with IVA and CCAdj

29 Long-run patterns (c0ntinued) MARs of both GDP and GDI are somewhat larger around cyclical points than in all quarters generally. MARs Around Cyclical Turning Points: Third to Latest Estimates [Percentage points] GDP$ GDI$ Prior Peak After Prior Trough After Recessions: , , 1980, , , 2001 Prior is the quarter before the turning point and after is the quarter after. Real GDP and GDI are not shown because both are calculated using the same price index. 29

30 Long-run patterns (c0ntinued) Comprehensive benchmark revisions incorporate both definitional revisions and new input output tables that are based on the most recent quinquennial economic censuses. For the revisions to current-dollar estimates of GDP in the most recent 5 comprehensive revisions: 1. The mean revision is 0.05 percentage point; a modest tendency toward upward revisions. 2. The MARs average somewhat more than 0.5 percentage point; largely due to definitional revisions. 3. Means for the components are both positive and negative, and the MARs range from 0.3 to 1.5 percentage point. 30

31 Long-run patterns (c0ntinued) Average Revisions to Quarterly Estimates of Current-Dollar GDP in Comprehensive Benchmark Revisions, and GDP Components in the 2009 Comprehensive Revision [Percentage points] Mean GDP Mean Avbsolute Year of Comprehensive Revisio Period Revision Revision :I-1991:III :I-1995:III :I-1999:II :I-2003:III :I-2009-I Average GDP Components Personal consumption expend 1983:I-2009-I Fixed nonresidential investmen 1994:I-2009-I Residential investment 1994:I-2009-I Exports 1986:I-2009-I Imports 1986:I-2009-I Government expenditures 1983:I-2009-I

32 Short-term patterns Focus now on the performance of the three current quarterly vintages Mean revisions are about 0.1 to 0.3 percentage point for all three vintages Mean absolute revisions mean revisions without regard to sign of current-dollar GDP are 1.1 to 1.2 percentage point, with a very slight reduction in the second and third estimates relative to the advance estimates. Mean average revisions of real GDP are about 1.3 percentage points, with no tendency to decline in the later vintages. 32

33 Short-term patterns (continued) The general lack of declines in later vintages has been the result of all of BEA s previous revisions studies. In comparison, the MARs for the average forecasts of the Philadelphia Federal Reserve Bank s Survey of Professional Forecasters, which are made early in the quarter, are 1.9 percentage points for current-dollar GDP, and 1.6 percentage points for real GDP. 33

34 Short-term patterns (continued) Revisions of Current Quarterly Estimates to Latest Estimates of GDP in [Percentage points] Mean Revision Mean Absolute Revision Current-dollar GDP Real GDP Current-dollar GDP Real GDP Advance Second Third

35 Short-run patterns (continued) Mean Absolute Revisions to Successive Vintages of Changes in Current-Dollar GDP; [Percentage points] Vintage of revision used as standard Vintage Second Third First annual Second annual Third annual Latest Advance Second Third First annual Second annual Third annual

36 Short-tern patterns (c0ntinued) In the table: 1. The MARs increase monotonically with later vintages of revisions used as standards. 2. All the MARs of all vintages to the latest are within about 0.2 of 1.0, with later vintages of estimates having smaller MARs. Comprehensive benchmark revisions incorporate both definitional revisions and new input output tables that are based on the most recent quinquennial economic censuses. 36

37 Short-term patterns (continued) Revisions to seasonal adjustment factors can also play a role Fixler and Grimm (January 2002 Survey of Current Business) examined the effects of revisions to seasonal adjustment factors on revisions to current-dollar GDP. 1. revisions to seasonal factors accounted for 1.0 percentage points, slightly larger than the MAR for GDP of 0.7 percentage point. 2. Seasonal adjustment factors are ultimately calculated using three years about the future to any given quarter and cannot be successfully anticipated. 3. This suggests that there is a non-zero asymptote that the MAR for GDP cannot go below. 37

38 Short-term patterns (continued) For the averages for revisions to GDI: 1. There are no advance estimates for GDI because some components are not yet estimated. 2. There is a fourth current quarterly estimate, appearing 2 months after the third and incorporating information from the Quarterly Census of Employment and Wages. 3. The mean revisions are quite small, and two have small negative values. 4. The MARs range from 1.2 to 1.4 percentage points, again with no tendency to get smaller with later vintages. 38

39 Short-term patterns (continued) Revisions to Quarterly Estimates of GDI; [Percentagve points] Mean revision Mean Absolute Revision Advance Second/1/ Third Fourth/2/ No fourth quarter estimates beginning in Estimates begin in 2002:I 39

40 Short-term patterns (c0ntinued) What about revisions to prices? Revisions of Current Quarterly Estimates to Latest Estimates of Price Indexes of GDP and Its Major Components; 1997:I to 2009:IV [Percentage points] Mean revision Standard deviation Mean absolute revision Gross domestic product Advance 0.25 * Second 0.24 * Third 0.20 * * Significant with p

41 Short-term patterns (c0ntinued) Revision to GDP price indexes. 1. The mean revisions for the GDP price index were small but positive. 2. The MARs were not much larger than the mean revisions, and much smaller than the MARs for GDP and GDI. 3. The mean revisions, however, are statistically significant. 4. The statistical significance results reflect statistically significant revisions to the price indexes for residential investment and for state and local government expenditures. 41

42 Results of 2012 Annual Revision

43 Revisions to the recession Revisions were unusually large Early estimates significantly understated the extent of the decline Revisions exposed gaps in data used for early estimates Some revisions came quickly as actual data replaced assumptions and indicators were revised 2008Q4 revised from 3.8% (Jan 30, 2009) to 6.2% (Feb 27, 2009) Other revisions came later with annual revisions In July 2011, 2008Q4 was revised to 8.9% Total revision from advance to latest (5.1 percentage points) is the largest downward GDP revision on record 43

44 Revisions to GDP contraction, 2007Q4 2009Q2 Release date July 2009 July 2010 July 2011 July 2012 Cumulative percent change [not at annual rate] Cumulative revision From first estimate for full contraction period to latest estimate, cumulative contraction in real GDP revised from 3.7 percent to 4.7 percent Revisions in July 2011 mostly reflected newly available annual surveys for

45 Sources of revisions to 2008Q4 GDP [percentage points] Advance Third Annual (Latest) Revision Gross domestic product Personal consumption Fixed investment Change in private inventories Exports Imports Government expenditures Revisions to inventory investment came in 2 nd estimate (Feb 2009) and in 1 st & 2 nd annual revisions (July 2009, 2010) Revisions to personal consumption came in 2 nd estimate (Feb 2009) and in 3 rd annual revision (July 2011) Revisions to fixed investment in 2 nd annual revision 45

46 Misperceptions about revisions BEA s practice of annualizing percent changes exaggerates perceived revisions GDP revisions Quarter 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 Percent change Annual rate Percent change Quarterly rate Advance Latest Revision Advance Latest Revision

47 Proposed major conceptual changes for 2013 Comprehensive Revision Capitalization of research and development (R&D) Capitalization of entertainment, literary, and artistic originals Accrual measures for defined-benefit pension plans Improved measures of banking services New approach for measuring the costs of ownership transfer 47

48 Summary of impacts: 2007 GDP Impacts (billions of dollars) Total GDP (as published in the NIPAs) 14,028.7 Plus: Business own-account and purchased R&D investment Less: Software overlap adjustment Plus: CFC on government and nonprofit R&D assets Total GDP, with R&D as investment 14,377.8 Impact on GDP of capitalizing R&D

49 Recent discussion about whether GDI is better than GDP Both GDP and GDI provide information about economic activity A simple average of the two was used in BEA articles in the SCB 49

50 Average of real GDP and real GDI 50

51 Average of real GDP and real GDI 51