Use of the Annual I O for productivity analysis at the Federal Reserve

Productivity Analysis at the Federal Reserve: The Role of the BEA’s Annual Industry Accounts Paul Lengermann Industrial Output Section Federal Reserve Board of Governors BEA Industry Accounts Users’ Conference, October 26, 2007 Outline Background and motivation FRB Productivity System • Conceptual Framework • Industry Measures & Empirical Challenges • Results for Basic Industry Results FRB Custom Aggregates • Construction • Key Findings • Extensions Concluding Remarks 2 Background & Motivation Late-1990s – • Sectoral Productivity Project –alternative, largely income based approach for studying productivity • High-tech Measurement– work on IP index extended into links to aggregate productivity Following this: • Detailed industry level analysis – emphasis on role of high tech and nonfinancial corporations • Work on industry origins of statistical discrepancy 3 Background & Motivation (cont’d) More recently, focus shifted to combining the industry results to illuminate sectoral trends Coincided with 2004 comprehensive revision to Industry Accounts Key challenges: • Customizing aggregates to provide a more meaningful view of productivity • Traversing breaks and noise in the data Resulting ―six sector‖ estimates are: • Flexible in terms of composition & construction • Readily updatable 4 FRB Productivity System Developed by Eric Bartelsman & Joe Beaulieu, it is both: • a collection of SAS macros designed to automate routines commonly used in productivity research • and a data repository – Hierarchical databases & Meta-data – Stored data files include: • BEA: Annual Industry Accounts (AIAs), Benchmark IO, selected NIPA data • BLS: MFP data, Current Population Survey • Census County Business Patterns • FRB Industrial production 5 Conceptual Framework We adopt the Domar or ―deliveries-to-final demand‖ framework to examine the role of our sectors: • MFP growth at any level of aggregation can be decomposed into contributions from underlying sectors or industries Framework has nice theoretical properties: • Maps productivity growth into the rate of expansion of the social production possibilities frontier Production modeled using the concept of ―sectoral output‖ • Equals gross output less amount produced and consumed within the sector • For detailed industries, close to gross output; for aggregates, closer to value added 6 Growth Accounting (Bolded letters denote real growth rates) Productivity growth can be expressed either as weighted average of underlying industries: (1) MFPk =  d MFPi , where ―Domar weights‖ are ik k i dik  Si  Sk  Or it can be calculated residually: (2) MFPk = Sk•  I k• . We calculate industry-level MFP with (2) and sectoral MFP with (1). Finally, we conduct our sectoral growth accounting using: (3) Sk• = skN Nk• +  d [ MFPi + s Li + ik k i L i siK Ki + siR Ri] . 7 Industry Measures Nominal Sectoral Output: • Use the annual IO accounts to subtract own-industry intermediate use from gross output Real sectoral output growth: • Built from quantity indexes that assume the price index for own industry intermediates equals the published gross output deflator Real growth of intermediates from other industries: • Derived by stripping real value of own intermediates from total intermediates Real capital input growth : • Two broad measures of capital services constructed from BEA’s detailed fixed assets: – IT Capital (computers, software, comm. equipment) – Other capital (equipment, structures, inventories, land) 8 Industry Measures (cont’d) Labor Input: • Captures changes in hours worked of all persons • Implicit labor quality adjustment made using the very detailed industry-level data from Census County Business Patterns • BEA’s NAICS-based employment begin in 1998 and 2000 respectively: – Using Census data, we developed a concordance between the 1997 SIC and the 1997 NAICS – Concordance applied to unpublished BEA SIC-based data on hours 9 Source Data Discrepancies Application of concordance complicated by large differences in industry classification between Census & BLS • Especially pronounced for Census employment in Management of Companies (N55) These inconsistencies complicate productivity analysis: • Industry output data (Census) are classified differently than labor input (BLS). • Integrated annual IO and GDPI data inherited the BLS compensation levels • Also apparent in the CPS which complicates labor quality adjustments 10 Source Data Discrepancies (cont’d) What do we do? • Ignore the problem for compensation shares • Adopt Census NAICS industry composition in 1998 and treat the BEA/BLS employment measures as indicators going forward • Re-balance our concordance to be consistent with the ―adjusted‖ labor data before applying it to the pre-1998 SIC-based data Goal: labor input changes which are consistent over time 11 Sources of growth for U.S. private industry and major industry groups Sectoral Output (1) A. 1987 to 2005 1. Private industries 2. Goods-producing industries 3. Agriculture, forestry, fishing, and hunting 4. Mining 5. Construction 6. Manufacturing 9. Services-producing industries 10. Utilities 11. Wholesale trade 12. Retail trade 13. Transportation and warehousing 14. Information 15. Finance, insurance, real estate, rental, and leasing 16. Professional and business services 17. Educational services, healthcare, and social assistance 18. Arts, entertainment, recreation, accomodation, and food services 19. Other services 3.4 2.4 1.6 .4 1.9 3.7 3.6 1.5 4.0 4.4 3.3 6.1 3.6 4.6 3.4 3.0 2.4 IT Capital (3) .6 .2 .0 .1 .1 .2 .6 .2 .4 .2 .4 1.0 .7 .7 .2 .1 .1 Other Capital (4) .7 .2 .0 .0 .1 .1 .8 .4 .2 .2 .1 .5 1.4 .3 .4 .3 .1 Purchased Inputs (6) .3 1.0 .3 .7 1.4 1.4 .6 .0 1.3 1.4 1.1 2.4 1.2 1.9 1.9 1.6 2.0 MFP (2) 1.2 1.2 1.4 -.2 -.4 2.4 .8 1.1 1.7 2.2 1.2 1.6 .0 .2 -.8 .1 .1 Labor (5) .7 -.2 -.2 -.2 .8 -.5 .9 -.2 .5 .3 .5 .6 .3 1.5 1.7 .8 .2 • • • Each row is a sources of growth decomposition for period 1987-2005: Estimates generated in Dec with Industry Accounts release In progress: ―preliminary‖ estimates following July’s NIPA annual revision 12 FRB Custom Aggregates Initial (based on NAICS) High-tech Computer and electronic product mfg. (NAICS 334) Software publishing (NAICS 5112) Telecommunications services (NAICS 5133) Information services (NAICS 5141) Computer systems design and related services (NAICS 5415) Housing Industrial Construction (NAICS 23) Real Estate (NAICS 53pt) Manufacturing (NAICS 31, 32, 33, excl. 334) Mining (NAICS 21) Utilities (NAICS 22) Forestry, fishing and related (NAICS 113-5) Distribution Wholesale trade (NAICS 42) Retail trade (NAICS 44, 45) Transportation (NAICS 48) Warehousing (NAICS 49) Finance and business Finance and insurance (NAICS 52) Real estate and rental and leasing (NAICS 53pt) Professional, scientific, and technical services (NAICS 54, excluding 5415) Management of companies (NAICS 55) Admin, waste and related (NAICS 56) Data processing services (NAICS 5142) Personal and cultural Health care and social assistance (NAICS 62pt) Arts, entertainment, and recreation (NAICS 71) Accommodation and food services (NAICS 72) Other services (NAICS 81pt) Newspaper, book, etc. publishing (NAICS 5111) Motion pictures and sound recordings (NAICS 512) Radio/TV broadcasting (NAICS 5131-2) Services and Cultural Products to Persons Services to Businesses Merchandisers and Transportation of Goods Production of Goods Primary Producing Function 13 FRB Custom Aggregates (cont’d) Key challenge: Disaggregate several BEA industries to define our sectors High-tech industries: • Publishing –> Software & Publishing • Broadcasting –> Telecommunications & Radio/TV/Cable Broadcasting • Information –> Information Services & Data Processing Services Nonprofits & Owner Occupied Housing: • Real Estate • Health Care • Other Services 14 Sources of growth in sectoral output for major and "intermediate" sectors of the U.S. economy1 Sectoral Output (1) C. Difference in Annual Averages, (1995 to 2000) vs. (1987 to 1995) 1. Private nonfarm business 2. Excl. high-tech 3. Construction and real estate 4. Industrial 5. Distribution 6. Finance and business 7. Personal and cultural 8. High-tech IT Capital (3) Other Capital (4) Purchased Labor Inputs (5) (6) MFP (2) 2.3 2.0 2.4 .8 1.2 4.0 .8 8.0 .3 -.1 -.2 -.4 .8 -.3 .1 2.7 .5 .4 .1 .1 .3 .8 .1 .9 .1 .1 -.2 .1 .2 .2 .0 .0 .7 .5 .9 .0 .2 1.0 -.2 2.1 .6 .9 1.7 1.0 -.3 2.3 .8 2.4 • Each row is a decomposition of the change in growth during the late-1990s: • Acceleration in aggregate MFP driven by high-tech and distribution • Pickup in IT capital services driven by finance and business services and distribution 15 Sources of growth in sectoral output for major and "intermediate" sectors of the U.S. economy1 Sectoral Output (1) D. Difference in Annual Averages, (2000 to 2005) vs. (1995 to 2000) 1. Private nonfarm business 2. Excl. high-tech 4. Construction and real estate 3. Industrial 5. Distribution 6. Finance and business 7. Personal and cultural 8. High-tech IT Capital (3) Other Capital (4) Purchased Inputs (6) MFP (2) Labor (5) -3.0 -2.6 -.9 -2.2 -2.6 -4.5 -1.4 -13.7 1.3 1.5 1.8 1.1 -.3 2.1 .9 -1.0 -.6 -.5 -.1 -.2 -.3 -.9 -.2 -1.2 -.5 -.4 -.5 -.2 -.3 -.7 -.1 -.4 -2.3 -1.8 -1.2 -1.2 -1.2 -2.4 -.5 -4.8 -1.0 -1.4 -1.0 -1.7 -.5 -2.6 -1.6 -6.3 •Each row is a decomposition of the change in growth during 2000-2005: • Faster aggregate MFP has been the major driver •High-tech and distribution are no longer the main drivers. •MFP accelerated in finance & business services, the industrial sector, and personal services 16 Key Findings By 2005, the productivity resurgence that began in mid-1990s was relatively broad based Late 1990s: U.S. productivity resurgence was a sectoral story • Increases for high tech and distribution partly offset by stepdowns elsewhere Since 2000: MFP the major contributor to economic growth • Strong MFP gains in the finance & business services • MFP growth for players in late-1990s remained robust 17 Extension: Modeling Aggregate Trend Productivity • Six sectors clearly have very different trends 3 Core Inflation: Trend MFP increases 3/4 ppt • Separate accounting and forecasting may intuition behind estimate of current trend in aggregate MFP • Sectoral trends can be linked to other elements of staff projections 2.5 2 1.5 1 0.5 0 4Q98 1Q99 2Q99 3Q99 MFP Increase 4Q99 1Q00 2Q00 3Q00 4Q00 MFP Decrease MFP Base 18 19 Extension: The Role of Imported Intermediates Total Purchased Inputs (Mk) Own Industry (Xk) Other Industries (Rk + Nk) Foreign (Rk) Domestic (Nk) • Imported intermediates derived from detailed measures obtained from the BEA • Price and quantity measures from other domestic industries derived residually • In progress: combine with BEA’s industry level KLEMS data 20 Table 3 Sources of growth in sectoral output for major and "intermediate" sectors of the U.S. economy1 Sectoral Output (1) D. Difference in Annual Averages, (2003 to 2005) vs. (1997 to 2002) 1. Private nonfarm business 2. Excl. high-tech 3. Construction 4. Industrial 5. Distribution 6. Finance and business 7. Personal and cultural 8. High-tech IT Capital2 (3) Other Capital3 (4) Purchased Purchased Domestic Inputs Foreign Inputs (6) (7) MFP (2) Labor (5) -.6 -.4 1.0 .7 -.2 .0 -.2 -3.0 1.6 1.5 .6 1.2 -.5 2.6 1.1 2.1 -.6 -.4 -.1 -.2 -.3 -.7 -.2 -1.3 -.4 -.3 -.1 -.1 -.2 -.6 -.1 -.4 -.5 -.3 -.4 .0 -.2 -.4 -.2 -2.1 -1.1 -1.1 .7 -.6 .7 -.9 -.7 -1.7 .3 .3 .3 .5 .1 .0 .1 .4 Each row is a decomposition of the change in growth during 2003-2005: • While import contribution has accelerated in in most sectors • Domestic inputs spurred growth in construction and distribution but subtracted from it elsewhere • Industrial sector growth due entirely to faster MFP and increased reliance on foreign inputs 21 Concluding Remarks Annual Industry Accounts are an essential input to FRB work on productivity • Sufficiently detailed to provide a rich—and highly customizable—assessment of recent trends Measurement: we know the least about the sectors with the best productivity performance • High-tech – Necessary to break out high-tech industries to work with the industry accounts. • Finance and Business Services – How reliable are the data for this sector? – Proposed Improvements to AIAs would really help (broader SAS coverage; expanded PPIs) 22 Concluding Remarks (cont’d) Industry classification differences: • Analysis compromised when industry output and inputs classified differently • Proposed reconciliation effort is important Flexibility: • To group industry-level results into vertically-integrated sectors further disaggregation of BEA industry data was necessary (high tech, nonprofits) • Plans to publish VA for more detailed industries would help a lot 23