Improving Measures of GDP and TFP

Transcription

1 Improving Measures of GDP and TFP Intangibles and Other Measurement Issues Jonathan Haskel, Imperial College Business School, CEP, CEPR and IZA How can we improve the way we measure GDP in ocial statistics? RSS/University of Manchester Technical workshop. Manchester, 27 February 2018 Haskel Improving Measures of GDP and TFP 1 / 13

2 Background Global labour and MFP productivity slowdown Need a framework to judge this and see what will happen in future e.g. AI This presentation Model of TFP slowdowns and mismeasurement, focus on intangibles Some data Haskel Improving Measures of GDP and TFP 2 / 13

3 GDP and growth accounting when intangibles are intermediates Upstream intangibles-producing N-sector: Upstream tangibles-producing I-sector: P N N = P X X N ; dn = σ N X dx N + da N (1) P I I = P X X I ; di = σ I X dx i + da I (2) Downstream tangible and intangibles-using C-sector, intangibles are intermediates: P C V C P C C P N N; dc = σ C X dx C + σ C Ndn C + da C (3) Economy-wide value added = sum of industry value-added P V V P C V C + P I I + P N N (4) = P V V GDP = P C C consump + P I I ; dv = σx V dx + da V invest output =share*input + TFP Haskel Improving Measures of GDP and TFP 3 / 13

4 GDP and growth accounting when intangibles are capital Upstream intangibles-producing N-sector: P N N = P X X N ; dn = σx N dx N + da N (5) Upstream tangibles-producing I-sector: P I I = P X X I ; di = σx I dx i + da I (6) Downstream tangible and intangibles-using C-sector, intangibles are capital: P C V C P C C; dc = σx C dx C + σr C dr C + da C (7) Economy-wide value added = sum of industry value-added P Q Q P C V C + P I I + P N N = P Q Q GDP = P C C consump + P I I tan invest + P }{{ N N } ; dq = σ Q X dx + σ Q R dr + da Q intan invest output =share* input X +share* input R+ TFP (8) Haskel Improving Measures of GDP and TFP 4 / 13

5 Comparison Intan intermediates P V V GDP Intan capital P Q Q GDP = P C C + P I I consump invest ; dv = σx V dx + dav = P C C + P I I + P N N consump tan invest intan invest output =share*input + TFP ; dq = σ Q X dx + σq R dr + daq output =share* input X +share* input R+ TFP Missing GDP levels; missing output and inputs from growth accounting Example: Tesco builds new shop and uses Big Data to predict product range to sell in it. New shop is in PiI, investment in new data is PnN. Assets are tangible in X, intangible in R. Example: Bank develops new app for internet banking. Does it show up in prices, quantities? Hints that even if count new product development in N or R, we need to measure price/output. Haskel Improving Measures of GDP and TFP 5 / 13

6 Output and input measurement problems Double deation. True value added growth is gross output less share-weighted intermediates But lack of double deation means measured variables are: where intermediates are mismeasured dv C = dg s z dz (9) dv C m = dg s z dz m (10) dz m = dz + m dz (11) Partial output quality-adjustment. Quality-adjustment implicit and explicit Partial input quality-adjustment dc m = dvm C + νdθ (12) dx m = dx + m dx (13) e.g. unobserved scrapping after a recession: dm X =K > 0. Non-quality adjusted ICT deators dm X =K(ICT ) < 0 Haskel Improving Measures of GDP and TFP 6 / 13

7 TFP Restatement of system (omit I sector for ease of notation) All sectors, total output dn = σ N X (dxn m m N dx ) + dan (14) dc m = σ C X (dxc m m C dx ) + σc R dr C intangibles Measured TFP, no intangibles Relation measured TFP to da + νdθ q adj s z m dz + da C doub de tech,spillovers dq ω Q C dcm + ωq N dn (15) dc m + ω Q N (dn dcm) dt m = dc m σ C X dxc m (16) dtm NoIntan = da ω Q N (dn dcm) + σ Q R dr C tech,spillovers missing intan invest intangibles + (σ Q X σc X }{{ )dxm } σ Q X m dx K, L share mismeas K, L mismeas + νdθ q adj s z m dz doub de Summary: measured TFP slows down if (a) intangible investment is growing faster than value added (b) dr slows down (c) dr slows down & da = γdr Haskel Improving Measures of GDP and TFP 7 / 13 (17)

8 Intangibles and tangible investment Haskel Improving Measures of GDP and TFP 8 / 13

9 Intangibles and intangible capital services Share of intangible-adjusted GDP Year Note: whole economy Tangible and intangible investment (Share of intangible-adjusted GDP, US+EU11) Intangible share Tangible share Annual growth rate Intangible capital services growth Non-farm business, US+EU Year Contrast investment tangibles and intangibles post-2007 Note slowdown in intangible capital services growth Haskel Improving Measures of GDP and TFP 9 / 13

10 Intangibles and spillovers? Multifactor productivity growth IT MFP and intangible capital services growth ( open diamonds; closed circles) DK FI DE ES UK US SE UK DE DK SENL FR IT FI AT NL FR USAT Intangible capital services growth Note: whole economy. MFP includes intangibles, but not labour composition ES Haskel Improving Measures of GDP and TFP 10 / 13

11 Mismeasurement dtm NoIntan = da C + B σ Q X m dx tech,spillovers bias due to missing intang K, L mismeas B = ω Q N (dn dcm) + σ Q R dr C + (σ Q X σc X }{{ )dxm } missing intan invest intangibles K, L share mismeas + νdθ We calculate B i.e. the bias due to the non-inclusion of non-na intangibles q adj s z m dz doub de (18) Bias to measured MFP Mismeasurement and intangible investment (Non-farm business sector, US+EU11) Year Bias to measured TFP Intangible share of GDP Intan invest/gdp So TFP slowdown is not quite as bad as we thought: but eect is small Haskel Improving Measures of GDP and TFP 11 / 13

12 Mismeasurement, UK and US UK US DlnTFP (NA) 2.29% 1.68% DlnTFP (all) 2.20% 1.47% Bias 0.09% 0.22% DlnTFP (NA) -0.82% 0.52% DlnTFP (all) -1.03% 0.32% Bias 0.22% 0.20% DlnTFP (NA) 0.42% 0.46% DlnTFP (all) 0.43% 0.45% Bias -0.02% 0.00% Slowdown v DlnTFP (NA) -1.87% -1.23% DlnTFP (all) -1.77% -1.02% Haskel Improving Measures of GDP and TFP 12 / 13

13 Back to framework Recap: So why might measured TFP slow down? dtm NoIntan = da C ω Q N (dn dcm) + σ Q R dr C tech,spillovers missing intan invest intangibles + (σ Q X σc X }{{ )dxm } σ Q X m dx K, L share mismeas K, L mismeas + νdθ q adj s z m dz doub de da slows because technology over. Gordon. Human capital growth slowing? Intang (see above) Mismeasured capital. UK follows US on computer deators. Was pre-recession comms capital growth too small m dx=dk < 0 so dt m too high, so TFP slowdown less? Double deation. UK does not double deate. China eect: we missed rise of cheap intermediates before recession so m dz < 0, so dt m too high. So TFP slowdown less. Important for manufacturing? Quality mismeasurement. Eect gets worse as service sector/free goods gets larger? AI, robots and Big Data. New wave of unmeasured tangible and intangible investment? New problems for quality-adjustment. Innovation in upstream sector (Cockburn, Henderson, Stern)? Innovation opportunities in health (Mandel,Swanson)? Overall assessment Measurement problems likely getting worse Measured TFP growth slowdown does not indicate slowdown in a If da driven by e.g. spillovers: da is a policy choice. Tax policy. Banking policy. Science Haskel Improving Measures of GDP and TFP 13 / 13 (19)