Bank Of Greece Sa · ATHEX:TELL
Bank Of Greece Sa — Technical Presentation (September 2022)
43 slides from Bank Of Greece Sa's technical presentation, September 2022
Slide 1: What Are the Labor and Product Market Effects of Automation? New Evidence from France
What Are the Labor and Product Market Effects of Automation? New Evidence from France Philippe Aghion, College de France Celine Antonin, OFCE Simon Bunel, Banque de France and PSE Xavier Jaravel, London School of Economics Workshop ESCB Research Cluster 2 September 2022 Any opinions and conclusions
Slide 2: Motivation
Motivation Why did major technological revolutions (steam engine, electricity) not generate mass unemployment anticipated by some? I Luddites in 1811-1812 I Keynes (1930): technological unemployment Policy implications: robot tax (Bill Gates, Andrew Yang, Benoît Hamon) Tradeoff: displacement vs. pro
Slide 3: This paper: Main Findings
This paper: Main Findings Consistent with productivity effect: I Automation ↑, marginal cost and prices ↓, demand ↑, employment ↑ Data: French manufacturing industry between 1995 and 2017 Estimates indicate that: I At firm-level, a 1% increase in automation at year t leads to a 0.2% increase in empl
Slide 4: Literature
Literature Labor Market / Industry-level studies find mixed results I Industrial robots: Acemoglu and Restrepo 2019, Michaels and Graetz 2018, Dauth et al. 2021 I Automation (patents): Dechezleprêtre et al. 2021, Mann and Puttmann 2020 Recent/ongoing work studies robots at firm level I Acemoglu et a
Slide 5: Data: Worker/Firm Data
Data: Worker/Firm Data Detailed information on workers and firms available from French administrative data I Matched employer-employee data (DADS) and balance sheet data (Ficus/Fare) covering all firms in French manufacturing industry from 1995 to 2017 I Firms: employment, sales, industry, etc. I Wo
Slide 6: Data: Measuring Automation
Data: Measuring Automation First measure: Balance sheet value of industrial machines I "All machines used for extraction, processing, shaping, packaging of materials or supplies" I Stock of industrial machines at firm level I Distinguishes between (i) industrial machines, (ii) land, (iii) buildings
Slide 7: Data: Measuring Automation
Data: Measuring Automation Second measure: Acemoglu-Restrepo (2022)’s automation measure I Defined as a "range of technologies that relate to industrial automation" I Based on imported intermediate goods, defined as products whose two-digit HS code correspond to : F Industrial robots F Dedicated mac
Slide 8: Examples
Examples (a) Chemicals (b) Paper (c) Food 7 / 23
Slide 9: Data: Measuring Automation
Data: Measuring Automation Advantages: I Covers broader set of automation technologies than IFR definition of an industrial robot, i.e. an "automatically controlled, reprogrammable multipurpose manipulator programmable in three or more axes" (ISO 8373) 50 20 40 Sectoral share, in euros (%) Sectoral
Slide 10: Event studies
Event studies Question: when a firm relies more extensively on automation, what happens to employment, prices and sales? Implementation of event studies: I Event defined as a major investment in automation technologies a given year 9 / 23
Slide 11: Event studies: Specification
Event studies: Specification n log Lit = α + ∑ βk 1t−Ei =k 1Investi >pX (Invest) + µi + λst + εit k=−n with employment Lit , firm F.E. µi and industry-year F.E. λst (mitigate potential correlated shocks) Specification allows for delayed response of employment to increased automation Pre-trends (lead
Slide 12: Event studies: Employment
Event studies: Employment Total Employment 1 .8 Estimated Semi-Elasticity -.2 0 .2 -.4 -.6 .4 .6 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Year relative to change Treated = Top 10% - Controlling for 5-digit-industry by year F.E. + Firm F.E. 11 / 23
Slide 13: Event studies: Employment - AR's automation measure
Event studies: Employment - AR’s automation measure 12 / 23
Slide 14: Event studies: Prices
Event studies: Prices Export Price .2 .1 Estimated Semi-Elasticity -.2 -.1 -.3 0 -4 -3 -2 -1 0 1 2 3 4 Year relative to change Treated = Top 10% - Controlling for CN8-Product by year F.E. + Country by year F.E. + Firm F.E. 13 / 23
Slide 15: Event studies: Sales
Event studies: Sales Total sales 1 .8 Estimated Semi-Elasticity -.2 0 .2 -.4 -.6 .4 .6 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Year relative to change Treated = Top 25% - Controlling for 5-digit-industry by year F.E. + Firm F.E. 14 / 23
Slide 16: Shift-Share IV
Shift-Share IV Limitation: event studies cannot rule out potential demand/supply shocks in contemporaneous period Ideal experiment would randomly assign purchasing prices for machines across firms Approximate with a shift-share research design, leveraging two components: 1 Variation in the cost of i
Slide 17: Shift-Share IV: Shocks and Exposure Shares
Shift-Share IV: Shocks and Exposure Shares Shocks are observed across trading partners by HS6 products: I We cannot directly observe changes in foreign machines’ quality-adjusted prices I gn,t is change in imports flows of machines from each trading partners (Germany, Italy, China, etc.) for each HS
Slide 18: Firm SSIV: First Stage
Firm SSIV: First Stage First Stage .06 5-year Log Change in Industrial Equipment for Linked Firms 0 .02 -.02 .04 -.4 -.2 0 .2 .4 Instrument (foreign supplier productivity shock) 17 / 23
Slide 19: Firm SSIV: Reduced Form
Firm SSIV: Reduced Form Reduced Form - Employment Reduced Form - Sales .01 -.06 Employment for Linked Firms -.065 0 Sales for Linked Firms 5-year Log Change in 5-year Log Change in -.07 -.01 -.075 -.02 -.08 -.085 -.03 -.4 -.2 0 .2 .4 -.4 -.2 0 .2 .4 Instrument (foreign supplier productivity shock) I
Slide 20: Firm SSIV: Employment
Firm SSIV: Employment ∆5 Employment (1) (2) (3) (4) (5) ∆5 Machines 0.426∗∗∗ 0.425∗∗∗ 0.424∗∗∗ 0.433∗∗∗ 0.433∗∗∗ (0.084) (0.100) (0.100) (0.098) (0.098) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X X X X X 2-digit Industry-period F.E. X X X
Slide 21: Firm SSIV: Sales
Firm SSIV: Sales ∆5 Sales (1) (2) (3) (4) (5) ∆5 Machines 0.325∗∗∗ 0.340∗∗∗ 0.340∗∗∗ 0.345∗∗∗ 0.346∗∗∗ (0.131) (0.123) (0.121) (0.114) (0.103) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X X X X X 2-digit Industry-period F.E. X X X X X Lagged
Slide 22: Evidence of business stealing
Evidence of business stealing 21 / 23
Slide 23: Industry-Level Employment and International Competition
Industry-Level Employment and International Competition ∆ Employment 1996-2017 International Competition All Industries Above Median Below Median (1) (2) (3) ∆ Machines 1996-2017 0.345∗∗∗ 0.404∗∗∗ 0.171 (0.059) (0.055) (0.133) 2-digit industry by year F.E. X X X ∆ Other types of capital 1996-2017 X
Slide 24: Conclusion
Conclusion Automation increases employment of firms that automate, which indicates that in practice the productivity effect of automation on employment tends to outweigh the displacement effect Automation also increases sales and profits, and reduces prices I Hence overall automation generates gains
Slide 25: Thank you!
Thank you! simon.bunel@banque-france.fr 24 / 23
Slide 26: Event studies: Employment - Heterogeneity?
Event studies: Employment - Heterogeneity? 25 / 23
Slide 27: Event studies: Wage
Event studies: Wage Mean Hourly Wage - p75 1 ° 1 81 1 \ 3 i 2 \ g 1 es cy \ Eo 5 1 a 1 Bg ! a a 381 1 ao 1 8 I El 7 1 \ 2 1 al Ll 8 7 6 5 4 3-2-1012 3 4 «5 6 7 8 Year relative to change Controlling for 4-digit year FE. + Firm FE. - Mean Hourly Wage - Ratio p90/p10 - p75 Mean Hourly Wage - Ratio Indu
Slide 28: Event studies: Job Creation & Job Destruction
Event studies: Job Creation & Job Destruction 27 / 23
Slide 29: Event studies: Job Creation & Job Destruction Placebo Test with Investments in Real Estate
Event studies: Job Creation & Job Destruction Placebo Test with Investments in Real Estate 28 / 23
Slide 30: Industry Level Employment
Industry Level Employment (a) Extensive margin (b) Intensive margin 29 / 23
Slide 31: Industry Level Sales
Industry Level Sales Total sales i} I I I I I © I I I I I I \ Estimated Elasticity 2 i -4 1 -6 T T T T T T T T T T T T T 6 5 4 3 2 -1 0 1 2 3 4 5 6 Year relative to change Controlling for year F.E. + 5-digit-industry F.E. + Other types of capital - 2630 observations 30 / 23
Slide 32: Industry Level Wages
Industry Level Wages Manufacturing Mean Hourly Wage - p50 | .06 .04 02 ae ann eaanee 0 b——1 -.02 Estimated Semi-Elasticity -.04 -.06 -.08 T T T T T T T T T 6 56 4 -3 -2 -1 0 1 2 3 4 5 6 Year relative to change Controlling for year F.E. + 5-digit Industry F.E. - 1716 observations 31/23
Slide 33: Industry Level Wage Inequality
Industry Level Wage Inequality 32 / 23
Slide 34: Shift-Share IV: Exposure Shares
Shift-Share IV: Exposure Shares Exposure shares of French firms: I sin is share of trading partner n in firm i’s total imports of machines in initial period (between 1995 and 1999) I Contemporaneous shares liable to reverse causality: use initial shares instead (and analyze outcomes from 2000 onward
Slide 35: Shift-Share IV: Identification Assumptions
Shift-Share IV: Identification Assumptions Relevance: need supplier relationships to be sufficiently persistent I Check power with first-stage F-statistic I Standard errors clustered by trading partner Exclusion restriction: firms linked to increasingly productive suppliers should not be unobservabl
Slide 36: Firm SSIV: Profits
Firm SSIV: Profits ∆5 Profits (1) (2) (3) (4) (5) ∆5 Machines 0.995∗∗ 0.824∗ 0.824∗ 0.827∗ 0.828∗∗ (0.448) (0.432) (0.432) (0.424) (0.412) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X X X X X 2-digit Industry-period F.E. X X X X X Lagged Fir
Slide 37: Firm SSIV: Labor Cost / Sales
Firm SSIV: Labor Cost / Sales ∆5 Labor Cost / Sales (1) (2) (3) (4) (5) ∆5 Machines 0.00453 0.00604 0.00607 0.00697 0.00686 (0.0164) (0.0173) (0.0172) (0.0166) (0.0157) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X X X X X 2-digit Industry-pe
Slide 38: Firm SSIV: Reduced Form - Falsification tests
Firm SSIV: Reduced Form - Falsification tests Reduced Form - Lagged Employment Reduced Form - Lagged Sales .06 .14 Lagged Employment for Linked Firms Lagged Sales for Linked Firms .13 .05 5-year Log Change in 5-year Log Change in .12 .04 .11 .03 .02 .1 -.4 -.2 0 .2 .4 -.4 -.2 0 .2 .4 Instrument (for
Slide 39: Falsification Test: Lagged Firm Employment
Falsification Test: Lagged Firm Employment Lagged ∆5 Employment (1) (2) (3) (4) (5) ∆5 Machines −0.180 −0.198 −0.199 −0.199 −0.200 (0.219) (0.220) (0.223) (0.220) (0.218) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X X X X X 2-digit Industry-
Slide 40: Firm SSIV: Competitors' Employment
Firm SSIV: Competitors’ Employment ∆5 Competitors’ Employment (1) (2) (3) (4) (5) ∆5 Machines −0.00578∗ −0.00920∗∗∗ −0.00920∗∗∗ −0.00914∗∗∗ −0.00913∗∗∗ (0.0032) (0.0033) (0.0032) (0.0033) (0.0033) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X
Slide 41: Falsification Test: Lagged Firm Sales
Falsification Test: Lagged Firm Sales Lagged ∆5 Sales (1) (2) (3) (4) (5) ∆5 Machines 0.0274 0.166 0.165 0.155 0.155 (0.202) (0.209) (0.218) (0.214) (0.211) First-Stage F 17.65 20.59 21.43 20.88 21.62 Partner-period F.E. X X X X X 4-digit Product-period F.E. X X X X X 2-digit Industry-period F.E. X
Slide 42: Industry Level SSIV: Employment
Industry Level SSIV: Employment ∆5 Employment (1) (2) (3) (4) ∆5 Machines 1.080∗∗∗ 1.076∗∗∗ 1.081∗∗∗ 1.091∗∗∗ (0.185) (0.186) (0.190) (0.193) First-Stage F 17.98 18.03 15.53 15.53 Partner-period F.E. X X X X 4-digit Product F.E. X X X X Lagged Industry Controls X X X X Lagged Machines X X X Lagged O
Slide 43: Industry Level SSIV: Sales
Industry Level SSIV: Sales ∆5 Sales (1) (2) (3) (4) ∆5 Machines 1.309∗∗∗ 1.312∗∗∗ 1.245∗∗∗ 1.207∗∗∗ (0.338) (0.338) (0.337) (0.327) First-Stage F 17.98 18.03 15.53 15.53 Partner-period F.E. X X X X 4-digit Product F.E. X X X X Lagged Industry Controls X X X X Lagged Machines X X X Lagged Other Capit
Bank Of Greece — Technical Presentation (September 2022) on 6ix