Structural Change within the Service Sector and the Future of Baumol s Disease

Transcription

1 Structural Change within the Service Sector and the Future of Baumol s Disease Georg Duernecker (University of Munich, CEPR and IZA) Berthold Herrendorf (Arizona State University) Ákos Valentinyi (University of Manchester, CEPR, and CERS HAS) December 4, 2017 Duernecker, Herrendorf, Valentinyi

2 1. Motivation Baumol s Disease: structural change reduces productivity growth Structural change causes reallocation to service sector. Average productivity growth is lower in service sector than goods sector. Many service industries have close to zero productivity growth. Does the service industry with the lowest productivity growth take over? This happens if there is little substitutability between different industries, which is often assumed in the literature on structural change. Duernecker, Herrendorf, Valentinyi 1

3 Our contribution We calculate the productivity effect of Baumol s disease in the postwar U.S. More than 1/3 of overall productivity growth slowdown. We capture the productivity effects of Baumol s disease with a model that disaggregates services into those with low and high productivity growth; uses non homothetic utility function with persistent income effects. We predict that the future effect of Baumol s disease will be smaller than in the past There is substitutability within the service sector. The service industries with low productivity growth won t take over. Duernecker, Herrendorf, Valentinyi 2

4 Related debate about the slowdown in productivity growth Big innovations have been made (sanitation vs. iphone, Gordon, 2016). Finite time endowments limit innovation and human capital accumulation (Fernald and Jones, AER, 2015) Secular stagnation Evidence on Slowdown Duernecker, Herrendorf, Valentinyi 3

5 2. Evidence on Baumol s Disease Data U.S. economy after World War II from WORLD KLEMS. Value added, prices, and efficiency units of labor ( quality adjusted hours). 65 industries. Duernecker, Herrendorf, Valentinyi 4

6 Quantifying the effects of Baumol s disease What would productivity growth have been without structural change? Answering this question involves two steps: fix the sectoral shares in 1947 or 2007; apply the actual annual growth rates of sectoral labor productivity. Having efficiency units is crucial for doing these counterfactual. Otherwise sector productivity differences include human capital differences. Reallocating a worker then implies that he gets human capital of new sector. Duernecker, Herrendorf, Valentinyi 5

7 Approach 1: average annual productivity growth in initial and final 20 year period Sector composition from Baumol s data 1947 disease % 2.45% % 1.87% difference Duernecker, Herrendorf, Valentinyi 6

8 Approach 2: average annual productivity growth Sector composition from Baumol s data disease 1.74% 2.03% 1.44% 0.59 Duernecker, Herrendorf, Valentinyi 7

9 Discussion Given that the productivity growth slowdown is a hot topic, why doesn t the structural change literature pay attention to Baumol s disease? Literature imposes BGP that rules out Baumol s disease from the start characterizing the equilibrium path requires having a BGP; obtaining a BGP requires measuring GDP growth in peculiar way (new numeraire in each period). In contrast, we measure GDP growth as it is done in the data. In a companion paper, Quantity Measurement and Balanced Growth in Multi Sector Models, we show that then structural change naturally causes Baumol s disease. Duernecker, Herrendorf, Valentinyi 8

10 Towards a model of Baumol s disease How much of the effects of Baumol s disease do different sector splits capture? all 65 industries goods & services 26 goods & services goods & 39 services 1947 composition 2.03% 1.88% 1.99% 1.92% 2014 composition 1.44% 1.58% 1.56% 1.46% difference Duernecker, Herrendorf, Valentinyi 9

11 A new two sector split for services Service Industries with Above Average Productivity Growth Annual Growth Rate Pipeline Transportation 5.6%... Broadcasting and Telecommunications 4.4%... Wholesale Trade 3.1%... Securities, Commodity Contracts, and Investments 2.8%... Insurance Carriers and Related Activities 1.6% Warehousing and Storage 1.5% Service Industries with Below Average Productivity Growth Management of Companies and Enterprises 1.3%... Educational Services 0.8%... Ambulatory Health Care Services 0.6%... State and Local General Government 0.4%... Food Services and Drinking Places 0.5% Transit and Ground Passenger Transportation 0.7% Duernecker, Herrendorf, Valentinyi 10

12 The new sector split captures most of the effects of Baumol s disease goods & goods & goods & goods & 39 services high/low services non/market services un/skilled services 1947 composition 1.92% 1.97% 1.91% 1.96% 2014 composition 1.46% 1.49% 1.53% 1.48% difference Definition Different Splits Duernecker, Herrendorf, Valentinyi 11

13 Relative Prices and Expenditures in Services 3.0 relative prices 1947= Value added of services with low relative to high productivity growth Price of services with low relative to high productivity growth Source: WORLD KLEMS, March 2017 release and own calculations. Duernecker, Herrendorf, Valentinyi 12

14 3. Model Goal: make it as simple and standard as possible 3 sectors producing goods and services with high and with low productivity growth. Sequence of static economies without capital. Exogenous growth in efficiency units of labor and sectoral productivity = income growth and price changes. Non homothetic CES utility = reallocation of production among sectors. Duernecker, Herrendorf, Valentinyi 13

15 Endowments and production H t efficiency units of labor that are inelastically supplied. Efficiency units can be used in all sectors to produce value added: C it = A it H it, i {g, l, h} Duernecker, Herrendorf, Valentinyi 14

16 Nested non homothetic CES utility C t = α gc εg 1 σc t σ c 1 σ C c gt εs 1 σc + α s C t C σ c 1 σ c st σ c σ c 1 C st = α lc ε l 1 σ s t σ s 1 σ C s lt ε h 1 σ + α h C s t C σ s 1 σ s ht σ s σ s 1 Evidence for non homothetic CES Duernecker, Herrendorf, Valentinyi 15

17 Key feature 1: Reduces to nested homothetic CES for ε i = 1 C t = α gc σ c 1 σ c gt ) σ c 1 σ c σ c 1 + α s C σ st c C st = α lc σ s 1 σ s lt + α h C σ s 1 σ s ht σ s σ s 1 Duernecker, Herrendorf, Valentinyi 16

18 Key feature 2: Long run income effects for ε i 1 C t = α gc εg 1 σc t σ c 1 σ C c gt εs 1 σc + α s Ct C σ c 1 σ c st σ c σ c 1 C st = α lc ε l 1 σ s t σ s 1 σ C s lt ε h 1 σ + α h C s t C σ s 1 σ s ht σ s σ s 1 Duernecker, Herrendorf, Valentinyi 17

19 4. Equilibrium Firm problems Implied equilibrium relationships max H it P it A it H it (1 + τ it )W t H it, i { g, l, h } P it Y it /H it P gt Y gt /H gt = 1 + τ it 1 + τ gt Choosing relative taxes allows us to match relative nominal labor productivities. Duernecker, Herrendorf, Valentinyi 18

20 Household problem Outer layer min C gt,c st P gt C gt + P st C st s.t. α gc ε g 1 σ c t σ c 1 σ C c gt ε s 1 σ + α s C c t C σ c 1 σ c st σ c σ c 1 C t where P st is the price index from the inner layer. Inner layer min C lt,c ht P lt C lt + P ht C ht s.t. α lc ε l 1 σ s t σ s 1 σ C s lt ε h 1 σ + α h C s t C σ s 1 σ s ht σ s σ s 1 C st Duernecker, Herrendorf, Valentinyi 19

21 5. Analytical Results Structural change between goods and services P st C st P gt C gt = α s α g ( Pst P gt ) 1 σc C ε s ε g t Assumptions Both C t and P st /P gt grow (true for the calibrated model). Goods and services complements: 0 < σ c < 1. Goods necessities and services luxuries: ε g < ε s. Bottom line: Business as usual between goods and services. Duernecker, Herrendorf, Valentinyi 20

22 Structural change within services P lt C lt P ht C ht = α l α h ( Plt P ht ) 1 σs C ε l ε h t Assumptions The two services are substitutes: σ s > 1 Service with high (low) productivity growth are necessities (luxuries): ε l > ε h Bottom line: Not clear which service sector, if any, takes over in the limit Duernecker, Herrendorf, Valentinyi 21

23 6. Quantitative Analysis Basic approach Calibrate model to match salient features of U.S. postwar economy (patterns of real and nominal labor productivities and of sectoral composition). Assume past growth rates of TFP, taxes, and efficiency units continue into future. Simulate model to predict by how much productivity growth will slow down in the future. Duernecker, Herrendorf, Valentinyi 22

24 Calibrated Parameters α g α h σ c σ s ε s ε g ε l ε h Key features Complementarity between goods and services: σ c < 1 Substitutability between services: σ s > 1 Services luxuries and goods necessities: ε s ε g > 0 Services with low (high) productivity growth luxuries (necessities): ε l ε h > 0 Calibration Details Duernecker, Herrendorf, Valentinyi 23

25 Findings about future productivity growth slowdown Productivity Growth (in %) Actual (model) (2.33) 1.45 (1.48) 1.74 (1.75) estimation period for Predicted exogenous variables , % A l = Duernecker, Herrendorf, Valentinyi 24

26 Why was past growth slowdown relatively large? Income and substitution effects: strong reallocation from C gt to C st C gt have high and C st low productivity growth Why is the future growth slowdown smaller? Income effect: shares of C st and C lt increase Substitution effect: share of C lt decreases Net effect: share of C lt remains roughly constant Further Findings Duernecker, Herrendorf, Valentinyi 25

27 Conclusions Our results Baumol s disease considerably reduced labor productivity growth in the postwar US. Nonetheless our model predicts that it will reduce it by less than half in the future. Services with slow productivity growth do not take over consumption in the limit. Our next steps are to endogenize sectoral productivity growth (Herrendorf and Valentinyi 2015); structural change within investment (Herrendorf, Rogerson, and Valentinyi 2017); changes in input output relationships. Duernecker, Herrendorf, Valentinyi 26

28 Appendix Duernecker, Herrendorf, Valentinyi 27

29 The slowdown in U.S. productivity growth Growth of real U.S. GDP per hour average of the previous 20 years Source: Total Economy Database, May 2017 Duernecker, Herrendorf, Valentinyi 28

30 The slowdown in productivity growth in other countries Growth of real GDP per hour average of the previous 20 years Australia Canada United Kingdom United States Source: Total Economy Database, May 2017 Back Duernecker, Herrendorf, Valentinyi 29

31 Different ways of splitting services Back 1 = High Growth Market Unskilled 2 = Low Growth Non Market Skilled Post and Telecom Wholesale Sale... of Motor Vehicles Retail Financial Transport Business Services Education Real Estate Health Private Households Personal Services Hotels Duernecker, Herrendorf, Valentinyi 30

32 Stylized facts of structural transformation Goods vs Services Relative nominal value added and efficiency units Relative productivities and prices relative prices 1947= Value added of services relative to goods Efficiency units of services relative to goods Price of services relative to goods Real value added per efficiency unit of goods relative to services Source: WORLD KLEMS, March 2017 release and own calculations. Source: WORLD KLEMS, March 2017 release and own calculations. Duernecker, Herrendorf, Valentinyi 31

33 Services with high vs low productivity growth Relative nominal value added and efficiency units Relative productivities and prices relative prices 1947= Value added of services with low relative to high productivity growth Efficiency units of services with low relative to high productivity growth Price of services with low relative to high productivity growth Real value added per efficiency unit of services with high relative to low productivity growth Source: WORLD KLEMS, March 2017 release and own calculations. Source: WORLD KLEMS, March 2017 release and own calculations. Duernecker, Herrendorf, Valentinyi 32

34 Persistent income effects within services Back Residuals of log difference of nominal value added of different services Partial correlation between relative value added within services and aggregate value added per efficiency unit Residuals of log of aggregate real value added per efficiency unit Residuals Linear fit Source: WORLD KLEMS, March 2017 release and own calculations Note: Residuals on the y-axis are from regressing the log difference of nominal value added of services with low and high productivity growth on the corresponding log difference of prices. Residuals on the x-axis are from regressing the log of aggregate real value added per efficiency unit on the same log differences of prices. Duernecker, Herrendorf, Valentinyi 33

35 Calibration Step 1: Calibrate { τ gt, τ ht, τ lt } 2014 t=1947 Set τ gt = 0 and compute { τ ht, τ lt } by using relative nominal productivities: 1 + τ jt = ṼA jt/ L jt ṼA gt / L gt Duernecker, Herrendorf, Valentinyi 34

36 Step 2: Calibrate { A gt, A ht, A lt } 2014 t=1947 Normalize A g1947 = 1 and obtain { A gt } 2014 t=1948 Obtain { A ht, A lt } 2014 t=1947 by using real productivity growth: A gt = A gt 1 ṼA gt /( P gt L gt ) ṼA gt 1 /( P gt 1 L gt 1 ) by using relative real productivities: A jt = (1 + τ jt )A gt P gt P jt Duernecker, Herrendorf, Valentinyi 35

37 Step 3: Choose ε g = 0.85 and ε h = 0.58, α s = 1 α g, and α l = 1 α h. Step 4: Calibrate {α g, α h, σ s, σ c, ε s, ε l } Match nominal value added shares { ṼAlt ṼA gt, ṼA lt } ṼA ht t=1947,...,2014 Duernecker, Herrendorf, Valentinyi 36

38 Statistics Implied by the Calibration 1.0 Distortions 5.0 Sector TFPs A g A h A l τ h τ l Total consumption and relative price of services 3.5 C P s /P g Duernecker, Herrendorf, Valentinyi 37

39 Value Added, Employment, and Aggregate Productivity Model and Data 2.5 Relative value added VA l /VA g 3.0 Aggregate labor productivity VA l /VA h Data Model Data Model Employment shares H l 0.3 H h H g Data Model Back Duernecker, Herrendorf, Valentinyi 38

40 Different Assumptions about Future Taxes baseline: average τ h, τ l τ h = max{τ h }, τ l = max{τ l } τ h = max{τ h }, τ l = min{τ l } τ h = min{τ h }, τ l = max{τ l } τ h = min{τ h }, τ l = min{τ l } Duernecker, Herrendorf, Valentinyi 39

41 P l = ω P l + (1 ω) P h Quality Mismeasurement and Aggregate Productivity Growth in% Data Model Model Predictions ω Back Duernecker, Herrendorf, Valentinyi 40