Structural Development Accounting

Transcription

1 Structural Development Accounting Fabrizio. Zilibotti Jerusalem - June 29, 2011 Fabrizio. Zilibotti () Structural Development Accounting Jerusalem - June 29, / 58

2 Technology Di usion New and more productive technologies di use slowly over time across both rms and time. Griliches s (1957) study of the adoption of hybrid corn in the US ( ndings con rmed by others): Slow di usion a ected by local economic conditions. Likelihood of adoption related to the contribution of the hybrid corn in a particular area, the market size and the skill level. S-shape of di usion. To what extent can slow technology di usion explain cross-country productivity di erences? Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

3 Technology Di usion Di usion of hybrid corn across US states Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

4 Technology Di usion Kiessling (2008) studies cross-country di usion of personal computers, cell phones and TV. Slow di usion, with large di erences across countries Additional insight: Quality of political institutions a ects the speed of di usion. McKinsey (2001) report on technology backwardness of ndian rms: Firms are too small to bene t from the best technologies. Such technologies require skills that the country does not possess. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

5 Technology Di usion V Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

6 Cross-Country Productivity Di erences n 2007, PPP-adjusted income per capita in 5% poorest countries in the world was less than 1:50 relative to the US. What share of such gap is due to slow technology di usion? The standard development accounting exercise (e.g., Hall and Jones (1999) and Caselli (2005)) attempts to provide an answer Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

7 Cross-Country Productivity Di erences t postulates that output per capita, y, can be represented by a production function, e.g. h i Y = AK α L σ σ 1 + γh σ σ 1 σ 1 σ 1 α Having data on the observables (Y, K, H, α, β, σ) one can compute A as a residual. One can also decompose the variance of output per worker into the contribution of factor accumulation and technology. Finding: factor accumulation accounts for 40%-50% of the observed disparities in income per capita. Productivity appears to be the main source of income di erences. Technological lag; Within-country resource misallocation. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

8 Structural Development Accounting Needed: a theory about how di erent types of technologies are developed and adopted across countries. Such a theory requires unbundling the concept of TFP into a set of heterogeneous technologies and to identify what country-speci c factors facilitate the adoption of certain innovations more than others: skill endowment market size barriers to technology adoption Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

9 Structural Development Accounting Two antecedents 1 Acemoglu and Zilibotti (2001, AZ) 2 Caselli and Coleman (2006, CC) Main focus today 1 Gancia, Mueller and Zilibotti (2011) Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

10 Structural Development Accounting Y i = K α i L σ σ 1 i + γh σ 1 σ i σ! 1 α σ 1 AZ, QJE 2001 model a world economy all countries use the same technology the model yields γ endogenously [directed technical change in the "North"] productivity di erences are due to technology inappropriateness (originating from PR enforcement problem) explaining a signi cant share of the cross-country produtivity di erences in 1985 auxiliary prediction: industry-speci c patterns of cross-country productivity di erences Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

11 Structural Development Accounting V Y i = K α i L σ σ 1 i + γ i H σ 1 σ i σ! 1 α σ 1 CC, AER 2006 estimate (country-speci c) γ i using skill-speci c wage data nd that the main di erence between poor (skill-scarce) and rich (skill-abundant) countries is the productivity of skilled labor consistent with the observations that models tend to overpredict skill premia in skill-scarce countries country use di erent technologies; barriers to adoption data quality issue Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

12 Structural Development Accounting V GMZ 2011 Y i Y j = Ki α K α j L σ σ 1 i + γ (ZH i ) σ σ 1 L σ σ 1 σ 1 j + γ ZH σ j σ σ 1 β σ σ 1 β (1) a structural theory that that yields a micro-founded version of (1) building blocks: directed technical change and technology adoption an "aggregation" result: although countries adopt (endogenously) di erent technologies, the productivity ratio between any two countries can be expressed as the ratio between two CES (à la CC 2006) with the same (endogenous) γ Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

13 Structural Development Accounting V More speci cally: An advanced economy, the North, develops (endogenously) the world technology frontier. Benchmark case: no trade in technology e.g., due to the lack of international protection of intellectual property rights (PR) so that new machines are sold in the North only. As a result, the equilibrium skill bias of the world technology frontier is proportional to the skill endowment of the North. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

14 Structural Development Accounting V All other countries can adopt existing technologies at a cost which is decreasing in their distance from the frontier. Technology adoption - just like innovation - is pro t-driven. t is directed, i.e., it depends on local conditions, such as the abundance of complementary factors (K, L and H) and the size of domestic markets. This combination of directed technical change with international knowledge spillovers leads to a tractable model of cross-country technology di erences suitable for quantitative analysis. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

15 Preview of Results A structural model allows us to tease out the relative importance of two distinct sources of low productivity: 1 technology inappropriateness (T) - the North produces technologies that are too skill bias for the factor endowment of the South 2 barriers to adoption (BA) - lag in technology adoption within each sector Main results Both T and BA were important determinants of productivity di erences in 1970 BA fell signi cantly within OECD between , but remained almost unchanged betwee OECD and non-oecd Trade increases wage inequality in both developed and developing nations Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

16 The Model CES production function (omitting time indexes): Y = K α J h Y ε 1 ε LJ + Y ε 1 ε HJ i ε ε 1 α 1, where ε > 1. The production functions at the sector level are: Z NL θ Y L = E L x L (ν) θ θ 1 θ 1 Z NH θ dν, YH = E H x H (ν) θ θ 1 θ 1 dν 0 0 where E L (N L ) θ 2 θ 1 and E H (N H ) θ 2 θ 1 in order for the model to be consistent with balanced growth. Producers of Y L and Y H are competitive. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

17 The Model The intermediate good sector is monopolistic, with each producer owning the patent for a single variety. The production function for each intermediate input, x L (ν) and x H (ν), is linear in the type of labor employed, x L (ν) = l (ν) and x H (ν) = Zh (ν), and is subject to the resource constraints Z NL 0 l (ν) dν L and Z NH Thus, the production level of each variety satis es x L = L N L and x H = ZH N H 0 h (ν) dν H. n equilibrium, sectorial output can be expressed as: Y L = N L L and Y H = N H ZH. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

18 The Model The optimal price of each monopolist is a markup over the marginal cost (skilled or unskilled wage) p L (i) = p L = (1 1/θ) 1 w L p H (i) = p H = (1 1/θ) 1 w H /Z This implies that pro ts per rm are a fraction 1/θ of revenues: π L = (p L w L ) x L and π H = (p L w L ) x L π L = p L θ L and π H = p H ZH N L θ N H Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

19 The Model V Hence, the resulting pro t ratio is π H π L = p H ZH p L L = NH N L 1 ε ε 1 ZH ε, L where p H /p L = (w H /Z ) /w L which in turns is pinned down by the relative value of the marginal product of the two types of labor. The expression has two components: a price e ect: rents are higher in sectors producing more expensive goods, a market-size e ect: rents are higher in larger sectors. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

20 The Model V nnovation takes the form of the introduction of new varieties of intermediate inputs (an increase in N H and N L ) and is directed. The development of a new variety in sector H (L) requires the investment of a xed outlay, µ 1/η. Free entry implies that along a balanced growth path V L = π L r = µ and V H = π H r Moreover, N L and N H grow at the same rate. = µ. The theory pins down the equilibrium skill bias of technology (N H /N L ) compatible with balanced growth (π L = π H ): N H N L = ZH ε 1. L Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

21 The Model V Balanced growth also requires a constant interest rate! r = πh /µ = π L /µ O balanced growth, there is only one type of innovation: if π H > π L, then Ṅ H > 0 and Ṅ L = 0 if π H < π L, then Ṅ H = 0 does not grow and Ṅ L > 0 Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

22 The Model V We can apply this model to study the process of technology adoption. M+1 countries: North (1) and South (M). Southern countries are skill scarce, H S /L S < H N /L N. The South can adopt, at a cost, the technologies developed in the North. Except for this, the North and the South are identical. Benchmark: no trade in goods nor international PR protection. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

23 The Model V For Southern rms, the cost of technology adoption replaces the cost of innovation. The cost of adoption is a decreasing function of the distance to frontier: ξ ξ NLN NHN c L = µ and c H = µ, ξ 0. N LS N HS ξ is an inverse measure of barriers to adoption in the South: Z NHN NHS µ 0 N HN Z NLN NLS µ 0 N LN ξ dn HS = µn HN 1 + ξ, ξ dn HS = µn LN 1 + ξ. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

24 The Model X n a balanced growth equilibrium, free entry implies V LS = π LS r = c L and V HS = π HS r n the South, balanced growth requires = c H. π HS π LS = c H c L This yields the skill bias of technology in the South: N HS = Z H ε 1 εξ " # ε 1 1+εξ S N 1+εξ HN = Z ε 1 H εξ 1+εξ S HN. N LS L S N LN L S L N Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

25 The Model X N HS N LS = Z ε 1 " HS HN L S L N εξ # ε 1 1+εξ. Technology adoption in the South depends on the skill endowment both in the North and in the South. A low skill endowment in the South gives a strong incentive to adopt low skill-complement innovations. A high skill endowment in the North means that skill-complement innovations are abundant and therefore relatively cheap to adopt. The skill-bias of the Southern technology, N HS /N LS, is increasing in ξ, capturing the speed of technology transfer. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

26 The Model X Extreme cases: f ξ = 0 (prohibitive barriers) the South develops technologies independently from the North: N ε 1 HS ZHS =. N LS L S f ξ! adoption is free (no barriers) so that the South is using the same technology as the North. This is the case studied by Acemoglu and Zilibotti (2001). N HS N LS! N ε 1 HN ZHN =. N LN L N Note: as long as ξ > 0, the world economy converges to a balanced growth (same interest rate and growth rate). The model yields then predictions for balanced-growth output and productivity di erences as function of factor endowments and of exogenous parameters. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

27 The Model X For any S, the output ratio relative to the frontier is Y S Y N = Ω S K α S K α N L σ σ 1 S + γ (ZH S ) σ 1 σ L σ σ 1 N + γ (ZH N ) σ 1 σ where K S /K N = (Y S /Y N ) / (χ S /χ N ) and σ 1 (ε 1) (1 + ξ) = σ 1 + εξ β = (1 α) 1 + ξ α + ξ γ = Z H N L N Note: Scale E ect in technology adoption ξ(ε 1) 2 1+εξ σ σ 1 β σ σ 1 β Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

28 Technology nappropriateness Even if developing countries had free access to all technologies, technologies developed in rich countries may be inappropriate for the needs of poor countries. Technology-skill mismatch: technologies are designed for the North, the South lacks the skill endowment required to operate them optimally. E.g., high literacy is necessary to bene t from T equipment. Acemoglu and Zilibotti (2001) show that even a model with ξ! is quite successful Our model allows us to disentangle the role of barriers (ξ) from that of technology inappropriateness Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

29 Technology nappropriateness Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

30 Quantitative Analysis Back to the model: Y S Y N = Ω S K α S K α N L σ σ 1 S + γ (ZH S ) σ 1 σ L σ σ 1 N + γ (ZH N ) σ 1 σ σ σ 1 β σ σ 1 β, where Ω S is an exogenous productivity di erential (error term) and σ 1 (ε 1) (1 + ξ) = σ 1 + εξ β = (1 α) 1 + ξ α + ξ γ = Z H N L N ξ(ε 1) 2 1+εξ Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

31 Quantitative Analysis Calibration we set α = 0.35 to match the non-labor share of GDP in industrialized countries; we set ε and Z so as to match the time evolution of the skill premium in the frontier economy (USA) using the predictions of our theory (see below); we estimate ξ so as to get the best t of cross-country productivity di erences in a repeated cross-section of up to 90 countries; - look at 1970 and Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

32 Quantitative Analysis More formally, the theory yields (in steady state) w w HN w LN = Z P HN N ɛ 2 HN = Z ɛ 1 HN Z ɛ 1 h ɛ 2 P LH N LN L N Then, we set ɛ and Z so as to match exactly the equation: log ( w US,t ) = (ɛ 1) log(z ) + (ɛ 2) log h US,t, where t 2 f1970, 2000g. Hence, ɛ = 2 + log ( w US,2000) log ( w US,1970 ). log h US,2000 log h US,1970 Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

33 Quantitative Analysis V The skill premia as well as the skill ratios are taken from the CPS. We use two alternative measures of skill: secondary and tertiary school. For high school graduates: w " from 1.40 in 1970 to 2.02 in 2000; for college graduates w " from 1.57 in 1970 to 1.88 in For high school graduates h increased from 2.59 to 9.30; for college graduates h increased from 0.21 to mplied estimates: secondary! ɛ = 2.29 and Z = 1.05 tertiary! ɛ = 2.25 and Z = 1.96 Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

34 Quantitative Analysis V Non OECD OECD Skill sec tert Table: Estimates of xi Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

35 Quantitative Analysis V Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

36 Quantitative Analysis V A formal measure of goodness of t < 2 = 1 n log(y Sj /y US ) j=1 2 n log(y c Sj /y US ) / j=1 log(y Sj /y US ) 2, where log(y Sj /y US ) is the log-di erence in output pw from; the US in the data and c log(y Sj /y US ) is the model prediction. < 2 measures the proximity to the 45-degree line < 2 would be equal to 1, if all points were aligned on the 45-degree Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

37 Quantitative Analysis V Baseline Estimation No Barriers Skill sec tert Table: Goodness of Fit Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

38 Quantitative Analysis X Figure: counterfactual, no barriers Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

39 Estimating ndividual Barriers Can alternatively use the model as an accounting framework Allow di erent ξ for each country Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

40 Adption Cost, 1/(1+xi) Estimating ndividual Barriers 1970, tert, OECD TUR KOR JPN GBR PRT TA MEX RL FN SWE AUS GRC DNK AUT BEL NOR CAN NLD USA Relative GDP pw Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

41 Adption Cost, 1/(1+xi) Estimating ndividual Barriers 1970, tert, All DN THA TZA MW PAK PHL LKA GHA PNG SDN KENZMB UGA TUR KOR BEN CAF NPL ZWE TGO EGY DZA GUY ECU PRT LSOMRT LBR BD BOL CMR HND TUN MYSMAR COL NC RWA SLE SYR COG JAM SEN DOM CHL PANMLT CYP PER SWZ URY PRY SGP FJ SLV MUS ZAF RQ CR JPN MEX RL FN SR RN TTO GBR TA SWE AUS GRC DNKVEN AUT BEL NOR BRB CAN NLD JOR Relative GDP pw USA Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

42 Adption Cost, 1/(1+xi) Estimating ndividual Barriers V 2000, tert, All JPN TUR MEX KOR POL SVK CZE HUN GER PRT NZL GRC ESP AUS CANCHE FN SWE NOR GBR DNK SL FRA NLD BEL TA AUT RL Relative GDP pw USA Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

43 Adption Cost, 1/(1+xi) Estimating ndividual Barriers V 2000, tert, All DN BGD RQ TZA VNM THA PHL DZA PAK BRA JPN ZMB PER ROM NPL MW GHA KEN NER NC MDA ECU PNG MNG COL MEX UGAGUY TUR MOZ SLE ZWE RN RWA BEN LSO HND YEM LAO SYR BOL ALB MAR KOR ML SEN MRT CMR PRY KHM LKA BGR JAM SLV VEN ARG KAZ PAN EST EGYCUB POL BWA JOR CZE FJ DOMHRV CRSVK HUN GER LVA LTU TUN URY NAM SWZ BLZ TTO MYS CHL PRT NZL CYP GRC ESP AUS CANCHE ZAF SVN LBY FN SWE NOR BHR BRB ARE GBR DNK SL FRA NLD MACMLT BEL MUSSAU QAT TA AUT RL Relative GDP pw USA Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

44 Trade Extend the analysis to economies that trade in goods. Under free trade, there is a single world price for p L and p H. Due to a standard Stolper-Samuelson e ect, P H " and P L # in the North; P H # and P L " in skill-scarce countries. nnovation: the North continues to be the relevant market for new frontier technologies, since there is no PR protection in the South. Thus, π HN " and π LN # implying increasing skill bias in the frontier technology. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

45 Trade The e ect of trade on the direction of technology adoption in the South is instead ambiguous. trade increases P L /P H in skill-scarce countries, accelerating the adoption low-skill technologies. however, the higher skill-bias at the frontier makes it cheaper to adopt skilled technologies. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

46 Trade We can obtain an expression for output and productivity di erences under free trade. Y S KS = Ω S Y N K N 0 α 1+ξ Z H N ξ LS + LN ε 1 ĥ 1+ξ ξ LN + Z H N LN ε 1 ĥ 1+ξ ξ HS 1+ξ ξ HN 1 C A 1 α (2) where ĥ 1+ n j=1 1+ n j=1 HSj 1+ξ ξ H N LSj 1+ξ ξ L N < 1 Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

47 Trade V No scale e ects in adoption (local market size no longer matters). Trade a ects the shape of countries aggregate production possibility frontier (see Ventura (2005) and Fadinger (2010)). n particular, for given technology, the elasticity of substitution between Y LS and Y HS (equivalently, between L (1+ξ)/ξ S and H (1+ξ)/ξ S ) is in nite, instead of ε, as all countries face the same world prices. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

48 Trade V We perform two exercises 1 Counterfactual: steady-state e ect of removing trade barriers in Re-estimate the model assuming free trade instead of autarky. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

49 Trade V Figure: counterfactual, free trade Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

50 Trade V Figure: model estimation assuming free trade Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

51 Licensing of Technologies Together with free trade, frontier technologies can be licensed from Northern to Southern (monopolist) rms in exchange of a perpetual royalty per unit produced in the South. The South adopts instantaneously all technologies introduced in the North. All pro ts made in the Southern market are transferred to Northern rms as royalties. Both costs and bene ts for the South the South must transfer to the North the pro t ow of intermediate producers. the South can adopt immediately all technologies. n addition, PR enforcement reduces the skill bias of the frontier technology. n the calibrated economy, we observe a sharp decrease of inequality in GDP per capita. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

52 Licensing of Technologies Figure: counterfactual, licensing Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

53 Changes in Skill Premia We use the calibrated economy to study the e ect on skill premia of perfectly removing barriers opening up to free trade f we remove barriers, all countries use the frontier technology. Thus, the South adopts more skill-biased technologies. Skill premia go up in non-frontier economies. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

54 Changes in Skill Premia Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

55 Changes in Skill Premia Switching to free trade yields two e ects: the skill premium falls due to the standard Stolper-Samuelson theorem the skill premium increases due to more skill-biased technical change at the frontier Calibrated economies: wage inequality increases in the majority of countries, especially in those with abundance of skills and low-barriers. Some exceptions: very few high-barrier (sub-saharan) countries nequality rises in ndia and China. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

56 Changes in Skill Premia V Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

57 Conclusions A model of the world income distribution based on cross-country di erences in factor endowments (and in cost of capital), factor-biased (directed) technical progress costly technology adoption. Estimating the model allows us to disentangle barriers to technology adoption, inappropriateness (excessive skill-bias) of frontier technologies Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58

58 Conclusions Despite the parsimonious speci cation, a good t of the world income distribution. Both barriers to adoption and the excessive skill-bias of frontier technologies are quantitatively important. Barriers are higher in less developed countries. They have fallen over time for OECD countries only. The complete removal of barriers would increase output per worker relative to the US and lead to higher skill premia. Trade leads to SBTC, higher income disparities, and rising skill premia in the majority of countries. These results are reverted if trade liberalization is coupled with international protection of PR. Fabrizio. Zilibotti (UZH) Structural Development Accounting Jerusalem - June 29, / 58