O shoring under Oligopoly Mitsuru Igami Yale April 2015 Igami (Yale) Strategic O shoring Apr 2015 1 / 24
O shoring under Oligopoly O shoring International/macro: Globalization of markets for goods & services Labor: Prospect of massive job destruction in North This paper: From Silicon Valley to Singapore Firm-level o shoring decisions of HDD makers (global oligopoly) Foreign direct investment (FDI), not outsourcing Understand rms incentives to o shore & impact of government interventions Figure 1: Computers and Electronics are the Hardest Hit Igami (Yale) Strategic O shoring Apr 2015 2 / 24
O shoring under Oligopoly Literature Theory & empirics based on monopolistic competition models Grossman & Rossi-Hansberg ( 08), Antras et al ( 06), Grossman et al ( 05), Antras & Helpman ( 04); Bernard et al ( 09), Eaton et al ( 04). Focus on labor-market impact Feenstra & Hanson ( 96, 97, 99, 03), Autor et a. ( 03), Hsieh & Woo ( 05), Feenstra ( 10), Ottaviano et al ( 10), Burstein & Vogel ( 11) Little is known about rms decision-making (none under oligopoly) From product-market perspectives, o shoring is: Cost-reducing investment ( process innovation ) Possibly drastic (Arrow 62) Complication: Computers & electronics are global oligopolies Rivals costs a ects own pro t & survival O shoring incentives depend on rivals & market structure Strategic, forward-looking decision Igami (Yale) Strategic O shoring Apr 2015 3 / 24
This Paper Goal To disentangle empirical relationship between o shoring & competition Questions 0. How do o shoring & competition evolve in the long run? 1. How does competition a ect o shoring incentives? 2. How does (the possibility of) o shoring a ect long-run market structure? 3. Should governments encourage or discourage o shoring? Who cares? Life & death of rms & industries Job destruction 2 creative destruction Main challenges Forward-looking behavior Endogenous market structure Igami (Yale) Strategic O shoring Apr 2015 4 / 24
This Paper Approach Describe Model Estimate Data: HDD Industry (1976 98) Dynamic oligopoly game of o shoring & entry/exit 1. Demand (global) 2. Production costs (north & south) 3. Sunk cost of o shoring Simulate 1. No o shoring 2. Unilateral intervention 3. Governments in Nash equilibrium Why bother with heavy tools? Life & death questions Simultaneous evolution of o shoring & market structure Need counterfactuals & welfare analysis Igami (Yale) Strategic O shoring Apr 2015 5 / 24
Data (1 of 6): Why Study Hard Disk? Relevant: Massive o shoring in high-tech Figure 2: Market Structure and O shoring Feasible Long panel (23 years) Global coverage (178 rms) Details on technology, products, & plant locations Igami (Yale) Strategic O shoring Apr 2015 6 / 24
Data (2 of 6): North vs South Figure 3: Average Output by Location O shorers produce more than non-o shorers More production after o shoring Igami (Yale) Strategic O shoring Apr 2015 7 / 24
Data (3 of 6): Descriptive Statistics Size, age, technology, & organizational type Table 1: Summary of 151 HDD Makers Variable Obs Mean Stdev Min Max Mean revenue from HDD sales 151 103.7 650.7 0 7764.8 First year in HDD market 151 1982.9 5.4 1976 1998 First year of o shoring 42 1988.9 4.3 1983 1998 Initial tech-generation of entry Indicator: 14 inch 151.23.42 0 1 Indicator: 8 inch 151.10.30 0 1 Indicator: 5.25 inch 151.39.49 0 1 Indicator: 3.5 inch 151.24.43 0 1 Indicator: 2.5 inch 151.04.20 0 1 Organizational type Indicator: Specialized HDD maker 151.42.49 0 1 Indicator: Computer maker 151.30.46 0 1 Indicator: HDD component maker 151.07.26 0 1 Indicator: Other electronics maker 151.21.41 0 1 Note: Major and fringe rms in the mainstream segment (non-captive, xed HDDs). Igami (Yale) Strategic O shoring Apr 2015 8 / 24
Data (4 of 6): Productivity-based Sorting? Common focus of trade literature Table 2: Do Better Firms Self-Select into O shoring? Quartile based on Number of % o shored by 1991 % exited by 1991 1976 85 market share Firms (without o shoring) 1st quartile 11 36.4 36.4 2nd quartile 11 27.3 63.6 3rd quartile 11 36.4 36.4 4th quartile 11 18.2 63.6 No obvious patterns in HDD industry Igami (Yale) Strategic O shoring Apr 2015 9 / 24
Data (5 of 6): Persistent Firm Heterogeneity? Rather random dynamics Figure 4: High Volatility of Firm Size High-tech rms far from steady state (almost by de nition) Igami (Yale) Strategic O shoring Apr 2015 10 / 24
Data (6 of 6): Preliminary Regressions Who o shores? When? Table 3: Preliminary Regression of O shoring Timing Dependent variable: Duration model (Cox proportional hazard estimates) Decision to O shore (1) (2) (3) (4) Firm size it 1.000 1.000 HDD entry year i 1.062.935 % o shore rms t 14.83 21.29 Initial tech-generation 8-inch 1.531 5.25-inch 2.204 3.5-inch 1.973 2.5-inch 3.873 Organizational type Specialized HDD maker 5.705 Computer maker 1.911 HDD component maker 1.218 Number of rms 151 151 151 151 Number of o shoring 42 42 42 42 Time at risk 772 772 772 772 Log likelihood 181.84 180.53 179.57 169.20 Note: Coe cients greater (less) than 1 indicate higher (lower) propensities to o shore. Firm size is measured by its revenue from HDD sales. o shore rms in the global market. % o shore rms measures the fraction of Igami (Yale) Strategic O shoring Apr 2015 11 / 24
Model (1 of 3): Intuition A static Cournot example Figure 5: Strategic Substitutability May Be Reversed in a Dynamic Setting Igami (Yale) Strategic O shoring Apr 2015 12 / 24
Model (2 of 3): Overview Dynamic discrete game N t rms in North 8 < V t (s t ) = π t (s t ) + max : N t rms in South Vt (s t ) = πt (s t ) + max 9 = ε 0 it, φ + βe V t+1 (s t+1 ) js t + ε 1 it, φ + βe Vt+1 (s t+1) js t κ + ε 2 ; it ε 0 it, φ + βe Vt+1 (s t+1) js t + ε 1 it Igami (Yale) Strategic O shoring Apr 2015 13 / 24
Model (3 of 3): Timeline In each year t 1. Potential entrants ( ): Observe market structure s t = (N t, N t ) Sequentially decide whether to enter: free entry max V t (s t ) Actual entrants become active in North κ ent t, 0 2. Each active rm i (incumbents + actual entrants): Observes updated s t & private cost shocks ε it Decides whether to: fexit, stay North, go Southg If already in South, whether to exit 3. Active rms earn period pro ts 4. Decisions implemented & state evolves π t (N t, N t ) & π t (N t, N t ) Igami (Yale) Strategic O shoring Apr 2015 14 / 24
Estimation (1 of 5): Demand Steps: (1) demand! (2) supply! (3) dynamics Industry demand: Di erentiated products msjt ln = α 1 p jt + α 2 g j + α 3 x j + ξ jt, ms 0t Table 4: Demand Estimates Market de nition: Broad Narrow Estimation method: OLS IV OLS IV (1) (2) (3) (4) Price ($000) 1.66 2.99.93 3.28 Diameter = 3.5-inch.84.75 1.75.91 Log Capacity (MB).18.87.04 1.20 Year dummies Yes Yes Yes Yes Region/user dummies Yes Yes Adjusted R 2.43.29.50.27 Number of obs. 176 176 405 405 Note: ***, **, and * indicate signi cance at the 1%, 5%, and 10% levels, respectively. IVs for p jt Prices in other region/user (Hausman-Nevo) Num. of product models/ rms (Bresnahan-BLP) Predictable & unpredictable changes in unobserved quality (Sweeting-Lee) Igami (Yale) Strategic O shoring Apr 2015 15 / 24
Estimation (2 of 5): Supply Cost of production Invert demand & use Cournot FOC: P t + c P Q q it = dmc it Figure 6: Cost Advantage of O shore Production Quick check: Labor share (10%) Wage discount (65%) = Cost cut (6.5%) Igami (Yale) Strategic O shoring Apr 2015 16 / 24
Estimation (3 of 5): Dynamic Game Cost of o shoring Full-solution approach (Rust 87, Benkard 04, Schmidt-Dengler 06, Goettler & Gordon 11, Lee 13) 1. Try some (κ, φ) 2. Solve for equilibrium Perfect Bayesian Equilibrium Sequential move Backward induction from year 1998 3. Pick (κ, φ) with maximum likelihood 4. Free entry: V t (N t, N t ) 6 ˆκ ent t 6 V t (N t 1, N t ) Data variation: Time-series of entry/exit/o shoring Table 5: Estimates of the Dynamic Parameters ($ Billion) Maximum Likelihood Estimates Fixed cost of operation (φ) h 0.07 i 0.09, 0.25 Sunk cost of o shoring (κ) 4.31 h 3.20, 5.85 Log likelihood 156.17 Note: The 95% con dence intervals, based on likelihood-ratio tests, are in brackets. i Igami (Yale) Strategic O shoring Apr 2015 17 / 24
Estimation (4 of 5): Sensitivity To accomodate nonstationarity, allow time trend in the o shoring cost Firms in year t pay δ t κ instead of κ Table 6: Sensitivity Analysis with respect to Time Trend ($ Billion) δ =.90 δ =.95 δ = 1.00 δ = 1.05 δ = 1.10 δ = 1.15 Fixed cost (φ) 0.05 0.07 0.08 0.08 0.06 0.05 Sunk cost (κ) 6.49 4.31 2.73 1.57 0.82 0.39 Log likelihood 158.65 156.17 158.73 167.08 178.82 190.23 Note: As a reminder, the baseline parameter value for δ is.95. Highest likelihood with δ =.95 (mildly decreasing trend) Igami (Yale) Strategic O shoring Apr 2015 18 / 24
Estimation (5 of 5): Fit Figure 7: Fit of Market Structure Dynamics Main patterns captured 1. N t peaks at 26 in the rst half 2. N t increases in the middle & dominates in the end Igami (Yale) Strategic O shoring Apr 2015 19 / 24
Finding (1 of 4): How Competition A ects O shoring Figure 8: E ects of Market Structure on Pro ts & Values Pro ts Values Decreasing in N (disproportionately for non-o shorers) Reasons: Downward pressure on global price & business stealing Decreasing in N =) Pr (exit) " in N Gap (V V ) " in N =) Pr (o shore) " in N (in a relevant region) Igami (Yale) Strategic O shoring Apr 2015 20 / 24
Finding (2 of 4): How Competition A ects O shoring Figure 9: O shoring Breeds O shoring How does Pr (o shore) change with N /N? Fix total N = N + N and vary N (& hence N /N) O shoring breeds o shoring: Fly or die Igami (Yale) Strategic O shoring Apr 2015 21 / 24
Finding (3 of 4): How O shoring A ects Competition 2 counterfactuals No Singapore: O shoring/trade cost prohibitively high at κ = 4ˆκ Unilateral intervention: κ US, κ Other = (4ˆκ, ˆκ) Figure 10: Counterfactuals with Higher O shoring Costs, κ Finding Pro-competitive & accelerates shakeout: drastic innovation (Arrow 62) Long-run unintended consequence of unilateral intervention: N US # Igami (Yale) Strategic O shoring Apr 2015 22 / 24
Finding (4 of 4): Governments in Nash Northern governments (US & Japan) with 2 hypothetical objectives Maximizing national producer surplus Maximizing national welfare Figure 11: Nash O shoring Policies Findings Mercantilist governments would end up in subsidy race, but Consumers would rather free-ride on foreign rms o shoring e orts Igami (Yale) Strategic O shoring Apr 2015 23 / 24
Conclusion O shoring under oligopoly Finding Hardest-hit industries are global oligopolies Rivals costs a ect own pro t & survival Firm-level analysis of incentives to o shore 1. O shoring breeds o shoring Via competitive pressure: y or die 2. O shoring is pro-competitive & triggers shakeout Plays the role of drastic innovation (Arrow 62) 3. Unilateral intervention kills domestic rms If you can t beat them, join them 4. Governments in Nash equilibrium Mercantilist subsidy race to drive out foreign rms Free-riding on foreign rms o shoring e orts Igami (Yale) Strategic O shoring Apr 2015 24 / 24