Organizational innovations and productivity: the case of foreign outsourcing José C. Fariñas (U. Complutense Madrid) Ana Martín-Marcos (UNED) Simposium on Business Dynamics and Innovation: The effects of agglomeration economies University of Barcelona (8 October, 2008) Introduction •Oslo Manual (OECD, 2005): •“… organizational innovations refer to the implementation of new organizational methods. These can be changes in business practices, in workplace organization and in the firm’s external relations” •Among of the most relevant organizational methods in firm’s external relations is outsourcing • We use the term foreign outsourcing (offshoring) to mean the acquisition of intermediate goods and services from foreign (affiliated and unaffiliated) firms •Fragmentation of production or foreign outsourcing has been a major factor in the growth of trade and FDI • The objective of the paper is to explore the relationship between foreign outsourcing and productivity at the firm level Outline • Theoretical framework • Descriptive evidence on foreign outsourcing • Testing procedure and productivity measure • Empirical results • Robustness checks •Other controls •Endogeneity issue • Main conclusions Theoretical framework • Antràs and Helpman (2004) offers a theoretical framework where firms make two endogenous organizational choices: - integration decision (integration vs. outsourcing) - location decision (domestic vs. foreign) • Two agents are engaged in production: final-good producers and firms producing components that can be located at home (D) or in the foreign country (F). Firms producing components can be vertically integrated (V) or not: outsourcing (O) • Fixed organizational costs of search, monitoring and communication are ranked as follows: fvF > foF > fvD > foD - regardless of ownership structure, fixed costs are higher in the foreign country - given location, costs of an V-firm > costs of an O-firm Industry equilibrium predictions: Location and integration decisions will depend on the level of firm productivity (λ) and on the relative input intensity of the industry, according to the following pattern: Component intensive industry: Exit Outsource in D OF OD Outsource in F (arm’s length trade) λ, Productivity level Headquarter intensive industry: Exit VF OF VD OD Either outsource or Outsource in F (arm’s length trade) integrates in D Integrates in F (FDI) (intrafirm trade) λ, Productivity level • Antràs and Helpman (2004) model predicts self selection: high productivity firms outsource in international markets, whereas low-productivity firms acquire intermediate materials at home. • However, from an empirical point of view, self-selection is not inconsistent with the fact that firms engage in foreign outsourcing in the expectation that this would improve productivity: firms reallocate relatively inefficient parts of their production process to another country where they can be produced more cheaply • Two-way relationship between foreign outsourcing and productivity Descriptive evidence on foreign outsourcing • Foreign outsourcing = the import of intermediate inputs (goods and services) by domestic firms (Feenstra and Handson 1996). • Measure (at the firm level): – Share of imported intermediate inputs in the total purchase of intermediate inputs. Os f imported intermedia te inputs of j by firm f j total purchases of intermedia te inputs by f • Estimates of foreign outsourcing at the firm level based on information reported by Encuesta sobre Estrategias Empresariales (ESEE): – Firms report the value of their total imports – Firms report the percentage of the capital stock purchased abroad and the annul value of their investment in capital goods – Firms report the value of imports of similar goods coming from affiliated companies Imports of intermediate importsi Total importsi - Imports of capital goodsi Imports of similar goodsi Percent of firms importing intermediate inputs 60 53,7 55 49,6 50 45 55 43,6 40 35 30 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 EXTENSIVE MARGIN: Large increase in the percentage of manufacturing firms performing foreign outsourcing over the period 1990-2002 Foreign outsourcing intensity: Imported intermediate inputs /Total intermediate inputs (only importers) 25,00 20,10 20,00 18,80 16,50 15,00 10,00 5,00 0,00 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 INTENSIVE MARGIN Firms already performing foreign outsourcing have increased their intensity, although less than the extensive margin Table 3 Foreign outsourcing and firm size Firm size Percent of firms performing foreign outsourcing Intensity of foreign outsourcing (intermediate imports/total intermediate inputs) Only intermediate importers All firms Firms with ≤ 200 employees 36,6 16,2 5,9 Firms with > 200 employees 80,2 20,1 16,1 Positive relationship between foreign outsourcing and the size of firms. The difference is higher for the probability than for the intensity, which suggest that performing this activity involves significant sunk costs Figure 2 The probability of outsourcing vs. the intensity of foreign outsourcing by industry (2 digit manufacturing NACE and ISIC classification) 25 15 17 16 12 18 20 9 7 6 4 14 10 20 2 13 15 8 3 19 10 11 5 5 1 .2 .4 .6 Number of outsourcing firms/ Total number of firms High degree of heterogeneity across industries. Industries where the intensive and the extensive margins are higher: 15 Office machinery, computers and precision instruments 16 Electrical machinery and communication equipment 17 Motor vehicles 18 Other transport equipment 9 Chemicals and chemical products .8 Testing procedure and productivity measurement • Perform comparisons of productivity distributions of different groups of firms • The procedure is based on non-parametric techniques and has been previously used in Delgado, Fariñas and Ruano (2002) • Let define two productivity distributions: F, productivity distribution of firms outsourcing in foreign countries G, productivity distribution of firms not performing foreign outsourcing (either they integrate or outsource at home) • According to predictions proposed in Antràs and Helpman (2004) for component intensive industries, the distribution F should dominate the distribution G. • Stochastic dominance of F relative to G requires two statistical conditions to be satisfied: - First, both distributions are not identical, i.e. the null hypothesis H0 : F(z)G(z) = 0 can be rejected; -Second, the sign of the difference is as expected, i.e. the null hypothesis H0 : F(z)-G(z) 0 cannot be rejected (the variable z represents independent random samples drawn from cdf’s F and G, of size n and m respectively) • These two-sided and one-sided tests can be performed respectively using the Kolmogorov-Smirnov test statistics: • To further illustrate the comparisons between the productivity of different groups of firms we have graphed their smooth sample distribution functions. Productivity measurement • Database: Encuesta sobre Estrategias Empresariales (ESEE) • Panel of manufacturing firms containing 21,098 observations over the period 1990-2002 that correspond to an average number of 3,462 firms per year. • The dataset is representative of the population of Spanish manufacturing firms Large firms (more than 200 employees) Small firm (between 10-200 employees), random sampling scheme • Firm productivity is defined by an index of total factor productivity for each firm over the period 1990-2002. The index proposed is an extension of the multilateral total factor productivity index proposed by Caves, Christensen and Diewert (1982). •The expression of total factor productivity for firm f, at time t, in a given industry is: ln ft ( ) 1 R ln y ft ln y ( rft r )(ln x rft ln xr ) 2 r 1 1 R ln y ln y ( r r )(ln xr ln x r ); 2 r 1 Where, yft is the output of firm f at time t; xrft is the quantity of input r labor, materials and the stock of capital, corresponding to firm f at time t; ωrτ is the cost share of input r. Firms are classified in two size groups of small and large firms, = 0 ,1 • The index measures the proportional difference of TFP for firm f relative to a given reference firm. The reference firm varies across industries (NACE classification at the two-digit level). When observations are pooled, average TFP differences across industries are removed. Empirical results 1. Comparisons between productivity distributions that have been performed: - Foreign outsourcing firms vs. non-foreign outsourcing firms 1.1 All manufacturing firms 1.2 Component intensive industries - Ex-ante comparisons: entering outsourcing firms vs. non-outsourcing firms 2. Robustness checks 2.1 Additional controls: productivity premium of foreign outsourcing is robust to other characteristics associated to productivity 2.2 Endogeneity issue Figure 2 Productivity differences between foreign outsourcing and non-outsourcing firms (Smooth sample distribution function) Table 2 Productivity level differences between foreign outsourcing firms and non- foreign outsourcing firms: hypotheses test statistics (All firms) Number of firms Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Foreign outsourcing 477 574 687 655 720 690 702 820 858 858 910 824 795 Non-foreign outsourcing 618 768 833 787 731 678 682 782 738 741 754 674 651 Difference (%) in productivity level between foreign outsourcing and nonforeign outsourcing firms in the: 25th 75th Median percentile percentile 4.8 6.4 4.9 8.7 10.4 4.8 7.7 8.8 6.9 7.5 7.5 4.4 8.7 10.3 5.5 8.6 10.7 7.4 9.5 10.9 8.8 8.8 8.6 8.5 8.4 8.0 5.6 9.0 8.9 7.7 8.2 8.7 7.9 5.5 6.2 4.2 4.3 6.4 4.3 Average difference at the median: 7.5 percent Equality of distributions Differences favorable to foreign outsourcing firms P-value P-value 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.826 0.814 0.960 0.939 0.988 1.000 0.998 1.000 0.996 0.986 0.962 0.969 0.985 Table 5 Productivity level differences between foreign outsourcing and non- foreign outsourcing firms: hypotheses test statistics (only component intensive industries) Number of firms Year 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 Foreign outsourcing 249 302 380 379 402 392 389 453 487 486 506 449 415 Non-foreign outsourcing 265 336 345 328 308 275 273 319 301 308 316 276 281 Percent difference in productivity level between foreign outsourcing and non- foreign outsourcing firms: 25th 75th Median Percentile Percentile 2.3 5.4 3.7 7.1 8.4 2.6 8.7 6.7 6.3 7.6 6.8 4.7 9.3 8.7 6.3 8.1 10.4 7.1 8.1 9.6 7.8 7.3 7.0 6.0 7.9 5.3 5.8 7.0 5.3 7.3 6.1 3.6 6.1 3.7 3.7 4.6 0.2 1.6 4.5 6.5 percent Equality of distributions1 Differences favorable to foreign outsourcing firms2 P-value P-value 0.059 0.001 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.026 0.124 0.708 0.612 0.895 0.965 0.996 1.000 0.998 1.000 0.974 0.994 0.928 0.915 0.864 Table 6 Ex-ante productivity differences between entering outsourcing firms and non-foreign outsourcing firms: hypotheses test statistics Number of firms Starters Year to foreign outsource 1990 74 1991 102 1992 81 1993 106 1994 59 1995 67 1996 74 1997 118 1998 78 1999 86 2000 76 2001 105 Non- starters to foreign outsource 516 644 620 552 574 546 548 585 590 595 587 542 Percent difference in ex-ante productivity level between starters to foreign outsource and nonstarters Equality of distributions1 Differences favorable to starters firms2 Median 25th percentile 75th percentile P-value P-value 4.6 8.0 0.8 6.4 1.2 3.7 5.8 7.8 4.8 2.6 3.9 3.2 0.7 1.8 5.1 3.7 3.9 4.4 3.8 7.9 4.3 5.3 5.5 2.8 3.8 6.2 -1.6 4.6 -0.7 1.0 3.4 6.6 6.6 2.2 0.5 3.9 0.403 0.009 0.807 0.044 0.516 0.129 0.018 0.000 0.080 0.319 0.064 0.139 0.936 0.882 0.793 0.883 0.862 0.720 0.998 0.919 0.968 0.984 0.954 0.911 4.4 percent. Robustness checks i) Additional controls: – We check whether the productivity premium of foreign outsourcing firms is robust to other characteristics that are associated with firm productivity. – Foreign outsourcing can be correlated with omitted variables such as size, innovation, human capital, which may be inflating the magnitude of the productivity premium. ii) Endogeneity issues – Potential endogeneity of foreign outsourcing could lead to biased estimates i) Additional controls ln TFPit FOit Z it eit • ln TFPit is the log of TFP, where average productivity differences across industries are removed • FOit is a dummy variable for the current outsourcing status (1 if firm i out sources in year t, 0 else); • Zit is a vector of control variables that includes: – the log of number of employees and its squared value to measure firm size – a dummy variable indicating Foreign ownership (1 if firm has 50 percent or more of equity owned by foreign capital) – the log of firm age – a full set of year dummies to control for common shocks to all firm – the log of the ratio of labour cost per hour to proxy human capital • Additionally, to control for unobserved firm heterogeneity due to timeinvariant firm characteristics, the equation is estimated with fixed firm effects. ii) Endogeneity issues • Potential endogeneity of foreign outsourcing could lead to biased estimates • More productive firms might self select into foreign outsourcing, and/or alternatively firms engage in this activity with the objective to improve productivity. • To address this issue: GMM estimation ln TFPit 1 FOI it 2 NFOit Z it eit 1, if the firm stops outsourcing abroad dNFOit 0, if the firm does not change their decision 1, if the firm starts outsourcing abroad FOI it FOI it 1 , if the firm continues outsourcing abroad dFOI it 0, otherwise • Total effect of foreign outsourcing on productivity using the coefficients from the GMM estimation: • Coefficient ≈ 0.16 • As foreign outsourcing increased by 0.025 percentage points over the sample (from 0.165 to 0.188), the coefficient implies that foreign outsourcing lead to an increase of 0.4 percent of TFP. • Given that firms increase their average TFP by 18 percent over the period, foreign outsourcing accounted for almost 2.5 percent of the growth of TFP at the firm level. Main conclusions • Productivity differences between firms engaged in foreign outsourcing of intermediate inputs and firms non-outsourcing abroad are consistent with the predictions of Antrás and Helpman (2004) • High productivity firms are more likely to engage in global production strategies. • Evidence is consistent with self-selection of the most productive firms into the practice of outsourcing abroad. The ex-ante productivity distribution of firms that engage into foreign outsourcing stochastically dominates the distribution of firms non-outsourcing abroad • The productivity premium of foreign outsourcing by firms is robust to other characteristics associated to productivity • Foreign outsourcing accounted for 2.5 percent of TFP growth at the firm level over the period. Final comment: Offshoring in the conventional sense of reallocation of processes to an external provider, has been very intensive in the footwear industry. A case study (Fuster, Martinez and Pardo, 2007) of this industry in the province of Alicante shows that firms that engage in global production strategies have higher productivity than firms nonoutsourcing from abroad.