## _wp05238 — 1. Offshoring Intensity: 1992–2000

## Source details

**Canonical URL:** [_wp05238 — 1. Offshoring Intensity: 1992–2000](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05238.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05238.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05238.pdf.json)

---

### Background and motivation
- New technologies enable firms to source service inputs from abroad (examples cited: call centers in India, computer software development).
- In the OECD, "the use of imported inputs in producing goods that are exported accounted for 21 percent of trade in 1990, and this grew by 30percent between 1970 and 1990."
- Service offshoring by U.S. manufacturing industries grew at "an average annual rate of6.3percent between 1992 and 2000."
- Political and public concern:
  - Support for free trade among white collar workers with incomes over $100,000 slid from "57 percent in 1999 to 28 percent in 2004."
  - "On March 4, 2004, the U.S. Senate passed restrictions on offshoring by barring companies from most federal contracts if they planned to carry out any of the work abroad."
- Research gap: empirical evidence on effects of service offshoring (on productivity and employment) is scant; this paper estimates whether offshoring leads to job losses and whether there are offsetting productivity benefits.

### Data and measurement (1992–2000)
- Time coverage: annual data for "1992 to 2000" from Bureau of Labor Statistics (BLS) I/O tables combined with trade data.
- Offshoring measures:
  - Service offshoring: share of imported services in industry inputs (includes imports from affiliated and unaffiliated firms), analogous to Feenstra and Hanson (1999) material offshoring measure.
  - Material offshoring: analogous measure for imported materials.
- Construction of ossi:
  - ossi = Σj [ input purchases of service j by industry i, at time t / total non-energy inputs used by industry i, at time t ] * [ imports of service j, at time t / (productionj + importsj − exportsj at time t) ]
  - Five service industries included as inputs to manufacturing: telecommunications; insurance; finance; business services; computing and information.
    - Average shares in 2000:
      - Business services: 12%
      - Finance: 2.4%
      - Telecommunications: 1.3%
      - Insurance: 0.5%
      - Computing and information: 0.4%
    - Example computation: U.S. economy imported 2.2 percent of business services in 2000 → average offshoring intensity of business services = 0.12 * 0.022 = 0.3 percent.
- Key aggregate statistics:
  - Average share of imported service inputs in 2000: 0.3 percent.
  - Average share of imported material inputs in 2000: 17.4 percent.
  - Growth over sample period (annual averages):
    - Service offshoring: 6.3 percent (annual average growth rate).
    - Material offshoring: 4.4 percent (annual average growth rate).
  - Change in offshoring intensity (Table 1 summary):
    - Share of Imported Material Inputs (OSM): 1992 = 11.72%; 2000 = 17.33%; 1992-2000 change reported as 4.38.
    - Share of Imported Service Inputs (OSS): 1992 = 0.18%; 2000 = 0.29%; 1992-2000 change reported as 6.26.
- Aggregation:
  - Data aggregated from "450 SIC manufacturing industries to 96 manufacturing industries" to match I/O tables that provide service input detail.
  - Analysis also uses the more disaggregated 450-industry division for employment effects.
- Capital and depreciation:
  - Capital stock series extended beyond 1996 using the perpetual inventory method with average depreciation rates from NBER (Bartelsman and Gray, 1996).
  - Depreciation rates: 7.7 percent for equipment and 3.5 percent for structures.
- Recognized measurement limitations:
  - Likely underestimation of offshoring intensity due to current-value measures and lack of quantities; applying economy-wide import shares to all industries is an imperfect proxy; measurement error can bias estimated effects downward.

### Identification and instruments
- Endogeneity concerns:
  - Reverse causality (productivity may drive offshoring choices) and measurement error in offshoring.
- Instrumental variables strategy:
  - Service offshoring instrument: country-level time-varying measures of technology that enable offshoring ("the number of internet users in the countries that the United States imports most of its service inputs"), interacted with the industry's initial share of services in output to create time/industry varying instruments.
  - Material offshoring instrument: "the freight cost of inputs" (freight-and-insurance rates weighted by input shares).
- Expected instrument properties:
  - Instruments should "only affect productivity through service offshoring, and have sufficient explanatory power in predicting changes in service offshoring."
- First-stage and diagnostics:
  - First-stage F(5,44) = 14.27 with p-value less than 0.01 for lagged service offshoring when only one endogenous variable included.
  - Shea partial R-squared for service offshoring instruments ranges 0.31 to 0.43; instruments for material offshoring weaker.
  - Sargan/Hansen tests reported across specifications; some pass overidentification while others fail certain tests.

### Conceptual framework and channels
- Production function: Yi = Ai(ossi, osmi) F(Li, Ki, Mi, Si)
  - Ai is a technology shifter depending on offshoring of services (ossi) and materials (osmi).
- Four channels through which offshoring affects productivity:
  - Static efficiency gain (compositional effect when less efficient stages are relocated).
  - Restructuring (remaining workers become more efficient; more likely from service offshoring).
  - Learning externalities (imported services, e.g., software, improve average productivity).
  - Variety effects (use of new material or service input varieties).
- Labor demand under cost minimization:
  - Li = g(wi, ri, qm, qs, Yi) / Ai(ossi, osmi) and in profit-maximizing output form Li = g(wi, ri, qm, qs, pi, ossi, osmi) / Ai(ossi, osmi).
- Net employment effect ambiguous: substitution and productivity effects may reduce labor demand; scale and competitiveness effects may increase labor demand.

### Estimation strategy
- Productivity estimating equation (first differences, logs):
  - ∆lnYit = α0 + α1 ∆ossit + α2 ∆osmit + β1 ∆lnLit + β2 ∆lnKit + β3 ∆lnMit + β4 ∆lnSit + δt Dt + δi Di + εit
  - Hypothesis: α1 and α2 are positive; one-period lags of offshoring variables included.
- Employment estimating equations (first differences, log-log):
  - Baseline: ∆lnlit = γ0 + γ1 ∆ossit + γ2 ∆osmit + γ3 ln∆wit + γ4 ∆lnYit + δt Dt + δi Di + εit
  - Scale-adjusted variant includes ∆lnpit.
- Estimation methods: OLS, Arellano-Bond GMM, robust regressions, two-stage least squares (IV) with clustering-robust standard errors.

### Main empirical findings — productivity
- Service offshoring has "a significant positive effect on productivity in the manufacturing sector."
- Contribution to labor productivity growth (1992–2000):
  - Service offshoring accounts for "around 11percent of labor productivity growth over the sample period."
  - Material offshoring accounts for "5 percent of total labor productivity growth between 1992 and 2000."
- Representative coefficients and magnitudes (preserve reported estimates):
  - Examples from Table 4: ∆oss i,t coefficients reported as 0.235***, 0.249***, 0.241***, 0.341***, 0.331***, 0.335***, 0.258*** across columns; joint tests p-values = 0.00.
  - IV/GMM-derived aggregate effect of service offshoring on productivity ranges reported as 0.3 (Arellano-Bond Table 10 column 6) to 0.38 (one-period differences 2SLS Table 9 column 4) and 0.46 (two-period differenced variables Table 9 column 8).
- Interpreted magnitudes:
  - Service offshoring increased by 0.1 percentage point over the sample period, from 0.18 to 0.29 (Table 1).
  - Interpreted implication: service offshoring led to an increase of 3 to 4.5 percent in labor productivity over the sample period.
  - Given value added per worker increased by an average of 35 percent over the sample period, service offshoring accounted for 11 to 13 percent of the average growth in labor productivity.
  - Material offshoring accounted for 3 to 6 percent of labor productivity growth (conclusion statement).
- Robustness:
  - Service offshoring results robust to controls (high-technology capital shares, import share), outlier exclusion (tobacco omitted), longer differencing, and alternative estimators.
  - Instrumental-variable estimates indicate a slightly larger positive productivity effect from service offshoring than OLS.
  - Material offshoring’s positive effect less robust and smaller in magnitude.

### Main empirical findings — employment
- Aggregation-dependent effects:
  - At the 96-industry aggregation, "service offshoring has no significant effect on employment."
  - At the 450 SIC 4-digit disaggregation, "service offshoring reduced manufacturing employment by around 0.4 of a percent" (weighted) and an unweighted implied loss of 3percent.
- Detailed estimates:
  - Two-period differenced specifications at 450-industry level show persistent negative service-offshoring effects with no offsetting scale effects.
  - Using estimates from Table 13 with scale effects, effect from service offshoring on employment equals 0.3. Given service offshoring grew by 0.1 percentage point, this implies a loss of 3percent employment; weighted by employment shares this number falls to 0.4of a percent.
- Material offshoring:
  - Material offshoring often shows positive effects on employment at the 96-industry level but becomes insignificant at the 450-industry disaggregation.
- Instruments and employment:
  - Instruments for service offshoring perform reasonably well in first-stage; instruments for material offshoring are weaker.
  - SIC-level IV/GMM employment estimations have mixed diagnostics; some overidentification tests fail and IV estimates are not uniformly significant.

### Magnitude summary and key statistics
- Service offshoring contribution to labor productivity growth (1992–2000): 11 to 13 percent.
- Material offshoring contribution to labor productivity growth (1992–2000): 3 to 6 percent.
- Offshoring intensity levels:
  - OSS (Share of Imported Service Inputs): 1992 = 0.18%; 2000 = 0.29%.
  - OSM (Share of Imported Material Inputs): 1992 = 11.72%; 2000 = 17.33%.
- Growth rates (annual averages):
  - Service offshoring growth: 6.3 percent.
  - Material offshoring growth: 4.4 percent.
- Depreciation rates used in capital stock construction: 7.7 percent for equipment and 3.5 percent for structures.

### Robustness, endogeneity, and limitations
- Econometric approaches to address endogeneity and measurement error:
  - Arellano-Bond dynamic panel (GMM), two-stage least squares with instruments, robust regressions, longer-differencing approaches.
- Instrument diagnostics:
  - Shea partial R-squared for service offshoring instruments: 0.31 to 0.43.
  - First-stage F-statistic reported: F(5,44) = 14.27, p < 0.01 for certain specifications.
  - Some GMM/IV specifications pass Sargan/Hansen tests while others do not; autocorrelation tests sometimes marginal (e.g., autocorrelation test z =1.85 Pr> z = 0.06).
- Data and inference limitations:
  - Offshoring measures at industry level may mask within-industry heterogeneity.
  - Application of economy-wide import shares to all industries is an imperfect proxy.
  - Measurement error and first-differencing can reduce precision and bias estimates toward zero.
  - Instruments may not remove all measurement error; some coefficients not consistently significant.

### Policy implications and suggested research directions
- Policy-relevant conclusions:
  - Sourcing service inputs from abroad by U.S. firms increased between 1992 and 2000 and has a robust positive effect on manufacturing productivity.
  - Service offshoring accounted for a nontrivial share (11 to 13 percent) of labor productivity growth during 1992–2000; material offshoring accounted for 3 to 6 percent.
  - Employment effects are small in aggregate but detectable at fine industry disaggregation (unweighted loss of 3percent, employment-weighted 0.4of a percent); aggregate 96-industry analysis shows no significant net employment loss.
  - Public concern and political actions (e.g., federal contracting restrictions) reflect perceived job risks, but empirical results show nuanced trade-offs between productivity gains and localized employment impacts.
- Policy interpretation:
  - Productivity gains from service offshoring are substantial and robust to multiple controls and estimators; policies might therefore weigh productivity benefits against localized employment displacements.
- Recommended avenues for future research:
  - Collect firm-level data distinguishing domestic input purchases from imports combined with detailed skill-level data to identify channels through which service offshoring increases productivity.
  - Investigate terms-of-trade and income-distribution effects of offshoring, including potential impacts on the wage skill premium (service offshoring likely more skill-intensive than material offshoring).
  - Use disaggregated skill data to study differential impacts across particular skill groups.

*Source: _wp05238 - 1. Offshoring Intensity: 1992–2000; Section III describes the data. Section IV presents the results and Section V concludes.*

### 1. Offshoring Intensity: 1992–2000 ..............................................................................20

### 1. Offshoring Intensity: 1992–2000

### Background and motivation
- New technologies enable firms to source service inputs from abroad (examples cited: call centers in India, computer software development).
- In the OECD, "the use of imported inputs in producing goods that are exported accounted for 21 percent of trade in 1990, and this grew by 30percent between 1970 and 1990."
- Service offshoring by U.S. manufacturing industries grew at "an average annual rate of6.3percent between 1992 and 2000."
- Political and public concern cited:
  - Support for free trade among white collar workers with incomes over $100,000 slid from "57 percent in 1999 to 28 percent in 2004."
  - "On March 4, 2004, the U.S. Senate passed restrictions on offshoring by barring companies from most federal contracts if they planned to carry out any of the work abroad."
- Research gap: empirical evidence on effects of service offshoring (on productivity and employment) is scant; this paper estimates whether offshoring leads to job losses and whether there are offsetting productivity benefits.

### Data and measurement
- Time coverage: annual data for "1992 to 2000" from Bureau of Labor Statistics (BLS) I/O tables combined with trade data.
- Offshoring measures:
  - Service offshoring: share of imported services in industry inputs (includes imports from affiliated and unaffiliated firms), analogous to Feenstra and Hanson (1999) material offshoring measure.
  - Material offshoring: analogous measure for imported materials.
- Aggregation:
  - Data aggregated from "450 SIC manufacturing industries to 96 manufacturing industries" to match I/O tables that provide service input detail.
  - Analysis also uses the more disaggregated 450-industry division for employment effects.
- Productivity measures: Total factor productivity (TFP) and labor productivity from BLS; note that netting out service inputs is important to avoid missing-input bias.
- Baseline productivity trend: "Labor productivity in manufacturing grew at an annual average rate of 4 percent between 1992 and 2000."

### Identification and instruments
- Endogeneity concerns:
  - Reverse causality: high-productivity firms may choose offshoring; alternatively low-productivity firms may offshore to improve productivity.
  - Measurement error in service offshoring could bias estimates downward.
- Instrumental variables strategy:
  - Service offshoring instrument: country-level time-varying measures of technology that enable offshoring (specifically "the number of internet users in the countries that the United States imports most of its service inputs"), interacted with the industry's initial share of services in output to create time/industry varying instruments.
  - Rationale: technological change (internet penetration) plausibly affects productivity only through enabling service offshoring.
  - Material offshoring instrument: "the freight cost of inputs."
- Expected instrument properties: should "only affect productivity through service offshoring, and have sufficient explanatory power in predicting changes in service offshoring."

### Main empirical findings — productivity
- Service offshoring has "a significant positive effect on productivity in the manufacturing sector."
- Contribution to labor productivity growth:
  - Service offshoring accounts for "around 11percent of labor productivity growth over the sample period."
  - Material offshoring accounts for "5 percent of total labor productivity growth between 1992 and 2000."
- Robustness:
  - Service offshoring results are robust to controls such as use of high-technology capital and the share of total imports.
  - Instrumental variables estimates indicate "a slightly larger positive productivity effect from service offshoring than those indicated by ordinary least squares (OLS)."
  - Material offshoring's positive effect on productivity was "not robust across all specifications" and its magnitude is lower than service offshoring.

### Main empirical findings — employment
- Aggregation and employment effects:
  - At the 96-industry aggregation, "service offshoring has no significant effect on employment."
  - At the more disaggregated 450-industry division, "service offshoring reduced manufacturing employment by around 0.4 of a percent."
  - The negative employment effect "washes out" with modest aggregation; detection requires finer industry disaggregation.
- Theoretical ambiguity noted: offshoring can reduce labor demand via productivity and substitution effects, or raise labor demand via scale and competitiveness effects — empirical results determine the net effect.

### Comparison to prior studies and external estimates
- Prior evidence:
  - Feenstra and Hanson (1996, 1999): focus on material offshoring and the skill wage premium; found material offshoring explained "over 40 percent of the increase in nonproduction wages in the 1980s."
  - Mann (2004): a "back of the envelope" calculation for IT industry finds offshoring led to "an annual increase in productivity of 0.3 percentage points for the period 1995 to 2002," cumulative effect "of $230 billion in additional GDP."
  - European studies show mixed evidence; some find positive productivity effects in specific industries or periods but results are not uniformly robust when extended.
- Contribution: "This is the first comprehensive study to find a link between service offshoring and productivity" for U.S. manufacturing across all industries.

### Implications and interpretation
- Productivity benefits from service offshoring are quantitatively meaningful (accounting for "around 11percent" of labor productivity growth), and instrumental-variable estimates suggest the effect may be larger than OLS implies.
- Employment effects are small in aggregate but detectable at fine industry disaggregation ("around 0.4 of a percent" decline in manufacturing employment in the 450-industry division).
- Policy relevance:
  - Public concern and political actions (e.g., legislation restricting offshoring in federal contracting) reflect perceived job risks, but rigorous empirical results show nuanced trade-offs between productivity gains and localized employment impacts.
  - Identification of channels (technology-enabled service offshoring, freight-cost-driven material offshoring) aids targeted policy and further research.

*Source: _wp05238 - 1. Offshoring Intensity: 1992–2000 (excerpt).*

### Section III describes the data. Section IV presents the results and Section V concludes.

### _wp05238 - Section III describes the data. Section IV presents the results and Section V concludes.

### Model: conceptual framework and channels
- Production function for industry i:
  - Yi = Ai(ossi, osmi) F(Li, Ki, Mi, Si)
  - Technology shifter Ai is a function of offshoring of services (ossi) and offshoring of material inputs (osmi).
- Four channels through which offshoring can affect productivity (Ai):
  - Static efficiency gain (compositional effect when less efficient stages are relocated).
  - Restructuring (remaining workers become more efficient; more likely from service offshoring).
  - Learning externalities (imported services, e.g., software, improve average productivity).
  - Variety effects (use of new material or service input varieties, as in Ethier (1982)).
- Timing and decision structure:
  - Firms choose total amount of each input in the first stage and the proportion imported in a second stage.
  - Fixed cost of importing material inputs, FMk, and fixed cost of importing service inputs, FSk, vary by industry k.
- Labor demand (conditional) under cost minimization:
  - Li = g(wi, ri, qm, qs, Yi) / Ai(ossi, osmi)
  - Offshoring affects labor demand via substitution (input price changes), productivity (Ai), and scale effects (output demand).
  - When substituting profit-maximizing output (pi), labor demand function becomes Li = g(wi, ri, qm, qs, pi, ossi, osmi) / Ai(ossi, osmi).
- Note on empirical expectations:
  - Net effect of offshoring on employment is ambiguous; scale effects may offset substitution and productivity effects.

### Estimation: productivity and labor-demand specifications
- Productivity estimating equation (first differences, logs):
  - ∆lnYit = α0 + α1 ∆ossit + α2 ∆osmit + β1 ∆lnLit + β2 ∆lnKit + β3 ∆lnMit + β4 ∆lnSit + δt Dt + δi Di + εit
  - Time differencing controls for time-invariant industry-specific effects; year fixed effects (δt) control for common time-varying shocks; some specifications also include industry fixed effects (δi).
  - Hypothesis: α1 and α2 are positive.
  - One-period lags of offshoring variables included; footnote: "Longer lags were insignificant."
- Econometric concerns and solutions for productivity:
  - Endogeneity of inputs → estimate using Arellano-Bond (1991) GMM (uses lags as instruments).
  - Alternative: redefine dependent variable as value added per worker to avoid including inputs as regressors.
  - Potential endogeneity of offshoring (self-selection, measurement error) → use two-stage least squares and Arellano-Bond GMM with additional exogenous instruments.
- Employment estimating equations (first differences, log-log):
  - Baseline:
    - ∆lnlit = γ0 + γ1 ∆ossit + γ2 ∆osmit + γ3 ln∆wit + γ4 ∆lnYit + δt Dt + δi Di + εit
    - Identification typically assumes wages are exogenous to industry (labor mobility); industry fixed effects absorb time-invariant industry rents.
  - Scale-adjusted:
    - ∆lnlit = γ0 + γ1 ∆ossit + γ2 ∆osmit + γ3 ln∆wit + γ5 ∆lnpit + δt Dt + δi Di + εit
    - In reduced-form variants pit may be omitted (pit is function of input prices).
  - Estimation method: OLS with robust standard errors corrected for clustering.
- Additional modeling choices and assumptions:
  - Firms are assumed to face same price for other inputs (imported inputs and rental on capital), modeled as r = f(t); in time-differenced equations these input price effects are captured by δt.
  - Offshoring intensity may proxy for lower price of imported service inputs, productivity effect, and/or substitution effect (reference to Amiti and Wei (2005) interpretation).

### Data and measurement of offshoring (1992 to 2000)
- Offshoring intensity of services (ossi,t) for industry i at time t:
  - ossi = Σj [ input purchases of service j by industry i, at time t / total non-energy inputs used by industry i, at time t ] * [ imports of service j, at time t / (productionj + importsj − exportsj at time t) ]
- Construction of components:
  - First bracketed term: annual input/output tables from 1992 to 2000 constructed by BLS, based on BEA 1992 benchmark tables (SIC1987 industry disaggregation, roughly 450 manufacturing industries, aggregated to 96 I/O manufacturing codes by BLS).
  - Five service industries included as inputs to manufacturing: telecommunications; insurance; finance; business services; computing and information.
    - Average shares in 2000:
      - Business services: 12%
      - Finance: 2.4%
      - Telecommunications: 1.3%
      - Insurance: 0.5%
      - Computing and information: 0.4%
    - Business services is the largest component of service inputs with an average share of 12% in 2000.
  - Second bracketed term: international trade data from IMF Balance of Payments yearbooks.
    - Imports and exports by industry unavailable → economy-wide import share applied to each industry.
    - Example: U.S. economy imported 2.2 percent of business services in 2000 → assume each manufacturing industry imports 2.2 percent of its business services that year → average offshoring intensity of business services = 0.12 * 0.022 = 0.3 percent.
  - Analogous measure constructed for material offshoring (osmi).
- Key aggregate statistics (Table 1 referenced in source):
  - Average share of imported service inputs in 2000: 0.3 percent.
  - Average share of imported material inputs in 2000: 17.4 percent.
  - Growth over sample period (annual averages):
    - Service offshoring: 6.3 percent (annual average growth rate).
    - Material offshoring: 4.4 percent (annual average growth rate).
- Additional detail (Table 2 referenced in source):
  - Business services grew from an average of 9.7 percent in 1992 to 12 percent in 2000 (first component of oss).
  - Large cross-industry variation: example in 2000 — "household audio and video equipment" industry: business services = 2 percent of total inputs; "greeting cards" industry: business services = 45 percent of total inputs.
  - Import share of all service categories, except communications, increased over the period (last column of Table 2).
- Data aggregation and capital stock:
  - Productivity estimated at the more aggregate BLS I/O industry level because service inputs by industry are only available from the I/O tables and must be subtracted from gross output.
  - Capital stock data available from the Annual Survey of Manufacturers (ASM) at the SIC level and aggregated up to the I/O level.
  - Capital stock series extended beyond 1996 using the perpetual inventory method with average depreciation rates applied in the NBER (Bartelsman and Gray, 1996) database.
- Recognized limitations of offshoring measures:
  - Likely underestimation because cost of importing services is likely lower than purchasing domestically; quantities are unavailable in U.S. data and current-value measures are used.
  - Applying the same import share to all industries is imperfect but follows Feenstra and Hanson (1996, 1999) methodology; without imports-by-industry data this is treated as the best available proxy.
  - Measurement error in offshoring intensity can contribute to downward bias in estimated effects.

*Italic: Source — _wp05238 - Section III describes the data. Section IV presents the results and Section V concludes.*

### 7.7 percent depreciation for equipment and 3.5 percent for structures.

### _wp05238 - 7.7 percent depreciation for equipment and 3.5 percent for structures.

### Data, sample, and estimation approach
- Sample period: 1992 to 2000 (industry-level estimations); offshoring measures constructed at BLS I/O classification comprising 96 manufacturing industries and analyses also at SIC 4-digit comprising 450 industries.
- Data sources: NBER Productivity database (Bartelsman and Gray, 1996) extended to 2000 using BEA and ASM; trade in services from IMF Balance of Payments; internet hosts/users from International Telecommunication Union (2003) Yearbook; bilateral BEA services shares used to weight country instruments.
- Variables: all variables entered in log first differences except offshoring (change in offshoring intensity); specifications include year fixed effects and in many cases industry fixed effects; errors corrected for clustering at the I/O level.
- Estimation methods: OLS, Arrellano-Bond dynamic panel (GMM), robust regressions, two-stage least squares (IV) using instruments based on ∆ln(Internet hosts) interacted with services intensity (service share in 1992), and freight-and-insurance rates for material offshoring.

### Total Factor Productivity (TFP) — key findings
- Service offshoring has a positive significant effect on total factor productivity across OLS and GMM specifications.
- Contemporaneous and lagged service offshoring are individually and jointly significant (joint p-value < 0.01 in many specifications).
- With industry fixed effects coefficients on service offshoring are larger.
- Material offshoring: positive and significant only in some specifications; lagged material offshoring becomes more prominent under GMM/IV.

Key reported coefficients and diagnostics (preserve reported estimates as stated in source tables):
- Examples from Table 4: ∆oss i,t coefficients reported as 0.235***, 0.249***, 0.241***, 0.341***, 0.331***, 0.335***, 0.258*** across columns; joint tests show strong significance (p-values = 0.00).
- Arrellano-Bond (GMM) column: service offshoring remains positive and significant; lagged dependent variable insignificant in that specification.
- Sargan overidentification test in column (9) estimation: χ2(20)=23.08, p-value=0.28; autocorrelation test z =1.85 Pr> z = 0.06.

### Labor productivity — effects and robustness
- Labor productivity measured as value added per worker (real output minus real materials and services, divided by employment).
- Main result: service offshoring positively correlates with labor productivity; material offshoring results are weaker and often insignificant once industry fixed effects are included.
- Additional controls: inclusion of high-technology capital shares (ex post and ex ante rental price measures) and import share:
  - High-tech capital (ex post) sometimes significant (10% level in some specs); ex ante measure generally insignificant.
  - Import share: tougher import competition has a positive effect on labor productivity in some specifications but inclusion does not erase service offshoring effect.
- Sensitivity checks:
  - Measurement error: two-period differences and long-difference (average of last three years less average of first three years) produce similar positive service-offshoring effects; material offshoring sometimes positive but with much smaller coefficients.
  - Outliers: robust regressions and excluding tobacco and high-tech outliers (computer and electronics) leave service offshoring effect positive; tobacco identified as main outlier and omitted from subsequent estimations.
  - Instrumental variables (IV) and GMM: instruments based on changes in internet hosts/users in major supplier countries interacted with initial services intensity provide good first-stage fit for service offshoring (Shea partial R-squared values between 0.31 to 0.43). IV/GMM maintain positive net effect of service offshoring on productivity though some individual coefficients vary in significance.
- Representative estimates and magnitudes:
  - IV and GMM-derived aggregate effect of service offshoring on productivity ranges reported as 0.3 (Arellano-Bond Table 10 column 6) to 0.38 (one-period differences 2SLS Table 9 column 4) and 0.46 (two-period differenced variables Table 9 column 8).
  - Service offshoring increased by 0.1 percentage point over the sample period, from 0.18 to 0.29 (Table 1).
  - Interpreted implication: service offshoring led to an increase of 3 to 4.5 percent in labor productivity over the sample period.
  - Given value added per worker increased by an average of 35 percent over the sample period, service offshoring accounted for 11 to 13 percent of the average growth in labor productivity.
  - Material offshoring accounted for 3 to 6 percent of labor productivity growth (statement in conclusions).

### Employment — effects, disaggregation, and instruments
- Aggregate (96-industry BLS I/O level) results:
  - Service offshoring has no significant effect on manufacturing employment in the 96-industry specifications (Tables 11 and 12).
  - Material offshoring often shows a positive effect on employment; significance varies across specifications.
  - Instrumental variables and GMM in employment specifications do not show a negative significant effect from offshoring; some IV specifications show positive effects from service offshoring and consistent positive effects from material offshoring.
- More disaggregated results (450 SIC 4-digit industries):
  - Using 450 industries, service offshoring shows a significant negative effect on employment in Table 13 across specifications (one- and two-period differences).
  - Two-period differenced specifications (Table 13 columns 4–6) show persistent negative service-offshoring effects with no offsetting scale effects.
  - Material offshoring becomes insignificant at the more disaggregated level.
  - Using estimates from Table 13 with scale effects, the effect from service offshoring on employment equals 0.3. Given service offshoring grew by 0.1 percentage point over the sample period, this implies a loss of 3percent employment. Weighted by employment shares this number falls to 0.4of a percent.
- Instrument concerns and IV performance:
  - Instruments based on internet host/user changes interacted with initial service intensity perform well for service offshoring first-stage; instruments for material offshoring (freight/insurance) provide weaker first-stage fit.
  - In the SIC-level IV/GMM employment estimations (Table 14) instruments fail some overidentification tests; IV estimates often negative for service offshoring but not always significant.

### Magnitude summary and interpretation
- Service offshoring contribution to labor productivity growth (1992–2000):
  - Service offshoring: accounts for 11 to 13 percent of labor productivity growth.
  - Material offshoring: accounts for 3 to 6 percent of labor productivity growth.
- Change in offshoring intensity (Table 1):
  - Share of Imported Material Inputs (OSM): 1992 = 11.72%; 2000 = 17.33%; 1992-2000 change reported as 4.38 (in table summary).
  - Share of Imported Service Inputs (OSS): 1992 = 0.18%; 2000 = 0.29%; 1992-2000 change reported as 6.26 (in table summary).
- Employment implications:
  - At fine industry disaggregation (450 SIC): estimated employment loss implied = 3percent (unweighted), falling to 0.4of a percent when weighted by employment shares.
  - At aggregate (96 industry) level: no significant net employment effect — negative effects in some industries offset by growth elsewhere within broad classifications.

### Robustness, endogeneity, and limitations
- Endogeneity and measurement error addressed via:
  - Arrellano-Bond dynamic panel (GMM).
  - Instrumental variables using ∆ln(Internet hosts/users) by partner country interacted with initial industry service intensity; freight-and-insurance rates weighted by input shares for material offshoring.
- Instrument diagnostics:
  - First-stage F(5,44) = 14.27 with p-value less than 0.01 for lagged service offshoring when only one endogenous variable included.
  - Shea partial R-squared for service offshoring instruments ranges 0.31 to 0.43; instruments for material offshoring weaker (Shea partial R-squared lower).
  - Hansen/J and Sargan tests reported across GMM/IV tables; some specifications pass overidentification tests while others fail certain tests at conventional levels.
- Data and inference limitations:
  - Offshoring measures are at the industry level (aggregation may mask within-industry heterogeneity).
  - Potential measurement error when taking first differences — addressed by longer differencing and averaging but sample size reduced.
  - Instruments may not capture all measurement error; some coefficients not highly significant possibly because instruments cannot remove all errors induced by first differencing.

### Policy-relevant conclusions and suggested research directions
- Main empirical conclusions:
  - Sourcing service inputs from abroad by U.S. firms increased between 1992 and 2000 and has a robust positive effect on manufacturing productivity.
  - Service offshoring accounted for a nontrivial share (11 to 13 percent) of labor productivity growth during 1992–2000; material offshoring accounted for 3 to 6 percent.
  - Employment effects depend on aggregation: finely disaggregated industries show a small negative employment effect (estimated unweighted loss of 3percent, weighted 0.4of a percent), whereas aggregate industry classifications (96 industries) show no significant net employment loss.
- Policy implications (from analysis):
  - Productivity gains from service offshoring are substantial and robust to multiple controls and estimators; policies might therefore weigh productivity benefits against localized employment displacements.
- Recommended avenues for future research (as stated):
  - Collect firm-level data distinguishing domestic input purchases from imports combined with detailed skill-level data to identify channels through which service offshoring increases productivity.
  - Investigate terms-of-trade and income-distribution effects of offshoring, including potential impacts on the wage skill premium (service offshoring likely more skill-intensive than material offshoring).
  - Use disaggregated skill data to study differential impacts across particular skill groups.

*Source: _wp05238 - 7.7 percent depreciation for equipment and 3.5 percent for structures. Canonical URL: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05238.pdf*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05238.pdf_