## _wp03248

## Source details

**Canonical URL:** [_wp03248](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03248.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03248.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03248.pdf.json)

---

### Context and motivation
- Productivity differences across countries largely reflect productivity variation (Hall and Jones, 1999; Trefler, 1995).
- Cross-country productivity variation arises from foreign as well as domestic innovations (Eaton and Kortum, 1999; Keller, 2002a), motivating study of international technology transfer.
- FDI and trade are suspected conduits of international technology transfer; both have grown faster than GDP recently.
- Foreign-owned companies account for almost one-sixth of U.S. manufacturing GDP.
- Example cited: the U.S. state of Alabama spent $230 million ($150,000 per newly created job) to attract a Mercedes plant in 1994 (Head, 1998).

### Data and sample
- Sample: about 1,300 U.S. firms for the years 1987 to 1996.
- Final cleaned sample: 1,277 U.S.-owned firms active between 1987 and 1996.
- Coverage: about 40 percent of U.S.-owned manufacturing employment and roughly 55 percent of U.S.-owned manufacturing research and development expenditures in the United States.
- Core data sources: Standard and Poor’s Compustat; Bartelsman and Gray (2001) NBER Productivity database; Feenstra (2002) imports; confidential BEA affiliate-level annual survey data for foreign employment.

### Mechanisms of technology spillovers
- Import-related mechanisms:
  - Learning from imported technologically advanced commodities enabling domestic production at lower cost.
  - Imported prices may not fully reflect quality due to buyer market power or seller appropriability problems.
- FDI-related mechanisms:
  - Worker mobility transferring MNE affiliate knowledge (Fosfuri, Motta, and Rønde, 2001).
  - Access to specialized intermediate inputs through multinationals (Rodriguez-Claré, 1996).
  - Productivity improvements via local intermediate suppliers learning from MNE activity.
- General point: market prices may not reflect full benefits and costs when externalities exist.

### Model and estimation framework
- Productivity measurement uses Olley and Pakes (1996) and Ericsson and Pakes (1995) framework addressing simultaneity and selection via investment and exit decisions.
- Production equation and estimation (notations preserved in source):
  - Equation (1): ,0ititkitmitlit ukmly++++=ββββ
  - Error decomposition (2): .ititit uηω+=
  - Investment inversion (3): ).,( tttt kih=ω
  - Reduced form for variable inputs (4): ,),(( tttttmtlt kimlyηφββ+++= with ),(),( 0ttttkttt kihkki++=ββφ.
  - Capital coefficient identification assumption: ,1−−= ttt ωωξ leading to final estimation equation (6): .)ˆ,ˆ( ˆˆ 11 tttttkttktmtlt Pkgkmlyηξβφβββ++−+=−− −−
- Estimation details:
  - Fourth-order polynomial in investment and capital; survival probability Ptˆ estimated by probit on a fourth-order polynomial in lagged capital and investment.
  - Baseline firm TFP: ititkitmitlit kmlytfpβββˆˆˆ −−−=.
- Empirical linking of TFP growth to foreign activity:
  - Equation (7): ,'21ititititit eFIIMXtfp+∆+∆+=∆γγβ
  - Baseline estimation equation (8): .21 25242321 ititit ititititittjit FIIM SMFMMSCUrtfp εγγ βββββαα +∆+∆+ ++∆+∆+++=∆ −−−
  - ∆ denotes one-year differences; itFI∆ is change in share of foreign-affiliate employment in total employment in consecutive years.

### Variables and measurement
- Firm-level variables:
  - Output y: net sales (Compustat data item 12), deflated by three-digit SIC deflators from Bartelsman and Gray (2001).
  - Labor l: Compustat data item 29.
  - Capital k: Compustat data item 8; capital deflators from Bureau of Labor Statistics / BEA satellite accounts.
  - Materials m: Compustat data item 41 + data item 189 − data item 14 − wage expenditures (wage expenditures = L * average industry wage from Bartelsman and Gray (2001)).
  - R&D r_it: Compustat data item 46; deflators from BEA satellite accounts until 1992, thereafter estimated.
- Industry-level foreign presence measures:
  - IM_it: share of U.S. imports in imports plus total shipments of the industry (Feenstra (2002) imports; Bartelsman and Gray (2001) shipments).
  - FI_it: share of foreign affiliate employment in total employment of the industry (confidential BEA affiliate-level data aggregated by activity).
- Controls: capacity utilization (CU); firm market share (MS); firm mark-up (FM); industry mark-up (SM); industry fixed effects αj and time fixed effects αt.

### Descriptive trends and key statistics
- Import share (imports to imports plus shipments) for sample firms:
  - 1987: 12.9 percent
  - 1996: 16.5 percent
  - Increase in import share was more or less monotonic over 1987–1996.
- Share of U.S. manufacturing employment accounted for by foreign-owned affiliates:
  - 7.7 percent (in 1987)
  - 11.7 percent (in 1996)
  - Change: 4.0 percentage points (1987 to 1996)
- FDI growth phases (1987–1996):
  - 1987 to 1993: FDI grew from 7.7 percent to 11.6 percent.
  - 1996: FDI share 11.7 percent.
- Industry examples (Import Share in %; FDI Share in %; change 1996/87):
  - Apparel and other textile: Import Share 24.7 (1987), 29.1 (1992), 33.4 (1996); Change 8.7. FDI Share 1.1, 3.2, 4.5; Change 3.4.
  - Motor vehicles: Import Share 29.3, 26.0, 26.7; Change -2.6. FDI Share 6.6, 11.0, 14.6; Change 8.0.
  - Electronics: Import Share 20.6, 25.2, 27.3; Change 6.7. FDI Share 12.0, 17.2, 18.6; Change 6.6.
  - Additional industry-level figures presented in the source (Table 2).

### Olley-Pakes production-function elasticities (preferred estimates)
- Preferred O-P specification O-P (1) (includes trend and trend*SIC357 for computers):
  - Capital elasticity: 0.188 (standard error (0.026))
  - Labor elasticity: 0.301 (standard error (0.011))
  - Materials elasticity: 0.594 (standard error (0.010))
  - Scale elasticity: 1.083
- O-P (2) (without trends):
  - Capital elasticity: 0.213 (0.029)
  - Labor elasticity: 0.295 (0.012)
  - Materials elasticity: 0.607 (0.012)
  - Scale elasticity: 1.115
- OLS first differences:
  - Capital elasticity: 0.041 (0.018)
  - Labor elasticity: 0.418 (0.016)
  - Materials elasticity: 0.467 (0.017)
  - Scale elasticity: 0.926
- Interpretation:
  - OLS estimates show lower capital and materials elasticities consistent with simultaneity and selection bias.
  - Olley-Pakes O-P(1) estimates used to compute baseline firm TFP measures.

### Baseline regression results (one-year differences) — foreign activity effects
- Sample for preferred specification: data for 1,149 firms with full contemporaneous and lagged foreign activity variables.
- Preferred specification (Table 4, column 1) FDI coefficients:
  - FDI Current: 0.213 (0.083)
  - FDI Lagged one: 0.303 (0.083)
  - FDI Lagged two: -0.049 (0.086)
  - Sum (total FDI effect): 0.516
  - Joint F-test: three FDI coefficients statistically significant as a group.
- Preferred specification (Table 4, column 1) Imports coefficients:
  - Imports Current: 0.480 (0.200)
  - Imports Lagged one: 0.755 (0.207)
  - Imports Lagged two: -0.236 (0.172)
  - Three coefficients jointly significant; two-year lagged import coefficient not individually significant.
- Interpretation:
  - Current and one-year lagged FDI growth associated with faster TFP growth; effect largely realized within two years.
  - Current and one-year lagged imports measures positive but import results judged less robust than FDI.
- Sensitivity:
  - Omitting industry fixed effects (column 5) makes FDI cumulative effect negative and import coefficients larger and individually significant — highlighting importance of industry fixed effects.
  - Excluding U.S. multinational firms modestly changes coefficients; cumulative impact of FDI somewhat larger when multinational firms are excluded.

### Robustness checks and alternative specifications
- Longer-difference specifications (Table 5):
  - FDI coefficients:
    - 1-year: 0.198 (0.076)
    - 2-year: 0.379 (0.134)
    - 3-year: 0.411 (0.177)
  - Imports coefficients:
    - 2-year: 0.341 (0.282) (not significant)
    - 3-year: 0.210 (0.297) (not significant)
  - Conclusion: FDI effect persists or increases in longer-difference specifications; imports less robust.
- Alternative measures normalizing by lagged total activity (Table 6, full sample column 1):
  - Change in FDI EMP Current: 0.151 (0.089); Lagged one: 0.205 (0.086); Lagged two: 0.067 (0.088).
  - Change Imports Current: 0.219 (0.120); Lagged one: 0.045 (0.125); Lagged two: -0.233 (0.122).
  - Change total EMP Current: -0.479 (0.094); Lagged one: -0.070 (0.079); Lagged two: -0.187 (0.078).
  - Change total sales Current: 0.421 (0.042); Lagged one: 0.036 (0.039); Lagged two: 0.123 (0.046).
  - Interpretation: FDI remains associated with TFP growth; imports weaker; increases in real sales associated with faster TFP growth while increases in total employment associated with slower TFP growth.
- Instrumental variables (Table 7):
  - Instruments: one-year differences in transport costs and tariffs for seven developed countries (1985–1996) plus lagged levels of import and FDI shares.
  - IV1 results (Baseline1 IV1, full sample):
    - FDI Current: 0.198 (0.076)
    - FDI Lagged one: 0.166 (0.077)
    - Imports Current: 0.071 (0.173)
    - Imports Lagged one: 0.554 (0.185)
    - Hansen J-statistic p-value: 0.740 (instruments not rejected)
  - Authors note instruments are weak and do not rely primarily on IV estimates; IV results broadly consistent with positive FDI effects and limited imports effects.
- Subperiod analysis (1990–1992 vs 1994–1996):
  - FDI coefficients larger in 1994–1996 than in 1990–1992.
  - FDI spillovers present throughout period but stronger in mid-1990s.

### Heterogeneity and measurement issues
- High-tech vs low-tech samples (Table 9):
  - High-tech sample FDI coefficients:
    - Current: 0.343 (0.125)
    - Lagged one: 0.413 (0.140)
    - Lagged two: 0.259 (0.187)
    - F-test (FDI) = 6.31, p-value (0.003)
  - Low-tech sample FDI coefficients:
    - Current: -0.012 (0.098)
    - Lagged one: 0.173 (0.097)
    - Lagged two: -0.135 (0.085)
    - F-test (FDI) = 1.72, p-value (0.162)
  - Full sample for comparison:
    - Current: 0.213 (0.083)
    - Lagged one: 0.303 (0.083)
    - Lagged two: -0.049 (0.086)
    - F-test (FDI) = 7.06, p-value (0.0001)
  - Interpretation: spillovers stronger in high-tech industries; sample composition partly explains aggregate spillover estimates.
- Measurement of FDI aggregation matters (Table 10):
  - Activities aggregation (preferred) FDI sum of significant coefficients: 0.516.
  - Main-line aggregation FDI sum of significant coefficients: 0.000.
  - Conclusion: mismeasurement of industry FDI via affiliate main-line of business biases spillover estimates toward zero; activity-level aggregation yields substantially larger and statistically significant FDI spillovers.

### Magnitude and economic importance of estimated spillovers
- Preferred specification (Table 4, first specification) significant FDI coefficients:
  - Current FDI: 0.213
  - One-year lagged FDI: 0.303
  - Sum (total FDI effect): 0.516
- Olley-Pakes implied average productivity growth in the sample: 0.19 over 1987–1996.
- Authors’ calculation of share of productivity growth accounted for by FDI spillovers: about 11 percent (calculation cited in text: 0.516*0.04/0.19).
- Analogous calculation for imports (based on specification one in Table 4) suggests imports account for about 9.5 percent of productivity growth over the sample period, though import estimates are noted as less robust.

### Interpretation relative to prior literature
- Conventional literature often finds FDI spillovers of, at best, minor importance; these results challenge that view.
- Authors attribute higher spillover estimates primarily to improved measurement of foreign multinational activity (activity-level BEA data) rather than sample selection.
- If improved measurement is the driver, results may generalize when similar data become available elsewhere.
- Literature review summary:
  - Imports-related spillovers: evidence mixed and often weaker with micro data and behaviorally explicit econometrics.
  - FDI-related spillovers: mixed case-study evidence; panel/micro studies generally find small or limited positive effects; this paper finds stronger positive effects using improved measurement and Olley-Pakes TFP.

### Policy implications and suggested research directions
- Economic significance:
  - FDI spillovers estimated here are economically important—accounting for about 11 percent of productivity growth in the sample—supporting the argument that technology spillovers associated with FDI can matter substantially for productivity growth and welfare.
  - Results provide the strongest evidence the authors are aware of that may support the provision of subsidies to attract FDI from a social welfare viewpoint.
- Cautions:
  - Whether a socially optimal policy is implemented depends on political-economic realities of local electoral competition.
  - Large public subsidies to attract multinationals are common and justified only if FDI or imports generate substantial positive externalities or technology spillovers for domestic firms.
  - Value of FDI incentives characterized as the present discounted value of the sum of direct subsidies (public infrastructure, tax reductions).
- Recommended future research:
  - Focus on specific industries and mechanisms given heterogeneity in spillover strength and motivations for FDI.
  - Broaden view of MNE entry effects to include inter-industry spillovers, longer-run effects (e.g., worker training programs), and signaling effects to other potential foreign investors.
  - Improve measurement of FDI at activity level in other settings to test generalizability.

### Data, sample construction, and variable definitions (high-level)
- Sales (Y): Compustat data item 12; deflated by industry-level price index from Bartelsman and Gray (2001).
- Labor (L): Compustat data item 29.
- Capital (K): Compustat data item 8; deflators from BEA satellite accounts.
- Materials (M): Compustat data item 41 + data item 189 − data item 14 − wage expenditures (wage expenditures = L * average industry wage from Bartelsman and Gray (2001)).
- R&D (R): Compustat data item 46; deflators from BEA satellite accounts until 1992, thereafter estimated.
- Import share (IM): Feenstra (2002) imports by industry over imports plus total shipments (Bartelsman and Gray 2001).
- FDI share (FI): Foreign affiliate employment by industry aggregated from activity divided by total U.S. employment by BEA industry; source: confidential BEA affiliate-level FDI data.
- Sample selection:
  - Initial universe: all manufacturing firms active between 1987 and 1996.
  - Foreign-owned firms removed; data cleaned for implausible year-to-year changes.
  - Final sample for firms reporting output and inputs including R&D expenditures: 1,277 firms.
  - Working sample observations vary by specification (examples in source: Number of observations 58,995; Firms 11,149 in some specifications; other splits reported for high-tech/low-tech).

*Source: Appendix I. Variable Definitions, Sources, and Data Construction, _wp03248*

### Appendix I. Variable Definitions, Sources, and Data Construction ........................................35

### Appendix I. Variable Definitions, Sources, and Data Construction ........................................35

### Context and motivation
- Most variation in living standards across countries can ultimately be traced back to differences in productivity (Hall and Jones, 1999 and Trefler, 1995).
- Cross-country productivity variation is due at least as much to foreign innovations as to domestic innovation (Eaton and Kortum, 1999 and Keller, 2002a), implying the need to understand international technology transfer.
- Foreign direct investment (FDI) and trade are long-suspected conduits of international technology transfer; both have grown faster than GDP recently.
- Foreign-owned companies account for almost one-sixth of U.S. manufacturing GDP.
- Governments spend substantial resources to attract foreign multinationals; one example cited: the U.S. state of Alabama spent $230 million ($150,000 per newly created job) to attract a new Mercedes plant in 1994 (Head, 1998).

### Data and sample
- Sample: about 1,300 U.S. firms for the years 1987 to 1996.
- Focus: measuring spillovers associated with imports and the activities of multinational enterprise (MNE) affiliates that constitute FDI.
- Emphasis on improved measurement of foreign multinational activity as a key methodological advance.

### Key findings and quantitative results
- According to the authors' preferred estimates, FDI spillovers accounted for more than 10 percent of U.S. productivity growth over the period 1987 to 1996.
- Evidence for imports-related technology spillovers exists, but overall the evidence on imports is described as less conclusive than for FDI.

### Interpretation and comparison with literature
- Conventional wisdom in the literature is that FDI spillovers are, at best, of minor economic importance (see Section II); the authors' results revisit and challenge this view.
- The authors argue that their higher spillover estimates relative to much of the prior literature are primarily due to improved measurement of foreign multinational activity rather than to idiosyncratic sample selection.
- If improved measurement is the driver, the authors suggest the results are more likely to generalize once similar data are available in other contexts.

### Policy-relevant observations
- Policy prescriptions from international organizations such as the World Bank and the World Trade Organization criticize discriminatory policies toward foreign investors and exporters.
- Large public subsidies to attract multinationals are common and are justified only if FDI or imports generate substantial positive externalities or technology spillovers for domestic firms.
- The value of FDI incentives is characterized as typically the present discounted value of the sum of direct subsidies, e.g., publicly provided infrastructure and tax reductions.

*Source: Appendix I. Variable Definitions, Sources, and Data Construction, _wp03248*

### Section III. Section IV gives an overview of the data, with more detail provided in the

### _wp03248 - Section III. Section IV gives an overview of the data, with more detail provided in the

### Technology spillovers through trade and foreign direct investment (FDI)
- Mechanisms discussed for import-related spillovers:
  - Importing a technologically advanced commodity may trigger learning enabling domestic producers to manufacture a similar good at lower cost.
  - Imported prices may not fully reflect quality because of buyer market power or seller appropriability problems.
- Mechanisms discussed for FDI-related spillovers:
  - Worker mobility: workers embodying MNE affiliate knowledge may be attracted to domestic firms (Fosfuri, Motta, and Rønde, 2001).
  - Access to specialized intermediate inputs through multinationals (Rodriguez-Claré, 1996).
  - Productivity improvements via local intermediate suppliers whose know-how increases through MNE activity.
- General point: market prices may not reflect the full benefits and costs when such externalities exist.

### Summary of existing empirical evidence (literature review)
- Imports-related spillovers:
  - Coe and Helpman (1995): productivity related positively to import-share weighted R&D of trade partners.
  - Xu and Wang (1999): strengthened results by focusing on machinery imports.
  - Keller (1998): obtained almost as strong results using counterfactual import data, suggesting limited robustness of import-based spillover evidence from these regressions.
  - Industry-level and micro-data results mixed: Keller (2002b) finds imports-related effects among G-7 industries; Kraay, IsolagA, and Tybout (2001) do not find such effects in three less developed countries.
  - Conclusion: some evidence for import-related spillovers, but not ubiquitous and weaker with micro data and behaviorally explicit econometrics.
- FDI-related spillovers:
  - Case studies produce mixed results (e.g., Larrain, Lopez-Calva, and Rodriguez-Claré (2000) vs. Hanson (2001)).
  - Panel/micro studies:
    - Aitken and Harrison (1999): increased presence of foreign-owned affiliates associated with lower productivity in Venezuelan plants (late 1970s–1980s); attributed to strong competition and average cost effects.
    - Girma and Wakelin (2001), Haskel, Pereira, and Slaughter (2001): find positive but small FDI spillovers in the United Kingdom; Haskel et al. estimate FDI spillovers account for about 5 percent of the TFP growth in British manufacturing in the two decades from 1973 to 1992.
  - Patent-citation approach: Branstetter (2000) shows FDI between the U.S. and Japan associated with higher knowledge flows measured by patent citations.
  - Inter-industry spillovers studied by Blalock and Gertler (2002) and Kugler (2002).
  - Overall conclusion in this literature: no evidence for strong positive technology spillovers associated with FDI; positive effects are limited and often small in economic magnitude.

### Model and estimation framework
- Objective: test whether domestic firm productivity is higher in industries with more foreign activity (imports and FDI) without specifying a single transmission mechanism.
- Productivity measurement approach: Olley and Pakes (1996) and Ericsson and Pakes (1995) framework to address simultaneity of input choice and selection biases via firm investment and exit decision modeling.
- Production equation (notations preserved):
  - Equation (1): ,0ititkitmitlit ukmly++++=ββββ
  - Error decomposition (2): .ititit uηω+=
  - Investment inversion (3): ).,( tttt kih=ω
  - Reduced form for variable inputs (4): ,),(( tttttmtlt kimlyηφββ+++= with ),(),( 0ttttkttt kihkki++=ββφ.
  - Capital coefficient identification uses assumption ωt is a random walk: ,1−−= ttt ωωξ leading to equation (5) and then accounting for selection via survival probability Ptˆ in equation (6):
    - Final estimation equation (6): .)ˆ,ˆ( ˆˆ 11 tttttkttktmtlt Pkgkmlyηξβφβββ++−+=−− −−
- Estimation details:
  - Use a fourth-order polynomial in investment and capital to approximate unknown functions (Robinson (1988) partially linear regression framework).
  - Survival probability estimated by probit on a fourth-order polynomial in lagged capital and investment; Ptˆ used to control for selection.
  - Baseline firm TFP measure: ititkitmitlit kmlytfpβββˆˆˆ −−−=.
  - Alternative productivity measures: use industry cost shares (,,,lmkv v = β) and a one-step OLS regression of outputs on inputs.
- Empirical specification linking TFP growth to changes in foreign activity:
  - Equation (7): ,'21ititititit eFIIMXtfp+∆+∆+=∆γγβ where IM∆ is change in import share and FI∆ is change in foreign-affiliate employment share; X'it are controls.
  - Baseline estimation equation (8) (preserved ordering and notation):
    - .21 25242321 ititit ititititittjit FIIM SMFMMSCUrtfp εγγ βββββαα +∆+∆+ ++∆+∆+++=∆ −−−
  - ∆ denotes one-year differences; itFI∆ is change in share of foreign-affiliate employment in total employment in consecutive years.

### Data
- Source: Standard and Poor’s Compustat database for an unbalanced sample of publicly traded U.S. manufacturing firms.
- Sample construction and coverage:
  - Sample period: 1987 to 1996.
  - Final cleaned sample: 1,277 U.S.-owned firms active between 1987 and 1996.
  - Coverage: about 40 percent of U.S.-owned manufacturing employment and roughly 55 percent of U.S.-owned manufacturing research and development expenditures in the United States.
- Variables and measurement:
  - Firm-level: (log) output y (net sales), (log) labor l, materials m, capital k; R&D expenditures r_it.
  - Output deflation: common deflator at the three-digit SIC level constructed from Bartelsman and Gray (2001) NBER Productivity database.
  - Capital deflators: Bureau of Labor Statistics.
  - Materials data: estimated netting out capital depreciation and labor costs from cost of goods sold and administrative and selling expenses.
  - Industry-level foreign presence measures:
    - IM_it: share of U.S. imports in imports plus total shipments of the industry (imports importance for a firm).
    - FI_it: share of foreign affiliate employment in total employment of the industry (FDI importance).
  - Industry classification: two- to three-digit SIC mapped to roughly 50 BEA manufacturing industries (Table 1 lists BEA codes and names).
  - Imports data source: Feenstra (2002).
  - Total shipments and employment by industry: Bartelsman and Gray (2001).
  - Foreign employment data: confidential BEA affiliate-level annual survey data aggregated to industry; employment figures are based on activity classification of individual affiliate employees.
- Controls included to isolate spillovers:
  - Capacity utilization (CU).
  - Firm market share (MS), firm mark-up (FM), industry mark-up (SM) to capture degree of competition (following Nickell (1996)).
  - Industry fixed effects αj and time fixed effects αt included.
- Endogeneity concerns:
  - FDI or imports may be endogenous to productivity trends (e.g., FDI attracted to fast-growing or weak industries).
  - Instrumental variable estimation used as robustness check, but instruments for industry-level changes are noted as not very powerful.
- Data limitations and measurement error:
  - Compustat covers only publicly traded firms; matching firms to industries is imperfect because large firms span several industries (measurement error discussed).
  - Zero-investment problem minimal in this sample (note on Levinsohn and Petrin (2001) alternative).

### Empirical analysis — main trends and descriptive facts
- Industry heterogeneity in labor input trends:
  - Three industries with average annual labor declines greater than 5 percent: Grain mill products [SIC 204], Beverages [SIC 208], and Apparel [SIC 230].
  - Four industries with average annual employment growth greater than 5 percent: Drugs [SIC 283], Metal cans [SIC 341], Farm and garden machinery [SIC 352], Specialized industry machinery [SIC 355].
- Import exposure trends for firms in the sample:
  - Import share (imports to imports plus shipments) in 1987: 12.9 percent.
  - Import share in 1996: 16.5 percent.
  - Annual growth of imports faced by these firms was almost twice as high as growth in industry shipments.
  - Increase in import share was more or less monotonic over the period.
  - Cross-industry variation example: between 1987 and 1996 the import share in apparel grew by 8.7 percentage points, whereas it fell for motor vehicles by about [text truncated in source].
- Authors’ preliminary synthesis:
  - Large cross-industry variation in inputs, imports exposure, and foreign presence motivates the econometric strategy that follows.
  - The Olley and Pakes method is used to construct TFP, while controls, fixed effects, and robustness checks (including IV where possible) are used to address endogeneity and measurement concerns.

*Source: _wp03248 - Section III. Section IV gives an overview of the data, with more detail provided in the*

### 2.6 percentage points, as Table 2 indicates.

### _wp03248 - 2.6 percentage points, as Table 2 indicates.

### Findings
- 2.6 percentage points, as Table 2 indicates.
- The share of U.S. manufacturing employment accounted for by foreign-owned affiliates has been growing over time as well, from 7.7 percent in 1987 to 11.7 percent in

### Key statistic
- 7.7 percent (in 1987)
- 11.7 percent

*Source: _wp03248 - 2.6 percentage points, as Table 2 indicates.*

### 1996. However, in this case, we can distinguish two separate phases of FDI dynamics.

### _wp03248 - 1996. However, in this case, we can distinguish two separate phases of FDI dynamics.

### FDI dynamics and import exposure (1987–1996)
- Between 1987 and 1993, FDI grew from 7.7 percent to 11.6 percent.
- In 1996 the FDI share was 11.7 percent.
- Industry differences (Import Share in %; FDI Share in %; change 1996/87):
  - Manufacturing: Import Share 12.9 (1987), 14.0 (1992), 16.5 (1996); Change 3.6. FDI Share 7.7 (1987), 11.5 (1992), 11.7 (1996); Change 4.0.
  - Food and kindred products: Import Share 3.7, 3.7, 4.1; Change 0.4. FDI Share 8.4, 11.9, 9.9; Change 1.5.
  - Textile mill products: Import Share 8.1, 8.8, 10.1; Change 2.0. FDI Share 3.7, 6.7, 7.3; Change 3.6.
  - Apparel and other textile: Import Share 24.7, 29.1, 33.4; Change 8.7. FDI Share 1.1, 3.2, 4.5; Change 3.4.
  - Wood and furniture: Import Share 7.6, 8.5, 11.2; Change 3.6. FDI Share 1.9, 2.6, 2.1; Change 0.2.
  - Paper: Import Share 8.4, 8.0, 9.0; Change 0.6. FDI Share 6.9, 7.5, 8.8; Change 1.9.
  - Printing and publishing: Import Share 1.1, 1.2, 1.5; Change 0.4. FDI Share 5.4, 6.6, 7.3; Change 1.9.
  - Chemicals: Import Share 7.6, 9.2, 11.4; Change 3.8. FDI Share 26.2, 32.1, 31.2; Change 5.0.
  - Rubber and plastic: Import Share 5.7, 7.5, 8.6; Change 2.9. FDI Share 6.6, 14.8, 15.4; Change 8.8.
  - Stone, glass, and mineral: Import Share 8.1, 9.5, 10.5; Change 2.4. FDI Share 14.5, 20.9, 21.6; Change 7.1.
  - Primary metals: Import Share 14.8, 15.0, 18.1; Change 3.3. FDI Share 12.2, 15.9, 14.4; Change 2.2.
  - Fabricated metals: Import Share 4.6, 5.6, 6.6; Change 2.0. FDI Share 4.1, 8.3, 9.4; Change 5.3.
  - Industrial machines: Import Share 17.9, 22.9, 24.5; Change 6.6. FDI Share 5.9, 11.3, 11.2; Change 5.3.
  - Electronics: Import Share 20.6, 25.2, 27.3; Change 6.7. FDI Share 12.0, 17.2, 18.6; Change 6.6.
  - Motor vehicles: Import Share 29.3, 26.0, 26.7; Change -2.6. FDI Share 6.6, 11.0, 14.6; Change 8.0.
  - Other transport: Import Share 7.4, 9.2, 12.9; Change 5.5. FDI Share 1.0, 4.9, 4.2; Change 3.2.
  - Instruments: Import Share 11.7, 12.5, 15.6; Change 3.9. FDI Share 7.4, 11.9, 13.3; Change 5.9.
- Overall correlation in Table 2 between changes in import and FDI tends to be negative, although not significantly so.
- Note: In year 2000, the share of foreign employment in U.S. manufacturing was 14.4 percent (source references within original text).

### A. Olley-Pakes production-function elasticities
- Preferred O-P specification O-P (1) (includes trend and trend*SIC357 for computers):
  - Capital elasticity: 0.188 (standard error (0.026))
  - Labor elasticity: 0.301 (standard error (0.011))
  - Materials elasticity: 0.594 (standard error (0.010))
  - Scale elasticity: 1.083
- O-P (2) (without trends):
  - Capital elasticity: 0.213 (0.029)
  - Labor elasticity: 0.295 (0.012)
  - Materials elasticity: 0.607 (0.012)
  - Scale elasticity: 1.115
- OLS first differences:
  - Capital elasticity: 0.041 (0.018)
  - Labor elasticity: 0.418 (0.016)
  - Materials elasticity: 0.467 (0.017)
  - Scale elasticity: 0.926
- Interpretation:
  - OLS estimates show significantly lower capital and materials estimates, consistent with simultaneity and exit leading to downward bias on the capital coefficient.
  - For this sample, increasing returns (scale elasticity > 1) is more plausible than constant or decreasing returns.
  - The Olley-Pakes O-P(1) estimates are used to compute baseline firm TFP measures.

### B. Baseline regression results (one-year differences)
- Estimation approach:
  - Equation (8) estimated using one-year differences.
  - Foreign activity measures: contemporaneous and lagged one and two years for imports and FDI.
  - Industry indicator variables included in preferred specifications; year dummies always included.
  - Standard errors heteroskedasticity consistent and clustered by firm.
- Sample size and panels:
  - Data for 1,149 firms for specification with full set of contemporaneous and lagged foreign activity variables.
- Control variables (coefficients and standard errors reported in Table 4; select findings):
  - R&D (lagged one period): coefficient positive but only marginally statistically significant.
  - Capital utilization: negative coefficient, not precisely estimated.
  - Two-year lagged change in firms’ market share: negative.
  - Firm mark-up: negative and significant.
  - Industry average markups (lagged two years): positive and statistically significant.
- Foreign activity (FDI) coefficients (rows six through eight in Table 4):
  - FDI Current: column 1 shows 0.213 (0.083).
  - FDI Lagged one: column 1 shows 0.303 (0.083).
  - FDI Lagged two: column 1 shows -0.049 (0.086).
  - Interpretation:
    - Current and one-year lagged FDI growth associated with faster TFP growth; two-year lagged growth associated with slower TFP growth.
    - Only current and one-year lagged variables statistically significant individually.
    - Joint test (F-test) indicates the three FDI coefficients are statistically significant as a group.
    - Suggests FDI-related spillovers have been fully reflected in domestic TFP within two years.
- Imports coefficients (rows nine through eleven in Table 4):
  - Imports Current: column 1 shows 0.480 (0.200).
  - Imports Lagged one: column 1 shows 0.755 (0.207).
  - Imports Lagged two: column 1 shows -0.236 (0.172).
  - Interpretation:
    - Current and one-year lagged measures positive; two-year lagged measure negative.
    - Three coefficients jointly significant per F-test.
    - Two-year lagged import coefficient not statistically significant on its own.
    - Consistent with technology spillovers through imports that occur fairly rapidly.
- Sensitivity to industry fixed effects and sample composition:
  - Column 5 (no industry fixed effects): FDI cumulative effect becomes negative (sum of three coefficients), while import coefficients become larger and individually significant—suggesting unobserved industry characteristics matter and industry fixed effects should be included.
  - Preferred specification includes industry fixed effects.
  - Excluding U.S. multinational firms (column 6) changes coefficients modestly; cumulative impact of FDI somewhat larger when multinational firms are excluded.
- Preferred summary magnitudes (column one, preferred specification):
  - Sum of the significant point estimates: FDI ≈ 0.52; Imports ≈ 1.24 (as stated by authors).

### C. Robustness checks
- Longer time differences (Table 5):
  - Two-year differences (column 1) and three-year differences (column 2) use full sample.
  - Coefficients on FDI:
    - 2-year: 0.379 (0.134).
    - 3-year: 0.411 (0.177).
    - 1-year (from Table 4 reproduced): 0.198 (0.076).
    - Interpretation: Point estimate for FDI increases in longer-difference specifications; FDI becomes statistically significant while imports do not.
  - Coefficients on Imports:
    - 2-year: 0.341 (0.282).
    - 3-year: 0.210 (0.297).
    - Imports not statistically significant in longer-difference specifications.
  - Conclusion: Baseline results not driven by short-term noise; evidence for FDI-related spillovers seems stronger than for imports.
- Accounting for changes in denominators and total activity (Table 6):
  - New measures: absolute yearly change in employment at foreign multinationals normalized by lagged total activity by industry; Total Employment defined as absolute yearly change in total employment by industry normalized by lagged total employment; analogous variables for real import growth and real sales growth by industry.
  - Key results (full sample, column 1 of Table 6):
    - Change in FDI EMP Current: 0.151 (0.089); Lagged one: 0.205 (0.086); Lagged two: 0.067 (0.088).
    - Change Imports Current: 0.219 (0.120); Lagged one: 0.045 (0.125); Lagged two: -0.233 (0.122).
    - Change total EMP Current: -0.479 (0.094); Lagged one: -0.070 (0.079); Lagged two: -0.187 (0.078).
    - Change total sales Current: 0.421 (0.042); Lagged one: 0.036 (0.039); Lagged two: 0.123 (0.046).
  - Interpretation:
    - FDI remains associated with TFP growth and effect appears to occur within two years.
    - Import growth results are weaker than baseline; any import spillover appears to occur in the first year only.
    - Increases in real sales growth associated with faster TFP growth; increases in total employment associated with slower TFP growth.
    - Suggests positive relationship between FDI and TFP growth not an artifact of variable construction; imports relationship more suspect.
- Instrumental variables for causation (Table 7):
  - Instruments constructed: one-year differences in transport costs and tariffs for seven developed countries (1985–1996), plus lagged levels of import and FDI shares.
  - IV results (Baseline1 IV1, Baseline2 IV2):
    - Baseline1 IV1 (contemporaneous and lagged one-year using IVs, full sample):
      - FDI Current: 0.198 (0.076)
      - FDI Lagged one: 0.166 (0.077)
      - Imports Current: 0.071 (0.173)
      - Imports Lagged one: 0.554 (0.185)
      - Hansen J-statistic p-value: 0.740 (fails to reject instrument validity)
      - Interpretation: FDI coefficient large and statistically significant; imports coefficients small and statistically insignificant overall.
    - IV results consistent across samples: FDI positive and significant, imports provide little support for spillovers.
    - Authors note instruments are quite weak; do not rely primarily on IV estimates.
- Subperiod analysis (1990–1992 vs 1994–1996; Table 8):
  - Split sample omitting 1993 to distinguish recessionary early 1990s and mid-1990s boom.
  - Findings:
    - FDI coefficients larger in 1994–1996 than in 1990–1992.
    - In early sample (1990–1992) only lagged one FDI coefficient marginally significant.
    - Results similar when excluding U.S. multinational firms: FDI coefficients much larger in mid-1990s.
    - Overall: FDI spillovers present throughout period, appear stronger during mid-1990s than during early-1990s recession.
- Overall robustness conclusion:
  - Evidence more robust for positive FDI spillovers than for imports-related spillovers.
  - Authors summarize: positive effect associated with FDI activity; fail to estimate a robust imports effect.

### D. Importance of spillovers in accounting for U.S. productivity growth
- Text begins assessment: "For FDI, the share of foreign employment in U.S. manufacturing rose between 1987 and 1996 from 7.7 percent to" (text truncated in provided content; further quantification and conclusions not available in the supplied excerpt).

*Source: Bureau of Economic Analysis (BEA) data and analyses presented in the supplied IMF working paper excerpt.*

### 11.7 percent, or by 4.0 percentage points. Our preferred estimate of the FDI spillover effect

### _wp03248 - 11.7 percent, or by 4.0 percentage points. Our preferred estimate of the FDI spillover effect

### Key findings on magnitude and economic importance
- Preferred specification (Table 4, first specification) significant FDI coefficients:
  - Current FDI: 0.213
  - One-year lagged FDI: 0.303
  - Sum (total FDI effect): 0.516
- Olley-Pakes input elasticity estimates (O-P (1) in Table 3) imply average productivity growth in the sample of 0.19 over 1987–96.
- Calculated share of productivity growth accounted for by FDI spillovers: about 11 percent (calculation cited: 0.516*0.04/0.19).
- Analogous calculation for imports (based on specification one in Table 4) suggests imports account for about 9.5 percent of productivity growth in the U.S. over the sample period, though import estimates are noted as less robust.
- ∑Sig Coeff (FDI) by measurement:
  - Aggregated from activities (preferred): 0.516
  - Aggregated from affiliate main line of business: 0.000

### Explanations for relatively strong FDI spillovers estimated
- FDI spillovers in the United States
  - U.S. firms have relatively high average productivity but substantial heterogeneity; foreign-owned affiliates often have higher productivity than average U.S.-owned firms.
  - Possibility of threshold effects: relatively high-productivity firms may be more likely to acquire FDI-related spillovers; U.S. may have many such firms.
  - Evidence suggests threshold effects cannot fully explain the magnitude of estimated spillovers.
- Estimation and measurement issues
  - Use of Olley-Pakes estimator vs. time-differencing:
    - Using sales regression analogous to equation (8) (equation (9) in text) yields partial correlations similar to Olley-Pakes TFP and FDI.
    - Olley-Pakes measures give a better estimate of in-sample productivity growth; Olley-Pakes implied average TFP growth: about 1.9 percent per year (0.19 over sample), whereas alternative input-elasticity based measure implied about 3.5 percent per year.
  - Controlling for market share and mark-up variables: removing competition controls changes FDI spillover estimates only minorly, suggesting change-in-competition effects are not driving the higher estimates here.
- Sample composition: large share of high-technology firms
  - Sample is disproportionately composed of large and high-technology firms; roughly half the firms are in eight high-tech industries (BEA codes 281, 283, 284, 289, 357, 367, 381, 384).
  - Table 9 (high-tech vs low-tech vs full sample) highlights heterogeneity:
    - High-tech sample (column 1): FDI coefficients
      - Current: 0.343 (standard error 0.125)
      - Lagged one: 0.413 (0.140)
      - Lagged two: 0.259 (0.187)
      - F-test (FDI) = 6.31, p-value (0.003) — FDI significant at less than the 1 percent level
    - Low-tech sample (column 2): FDI coefficients
      - Current: -0.012 (0.098)
      - Lagged one: 0.173 (0.097)
      - Lagged two: -0.135 (0.085)
      - F-test (FDI) = 1.72, p-value (0.162) — FDI not significant at standard levels
    - Full sample (column 3) FDI coefficients (for comparison)
      - Current: 0.213 (0.083)
      - Lagged one: 0.303 (0.083)
      - Lagged two: -0.049 (0.086)
      - F-test (FDI) = 7.06, p-value (0.0001)
  - Interpretation: spillovers are stronger in high-tech industries; sample composition partly explains high aggregate spillover estimates but does not negate the calculation that FDI spillovers account for about 11 percent of TFP growth within the sample.
  - Sample TFP growth contrast: Bureau of Labor Statistics multi-factor productivity in U.S. manufacturing 1987–96 = 6.7 percent; Olley-Pakes estimated average TFP growth in sample = 19 percent (0.19).
- Measurement error: FDI by main line of business versus by activity
  - Preferred FDI measure: foreign affiliate employment aggregated by activity (employees engaged in particular activities, below affiliate level).
  - Alternative (flawed) measure: foreign employment based on affiliate main line of business (entire affiliate workforce allocated to one industry).
  - Comparison (Table 10) shows measurement matters:
    - FDI coefficients using activities aggregation (preferred):
      - Current: 0.213 (0.083)
      - Lagged one: 0.303 (0.083)
      - Lagged two: -0.049 (0.086)
    - FDI coefficients using main line of business aggregation:
      - Current: 0.069 (0.047)
      - Lagged one: 0.011 (0.031)
      - Lagged two: -0.039 (0.032)
    - F-test (FDI):
      - Activities aggregation: 7.06, p-value (0.0001)
      - Main-line aggregation: 1.38, p-value (0.249)
    - Sum of significant FDI coefficients:
      - Activities: 0.516
      - Main-line: 0.000
  - Conclusion: mismeasurement of industry FDI biases spillover estimates toward zero; detailed activity-level measures produce substantially larger and statistically significant FDI spillovers.

### Robustness, heterogeneity, and limitations
- Import-related spillovers:
  - Some evidence of import-related spillovers, but import estimates are less robust than FDI estimates.
  - Table 9 shows import coefficient patterns differ across high-tech and low-tech samples; e.g., in high-tech sample lagged two-year import variable = -1.361 (0.442), significant and negative.
- Heterogeneity across firms:
  - Evidence (and supplementary literature cited) suggests less productive (and smaller) plants may receive stronger FDI spillovers than more productive (and larger) ones.
  - Excluding U.S.-owned MNEs tends to increase estimated FDI coefficients slightly.
- Generalizability:
  - Authors argue that accurate measurement of FDI activity likely explains much of the discrepancy with earlier studies and that better data in other settings could revise FDI spillover estimates upward.
  - Nonetheless, there does not appear to be a single “average” FDI spillover effect across both high- and low-tech industries.

### Policy implications and suggested research directions
- Economic significance:
  - FDI spillovers estimated here are economically important—accounting for about 11 percent of productivity growth in the sample—supporting the argument that technology spillovers associated with FDI can matter substantially for productivity growth and welfare.
  - Results provide the strongest evidence the authors are aware of that may support the provision of subsidies to attract FDI from a social welfare viewpoint.
- Cautions and open questions:
  - Whether a socially optimal policy is implemented depends on political-economic realities of local electoral competition.
  - Future research directions recommended:
    - Focus on specific industries and mechanisms, given heterogeneity in spillover strength and heterogeneity in FDI motivation.
    - Broaden view of MNE entry effects to include inter-industry spillovers, longer-run effects (e.g., worker training programs), and signaling effects to other potential foreign investors.
    - Improve measurement of FDI at activity level in other settings to test generalizability.

### Data, sample construction, and variable definitions (high-level)
- Core data sources and variable construction highlights:
  - Sales (Y): Compustat data item 12; deflated by industry-level price index aggregated from Bartelsman and Gray (2001).
  - Labor (L): Compustat data item 29.
  - Capital (K): Compustat data item 8; deflators from BEA satellite accounts.
  - Materials (M): Compustat data item 41 + data item 189 − data item 14 − wage expenditures (wage expenditures = L * average industry wage from Bartelsman and Gray (2001)).
  - R&D (R): Compustat data item 46; deflators from BEA satellite accounts until 1992, thereafter estimated.
  - Import share (IM): Feenstra (2002) imports by industry over imports plus total shipments (Bartelsman and Gray 2001).
  - FDI share (FI): Foreign affiliate employment by industry of activity aggregated from affiliate level to BEA industry level divided by total U.S. employment by BEA industry; source: confidential affiliate level FDI data at the BEA.
- Sample selection:
  - Initial universe: all manufacturing firms active between 1987 and 1996.
  - Foreign-owned firms removed; data cleaned for implausible year-to-year changes.
  - Final sample for firms reporting output and inputs including R&D expenditures: 1,277 firms (see Table 4, (2)).
  - Full working sample used for regressions: firms and observations vary by specification (examples from tables: Number of observations 58,995; Firms 11,149 in some specifications; other splits reported for high-tech/low-tech).

*Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03248.pdf*

### References

### _wp03248 - References

### Journal articles and book chapters
- Aitken, B., and A. Harrison, 1999, “'Do Domestic Firms Benefit from Foreign Direct Investment? Evidence from Venezuela'', American Economic Review Vol. 89, pp. 605–18.
- Branstetter, L., 2000, “Is Foreign Direct Investment a Channel of Knowledge Spillovers? Evidence from Japan’s FDI in the United States,” NBER Working Paper 8015.
- Coe, D. T., and E. Helpman, 1995, “International R&D Spillovers,” European Economic Review Vol. 39, pp. 859–87.
- Doms, M., and B. Jensen, 1998, “Comparing Wages, Skills, and Productivity between Domestically and Foreign-Owned Manufacturing Establishments in the United States,” in R. Baldwin, R. Lipsey, and D. Richardson (eds.), Geography and Ownership as Bases for Economic Accounting, pp. 235–58 (Chicago: Chicago University Press for the NBER).
- Eaton, J., and S. Kortum, 1999, “International Patenting and Technology Diffusion: Theory and Measurement,” International Economic Review Vol. 40, pp. 537–70.
- Ericson, R., and A. Pakes, 1995, “Markov-Perfect Industry Dynamics: A Framework for Empirical Work,” Review of Economic Studies, Vol. 62, pp. 53–82.
- Fosfuri, A., M. Motta, and T. Rønde, 2001, “Foreign Direct Investment and Spillovers through Workers’ Mobility,” Journal of International Economics Vol. 53, pp. 205–22.
- Griliches, Z., 1995, “R&D and Productivity: Econometric Results and Measurement Issues,” in P. Stoneman (ed.), Handbook of the Economics of Innovation and Technological Change, pp.52-89 (Oxford: Blackwell).
- Griliches, Z., and J. Mairesse, 1984, ''Productivity and R&D at the Firm Level,” in Z. Griliches (ed.), R&D, Patents, and Productivity (Chicago: Chicago University Press for the NBER).
- Griliches, Z., 1995, “Production Functions: The Search for Identification,” NBER Working Paper 5067 (Cambridge, MA: NBER, March).
- Hall, R., and C. Jones, 1999, ''Why Do Some Countries Produce So Much More Output per Worker than Others?” Quarterly Journal of Economics Vol. 114, pp. 83–116.
- Haskel, J., S. Pereira, and M. Slaughter, 2001, “Does Inward Foreign Direct Investment Boost the Productivity of Domestic Firms?” paper presented at the NBER Summer Institute, August.
- Jovanovic, B., and P. Rousseau, 2002, “The Q-Theory of Mergers,” American Economic Review Vol. 92 (May), pp. 198–204.
- Keller, W., 1998, “Are International R&D Spillovers Trade Related? Analyzing spillovers Among Randomly Matched Trade Partners,” European Economic Review Vol. 42, pp. 1469–81.
- Keller, W., 2002a, “Geographic Localization of International Technology Diffusion,” American Economic Review Vol. 92, pp. 120–42.
- Keller, W., 2002b, “Knowledge Spillovers at the World’s Technology Frontier,” Working Paper, University of Texas (March).
- Keller, W., 2003, “International Technology Diffusion,” unpublished, Brown University, July; revision of NBER Working Paper 8573.
- Levinsohn, J., and A. Petrin, 2001, “Estimating Production Functions Using Inputs to Control for Unobservables,” forthcoming, Review of Economic Studies.
- Nickell, S., 1996, “Competition and Corporate Performance,” Journal of Political Economy Vol. 104, pp. 724–46.
- Olley, S., and A. Pakes, 1996, “The Dynamics of Productivity in the Telecommunications Equipment Industry,” Econometrica Vol. 64, pp. 1263–97.
- Robinson, P., 1988, “Root-N Consistent Semiparametric Regression,” Econometrica Vol. 55, pp. 931–54.
- Rodriguez-Claré, A., 1996, “Multinationals, Linkages, and Economic Development,” American Economic Review Vol. 86.
- Trefler, D., 1995, “The Case of Missing Trade and other Mysteries,” American Economic Review Vol. 85, pp. 1029–46.
- Xu, B., and J. Wang, 1999, “Capital Goods Trade and R&D Spillovers in the OECD,” Canadian Journal of Economics Vol. 32, pp. 1258–74.

### Working papers, research papers, and case studies
- Bartelsman, E., and W. Gray, 2001, NBER Productivity Database, available at www.nber.org.
- Blalock, G., and P. Gertler, 2002, “Technology Diffusion from Foreign Direct Investment through Supply Chains,” Working Paper, Haas School of Business, University of California, Berkeley ( June).
- Görg, H., and D. Greenaway, 2002, “Much Ado About Nothing? Do Domestic Firms Really Benefit from Foreign Direct Investment?” CEPR Working Paper 3485 (London: CEPR).
- Griffith, R., S. Redding, and H. Simpson, 2003, “Productivity Convergence and Foreign Ownership at the Establishment Level,” CEPR Working Paper 3765 (London: CEPR).
- Kraay, A., I. Soloaga, and J. Tybout, 2001, “Product Quality, Productive Efficiency, and International Technology Diffusion: Evidence from Plant-Level Panel Data,” paper presented at the National Bureau of Economic Research Summer Institute, July.
- Kugler, M., 2002, “The Diffusion of Externalities from Foreign Direct Investment: Theory ahead of Measurement,” Working Paper, University of Southhampton, Southhampton, United Kingdom.
- Larrain B., F., L. Lopez-Calva, and A. Rodriguez-Claré, 2000, “Intel: A Case Study of Foreign Direct Investment in Central America,” Working Paper No.58, Center for International Development, Harvard University, December.
- Fosfuri, A., M. Motta, and T. Rønde, 2001, “Foreign Direct Investment and Spillovers through Workers’ Mobility,” Journal of International Economics Vol. 53, pp. 205–22.
- HANSEN, G., 2001, “Should Countries Promote Foreign Direct Investment?” G-24 Discussion Paper Series, United Nations, New York and Geneva.
- Head, K., 1998, “Comment on Doms and Jensen,” in Robert Baldwin, Robert Lipsey, and J. David Richardson (eds.), Geography and Ownership as Bases for Economic Accounting, pp.255–58 (Chicago: University of Chicago Press).
- Levinsohn, J., and A. Petrin, 2001, “Estimating Production Functions Using Inputs to Control for Unobservables,” forthcoming, Review of Economic Studies.

### Data sources and government publications
- Feenstra, R., 2002, World Trade Flows, 1980-1997 (Davis: Center for International Data, University of California).
- Feenstra, R., J. Romalis, and P. Schott, 2002, “U.S. Imports, Exports, and Tariff data 1989–2001,” NBER Working Paper 9387.
- United States, Bureau of Labor Statistics (BLS) ,2002, Data on Multi-factor Productivity (Washington: Bureau of Labor Statistics). Available online at www.bls.gov.
- United States, Department of Commerce, Bureau of Economic Analysis (BEA), 2002, Affiliate-level data on U.S Foreign Direct Investment (Washington: U.S. Department of Commerce, Bureau of Economic Analysis (International Division)).
- Zeile, W., 2002, “U.S. Affiliates of Foreign Companies. Operations in 2000,” Survey of Current Business August 2002, pp.149–66.

*Source: _wp03248 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2003/_wp03248.pdf_