## wp17253

## Source details

**Canonical URL:** [wp17253](https://www.imf.org/-/media/files/publications/wp/2017/wp17253.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17253.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17253.pdf.json)

---

### Introduction and scope
- Research question: How has import competition from China affected U.S. firms’ innovative activities (R&D) and the allocation of R&D inputs across firms and industries?
- Theoretical tension: rising competition can (a) induce more R&D to improve product quality/differentiation or (b) reduce R&D by lowering domestic firms’ market shares. Aghion et al. (2005) predict an inverted-U relationship between competition and innovation.
- Main identification: industry-level variation in Chinese imports (four-digit SIC) measured as Chinese imports as a share of U.S. consumption/value added; instrumented by Chinese exports to eight other high-income countries (Denmark, Finland, Germany, Japan, New Zealand, Spain and Switzerland) with denominators fixed at U.S. 1995 consumption/value added.

### Data and measurement
- Trade period: 1995 to 2009.
- Firm data: Compustat North America (public firms), 1995–2009; firms with zero R&D from 1995–2009 are dropped.
- Winsorization: variables that can be positive or negative winsorized at 1 percent each tail; nonnegative variables top 1 percent only.
- Productivity and markup measures: TFP (Olley-Pakes), labor productivity (value added per employee), Lerner index (operating profit over sales).
- Firm-level variables lagged five years in regressions.
- Import competition measures:
  - Import Competition 1_US_i,t = Trade Flow China→US_i,t / Consumption_US_i,t
  - Import Competition 2_US_i,t = Trade Flow China→US_i,t / Value Added_US_i,t
  - Instruments: sum of China→c_j,t trade flows divided by Consumption_US_j,1995 or Value Added_US_j,1995.
- Sample construction: overlapping (rolling) five-year differences (e.g., 2005–2000, 2004–1999) for main specification; non-overlapping samples used in robustness checks.

### Key sample summary statistics (Table 1)
- Total assets in mil. USD: Obs 28799; Mean 1744.954; Std. Dev. 11626.4; P50 87.469
- Sales (net) in mil. USD: Obs 28800; Mean 1547.598; Std. Dev. 9442.648; P50 69.263
- R&D expense in mil. USD: Obs 28800; Mean 72.7144; Std. Dev. 413.286; P50 6.611
- R&D expense/sales: Obs 27452; Mean 2.512; Std. Dev. 15.698; P50 0.081
- Employment (thousands): Obs 28800; Mean 4.8671; Std. Dev. 9.6; P50 2.332
- Market-to-book ratio: Obs 24549; Mean 3.634; Std. Dev. 7.009; P50 1.856
- Kaplan-Zingales index: Obs 24176; Mean .922; Std. Dev. 3.144; P50 .565
- Imp. Comp.1, US: Obs 1503; Mean .078; Std. Dev. .145; P50 .02
- Five-year change in Imp. Comp.1, US: Obs 710; Mean .031; Std. Dev. .087; P50 .011

### Industries most affected by increased import competition (Table 2: change 1997–2007)
- Printing trades, machinery & equipment: .69
- Computer peripheral equipment: .47
- Electronic computers: .36
- Optical instruments & lenses: .32
- Games, toys & children’s vehicles: .32
- Note: top four out of five are technically advanced industries; textile and shoe industries excluded because firms with no R&D were dropped.

### Firm-level empirical strategy
- Dependent variable: five-year change in log(R&D) for firm j.
- Baseline regression (five-year differences): Δ log(R&D)_j,t = β1 · Δ Import Competition_US_i,t + firm controls_{j,t-5} + industry and year fixed effects + ε_j,t
- IV: instrument Import Competition_US_i,t with Import Competition_Other_i,t (constructed from other high-income countries, denominators fixed at U.S. 1995).
- Controls include five-year lags of log assets, market-to-book ratio, Kaplan-Zingales index.
- Standard errors clustered at SIC three-digit level.

### Average (aggregate) effects on firm R&D (Table 3)
- OLS and 2SLS results reported; first-stage F statistics reported.
- Key baseline IV estimate (column 4): Imp. Comp.1, US coefficient = -0.640 (standard error 0.238); first-stage F statistic 15.101.
- Stated textual interpretation: "A 10 percentage point increase in import competition leads to a 6 percent drop in R&D expenditures, on average."
- Baseline numerical statement based on column (4): "a 10 percentage point increase in industry-level import competition ... would lead to a 6.4 percent drop in R&D investment, on average."
- High-tech firms (column 5): coefficient -0.730 (0.267) → "10 percent increase ... leading to a 7.3 percent decrease in R&D expenditures on average."
- Balanced panel (column 6): coefficient -0.734 (0.236).

### Within-industry reallocation effects (Tables 4–6)
- Specification adds interaction Δ Import Competition × (lagged firm characteristic in deviations from industry mean); instruments constructed as interaction between original instrument and firm characteristic.
- Reallocation toward more productive and higher-markup firms is estimated and quantified.

Reallocation by TFP (Table 4)
- Baseline: Imp. Comp.1, US = -0.640 (0.238)
- L5.TFP (Olley-Pakes): 0.126 (0.035)
- Interaction Imp. Comp × L5.TFP: 0.683 (0.270), statistically significant
- Interpretation: "For a U.S. firm whose TFP is 1 standard deviation above the industry mean, a 10 percentage point rise in import competition would lead to a 6.8 percent additional increase in R&D."

Reallocation by labor productivity (Table 5)
- Baseline: Imp. Comp.1, US = -0.640 (0.238)
- L5.Labor productivity: 0.109 (0.041)
- Interaction Imp. Comp × L5.Labor productivity: 1.116 (0.483), statistically significant

Reallocation by profit margin / Lerner index (Table 6)
- Baseline: Imp. Comp.1, US = -0.640 (0.238)
- L5.Profit margin: 0.238 (0.028)
- Interaction Imp. Comp × L5.Profit margin: 0.656 (0.227), statistically significant
- Interpretation: firms with higher initial profit margins increase R&D when import competition rises; consistent with inverted-U relationship of competition and innovation.

### Aggregate decomposition of average and reallocation effects
- Aggregation formula (Equation 7) combines average effect and reallocation effect using firm-level coefficients and initial firm R&D shares.
- Coefficients used in decomposition example:
  - 1,avg = -0.640 (from Table 3)
  - 1,TFP = -0.584 and 4,TFP = 0.683 (from Table 4)
- Empirical decomposition (2000–2007): reallocation offsets between 57 percent (in 2007) and 90 percent (in 2005) of the average effect. Total effects remain negative but very small in magnitude in those years.
- Decomposition starts in 2000 due to five-year lags; 2008–2009 decomposition not included because of financial crisis anomalies.

### Industry-level results and researcher reallocation (Table 7 and subsequent analysis)
- Industry-level R&D computed by aggregating Compustat firm R&D within SIC industries (balanced panel 1997–2009) controlling for industry-level medians of log assets, market-to-book, book leverage.
- Table 7 summary: Imp. Comp.1 coefficients across specifications show no robust negative effect on industry-level R&D expenditure; point estimates include -0.068 (OLS), -0.176 (2SLS), 0.179 (2SLS with SIC3 FE).
- Text conclusion: "import competition does not reduce industry-level R&D expenditures. On the contrary, analysis using employment data suggests that demand for researchers increases as a result of rising import competition, as evidenced by the higher employment shares and wages of researchers."
- Using U.S. Decennial Census (2000) and American Community Survey (since 2006), researchers (Science and Engineering occupations) are analyzed by industry: import competition in manufacturing industries speeds structural transformation; researchers are reallocated from affected manufacturing industries to three service industries: business and repair services, financial services and personal services.

### Reallocation of researchers and R&D: main findings
- Researchers defined: prime age workers (employed full-time and aged between 25 to 65) who hold at least one college degree and work in a Science and Engineering (S&E) occupation; 47 S&E occupations identified.
- Data merged: U.S. population censuses (2000, 2006–2012) with industry-level import competition measures (1997, 2003–2009). Industry codes used: IND1990.
- A 10 percentage point increase in import competition:
  - leads to a 1 percentage point increase in the researcher share (share of researchers among all workers in each industry).
  - leads to a 7 percent increase in wages paid to researchers.
- Given that the total share of researchers in employment is less than 5 percent, the 1 percentage point increase in researcher share is economically significant.
- Combined effect: rising import competition increases researchers’ total payroll through higher researcher share and higher researcher wages.

Quantitative estimates (industry-level researcher effects; Table 8)
- Dependent variables: Researcher/Employment and Log wage.
- OLS and 2SLS estimates for L3.Imp Comp.1,US:
  - Researcher/Employment: 0.082*** (0.021) in column (1); 0.098*** (0.032) in column (2).
  - Log wage: 0.379*** (0.080) in column (3); 0.680*** (0.123) in column (4).
- Instrument in 2SLS: Imp. Comp.1 Other_{i,t-3} for Imp. Comp.1 US_{i,t-3}.
- F statistic (K-P): 360.926.
- N = 552.

### Reallocation across industries: movement to service industries (specification and findings)
- Regression (Equation 9): Number of researchers_{s,o,t} / Total number of researchers_{o,t} = α_s · Import Competition China→US_{o,t-3} + o + μ_t + ε_{s,o,t}.
- Construction of occupation-level import competition (Equations 10–11): weighted sums of industry-level import competition using s_{j,o,t} = number of workers in industry j, occupation o at time t / total number of workers in occupation o at time t.
- Significant reallocations detected to three service industries:
  - Business and repair services.
  - Financial services.
  - Personal services.
- Magnitude example: a one standard deviation increase in occupation-level import competition leads to an 11.5 percentage point increase in the share of researchers working in business and repair services.

Quantitative estimates (service-industry reallocation; Table 9)
- L3.Imp Comp.1(Occ) coefficients by service industry:
  - finance: 0.072* (0.036)
  - business and repair: 0.115** (0.047)
  - personal: 0.111* (0.057)
- F statistic (K-P): 1324.621.
- N = 341.

### Mechanisms and interpretation
- Within-industry reallocation:
  - Import competition reallocates R&D expenditures towards more productive and more profitable firms within an industry; this reallocation can offset the negative average effect on R&D at the firm level and explain the insignificant aggregate industry-level effect.
  - Results consistent with theories where more productive firms increase R&D to differentiate products and survive competition.
- Market-power heterogeneity:
  - Incorporating an inverted U-shaped relationship (Aghion et al. (2005)): firms with higher initial market power capture reallocated R&D; firms already facing high competition reduce R&D as import competition intensifies.
- Cross-industry reallocation:
  - Import competition in manufacturing mobilizes researchers from manufacturing into service industries (business and repair, finance, personal services), consistent with a shift in employment composition.
- Broader labor-market implication: rising import competition increased demand for high-skilled researchers, raising both employment and wages of researchers in affected industries.
- Manufacturing employment share decline noted: from 22 percent in 1980 to 8 percent in 2016.

### Robustness checks — overview and selected results
- Robustness dimensions: alternative measures of import competition; non-overlapping sample estimation; controlling for sales in foreign markets and additional firm controls; six-year differences.
- Alternative measures (Measure 2: Chinese imports as a share of U.S. production; Measure 3: Chinese imports as a share of total imports to the U.S.):
  - With both alternative measures, import competition does not significantly reduce firm-level R&D expenditures on average.
  - Within-industry reallocation effects remain significant.
  - Selected coefficients (preserving source labels):
    - Table C1 (Measure 2): Imp. Comp.2,US reported values include -0.362, -0.464, -0.921, -1.129, -1.259, -1.335.
    - Table C2 (Measure 3): Imp. Comp.3,US reported values include -0.305, -0.311, -0.200, -0.267, -0.451, -0.501.
    - Table C3 (Measure 2 within-industry): Imp. Comp×L5.TFP = 0.452⇤; Imp. Comp×L5.Labor productivity = 0.796⇤; Imp. Comp×L5.Profit margin = 0.621⇤⇤.
    - Table C4 (Measure 3 within-industry): Imp. Comp×L5.TFP = 0.847⇤⇤⇤; Imp. Comp×L5.Labor productivity = 1.148⇤⇤; Imp. Comp×L5.Profit margin = 1.016⇤⇤⇤.
- Non-overlapping five-year differences:
  - Average and reallocation effects robust; selected coefficients:
    - Table C5 (Average Effects): Imp. Comp.1,US = -0.912⇤⇤⇤, -0.889⇤⇤⇤, -0.654⇤⇤, -0.626⇤⇤, -0.742⇤⇤, -0.554⇤.
    - Table C6 (Within-Industry): Imp. Comp×L5.TFP = 1.526⇤⇤; Imp. Comp×L5.Labor productivity = 2.035⇤⇤⇤.
- Additional controls (firm log sales, share of foreign sales):
  - Average and reallocation effects remain unchanged.
  - Firms with increasing foreign presence tend to invest more in R&D.
  - Selected coefficients:
    - Table C7 (Average Effects, New Controls): Imp. Comp.1,US = -0.570⇤⇤⇤, -0.567⇤⇤⇤, -0.572⇤⇤, -0.595⇤⇤, -0.689⇤⇤, -0.688⇤⇤⇤.
    - Share of Foreign Sales coefficients reported (e.g., 0.339⇤⇤⇤).
- Six-year differences:
  - Baseline results robust to six-year differences.
  - Reported magnitudes: a 10 percentage point increase in import competition would reduce average R&D expenditures by six percent (similar to 6.4% main estimate).
  - Selected coefficients:
    - Table C9 (Average Effects, Six-Year Difference): Imp. Comp.1,US = -0.682⇤⇤, -0.679⇤⇤⇤, -0.601, -0.602⇤, -0.767⇤⇤, -0.747⇤⇤.
    - Table C10 (Within-Industry, Six-Year Difference): Imp. Comp×L6.TFP = 0.699⇤⇤; Imp. Comp×L6.Labor productivity = 1.023⇤; Imp. Comp×L6.Profit margin = 0.952⇤⇤⇤.

### Policy-relevant implications and conclusions
- Average firm-level effect: rising import competition from China leads to a negative average effect on firm R&D (baseline IV: Imp. Comp.1, US = -0.640; textual equivalents: "A 10 percentage point increase ... leads to a 6 percent drop" and "would lead to a 6.4 percent drop").
- Reallocation effect: import competition induces substantial reallocation of R&D toward more productive firms (TFP and labor productivity) and higher-markup (more profitable) firms; reported interaction coefficients include 0.683 (TFP), 1.116 (labor productivity), 0.656 (profit margin).
- Aggregate outcome: reallocation offsets a large portion of the negative average effect; in 2005–2007 decomposition, reallocation offsets between 57 percent and 90 percent of the average effect, leaving small negative total effects.
- Industry-level R&D and researcher inputs: no evidence of a reduction in industry-level R&D expenditure; employment and wages of researchers increase in affected industries, and researchers relocate to specific service industries — indicating allocative improvement and structural transformation rather than a simple aggregate decline in innovative capacity.
- Conceptual policy implication: policies assessing trade shocks should distinguish average firm-level declines from reallocative gains that concentrate R&D in more productive/profitable firms and shift researcher labor across industries. Policymakers should invest in educating and supplying more well-trained engineers and scientists domestically to support this reallocation.

*Source: wp17253 - References27 (IMF Working Paper, content pages and tables as provided).*

### References27

### References27

### Introduction and scope
- Research question: How has import competition from China affected U.S. firms’ innovative activities (R&D) and the allocation of R&D inputs across firms and industries?
- Key theoretical tension: rising competition can (a) induce more R&D to improve product quality/differentiation or (b) reduce R&D by lowering domestic firms’ market shares. Aghion et al. (2005) predict an inverted-U relationship between competition and innovation.
- Main identification: industry-level variation in Chinese imports (four-digit SIC) measured as Chinese imports as a share of U.S. consumption/value added; instrumented by Chinese exports to eight other high-income countries (Denmark, Finland, Germany, Japan, New Zealand, Spain and Switzerland) with denominators fixed at U.S. 1995 consumption/value added.

### Data and measurement
- Trade period: 1995 to 2009 (trade data obtained from Comtrade; extended to 2009).
- Firm data: Compustat North America (public firms), 1995–2009; firms with zero R&D from 1995–2009 are dropped; winsorization rules: variables that can be positive or negative winsorized at 1 percent each tail; nonnegative variables top 1 percent only.
- Productivity and markup measures: TFP (Olley-Pakes), labor productivity (value added per employee), Lerner index (operating profit over sales); firm-level variables lagged five years in regressions.
- Import competition measures:
  - Import Competition 1_US_i,t = Trade Flow China→US_i,t / Consumption_US_i,t
  - Import Competition 2_US_i,t = Trade Flow China→US_i,t / Value Added_US_i,t
  - Instruments use sum of China→c_j,t trade flows divided by Consumption_US_j,1995 or Value Added_US_j,1995.
- Sample construction: overlapping (rolling) five-year differences (e.g., 2005–2000, 2004–1999) for main specification; non-overlapping samples used in robustness checks.

### Key sample summary statistics (Table 1)
- Total assets in mil. USD: Obs 28799; Mean 1744.954; Std. Dev. 11626.4; P50 87.469
- Sales (net) in mil. USD: Obs 28800; Mean 1547.598; Std. Dev. 9442.648; P50 69.263
- R&D expense in mil. USD: Obs 28800; Mean 72.7144; Std. Dev. 413.286; P50 6.611
- R&D expense/sales: Obs 27452; Mean 2.512; Std. Dev. 15.698; P50 0.081
- Employment (thousands): Obs 28800; Mean 4.8671; Std. Dev. 9.6; P50 2.332
- Market-to-book ratio: Obs 24549; Mean 3.634; Std. Dev. 7.009; P50 1.856
- Kaplan-Zingales index: Obs 24176; Mean .922; Std. Dev. 3.144; P50 .565
- Imp. Comp.1, US: Obs 1503; Mean .078; Std. Dev. .145; P50 .02
- Five-year change in Imp. Comp.1, US: Obs 710; Mean .031; Std. Dev. .087; P50 .011

### Industries most affected by increase in import competition (Table 2: change 1997–2007)
- Printing trades, machinery & equipment: .69
- Computer peripheral equipment: .47
- Electronic computers: .36
- Optical instruments & lenses: .32
- Games, toys & children’s vehicles: .32
- Note: top four out of five are technically advanced industries; textile and shoe industries excluded because firms with no R&D were dropped.

### Firm-level empirical strategy
- Dependent variable: five-year change in log(R&D) for firm j.
- Baseline regression (five-year differences): Δ log(R&D)_j,t = β1 · Δ Import Competition_US_i,t + firm controls_{j,t-5} + industry and year fixed effects + ε_j,t
- IV: instrument Import Competition_US_i,t with Import Competition_Other_i,t (constructed from other high-income countries, denominators fixed at U.S. 1995).
- Controls include five-year lags of log assets, market-to-book ratio, Kaplan-Zingales index; standard errors clustered at SIC three-digit level.

### Average (aggregate) effects on firm R&D (Table 3)
- OLS and 2SLS results reported; first-stage F statistics reported.
- Key baseline IV estimate (column 4): Imp. Comp.1, US coefficient = -0.640 (standard error 0.238), with first-stage F statistic 15.101, N 1320911230  // (table formatting in source).
- Stated interpretation in text (consistent with baseline): "A 10 percentage point increase in import competition leads to a 6 percent drop in R&D expenditures, on average."
- Baseline numerical statement based on column (4): "a 10 percentage point increase in industry-level import competition ... would lead to a 6.4 percent drop in R&D investment, on average."
- High-tech firms (column 5): coefficient -0.730 (0.267) → text: "10 percent increase ... leading to a 7.3 percent decrease in R&D expenditures on average."
- Balanced panel (column 6): coefficient -0.734 (0.236).

### Within-industry reallocation effects (Tables 4–6)
- Specification adds interaction Δ Import Competition × (lagged firm characteristic in deviations from industry mean), with instruments constructed as interaction between original instrument and firm characteristic.
- Reallocation toward more productive and higher-markup firms is estimated and quantified.

Reallocation by TFP (Table 4)
- Baseline: Imp. Comp.1, US = -0.640 (0.238)
- L5.TFP (Olley-Pakes): 0.126 (0.035)
- Interaction Imp. Comp × L5.TFP: 0.683 (0.270), statistically significant
- Interpretation in text: "For a U.S. firm whose TFP is 1 standard deviation above the industry mean, a 10 percentage point rise in import competition would lead to a 6.8 percent additional increase in R&D."

Reallocation by labor productivity (Table 5)
- Baseline: Imp. Comp.1, US = -0.640 (0.238)
- L5.Labor productivity: 0.109 (0.041)
- Interaction Imp. Comp × L5.Labor productivity: 1.116 (0.483), statistically significant
- Qualitative conclusion: similar to TFP results; stronger interaction magnitude possibly because labor productivity is more general.

Reallocation by profit margin / Lerner index (Table 6)
- Baseline: Imp. Comp.1, US = -0.640 (0.238)
- L5.Profit margin: 0.238 (0.028)
- Interaction Imp. Comp × L5.Profit margin: 0.656 (0.227), statistically significant
- Interpretation: firms with higher initial profit margins (higher Lerner index, implying lower initial competition) increase R&D when import competition rises; consistent with inverted-U relationship of competition and innovation.

### Aggregate decomposition of average and reallocation effects
- Aggregation formula (Equation 7) combines average effect and reallocation effect using firm-level coefficients and initial firm R&D shares; coefficients used:
  - 1,avg = -0.640 (from Table 3)
  - 1,TFP = -0.584 and 4,TFP = 0.683 (from Table 4) as applied in decomposition example
- Empirical decomposition (2000–2007): reallocation offsets a large fraction of the negative average effect; reallocation offsets between 57 percent (in 2007) and 90 percent (in 2005) of the average effect. Total effects remain negative but very small in magnitude in those years.
- Note: decomposition starts in 2000 due to five-year lags; 2008–2009 decomposition not included because of financial crisis anomalies.

### Industry-level results (Table 7) and researcher reallocation
- Industry-level R&D computed by aggregating Compustat firm R&D within SIC industries (balanced panel 1997–2009) and controlling for industry-level medians of log assets, market-to-book, book leverage.
- Table 7 summary: Imp. Comp.1 coefficients across specifications show no robust negative effect on industry-level R&D expenditure; point estimates include -0.068 (OLS), -0.176 (2SLS), 0.179 (2SLS with SIC3 FE) with standard errors reported.
- Text conclusion: "import competition does not reduce industry-level R&D expenditures. On the contrary, analysis using employment data suggests that demand for researchers increases as a result of rising import competition, as evidenced by the higher employment shares and wages of researchers."
- Using U.S. Decennial Census (2000) and American Community Survey (since 2006), researchers (Science and Engineering occupations) are analyzed by industry:
  - Finding: import competition in manufacturing industries speeds structural transformation; researchers are reallocated from affected manufacturing industries to three service industries: business and repair services, financial services and personal services.
  - Interpretation: import competition mobilizes researchers from less efficient incumbents to younger private companies in the same industry and to growing service industries.

### Main conclusions and policy-relevant implications
- Average firm-level effect: rising import competition from China leads to a negative average effect on firm R&D (baseline IV: Imp. Comp.1, US = -0.640; equivalent textual statements: "A 10 percentage point increase ... leads to a 6 percent drop" and "would lead to a 6.4 percent drop").
- Reallocation effect: import competition induces substantial reallocation of R&D toward more productive firms (TFP and labor productivity) and higher-markup (more profitable) firms; interaction coefficients reported (TFP interaction 0.683; labor productivity interaction 1.116; profit margin interaction 0.656).
- Aggregate outcome: reallocation effect is sizable and offsets a large portion of the negative average effect; in 2005–2007 decomposition, reallocation offsets between 57 percent and 90 percent of the average effect, leaving small negative total effects.
- Industry-level R&D and researcher inputs: no evidence of a reduction in industry-level R&D expenditure; employment and wages of researchers increase in affected industries, and researchers relocate to specific service industries — indicating an allocative improvement and structural transformation rather than a simple aggregate decline in innovative capacity.
- Conceptual implication for policy: policies assessing trade shocks should distinguish average firm-level declines from reallocative gains that concentrate R&D in more productive/firmer firms and shift researcher labor across industries; aggregate R&D measures and researcher employment patterns provide complementary perspectives.

*Source: wp17253 - References27 (IMF Working Paper, content pages and tables as provided).*

### 2009. Year fixed effects are controlled in all columns. SIC three-digit industry fixed effects are added in the

### wp17253 - 2009. Year fixed effects are controlled in all columns. SIC three-digit industry fixed effects are added in the

### Reallocation of researchers and R&D: main findings
- Researchers are defined as prime age workers (employed full-time and aged between 25 to 65) who hold at least one college degree and work in a Science and Engineering (S&E) occupation. Following the NSF classification, 47 S&E occupations are identified.
- Data merged: U.S. population censuses (2000, 2006–2012) with industry-level import competition measures (1997, 2003–2009). Industry codes used for the merge are IND1990.
- A 10 percentage point increase in import competition:
  - leads to a 1 percentage point increase in the researcher share (share of researchers among all workers in each industry).
  - leads to a 7 percent increase in wages paid to researchers.
- Given that the total share of researchers in employment is less than 5 percent, the 1 percentage point increase in researcher share is economically significant.
- Combined effect: rising import competition increases researchers’ total payroll through higher researcher share and higher researcher wages.

### Quantitative estimates (industry-level researcher effects; Table 8)
- Dependent variables: Researcher/Employment and Log wage.
- OLS and 2SLS estimates for L3.Imp Comp.1,US:
  - Researcher/Employment: 0.082*** (0.021) in column (1); 0.098*** (0.032) in column (2).
  - Log wage: 0.379*** (0.080) in column (3); 0.680*** (0.123) in column (4).
- Instrument in 2SLS: Imp. Comp.1 Other_{i,t-3} for Imp. Comp.1 US_{i,t-3}.
- F statistic (K-P): 360.926.
- N = 552.
- Standard errors reported in parentheses. Significance: * p<0.10, ** p<0.05, *** p<0.01.
- Note: Industry fixed effect and year effect are controlled. Researchers defined as college graduates working in S&E occupations.

### Reallocation across industries: movement to service industries (specification and findings)
- Regression specification (Equation 9): Number of researchers_{s,o,t} / Total number of researchers_{o,t} = α_s · Import Competition China→US_{o,t-3} + o + μ_t + ε_{s,o,t}, where s is one of 9 service industries, o is one of 47 S&E occupations, and t is year.
- Construction of occupation-level import competition (Equations 10–11): weighted sums of industry-level import competition using s_{j,o,t} = number of workers in industry j, occupation o at time t / total number of workers in occupation o at time t.
- Significant reallocations detected to three service industries:
  - Business and repair services.
  - Financial services.
  - Personal services.
- Magnitude example: a one standard deviation increase in occupation-level import competition leads to an 11.5 percentage point increase in the share of researchers working in business and repair services.

### Quantitative estimates (service-industry reallocation; Table 9)
- L3.Imp Comp.1(Occ) coefficients (occupation-level import competition) by service industry:
  - finance: 0.072* (0.036)
  - business and repair: 0.115** (0.047)
  - personal: 0.111* (0.057)
- F statistic (K-P): 1324.621.
- N = 341.
- Standard errors in parentheses. Significance: * p<0.10, ** p<0.05, *** p<0.01.
- Occupation fixed effect and year effect are controlled. Only significant service industry results presented; researchers in service industries overall and other service industries not listed are not affected.

### Mechanisms and interpretation
- Within-industry reallocation:
  - Import competition reallocates R&D expenditures towards more productive and more profitable firms within an industry; this reallocation can offset the negative average effect on R&D at the firm level and explain the insignificant aggregate industry-level effect.
  - Results are consistent with theories where more productive firms increase R&D to differentiate products and survive competition (De Loecker (2011); Hombert and Matray (2015)).
- Market-power heterogeneity:
  - Incorporating an inverted U-shaped relationship (Aghion et al. (2005)): firms with higher initial market power capture reallocated R&D; firms already facing high competition reduce R&D as import competition intensifies.
- Cross-industry reallocation:
  - Import competition in manufacturing mobilizes researchers from manufacturing into service industries (business and repair, finance, personal services), consistent with a shift in employment composition.

### Broader context, magnitudes, and implications
- Manufacturing employment share decline noted: from 22 percent in 1980 to 8 percent in 2016.
- The reallocation effect resembles the “cleansing” effect after negative shocks (Davis and Haltiwanger (1992, 1999)), with reallocation potentially productivity-enhancing.
- Labor-market implication: rising import competition increased demand for high-skilled researchers, raising both employment and wages of researchers in affected industries.
- Policy implication: at the industry level, import competition has no significant negative impact on R&D expenditures among public firms; more intense competition directed R&D towards more productive and more profitable firms (and potentially towards young private firms). Policymakers should not be overly concerned about import competition with regard to innovation and U.S. future growth, but should invest in educating and supplying more well-trained engineers and scientists domestically.

### Methodology notes (TFP and firm-level measures referenced)
- TFP measurement follows Olley and Pakes (1996) and İmrohoroğlu and Tüzel (2014) using Compustat and NBER Manufacturing Database; two-factor production function with labor and capital inputs and value-added output. SIC3 industry-specific year fixed effects μ_st included in the first-stage regression.
- Kaplan-Zingales (KZ) index construction follows Almeida et al. (2004) with winsorization procedures described for component variables.

*Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17253.pdf*

### 1. Cash flow: it is measured as income before extraordinary items (IB) plus depreciation

### 1. Cash flow: it is measured as income before extraordinary items (IB) plus depreciation

### Variable definitions and construction
- Cash flow: measured as income before extraordinary items (IB) plus depreciation and amortization (DP) normalized by the total asset (AT) in the previous year.
- Tobin’s Q: measured as total asset (AT) less book value of equity plus market value of equity normalized by total asset, where:
  - book value of equity = total asset (AT) less total liability (LT) and liquidation value of preferred stock (PSTKL)15 plus deferred taxes and investment tax credit (TXDITC);
  - market value of equity = annual closing price (PRCC from CRSP) of last year multiplied by common shares outstanding (CSHO from CRSP).
- Leverage: measured as total debt in current liabilities (DLC) plus total long-term debt (DLTT) normalized by the total asset (AT).
- Dividends: measured as common/Ordinary dividends plus preferred/preference dividends normalized by the total asset (AT) in the previous year.
- Cash Holding: measured as cash and short-term investments (CHE) normalized by the total asset (AT) in the previous year.
- Lerner Index (following Datta et al. (2013) and Lerner (1934), using Compustat variables):
  - Lerner Index = Net operating profits / Sales = (SALE − COGS − XSGA) / SALE
  - Winsorized to the range of[1,1]to eliminate the effect of extreme values.

### Robustness checks — overview
- Four robustness dimensions reported:
  1. Alternative measures of import competition.
  2. Non-overlapping sample estimation.
  3. Controlling for sales in foreign markets (share of foreign sales) and additional firm controls.
  4. Six-year differences (alternative time differencing).

### C.1 Alternative measures of import competition — key findings
- Two alternative industry measures tried (following Bloom et al. (2016)):
  - Chinese imports as a share of U.S. production (Measure 2).
  - Chinese imports as a share of total imports to the U.S. (Measure 3).
- Main empirical findings:
  - With both alternative measures, import competition does not significantly reduce firm-level R&D expenditures (Tables C1 and C2).
  - Despite insignificant average effects, within-industry reallocation effects remain significant: firms with productivity and profitability above industry mean increased R&D expenditures in response to intensified competition from China (Tables C3 and C4).
  - Reallocation effects are especially strong when scaling Chinese imports by total imports.

- Selected reported coefficients (measure labels preserved as in source):
  - Table C1 (Measure 2, OLS/2SLS columns across samples): Imp. Comp.2,US = -0.362, -0.464, -0.921, -1.129, -1.259, -1.335 (standard errors shown in table).
  - Table C2 (Measure 3): Imp. Comp.3,US = -0.305, -0.311, -0.200, -0.267, -0.451, -0.501 (standard errors shown in table).
  - Table C3 (Within-industry, Measure 2): Imp. Comp.2,US = -1.129, -0.818⇤⇤, -0.822⇤⇤, -0.966⇤⇤; Imp. Comp×L5.TFP = 0.452⇤; L5.TFP (Olley-Pakes) = 0.117⇤⇤⇤; Imp. Comp×L5.Labor productivity = 0.796⇤; L5.Labor productivity = 0.093⇤⇤; Imp. Comp×L5.Profit margin = 0.621⇤⇤; L5.Profit margin = 0.222⇤⇤⇤.
  - Table C4 (Within-industry, Measure 3): Imp. Comp.3,US = -0.267, -0.362, -0.337, -0.562; Imp. Comp×L5.TFP = 0.847⇤⇤⇤; L5.TFP (Olley-Pakes) = 0.108⇤⇤⇤; Imp. Comp×L5.Labor productivity = 1.148⇤⇤; L5.Labor productivity = 0.098⇤⇤; Imp. Comp×L5.Profit margin = 1.016⇤⇤⇤; L5.Profit margin = 0.189⇤⇤⇤.

### C.2 Non-overlapping differences — key findings
- Using non-overlapping five-year differences (samples in 1997, 2002, 2007):
  - Both average effects (Table C5) and reallocation effects (Table C6) are robust to non-overlapping differences.
  - Exception: insignificant reallocation effect towards more profitable firms, likely due to smaller samples.
- Selected reported coefficients:
  - Table C5 (Average Effects, Non-Overlapping Sample): Imp. Comp.1,US = -0.912⇤⇤⇤, -0.889⇤⇤⇤, -0.654⇤⇤, -0.626⇤⇤, -0.742⇤⇤, -0.554⇤.
  - Table C6 (Within-Industry, Non-Overlapping Sample): Imp. Comp.1,US = -0.626⇤⇤, -1.069⇤⇤⇤, -0.948⇤⇤, -0.896⇤⇤; Imp. Comp×L5.TFP = 1.526⇤⇤; Imp. Comp×L5.Labor productivity = 2.035⇤⇤⇤; L5.Profit margin = 0.211⇤⇤⇤.

### C.3 Other controls — key findings
- Adding firm log sales and share of foreign sales as controls:
  - Average effects and reallocation effects remain unchanged with these new controls.
  - Firms with increasing foreign presence tend to invest more in R&D.
  - Firm-specific changes in foreign sales do not interfere with industry-level import competition estimates.
- Selected reported coefficients:
  - Table C7 (Average Effects, New Controls): Imp. Comp.1,US = -0.570⇤⇤⇤, -0.567⇤⇤⇤, -0.572⇤⇤, -0.595⇤⇤, -0.689⇤⇤, -0.688⇤⇤⇤; in Share of Foreign Sales = 0.339⇤⇤⇤, 0.236⇤⇤, 0.339⇤⇤⇤, 0.236⇤⇤, 0.203⇤, 0.293⇤⇤.
  - Table C8 (Within-Industry, New Controls): Imp. Comp.1,US = -0.595⇤⇤, -0.536⇤⇤, -0.512⇤⇤, -0.653⇤⇤⇤; in Share of Foreign Sales = 0.236⇤⇤, 0.233, 0.239⇤, 0.245⇤; Imp. Comp×L5.TFP = 0.637⇤⇤; L5.TFP (Olley-Pakes) = 0.083⇤⇤⇤; Imp. Comp×L5.Labor productivity = 1.078⇤⇤; Imp. Comp×L5.Profit margin = 0.631⇤⇤⇤; L5.Profit margin = 0.236⇤⇤⇤.

### C.4 Six-year differences — key findings
- Baseline results robust to using six-year differences.
- Reported magnitudes:
  - A 10 percentage point increase in import competition would reduce average R&D expenditures by six percent, similar to the 6.4% estimated in main results.
- Selected reported coefficients:
  - Table C9 (Average Effects, Six-Year Difference): Imp. Comp.1,US = -0.682⇤⇤, -0.679⇤⇤⇤, -0.601, -0.602⇤, -0.767⇤⇤, -0.747⇤⇤.
  - Table C10 (Within-Industry, Six-Year Difference): Imp. Comp.1,US = -0.602⇤, -0.467, -0.425, -0.766⇤⇤; Imp. Comp×L6.TFP = 0.699⇤⇤; L6.TFP (Olley-Pakes) = 0.098⇤⇤; Imp. Comp×L6.Labor productivity = 1.023⇤; Imp. Comp×L6.Profit margin = 0.952⇤⇤⇤; L6.Profit margin = 0.216⇤⇤⇤.

### Aggregation of firm-level effects — derivation and parameter values used
- Aggregate growth decomposition of R&D (no entry/exit assumption) separates:
  1. Average effect (average change of firms’ R&D investment).
  2. Reallocation effect (shift of R&D from ex ante low R&D firms to high R&D firms).
  3. Entry component (increase in R&D from new entrants).
  4. Exit component (decrease in R&D from exiting firms).
- Approximation used: (X1 − X0)/X0 ≈ ln(X1/X0) and (xi1 − xi0)/xi0 ≈ ln(xi1/xi0).
- Aggregate impact of Chinese import competition on growth of aggregate R&D written as:
  - (X1 − X0)/X0 China = 1/N Σ_{i=1}^N α_avg · (−Imp_j) + 1/N Σ_{i=1}^N (α_TFP · (−Imp_j) + −TFP · (−Imp_j) · TFP_i) · (xi0 − x̄0)/x̄0 − 1/N Σ_{i∈Ω0} (−α_Surv. · (−Imp_j) + −Surv. · (−Imp_j) · TFP_i) xi0/x̄0
- Coefficient values reported and used in this aggregation:
  - α_avg = −0.640 (according to Table 3)
  - α_TFP = −0.584 (according to Table 4)
  - −TFP = 0.683 (according to Table 4)
- Note: Coefficients α_Surv. and −Surv. in the survival term are estimated from regressions of a survival indicator on import competition, starting TFP, and their interaction.

*Source: wp17253 - 1. Cash flow: it is measured as income before extraordinary items (IB) plus depreciation (IMF working paper content provided in the source PDF).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17253.pdf_