## Appendix A. Nokia’s expenditure on R&D and the net sales in 1999–2014

## Source details

**Canonical URL:** [Appendix A. Nokia’s expenditure on R&D and the net sales in 1999–2014](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021258-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021258-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021258-print-pdf.pdf.json)

---

### Introduction and study contribution
- Knowledge spillovers from labor mobility transmit tacit knowledge that is not easily codified (Polanyi 1958).
- Study exploits the closure and decline of Nokia’s mobile unit as a quasi-natural experiment to identify causal knowledge spillovers from a superstar technology firm to hiring firms.
- Three major contributions:
  - Causal identification of spillovers via a quasi-experiment created by Nokia’s closure (mitigating reverse causality).
  - Use of job tenure as a novel measure of spillover intensity (mitigates endogeneity from initial selection of skilled workers).
  - New empirical evidence using a comprehensive dataset covering practically all companies operating in Finland and tracking Nokia employees directly; results highlight positive implications of public R&D funding when social returns exceed private returns.

### Empirical background and motivation
- Nokia’s R&D and market context:
  - Nokia’s R&D activities conducted in Finland in 2008 were approximately 49 percent of total business sector R&D expenditure and 37 percent of the total R&D expenditure in Finland (Ali-Yrkkö 2010, Ali-Yrkkö et al. 2021a).
  - Annual R&D expenditure increased from 1.8 billion euros in 1999 up to about 6 billion euros in 2008.
  - Nokia lost its position as world’s largest mobile phone manufacturer in 2012 and exited the mobile device market in 2014.
- Worker departures and reallocation:
  - Approximately 25 thousand of Nokia employees left the company in Finland during the period 2008-2017, with a bit more than fifty percent moving to ICT and other companies in Finland and the rest to the public sector or elsewhere (Ali-Yrkkö et al. 2021b).
  - Between 2004-2016, approximately 30 thousand employees left Nokia, of which 57 percent were workers from senior positions.
  - Specialist positions include senior officials and employees in research and planning (category 3 Upper-level employees with administrative, managerial, professional, and related occupations).

### Key statistics on Nokia worker exits (2004–2016)
- Per-year worker exit figures (exactly as presented):
  - 2004: Total employment 23,938; Share of departing workers in total nr of employees, % = 5.4; Nr of departing workers = 1,274; Nr of departing specialist workers = 362; Nr of departing other workers = 912
  - 2005: Total employment 24,452; Share = 5.9; Nr of departing workers = 1,427; Specialist = 528; Other = 899
  - 2006: Total employment 25,044; Share = 4.9; Nr of departing workers = 1,207; Specialist = 439; Other = 768
  - 2007: Total employment 24,661; Share = 8.6; Nr of departing workers = 2,147; Specialist = 944; Other = 1,203
  - 2008: Total employment 25,265; Share = 6.4; Nr of departing workers = 1,598; Specialist = 725; Other = 873
  - 2009: Total employment 22,763; Share = 12.5; Nr of departing workers = 3,000; Specialist = 1,340; Other = 1,660
  - 2010: Total employment 20,810; Share = 12.6; Nr of departing workers = 2,752; Specialist = 1,520; Other = 1,232
  - 2011: Total employment 18,178; Share = 15.9; Nr of departing workers = 3,108; Specialist = 1,878; Other = 1,230
  - 2012: Total employment 12,961; Share = 35.0; Nr of departing workers = 5,454; Specialist = 2,866; Other = 2,588
  - 2013: Total employment 10,967; Share = 21.9; Nr of departing workers = 2,630; Specialist = 1,819; Other = 811
  - 2014: Total employment 6,955; Share = 48.9; Nr of departing workers = 4,386; Specialist = 3,862; Other = 524
  - 2015: Total employment 6,705; Share = 7.6; Nr of departing workers = 521; Specialist = 385; Other = 136
  - 2016: Total employment 5,872; Share = 14.6; Nr of departing workers = 917; Specialist = 564; Other = 353
- Additional cited statistics:
  - In 2010, the share of specialist to the total number of workers leaving the company constituted 88 percent, while in 2004 the share was only 28 percent.
  - Net sales and total expenditure on R&D for 1999–2014 are reported in Appendix A (figures referenced).

### Identification strategy and econometric approach
- Core challenges:
  - Distinguish direction of causality between ex-Nokia hires and hiring-firm performance.
  - Develop appropriate measures of spillover intensity.
- Research design:
  - Quasi-natural experiment: treatment defined as firms that hired their first former Nokia employee during 2004-2015.
  - Pre-treatment period: 2000-2003; Post-treatment period: 2016-2017.
- Two-stage empirical procedure:
  - Stage 1: Coarsened exact matching (CEM) to match treated and non-treated firms and reduce selection bias.
  - Stage 2: Difference-in-differences estimation exploiting heterogeneity in treatment intensity proxied by exogenous Nokia spillovers.
- Treatment intensity specifications (exactly as stated):
  - (i) Number of hired former Nokia employees,
  - (ii) Years of job tenure at Nokia (1990-2017),
  - (iii) Years of job tenure at Nokia during its growth (1990-2008) and decline (2009-2017) periods.
- Rationale for tenure-based measures:
  - Job tenure at an R&D-intensive firm is operationalized analogous to years of schooling (Mincer) to capture accumulated human capital; specifications (ii) and (iii) sum years of Nokia experience over hired former Nokia employees.

### Matching, samples, and data
- Data and coverage:
  - Source: population- and nationwide administrative datasets gathered by Statistics Finland; linked employer-employee data; Nokia identified business units to trace former Nokia employees.
  - Total sample covers 2000-2017; estimation sample limited to 2004-2015.
  - Firms that recruited former Nokia employees in 2000-2003 were dropped: 971 firms that hired 3,282 employees were excluded.
- Treated vs control definitions:
  - All hires: treated firms hired any former Nokia employees; controls did not hire former Nokia employees.
  - Specialist hires: treated firms hired former Nokia employees to specialist positions; controls recruited workers to similar specialist positions.
- Selected sample summary statistics (means and standard deviations one year prior to recruitment or control):
  - Number of observations: Treated group 1,403; Control group 51,698 (before CEM). After CEM: Treated group 1,214; Control group 40,197.
  - Firm-level means (Before CEM / After CEM):
    - Nr of workers: 124.33 (376.62) / 55.23 (97.48) for treated; 14.70 (46.52) / 32.46 (95.10) for control.
    - Net sales, mill. €: 38.76 (214.43) / 13.56 (45.34) for treated; 3.10 (17.14) / 7.95 (36.09) for control.
    - Value added, mill. €: 9.51 (36.78) / 3.98 (7.99) for treated; 0.84 (3.37) / 1.98 (5.94) for control.
    - Labor productivity, 10^3 €: 88.30 (346.25) / 85.99 (363.71) for treated; 56.24 (116.67) / 67.21 (96.51) for control.
    - Operating profit, 10^3 €: 2,347 (20,532) / 784.13 (3,439) for treated; 164.81 (1,393) / 285.7 (2,022) for control.
    - Fixed assets per worker, 10^3 €: 187.55 (1,467) / 93.62 (817.16) for treated; 99.05 (1,608) / 105.88 (670.38) for control.
    - Share of firms operating in same region as Nokia: 0.90 (0.30) treated; 0.61 (0.49) control before CEM; 0.90 (0.30) treated; 0.67 (0.47) control after CEM.
    - Age of firm, years: 15.32 (15.18) treated; 11.07 (10.90) control before CEM; 13.8 (11.8) treated; 13.4 (11.5) control after CEM.
  - Worker-level means (Before CEM / After CEM):
    - Average education, years: 13.55 (1.74) treated; 12.10 (1.57) control before CEM; 13.6 (1.8) treated; 13.4 (1.7) control after CEM.
    - Average age, years: 38.11 (6.02) treated; 37.48 (6.98) control before CEM; 37.7 (6.1) treated; 38.9 (6.8) control after CEM.
    - Share of females: 0.36 (0.27) treated; 0.34 (0.34) control before CEM; 0.35 (0.27) treated; 0.39 (0.34) control after CEM.
    - Education shares (After CEM): Academic 0.16 (0.21) treated; 0.12 (0.20) control. College 0.70 (0.20) treated; 0.76 (0.22) control. Lower 0.13 (0.15) treated; 0.12 (0.16) control.
  - Industry shares (After CEM):
    - Other services: 0.38 (0.49) treated; 0.51 (0.50) control.
    - ICT services: 0.21 (0.40) treated; 0.05 (0.22) control.
    - Non-ICT business services: 0.17 (0.38) treated; 0.19 (0.39) control.
    - Non-ICT manufacturing: 0.17 (0.37) treated; 0.13 (0.34) control.

### Empirical results — main findings and elasticities
- Outcomes studied: employment, value added, labor productivity, net sales, operating profit (EBIT).
- General finding: empirical evidence supports positive spillovers from Nokia on firm performance; more than half of estimated effects are statistically significant.
- Significance pattern summary:
  - Significant positive effects for most specifications on employment (except a negative effect from job tenure in the growth period), value added, net sales, and operating profit.
  - Largest spillover effects on operating profit and net sales across specifications.
  - Labor productivity coefficients insignificant in most specifications.
- Table 3 — All hires sample (coefficients interpreted as elasticities; standard errors in parentheses; asterisks denote significance at 10%, 5%, and 1% levels):
  - (i) Nr of Nokia ex-employees:
    - Employment: 0.020 (0.014)
    - Value added: 0.071** (0.030)
    - Labor prod.: 0.019 (0.024)
    - Net sales: 0.077 (0.053)
    - Oper. profit: 0.110** (0.044)
  - (ii) Job tenure, total:
    - Employment: 0.009* (0.005)
    - Value added: 0.035*** (0.011)
    - Labor prod.: 0.012 (0.009)
    - Net sales: 0.043** (0.019)
    - Oper. profit: 0.046*** (0.016)
  - (iii) Job tenure, growth period:
    - Employment: -0.019*** (0.006)
    - Value added: -0.004 (0.014)
    - Labor prod.: 0.004 (0.011)
    - Net sales: 0.013 (0.024)
    - Oper. profit: 0.015 (0.023)
    Job tenure, decline period:
    - Employment: 0.071*** (0.011)
    - Value added: 0.101*** (0.024)
    - Labor prod.: 0.022 (0.019)
    - Net sales: 0.078* (0.041)
    - Oper. profit: 0.085** (0.039)
- Interpretations (text excerpts):
  - Specification (i): “a one percentage change in the number of hired ex-Nokia employees on average causes about 0.08 percent and 0.11 percentage change in net sales and operating profit, respectively.” Also positive impacts on value added and employment: “A one percentage change in treatment intensity causes 0.02 and 0.07 percent change in employment and value added, respectively.”
  - Specification (ii): “a one percentage change in treatment intensity results in 0.05, 0.04, 0.04, and about 0.01 percentage change in operating profit, net sales, value added, and employment, respectively.”
  - Specification (iii): tenures at declining Nokia show consistently positive and higher spillover impacts than tenures at growing Nokia. Example: “A one percent change in intensity of ‘tenure at declining Nokia’ results in 0.07 percent change in employment, whereas a one percent change in intensity ‘tenure at growing Nokia’ results in 0.02 percentage decrease of employment.”

### Specialist hires — regression results (Table 4)
- Specialist hires sample — estimated elasticities (standard errors in parentheses):
  - (i) Nr of Nokia ex-employees:
    - Employment: 0.014 (0.017)
    - Value added: 0.066** (0.032)
    - Labor prod.: -0.000 (0.022)
    - Net sales: -0.005 (0.068)
    - Oper. profit: 0.019 (0.047)
  - (ii) Job tenure, total:
    - Employment: 0.001 (0.006)
    - Value added: 0.026** (0.012)
    - Labor prod.: 0.001 (0.009)
    - Net sales: 0.012 (0.024)
    - Oper. profit: 0.001 (0.019)
  - (iii) Job tenure, growth period:
    - Employment: -0.037*** (0.008)
    - Value added: -0.012 (0.016)
    - Labor prod.: -0.006 (0.012)
    - Net sales: -0.008 (0.030)
    - Oper. profit: -0.032 (0.026)
    Job tenure, decline period:
    - Employment: 0.092*** (0.014)
    - Value added: 0.095*** (0.027)
    - Labor prod.: 0.016 (0.020)
    - Net sales: 0.043 (0.052)
    - Oper. profit: 0.083* (0.044)
- Interpretation:
  - Specialist hires exhibit statistically significant positive spillovers on employment and value added; effects on labor productivity, net sales, and operating profit are less consistently significant.
  - Magnitudes for employment and value added align with All hires sample; decline-period tenure again produces larger positive impacts.

### Average spillover effects (ASE) — methodology and key estimates
- ASE computed using equation (3), average treatment intensities from Appendix D, and coefficients from Tables 3 and 4; reported as percent changes in outcomes due to average treatment intensity.
- All hires sample (Table 5) — ASE (in percent):
  - (i) Nr. of Nokia ex-employees: Employment 1.59; Value added 5.21; Labor prod. 1.37; Net sales 5.12; Oper. profit 8.00
  - (ii) Job tenure, total: Employment 2.61; Value added 10.43; Labor prod. 3.47; Net sales 12.79; Oper. profit 14.05
  - (iii) Job tenure, growth period: Employment -5.26; Value added -1.02; Labor prod. 1.02; Net sales 3.37; Oper. profit 4.01
    - Job tenure, decline period: Employment 15.19; Value added 20.40; Labor prod. 4.04; Net sales 14.53; Oper. profit 17.28
- Highlights (All hires):
  - Largest ASE on operating profit when treatment modeled by number of ex-employees or years of tenure.
  - Average treatment by number of former Nokia employees (2 former Nokia employees) ⇒ operating profit ≈ 8 percent increase.
  - Average treatment by job tenure (17 years of Nokia experience) ⇒ operating profit ≈ 14 percent increase.
  - Modeling by years of tenure yields ASEs almost two times larger than by number of ex-employees.
  - Job tenure during decline period (about 6 years in 2009-2014) ⇒ 20 percent increase in value added for a treated firm receiving average treatment.
- Specialist hires sample (Table 6) — ASE (in percent):
  - (i) Nr. of ex-Nokia employees: Employment 1.30; Value added 6.69; Labor prod. -0.03; Net sales -0.47; Oper. profit 1.65
  - (ii) Job tenure, total: Employment 0.37; Value added 8.62; Labor prod. 0.40; Net sales 3.78; Oper. profit 0.45
  - (iii) Job tenure, growth period: Employment -10.35; Value added -3.45; Labor prod. -1.74; Net sales -2.34; Oper. profit -8.79
    - Job tenure, decline period: Employment 20.79; Value added 22.26; Labor prod. 3.57; Net sales 9.29; Oper. profit 17.92
- Highlights (Specialist hires):
  - Largest ASE on value added.
  - Average treatment ~3 hired former Nokia employees ⇒ employment increase ≈ 7 percent.
  - Average treatment by years of Nokia tenure (~25 years in 1990-2008) ⇒ employment increase ≈ 9 percent.
  - Decline-period tenure produces the largest positive ASEs: Employment 21 percent; Value added 22 percent; Oper. profit 18 percent.

### Marginal spillover effects over time (treatment-year dynamics)
- Estimation of treatment-year dynamics using equation (2); treatment-year axis: year 0 = first treatment year; maximum duration 13 years.
- Time-path findings for specification (ii), All hires:
  - Employment:
    - Immediate effect; highest in treatment year 1: a one percentage change in treatment intensity ⇒ about 0.02 percent change in employment in treatment year 1.
    - Treatment effect decreases over time and becomes statistically insignificant about four years after treatment.
  - Value added:
    - Effects fluctuate but remain positive and statistically significant for about 6 years.
    - Largest treatment effect immediately after treatment: 0.06 percent in treatment year 0.
  - Labor productivity:
    - Peak at treatment year 12: a one percent change in tenure intensity ⇒ 0.05 percent change in labor productivity at the 1% significance level.
  - Net sales:
    - Statistically significant positive spillover observed until the 4th treatment year.
  - Operating profit:
    - Spillover effect becomes statistically significant again in treatment year 13: 0.31 at the 5% significance level.
- Overall: spillovers can persist long after treatment (up to 13 years), with varying magnitudes and significance across outcomes and years.

### Conclusion and policy implications
- Study contributions reiterated:
  - First identification of knowledge spillovers from Nokia using linked employer-employee data tracking former Nokia employees over time.
  - Modeling spillovers by both number of workers and years of Nokia experience (job tenure) to mitigate endogeneity.
- Main empirical conclusions:
  - Knowledge spillovers from Nokia to hiring firms are generally positive across employment, value added, sales, and operating profits.
  - Spillovers are especially pronounced for employees who worked at Nokia during its decline period.
  - Magnitude and detectability of spillovers depend on treatment specification and choice of control group (All hires vs Specialist hires).
  - All hires sample: largest impacts on net sales and operating profit (when less restrictive).
  - Specialist hires sample: largest impacts on employment and value added (when more restrictive).
  - Little consistent evidence of direct spillover impact on labor productivity; potential offsetting effect due to simultaneous increases in employment and value added.
- Policy implication:
  - Results support that social returns to R&D exceed private returns due to labor-driven knowledge spillovers; public R&D funding can be justified because society benefits via spillovers even when workers leave the original firm.

*Source: wpiea2021258-print-pdf - Appendix A. Nokia’s expenditure on R&D and the net sales in 1999–2014.*

### Appendix A. Nokia’s expenditure on R&D and the net sales in 1999–2014.

### Appendix A. Nokia’s expenditure on R&D and the net sales in 1999–2014

### Introduction and study contribution
- Knowledge spillovers from labor mobility transmit tacit knowledge that is not easily codified (Polanyi 1958).
- The study exploits the closure and decline of Nokia’s mobile unit as a quasi-natural experiment to identify causal knowledge spillovers from a superstar technology firm to hiring firms.
- Three major contributions:
  - Causal identification of spillovers via a quasi-experiment created by Nokia’s closure (mitigating reverse causality).
  - Use of job tenure as a novel measure of spillover intensity (mitigates endogeneity from initial selection of skilled workers).
  - New empirical evidence using a comprehensive dataset covering practically all companies operating in Finland and tracking Nokia employees directly; results highlight positive implications of public R&D funding when social returns exceed private returns.

### Empirical background and motivation
- Nokia was a flagship Finnish ICT firm and invested heavily in R&D:
  - Nokia’s R&D activities conducted in Finland in 2008 were approximately 49 percent of total business sector R&D expenditure and 37 percent of the total R&D expenditure in Finland (Ali-Yrkkö 2010, Ali-Yrkkö et al. 2021a).
  - Annual R&D expenditure increased from 1.8 billion euros in 1999 up to about 6 billion euros in 2008.
- Nokia’s difficulties began around 2008 with the rise of Android and iPhone; Nokia lost its position as world’s largest mobile phone manufacturer in 2012 and exited the mobile device market in 2014.
- Worker departures and reallocation:
  - Approximately 25 thousand of Nokia employees left the company in Finland during the period 2008-2017, with a bit more than fifty percent moving to ICT and other companies in Finland and the rest to the public sector or elsewhere (Ali-Yrkkö et al. 2021b).
  - Between 2004-2016, approximately 30 thousand employees left Nokia, of which 57 percent were workers from senior positions.
  - Specialist positions include senior officials and employees in research and planning (category 3 Upper-level employees with administrative, managerial, professional, and related occupations).

### Key statistics on Nokia worker exits (2004–2016)
- Table 1 (Worker Exits from Nokia) — per-year figures exactly as presented:
  - 2004: Total employment 23,938; Share of departing workers in total nr of employees, % = 5.4; Nr of departing workers = 1,274; Nr of departing specialist workers = 362; Nr of departing other workers = 912
  - 2005: Total employment 24,452; Share = 5.9; Nr of departing workers = 1,427; Specialist = 528; Other = 899
  - 2006: Total employment 25,044; Share = 4.9; Nr of departing workers = 1,207; Specialist = 439; Other = 768
  - 2007: Total employment 24,661; Share = 8.6; Nr of departing workers = 2,147; Specialist = 944; Other = 1,203
  - 2008: Total employment 25,265; Share = 6.4; Nr of departing workers = 1,598; Specialist = 725; Other = 873
  - 2009: Total employment 22,763; Share = 12.5; Nr of departing workers = 3,000; Specialist = 1,340; Other = 1,660
  - 2010: Total employment 20,810; Share = 12.6; Nr of departing workers = 2,752; Specialist = 1,520; Other = 1,232
  - 2011: Total employment 18,178; Share = 15.9; Nr of departing workers = 3,108; Specialist = 1,878; Other = 1,230
  - 2012: Total employment 12,961; Share = 35.0; Nr of departing workers = 5,454; Specialist = 2,866; Other = 2,588
  - 2013: Total employment 10,967; Share = 21.9; Nr of departing workers = 2,630; Specialist = 1,819; Other = 811
  - 2014: Total employment 6,955; Share = 48.9; Nr of departing workers = 4,386; Specialist = 3,862; Other = 524
  - 2015: Total employment 6,705; Share = 7.6; Nr of departing workers = 521; Specialist = 385; Other = 136
  - 2016: Total employment 5,872; Share = 14.6; Nr of departing workers = 917; Specialist = 564; Other = 353
- Additional statistics cited:
  - In 2010, the share of specialist to the total number of workers leaving the company constituted 88 percent, while in 2004 the share was only 28 percent.
  - The net sales of Nokia in period 1999-2014 are reported in Appendix A (Figure A.2).
  - The total expenditure on R&D by Nokia during the period 1999–2014 is presented in Appendix A (Figure A.1).

### Identification strategy and econometric approach
- Core identification challenges:
  - Direction of causality: distinguishing whether ex-Nokia employees improved hiring-firm performance or productive firms selectively hired ex-Nokia employees.
  - Measurement of spillovers: need for an appropriate metric of knowledge transfer.
- Research design:
  - Mimics an experimental design via a quasi-natural experiment and observational data.
  - Treatment definition: firms that hired their first former Nokia employee during the treatment period 2004-2015.
  - Pre-treatment period: 2000-2003; Post-treatment period: 2016-2017.
- Two-stage empirical procedure:
  - Coarsened exact matching (CEM) to match treated firms with similar non-treated firms and reduce selection bias.
  - Difference-in-differences (diff-in-diff) estimation exploiting heterogeneity in treatment intensity proxied by exogenous Nokia spillovers.
- Treatment intensity (heterogeneous treatment) — three specifications considered exactly as stated:
  - (i) Number of hired former Nokia employees,
  - (ii) Years of job tenure at Nokia (1990-2017),
  - (iii) Years of job tenure at Nokia during its growth (1990-2008) and decline (2009-2017) periods.
- Rationale for tenure-based measures:
  - Job tenure at an R&D-intensive firm is operationalized in the spirit of the Mincer returns-to-schooling approach; years of Nokia experience are used analogously to years of schooling to capture accumulated human capital and refine spillover measurement.
  - Specification (ii) and (iii) measure intensity as the total years of accumulated Nokia experience summed over hired former Nokia employees.

*Source: wpiea2021258-print-pdf - Appendix A. Nokia’s expenditure on R&D and the net sales in 1999–2014.*

### 2017. If workers are hired during the decline period, their job tenure is split into growth and decline

### wpiea2021258-print-pdf - 2017. If workers are hired during the decline period, their job tenure is split into growth and decline

### Context
- Purpose of the growth/decline split: distinguish if spillover effects through hired former Nokia employees differ between Nokia’s growth and decline phases, motivated by the rise and fall of Nokia’s net sales.
- Conjecture: working at Nokia during its “golden years” differed from working there during its most challenging times.

### Econometric approach
- Two-stage conditional difference-in-difference approach augmented by intensity of treatment.
- Stage 1: Coarsened exact matching (CEM) on an annual basis for each cohort (years 2004-2015), matching one year prior to treatment. Matching done without replacement after temporary coarsening.
  - Variables used to form strata: firm size (net sales and employment), firm age, capital intensity (fixed assets/employment), human capital in firms (average education level of workers), and industry.
  - Matching also included dependent variables appearing in the second stage plus a one-year proportional change of the dependent variable to account for pre-treatment development.
  - Matching performed separately for two treated groups: All hires and Specialist hires.
  - CEM produces weights used in stage 2.
- Treatment intensity measures:
  - (i) Number of ex-Nokia employees.
  - (ii) Job tenure (total).
  - (iii) Job tenure split into growth period and decline period.
  - Treatment intensity accounts for firms that hired their first former Nokia employee in 2004-2015 and may hire additional ex-Nokia employees during post-treatment 2016-2017.
  - Job tenure accumulation starts in 1990 (start date of dataset).
- Stage 2: Quasi-experimental diff-in-diff with treatment intensity interacted with treatment dummy.
  - Treatment dummy Di = 1 if firm hired at least one former Nokia employee over a certain period; zero otherwise.
  - Parameter δ interpreted as elasticity: “a one percent increase in the treatment intensity increases outcome y by δ percent.”
  - Year-by-treatment-year specification estimated so elasticity can vary by treatment time (τ), with τ = 0 for the first year of treatment during 2004-2015; negative τ denote pre-treatment years; maximum duration of treatment is 13 years (if treatment begins in 2005 and continues to 2017, consistent with indexing described).
  - Average spillover effect (ASE) defined as percentage change in output due to the average intensity of treatment TI, evaluated at sample means so ASE depends on TI and δ.

### Data
- Source: Various population- and nationwide administrative datasets gathered by Statistics Finland; linked employer-employee data; Nokia identified business units in the data to trace former Nokia employees.
- Coverage and sample decisions:
  - Total sample covers 2000-2017.
  - Estimation sample limited to 2004-2015 to observe firms’ performance four years prior and two years after recruitment (treatment).
  - Firms that recruited former Nokia employees in 2000-2003 were dropped: 971 firms that hired 3,282 employees were excluded.
- Treated vs control sample definitions:
  - All hires sample: treated firms hired any type of former Nokia employees; control firms did not hire former Nokia employees.
  - Specialist hires sample: treated firms hired former Nokia employees to specialist positions; control firms recruited workers to similar specialist positions.
- Selected sample summary statistics (means and standard deviations one year prior to recruitment or being in control group):
  - Number of observations: Treated group 1,403; Control group 51,698 (before CEM). After CEM: Treated group 1,214; Control group 40,197.
  - Average firm's characteristics (Before CEM / After CEM means shown):
    - Nr of workers: 124.33 (376.62) / 55.23 (97.48) for treated; 14.70 (46.52) / 32.46 (95.10) for control.
    - Net sales, mill. €: 38.76 (214.43) / 13.56 (45.34) for treated; 3.10 (17.14) / 7.95 (36.09) for control.
    - Value added, mill. €: 9.51 (36.78) / 3.98 (7.99) for treated; 0.84 (3.37) / 1.98 (5.94) for control.
    - Labor productivity, 10^3 €: 88.30 (346.25) / 85.99 (363.71) for treated; 56.24 (116.67) / 67.21 (96.51) for control.
    - Operating profit, 10^3 €: 2,347 (20,532) / 784.13 (3,439) for treated; 164.81 (1,393) / 285.7 (2,022) for control.
    - Fixed assets per worker, 10^3 €: 187.55 (1,467) / 93.62 (817.16) for treated; 99.05 (1,608) / 105.88 (670.38) for control.
    - Share of firms operating in same region as Nokia: 0.90 (0.30) treated; 0.61 (0.49) control before CEM; 0.90 (0.30) treated; 0.67 (0.47) control after CEM.
    - Age of firm, years: 15.32 (15.18) treated; 11.07 (10.90) control before CEM; 13.8 (11.8) treated; 13.4 (11.5) control after CEM.
  - Average worker's characteristics (Before CEM / After CEM):
    - Average education, years: 13.55 (1.74) treated; 12.10 (1.57) control before CEM; 13.6 (1.8) treated; 13.4 (1.7) control after CEM.
    - Average age, years: 38.11 (6.02) treated; 37.48 (6.98) control before CEM; 37.7 (6.1) treated; 38.9 (6.8) control after CEM.
    - Share of females: 0.36 (0.27) treated; 0.34 (0.34) control before CEM; 0.35 (0.27) treated; 0.39 (0.34) control after CEM.
    - Education shares (After CEM): Academic 0.16 (0.21) treated; 0.12 (0.20) control. College 0.70 (0.20) treated; 0.76 (0.22) control. Lower 0.13 (0.15) treated; 0.12 (0.16) control.
  - Industry shares (After CEM):
    - Other services: 0.38 (0.49) treated; 0.51 (0.50) control.
    - ICT services: 0.21 (0.40) treated; 0.05 (0.22) control.
    - Non-ICT business services: 0.17 (0.38) treated; 0.19 (0.39) control.
    - Non-ICT manufacturing: 0.17 (0.37) treated; 0.13 (0.34) control.

### Empirical results — main findings
- Outcomes of interest: employment, value added, labor productivity, net sales, operating profit (EBIT).
- General result: empirical evidence supports positive spillovers from Nokia on firm performance. More than half of the estimated effects are statistically significant.
- Significance pattern:
  - Significant positive effects for most specifications on employment (except a negative effect from job tenure in the growth period), value added, net sales, and operating profit.
  - Spillover effects are largest on operating profit and net sales across specifications.
  - Coefficient estimates on labor productivity are insignificant in most specifications.
- Table 3 — Treatment effects estimated on the sample of All hires (coefficients interpreted as elasticities; asterisks denote significance at 10%, 5%, and 1% levels):
  - Treatment intensity (i) Nr of Nokia ex-employees:
    - Employment: 0.020 (0.014)
    - Value added: 0.071** (0.030)
    - Labor prod.: 0.019 (0.024)
    - Net sales: 0.077 (0.053)
    - Oper. profit: 0.110** (0.044)
  - Treatment intensity (ii) Job tenure, total:
    - Employment: 0.009* (0.005)
    - Value added: 0.035*** (0.011)
    - Labor prod.: 0.012 (0.009)
    - Net sales: 0.043** (0.019)
    - Oper. profit: 0.046*** (0.016)
  - Treatment intensity (iii) Job tenure, growth period:
    - Employment: -0.019*** (0.006)
    - Value added: -0.004 (0.014)
    - Labor prod.: 0.004 (0.011)
    - Net sales: 0.013 (0.024)
    - Oper. profit: 0.015 (0.023)
    Job tenure, decline period:
    - Employment: 0.071*** (0.011)
    - Value added: 0.101*** (0.024)
    - Labor prod.: 0.022 (0.019)
    - Net sales: 0.078* (0.041)
    - Oper. profit: 0.085** (0.039)
- Interpretations provided in the source:
  - For specification (i): “a one percentage change in the number of hired ex-Nokia employees on average causes about 0.08 percent and 0.11 percentage change in net sales and operating profit, respectively.” Also positive impacts on value added and employment: “A one percentage change in treatment intensity causes 0.02 and 0.07 percent change in employment and value added, respectively.”
  - For specification (ii): “a one percentage change in treatment intensity results in 0.05, 0.04, 0.04, and about 0.01 percentage change in operating profit, net sales, value added, and employment, respectively.”
  - For specification (iii): tenures at declining Nokia show consistently positive and higher spillover impacts than tenures at growing Nokia. Example:
    - “A one percent change in intensity of ‘tenure at declining Nokia’ results in 0.07 percent change in employment, whereas a one percent change in intensity ‘tenure at growing Nokia’ results in 0.02 percentage decrease of employment.”
    - “The largest significant positive spillover effect during the decline period is on value added – a one percentage change in job tenure of Nokia workers during its decline period causes a” (text truncated in source).

*Italic: Source: wpiea2021258-print-pdf (2017) — extracted content provided in the input.*

### 0.1 percentage change in value added. These results suggest that Nokia spillovers from employees

### wpiea2021258-print-pdf - 0.1 percentage change in value added. These results suggest that Nokia spillovers from employees who have worked at Nokia during its challenging years are larger than the experience obtained

### Specialist hires — regression results (Table 4)
- Regression framework: Diff-in-diff estimating elasticity of dependent variables (employment, value added, labor productivity, net sales, operating profit) with respect to treatment intensity measures (i)-(iii). Asterisks denote significance at the 10%, 5%, and 1% levels.
- Estimated coefficients (standard errors in parentheses):
  - (i) Nr of Nokia ex-employees
    - Employment: 0.014 (0.017)
    - Value added: 0.066** (0.032)
    - Labor prod.: -0.000 (0.022)
    - Net sales: -0.005 (0.068)
    - Oper. profit: 0.019 (0.047)
  - (ii) Job tenure, total
    - Employment: 0.001 (0.006)
    - Value added: 0.026** (0.012)
    - Labor prod.: 0.001 (0.009)
    - Net sales: 0.012 (0.024)
    - Oper. profit: 0.001 (0.019)
  - (iii) Job tenure, growth period
    - Employment: -0.037*** (0.008)
    - Value added: -0.012 (0.016)
    - Labor prod.: -0.006 (0.012)
    - Net sales: -0.008 (0.030)
    - Oper. profit: -0.032 (0.026)
  - Job tenure, decline period
    - Employment: 0.092*** (0.014)
    - Value added: 0.095*** (0.027)
    - Labor prod.: 0.016 (0.020)
    - Net sales: 0.043 (0.052)
    - Oper. profit: 0.083* (0.044)
- Interpretation and significance:
  - Specialist hires sample: statistically significant positive spillovers on employment and value added persist; labor productivity, net sales, and operating profit effects are less statistically significant.
  - Magnitudes similar to All hires sample for employment and value added.
  - Specification-level elasticities (as described in text):
    - (i) A one percent change in treatment intensity ≈ 0.07 percent change in value added.
    - (ii) A one percent change in treatment intensity ≈ 0.03 percent change in value added.
    - (iii) Results align with All hires sample: differential impact between expansion and contraction years.

### Average spillover effects (ASE) — methodology and key estimates
- ASE computed using equation (3), average treatment intensities from Appendix D, and coefficients from Tables 3 and 4; reported as percent changes in outcomes due to average treatment intensity.
- All hires sample (Table 5) — ASE (in percent):
  - (i) Nr. of Nokia ex-employees: Employment 1.59; Value added 5.21; Labor prod. 1.37; Net sales 5.12; Oper. profit 8.00
  - (ii) Job tenure, total: Employment 2.61; Value added 10.43; Labor prod. 3.47; Net sales 12.79; Oper. profit 14.05
  - (iii) Job tenure, growth period: Employment -5.26; Value added -1.02; Labor prod. 1.02; Net sales 3.37; Oper. profit 4.01
    - Job tenure, decline period: Employment 15.19; Value added 20.40; Labor prod. 4.04; Net sales 14.53; Oper. profit 17.28
- Highlights from All hires:
  - Largest ASE on operating profit when treatment modeled by number of ex-employees or years of tenure.
  - Average treatment by number of former Nokia employees (2 former Nokia employees) ⇒ operating profit ≈ 8 percent increase.
  - Average treatment by job tenure (17 years of Nokia experience) ⇒ operating profit ≈ 14 percent increase.
  - Modeling by years of tenure yields ASEs almost two times larger than by number of ex-employees.
  - Job tenure during decline period (about 6 years in 2009-2014) ⇒ 20 percent increase in value added for a treated firm receiving average treatment.
- Specialist hires sample (Table 6) — ASE (in percent):
  - (i) Nr. of ex-Nokia employees: Employment 1.30; Value added 6.69; Labor prod. -0.03; Net sales -0.47; Oper. profit 1.65
  - (ii) Job tenure, total: Employment 0.37; Value added 8.62; Labor prod. 0.40; Net sales 3.78; Oper. profit 0.45
  - (iii) Job tenure, growth period: Employment -10.35; Value added -3.45; Labor prod. -1.74; Net sales -2.34; Oper. profit -8.79
    - Job tenure, decline period: Employment 20.79; Value added 22.26; Labor prod. 3.57; Net sales 9.29; Oper. profit 17.92
- Highlights from Specialist hires:
  - Largest ASE on value added.
  - Average treatment ~3 hired former Nokia employees ⇒ employment increase ≈ 7 percent.
  - Average treatment by years of Nokia tenure (~25 years in 1990-2008) ⇒ employment increase ≈ 9 percent.
  - Decline-period tenure produces the largest positive ASEs: Employment 21 percent; Value added 22 percent; Oper. profit 18 percent.

### Marginal spillover effects over time (treatment-year dynamics)
- Estimation uses equation (2); Figure 1 displays treatment-year effects for specification (ii) (job tenure) on All hires sample.
- Treatment-year axis: year 0 = first treatment year; maximum duration 13 years.
- Time-path findings (specification (ii), All hires):
  - Employment:
    - Immediate effect; highest in treatment year 1: a one percentage change in treatment intensity ⇒ about 0.02 percent change in employment in treatment year 1.
    - Treatment effect decreases over time and becomes statistically insignificant about four years after treatment.
  - Value added:
    - Effects fluctuate but remain positive and statistically significant for about 6 years.
    - Largest treatment effect observed immediately after treatment: 0.06 percent in treatment year 0.
  - Labor productivity:
    - Peak at treatment year 12: a one percent change in tenure intensity ⇒ 0.05 percent change in labor productivity at the 1% significance level.
  - Net sales:
    - Statistically significant positive spillover observed until the 4th treatment year.
  - Operating profit:
    - Spillover effect becomes statistically significant again in treatment year 13: 0.31 at the 5% significance level.
- Overall: duration of spillovers may persist long after treatment (up to 13 years), though significance and magnitudes vary by outcome and year.

### Conclusion and policy implications
- Study contributions:
  - First identification of knowledge spillovers from Nokia using linked employer-employee data tracking former Nokia employees over time.
  - Models spillovers by both number of workers and years of Nokia experience (job tenure), helping mitigate endogeneity from initial skills premia.
- Main empirical conclusions:
  - Knowledge spillovers from Nokia to hiring firms are generally positive across employment, value added, sales, and operating profits.
  - Spillovers are especially pronounced for employees who worked at Nokia during its decline period.
  - Magnitude and detectability of spillovers depend on the treatment specification and choice of control group (All hires vs Specialist hires).
  - All hires sample: largest impacts on net sales and operating profit (when less restrictive).
  - Specialist hires sample: largest impacts on employment and value added (when more restrictive).
  - Little consistent evidence of direct spillover impact on labor productivity; potential offsetting effect due to simultaneous increases in employment and value added.
- Policy implication:
  - Results support the argument that social returns to R&D exceed private returns due to labor-driven knowledge spillovers; public R&D funding can be justified because society benefits via spillovers even when workers leave the original firm.

*Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021258-print-pdf.pdf*

### Part 2), 9-49.

### Part 2), 9-49.

### Literature cited (selected studies and themes)
- Classic human capital and evaluation foundations:
  - Becker, G.S. 1964. Human capital; a theoretical and empirical analysis, with special reference to education.
  - Becker, G.S. 1967. Human capital and the personal distribution of income: An analytical approach (No. 1).
  - Mincer, J. 1958. Investment in human capital and personal income distribution.
  - Mincer, J. 1974. Schooling, Experience, and Earnings.
  - Heckman, J. 2000. Micro data, heterogeneity, and the evaluation of public policy: Nobel lecture.
  - Heckman, J. and E. Vytlacil 2007a; 2007b on econometric evaluation of social programs.
- Knowledge spillovers, agglomeration, and productivity:
  - Jaffe, A.B., Trajtenberg, M. and Henderson, R. 1993. Geographic localization of knowledge spillovers as evidenced by patent citations.
  - Griliches, Z. 1998. The Search for R&D Spillovers.
  - Greenstone, Hornbeck, and Moretti. 2010. Identifying Agglomeration Spillovers.
  - Romer, P.M. 1990. Endogenous technological change.
  - Döring, T. and Schnellenbach, J. 2006. Survey on geographical knowledge spillovers and regional growth.
- Worker mobility and firm/industry channels of spillovers:
  - Filatotchev et al. 2011. Knowledge spillovers through human mobility across national borders.
  - Maliranta, Mohnen and Rouvinen 2009. Inter-firm labor mobility as a channel of knowledge spillovers.
  - Stoyanov and Zubanov 2012. Productivity spillovers across firms through worker mobility.
  - Cardoza et al. 2020. “Worker Mobility and Domestic Production Networks.” IMF Working Paper 20/205.
  - Serafinelli 2019. "'Good' Firms, Worker Flows, and Local Productivity."
- Firm-, industry-, and region-level evidence on spillovers, R&D cooperation, and clustering:
  - Fritsch and Franke 2004; Ponds, Oort and Frenken 2009; Niosi and Zhegu 2005; Isaksson, Simeth and Seifert 2016.
- Methodology and causal inference relevant references:
  - Florens et al. 2008. Control functions for continuous, endogenous treatment.
  - Iacus, King, and Porro 2011; 2012. Coarsened exact matching and multivariate matching.
  - Keller 2021. “Knowledge Spillovers, Trade, and Foreign Direct Investment.” NBER Working Paper 28739, April.

### Appendix A — Nokia’s expenditure on R&D and net sales (1999–2014)
- Figures reported (descriptive):
  - Figure 2: The Total Expenditure on R&D by Nokia in 1999–2014, in Billion Euros (in Current Prices). (Authors’ calculations based on Nokia’s annual reports.)
  - Figure 3: The Net Sales of Nokia in 1999–2014, in Billion Euros (in Current Prices). (Authors’ calculations based on Nokia’s annual reports.)
- Axes/ticks indicated in source graphics (textual reproduction):
  - R&D figure vertical ticks: 0, 1, 2, 3, 4, 5, 6, 7.
  - R&D figure years: 1999 through 2014.
  - Net sales figure vertical ticks: 0, 10, 20, 30, 40, 50, 60.
  - Net sales figure years: 1999 through 2014.

### Appendix B — Control variables included in the analysis
- Employment and hiring measures:
  - Other hires: The logarithm of the cumulative sum of hired non-Nokia workers.
  - Other specialist hires: The logarithm of the cumulative sum of non-Nokia workers hired for specialist positions.
- Firm-level characteristics:
  - Capital to labor ratio: Logarithm of the ratio of fixed assets to employment.
  - Firm age: Logarithm of firm’s age, years.
  - Firm size dummies:
    - EMP_FIRM_10_49: Dummy: firm has 10-49 workers (omitted group is size class 0-9 workers).
    - EMP_FIRM_50_249: Dummy: firm has 50-249 workers.
    - EMP_FIRM_250_+: Dummy: firm has 250+ workers.
- Workforce composition (shares, scale 0-1 unless noted):
  - Education: Academic (share of academic-level educated workers, omitted group is share of less than college-level educated workers), College (share of college-level educated workers).
  - Age cohorts: AGE_EMP25_34 (share of 25-34 years old workers, omitted group 16-24 years), AGE_EMP35_44, AGE_EMP45_54, AGE_EMP55_70.
  - Share of R&D-workers: Share of R&D-workers based on socio-economic classification (class 32 Senior officials and employees in research and planning), scale 0-1.
  - Share of females: Share of females, scale 0-1.
- Ownership and geography:
  - FOREIGN_OWNED: Dummy for foreign-owned firms.
  - GOV_OWNED: Dummy for state-owned firms.
  - Same region: Dummy: receives a value of one if a firm is in the same geographical region as Nokia's business unit.
  - Industry: Industry dummy based on NACE Rev. 2 (at the 2-digit level).
  - Region: Dummy for the region (21 regions).
- Note on specialist positions:
  - Specialist position refers to Category 3 in the Classification of Socio-economic Groups 1989 (Statistics Finland). This category includes upper-level employees with administrative, managerial, professional, and related occupations.

### Appendix C — Pre-trends
- The appendix contains visual/analytical material on pre-trends of outcome variables prior to the treatment time (figures referenced in source).

### Appendix D — Average treatment intensities for specifications (i)-(iii)
- Treatment intensity units reported for All hires and Specialist hires samples:
  - All hires:
    - i) Nr. of Nokia ex‐employees: 2.21 (Employment), 2.04 (Value added), 2.02 (Labor prod.), 1.91 (Net sales), 2.01 (Oper. profit)
    - ii) Job tenure, total: 19.59 (Employment), 17.37 (Value added), 17.16 (Labor prod.), 16.48 (Net sales), 17.14 (Oper. profit)
    - iii) Job tenure, growth period: 15.95 (Employment), 14.39 (Value added), 14.18 (Labor prod.), 13.77 (Net sales), 14.27 (Oper. profit)
      - Job tenure, decline period: 7.32 (Employment), 6.29 (Value added), 6.29 (Labor prod.), 5.68 (Net sales), 6.54 (Oper. profit)
  - Specialist hires:
    - i) Nr. of Nokia ex‐employees: 2.57 (Employment), 2.67 (Value added), 2.71 (Labor prod.), 2.59 (Net sales), 2.40 (Oper. profit)
    - ii) Job tenure, total: 23.69 (Employment), 24.79 (Value added), 25.18 (Labor prod.), 24.22 (Net sales), 21.75 (Oper. profit)
    - iii) Job tenure, growth period: 19.39 (Employment), 20.26 (Value added), 20.57 (Labor prod.), 19.82 (Net sales), 18.12 (Oper. profit)
      - Job tenure, decline period: 7.82 (Employment), 8.38 (Value added), 8.54 (Labor prod.), 8.06 (Net sales), 7.28 (Oper. profit)

### Appendix E — Spillover effects over time (estimation and outcomes)
- Estimation approach:
  - Regression results from the estimation of equation (2) using the number of former Nokia employees hired by treated firms as a measure of treatment intensity (specification (i)) and, in one specification, using job tenure as a measure of treatment intensity (specification (ii)).
- Outcome variables for which elasticities of treatment effects are estimated:
  - a) employment
  - b) value added
  - c) labor productivity
  - d) net sales
  - e) operating profit
- Figures described:
  - Figure 4: Spillover Effects by Treatment Year Estimated on the Sample of All Hires (specification (i)).
  - Figure 5: Spillover effects by treatment year estimated on the sample of Specialist hires (specification (i)).
  - Figure 6: Spillover effects by treatment year estimated on the sample of Specialist hires using job tenure as treatment intensity (specification (ii)).

*Source: Part 2), 9-49. (Source PDF filename: wpiea2021258-print-pdf - Part 2), 9-49.)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021258-print-pdf.pdf_