Why innovation and wages?

• Innovation is the main driver of long run economic growth
  • and competitiveness and profits, too …
• Innovation → wage link is important for the distribution of income
  • Inequalities between labor and capital
  • Inequalities between wage earners
• Inclusive growth

Outline

• Link between innovation, wages, and profits
• Methodology
• Data source
• Main findings

Innovation and wages

Wage bargaining model

\(w = \theta lp + (1 - \theta) w^*\)

• \(lp\) labor productivity
• \(w^*\) workers’ outside wage option
• \(\theta\) workers’ bargaining power

Innovation and wages

\(w = \theta lp + (1 - \theta) w^*\)

Innovation will lead to an increase in wages because of

• an increase in labor productivity, \(lp\)
  • if workers have any bargaining power, \(\theta\)
• an increase in workers’ bargaining power, \(\theta\)
• an increase in workers’ outside option, \(w^*\)

Innovation and wages

Innovation may not have an equal effect on all workers.
Consider two types of labor, high-skill (H) and low-skill (L)

H get higher wages because they have

• higher outside option (higher human capital)
• higher bargaining power (specific skills and tacit knowledge)

H will benefit more from innovation because innovation will increase

• labor productivity
• H workers’ bargaining power (tacit knowledge)
• H workers’ outside option (increase in human capital and demand for skilled labor)
  • Skilled-biased technical change

Innovation and profits

Labor productivity will increase

• Wages of high-wage workers, and
  • Any change in low-wage workers? They earn the minimum wage
• Profits
  • More than average wages?

Innovation is likely to make income distribution unequal

Innovation, labor turnover and spillovers

• Some of high skill workers may leave the firm
  • To get higher wages in other firms
    • Spillover - other firms will benefit from innovation
• Hiring rates may
  • increase (product innovation)
  • decrease (process innovation)

Effect of innovation on wages

Effect of innovation on wages

Estimation method

• Assumptions
  • Common trend
    • The difference between innovators and others before innovation is constant
  • No anticipation period
    • No effect on wages before innovation
• Difference-in-differences
  • Fixed effects model: add a dummy for the post-innovation period
• Problems
  • Staggered innovation
  • Common trend
  • No anticipation period

Problems

• Staggered innovation

Problems

• Common trend
  • Non-linear common trend
  • Conditional common trend
• No anticipation period
  • Test for no anticipation period & common trend

Model

• Firms are observed for years t = 1, …, T
  • All firms are observed in all years (balanced panel)
• The first innovation is observed in year q
  • No innovation before q, t = 1, …, q-1
• The impact of innovation goes from q to T
  • Staggered treatment model
  • \(d_t\) (t ≥ q) is the t^th cohort of innovators
  • \(z_{it}\) innovation dummy variable

Model

• The effect of innovation in s on wages in t

\(te_t(s) = w_t(s) - w_t(∞), s = q, ..., T\)

• Average innovation effect on innovators (ATT, average treatment effect on treated)

\(\tau_{st} = E[te_t(s) | d_s = 1], s = q, ..., T; t = s, ..., T\)

Wooldridge method

• Assumptions
  • No anticipation period
    • \(E[w_t(s) - w_t(∞) | d] = 0, t < s\)
  • Common trend
    • \(E[w_t(∞) - w_1(∞) | d_q, ..., d_T] = E[w_t(∞) - w_1(∞)], t = 2, ..., T\)
• Expected wage (conditional on d)
  • \(E(w_t | d) = E[w_t(∞) + d_q \tau_{qt} + ... + d_T \tau_{Tt}]\)

Wooldridge method

Expected wage can be estimated by fixed-effects model

• \(w_{it} = f_i + T_t + \tau z_{it} + \beta X_{it} + \epsilon_{it}\)
  • \(\tau\) can be estimated as negative even if the real effect is positive but heterogeneous
  • \(X_{it}\) could be endogenous
• \(w_{it} = f_i + T_t + \tau_{st}z_{it} d_{is} T_t + T_t X_i \pi_t + \rho_{st}z_{it} d_{is} T_t \dot X _{ir}\)
  • \(X_i\) are firm-specific variables that effect the trend (common trend conditional on \(X_i\))
  • We impose a restriction: The impact depends on treatment time (same for all cohorts)
  • \(\tau_{st} = \tau_{t-s}, s = q, ..., T, t = s, ..., T\)
• For more information: Jeffrey M. Wooldridge (2021), Two-Way Fixed Effects, the Two-Way Mundlak Regression, and Difference-in-Differences Estimators [https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3906345]

Data sources

• The Ministry of Industry, Entrepreneur Information System
  • Social security data
    • Employee data, quarterly
  • Balance sheets and income statements
  • Patent applicants
  • R&D performers

Wages

• Minimum wage
  • 2022 January: net 4253 TL/month, gross 5004 TL, labor cost 5880 TL
• Social security contributions
  • Floor (minimum wage)
  • Ceiling (gross 37500 TL)
    • Ceiling/floor ratio
    • 2017-2022: 7.5
    • 2009-2016: 6.5

The data set

• Group 1: All manufacturing firms
  • Has balance sheet
  • Exists for a certain time period
  • Employs at least one person
• Group 2: Balanced panel (2007-2020 period)
• Group 3: Control group (CEM, Coarsened Exact Matching)
  • Variables: Number of employees in 2008 (log), firm age (log), capital intensity (log), export status, sector (NACE Rev2, 2-digit level)
• Subgroups - “new innovators”
  • Patent applicants since 2009
  • R&D starters since 2009
    • Some overlap between patent applicants and R&D performers

Number of firms, 2020

• All manufacturing: 459600
  • Has balance sheet: 216503
  • Employs at least 1 person: 135113
    • Patent applicant: 851
    • R&D performer: 3022
• Balanced panel 32274
  • Patent applicant: 429
  • R&D performer: 1470
• All patent applicants in the 2009-2020 period: 1095
  • Control group: 1095
• All R&D performers: 2010
  • Control group: 2010

Variables

• Innovation
  • Patent application
  • R&D
• Wages
  • w10 10^th percentile (the lowest 10^th wage in the firm)
  • w50 50^th percentile (the median wage)
  • w90 90^th’ percentile (the highest 10^th)
• Labor productivity (value added per employee)
• Profit (gross operating margin)
• Labor turnover
  • Hiring and separation rates
• Average wages of new hires and leavers

Average firm size

Average firm size

Relative median wages

Intra-firm wage differentials

Estimation results

• Method: DiD/Fixed effects model - Wooldridge (2021)
• Comparisons
  • Innovators vs all firms (balanced panel)
  • Innovators vs control group
• Anticipation period: 3 years
• Software: R program, data.table, fixest and cem packages

Effect of patents on wages

Effect of R&D on wages

Effect of patents on gender wage differentials

Effect of R&D on gender wage differentials

Effect of patents and R&D on productivity

Effect of patents and R&D on profits

Effect of patents and R&D on turnover

• No significant effect on
  • Hiring rates
  • Separation rates
  • Wages of new hires
  • Wages of leavers

Results

Innovation leads to a long term, permanent, increase in

• wages and productivity
  • The effect is stronger in the case of patents
• inequality between labor and capital
• intra-firm wage differentials
  • no effect on low wage earners

• Weak effect on gender wage differentials
• No effect on labor turnover
• No effect on wages of new hires and leavers

Policy implications

• High wage workers benefit from innovation
  • But less than capital
• Low wage workers do not benefit at all
  • No bargaining power
  • No change in human capital - outside option
• How to achieve inclusive growth?

Limitations

• Mechanisms?
  • Productivity, bargaining power, outside option?
• Patent applications?
  • Granted? Quality?
• R&D intensity?
• Repeat patents?
• Patent-R&D interactions?
• Labor turnover?
• Exits?
• Other variables - exports?
• Effects of technology policies?
  • Patents vs R&D?

Many thanks…