Providing Protection or Encouraging Holdup? The Effects of Labor Unions on Innovation

Transcription

1 Providing Protection or Encouraging Holdup? The Effects of Labor Unions on Innovation Daniel Bradley, University of South Florida Incheol Kim, University of South Florida Xuan Tian, Indiana University 1 st CSEF Conference on Finance and Labor August, 2013

2 What do people think of unions? 2

3 How do unions affect innovation? Labor unions have two faces (Freeman and Medoff, 1979) Collective voice view: reduce employee turnover, improve cooperation, and implement better policies Monopoly view: raise wages, increase income inequality, and lower society s output Unionization rates in the US have keep dropping Union membership peaked at 28% in 1954, dropped to 11% in 2003, and is 11.3% now (Bureau of Labor Statistics) 3

4 The employee protectionism hypothesis Motivating innovation is important Economic growth (Solow, 1957, REStat) and competitive advantage (Porter, 1992, HBR) but motivating innovation is difficult Long, risky, and idiosyncratic process Unions that provide employees protection against termination in bad faith may motivate innovation Manso (2011, JF) Acharya et al. (2012,WP; 2013, RFS) 4

5 The holdup hypothesis Unions could encourage holdup and impede innovation Contracts motivating innovation are almost always incomplete Workers have incentives to expropriate rents after the innovation process begins Potential ex-post holdup leads to ex-ante underinvestment Unionization may raise wages, encourage shirking, and lower output. Therefore, unionzation could reduce innovation productivity 5

6 This paper We examine the effects of unionization on innovation Using union election data, we compare innovation output for firms that elect to unionize to those that do not To establish causality, we use a regression discontinuity design (RDD) relying on locally exogenous variation in union status For close-call elections, passing is very close to a random event and unlikely correlated with firm unobservables We further identify underlying economic mechanisms 6

7 Main findings Unionization negatively affects firm innovation After successful union elections, patent counts (citations) drop by 22% (14%) three years post-election After successful deunion elections, patent counts increase by 12.3% Effects are economically larger with the RDD The main channel is declined productivity rather than underinvestment in R&D Event studies show that the market reacts negatively to innovative firms that elect to unionize 7

8 Literature Motivating innovation Holmstrom (1989, JEBO), Aghion and Tirole (1994, QJE), Manso (2011, JF), Aghion et al. (2005, QJE), Lerner et al. (2011, JF), Nanda and Rhodes-Kropf (2012, JFE), Seru (2012, JFE), Aghion et al. (2012, AER), Atanassov (2013, JF), Tian and Wang (2012, RFS), Fang, Tian, and Tice (2013, JF), He and Tian (2013, JFE), Hsu, Tian, and Xu (2013, JFE), Cornaggia, et al. (2013, JFE), Mao, Tian, and Yu (2013, WP), etc. Most related work Acs and Audretsch (1988, AER), Hirsch and Link (1987, JLR) Acharya et al. (2012, WP; 2013, RFS) 8

9 Data Union election data from the National Labor Relations Board (NLRB) over 1980 to 2006 We manually match firms with Compustat by name, location, and industry Innovation data from the NBER patent database We fix the patent and citation truncation problems following Hall, Jaffe, and Trajtenberg (2001, WP) Two main proxies for innovation: patent counts; citations per patent Two alternative proxies for innovation: patent generality; patent originality 9

10 Summary statistics Obs. Mean Std. P25 Median P75 Panel A. In case of unionization Vote for union (%) Passage (%) Panel B. In case of de-unionization Vote for de-union (%) Passage (%) Panel C. Innovation variables Patents 12, Citations/Patents 12,

11 Baseline specification Ln( Innovation Unionization Z Year Firm i, tn ) Z is a vector of controls that includes firm assets, ROA, leverage, PPE, capital expenditures, Tobin s Q, KZ index, and Herfindahl i, t (1) (2) (3) (4) (5) (6) i, t t i i, t Ln (Patents) t+n Ln (Citation/Patents) t+n Unionization *** *** ** *** ** ** (-2.67) (-2.70) (-2.50) (-2.58) (-2.33) (-2.03) Controls Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 12,244 12,244 12,244 12,244 12,244 12,244 R-squared

12 Deunionization (1) (2) (3) (4) (5) (6) Ln (Patents) t+n Ln (Citation/Patents) t+n Deunionization 0.105** 0.107** 0.123** (2.06) (2.11) (2.39) (0.57) (0.82) (1.21) Controls Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 3,863 3,863 3,863 3,863 3,863 3,863 R-squared

13 Identification Union election results are more likely exogenous They are collectively determined by all union members However, there are still concerns Omitted variables related to both innovation and election results Firms with low innovation potential may be more likely to pass union elections Regression discontinuity design (RDD) Relying on locally exogenous variation in union status generated by elections that pass or fail by a small margin Passing is almost random and unlikely correlated with firm unobservable characteristics 13

14 RDD: A plot support support 14

15 RDD: Polynomial Ln( Innovation Year Industry i, tn ) Unionizationi, t Pl ( vi, c) Pr ( vi, c) (1) (2) (3) (4) (5) (6) VARIABLES Ln (Patents) t+n Ln (Citations/Patents) t+n t j i, t Unionization *** ** *** ** * ** (-2.92) (-2.54) (-3.03) (-1.97) (-1.82) (-2.12) Polynomial (3) Yes Yes Yes Yes Yes Yes Year/ Industry Yes Yes Yes Yes Yes Yes Observations R-squared

16 RDD: Close elections (1) (2) (3) (4) (5) (6) [48%, 52%] Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization * * (-1.27) (-1.42) (-1.69) (-0.67) (-1.39) (-1.76) Observations R-squared

17 RDD: Placebo tests (1) (2) (3) (4) (5) (6) [38%, 42%] Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization (-0.03) (-0.53) (-0.34) (-0.41) (-0.94) (-0.65) Observations R-squared (1) (2) (3) (4) (5) (6) [43%, 47%] Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization (0.51) (0.35) (0.26) (0.33) (0.16) (-0.47) Observations R-squared

18 RDD: Placebo tests (1) (2) (3) (4) (5) (6) [53%, 57%] Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization (-0.78) (0.21) (-1.44) (-0.46) (0.01) (-1.45) Observations R-squared (1) (2) (3) (4) (5) (6) [58%, 62%] Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization (0.51) (0.16) (0.24) (-0.21) (-0.97) (-0.61) Observations R-squared

19 Private firms There appears a causal, negative effect of unions on innovation for public firms What about private firms? Limitations: we don t observe accounting and ownership information We match private firms by name, location, and industry 19

20 Private firms (1) (2) (3) (4) (5) (6) Polynomial order 3 [48%, 52%] VARIABLES Ln (Patents) t+n Ln (Patents) t+n Unionization * * * * (-1.20) (-1.76) (-1.81) (-1.61) (-1.69) (-1.85) Observations 7,234 7,234 7, R-squared

21 Possible Mechanisms Underinvestment in R&D Workers have incentives to expropriate rents after innovation starts Ex-post hold-up leads to ex-ante underinvestment in R&D Reduced innovation productivity Unionization may encourage shirking, raise wages, and lower outputs Investment may be the same, but productivity drops 21

22 Underinvestment in R&D? (1) (2) (3) (4) (5) (6) (R&D/Assets) t+n (R&D/Assets) t+n Unionization ** ** ** (-2.27) (-2.17) (-1.98) (-0.60) (-0.25) (0.32) Controls Yes Yes Yes Yes Yes Yes Industry FE Yes Yes Yes No No No Firm FE No No No Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 12,244 12,244 12,244 12,244 12,244 12,244 R-squared

23 Underinvestment in R&D? (1) (2) (3) (4) (5) (6) Polynomial order 3 [48%, 52%] VARIABLES (R&D/Assets) t+n (R&D/Assets) t+n Unionization (-0.41) (-0.21) (-1.11) (-1.06) (-0.68) (-0.05) Observations R-squared

24 Market reactions (1) (2) (3) (4) (5) CAR [-1, +1] Unionization (-0.63) (1.06) (1.18) (1.23) (1.11) Unionization * Past Innovation ** (-2.12) ** (-2.21) ** (-2.24) ** (-2.00) Past Innovation (0.89) (0.86) (0.87) (0.77) Controls No No No No Yes Industry FE No No Yes Yes Yes Year FE No No No Yes Yes Observations R-squared

25 Conclusion Unions impede firm innovation Effects are economically larger when the RDD is used Unions reduce innovation productivity but not underinvestment in R&D Market is aware of the potential damage unionization has on firm innovation 25

26 Diagnostic check for RDD WIN=0 WIN=1 Diff p-value Ln (Assets) Ln (1+BM) ROA PPE/Assets Debt/Assets Capx/Assets Ln (1+Firm Age) HHI

27 Robustness checks Alternative econometric model: Negative binomial model Alternative innovation proxies: Originality and generality. Union size: No. of union members/no. of employees Results are stronger with 10% than 5% Type of workers: blue collar and white collar Results are similar across both groups 27

28 Alternative model: Negative binomial (1) (2) (3) (4) (5) (6) No. of Patents t+n No. of Citations t+n Unionization *** *** *** *** *** *** (-2.66) (-2.63) (-2.87) (-3.29) (-3.14) (-3.51) Controls Yes Yes Yes Yes Yes Yes Industry FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 12,244 12,244 12,244 12,244 12,244 12,244 Chi-squared back 28

29 Alternative proxy: Generality and Originality (1) (2) (3) (4) (5) (6) Gnerality Originality Unionization ** * * * (-2.39) (-1.84) (-1.41) (-1.68) (-1.40) (-1.78) Controls Yes Yes Yes Yes Yes Yes Firm FE Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Observations 12,244 12,244 12,244 12,244 12,244 12,244 Chi-squared back 29

30 Union size Ratio > 5% Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization *** *** ** ** * (-2.84) (-2.59) (-2.32) (-2.25) (-1.87) (-1.62) Crls, firm & year FE Yes Yes Yes Yes Yes Yes Observations 4,361 4,361 4,361 4,361 4,361 4,361 Chi-squared back Ratio > 10% Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization *** *** *** ** * * (-3.55) (-3.23) (-3.02) (-2.55) (-1.78) (-1.87) Crls, firm & year FE Yes Yes Yes Yes Yes Yes Observations 2,959 2,959 2,959 2,959 2,959 2,959 Chi-squared

31 Worker types Blue collar Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization ** ** ** ** ** ** (-2.23) (-2.29) (-2.22) (-2.58) (-2.32) (-2.04) Crls, firm & year FE Yes Yes Yes Yes Yes Yes Observations 9,406 9,406 9,406 9,406 9,406 9,406 Chi-squared White collar Ln (Patents) t+n Ln (Citations/Patents) t+n Unionization ** ** (-2.15) (-2.13) (-1.58) (-0.44) (-0.15) (-0.11) Crls, firm & year FE Yes Yes Yes Yes Yes Yes Observations 2,838 2,838 2,838 2,838 2,838 2,838 back Chi-squared