Can innovation be induced by state involvement in the market? Evidence within an expanded framework of Hall & Soskice (2001)

eunhee-kim, 김연배 (2023) · Structural Change and Economic Dynamics 66:264-284 · DOI ↗

Moving beyond the dichotomy of Hall & Soskice (2001): The state's role in economic growth (EAP 2021) 의 expanded VoC (LME, CME, SME) framework 를 innovation 차원으로 확장. 29 OECD/선진국 × 1998-2017 panel (T=20) 의 주성분 분석 로 C_factor (corporate governance), L_factor (labor relations), S_factor (state involvement) 의 3-축 구성 후 35 patent technology field + high/medium-high/medium-low tech RCA 의 technology-specific 추정. 결과: state involvement 가 #7 IT methods for management (S β=0.0713 → 0.2643**), #4 Digital communication, #6 Computer technology 등 strategic technology 에서 patent quality 향상. Institutional complementarity (L×S interaction) 가 대부분 technology field 에서 양 유의 — labor protection + state involvement 의 combined effect 가 single channel 보다 큼. 무역 차원에서 RCA in medium-high tech 가 cointegrated (Pt=-12.59**), RCA in medium-low tech 도 (Pt=-13.85***).

RQ: Hall-Soskice (2001) 의 CME = incremental, LME = radical binary 명제가 technology field 별 로 성립하는가? 3rd type SME 의 innovation pattern + state involvement 의 radical vs incremental differential effect 는?
방법론: fmols / dols panel-cointegration (Moving beyond the dichotomy of Hall & Soskice (2001): The state's role in economic growth 와 동일 framework), institutional-complementarity-interaction (C×S, L×S, C×L) 추가, 35 patent technology field (digital comm / computer / IT / bio / micro-nano / machine tools / engines / mechanical / transport 등) + RCA (high/medium-high/medium-low tech), patent quality index + patent count/GDP dual outcome
데이터: 29 OECD + selected 국 × 1998-2017 (T=20), patent: WIPO/EPO 35 IPC field, 무역: RCA at 3 tech intensity level, 통제: globalization, IPR, ln GDP, ln Pop, inflation, R&D, education, defense
주요 발견: (1) Strategic technology field 에서 state involvement positive: #7 IT methods β_S=0.2643** (complementary model), #6 Computer −0.0387* (base) → +0.0995 (complementary), #4 Digital comm S×C=−0.0939*, L×S=+0.1125** — 3-축 interaction 유의. (2) L×S complementarity 의 universal positive: Table G.5 summary 에서 35 fields 중 25+ field 에서 L×S 양 유의 (patent count/GDP). (3) Incremental field (machine tools, engines): S_factor base β=-0.0623** → complementary β=+0.0561 — interaction 추가 시 부호 반전. (4) 무역 (RCA medium-high tech): Pedroni Pt=-12.59*, cointegrated. RCA medium-low tech: Pt=-13.85*** — state involvement 가 기존 산업의 RCA 도 induce. (5) #15 Biotech: S_factor β=+0.0077 (n.s.) → complementary β=-0.0117 — biotech 는 state-driven 미흡
시사점: (a) Hall-Soskice 의 binary 가설 의 부분 반박 — state involvement + labor protection 의 complementarity 가 radical innovation 도 가능. (b) Korean developmental state 의 strategic technology 영역 (반도체, 배터리, AI) 의 국가 R&D + 시장 결합 모델의 정량 정당화. (c) 한계: small panel (29국), C/L/S factor 의 measurement complexity, complementary model 의 3-way interaction 의 추정 정밀도

29 국의 3-axis VoC (C_factor corporate governance, L_factor labor relations, S_factor state involvement) 의 3D cluster map. Hall-Soskice (2001) 의 LME + CME 이분법 외 SME (State-permeated Market Economy) 의 3rd cluster 의 empirical separation 시각화

요약

Moving beyond the dichotomy of Hall & Soskice (2001): The state's role in economic growth (EAP 2021) 의 expanded VoC + economic growth 분석 의 innovation 차원 확장 — 같은 first-author EunHee Kim, 같은 framework, dependent variable 만 patent quality index + RCA 로 변경. 김연배 author page 의 제3기 (2018-2025) 의 국가-시장 관계 + 혁신 제도 라인의 capstone (author page 분류). 기존 VoC-innovation 문헌 (Akkermans-Castaldi-Los 2009 Research Policy, Witt-Jackson 2016 J Int Bus Stud, Hall-Soskice 2001) 의 LME = radical / CME = incremental binary 의 empirical 반박 + state involvement 의 radical 가능성 식별.

방법론은 4-stage. (i) 주성분 분석 로 C_factor (corporate governance flexibility, financial system), L_factor (employment protection, collective bargaining), S_factor (state ownership, regulation depth) 구성. (ii) Hierarchical clustering 으로 LME / CME / SME 3-cluster (Korea, Norway, Greece 등이 SME). (iii) panel-cointegration test (Pedroni Gt, Ga, Pt, Pa) — radical innovation outcome (patent quality index) 에서 Pt=-14.28** (radicalness), Pt=-3.22 (backward citation). (iv) fmols + dols long-run estimation with C, L, S linear terms + L² + C×S, L×S, C×L complementarity interactions. 35 patent technology field × 2 outcome (quality index, count/GDP) × 2 spec (base, complementary) 의 grid 로 field-specific differential effect 추출.

핵심 정량 결과: Table G.5 summary 에서 patent count/GDP outcome 의 L×S interaction 이 35 field 중 25+ field 에서 양 유의 — labor protection + state involvement 의 combined effect 가 single channel 보다 systematically 크다는 evidence. Strategic technology field 의 detail: (a) #7 IT methods for management: S_factor base β=0.0713* → complementary β=0.2643** (3.7배 증가); (b) #4 Digital communication: C_factor base −0.0209 → complementary +0.0995, C×S=−0.0939*, L×S=+0.1125** — 3-way interaction 유의; (c) #6 Computer technology: C_factor 부호 반전 (−0.0387* → −0.0034); (d) #15 Biotech: S_factor β=+0.0077 (n.s.) → complementary β=-0.0117 — biotech 는 state-driven 미흡, L_factor 가 0.0652** 로 더 important. Incremental field (#26 machine tools): S_factor base β=-0.0623** → complementary +0.0561 — interaction 추가 시 부호 반전, L×S=−0.1807*** 의 negative 가 net effect 결정. 무역 (RCA medium-high tech): Pt=-12.59*, cointegrated. RCA medium-low tech: Pt=-13.85*** — state involvement 가 기존 산업의 RCA 까지 induce.

이 결과는 Hall-Soskice 의 binary 가설 의 partial 반박: (i) State involvement 가 radical innovation field (digital comm, IT methods, computer) 에서 quality 양의 효과, (ii) Labor protection + state involvement 의 complementarity 가 대부분 technology field 에서 universal positive, (iii) Bank-based corporate governance (C_factor) 가 radical innovation 의 일부 field 에 양 — bank-based = incremental only 명제도 부분 반박. 정책 함의: Korean developmental state 의 strategic technology (반도체, 배터리, AI) 의 국가 R&D + 시장 결합 모델의 정량 정당화. Picking winner risk 보다 enabling ecosystem 의 3-axis institutional complementarity 활용이 핵심. 한계: (i) small panel (n=29 × T=20 = 580 obs), (ii) Complementary model 의 3-way interaction 의 추정 정밀도 (multicollinearity), (iii) Reverse causality (state involvement 의 crisis response endogeneity), (iv) 1998-2017 의 single VoC era.

핵심 결과

Cointegration test (radical innovation outcomes, Pedroni)

Dependent	Gt	Ga	Pt	Pa
Radicalness	-2.34*	4.17	-12.91***	-4.95
Backward citations	-6.92***	-0.10	-3.22	-0.27
RCA in high tech	-2.54*	-3.37	-11.04	-3.32
RCA in medium-high tech	-2.91***	-3.13	-12.59*	-3.11
RCA in medium-low tech	-2.96***	-4.52	-13.85***	-3.41

Strategic radical technology field (patent quality index outcome)

Field	C_factor	L_factor	S_factor	C×S	L×S
#4 Digital communication (comp.)	+0.099	-0.100	+0.044	-0.094*	+0.113*
#6 Computer (base)	-0.039*	+0.111***	-0.019	—	—
#7 IT methods (comp.)	+0.082	+0.250	+0.264*	-0.091	-0.030
#15 Biotech (base)	-0.023	+0.065*	+0.008	—	—
#22 Micro/Nano (base)	+0.171*	+0.284*	+0.094	—	—

Incremental field (sign reversal pattern, patent quality)

Field	Spec	S_factor	L×S
#26 Machine tools	Base	-0.062*	—
#26 Machine tools	Complementary	+0.056	-0.181*
#27 Engines/pumps/turbines (comp.)		+0.056	+0.112*
#30 Thermal (comp.)		-0.053	-0.130*
#32 Transport (comp.)		+0.126*	-0.056*

정량 결론 (i) Strategic radical technology 의 S_factor positive 가 complementary model 에서 강화 — L×S, C×S interaction 통해. (ii) Field-specific differential — biotech 는 L_factor 주도, IT methods 는 S_factor 주도, micro-nano 는 C×L 주도. (iii) Incremental field 에서 interaction 추가 시 부호 반전 — base model alone 은 misleading.

방법론 노트

핵심 추정 식은 base model (single-channel) 과 complementary model (3-way interaction) 의 paired 비교:

\text{Innovation}_{it} = \alpha + \beta_C C_{it} + \beta_L L_{it} + \beta_S S_{it} + \gamma' X_{it} + u_{it} \quad \text{(base)}

\text{Innovation}_{it} = \alpha + \beta_C C + \beta_L L + \beta_S S + \beta_{CS}(C \times S) + \beta_{LS}(L \times S) + \beta_{CL}(C \times L) + \gamma' X_{it} + u_{it} \quad \text{(complementary)}

여기서 $X$ = (globalization index, IPR strength, ln GDP, ln population, inflation, R&D/GDP, education, defense spending). Outcome 은 (i) patent quality index (radicalness, backward citation 등 4 measures), (ii) patent count/GDP, (iii) RCA at 3 tech intensity levels. fmols (Fully Modified OLS, Pedroni 2001) 과 dols (Dynamic OLS, Stock-Watson 1993) 의 long-run cointegration coefficient 가 핵심 — serial correlation + endogeneity 자동 보정. 식별 가정: (i) Panel cointegration 의 long-run equilibrium, (ii) C/L/S factor 의 exogeneity (PCA 후 standardized), (iii) Interaction term 의 meaningful multicollinearity 범위 (mean-centered).

Identification 핵심: Base vs complementary spec 의 paired 비교 가 institutional complementarity hypothesis (Hall-Gingerich 2009) 의 핵심 검정. Base model 만 보면 S_factor 의 negative effect 가 machine tools 에서 관찰되지만, complementary model 추가 시 L×S interaction 이 main negative source — 즉 S 자체가 아닌 L 과의 상호작용 이 incremental field 의 driver.

연구 계보

김연배 author page 의 제3기 (2018-2025) 국가-시장 관계 + 혁신 제도 라인의 capstone (author page 분류). 직접 predecessor 는 Moving beyond the dichotomy of Hall & Soskice (2001): The state's role in economic growth (EAP 2021 72:530-548) — expanded VoC framework 의 economic growth 분석 — 의 innovation 차원 후속. 같은 first author, 같은 PCA + cointegration framework, dependent variable 만 변경. Sibling: Socio-cultural factors explaining technology-based entrepreneurial activity: Direct and indirect role of social security (Tech in Society 2020) 의 welfare-entrepreneurship 의 macro VoC 통합 라인. 이론적 뿌리: (a) Hall & Soskice (2001) Varieties of Capitalism — LME/CME 이분법 정초, (b) Hall & Gingerich (2009) BJPS 의 institutional complementarity, (c) Amable (2003) Diversity of Modern Capitalism, (d) Schneider (2009), Nölke-Vliegenthart (2009), Nölke-ten Brink-Claar-May (2019) 의 State-Permeated Market Economy, (e) Akkermans-Castaldi-Los (2009) Research Policy 의 VoC-innovation pattern 실증 라인, (f) Witt & Jackson (2016) J Int Bus Stud 의 VoC + innovation 메타-비판, (g) Pedroni (2001) ReStat, Stock & Watson (1993) Econometrica 의 FMOLS/DOLS 계량. RCA 측정: (h) Balassa (1965) 의 revealed comparative advantage 정의, patent quality: (i) Lanjouw & Schankerman (2004) Econ J 의 patent quality index.