Poster #68 - Does Regional R&D Specialization Drive Cross-Metropolitan Collaboration? Evidence from Korean National R&D Administrative Data

• In Event: Friday Poster Reception

Abstract

A central premise of place-based innovation policy is that concentrating R&D investment in regional strengths builds competitive advantage—but does specialization foster or constrain knowledge exchange beyond metropolitan boundaries? This study examines how regional R&D specialization shapes cross-metropolitan collaboration patterns across Korea's 231 county-level administrative districts over 2012–2024.

We construct a novel panel dataset linking national R&D project records with directed cross-metropolitan expenditure flows—capturing sub-contract and joint research spending between institutions across Korea's 17 upper-tier administrative units. Regional specialization is measured using the peak Location Quotient (LQ_max)—the maximum LQ across 19 scientific fields—capturing a district's dominant comparative advantage. This operationalization is supported by an aggregation sensitivity analysis: broadening the average from the peak field to the top-2, top-3, and top-5 fields yields progressively weaker statistical precision (p = 0.025, 0.039, and 0.067, respectively), with the top-5 average falling below conventional significance thresholds, consistent with the specialization signal being sharpest at the single peak field.

Two-way fixed effects panel regressions show that a one-unit increase in LQ_max predicts a 0.21 percentage point increase in the cross-metropolitan collaboration rate after controlling for district/year fixed effects and total R&D investment scale (β = 0.209, p < 0.05). Directly controlling for investment scale strengthens the LQ coefficient, ruling out a pure size-driven explanation. Crucially, applying the identical model to within-metropolitan collaboration—between districts sharing the same metropolitan region—yields a near-zero, statistically insignificant coefficient (β = −0.013, p = 0.82). This contrast serves as an identification check: if LQ_max simply proxied for overall R&D activity or regional size, both collaboration types would respond similarly. Instead, specialization selectively promotes knowledge-seeking across—but not within—administrative boundaries, consistent with a mechanism in which peak specialization creates external visibility and absorptive capacity for partners beyond the home region. Descriptively, the most specialized district-field pairs exhibit cross-metropolitan collaboration participation rates 34 percentage points higher than the least specialized (91% vs. 58%), with median outflow rates rising from 3.2% to 13.5%.

Gravity model estimates using Poisson Pseudo-Maximum Likelihood (PPML) on the full origin-destination matrix (49,032 directed pairs, 75% structural zeros) confirm that road distance imposes significant spatial friction (elasticity = −0.307 to −0.363, p < 0.001), while research portfolio similarity strongly amplifies collaboration intensity (β = 1.410–1.650, p < 0.001), suggesting districts gravitate toward similar rather than complementary partners. A notable structural exception is Daejeon—Korea's science capital hosting over 60% of government-funded research institutes in a single cluster—where distance bears no statistically significant relationship to collaboration flows, consistent with its role as a distance-insensitive national knowledge hub. Results are robust across three alternative sample construction criteria.

These findings carry direct policy implications. Selective R&D specialization investments appear to expand cross-metropolitan collaboration networks rather than concentrating activity locally, supporting place-based concentration strategies. Portfolio similarity between districts provides an actionable basis for designing targeted cross-metropolitan R&D partnership initiatives beyond geographic proximity alone.

Author

• Hanul Kim, Korea Institute of S&T Evaluation and Planning
  Presenting Author