This chapter empirically investigates the Borrowed Size hypothesis using a two-stage Data Envelopment Analysis (DEA) on an industry-level panel dataset covering all 47 prefectures of Japan. In the first stage, an output-oriented DEA with variable returns to scale is applied to estimate regional technical efficiency scores. In the second stage, Tobit regression examines the statistical relationship between spatial externalities, captured by a Borrowed Size index, and population density and inefficiency scores. While procedures such as bootstrap correction and truncated regression are theoretically preferable, this chapter adopts a simplified but theoretically informed approach that reflects practical constraints common in regional economic research, including software limitations and computational cost. This approach does not aim at causal inference; rather, it represents a transparent and pragmatic adaptation to the empirical context. By explicitly acknowledging these limitations and the associated trade-offs, the chapter maintains both analytical coherence and empirical feasibility. The results indicate that in the manufacturing sector, Borrowed Size is positively and significantly associated with efficiency, suggesting that geographic proximity enhances regional performance. In the nonmanufacturing sector, however, proximity to large urban centers may reduce efficiency due to excessive agglomeration and intensified interregional competition, consistent with the concept of “competitive crowding.” By integrating spatial context into the assessment of regional productive efficiency, this chapter proposes a novel analytical framework that offers insights for rethinking regional development strategies and locational policy design.