Different Outputs, Different Benchmarks: DEA Benchmarking of National Entrepreneurial Systems Across 89 Economies, 2015–2022

National entrepreneurship benchmarks depend on the outputs they assess. This study examines whether combining overall Total early-stage Entrepreneurial Activity (TEA) with its nested subsets yields a balanced composite. Data envelopment analysis (DEA) evaluates 417 country–year observations from 89 economies during 2015–2022 using four aggregate antecedent conditions as inputs. Output-oriented Charnes–Cooper–Rhodes (CCR) models assess overall TEA (Model A), internationally oriented and high-growth-expectation TEA (Model B), and all three outputs as a diagnostic composite (Model C). Banker–Charnes–Cooper (BCC) models provide sensitivity estimates. The country-mean ranks from Models A and B have a Spearman correlation of 0.231, and 28 of 63 economies fall into opposite median groups. Model C ranks correlate more strongly with Model A (0.909) than Model B (0.540), without establishing substantive equivalence. Mean minimum and maximum feasible virtual-output shares for overall TEA are 78.5% and 80.8%. The main ordering persists in a complete eight-year panel of 25 economies. Exploratory fractional-probit estimates indicate a positive between-economy association between income and Model B. The study combines output-specific benchmarks, rank comparisons, and feasible-share intervals to diagnose imbalance in nested-output composites. Such composites require an explicit policy rationale for the included outcomes and their trade-offs.