Construction and Validation of a Parsimonious Indicator Set for SDG-Oriented Urban Sustainable Development: Evidence from District-Level Panel Data in Shanghai

To balance indicator coverage, information redundancy, and the costs of annual data updating in urban sustainability assessment, this study constructs and validates a parsimonious district-level indicator set for SDG localization. Official statistical data for Shanghai’s 16 districts from 2016 to 2024 were used to establish a balanced panel dataset comprising 21 candidate indicators and 144 district-year observations, with no missing or interpolated values. Of these indicators, 18 with unambiguous directional interpretations were included in the screening process. Using 2016–2020 as the training period, deterministic exhaustive searches were conducted for combinations of 6–10 indicators, subject to six-dimensional coverage and a maximum variance inflation factor of 5. Data from 2021 to 2024 were reserved as an independent temporal test set. The results identified a seven-indicator set covering economic vitality, population and inclusion, education, health, ecological greening, and urban form. In the training period, the Pearson correlation, Spearman correlation, and calibrated R2 between the parsimonious and full reference indices were 0.901, 0.908, and 0.812, respectively; the corresponding values in the independent test period were 0.867, 0.889, and 0.698. Leave-one-year-out validation further supported the temporal information-retention capacity of the indicator set. By contrast, leave-one-district-out validation and indicator reselection revealed substantial spatial context dependence, indicating that the selected indicators are suitable as a parsimonious core set for annual district-level monitoring in Shanghai rather than as a uniquely defined and universally invariant minimum set. Additive assessments based on equal, entropy, and CRITIC weighting exhibited relatively high consistency, and excluding 2022 had almost no effect on the results. However, geometric aggregation and alternative normalization produced appreciable differences in district rankings. Accordingly, this study proposes a hierarchical monitoring framework consisting of a stable core indicator set supplemented by context-specific indicators.