Synthetic dataset: Energy intelligence and residential property valuation (US market)
Description
Supplementary data for the paper: Ratic, M., & Filipovic, S. - Determining the market value of residential property by incorporating intelligent energy management (The European Journal of Applied Economics). The empirical analysis relies on a synthetically generated dataset that simulates realistic distributions of the US residential market, calibrated on publicly available sources (RECS, US Census Bureau, Zillow Research, Redfin Data Center, US DOE reports) and generated in Python with a fixed random seed for full reproducibility. Contents: (1) properties.csv / dataset.xlsx - 50 residential properties across five regions (California, Florida, Texas, New York metro, Colorado) with structural attributes, HERS index, the four energy intelligence components and the composite attribute E_int, and transaction prices, used to estimate the extended hedonic model; (2) scenarios.csv - 50 operational scenarios (25 classical deterministic, 25 intelligent control: RL/MAS/MPC) with simulated monthly electricity costs under five tariff regimes and a price shock column, used to test hypotheses on operating savings and price volatility protection; (3) gen_props.py, gen_scen.py - generator scripts; (4) analysis.py - reproduction script that re-estimates the extended hedonic model (OLS with regional fixed effects; delta = 0.082, R2 = 0.847) and the scenario analysis (21% average saving, 47% variance reduction, 38% lower shock cost). See README.md for the column dictionary and step-by-step reproduction instructions.
Files
Steps to reproduce
Install Python 3 with pandas, numpy, statsmodels and scipy (pip install pandas numpy statsmodels scipy), place analysis.py in the same folder as properties.csv and scenarios.csv, and run: python analysis.py. The script re-estimates the extended hedonic model, ln(P) = a + b'X + gamma*E_eff + delta*E_int + zeta*(E_eff x E_int) + regional fixed effects + e, by OLS and prints the point estimates reported in Table 2 of the paper (delta = 0.082, R2 = 0.847), together with the scenario analysis: 21% lower average monthly costs under intelligent control (regime-level savings 12-30%, median 18%; two-tailed t = 6.42), a 47% reduction in the average within-scenario variance of monthly costs (F = 3.89), and a 38% lower maximum monthly cost under the simulated price shock. The datasets themselves can be regenerated from scratch with gen_props.py (fixed seed = 34) and gen_scen.py (deterministic construction). See README.md for the column dictionary and definitions.