Capital expenditures data is critical in accurately calculating commercial real estate (CRE) property return. For example, price indices and related benchmarks that only rely on transaction information may not accurately reflect price appreciation returns. Capital expenditure details are also important in understanding the benefits to investing in various property improvements, in predicting operational risk, and in assessing the impact of changes to a structure on neighboring properties as well as the local economy. Unfortunately, few data sources capture capital expenditures. We explore a statistical solution to these issues by studying the relationship between permitting data, acquired from BuildFax via county-level sources, and known outlays reported in the NCREIF property-level dataset. Our model is able to predict CapEx out of sample and captures significant time-series and cross-sectional variation. We demonstrate the model’s utility by applying its out of sample predictions to correcting a repeat sales index which, in the absence of adjustment for capital investment, results in a 2% bias per year in true capital gains.