Capturing the increase in land value attributable to transit accessibility has become an increasingly examined alternative to fund the transit system, one which demands reliable estimates of the impact of transport improvement on property values. In the context of value capture, this study presents a new method to value transit accessibility. We integrate the Heckman selection model and spatial econometric techniques to account for two issues that are often overlooked in the conventional hedonic price analysis: sample selection and spatial autocorrelation. The modelling framework is applied to the city of Boston to assess the impact of the subway system on single-family property values and gauge the potential for value capture. We find that failing to correct for sample selection and spatial autocorrelation results in significant bias in valuing transit accessibility. This bias might distort estimates of the value-added effect of transit infrastructure investment and misguide policy design for value-capture programmes.