Fuel economy (FE) and greenhouse gas (GHG) emissions measured via chassis testing under laboratory conditions were never intended to represent the wide range of real-world driving conditions that are experienced during a vehicle's lifetime. Comprehensive real-world information is needed to better assess US FE label adjustments, determine off-cycle credits for FE standards, and forecast real-world driving behavior, fuel consumption, and CO2 emissions. This paper explores a cost effective method to collect in-use fuel consumption data using the on-board diagnostics (OBD) data stream in light-duty vehicles (LDVs). The accuracy of fuel consumption calculated from the OBD data was analyzed in two ways. First, fuel rates calculated from standard OBD Parameter IDs (PIDs) were compared with fuel rate estimates based on enhanced PID (OEM fuel injector fuel rate) data in two different vehicles. The cumulative fuel rate derived from standard PIDs was lower than the fuel rate values obtained from injector-based data (Enhanced PID) for the tested vehicles by approximately 3% and 13%. A second analysis compared fuel rates calculated from OBD inputs, including mass air flow (MAF) sensor and narrow-band O2 sensors, against fuel rates determined through carbon balance measurements from dynamometer chassis testing. The OBD MAF data were sufficient to closely match the fuel consumption measured by chassis testing during stoichiometric operation. Vehicle-specific algorithms were developed for cold start and deceleration fuel cut-off events, identified from the narrowband O2 sensor signal, resulting in cumulative fuel consumption within 3% of the dynamometer results.