We conduct a discrete choice experiment on 931 solo-driving commuters in Lyon, France to estimate the values of end-to-end travel time (VoTT) in the presence of an HOV lane for four modes: Solo Driver, Carpool Driver, Carpool Passenger and Public Transport. Mixed and latent class logit models are estimated. We find that Carpool Passenger, Carpool Driver and Public Transport median VoTTs are respectively around 20%, 40% and 60% higher than Solo Driver VoTT. The analysis of individual heterogeneity distinguishes three classes of behavior in our sample: open to carpool as a driver (41%), open to passenger modes (32%) and resistant to all alternatives to solo driving (28%). These three categories allow to identify solo drivers who could switch to carpool as drivers. We show that encouraging current solo drivers to switch to carpool as passengers will be more sensitive if public transport services are also improved.