Cars are horrible to try to use as shelter - they're really good at conducting heat right out of themselves due to things like poor sealing, the gap between the floor and the road, etc.
But unless this is in a very, very cold place, temps are more often in the wet cold range than the hard and dry cold range (of course, as I type this, it's twenty degrees or so, and I about froze my face off running in seven degrees last night). If it's wet out, a dry car will be safer than a wet ditch/field, maybe even than a wet wood. With the right insulating/heating stuff, a car will be ok. In deeper cold, of course, one can always take the heating gear outside.
And I know virtually nil about heat dynamics, particularly compared with other people here, but that figure of a candle raising car temp by ten degrees is so imprecise as to be impossible. Raise it from what? How heated is the car already? How cold is it outside. I could be wrong, but I think that both of those factors would affect how fast the car gets cold. Are we starting our candle test when the car is cold? If so, what's the point, unless it's a dead car in which someone has taken up residence? Is it a well-built and insulated small car (can't think of any off the top of my head, but there's gotta be some such critter), a utility van, or an SUV? Seems that with so many variables at play, any figure like ten degrees has gotta be pure invention.
