I personally like the second classification:
__ Rail - streetcar or trolley
__ Rail - subway, metro, or light rail
__ Rail - commuter railroad or Amtrak
My reasons are the following:
1) As a survey respondent it removes the confusion regd. what to answer faced with this question. A regular respondent will say that they take the train, they do not distinguish between subway/light rail vs commuter rail and calling it out in this manner - where we we have rail broken into subcategories will help the respondent choose the right type of rail.
2) As an analyst, having this classification makes it easier to compare across and within modes and geographies. What i mean is this - lets say we did not distinguish that these are 3 rail modes which are classified further as streetcar/light/commuter and geog A has all 3 rail modes available, geog B has only 2 rail modes, and geog C has only one rail mode available. Doing this classification allows the analyst to determine that in all 3 geographies, when a respondent choice is one common mode, it means the same thing and not different things to different respondents in different geographies. It will also allow for better mode choice estimation if the nest definitions are directly from the survey data and not solely reliant on analyst judgement. Further, when looking at results/doing studies later on it might be easier to consider what the aggregate rail mode comprises of when compared to auto and bus.
Krishnan