Representations offer many controls over how the data is displayed. They are time consuming to set up, but can be well-worth it for if you are going to displaying the same data over and over. Representations can help one avoid having to edit the data just to make it display correctly. The underlaying data stays the same and the Representation instead handles its display. For example, if you have two coincident features, you can use the controls in Representations to show them side-by-side instead of overlaying each other, which may make the idea clearer to users if it is showing subject matter like bus routes both on the same street.
Layer files are far easier to set up, but are more limited in how they can affect the display of data. For the data to be displayed clearly, you may end up having to edit you base data just for display clarity. While this can be workable for a one-time project, if there are a succession of maps of the same basic material that need to be made and there are data changes, one will have to go back and re-edit the data not just for the changes but also to make it display correctly again. Plus you can end up with multiple data sets, a source layer and one or more display layers. This offers potential for confusion when data edits need to be made - which parts of the data are there only to make it display correctly? You have to remember what was edited for display and why, and then redo it every time.
So it usually comes down to how complex the display is of your data, and whether you will be using the data over and over for maps in comparison to a one-shot project.
Chris Donohue, GISP
