You should use dummies where you have dichotomous (e.g. male, female) or categorical data (e.g. married, single, divorced)
In the latter case you always have 1 less dummy than you have categories. So if you have 7 categories, you should have 6 dummy variables to represent it - which is why it is always easier to measure age continuously, rather than in age brackets! (as I found out to my cost)