To build on Eric's comment,

SQL joins give you a bit more flexibility in joining datasets but are more limited in the operations that can be preformed at the time of execution.

One word of advice, in most situations a DATA step is sufficient for merging datasets using "BY" and "IN=" statements/options, and the logs gives you valuable information about what has happened in the datastep. However, with SQL the log is of very little help, it assumes that you know exactly what you wanted and coded it as such. So, be VERY careful to make sure that you know if/when you are dealing with Multi-to-Multi joins.

My rule of thumb is if I'm just joining two sets, I use SQL; it's usually faster and requires no sorting. If I need manipulate the records, I'll usually use a DATA step as it's got a bit more flexibility, not to mention the increased readability.