This is a question not easily answered, but Noah Smith, a candidate for a Ph.D. in economics, makes a really good effort here in his Noahpinion blog. He discusses both clearly and at length the scientific deadend of deducing theories that don't fit real-world data and the equally sterile pursuit of mining data but inducing no explanatory theories. An economist--or other scientist--who does only one or only the other is stuck, which is what economics has been for a long time.  Smith points to some green shoots he hopes mean economists may be getting unstuck.