Toronto Centre includes Rosedale, which is insanely wealthy, where the NDP vote is always in the single digits. While Rosedale makes up only a small portion of the riding, it is enough to ensure the NDP can never win the seat. South of Bloor, the riding is much poorer, but the seat's Liberal (and Red Tory) history has meant the Liberals are strong there as well.
In the federal election, we saw the Liberals finish 2nd in many polls across the riding (to the Tories in the north and the NDP in the south), which was enough for them to win.
This explanation has never convinced me, because I don't know how Liberal/Tory popularity in Rosedale can somehow sway voters south of Bloor. If the NDP is genuinely appealing to voters in the south part of the riding, those voters would be going NDP in much larger numbers than they historically have. The party has not typically fielded good candidates, but they must claim the blame for that. You're right about the 2011 federal election, but on the new boundaries (which exclude Rosedale altogether) Rae would have won by 3%. And none of what you cite can account for the fact that the NDP's performance in the provincial election was so dramatically worse than it was in nearby ridings. The party won virtually no polls, and Krago's numbers show that boundary changes wouldn't have helped.
To me, this question is difficult to answer. Voters in Regent Park, for example, voted Liberal even in 2011. Surely the NDP thinks of those voters as its natural constituency, but it could say the same about Jane and Finch--another Grit stronghold, albeit a weakening one. Demographic analysis can only take you so far; at a certain point, a riding's political history starts to matter. This gets to your point that the Liberal institutional presence, buttressed by Rosedale, is an important contributory factor to NDP weakness in Centre. But it can't just be Rosedale that created that history.
