Interesting but I'm not sure these numbers tell us a lot about the power of incumbency in Canada.
Some thoughts:
1) Throw out 1993, it is an outlier
2) You can't really compare the Liberals to the other parties because they have held more seats over a longer period of time; so there is more opportunity for them to lose/pick up seats. You would have to normalize the data some how to do a fair apples to apples comparison between the Liberals and other parties
3) Look at standard deviations to measure volatility (the sample sets are too small to be meaningful, but this is for fun)
4) BQ has the highest SD by far, so a lot of the volatility is in Quebec, which may mean the Liberals and BQ are constantly changing seats between them. This dynamic doesn’t really affect any of the conservative parties
5) NDP also has a high SD. So if you’re an NDP, you’re constantly trading seats (again with the Liberals?)
6) So my guess is, each party has a strong-hold region in the country, and where this is no dominating party it is a battle between Lib/NDP; Lib/BQ
7) If you’re a PC between 88 and 00 (ex 93) you have the exact same low volatility as if you’re a Liberal between 04 and 08. What does that say? I have no idea but either way you’re not losing your seat.

Bottom line, incumbency is important, sure… I think we know that intuitively. But I’m not sure it is as important as a “96%” next to the Conservatives row would suggest. If you factored out Quebec, the difference in volatility between the parties would probably drop and the Quebec issue probably depends more on the waxing and waning of the separatist movement than on incumbency per se. Also, Canadians vote for an MP, but they often decide based on the leader and would-be PM, so it would be interesting to see the correlation between the party leader’s popularity in a particular region and seats held and compare that with some incumbency statistic.
 

Ron, the lion's share of the seats the NDP lost in 1993 was to the Reform Party out west. They've won many of them back, notably in BC, since then. It's a real Ontario bias to only think of an NDP/Lib switch.

Of course you're correct to say that 96% is high for the Conservatives right now, because we haven't had a similar watershed election for them yet, as with the PCs in 1988.

In a separate blogpost, I recently compiled counts of MPs by year of first election, and concluded that with the number of (primarily Liberal) MPs who were first elected in 1993 or earlier, we may be set for a big non-incumbency-related turnover in the House of Commons soon.

http://www.punditsguide.ca/2009/11/no-commons-for-old-mps-2008-mps-by-year.php

Thanks to Samara for the very kind citation, and for the very interesting work on retired MPs.
 

“In a separate blogpost, I recently compiled counts of MPs by year of first election, and concluded that with the number of (primarily Liberal) MPs who were first elected in 1993 or earlier, we may be set for a big non-incumbency-related turnover in the House of Commons soon.”

With all due respect, I’m struggling to understand your point. If you’re saying that there is a “big non-incumbency-related turnover” set to happens because a bunch of MPs that are old and are getting ready to retire, I’m sure that is true. But the phenomena of people getting older and retiring is found in many industries and not just politics.

If you’re saying that a “big non-incumbency-related turnover” is related to one party set to lose a bunch of seats, again, I’m sure that is true although I’m not sure how you can be “concluded” based on historical data. I think the more likely scenario is an election of a majority government that takes away seats from the other parties. But again, I’m not sure it is anything new to say that an election can change a government in a democracy.

If you’re saying that incumbency is important, again, that is probably true to some extent. But exactly how important cannot be derived just by looking at average number of MPs returning to their seats after an election.

Again, with all due respect, holding up one number (an average) and saying it proves a point (incumbency is important) is bad analysis. What other variables have been considered?
 

Ron, The Samara blog is clearly directed at academics. The Pundits' Guide is a web database of statistics that can be used for decision support by political professionals. They have two different purposes and two different audiences.

I'm not making a "point" in the sense of a graduate seminar. Incumbency is one factor to consider when analyzing how vulnerable a seat might be, and how high to prioritize it in a list of target seats. So would be the traditional voting patterns and demographics in those seats, along with the parties' candidate recruitment, and spending patterns. My point about the incumbents set to retire, if you read the blogpost I referenced, was that many had been elected in non-traditional seats for the Liberal party, but were elected in 1993 when both the right was split and the NDP vote fell well below average. In spite of those two conditions changing since then, many of those incumbents have held on. But once they leave all bets are off in the demographically non-traditional seats.

All political parties pay special attention to open seats in their election planning, especially if the seats are potentially competitive. One would think that in the wake of the Massachucetts Senate race, the point should be well understood that under the right conditions no seat is safe, but particularly no open seat.

You ask about other variables, and indeed I've also studied with Professor Stanbury the correlation of candidate spending levels with their vote share. All these factors have an impact. It's the role of a political professional to take them into account in devising a strategy for effecting political change. And it's the role of the academic to measure how they did afterwards.