The phantom (elector) menace?

A curious little anomoly in the by-election related statistics led this corner to produce the following pretty pretty graph.

It shows the percentage change in the total number of voters on the electors list in each provincial by-election since January 1, 2000, as compared to the number of electors on the list in the same district in the preceding general election. Green bars indicate growth, red is decline:


Why on earth were the lists almost always smaller in by-elections, prior to the last general election? Why have they almost always been larger in each by-election since? Given that the strange little trend applies to most suburban St. John's districts as well as to rural areas that are experiencing population decline, what can account for this picture? And why is the "growth" in electors so large as of late? Theories, or even wild and uninformed speculation, is more than welcome.

By way of comparison, here is the equivalent graph for all federal by-elections during the same period, with the exception of the three most recent ones. (Final tallies of electors in those three have not yet been published by Elections Canada.) The vertical scale is the same in this graph as in the previous one. Federal by-elections in Newfoundland and Labrador are marked with an asterix.



Only two federal by-elections during this period have had voters lists that varied in size by more than 5% from the previous general election.

Only five provincial by-elections have had voters lists that so varied by less than 5%.