Look at the following table:
|Oliver Heald||Conservative||26,995||53.5|| +5.5|
|Hugh Annand||Liberal Democrat||11,801||23.4|| +3.1|
|Adrianne Smyth||UK Independence Party||2,075||4.1|| +0.8|
|Rosemary Bland||Green||875||1.7|| +1.7|
|Richard Campbell||Independent||209||0.4|| +0.4|
|David Ralph||Your Right To Democracy Party Limited||143||0.3|| +0.3|
|Philip Reichardt||Independent||36||0.1|| +0.1|
Yes, this is a case where the leading candidate increased his vote, the former second place collapsed into third place, and the previous third place rose to second place.
So how does the Swingometer report this multifaceted situation?
A 1.2% swing from LD to CON.
So how is this calculated?
It's half the difference in the winning party's change in vote and the second party's change in vote. In the case where two main parties contest a constituency, if the winner has a vote rise of 5%, and the loser has a loss of 4% (with the rest disappearing off to the marginal parties), then the difference is 9%, and the swing is 4.5%. This is a reasonably meaningful measure, giving some indication of the proportion of voters deserting one party for the other.
But what we have here is not a two party fight, but more a three party one. Both first and second place parties increased their vote. So although the calculation is (5.5% - 3.1%)/2, giving that 1.2%, the implication of the calculated swing that 1.2% of voters deserted the LDs for the Tories is quite blatantly false.