Written evidence submitted by Professor
Michael Thrasher (PVSCB 11)
A short note on electoral bias:
Elections fought under FPTP rules often result
in outcomes that are disproportional. This often involves a "winner's
bonus" whereby the largest party overall in terms of vote
share receives a larger share of seats. FPTP often results in
a small bonus for the second placed party also. It is those smaller
parties whose votes are broadly rather than narrowly (in terms
of geography) distributed that suffer most in the translation
of votes into seats.
Proportionality/disproportionality is not equivalent
to electoral bias. Bias occurs when two (or more) parties that
obtain similar levels of voter support nevertheless receive markedly
different seat shares. There are a number of factors in any electoral
outcome that combine together to explain the distribution of bias.
In recent elections the bias has favoured the Labour party but
this pattern is by no means fixed.
There is a common misconception that periodic
boundary reviews should remove electoral bias. This view is mistaken
because such reviews are only concerned with one element that
contributes towards bias, viz., unequal electorate size
(malapportionment). Other elements are contributing towards overall
bias. Apart from malapportionment these remaining elements are,
vote distribution (geography); differential turnout (abstention);
and the effects produced by competition from smaller parties.
There are, in addition, the interaction effects that result from
two or more of these components interacting with one another,
for example, a party wins its seats in small electorate areas
where abstention is also high.
Using a new method for decomposing bias for
three-party systems (Borisyuk, Johnston, Thrasher and Rallings)
the following three Tables reveal the size and distribution of
the separate bias components for the actual (Table 1) and estimated
(Table 2) 2005 general election and finally the recent 2010 general
election (Table 3). The total bias (positive and negative) for
the three main parties is also shown.
Table 1 shows the decomposition of electoral
bias for the actual 2005 general election. Half of Labour's 83
seat positive bias is a function of its effective vote distributioneffective
in the sense that it does not accumulate excessively large numerical
majorities and neither does it acquire a large share of votes
in seats that it does not win. The rule, "win small, lose
big" operates in FPTP systems for parties wishing to optimise
vote distributions. Labour is also benefitting relative to its
competitors from low turnout. It is also apparent that malapportionment
although a significant contributing factor to Labour's bias advantage
in 2005 was responsible for only 11 of its 83 (13%) seat advantage.
However, it is worth noting that the electoral component accounted
for 12 seats of the Conservative 30 seat negative bias (40%).
COMPONENTS OF THREE-PARTY BIAS FOR ACTUAL
2005 ELECTION RESULTS
Table 2 shows the decomposition of bias following the 5th
Periodic Boundary Review. The overall bias advantage to Labour
was reduced to 75 seats while the Conservative party benefitted
overall. It is most important to note that the Review Process
reduced the bias resulting from unequal electorates. Labour's
advantage fell from 11 to four seats; the Conservative disadvantage
was halved (from -12 to -6). The review process assisted
the Liberal Democrats in terms of electorate bias but a spillover
effect was an increase in negative bias from a recalibration of
its vote distribution. Arguably, if the Review had been able to
equalise electorates to an even greater degree then the malapportionment
component may have disappeared altogether.
A crucial point is that the boundary review process takes
no account of either turnout (abstention) or, formally speaking,
the partisan nature of any constituency re-drawing that takes
place. It is not, therefore, impacting directly upon the abstention
or geography components.
COMPONENTS OF THREE-PARTY BIAS FOR ESTIMATED ELECTION
A major part of the criticism of the boundary review process
is that the electorates that are used to calculate the new constituency
boundaries are out of date by the time the review is completed
and even more so by the time of the election fought on those new
If this criticism is substantiated then we should expect
the electorate bias component to feature prominently in the distribution
of bias following the 2010 general election result. Table 3 shows
that the electorate bias component is indeed larger. In absolute
terms there is a two seat greater advantage for Labour, an additional
one seat disadvantage for the Conservatives while the two seat
positive bias for the Liberal Democrats reduces to one seat. However,
most bias is again contributed by the geography and abstention
DECOMPOSITION OF BIAS (2010 GENERAL ELECTION)
More generally, the overall bias favouring Labour reduces
to 63 seats and a negative bias for the Conservatives now becomes
a positive one of 12 seats. The bias against the Liberal Democrats
rises from -52 to -76 seats, reflecting the party's
vote and seat distributions in 2010.
20 July 2010