Regional Patterns Of Swing

A few weeks ago I wrote here about how many seats each of the parties could hope to win in the election, assuming various levels of swing from the 1997 result; but I entered a caveat that this assumed uniform national swing, which in practice is unlikely.

One of the certain factors which will disrupt uniform swing is tactical voting; but at this stage of the election that is still an imponderable. Until the starter's gun is fired in earnest, and the mass of electors begin to be made aware of the tactical situation in their own constituencies, we cannot hope to make a realistic measure of how prevalent it is likely to be, let alone of how many seats it might swing (which can really only be measured with constituency polls in any case).

The second obvious effect is which may upset the calculations is regional differences in swing. These patterns are already emerging (although, of course, there is nothing to say that they are set in stone, and the regional variations which we can see in the current polls may not survive to polling day). We have analysed the voting intention figures in the ten fortnightly MORI Omnibus polls between the start of October 2000, when the fuel crisis effect was beginning to die down, and last weekend, the poll published in yesterday's Times. Altogether this gives us an aggregate of almost 20,000 interviews, big enough to give us robust sample sizes in all the regions. The results are shown in the table. (The analysis is by Government Office Region, which are also the regions used as constituencies in the European Parliament elections.)

Source: MORI Base: 19,979 British adults 18+, October 2000-February 2001

The general pattern is clear. In the South and especially London, the swing since 1997 has been to the government; in the Midlands and parts of the North, the Tories have made limited progress. What are the implications of this pattern?

This depends, in fact, on the precise national levels of support to which this pattern of regional variation is added. This is because regions vary in the number of marginal seats that they have, and in how marginal those seats are. At some levels of support, the boost of extra swing in a particular region over and above the national swing might be enough to deliver some seats which would not fall on the national swing alone, but if the national swing is big enough to win those seats anyway, or the regional boost is not enough, then there is nothing to be gained from the pattern. At the same time of course, in other regions of the country where the Tories are doing worse than the national average, the converse applies.

When we try applying this pattern at various different levels of swing, a curious fact emerges: as the table shows, if the swing is 3% or less, or if it is 8% or more, the current regional pattern benefits the Tories. However, in the band between 4% and 7% swing, the Tories win fewer seats on this pattern of support than they would do if there were a uniform national swing. Very awkwardly for Mr Hague, this band falls right across the target of 60-70 gains that is currently being bandied about as a reasonable criterion for his survival as leader. On the other hand, if he confounds the current state of the polls and does well enough to find himself on the verge of Downing Street, suddenly the regional variation kicks in to help him, with the biggest bonus being precisely at the key swing of 11%, which with the regional bomus would be enough to carry him past the 330-seat threshold.

Of course, it must be emphasised that uniform regional swing is as unrealistic an assumption as uniform national swing; even when there are broad regional variations, they are unlikely to fall conveniently into the neat geographical pigeonholes into which we divide up the country. It is sometimes possible in retrospect to look at an election and identify the local trends that occurred, regardless of arbitrary boundaries, in small groups of constituencies; but opinion polls are not a sufficiently sensitive instrument to measure at any finer level than the regional, unless there are dedicated polls designed for the purpose. Nevertheless, the broad pattern of support over the last five months is an intriguing one, and demonstrates that matters are rarely so simple as some particular factor such as regional variation always working in one party's favour.