The Referendum Results
By far the best 'predictor' of referendum turnout, however, is estimated turnout in the preceding general election. This alone accounts for 75.3 per cent of the variation in referendum turnout and if it is introduced into a multiple regression analysis then no other variable significantly increases the proportion of variation explained. This is not to say that election turnout 'influenced' referendum turnout in the same way as the social composition variables. It is simply that the factors which affected general election turnout (with the exception of constituency marginality) also affected referendum turnout with the result that the two are closely related.6 Nonetheless, the regression equation predicting referendum turnout on the basis of election turnout can be used to discover those areas which deviated from their 'normal' turnout level in the referendum.7 Those authorities which deviated by more than three percentage points are shown in Table 6.4. Of the five listed as having lower than expected turnout, four can be easily explained in terms of the expected marginality of constituencies at the general election. Aberdeen contains the highly marginal Aberdeen South and fairly marginal Aberdeen North; Dumfries was a key Labour target seat; Stirling, formerly held by the Scottish Secretary of State, Michael Forsyth, was very keenly contested while Aberdeenshire includes the marginal
% of votes cast
% of adjusted electorate
The former Regional Council areas, while allowing comparisons between the 1979 and 1997 referendums, were large and often internally heterogeneous. As a result, they conceal substantial internal variations. Strathclyde, for instance, was in its day the largest local authority in Europe, and covered areas as diverse as Glasgow's inner city, the depressed industrial towns and ex-coalfields of Lanarkshire and Ayrshire, as well as very affluent commuter suburbs and villages such as Bearsden and Drymen, and rural areas (including remote Highland areas in Argyll and islands such as Islay, Mull and Jura). Politically, too, the Regional Councils covered a wide range. Again in Strathclyde, for instance, in addition to numerous Labour strongholds, the region contained the constituency of Eastwood, by a considerable margin the safest Conservative seat in Scotland after the 1992 general election, as well as the Liberal Democrat seat of Argyll and Bute. We need, therefore, a more detailed map to enable us to draw conclusions about the geography of the referendum results. The 1997 results were issued separately for each of the 32 current local authorities, and although this is smaller than the number of parliamentary constituencies (72) it does provide a finer scale of aggregation and allows more scope for analysis than was possible in 1979.