One question or two? Let's have a referendum

The political and legal row over the Scottish referendum – when it should be held, and whether it should consist of one question or two – is unlikely to be quickly or easily settled. It raises issues important to all parties, north and south of the border.

But, beyond the merely political issues, there are some statistical ones. If, as some suggest, the referendum offers two questions to the electorate, there needs to be absolute, analytical clarity about how the “yes” votes are to be publicly reported and counted. 

If there are two questions, they would ask voters to provide yes/no answers to (i)  independence; and (ii) greater devolution of powers,  in shorthand “devo-max”. Current opinion polls suggest that the latter is more likely to command a majority than the former.

The analytical options are to analyse the two questions separately, or cross-tabulation of the respondents’ dual answers as:

  • a)  Yes to (i), Yes to (ii)  
  • b)  Yes to (i), No to (ii).
  • c)   No to (i), Yes to (ii)
  • d)   No to (i), No to (ii)

In the case of cross-tabulation, the count has to take cognisance also of 4 other combinations because some respondents may opt to answer one only of the two questions on which opinion was sought. If the return remains valid when a respondent answers only one of two posed questions, we have also to count, and report publicly, the following partial responses:

  • e). No response to (i), Yes to (ii)
  • f)   No response to (i),   No to (ii)
  • g)               Yes to (i), No response to (ii)
  • h)                 No to (i), No response to (ii).

The decision on validity affects whether the count for independence is the sum of votes a)+b)+g) versus just a)+b); and whether the count against independence is c)+d)+h) versus c)+d).

The UK Government’s consultation paper, published this week, takes the view that a two-question referendum would be wrong. “If these two questions were taken together, there would be four possible outcomes, and potentially four campaigns, each arguing for a different result” it says. “Having four campaigns would not help to generate clarity.” So the UK Government’s appears already to have discounted e) to h) . . .

There is, in addition, the issue raised by the Liberal Democrat peer Lord Steel. Suppose that a small majority voted for independence, and a large majority for devo-max. Then the more popular option would be defeated by a less popular one. Because his electoral mandate is for a referendum on Scottish independence, Alex Salmond, the SNP leader and Scotland’s First Minister, has made it clear that a majority for independence would trump a larger majority for devo-max.

By contrast, the Constitution Unit at University College London has proposed a two-stage referendum on independence: in principle, and in practice. The first stage would establish if a majority of the Scottish electorate favoured independence: if yes, the terms of such independence would be negotiated. The second referendum would determine if a majority favoured the terms reached.

Without such a process, Scots would be making a decisive decision on the basis of pretty slender information, says Professor Robert Hazell of the UCL unit. But neither side in the debate appears persuaded: Alex Salmond has never accepted the need for a second referendum, and now the UK Government has stated its preference for a single referendum asking a single question.

That avoids the complexities of a two-question referendum but, as the Coalition Government doubtless calculates, forces voters into a trustful choice, some practicalities of which emerge in detail only later.