A re-run of Andy Cooke’s December 2007 guest slot

(This analysis first appeared on PB in December 2007 and has been much quoted both here and in the mainstream media since. It’s even more timely today and I thought it was worth re-running – MS)

One of the most widely held items of received wisdom is “The opinion polls always swing back towards the Government as the Election approaches”. That mid-term blues will always overemphasise an Opposition’s performance in the polls and these will wear off as the Parliament winds down towards the next election.

Received wisdom can well be right – there is always a reason somewhere along the line that any given fact ends up as received wisdom. But sometimes it can be wrong. Usually when this happens it was a truth that has worn off, that has gradually ceased to be so, leaving only the story of its presence behind.

The thing is – in cases such as these, many people still end up betting on defunct received wisdom, which leaves a whopping great opportunity for the more questioning among us. The trick is, of course, to identify which items in the pantheon of received wisdom remain valid and which are now dead. Whether you agree with the thrust of this article or you don’t, at least it should cause you to question this belief – so if you remain a believer, at least you will now be more confident in your belief.

So – let’s have a look at “The opinion polls always swing back to the Government.” The main reason that this popped up a bit of a flag for me is that when I noticed this argument, I thought – hang on. I don’t thing that they did last time. The long period of Tory rule meant that a swing towards the Government was exactly equivalent to a swing towards the Tories. If the rule was valid, there should have been such a swing towards Labour in the last two elections – which doesn’t appear to be so.

Let’s look at the facts. As is frequently pointed out, it’s very difficult these days to compare polls from before the “great methodology change” of 1992-1997. However, MORI have pretty much sustained their “all adults naming a party” methodology and to their great credit have made available their archives back to 1983. This provides a decent baseline for the traditional methodology (the one under which the “rule” was coined).

ICM were the pioneers of the new weighting methodologies, with “spiral of silence” adjustments, first applied in 1994. They have made available their archives since 1990 and they have retrospectively adjusted their 1992-1994 figures to meet the new methodology (Thanks to Anthony Wells at ukpollingreport for making all of these figures available in a very convenient form). Accordingly, we can compare equivalent polls from a Tory Government to the last two Labour Governments in order to see correlations in swingback between midterm and polling day.

For each, what we’d expect to see is the Government (after a short honeymoon period) losing support until the mid-term and then turning it around and steadily recovering until the polls. Individual polls should vary about this tendency, but comparing the average score for the middle year of the Parliament with the following election should provide the data that we’re after.

MORI – LONG BASELINE (1983-2005), “TRADITIONAL” METHODOLOGY

1979-1983 Parliament

Mid-term (calendar year 1981) average: Con 29.5, Lab 36.5, Lab lead 7%

1983 Result: Con 44, Lab 28, Con Lead 16%

So a swingback to the Government (or the Tories) of 11.5% as against the mid-term polls

Note: The “Falklands Effect” must have had some level of influence upon this result – so at least some of the swing to the Conservatives would be over and above the “underlying” recovery

1983-1987 Parliament

Mid-term (calendar year 1985) average: Con 35.5, Lab 35.5, Lab lead 0%

1987 result: Con 43, Lab 32, Con lead 11%

A swingback to the Government (or the Tories) of 5.5% as against the mid-term polls

1987-1992 Parliament

Mid-term (June 1989-June 1990): Con 36.5, Lab 48, Lab lead 11.5%

1992 result: Con 43, Lab 35, Con lead 8%

A swingback to the Government (or the Tories) of 9.75% as against the mid-term polls

Note: The near death of the third Party and its recovery as the Parliament wound on has to have had some effect. Again, I suggest that the recovery of the Tories was enhanced by this

1992-1997 Parliament

Mid-term (June 1994-June 1995): Con 24.5, Lab 56, Lab lead 31.5%

1997 result: Con 31, Lab 44, Lab lead 13%

A swingback to the Government (or the Tories) of 9.25% as against the mid-term polls

So far … so what? The rule seems to be working fine. Strong correlation between the rule and reality – at least 5% swingback for the Governing Party, enhanced by winning wars or events such as near-demise of the third Party mid-term. However – we’ve only explored the situation for a Tory Government. We could equally argue that a rule stating “there is aways a swingback to the Tories from mid-term to polling day would be equally valid. To seperate the two conditions, we need to explore what happens with a change of Government. And now we have a change of Government and we can “press to test”

1997-2001 Parliament

Mid-term (calendar year 1999) average: Con 27, Lab 53, Lab lead 26%

2001 Result: Con 33, Lab 42, Lab Lead 9%

So a swing away from the Government (or towards the Tories) of 8.5% as against the mid-term polls. Hmm.

2001-2005 Parliament

Mid-term (calendar year 2003) average: Con 29, Lab 43, Lab lead 14%

2005 result: Con 33, Lab 36, Lab lead 3%

So a swing away from the Government (or towards the Tories) of 5.5% as against the mid-term polls. Again.

Maybe it’s isolated to MORI. Let’s look at the ICM results.

ICM – SHORTER BASELINE (1992-2005), “MODERN” METHODOLOGY

1992-1997 Parliament

Mid-term (June 1994-June 1995): Con 29.5, Lab 49, Lab lead 19.5

1997 result: Con 31, Lab 44, Lab lead 13%

A swingback to the Government (or the Tories) of 3.25% as against the mid-term polls

Okay – still a swingback to the Government – if not as strong as the MORI polls showed.

1997-2001 Parliament

Mid-term (calendar year 1999) average: Con 30, Lab 48, Lab lead 18%

2001 Result: Con 33, Lab 42, Lab Lead 9%

So a swing away from the Government (or towards the Tories) of 4.5% as against the mid-term polls. Again – a swing away from the Labour Government – less pronounced than the MORI figures show, but definitely not the 5%+ swingback to the Government that we’d expect and that those that cite the rule tend to imply.

2001-2005 Parliament

Mid-term (calendar year 2003) average: Con 32, Lab 38, Lab lead 6%

2005 result: Con 33, Lab 36, Lab lead 3%

So a swing away from the Government (or towards the Tories) of 1.5% as against the mid-term polls. Again a swing away. Again, noticeabley less in magnitude than under the traditional polling system (so we could assume that the “new methodology” acts to dampen large swings in public opinion). But that swingback that we expected is once again absent – any swing here is away

FURTHER BACK

Getting data from further back is very difficult. It appears that the 1974-1979 Parliament had very variable poll scores – to the point where Labour was level in the polls with the Tories in 1978 – but 7 points behind by the election in 1979. This doesn’t really fit with the rule – which tends to be quoted as implying an inexorable pendulum, shifting opinion away from the Government until it bottoms out in mid-term and then swinging steadily back towards the incumbents. Of course, there would be some variation around this central pattern – but the fluctuations between 1978 and 1979 would seem too large to be a variation around a trend of swingback to the Labour Government.

The Feb-Oct 1974 Parliament was rather abbreviated – the “mid-term” would have been May/June 1974. The only data that I’ve got is that before the campaign (end of August/start of September) “Opinion polls showed Labour running 10 points clear of the Tories” (from the BBC website). Their 3.5 point lead in the actual election shows a swing away from the Labour Government from that point of 3.25%.

The last time that I can say with confidence that the “rule” worked for a Labour Government was the 1966-1970 one. And even that one had a surprise swing away from Labour at the end.

CONCLUSION

My conclusion? The rule came about because it was true – at least until 1970. Maybe even to 1979 – with some exceptions that tested it somewhat. It has broken since then. Every election since Maggie Thatcher walked in to Downing Street has seen a stronger showing by the Tories at the ballot box than their mid-term polling levels. And a weaker performance for Labour. Under “traditional” polling methods, swings to the Tories of 5.5%-12.5% have been recorded – averaging over 8% swing in the Tories favour (9% when in Government, 7.5% when out of Government). Under the “new” polling methods, the size of these swings is less – but the direction is the same.

Why has it changed? I don’t know. That’s the realm of speculation. It would be very nice to know why, however. These figures have shown the death of the “Swing to the Government” rule and its regeneration (like a wounded Time Lord) into a “Swing to the Tories” rule. It would be nice to know if this rule also comes with a sell-by date as well.

There may yet be a swingback. But I wouldn’t put my faith in the rule that “Governments always recover from mid-terms”. They don’t – especially, it seems, if the rosettes worn are red. Bet with your head on the trends you see, and don’t assume that there is a pendulum acting for the Government (or even, despite the apparent trend that we’ve seen above, for the Tories).

1 – The swingback rule has not worked for recent Labour Governments; instead there has been a “swingaway” effect.

2 – Even if this has been an aberration, the “new polling methodology” tends to dampen such swings; a swingback to the Tories under MORI of 9.25% (1992-1997) was only 3.25% under ICM.

Andy Cooke