Since 1979, British governments have tended to hold elections in the Spring.  According to Nigel Lawson’s memoirs, this was because Mrs Thatcher, having won in the Spring of 1979, thought that May and June were her lucky months, but subsequent Prime Ministers continued the practice.  Of the last 11 elections, only one (the most recent) was not held in April, May or June.  Subject to the five-year limit, the Prime Minister controlled the timing of the elections until the passage of the 2011 Fixed Term Parliaments Act, and are likely to do so again if the current government carries out its pledge to repeal that law.

I have been wondering for some time if holding elections in the Spring is sensible on the Government’s part?  In other words, is the public likely to look most favourably on its government in the Spring?  Or is there no particular seasonal patten to government support?  The most recent election was held in December, and it was a triumph for the government, and the election before that was in June, and was widely regarded as a disappointment.  

Obviously, we cannot know what would have happened had the 2019 election been held in the previous, or following, Spring.  However, as I have argued before on this site, final opinion polls are an excellent predictor (around 90% correlation) with general election results.  So if opinion polls are on average more favourable to the government in the spring, we can say that conventional wisdom is correct, and government popularity increases as the days lengthen and the weather gets warmer.

Conclusion

From looking at the polling data, I have concluded that the raw data seems to support the idea that Spring is the best time to call an election.  However, correcting for the advantage that the government can enjoy to get favourable ratings just before an election eliminates this finding.

Data

The UK polling report site has extremely comprehensive polling data from 1970-2020.  Over such a large sample (>5,400 polls and 50 years) any outlying years should even out.  I have calculated the average government lead (Conservative 1970-4, 1979-97, 2010-2020; Labour 1974-9, 1997-2010) by season (Winter – Jan-Mar, Spring – Apr-June, Summer Jul-Sept, Autumn Oct-Dec) and by month.

Results

	Averagegvt lead (%)
Winter	-2.4
Spring	-1.2
Summer	-2.7
Autumn	-1.9
Total	-2.1

	Averagegvt lead(%)
Jan	-2.15
Feb	-2.08
Mar	-2.94
Apr	-2.97
May	0.34
Jun	-1.11
Jul	-3.22
Aug	-2.38
Sep	-2.48
Oct	-2.76
Nov	-1.05
Dec	-1.74
Total	-2.05

Looking at the seasonal data, the conventional wisdom seems to be correct: there is a small but significant advantage to a government in facing the voters in the Spring if opinion polls are to be believed.  Holding an election in the Spring (lead -1.2%) rather than in the Summer (-2.7%) might increase the government’s lead by 1.5%.  This may not seem very much, but according to electoralcalculus.co.uk, it could correspond to an increase in the government’s majority of 16 seats, making the difference between a hung Parliament and a small majority, or a small majority and a working one.  I was surprised to note that Summer seems to be worse than the Spring.  This means that it is unlikely to be good weather that makes for higher government support.

Turning to the monthly data, May seems to be the best month – indeed the only month in which the government is ahead on average.  June is also a good month to go.  April is, however, the worst in the year.  The 3.3% difference can lead to a change in the government’s majority of 36 seats.  

Is the Government’s ability to choose the month of the election a source of bias?

The main weakness in this approach is the government’s ability (until 2011) to choose the date of an election.  Might that cause bias in the above results?  It is often argued that governments can engineer favourable conditions, and so poll leads, at the anticipated time of the election to win more seats.  If governments mostly anticipate going in the Spring, this could mean that favourable ratings in the Springs of election years could be caused by deliberate government action, rather than any particular change of the country’s mood.

To test this, I eliminated from the ukpollingreport data all polls three months before elections.  Given the large size of the data set, this would not result in any significant loss of robustness in the overall results.  This should have the effect of reducing any bias from the government’s ability to favour the Spring in its timing of elections.  

This changed the results significantly, as can be seen from the table below.  

	Averagegvt lead (%)
Winter	-2.6
Spring	-3.7
Summer	-2.7
Autumn	-2.8
Total	-2.9

	Averagegvt lead(%)
Jan	-2.2
Feb	-2.3
Mar	-3.2
Apr	-3.3
May	-4.5
Jun	-3.3
Jul	-3.2
Aug	-2.4
Sep	-2.5
Oct	-3.3
Nov	-2.6
Dec	-2.6
Total	-2.9

Eliminating the months before general elections in fact makes the Spring months the worst in which to hold an election, and, counter-intuitively, the winter months are significantly better than average and May is now the worst of all.  There is no support from the data for the notion that the public is more favourably disposed to the government in the spring than at any other time of year.

Betting implications

The most obvious betting implications for these results are for those who want to bet on whether Labour will have an opinion poll lead by the end of the year.  As no general election is likely, the latter table seems the more appropriate.  A government lead seems slightly likelier than average in November and December, though the difference is not very great.  Those expecting the Opposition to benefit significantly from the approaching dark nights seem likely to be disappointed.

Fishing