Lewis Baston is author of Reggie: The Life of Reginald Maudling and several books about British general elections. He is a consultant on politics, elections and constituencies.

Harold Wilson once described a Prime Minister’s decision to go for a general election as being the loneliest one in politics. Having led his party into three in office – one that went brilliantly, one that went OK and one that backfired – he had more experience in this than any other Prime Minister except Lord Salisbury.

David Cameron, Winston Churchill and Neville Chamberlain are members of a rare club of Prime Ministers who did not have to make any decisions about election timing. Cameron was locked in to a five-year term under the coalition agreement and the Fixed Term Parliaments Act. Perhaps his relaxed style of leadership from 2010-15 owed something to the absence of the constant existential stress of thinking about election timing, although the timing of the EU referendum overshadowed his second government.

Theresa May, having discarded the comforting straitjacket of fixed terms, joined a long line of Prime Ministers who have faced this solitary dilemma, and a somewhat smaller group who were damaged by their choice. We can break general elections since 1918 down into four slightly fuzzy categories – ones that Prime Ministers can regard as complete successes; ones over which Prime Ministers had limited or no choice, debatable ones (including counterfactual elections, such as those of October 1978 or November 2007)…and ones that go completely wrong.

Elections that go according to plan: 1922, 1931, 1935, 1955, 1959, 1966, 1983, 1987, 2001, 2005

In ten elections, the incumbent party was re-elected with its authority enhanced, and with at the very least a comfortable majority. Four Prime Ministers can therefore claim to have an unblemished record in choosing election dates: Margaret Thatcher and Tony Blair both had two election calls and in each case won three-figure majorities.

Forced elections: 1918, 1929, 1945, 1950, 1964, 1997, 2010, 2015

Seven elections were forced on Prime Ministers at or approaching the end of a Parliament’s term. Two – 1918 and 1945 – were overdue. In 1945, Churchill did not have a choice, since his Labour partners would not extend the coalition and prolong the Parliament any further after VE Day.

The elections of 1964 and 1997 came at the very end of five-year terms during which unpopular governments had lacked opportunities to call earlier elections. There was perfunctory discussion of a spring election in 1964, but it was not a serious prospect. The Conservatives had a good summer and lost only very narrowly in October: Alec Douglas Home’s timing was as good as anyone could have demanded.

The election of February 1950 is more uncertain because it would have been possible for Clement Attlee to have gone on until summer 1950. The Cabinet, with the exception of Herbert Morrison who was more attuned than most to the mood of swing voters, were agreed. The advice of Stafford Cripps, the Chancellor, who did not want to deliver a pre-election Budget and was fretting about sterling, proved decisive: Chancellors are often the most hawkish about going for early elections.

The debatable cases: 1924, October 1974, and the elections that never were of 1978, 1991, and 2007

October 1924: The Labour government of 1924 made the choice to lose a confidence vote rather than be nice to the Liberals. Although the Tories won a larger than expected majority, Labour achieved its strategic aim of creating a two-party system in which the Liberals ceased to be a credible contender for government.

October 1974: Wilson attempted to replicate his 1964-66 consolidation of power, but it did not work out so well. Labour’s majority was only three seats. The Parliament elected in February was unsustainable, and another election was clearly needed; the question is whether the election should have come in May or June, when Labour’s poll lead was larger.

October 1978: As in 2007, election speculation ran strong, but Jim Callaghan did his Duke of York act more elegantly than Gordon Brown, at one stage breaking into music-hall song about “waiting at the church’” Callaghan himself felt that the best that Labour could do was largest party in a hung parliament, and he was already tired of the grinding stress of governing without a majority.  For what it’s worth, I think Labour would have won in 1978. Callaghan was personally popular and the economic news was positive. Win properly or lose big was Callaghan’s gamble in going on to 1979, rather to the horror of his colleagues. After the winter of discontent, a death wish seems to have overcome Callaghan: a bit more chicanery could have won the confidence vote and put the election back to June or October, and every month further from the winter would surely have saved a few Labour seats.

1991: Had the Conservatives lost in 1992, this would have gone down as a missed opportunity for a new mandate. Had Michael Heseltine won the leadership, he would have gone to the polls quickly to assert his control over the Parliamentary Party. John Major was less of a risk-taker, and a 1991 election would have been complicated by economic recession, party splits, the Gulf War and the prelude to the Maastricht Treaty. It is doubtful that an election, other than one under Heseltine in February, could have produced a better result for the Conservatives than did April 1992.

2007: Ten years on, Gordon Brown’s eventual decision not to have an election in autumn 2007 looks more sensible than it did. As a new Prime Minister, Brown was popular and Labour were bounding ahead in the polls. A campaign stressing his workmanlike merits – ‘Not Flash, Just Gordon’ – was on the stocks, and it looked as if an election would provide a Labour win with a reduced majority. If successful, Labour would have gained another two years of office in 2010-12 – useful in riding out the coming economic troubles.

However, a 2007 election did have some pitfalls. We have now seen how an unflashy, introverted Prime Minister can fail to carry a personalised election campaign on her shoulders, and the same could easily have happened to Brown in 2007. As the successful Conservative conference of that year showed, it would have been foolish to underestimate the tactical wiliness of Cameron and George Osborne. There were certainly some people at the time, such as Chris Mullin, who thought an election unnecessary:

“Arrived [at Labour conference in Bournemouth] to find the place in the grip of election fever. Until now I had assumed it was all got up by the media but it appears that the clever young master strategists around Gordon are furiously talking up the possibility on the strength of a couple of good polls and some crap about Gordon needing his own mandate. Pure insanity.” (23 September 2007)

Although the decision was a closely balanced one, Brown failed by allowing expectation to reach near-hysteria. This build-up and the unconvincing denial that the decision had anything to do with the polls smashed his image as a steady, serious leader. If one is going to call a snap election, it should be done Theresa May-style, with no preparatory softening-up, even if that is also a risk to one’s reputation for telling it straight.

The botched elections: 1923, 1951, 1970, February 1974, 2017

1923: The 1923 election was called only a year into its term by a government that wanted to pursue a radical change in international trade policy, but it resulted in the loss of the first majority that the Prime Minister’s party had won in over 20 years.

1951: Rather like 1979, this election was brought on by governmental exhaustion. It came after a year and a half with a majority of six seats, war in Korea and a deteriorating economic situation – and out of courtesy to allow the King to go abroad and not leave a potential crisis behind him. Clement Attlee might have done better to hang on longer, and perhaps stay in power for another decade of greater prosperity. Electorates tire of austerity, and in 1950 and 1951 Labour were the party of austerity.

1970: Harold Wilson called an election soon after the first polls and then local elections showing Labour doing well after a deep mid-term trough. Although it was Wilson’s personal preference, he did take it to a political Cabinet: “everyone agreed? Right: then no one will be able to claim the virtue of hindsight”, Barbara Castle recorded him saying. Wilson escaped much personal blame when it all went wrong. It might have made sense for him to ensure that Labour’s recovery was firmly established before going to the polls, but there was also the risk of strikes or economic bad news over the summer of that year. Although probably a wrong decision, Wilson’s June call was based on logic and calculation rather than the muddle that enveloped governments in 1951 and 1974.

February 1974: I have written previously on this site about February 1974, which was a composite of every election timing mistake that one could make; Heath might well have won had the election come three weeks earlier; the attempt to keep the agenda on a single subject was doomed to failure, and the opposition was assumed to be too disorganised and extreme to mount a good campaign.

Bad election calls are often the result of underestimating the opposition. Fighting an election is a great unifier of divided parties; the guns deployed in party infighting are quickly trained on the external enemy. In 1905-06 and 1923 the coming of an election united the Liberals. In February 1974 and 2017, the election provoked a sudden outbreak of fraternal feeling within the Labour Party. Even the most divided opposition can unite around a wish list of policies and a negative critique of the incumbent government, even if these are not sufficient to win outright.

The other problem is that famous Harold Macmillan trope about ‘events’. An election campaign is a concentrated series of events. Bad economic figures and the machinations of Enoch Powell upset the calculations of Wilson and Heath in 1970 and February 1974. A further problem hits campaigns that try to be about a single issue, as with Heath’s ‘Who Governs?’ election in 1974 and May’s bid for an enhanced Brexit mandate in 2017 (and Malcolm Turnbull’s election in Australia in 2016). The public’s attention will wander during the election, and opinion can easily sour on a government if living standards are under pressure. However, ‘events’ can also make fools of Prime Ministers who pass up a opportunity: Callaghan ran into the Winter of Discontent, and Brown into the great financial crisis.

Prime Ministers tend not to repeat the mistakes of their predecessors, but instead to make the opposite error. So it was that Callaghan decided not to go for an election as soon as he saw a gap in the clouds in the second half of 1978 because he had seen what had happened when Harold Wilson rushed to the polls in 1970.

Brown fuelled election speculation in autumn 2007 –  and then lost credibility by not following through.

May, having seen what happened to Brown, refused to encourage any election speculation in 2017 but then took the initiative. In contrast to all the other election botches, the indicators at the time she called the election were pointing to a comfortable win. She might have stumbled because the underlying rule that campaigns don’t matter has been quietly revoked by the electorate during the last two years rather than because of any particular obtuseness in her own judgement.