Four days to go, and it’s about now that election coverage typically reaches its low point, with pundits having lost all sense of perspective and become totally immersed in the minutiae of the campaign. Hence their obsessive mantra that “it’s going to be close”, blissfully unaware that they said the exact same thing only six years ago, in 2004.

So this is probably a good time to stand back a bit and try to say some very general things about what we know about Australian electoral behavior, and particularly about the election record of first-term governments.

First rule: first-term elections swing against the government. This is easily the most well-confirmed generalisation; of the 10 first-term governments to face the polls since 1910 (when the two-party system was established), every one has lost ground in votes and seats (you could argue about the two-party-preferred vote in 1919, but the government certainly lost seats).

Second rule: despite the first rule, first-term governments win re-election. This is of more recent vintage, but still pretty well-established; first-term governments went down in 1913, 1914 and 1931, but none of the six since then has been defeated.

Third rule: close elections go the government. This one is a bit harder to verify, because you have to define what you mean by “close”, but averaging the measures of seats and (two-party-preferred) vote I came up with the following list, in descending order of closeness: 1961, 1954, 1998, 1913, 1940, 1969, 1990, 1974, 2001, 1951, 1993, 1980, 1987. Only one of those 13 was won by the opposition, and that was way back in 1913.

There’s clearly some relationship between these three rules. Given the third rule, for example, you expect first-term governments to start out with a substantial majority, making it more likely that they will lose some ground next time and also that they will have enough of a margin to survive. But you could equally well argue the other way around; the direction of causation, if any, is uncertain.

Another interesting thing about the rules is that they don’t seem to apply to state elections. The first rule has held up reasonably well in recent times, although there are a few exceptions (Queensland 1998, South Australia 1982), but the other two carry no weight. Lots of state oppositions have won close elections, and most have then gone on to win swings in their favor the next time around — this has been a common pattern over the past 20 years.

So could the rules be just a matter of coincidence? I calculated that the odds against the run of government successes in close elections were about 43-1. Of course that’s sensitive to just how you set the parameters, but it’s a large enough number to suggest that something more than chance is at work.

But it’s quite mysterious what that might be. Many commentators talk as if the third rule was easy enough to explain by the way the government benefits from the advantage of incumbency and superior campaign resources in marginal seats. That would account for cases where the government does better in seats than in votes (for example, 1990 and 1998), but it doesn’t explain why those swings happen to stop just short of the point where the government would lose office.

Nor does it tell us about elections that go the other way, such as 1974.

Or why the same advantages don’t work for state governments.

It’s unsatisfying to have to say that something is going on but we don’t know what. I’m rather a sceptic about induction to start with, and I tend to think chance has more of an influence than we like to admit; it would be unwise to disregard the three rules entirely, but I don’t think they should be treated as gospel truth.

So don’t be surprised if one or more of them falls over before long.

Indeed, I think there’s a fair chance that the first rule will be falsified on Saturday, but that’s still an unpopular view.