When South Australians go to the polls on March 15 they should be secure in the knowledge that their voting system promises “electoral fairness”, and that the party that wins a majority is the party that wins government. But that’s not what actually happens.

“Electoral fairness” is laid out in section 83 of the state’s constitution, which requires that boundaries be redrawn after each election so “if candidates of a particular group attract more than 50% of the popular vote, they will be elected in sufficient numbers to enable a government to be formed”. This avoids the not uncommon spectacle of parties winning election from a minority of the two-party vote thanks to fortuitous results in marginal seats, as occurred at federal level to the Coalition’s advantage in 1998 and to Labor’s in 1990.

Or at least, that’s the theory. Section 83 hints at the actual reality of the situation by adding the words “as far as practicable” — which, as experience over five elections shows, really isn’t very far at all. Since the measure was introduced by a Labor government that was limping through a disastrous final term after winning the 1989 election off 48% two-party preferred, the “wrong” party has emerged victorious on two occasions out of five.

The difficulty is that electoral boundaries are an extremely blunt instrument for ensuring the majority party wins office, and they can only promise to deliver if the overall swing is precisely uniform. The last election in 2010 made a mockery of that assumption, with Labor copping huge but ultimately harmless double-digit swings throughout its Adelaide heartland, while fighting brilliantly successful rearguard actions where it mattered most.

The final score was a fairly comfortable Labor majority of 26 seats out of 47 off just 48.4% of the two-party vote.

So great was the gap between the theory and the reality that, this time around, the Electoral Districts Boundaries Commission jumped straight through the “as far as practicable” escape hatch and did not even attempt to give effect to the directive of Section 83, which under the circumstances would have required some highly creative map-making. It has instead contented itself with clipping Labor’s wings somewhat in its four most marginal seats, only one of which has moved all the way to the notional Liberal column.

Furthermore, as a new set of seat margins calculated by Antony Green indicates, it’s debatable whether the EDBC’s methods would have delivered on their theoretical premise even if Section 83 had been scrupulously observed.

The heart of the problem is that creating hypothetical margins for new boundaries from previous election results requires knowledge of how people voted in areas transferred from one electorate to another. The only available tool for doing so is results from polling booths located in or near the relevant areas, which is somewhat imprecise even in the best of circumstances and is being rendered even less satisfactory by the growing popularity of postal and pre-poll voting.

Whereas ordinary polling booth voting accounted for 88.5% of the total at the 1993 election, by 2010 it had fallen to 78.6%. That leaves something approaching a quarter of the vote that can’t be identified with a particular location in the electorate, an issue the South Australian Electoral Districts Boundaries Commission (EDBC) deals with in the crudest possible way.

Under the EDBC method, if 10% of an electorate’s voters are transferred to another seat, 10% of its non-polling booth total gets transferred as well, with no mind paid to the fact that this total says nothing about the peculiarities of the affected area.

By contrast, Green’s method makes appropriate adjustments to non-ordinary vote transfers. This causes him to calculate a Labor margin of 2.8% in the electorate of Light on Adelaide’s northern fringe, which has lost some territory in overwhelmingly Labor-voting outer suburbs. By the reckoning of the EDBC, the Labor margin is a more comfortable 4.2%.

Green’s margins suggest the uniform swing required to cost Labor its majority is 0.6%, whereas the EDBC has it at 1.5%. In other words, a system that supposedly enshrines electoral fairness has the Liberals shooting for either 52.2% or 53.1% merely to reduce the other side to a minority.

Assuming the three sitting independents are re-elected, the target facing the Liberals for majority government is a daunting 54.2% on Green’s numbers, or 54.3% on the EDBC’s. But that, of course, assumes an alternative reality in which swings are perfectly uniform. As history shows, the reality for the Liberals could be quite a lot better — or worse.