When measuring an abstract concept, occasional oddities are probably unavoidable. Take inequality, which was a hotly debated topic during the last election campaign. The opposition claimed inequality had increased under Labour but Labour denied it. Both were, in a sense, right. The very top and the very bottom of the income distribution had pulled further apart, but the rest of the distribution had become more compressed. Inequality had decreased, but it had also increased.

However, when a measure frequently produces results that go flatly against common sense, then it fails its task. In poverty measurement, this is precisely what happens. We don’t have final data on the recession’s effect on poverty yet, because these only go until 2009 so far. But judging from the latest forecasts by the Institute for Fiscal Studies, it seems that the recession has decreased poverty. Between 2008 and 2011, the rate of relative child poverty has fallen by 2.6 percentage points and the rate of relative working age poverty by 0.4 percentage points. This is a recurring phenomenon: the recessions of the early 1970s, the early 1980s and the early 1990s have also been effective anti-poverty programmes, at least statistically. But an index of poverty which falls during recessions is like an index of binge-drinking which falls on Friday nights.

The phenomenon is explained by the following pattern: during recessions, incomes fall across the board, but they fall at different paces. For households at the bottom of the distribution, the most important income source is government transfers, while for households in the middle of the distribution, it is wages. Transfers are not as exposed to the recession as wages, so the middle often takes a harder hit. When middle incomes fall, they automatically take the poverty line with them, and poverty ‘falls’.

The poor are still poorer in terms of their living standards, of course. As the IFS report also shows, poverty has increased when compared to a fixed threshold. From next year on though, the difference between the two measures will diminish. Median incomes will resume growth at a sluggish pace, while the slower uprating of benefits will dampen real income growth at the bottom. Thus, poverty will rise again in relative terms, while stagnating when measured against a fixed threshold. It’s bad news either way.

The sensitivity analysis in the report is also revealing. If job creation turns out to be stronger than forecast, the increase in relative poverty will be somewhat more pronounced. The reason is the same as above: more jobs mean higher median incomes, and thus a faster rise in the poverty line. Unless employment increased disproportionately among the poorest, more jobs mean more poverty.

Even if earnings growth was most pronounced in the lower half of the pay scale, relative poverty would increase. This is because work levels at the bottom of the distribution are too low for an increase in wages to have a strong effect on incomes. Those in low-paid employment would benefit, but those without or with only sporadic employment would not, and the former are further up the income distribution than the latter. The median income would rise, and so would the poverty line.

Many will lament the government’s failure to reach the headline child poverty target. It will be a barking up the wrong tree. Progress against a nonsensical target is not an achievement. The coalition should now admit that it was a mistake to have signed up to this target in the first place, and abandon it. What we need instead is an anti-poverty strategy based on lowering the cost of living through supply-side reforms, and raising work levels among those with weak labour market attachment.