Earlier in the week when the Government announced that it was to ban the sale of alcohol at below cost price one particular claim caught the collective attention of Full Fact.

Not only in the video above but in other media appearances, Home Office Minister James Brokenshire estimated that the policy “will reduce crime by 7,000 incidents a year 2,000 of which are violent crimes”

Curious as to how the Home Office could pinpoint the numbers in such a way, and concerned that there was apparently little explanation of where the figures came from we decided to investigate.

Unsurprisingly, given the variation in drinks, drink prices and drinkers that any forecast would involve, the answer was not straightforward – and the first document we were referred to only served to raise more questions.

An exploration of studies on the social effects of changes in alcohol prices and crime suggested an association between the two, but made clear “this relationship is not straightforward and linear and the evidence base is not able to support a causal relationship between alcohol pricing and crime.”

So if the evidence does not support a causal relationship, how can the evidence support specific forecasts resulting from the changes? We went back to the Home Office again but were told that the figure was based on modelling from academics at Sheffield University.

Some findings were published in this document (see the tables from P117 onwards), but they do not consider the specific policy now being implemented by the Government, so therefore get us no closer to the method behind Mr Brokenshire's figures.

At the third time of asking we were told that the figures were produced by assessing what policy would mean in terms of a minimum price per unit for beer, wine and spirits. This was then compared to sales data to work out how much of the drinks market would be affected by the change and thus suggest changes in consumption.

The effect on crime was then estimated using the sort of modelling mentioned above – to produce the statistic used by the minister.

The only problem is that the Home Office modelling is not publicly available meaning there is no option but to take their word for it.

However given that 7,000 incidents represents just 0.17 per cent of the over 4 million incidents reported to police per year, no one could accuse the Home Office of claiming a huge impact for the policy – even if the maths behind the figures remains murky.