don't know, is whether those budgets would have been smaller if they'd just used public money in the first place.

Part of the problem is that no one's come up with a working mechanism to test which route would be best for any project. There was one, once - the zippily named 'public sector comparator'. But officials rapidly worked out that there wasn't any public funding, and if they didn't fiddle the figures in favour of PFI, they didn't get their project. This rendered the entire exercise rather pointless, so it was abandoned. Since 2004, the only thing that decides whether to use PFI or not is Treasury fiat.

More worrying is that the government auditors have never tested whether PFI offers better value than public money. Since 2003 the Treasury has said it has research which makes the case, but it's never been made public. Only one report has ever even attempted a direct comparison - a 1999 report by consultancy Mott McDonald, that considered so few projects as to be almost meaningless.

The result, in the word of one

insider-turned-academic, is "a multi-billion pound policy built on three reports with about 20 bits of information in them".

Given the widespread view that PFI exists purely to funnel tax money to fat-cat executives, you have to wonder why the government hasn't done more to make its case. After all, one decent, independent report could dispel the doubts about the scheme forever. So where is it?

It's hard to escape the feeling that the Treasury is petrified to do the research in case it finds out it's wrong and has spent the last 15 years enthusiastically wasting public cash. And it gets away with it. All it has to do is endure a few bad headlines and angry quotes from the Public Accounts Committee once every few months, and it all goes away again.

The result of all this is we might be spending the next 30 years paying through the nose for buildings we should have got at half the price. Or we might have been saved from massive public sector inefficiency. I honestly don't know. Neither does anyone else.