Bez wrote: ... There are other objective measures: rate of reporting of close pass incidents (which, if conditions improve, should fall per mile cycled), and travel survey results for cycling journeys (which, if conditions improve, should rise, though clearly other factors will affect this measure).
I believe WMP claim to have already recorded a large drop in the former; I also believe Birmingham are monitoring the latter. We'll see some post-intervention casualty data in September and more next year, of course.
If the police are perceived to be responding positively to reports of close passes, it's almost inevitable that the reports of close passes will increase, even if the real trend is for close passes to fall. And so on. Also, I'd be surprised if anybody was doing much in the way of a truly robust statistical analysis. eg Are there reliable data for the numbers of reports where the complainant was previously fobbed off? Ultimately, what you are trying to measure is whether riders feel more comfortable on the road.
I'm not a statistician but over the years I've understood enough to see where a lot of the stats are unreliable.
I also think that it's a matter of luck which significant events occur. Eg we've had somebody choose to contest a case - not actually part of the project with the police officer riding the bike - who has nevertheless been convicted and that has happened on a day when no men bit a dog so it made it onto the BBC National News www. Valuable publicity which I fancy will have done more than anything else could to influence attitudes on both sides. Had it gone the other way, things might well be worse. Had the driver concerned accepted they were wrong, it would have been one more case in the numbers but having no wider effect.
There are lots of people with degrees in media studies who could explain this better than I can.