How much do the unemployment figures tell us? The Bank of England thinks the answer is: quite a lot - hence its use of the unemployment rate as an intermediate threshold. But I'm not so sure.

By this I don't mean just that there's huge amounts of hidden unemployment. Today's figures show that there are 2.3m "economically inactive" people who'd like a job, and 1.46m part-timers who'd like full-time work. And this is not to mention full-timers who'd like more hours.

Nor do I just mean that there's sampling error in the unemployment numbers. The ONS puts this at 87,000. This means we should understand today's news that there are 2.47m unemployed as meaning there's a 95% chance that the number of unemployed - on the LFS definition - is between 2.38m and 2.55m.

Instead, I'm thinking of the data on labour market flows. The ONS puts people into three categories: employed; unemployed (people without work who have been actively seeking work within the last four weeks and are available to start work within the next two weeks); and inactive. There are six possible transitions between these three states: employment to unemployment, employment to inactive, and so on. My chart plots four of these flows, since current data began in 2001. These are taken from table X02 here.

This picture is messy. Which is big fact in itself. Labour market flows are high and volatile. On average in this time, 913,000 people have moved out of employment each quarter, and 970,000 have moved in. I'd highlight three points here:

1. Recently, flows from unemployment to employment have been very high. Before the crisis, most people entering employment did so from inactivity, not unemployment. Recently, though, the opposite has been the case. This has tended to depress measured unemployment lately.

2.Since the mid-00s, employment to inactive flows have fallen relative to flows from employment to unemployment. This has tended to raise unemployment, relative to the pattern we'd expect to see on the basis of pre-crisis flows. Before the crisis, most people leaving employment became inactive - say because they retired or became home-makers. This is still the case, but to a lesser extent.

3.Recently, flows from inactivity to unemployment have exceeded flows from unemployment to inactivity by an unusually wide margin. The net inflow to unemployment on this basis was 133,000 in Q3, compared to an average of 86,000 between 2001 and 2008.This makes the unemployment numbers look worse.

I don't say any of this to draw a big conclusion about whether unemployment will or won't fall quickly. My point is merely that unemployment can and does change for reasons other than changing employment levels. This is especially true when the economy is growing normally. Between 2001 and 2007, the correlation between changes in unemployment and changes in employment was only minus 0.43. For this reason, the headline unemployment numbers aren't always as informative as you might think.