It was an early start at Roy Morgan Research in Melbourne this morning with the release of its latest State of the Nation report, presented by the firm’s CEO, Michele Levine, assisted by the old stager Gary Morgan. You can browse the highlights, or readers with a spare $3500 can buy the whole lot.

I found most of the conclusions fairly uncontroversial. Country people are a bit more conservative in their attitudes than city people, don’t use new technology as much and have more health problems — all pretty much as expected. Interesting to see, though, that 74% (78% in the country) disapprove of foreign investors buying Australian farmland, and even more interesting to see that disapproval is slightly higher among Coalition voters than Labor voters. (Which is supposed to be the free market party again?)

But the discussion got interesting when people started talking about unemployment, one of Morgan’s favourite hobby horses. For many years Morgan has been producing its own estimates of the unemployment rate, and they’re invariably higher than those released by the Australian Bureau of Statistics. You can see the full-time series here, going back to 1992.

The two, of course, are not measuring the same thing. The ABS uses a standard international definition according to which you have to have done no paid work in the last week and be actively seeking and available for work. Morgan instead counts all those who “are looking for work”, which it argues — with considerable justice — is a better fit for most people’s intuitive idea of unemployment.

How much the difference matters is another question. Gary Morgan’s case is that understating the seriousness of problem is callous towards the unemployed and leads to policy mistakes. Specifically, he argues for further interest rate cuts and further deregulation of the labour market.

But for policy responses, what matters most is the trend, and if the trend in two sets of figures is the same it doesn’t much matter which one you use. Your static picture might be wrong, but you’ll still get the dynamics right (particularly if the rest of the world is relying on the same measure).

And here’s where it gets interesting. Most of the time, Morgan’s figures match the ABS trend pretty well. They’re consistently higher, to the tune of maybe about a third, and they’re more prone to sudden movement, tending to exaggerate the trend. But over the years they’ve generally moved in the same direction and the gap between them hasn’t been subject to huge variation.

But in the last year that’s changed. The ABS unemployment figure (whether raw or seasonally adjusted) has been basically flat since the middle of 2010. Morgan’s figures, however, are not just higher but have continued rising. The gap between them for the March quarter, 9.7% versus 5.3%, is completely unprecedented. (The previous highest I can find is in March 2001, 10.5% to 6.8%.)

This could be just bad survey technique. Morgan’s polling certainly has some blemishes on its record, most famously perhaps its prediction of a decisive ALP victory just before the 2001 federal election. But its mistakes are usually due to short term problems: if you look at a Morgan poll series it’s extremely jumpy, so any one result is as likely as not to be inaccurate.

Once you smooth out the noise, however, the trends nonetheless match the results from other pollsters quite well.

The unemployment question looks like something different. It looks as if something has been going on that for some reason the ABS hasn’t picked up, which would be a worrying sign. Morgan’s most recent monthly results, however, do show downward movement, so it’s possible that the statistical blip, whatever it was, is now over.

There’ll still be an interesting argument to have over what really counts as unemployment, but if on either measure it’s coming down then at least that’s one less thing to worry about.