Here's the first of them.
Okay, when I saw this graph at first, I didn't know what was going on, either.
Let's start with those red ("Unemployment at 5%") and green ("Unemployment at 10%") lines that look like isoquants (and are). Ignore the country data points just for a moment.
Each of these isoquants traces out those combination of inflows into unemployment (as a percentage of the labour force per month, on the Y-axis) and the duration of unemployment (in months, on the X-axis), that keep the unemployment rate steady.
The red line shows, for example, up the left hand side of the curve, that a 5% unemployment rate could be maintained indefinitely if 3% or so of the labour force lost their jobs in any given month, but only spent about a month or a month and a bit in unemployment. Or, out the right hand side, that there would be a steady 5% unemployment rate when only about 0.5% of the labour force lost their jobs in any given month, but stayed unemployed for about a year.
And the green line shows the same thing for a 10% rate of unemployment, obviously lying well outside the 5% one. You can imagine a whole series of isoquants for different rates of unemployment, but only two are shown here.
Okay - now we know what the isoquants are, we can start thinking about the data points, which represent where countries have actually been over the period 1995-2007.
Note first that you ideally want to be on an isoquant closer to the origin rather than further away. On that basis we scrub up reasonably well. In this data set (which is the G20 plus a few odds and sods like us), there aren't many countries on isoquants clearly closer to the origin than us (other than Norway), quite a few similar, and a fair number markedly worse.
What about the significance of where you are along any given isoquant?
Here it's probably best to look at the labour market theory behind all of this. You'll see that 'Perez and Yao' are credited in the graph, and that's a reference to Can Institutional Reform Reduce Job Destruction and Unemployment Duration? Yes It Can. Perez and Yao call each isoquant an 'IsoUnemployment Curve', or IUC.
The isoquants can be used (they say, p21) "to classify countries according to their preferences over the job destruction-unemployment duration trade-off", i.e. where they are along any one isoquant shows "each economy’s revealed social preferences over the destruction-duration mix". And different countries clearly prefer (or at a minimum end up with) particular combinations: "Some countries seem to tolerate relatively high destruction rates as long as unemployment duration is short [that includes us, and Australia]. Others are biased towards job security and do not mind ﬁnancing longer job search spells. A few unfortunate countries [Spain, for example] are trapped in a high inﬂow-high duration combination, seemingly condemned for long periods of high unemployment".
Which combination a country chooses is not necessarily linked to its unemployment rate. "The upshot of this analysis is that labor markets characterized by high levels of job destruction but brief unemployment spells [like us] do not necessarily outperform countries characterized by the opposite behavior". Canada, for example, is on a worse isoquant than Japan.
Personally, I quite like where we have been revealed to be - some unemployment, more widely but more briefly spread, seems better to me than some unemployment, heavily borne by the long-term unemployed.
I said there were two interesting graphs: here's the other one.
In the original Vox paper, the IMF economists were saying that the IMF probably ought to "tread carefully" if tempted to advise countries on how to run their collective bargaining arrangements, because "trusting partners can make widely differing combinations of institutions work well", and they use the graph to show that it is the level of trust that can be as important as anything else.
That makes broad sense, even if you're immediately tempted to observe that trust isn't an exogenous manna from heaven and to wonder about chains of causation (does trust make structures and processes work better? Yes. But do good structures and processes engender more trust? That too).
In any event, we again show up down the desirable end of the spectrum.