Wednesday, 4 March 2015

Greece's shrinking working-age population: response to a critic (UPDATE)


In discussing my recent post on Greece's alleged humanitarian crisis, I was challenged by one reader for saying that Greece's poverty stats could not be much worse in 2014 than they had been in 2013. I based this on both GDP rising and unemployment falling in 2014, and my critic took issue with the latter point on a factual basis.

His argument was that Greece's 2014 fall in unemployment was due to the working age population shrinking, and that gains in employment were in fact cancelled out by a fall in the working age population - this, he explained, was due to people leaving the country in search of work.

The argument has been played out over Twitter; I don't intend to reproduce it here just to score points. The end result is this: As of Nov 2014, the Greek Labour Force Survey (source of almost all labour market statistics) told us that unemployment was down 107k yoy. Employment was up 53K yoy. Economic inactivity was down 3K yoy. The size of the working-age population implied by LFS (I assume this is ages 15-64) was down 57K yoy. The ratio of employed to total working age population was up by about 1pc point.

Where are those extra 57k working-age folks? The quick answer is that this is a meaningless number - LFS is conceptually one of the worst means possible of measuring the size of the working age population; particularly if one is trying to test a hypothesis related to migration. The reason for this is simple: LFS uses the latest Greek Census data as its sampling frame, and uses age groups for the purposes of post-stratification. That is to say, the LFS starts by assuming that the size and age structure of the overall population at even the local level is more or less the same as it was during the last Census (May 2011). Until as late as mid 2013, ELSTAT was using the 2001 Census as its sampling frame, but estimates were rebased in early 2014 to reflect the 2011 Census and every number you can find on ELSTAT's database now reflects this (though past press releases may not). There isn't another Census planned until 2021, so we're stuck with this sampling frame for quite a while. The problem is that the further away we go from 2011, whether in terms of time and in terms of economic circumstance, the less adequate the LFS figures will be. The rebased working age population figures for Q2 2011 were only 1% off Census - a mere statistical error. But how far off the 2014 estimates will be, is simply unknowable.

Since we can't have a Census each year, the next best thing to help us understand what has happened to the Greek working age-population is administrative data. You can find admin-data-based estimates of the Greek working age population here, and detailed regional estimates here. In the pre-Census years, the estimates were hopelessly off, but from 2013 onwards they were brought back in line with Census data. So while it's not clear what happened in between, and whether the 2009 figure is in any way accurate, admin data suggest that the working age population in Greece shrank by 6%, or 482k, between 2009 and 2013, 31% of which was due to workers who were not Greek nationals leaving the country.

On the same basis, Eurostat has put together some slightly iffy immigration and emigration estimates for 2011 and 2012. The reason I call them iffy is because they are derived from comparing admin data with the natural changes expected due to demographics, and they mostly refer to pre-2013 data which are nowhere near as accurate. Anyway, these data tend to show that pre-crisis Greece was routinely sending ca. 100K working-age people abroad, and bringing slightly more than that in from abroad. Between 2010 and 2012, however, immigration fell from 103K to 84K, while emigration rose from 99K to 135K - a net loss of 60K work-age individuals in the three years on which we have data.

Importantly, the 2011 Census figures for our working age population came in much lower than previously estimated or forecast - at a mere 7.1m. Where did those people go? Were they Memorandum refugees? To test this, we need a good counterfactual. Amazingly, we have three, though quality tends to vary.

First, the UN produces national population forecasts on a regular basis, based on projected patterns in fertility, longevity and migration. The detailed 2012 forecasts can be easily accessed here. It's no longer very easy to get hold of the 2010 vintage of the UN projections, but you can see the central forecasts here. Crucially, both sets of forecasts use projected 2010 figures as their baseline, and neither of them incorporate economic factors in their analysis. This is because the UN forecasts are focused on building credible scenarios of population growth for the very long-term, not the next economic cycle. While I still think this is a weakness generally, it is a blessing in this particular case, as the forecasts are unaffected by austerity and provide a good counterfactual. Extrapolating from the 2012 estimates, it seems to me that Greece should have lost ca. 157k working-age people between 2009 and 2013, assuming zero net migration.

* note: the economist Edward Hugh, whom I follow with interest on Twitter, has had a brush or two with the UN figures and is very skeptical. I trust his judgement and hope to quote him on more detail soon.

Second, Eurostat produces its own estimates, called EUROPOP, every three years. The EUROPOP 2010 figures are now hard to find (try here) but the 2013 estimates are easily accessible here. A little extrapolation using the latest of the two datasets shows that Greece was meant lose about 254K working-age employees anyway between 2009 and 2013, regardless of policy, assuming no net migration. Even the more accurate UN estimate, however, was still more than 200K people off the actual Census figures in 2011. Both of these sets of estimates are constrained by their respective methodologies and the data they relied on. The Eurostat figures over-rely on admin data, which in Greece's case led to an over-estimation of the population. The UN data over-relied on Census data, which, by the time the crisis hit Greece, were hopelessly out of date.

The table below summarises all of the population estimates and counterfactuals for Greece, so you can check for yourselves.

Finally, we have a counterfactual on emigration that can shed light on working-age Greeks' intentions pre-crisis. In December 2009, there was a Eurobarometer study of EU citizens' intentions to work abroad. You can see the full results here, but pay special attention to pp. 178-188. As of late 2009, 8% of working-age Greeks had hoped to move abroad for work purposes. Not all of these will have moved of course, but we have two ways of pinpointing the ones who did. First, 27% of that 8% said they were planning to move in the next 2 years or less - taking us to just after the 2011 Census. Another way of looking at this is that 44% of those who said they would consider moving had taken some concrete steps to facilitate the move. Without access to the raw data, my estimate would be that between 2% and 3.5% of Greece's working age population were planning to move pretty seriously as of late 2009 - pre-austerity. Assuming they left the country evenly between 2010 and 2012, that would account for 49K-86K per year - enough to explain between 43% and 75% of the admin-estimated emigration in the coming years. But clearly the real emigration figures must have been higher, swelled by unconventional net emigrants, such as students who had meant to return after their studies and found they could not.

And what of the post-Census period? LFS (which we must treat with care as explained) suggests an average loss of 61K working age people per year between 2011 and 2014. The UN figures suggest a loss of 39K per year would have been natural over this period. The EU estimates suggest a loss of 81K per year would have been.

UPDATE 7/5/15: The 2013 im/emigration data are finally in, but there's a catch: a break in the series as the admin data were rebased in 2013 in order to bring them in line with the Census. Taking the data at face value, immigration fell off a cliff, with only 38k working-age people coming into Greece, down from 84K. But emigration also fell, from 135K to 104K, lower even than in 2011. Overall, in the years from 2010 up to and including 2013, Greece lost 126K working age individuals (31.5K per year) to net emigration. Working backwards from the net emigration figures, the administrative population figures and an average of the UN and EU demographic estimates, I think this rebasing of the admin data removed some 118K working-age people from the final figures. These are clearly people who net emigrated over the previous years, but it's impossible to attribute them to individual years properly.

If I'm right, then the actual figures would look as follows:

IMHO all of this means the following:

1. The Greek working-age population has been shrinking for the last four years or so, though LFS figures alone are a very poor means of testing this. We do not know how many people were lost but something in the region of 350K is reasonable.
2. Absent a strong immigration trend, the Greek working-age population was going to shrink anyway  post-2010, regardless of policy, as a result of demographics. In fact even if we assume that all of the people cleared out of the admin data in 2013 were crisis-era emigrants, half of the change in the working age population over this period would still be down to demographics.
3. The bulk of post-crisis emigration by Greeks probably occured pre-2011 as a result of emigration plans that pre-dated austerity. Some 150K to 250K people left the country in this way between 2009 and 2011, and they were arguably the ones with the most marketable skills.
4. On top of these, Greece was also losing at least 76K working-age individuals per year to net emigration as of 2013, compared to the counterfactual of 10K net immigration. This 'extra' loss continues to widen, driven by falling immigration, as opposed to rising emigration. This could be due to Greeks stuck abroad after their studies, but it should be noted that foreign nationals continue to account for some 40% of the loss of working-age people year-on-year.
5. In previous editions of this blog I have said that the emigration trend has been tapering off, as Greece slowly runs out of exportable people. On reflection, I think I was right. Emigration is tapering off. But without at least another year's worth of data it will be hard to say for certain.

UPDATE 27/8/15:

UK National Insurance registration statistics are a good place to start looking for corroborating evidence (H/T to Mike Bird @birdyword). The UK is a bellwether for Greek emigration appetites because it puts up the fewest barriers to Greeks wanting to work; it's accessible, travel and work here do not require a visa, and the language is not a problem. New NI registrations by Greek adults rose from ca. 3,000 per annum in the good days to just short of 10,000 in 2013 and 2014 - and although a tentative recovery in early 2014 showed signs of reversing the trend, early 2015 data suggest we're in for a record-breaking year in 2015.  A quick search through the data suggests that UK university towns are a major draw - so my tentative guess is that graduates settling in their host country may be driving emigration figures.