The comment is a response to Michael Taft’s PE post on the comparisons between the public and private sector in which he suggests that comparisons are possible, based on the size of enterprise, and when they are, it shows the public sector pay is not above private sector pay in the main. As JG argues, the debate itself is getting stymied by this bald opposition of the two sectors.

A very useful alternative starting point, Michael.
I agree that statistical analyses have been deployed in very political ways in the current debate, but I think it’s also important not to give the impression that there are ‘lies, damned lies and statistics’! There are very good scientific reasons to expect that attributes of sectors, firms and organizations (like size) predict differences in wages, as well as attributes of individuals (like years of experience). The problem with the debate so far is that it has pretended that there are only two sectors in the Irish labour force - public and private. It may well still be true that some clusters of occupations within the public sector are more highly remunerated than comparable clusters within the private sector - but it seems highly unlikely that this is true of all (or perhaps even most) public sector jobs.

Not that we can even apply anything more than very basic scrutiny, as the benchmarking reports hide behind a veil of secrecy on the spurious grounds that actually publishing their data would lead to lengthy and futile disputes.

But lets make a leap of faith and assume for a second that the conclusions of the first (2002) report were kosher, and the average 9% increase was really needed to allow the public sector catch up with their private sector counterparts. So once this increase was applied, the playing field is fully leveled, right?

Then along comes the second report 5 years later, applies a discount of 12% in respect of the PS pension value, before concluding that most grades should get little or no increase.

In order for their own conclusions to make any sense at all, the public sector would have had to have fallen further behind the private sector by at least 12% in the period from 2002 to 2007, as the pension value was not factored into the original comparison. Yet all the data suggests otherwise, i.e. that the gap widened significantly in the opposite direction during this period.

Did the famed structural differences between the sectors only emerge in this period? Were there a program of mass-lobotomization across the private sector, with concurrent accelerated evolution in the public sector over those years? Was there an outbreak of time dilation so that public servants acquired experience at a much faster rate than their private sector counterparts? I think not.

I’m afraid benchmarking is damned by the internal contradictions in its own conclusions. The 2002 and 2007 reports cannot *both* be accurate, and the suspicion is that neither was based on reality (though the lack of transparency renders the latter determination impossible).

And the methodologies undertaken by the Benchmarking board are just as sophisticated, and just as mainstream - as the ESRI working paper acknowledges - as the methodologies undertaken by the authors of the ESRI working paper. The authors of the ESRI working paper do not agree with the benchmarking pay awards, but what they do agree with is the fact that there are structural differences between the public and private sector. Indeed, the very reason why the authors of the ESRI working paper engage with a multivariate estimation strategy in assessing the levels of the benchmarking pay awards is BECAUSE there are structural differences between the public and private sector, that it is not “like for like.”

But the benchmarking board didn’t pull their methodologies out of their ass, no more than the authors of the ESRI working paper pulled their methodologies out of their ass. In both cases, we’re talking mainstream.

RTE and Morning Ireland, however, saw it differently, and spent almost the entire interview with Tom Geraghty hammering home their point that the ESRI and the CSO were comparing like with like. and that’s not the case. Neither the authors of the ESRI report, nor the working paper, nor the authors of the NES 2007, made such a point - despite the strong insistence of Morning Ireland that they did - nor would any professional make such a point, for the silly reason that the facts say otherwise.

http://www.irisheconomy.ie/index.php/2009/07/14/public-sector-pay-differentials-regressions-can-actually-be-useful/

All good and valid points on the ESRI’s methodology versus raw or bilateral comparisons.

The point of my original post of course was Tom Geraghty fails to see any difference between the raw unadjusted figures released by the CSO and the much more sophisticated approach undertaken by the ESRI, who controlled for differences in human capital etc.

Karl (and Conor too, in a round-about way) makes it clear that the latter approach is much more trustworthy than the simple long division of the former. Of course no statistical method is perfect, its always a case of reducing these data to a much smaller set of those numbers, a transformation that cannot ever be loss-less.

But its not really the ESRI methodology that Tom Geraghty is attacking, as he’s clearly unfamiliar with that. Rather its the conclusion he doesn’t like, and hence the attempt to conflate the ESRI’s conclusions with the statistical dubiousness of the CSO’s averages.

Ironically, Tom would point to the Benchmarking exercises as the model of statistical probity. Whereas in fact one wouldn’t know were to even start criticizing the benchmarking methodology, its so full of holes. Well I guess the fact that their conclusion was predetermined would be as good as place as any.

As you seem to think that the ESRI working paper is the last word on this, well, Karl Whelan offers a caveat: “Of course, these regressions are hardly perfect either since they likely hide substantial unobserved heterogeneity but they certainly beat a series of pair-wise comparisons.”

So. The ESRI recognises the structural differences, while the methodology utilized by the ESRI is “hardly perfect” (Professor Whelan makes it clear in his post that he sees the ESRI working paper an improvement on the current public debate of comparing the public and private sector on a straight basis.)

The authors urge caution on the figures, and they also urge caution on direct comparisons between the public and private sector, because there are structural difference in employment between the public and private sector. These include differences in levels of education, occupation structure and length of work experience.

Which is what I said in my post on wages and employment structure in the public sector, NES, 2007 - points raised by both the authors of the working paper and by Professor Karl Whelan of UCD - and points that you spent the last two days criticising.

The figures on which the authors counsel caution are the raw, unadjusted mean private versus public sector earnings. Even if you’d continued your copy’n'paste just to the end of that individual paragraph, quite a different meaning would emerge. Its what’s know in the business as a justification of one’s methodology.

Usually it goes thusly: “Just considering A wouldn’t give the full picture. So clever clogs that I am, I also considered B.” You quoted the initial bait, but omitted the subsequent switch.

So I’ll save your readers the trouble of firing up acrobat themselves, and continue just that paragraph:

“By adopting a multivariate estimation strategy, we can assess the extent to which higher earnings in the public sector go beyond the level that can be attributed to characteristic effects i.e. the framework allows us to accurately estimate the extent of any public sector wage premium, controlling for differences in educational levels, work experience and so on.”

Puts a whole other shine on it, no?

Cheers, Prop.

Well Tom is straying from the shining path of truth and righteousness right there.

In fact, the ESRI study did a lot more than a bit of auld long division. Their analysis controls for a variety of wage determining characteristics including differences in educational levels and work experience, yet still finds a wage premium of 23.5% existed in 2006.

Though apparently they did not control for the incidence of Fáinne-wearing, nor for the frequency of the name Gobnait. Maybe that’s at the root of the dubious statistical value to which our hero refers.

In any case, the ERSI report is worth a looksee: http://www.tara.tcd.ie/bitstream/2262/27906/1/WP270.pdf