Slí Eile: In a recent paper to the Statistical and Social Inquiry Society of Ireland, Statisticians at the CSO have presented further analysis of differences in earnings for employees in the public and private sectors (note that nobody is attempting to compare all types of income including shares, rent etc for all types of households). The paper is entitled "Investigating the Public-Private Wage Gap in Ireland using data from the National Employment Survey 2007" and may be downloaded here
The CSO paper may be viewed as a dampener on the claims by various commentators, including some economists at the ESRI, that the gap in pay 'comparing like with like' is over 20% using 2006 data. The data sources are the same (in CSO and ESRI papers) but the modelling and the exclusion of various sub-groups makes a big difference to the results - really big as the latest CSO paper shows. Readers will recall that the ESRI confidently announced a premium as if there were one way only of estimating and modelling pay in each sector. The ESRI paper entitled "Benchmarking, Social Partnership and Higher Remuneration: Wage Settling Institutions and the Public-Private Sector Wage Gap in Ireland" (Kelly et al. 2009) can be downloaded here.
They confidently concluded that:
The results indicate that the public sector pay premium increased dramatically from 9.7 to 21.6 per cent between 2003 and 2006. Furthermore, we found that by 2006 senior public service workers earned almost 8 per cent more than their private sector counterparts, while those in lower-level grades earned between 22 and 31 per cent more.
The CSO paper does not overturn the key ESRI claims. It confirms that:
* There is, indeed, a positive premium to public sector workers in all these various studies and regression specifications; and
* The premium in favour of workers in the public sector is higher at the lower end of the pay scale
Where the CSO paper differs is that:
* Different model specifications give different (and significantly different) results
* If enterprise size is included along with a focus on 25-59 yr olds permanent employee and particular occupational groups excluded you can get a very significantly reduced premium.
CSO is not convinced about the propensity score matching methodology beloved of the ESRI researchers. They state that:
'Due to the lack of systematic guidelines on the selection of a comparison group, this study does not use propensity score matching to estimate the public-private pay gap. We have found that the estimated premium using propensity score matching is highly sensitive to the criteria used to discard sub-samples without a common support.'
The bottom line in the CSO paper is as follows: 'we advise caution in attempting to estimate one definitive “answer” for the average premium'
Perhaps the most telling finding of the entire paper was the following (I have added bold):
Furthermore, in the Irish context, there are a number of occupations within the public sector that really have no comparable occupation in the private sector, and vice versa (most noticeably Gardaí, Prison Officers, and members of the Defence Forces). To highlight the consequences of ignoring this lack of comparability in some occupation sectors, this study also estimated the public-private wage gap on a sub-sample excluding personal and protective services employees.
The impact of excluding this sector was to dramatically reduce the average public-private wage gap, especially for males (from 7.2% to 2.7% using OLS). The size of this reduction was even larger when the analysis was conducted using weighted data.
Table 3 is the key table. It shows a premium of 2.7% for males when personal and protective service employees are excluded. This is hardly anywhere near John O'Hagan's presumption of a 20% cut (surely linked to the ESRI factoid). For women the premium was just over 11%. In other words to compare employees in public and private sectors a number of significant complicating factors cut across a crude comparison:
* Enterprise size (disputed by ESRI)
* Gender differences within each sector (with greater inequality in the private sector)
* Particular differences applying to groups such as Garda, prison officers.
Aside from all this various types of income are not factored in. Essentially, rental income, income on shares, dividends and other property are not included as are 'irregular bonses' - a major difference in some private sector occupations and sectors. Also worthy of note is the fact that 'irregular bonsuses' or 'benefits in kind' are not included.
In the CSO published supplementary analysis of data from the 2007 National Employment Survey (download here) P21
Earnings are 'defined as gross earnings (before the deduction of tax, PRSI, superannuation) payable by organisations to its employees. It includes normal wages, salaries and overtime, taxable allowances, regular bonuses and commissions, holiday and sick pay. It does not include irregular bonuses and commissions, employer’s PRSI, redundancy payments and back pay.'
CSO state that 'A regular bonus is defined as a bonus received every pay period although the amount may vary from period to period.'
A useful addition to the analysis is the work by John Geary of UCD and Anthony Murphy of University of Oxford. In a comment in a long thread on their research on irisheconomy.ie here
we get a glimpse of an alternative view. The main article appeared in the current issue of Industrial Relations News. (Cutting public sector pay: an alternative view - …THE JOB DESCRIPTORS USED BY CSO ARE NOT RICH AND REFINED ENOUGH…). They state that:
The risk of imposing wage cuts across the board in the public sector is that it will result in industrial strife with huge costs both in terms of the country’s economic fortunes and social cohesion. There is no single, best measure of the public sector wage premium. The estimated premium varies over time and across occupations. It also varies across the income distribution and at the upper end of the distribution, is often a discount. It is also difficult to find good “like-for-like” comparison groups for some public sector occupations. Thus, the basis on which public sector pay might be cut is not as sound as some people claim on the basis of headline figures.
It can be concluded that any generalised claims based on crude comparisons and 'one methodology-fits-all' are suspect especially when you know that a particular agenda is at stake.