Central Statistics Office

Earnings Analysis 

Central Statistics Office, Skehard Road, Cork, Ireland

This report looks at the quality of the Labour Cost Survey 2012 (LCS), as defined by Annex 1 of the Commission Regulation (EC) No 698/2006 of 5 May 2006. The LCS results in the Irish case has been compiled from a quarterly survey Earnings, Hours & Employment Costs Survey (EHECS) that collects comprehensive data on earnings, employment, hours worked, job vacancies and job turnover. This information is collected by the Central Statistics Office (CSO) under the S.I. No. 314/2008 Statistics (Labour Costs Survey) Order 2008. The survey results meet the requirements for labour costs statistics set out in Council Regulation (EC) No. 530/1999 but is also used to meet requirements for data on job vacancies, the Labour Costs Index (LCI)  and certain short term statistics (STS) data.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

-

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

Not available.
New concept added with the migration to SIMS 2.0.
Information (content) will be available after the next collection.

The Earnings, Hours and Employment Costs Survey (EHECS), which is the basis of the LCS, is regularly assessed for quality and improvements using a Lean Six Sigma process. 

A quality report is produced each year and is evaluated by Eurostat to assess the quality of the production processes.

The overall quality is deemed to be good, with very limited revisions in the European aggregates.

-

The purpose of the quarterly survey is to produce indices for the purpose of monitoring change in labour costs in Ireland.

The national users of the quarterly data derived from the survey include:

• Other government departments
• The Central Bank of Ireland
• Social partners such as trade unions and employer’s associations
• Research bodies (including universities etc) and private institutions
• Economists in organisations such as Irish banks
• The media

International users comprise organisations such as the Organisation for Economic Co-Operation and Development (OECD), International Labour Organisation (ILO)  and the United Nations (UN).

All the main elements of labour costs are disseminated both on a quarterly and an annual basis.

N/A

Regulation (EC) No 698/2006 of the European Parliament and of the Council concerning the labour cost index is fully implemented.

100%

Sampling methodology

A census of all enterprises with 50 or more employees is undertaken each quarter.  A stratified sample by NACE 2-digit sector and size of enterprise classifications is used for enterprises with less than 50 employees, with the following proportions by size class for all sectors: 

Table 1: Enterprises proportions by size class  (2012)

The overall sample is approximately 6,500 enterprises which represent around 12% of all enterprises and accounts for 66% of all employees across NACE sectors B-S.

Survey technique/data transfer

EHECS consists of two questionnaires (Form A and Form B) and also utilises an electronic facility to enable enterprises to extract the required data directly from their payroll systems.  The Form A questionnaire is issued to all enterprises with registered employment of 100 or more persons.  The Form B questionnaire is a summarized version of the Form A and is issued to enterprises with less than 100 persons employed.  The EHECS also utilises software functionality on an enterprises payroll system which extracts the required EHECS data (as for Form A) from an enterprises payroll system and puts the EHECS data in an XML file. Any enterprise regardless of employment size class can submit an XML file. The only requirement is that the enterprise has the necessary software functionality on their payroll system.

Questionnaire (including explanations)

Form A and Form B questionnaires and explanatory notes can be found at the following link:

http://www.cso.ie/en/surveysandmethodology/earnings/earningshoursandemploymentcostssurvey

XML files are generated via the CSO module included in some payroll systems used by enterprises.  On average 40.5% of returns were XML files in 2012.  Table 2 provides details of employer and employee coverage within this response.  

 Table 2: Form type response with employer and employee coverage

Data editing

Extensive edits rules are applied to check unit level (i.e. enterprise) data each time a dataset is compiled, with additional quarterly checks against previous returns for all enterprises with 250 or more employees.

The main checks performed on the data include:

• Check that all relevant variables are filled in correctly for each category received:

• Employment is present both at the beginning and the end of the quarter
• Both wages and hours variables are present (regular/overtime)
• Both wages and Pay Related Social Insurance (PRSI) variables are present

• Check that the hourly rate of wage (regular/irregular/overtime) falls within a reasonable range for the sector
• Check that the average number of hours worked per week (full time/part time/apprentices) fall within a reasonable range for the sector
• Check that the received figures tally up (e.g. that the PRSI value provided is within correct parameters for the wages and salaries indicated)
• Check that benefit in kind (BIK) and pension figures provided are within a reasonable range
• Ensure that the hired and vacancies rates received are correct

Other quality assurance techniques used

• Metadata such as response rates, coefficients of variation (CVs) estimates, non-respondent calls and key company updates is compiled and checked each quarter
• Coherence checks are undertaken on individual enterprises across quarters.  Matched sample analysis is undertaken to determine the effects of sample changes
• Time-series analysis across NACE sectors is used to study quarter changes within broader longer-term trends in earnings

Sampling errors cannot be directly estimated for total annual labour costs (D) due to the manner of compilation of the estimates whereby aggregate quarterly estimates are summed across the 4 quarters to give an annual total (i.e. annual labour costs are not estimated at the microdata level).

Sampling errors cannot be directly estimated for total annual labour costs (D) due to the manner of compilation of the estimates whereby aggregate quarterly estimates are summed across the 4 quarters to give an annual total (i.e. annual labour costs are not estimated at the microdata level). However sampling errors have been estimated for each quarter and the table below presents the average CVs by NACE and size class across the four quarters of 2012. Given the high percentage of enterprises a relatively low variation was seen in the CVs across the quarters. The higher value for NACE L is attributed to the number of smaller enterprises within this sector and a lower response rate when compared with other sectors. As expected CVs on total labour costs are either equal to or higher than the CV on average hourly labour costs in each case. While these CVs are not directly relevant to the annual totals and overall annual averages they are highly indicative of the relative level of precision.

Other non-sampling effects (than the ones mentioned below)

Following a Lean Six Sigma project response rates for preliminary and final data are set to ensure that the change from preliminary to final data remains broadly in the range of plus or minus 5%.  In reality, the average rate of change generally falls within the order of plus or minus 2%.  Other non-sampling effects such as errors on the CSO’s CBR, questionnaire errors or processing errors are all examined as part of the quality assurance techniques used.

All enterprises were allocated by NACE according to the NACE recorded on the CSO’s central business register (CBR) and this is assumed to be correct.

All known active enterprises with 3 or more employees were included in the sampling frame so no non-sampling errors outside the minor non-coverage within the CSO’s CBR are known to exist for enterprises in that size class. Enterprises with 1 or 2 employees were not sampled and so are not reflected in the results for the 1-9 employee size class.

With specific reference to the quarterly survey which is the base for the LCS the local STS register is updated on an ongoing basis to ensure continuing relevance however there can be a time lag in updated information being reflected in the CSO’s CBR. As such there can be limited under coverage where new births of enterprises are not reflected for some time and also over coverage where closures are not immediately reflected. If such ceased enterprises are included in the quarterly sample and found to be ceased the CSO’s CBR is updated accordingly. The exact level of over and under coverage cannot be estimated.

Cannot be estimated

[Not requested]

Two methods of data collection were used. For approximately 40% of respondents, data was returned electronically using an XML file directly generated from the payroll system. This was possible because of an extensive project whereby the EHECS data requirements were incorporated into all of the main payroll software systems. The system allocated each payment type for an enterprise to each of the different elements of labour costs. As a result following an initial setup phase the enterprise could generate the large majority of the quarterly return including all the LCS elements automatically through a module in the payroll software. A limited amount of information such as job vacancy data had to be manually entered onto the XML file. Once these manual elements were added the XML file was electronically submitted.

 For other respondents (60% approximately) all data was collected via a paper questionnaire (Forms A and B) which was scanned to capture the data.

 The survey form may be accessed on the CSO website at:

http://www.cso.ie/en/surveysandmethodology/earnings/earningshoursandemploymentcostssurvey/

All enterprises with 50 or more employees were included in the survey and a sample of enterprises with 3 to 49 employees.

On average across the year the response level was a slightly higher than 60% of sampled enterprise. The respondents covered almost 80% of the estimated total number of employees in the population. Details of the response for the NACE sectors are given below.

Where enterprises with more than 50 employees did not respond on time the data was imputed, typically with reference to data submitted for the previous quarter for the enterprise extrapolated with reference to trends among respondents within the same NACE sector. In the absence of previous data, imputation was done with respect to respondent enterprises in the same size class and NACE sector. Analysis has shown that the level of consistency between imputed values and true data subsequently submitted is high.

The overall level of either partial or complete imputation in each of the 4 quarters of 2012 was approximately 20% of records – however of those over 80% were ratio imputations meaning that actual returns were received for at least part of the year for the large majority of imputed records. Imputation rates by variable are not available.

The high level of imputation reflects the fact that data for all enterprises with 50 or more employees were imputed if no response was received. No grossing was required for enterprises with 50 or more employees. For enterprises with less than 50 employees, the total number of employees in these enterprises, classified by 2-digit NACE on the CSO’s CBR, was used as the control total and grossing factors were estimated accordingly.

[Not requested]

[Not requested]

For both the scanned and electronically submitted data an extensive range of edit checks were undertaken and respondent enterprises were contacted with follow up queries to identify corrections to the data. As the LCS is sourced from the quarterly survey then a significant aspect of the edit checking refers to consistency with previous periods in terms of average hourly rates (both earnings and other labour costs in particular). There are also absolute level checks. Any questionable values were followed up with respondent enterprises with corrections applied as necessary.

Collection of data to produce B1 ( Hours worked by all employees) and associated sub-components was discontinued with effect from Q2 2010.  Estimated values for this variable were generated by applying the ratio change for usual hours worked from the Labour force survey (LFS) to the 2008 B1 value to derive the rate of change to be applied to the C1 values.   The subcomponents of B11 (hours worked by full-time employees, excluding apprentices), B12 (hours worked by part-time employees, excluding apprentices), and B131 (hours worked by apprentices) were also generated based on ratio change of full-time, part-time and apprentices from the LFS.  The same process was used to estimate values for D1113 (payments for days not worked) variable.

Imputation rate is 25%

The annual data was essentially the sum of the four quarters in 2012 for which quarterly data were collected. The full set of submitted variables were available in each quarter so the first step was the estimation of aggregate totals for each variable for each of the 6 size classes at the 2 digit NACE level on a quarterly basis. (See Tables 3B and 4B for details of size classes). These aggregate totals were summed across quarters to give annual estimates.

[Not requested]

[Not requested]

[Not requested]

[Not requested]

[Not requested]

-

Preliminary results are published by the CSO within T+56 days after the end of the reference quarter.  Further responses continue to be sought and edits are carried out.  The final results are published by the CSO within T+112 days.

T+56 days after the end of the reference quarter.

The final results are published by the CSO within T+112 days.

As referred to in other sections the LCS data were collected via a quarterly survey (EHECS). Survey forms are issued in the last week of the end of the reference quarter and respondents are asked to respond by the 15^th of the next month. So for quarter 4 2012 responses were requested by 15^th January, 2013. If the 15^th of the month in question falls on a weekend the response deadline is set as the following Monday. A posted reminder is sent in the week following the first response deadline and if responses are not received within 2 weeks, responses are sought by telephone by CSO staff.

[Not requested]

-

-

National concepts have been defined as close as possible to European concepts, all classifications and definitions are as stated in the Regulation including reference period, population, classifications and variables covered.  The main differences are that enterprises rather than local units have been surveyed and enterprises with 1 or 2 employees are excluded from the sample.

[Not requested]

The production of the 2012 data followed the same processes used in producing the 2008 data, i.e.  A separate annual LCS has not been undertaken and the results have been derived from the quarterly EHECS. The EHECS was designed to cover the components required for the LCS so at an aggregate results level estimates should be comparable with previous periods.

 Notwithstanding this the main other changes impacting upon comparability over time are:

 Up to 1981, the surveys covered establishments with 10 or more employees. From 1984 on, the surveys have covered enterprises. 

• In 1984 and 1988, the coverage was restricted to enterprises with 20 or more employees; while in 1992, 1996, 2000 and 2004 enterprises with 10 or more employees were covered.
• From 1988 on, only enterprises that traded throughout the year are included.
• Prior to 1992, reports of labour cost surveys classified results by activity according to the General Industrial Classification of Economic Activities within the European Community (1970) (NACE 70). For 1992, results were produced using both NACE 70 and a new classification known as Statistical Classification of Economic Activities in the European Community, Rev. 1(1990).  Results for 1996, 2000 and 2004 are classified in accordance with NACE Rev. 1 only.
• Sectors F (Construction), L (Public administration and defence), M (education), N (Health and social Work) and O (Other community, social and personal services activities) were only covered in 2004.
• Sectors H (Hotels and Restaurants), I (Transport, storage and communication) excluding travel agents and NACE 67 (Activities auxiliary to financial intermediation) were covered from 2000 onwards.
• Sectors G (Wholesale and retail trade; repair of motor vehicles, motorcycles and personal and household goods), K (Real estate, renting and business activities), NACE 65 (Financial intermediation, except insurance and pension funding) and 66 (Insurance and pension funding, except compulsory social security) were covered from 1996 onwards.

[Not requested]

Coherence with statistics from the Labour Force Survey (LFS)                            

The table 6 below shows the average actual hours per week from the LFS compared with the average hours worked from the LCS (calculated as B1 (paid hours or all employees) divided by A1 (total number of employees) divided by 52 weeks).

At the total level the level of difference is low; however there are differences in certain sectors. In explanation of this a number of things can be noted:

• In the Education sector (P) there is a difference in the way hours are recorded in payroll software which yields a consistent low level of hours worked when compared with LFS averages. These low levels primarily reflect the low levels of contractual hours in that sector which may not match with self- reported hours worked.

•  In Accommodation and food sector (I) there is very heavy variability in average hours reported in the LFS across the year. This sector has a high level of second jobs and part time work which arguably could cause LCS estimates to be lower than LFS estimates as second jobs are not reflected in the LFS. A similar explanation can be made for other sectors with larger differences where second jobs would be more prevalent (the Administrative and support service activities sector (N) would be another example). 

Coherence with structural business statistics (SBS)

The table 7 below shows the comparison of SBS and LCS wages and salaries per employee. While there are some larger percentage differences in some sectors, notably D, H, L and N, the different methodologies utilised in deriving final data must be considered.  The SBS average annual estimate is based on greater than 3 employees, while the LCS is based 10 plus employees.   In addition, the SBS, is based on an Annual survey, while the LCS is based on quarterly data which is annualised.

Note: SBS comparisons can only be done for sectors B to N as no SBS data is available for sections O to S.  Data for 2012 was not available from the SBS for sector F at the time of writing.  K is not surveyed by SBS.

Coherence with LCI

The EHECS data is used to calculate the LCI.

[Not requested]

[Not requested]

The EHECS data is utilised in National Accounts calculations. The National Accounts department validate the earnings and labour cost data by scrutinizing the trends in the time series. 

[Not requested]

-

The Earnings and Labour Costs release is published electronically each quarter by the CSO. There is no press conference relating to the release.

The Earnings and Labour Costs release is published quarterly and outline trends in earnings, hours, employment and labour costs indicators for Ireland. 

This data is utilised in producing the Labour Cost Index, which is published by Eurostat on a quarterly basis.

Annualalised earnings and labour cost data is published each year nationally and provides estimates of total earnings and labour costs, as well as annual average earnings and labour costs.

A separate annual labour costs publication was introduced by the CSO in 2011 with annual data available from 2008.  The data source for this publication is the EHECS. This differs from the LCS data supplied to Eurostat in that it includes enterprises with 3 or more employees,   whereas the LCS includes 10 or more employees.  The main variables, all at NACE 1-digit level,  are  average annual earnings, average annual other labour costs, average annual total labour costs, total annual earnings and total annual labour costs.  Further breakdowns for regular earnings, overtime, irregular earnings and other labour costs are available on the CSO website.  See links:

 http://www.cso.ie/px/pxeirestat/Statire/SelectVarVal/Define.asp?maintable=EHA04&PLanguage=0

http://www.cso.ie/px/pxeirestat/Statire/SelectVarVal/Define.asp?maintable=EHA05&PLanguage=0

[Not requested]

[Not requested]

Other publications that include earnings and labours cost data from EHECS include;

- CSO's yearbook,

- outouts from the ILO, OECD and UN.

The quality report for the EHECS is updated annualy and outlines the quality procedures and practices in place. It is available here:

http://www.cso.ie/en/surveysandmethodology/standardreportsonmethodsquality/

Metadata relating to the production of STS, LCI and JVS statistics, all of which are based upon EHECS are updated annually and are assessed by Eurostat.

Metadata relating to the Labour Cost Survey is compiled for each LCS to reflect the policies and procedures employed in compiling the LCS.

The quality report for the EHECS is updated annualy and outlines the quality procedures and practices in place. It is available here:

http://www.cso.ie/en/surveysandmethodology/standardreportsonmethodsquality/

100%

[Not requested]

[Not requested]

-

[Not requested]

[Not requested]

Additional documentation and publications

•  CSO Survey and Methodology

http://www.cso.ie/en/surveysandmethodology/earnings/

• Labour Cost Survey / Labour Cost Index

http://epp.eurostat.ec.europa.eu/portal/page/portal/labour_market/labour_costs

• Short Term Statistics

http://epp.eurostat.ec.europa.eu/portal/page/portal/short_term_business_statistics/introduction/sts_in_brief

• Job Vacancy Statistics

http://epp.eurostat.ec.europa.eu/statistics_explained/index.php/Job_vacancy_statistics