National Institute for Statistics and Economic Studies (INSEE)

DIRECTORATE OF DEMOGRAPHIC AND SOCIAL STATISTICS

Department for Employment and Working Income

Earnings and Working Income Division

TIMBRE DG75-F240 - 18 bd Adolphe Pinard - 75675 PARIS CEDEX 14 – FRANCE

The Structure of Earnings Survey (SES2014), supplied by INSEE to Eurostat the 8th of July 2016, results, as for 2010, from the concatenation of the data from the annual structure of earnings surveys (ESS) 2013 and 2014, after updating the financial information collected in 2013 in such a way as to be representative of 2014. The scope of these annual surveys includes the private sector, public hospital services and local and regional public services.

In addition, a complementary survey of central public service employees (FPE 2014) was carried out for the year 2014. This survey completes the scope of SES2014 in accordance with the Structure of Earnings Survey 2014: Eurostat’s arrangements for implementing Council Regulation 530/1999, and Commission Regulations 1916/2000 and 1738/2005

The classifications used are those set out in the implementation rules for structure of earnings statistics in “Structure of Earnings Survey 2014: Eurostat’s arrangements for implementing Council Regulation 530/1999 and Commission Regulations 1916/2000 and 1738/2005”.

The scope of SES2014 covers the employees and local units of enterprises with at least 10 employees located in France (excluding Mayotte), in the areas of economic activity defined by sections B to S of NACE rev. 2.

In comparison with 2010, the scope has thus been extended to include local units and employees in the Overseas Departments excluding Mayotte (Guadeloupe, French Guiana, Martinique and Reunion).

The concepts and definitions of the variables follow the implementation rules for structure of earnings statistics in the “Structure of Earnings Survey 2014: Eurostat’s arrangements for implementing Council Regulation 530/1999 and Commission Regulations 1916/2000 and 1738/2005”.

The statistical units are local units (“établissements”) and employees (“salariés”) defined in accordance with the “Structure of Earnings Survey 2014: Eurostat’s arrangements for implementing Council Regulation 530/1999 and Commission Regulations 1916/2000 and 1738/2005”.

SES2014 contains 40,360 local units and 270,248 employees.

The reference population covers employees receiving positive remuneration in 2014 during a reference month. For France, the reference month is an average month in the year.

Not applicable.

Not applicable.

1 – Sources

The SES2014 is based on the following sources: the annual structure of earnings surveys (ESS 2013 and ESS 2014), the complementary four-yearly survey of central public service employees (FPE 2014) and exhaustive administrative sources.

The ESS 2013 and 2014 surveys

The ESS surveys are annual surveys, constituting the first part of the Ecmoss device (two years of structure of earnings surveys + two years of labour cost surveys). These surveys cover local units and employees from the private sector, public hospital services and local and regional public services.

Complementary survey of central public service employees 2014 (FPE 2014)

To cover the entire scope required by Eurostat, INSEE conducted in 2015 a complementary survey of central public service employees upon their earnings and working time/conditions in 2014. These employees concentrate in (non-market) education (section P) and public administration (section O).

Administrative sources

The exhaustive administrative databases serve both as a sample frames for the surveys and as complementary databases of information, in order to reduce the survey response burden on firms. These administrative databases and registers are as follows:

Firm register: the samples of local units for the ESS surveys are drawn from the “identification system for the register of statistical units” (SIRUS) register, a statistical register listing all firms and local units (establishments) and supplying information about the size of the local unit, the size of the firm, the economic activity, etc..

Annual Declarations of Social Data (DADS): these declarations include any firm employee receiving wages, whatever the amount perceived or the duration of the employment spell. These declarations are intended first for the social security organisms and the tax authorities. The social security organisms use them to calculate social contributions owed by employers, to check that employers have paid all the contributions that are due and determine employees’ retirement pension and health cover entitlements. INSEE is also an official recipient of these declarations to produce statistical databases about employment and earnings, after quality control of the administrative information and adjustment

The Public Service Employee Information System (SIASP): this exhaustive statistical file is based on the State employee payroll and on the DADS for public hospital service and local and regional public service employees. It is the reference source for measuring the number of employees, the wage costs and the number of paid hours in the public sector.

The information about earnings contained in the DADS and the SIASP serve as a basis for calculating taxes and social contributions, and is therefore of good quality.

These databases are used upstream of the survey to determine the number of local units and employees to sample in each stratum (see point 2). They also act as samplingbases for the employee-level survey, as calibration bases , and provide additional information not requested in the surveys or which is compared with the responses to the survey (see point 3). For employees, this information includes:

 -   age, sex, duration of employment in days

 -   socio-professional category

-    gross earnings in the year

-    type of employment, type of employment contract

2 – Sampling

The samples for the ESS 2013 and 2014 surveys

The sampling method for ESS 2013 and 2014 is the same as for the previous edition. It relies on a 2-level stratified sampling design, the first level being made up of local units, and the second of employees within those units. This stratification improve estimate precision by creating groups that differ from each other but in which the observed variables of interest are relatively homogeneous.

The sampling designs minimise the variance of the average of a proxy variable for the hourly gross earnings (this proxy is the ratio of the gross earnings such as recorded in DADS/SIASP divided by the number of paid hours such as recorded in DADS/SIASP), allocating more statistical units to be interviewed in heterogeneous (in terms of the proxy distribution) strata than in homogeneous ones in order to improve the precision of the average estimate of the hourly gross earnings, both at each stratum level and globally

More precisely, the sampling allocation between strata is calculated by minimising the variance of the stratum average of the proxy subject to a fixed minimum number of local units to be interviewed per stratum (constrained Neyman allocation). It requires to successively compute a minimum number of local units in each stratum for a given level of precision, a Neyman allocation for the number of local units (1st degree), the minimum number of employees to be interviewed in each stratum to achieve the precision objective and, lastly, a Neyman allocation subject to the numbers of employees constraints (2nd degree). For the employees, the allocation distinguishes management (cadres) from non-management job positions within each previously-defined stratum.

The exhaustive statistical databases are thus used to define the allocations of the number of local units and employees to be sampled per stratum (validity n-2 for survey n), and then to carry out the sampling (validity n-1 for survey n).

For ESS 2013, the sample for employees is drawn from the employees employed on 31/12/2012 in the DADS 2012 (who a priori would be employed for at least a part of 2013) . In the same way, the employee sample for ESS 2014 is drawn from employees employed on 31/12/2013 in the DADS 2013.

The sampling design strata are based on the combination of four variables (economic activity, firm size , size of the local unit and geographical location). The combination of these different factors led to a total of 826 strata.

Economic activity (sub-sections of NACE rev. 2 or groupings of sub-sections)

"BB" = Extracting industries (section B)

"AT" = Food, beverages, tobacco (sub-section CA)

"TH" = Textiles, clothing, wood, paper (CB+CC)

"RC" = Refining, chemicals, pharmaceutics, plastic, other minerals (CD+CE+CF+CG)

"MM" = Metals, machines-electronics, electrics, IT, automobile (CH+CI+CJ+CK+CL)

"MR" = Furniture, repairs (CM)

"ED" = Electricity, gas, water, sanitation, waste (D+E no distinction is made here between sanitation and depollution of the remaining waste)

"FF" = Construction (F)

"GG" = Trade (G)

"HH" = Post and transport (no communication - H)

"II" = Accommodation, food services (I)

"JJ" = Information, communication (J)

"KK" = Finance (K)

"LL" = Property (L)

"MA" = Legal, accounting, engineering and technical activities (MA + part of MC)

"MB" = Scientific research and development, specialised and veterinary scientific activities (MB + part of MC)

"NN" = Administrative and support service activities (N)

"OO" = Administration (O)

"PP" = Education (P)

"QQ" = Human health and social action (Q)

"SR" = Arts, entertainment (R)

"SR" = Other services (S)

Geographical location

“IDFR” = Île-de-France (Z1)

“NEST” = North and East (Picardy, Champagne-Ardenne, Burgundy and Z3 and Z4)

“OUES” = West and Southwest (Centre and Normandy and Z5 and Z7)

“SUDE” = Centre-East and Mediterranean (Z8 and Z9)

“DOMS” = Overseas Departments (Z0)

Some combinations are grouped into “TTZT” = all ZEATs, or with the single distinction Île-de-France (“IDFR”) / Province (“PROV”).

Size of firm

“TEG1” = from 10 to 49 employees

“TEG2” = from 50 to 499 employees

“TEG3” = 500 employees or more

Size of local unit

“A1” = from 1 to 9 employees

“A2” = from 10 to 19 employees

“A3” = from 20 to 99 employees

“A4” = from 100 to 499 employees

“A5” = 500 employees or more

The sample for the FPE 2014 survey

The sampling design of the 2014 FPE survey is balanced and stratified (in 2010 it was only stratified).

The balancing variables are the age of the employee (above or below 35 and above or below 50), the supervising ministry of the employer (17 modalities), the region in which the employee works (24 modalities), the employee’s status (civil servant, fonctionnaire, or not), the statutory category (A, B, C, unknown) and the gross hourly wage (such as observed in the SIASP).

The strata used in the sampling design were age (3 modalities), status (2 modalities), sex (2 modalities), the supervising ministry for the employer (5 modalities), the region (2 modalities) and whether or not the individual has a supervisory position (cadre).

Being balanced, the sample is also representative of the population in terms of averages or total amount of continuous variables (for example total wage costs).

3 – Enhancement of the surveys

The DADS and the SIASP statistical databases (from validity n for survey n) are used to provide complementary information to the survey responses. For central public service employees (FPE), all variables relating to earnings and paid hours come from the SIASP.

For the ESS surveys, some (few) variables extracted from the DADS, are also present in the survey questionnaire (gross earnings, number of paid days, working time of part-time employees (%)). In such a case, they are used to check the quality of the variables collected, to correct them during the adjustment phase or to impute them in case of partial non-responses.

The complementary FPE survey is carried out every four years. The ESS surveys are conducted over two consecutive years every four years. The statistical databases DADS and SIASP each year provide annual information.

The ESS 2013 and 2014 survey employers (local units, establishements), to whom paper questionnaires are sent. Large firms can also respond using a computer file. Local units receive a “local unit” questionnaire and questionnaires concerning between one and 24 named employees from that local unit. It is therefore the employers who respond to “employee” questionnaires. Data collection runs from May to December of each relevant year (2014 for ESS 2013 and 2015 for ESS 2014).

For the FPE 2014 survey, the selected central public service employees respond to the questionnaire themselves (not their employers). Data collection is initially carried out over the internet, with the possibility of filling in a paper questionnaire when reminders are sent to those who have not responded yet. Data collection for the 2014 FPE took place from 26 September to 10 December 2015.

To reduce risks of measurement and processing errors, INSEE has included a data checking program in the computer application used to monitor the survey. This program checks the data consistency in order to detect errors during data collection. The system automatically checks that the elements included in an overall total are lower than that total. In addition, orders of magnitude are checked, using the distribution of variables observed in the previous survey. This allows to detect and correct data entry and optical reading errors, totals reporting to firms rather than local unit, or respondents’ calculation errors, which can go as far as multiplying or dividing a total by 10.

Once data collection is complete, INSEE carries out adjustments, non-response correction, calibration, and a detailed validation based on comparison with the previous edition and with other surveys and administrative statistical databases (see section 8).

Not applicable.

Updates of 2013 data to 2014 levels

The Eurostat database contains observations from surveys relating to 2014, and others relating to 2013 but updated to be representative of 2014. Variations in earnings observed in 2013 are indeed updated by multiplying them by the change in the mean gross hourly wage (according to a certain stratum) estimated from the DADS between 2013 and 2014.

The strata used to calculate changes in the mean gross hourly wage are composed as follows:

-          Three different firm sizes (fewer than 50 employees, 50 to 499, 500 and more)

-          Economic activity in NACE A21.

-          geographical location : Île-de-France versusother regions

The economic activity sections are grouped together when the numbers of employees are too small (NACE B-D-E, NACE K-L, NACE R-S), the locations are grouped together for the section R , firms with 50 to 499 employees are groupes with those with500 employees or more for the Île-de-France and the other regions for the section P.

The earnings variables updates are the gross annual earnings in the reference year (B41), the annual bonuses and allowances not paid at each pay period (B411), gross earnings in the reference month (B42), overtime earnings (B421), special payments for shift work (B422).

Neither variables relating to time worked (paid hours, hours worked by a part-time employee or length of presence in the local unit), nor variables that are characteristics of the individual, are adjusted.

The other adjustments carried out are described in section 8.

In order to evaluate their relevance, the ESS and FPE surveys are submitted every four years to the National Council for Statistical Information (CNIS) and the Quality Label Committee for Official Statistics. This council and this committee evaluate the public interest and the statistical quality of the project and check the quality of the operation, its process, statistical and accounting standards and tests carried out on the questionnaire before the survey. They also check that consultation has taken place with the partners concerned. Where appropriate the Quality Label Committee can grant the surveying body the right to oblige participants to respond, which insures a better representativeness of the survey results.

The CNIS (National Council for Statistical Information) issued a positive opinion on the ESS 2013 and 2014 surveys and the FPE 2014 survey (opinions n°52 and 53 dated 6 May 2013). The Quality Label Committee for Official Statistics designated these surveys as being in the public interest, of high statistical quality and of an obligatory nature, with visa numbers 2015X095EC for the FPE 2014 survey and 2015A065EC for the ESS 2013 and 2014 surveys.

In addition to this ex-ante quality label accordance, SES 2014 was also evaluated by different departments of INSEE or of the Statistical Office of the Employment Ministry, during the imputation/adjustment phase, the result validation one (see sections 3.4, 3.6, 6 and 8) and through exchanges with INSEE’s own quality unit and statistical methodology departments. After data collection, INSEE and the Statistical Office of the Employment Ministry carried out imputations and adjustments (see section 3.4, 3.6, 6). INSEE then carried out in-depth validation of the data by comparisons with the previous edition, and with external sources (National Accounts, DADS/SIASP administrative data, Labour Force Survey, see section 8).

Many national organisations use SES 2014: INSEE, the Labour Ministry (in particular its Surveys and Statistics Department, the DARES) and researchers.

The DARES uses the ESS surveys to respond to many requests relating to different components of remuneration and the organisation of working time (statistics by type of collective agreement, the characteristics of employees earning the minimum wage, employee savings schemes, overtime, working time…).

Information from these surveys is also used for national accounting at INSEE and the welfare protection accounts department at the DREES (Directorate of Research, Surveys, Evaluation and Statistics of the Ministry for Health). Social science researchers also use this data source that includes both local units and their employees, to study companies’ wage practices.

Not applicable.

All the obligatory variables required by Eurostat regulations are provided.

Not applicable.

Not applicable.

Refer to 'Sampling error indicator'.

The SES 2014 is made up of variables at the employee level (270,248 employees) and variables at the local unit level (40,360 local units). The variances of the average estimators of the variables of interest are computed empirically by bootstrap.

The tables below report empirical precision information obtained by bootstrap with 1,000 replications of identical size (CV denotes the coefficient of variation, p_2_5 and p_97_5 are respectively the 2.5% and 97.5% percentiles of the distribution of the bootstrap replicates, i.e. the lower and upper limits of the 95% confidence limits for the estimators of the averages of the main variables of interest, that is the gross earnings in the reference month and the gross hourly earnings in the reference month (as required by the Eurostat regulation).

For the gross hourly earnings (variable B43)

For the gross earnings in the reference month (variable B42):

Refer to sub-part of non sampling error for more information.

The central social security organisms (CNAM, CNAV, etc.) are not in the survey database; neither are public employees of the French armed forces. The samplings upon employees rely on employees having a position the 2013/12/31 (resp. 2012/12/31 for ESS2010), those who enter a firm later in 2014 cannot be interviewed. However, the coverage default is corrected by the final calibration of the data upon the 2014 information issued from the exhaustive statistical databases (DADS, SIASP) at the end of the process.

To avoid errors of over-coverage, employees who are out of the scope of the survey are eliminated during the adjustment phase,for instance those whose employment contract ended the 2013/12/31 (approximately 2% of cases) (resp. before 2013 for ESS 2013).

Not applicable.

To reduce risks of measurement errors, INSEE has included a data checking program in the computer application used to monitor the survey. This program checks the data consistency in order to detect errors during data collection. The system automatically checks that the elements included in an overall total are lower than that total. In addition, orders of magnitude are checked, using the distribution of variables observed in the previous survey. This allows to detect and correct data entry and optical reading errors, totals reporting to firms rather than local unit, or respondents’ calculation errors, which can go as far as multiplying or dividing a total by 10.

The central variables of the survey (gross earnings and number of paid hours), are checked by comparison with individual information coming from the exhaustive statistical databases (DADS and SIASP). For ESS 2014, the survey and administrative data were comparable (less than 5% difference) in 80% of cases, significantly different in 5% of cases and survey non-responses concerned 15% of cases. When the data were not consistent, variables were adjusted using information coming from the administrative sources or from other parts of the questionnaire. These adjustments lead to a reduction of 20% in the average gross hourly earnings in ESS 2014 (reduction of 23% in the Ecmo 2012), while the median remained unchanged.

A second adjustment was then carried out after the imputation and the adjustment phases, to satisfy the requirements imposed by Eurostat: strict respect of limits for several variables (for example working time and overtime payments), and rejection of of partial non-responses (deletion of observations for whom certain variables remain missing or not coded with enough details after the imputation phase).

Response rate

For the three surveys, ESS 2013, ESS 2014 and FPE 2014 , 360,000 employees were questioned, 313,216 useable responses were received, representing a response rate of 87.0%. Total non-responses were more frequent in the Overseas Departments than in Metropolitan France (26% compared to 14%).

Processing operations for total non-response

The initial survey weights were corrected for total non-response and modified by calibration of the employee sample structure. The weights of the local units were obtained in a similar way.

Total non-response correction

The ESS 2013 and 2014 files were corrected for total non-response by post-stratification: when a unit did not respond, its weight was re-allocated homogeneously over the units from the same stratum that did respond.

Influential unit weights correction

A programme of processing of the most influential units was then applied in order to control for the influence of individuals who, by virtue of their response and their weight, may biased the statistics of the group to which they belong but without their response being incorrect. The processing use the Kokic and Bell’s method andmodifies the individual’s weight in order to reduce the risk of bias without losing the information from the response provided by the (influential) individual. The processing of influential units was not carried out in 2010. It improves the precision of the estimators of gross hourly earnings, in particular in groups where the gross hourly earnings variance is high and in groups with few individuals.

The weights corrected for these influential units were then calibrated using the margins obtained from the exhaustive statistical databases (DADS 2013 and 2014).

For the FPE 2014 survey, the weights were also corrected for total non-response by post-stratification (the new ponderations were obtained by dividing the initial weights by the response rate in the stratum). The strata to be used were obtained by combining status, category, gender, age (same groups as for the adjustment) and the employing ministry.

Calibration

For ESS 2013 and 2014, the calibration process contains two steps.

1 – Variables extracted from the administrative statistical databases, DADS, are calibrated upon the margins reporting to the same variable computed from the DADS database corresponding to the year of the survey. These variables are the number of paid hours, the gross wage and the employment duration. The calibration is done by:

-           Socio-professional category xsex

-           Geographical location (9ZEATs).

-           Firm size (< 50, 50-499 and > 500) x economic activity ( 21 NACE categories), with grouping NACE sections with too few observations.

-           Full time/part-time

The number of local units are also calibrated by:

-           Geographical location (9 ZEATs)

-           Firm eize (< 50, 50-499 and > 500) x economic activity (21 NACE categories ), with grouping NACE sections with too few observations.

-           Public/private control : the financial control variable, which is constructed from the Register of companies under State majority control.

2 – The ESS 2013 and 2014 files were then merged and calibrated again using the margins from the DADS 2014, so that the SES 2014 is eventually representative of 2014.

For the FPE 2014 survey, the calibration margins are computed from the SIASP 2014.

The numbers of employees, hours paid, gross earnings and employment duration, taken from the SIASP, were calibrated to the margins by:

-           Category (A, B or C) x Status (civil servant or not) * sex

-           Employing ministry x location

-           Age group (below 35, 35 to 49, 50 and over)

-           Full time / part time

All those calibrations were carried out using the Calmar¹ macro.

Processing operations for partial non-response

Eurostat does not accept partial non-response. In order to keep as many observations as possible in the SES2014, observations with partial non-responses are imputed when possible.

Usually the imputation procedure relies on a consistency analysis that use information extracted from the administrative statistical databases (DADS, SIASP). When this is not possible, other imputation methods are used :

- imputation using regulation information (on working time duration, number of days of holiday leaves, of working time reduction)

- predictions using models estimated on respondents (employment duration, % of working time for part-timers)

- imputation by the mean (overtime, and related earnings)

- imputation by hot-deck (using values from respondent individuals with similar characteristics)

Hot-deck imputation is used for the length of service in the firm. Respondents considered as similar to the partially non-respondent are in the same

economic activity, age , socio-professional category, geographical location (Île-de-France /Other regions) groups for employees with a indefinite-duration contract, ;

type of contract, economic activity and age groups for apprentices or those with a fixed-term contract.

Hot-deck imputation is also used for the highest successfully completed level of education. Respondents considered as similar to the partially non-respondents are in the samesame economic activity, size of local unit, geographical location, socio-professional category, same earnings bracket groups.

See section “6.3.4.1. Imputation rates” for detailed tables of imputation rates by variables.

1CALibration on MARgins, iterative method enabling modification of the weighting of the cells of a pivot table to obtain a priori fixed margins.

Not relevant.

Not relevant.

See below.

The rates of adjustment or imputation compare the number of individuals (here, unweighted) whose response has been adjusted/imputed to the total number of individuals.

The percentages are close to those observed for the ESS 2010.

Imputation rates in ESS 2014 and FPE 2014

* the imputation rate is reported only for variables using survey responses (or both survey and administrative data),

it equals 0 by construction for variables constructed only from administrative data

** only for employees who have not an annual working day contract

*** only for non-teachers

Source : ESS 2014, FPE 2014

Adjustment rates in ESS 2014 and FPE 2014

*data edition of non-missing variables

** only for employees who have not an annual working day contract

Source : ESS 2014, FPE 2014 ; scope : non missing variables

Note : 21.33 % of non missing data are adjusted to correct their number of hours actually paid during the reference month.

Not applicable.

Not applicable.

Not applicable.

Not applicable.

Not applicable.

The SES2014 will be available for use and will generate publications from the end of 2016.

The first publications are realised by the DARES and the INSEE (INSEE-Première, INSEE-Résultats, which will contain detailed tables of results) by the end of 2016 and in 2017.

Not applicable.

The schedules were adhered to.

More precisely, for ESS 2014, the schedule for conducting the survey was as follows:

-           letters announcing the survey were sent out in March 2015;

-           the questionnaires (paper or electronic) arrived in local units in May 2015;

-           first reminder for local units that had not responded yet to either the local unit questionnaire or the employee questionnaires sent in July 2015;

-           second reminder in September 2015, and then, if necessary, a formal notice to respond within ten days;

-           statement of non-response drawn up in October 2015 for the last local units that had not responded.

-           data collection ended in mid-December 2015

From the mailing of the questionnaires to the end of data collection, devoted staff dealt with the employers' requests (answers to questions, agreement to additional response time, etc.). They also checked the questionnaires, contacting respondents in the case of significant errors, all during the data collection.

For ESS 2013, the survey schedule was very close to that for ESS 2014 described above, with one year difference.

For FPE 2014, the schedule was :

-           end of September 2015: letters announcing the survey sent out and Internet site for responses opened;

-           end of October 2015: first reminder accompanied by a paper questionnaire;

-           mid-November 2015: second reminder;

-           mid-December 2015: end of data collection on paper (10 December) closure of the Internet site for data collection (14 December);

The first databases (ESS and FPE) were delivered in February 2016 for data editing, coding, imputation, adjustment, calibration and validation processes . The SES2014 was eventually sent to Eurostat on 8th July 2016. The exhaustive statistical databases (DADS 2014, SIASP 2014) required for the final calibration were only available mid-May 2016.

Not applicable.

European comparison

The variables NUTS (region where the local unit is situated), size of the firm and economic activity (NACE rev. 2 at the division level) come from the firms' register.

For employees concerned by the ESS survey (excluding those working for central public services concerned by the FPE survey), age, sex and number of paid days come from the DADS. For central public service employees, the earnings variables and the number of paid hours come also from the SIASP.

The other variables are taken from the local units’ responses to the employee questionnaires for the private sector, public hospital services and local and regional public srvices (ESS surveys), and from the responses provided by the employees themselves for those working for central public services (FPE survey).

The annual structure of earnings surveys (ESS 2013 and ESS 2014) offer local units two modes of data collection: paper questionnaires or computer files. The FPE survey also offers two modes of data collection: website and paper questionnaires. Whatever the data collection mode used, the questionnaires contain the same information.

The reference month is not a specific month, but corresponds to an average month in the year.

The number of days of holiday leave in the year (B33) is the number of days of holiday leave actually taken and not the legal number of days to which the individual is entitled, in order to remain consistent with the 2010 survey.

Not applicable.

Comparability between the SES2014 and SES2010 surveys

1- Scope extension

The scope of the survey was extended between 2010 and 2014:

-           to local units and employees in the Overseas Departments (except Mayotte)

-           to employees employed by central public services and paid by public administrative establishments: for example national public establishments of a scientific, cultural and professional nature (higher education), local public teaching establishments (secondary education) …

-           to employees of social security organisms, who were considerably under-represented in the 2010 edition.

Moreover, researchers in public research institutions were deleted from the SES2010 because their profession was coded with a single digit only. This is no longer the case in SES2014.

2- Changes in data treatment

In the ESS surveys

-           improvement in the calculation of hours worked and the hours paid

The SES2014 survey contains only the number of paid hours, but the number of hours worked is also calculated and used to test data consistency. This calculation and consistency analysis have been improved since the 2010 edition. The new treatment follows the same methodology as the 2012 Labour Cost Survey. In particular, the survey data relating to employees with an annual working days contract are compared with those coming from the 2014 Labour Force Survey (number of hours worked per day per social category, and by economic activity for managers (”cadres”)). In addition, the number of days worked is measured better thanks to new questions about the number of days of working time reduction and days placed in time savings accounts. However, the 2014 questionnaire no longer allows the distinction to be made between paid absences and unpaid absences as it was in 2010, the percentages observed in 2010 have thus been used.

-           changes in ISCO coding

There is no longer information about the main tasks and duties carried out by the employee in the ESS 2013 and 2014 questionnaires, unlike the 2010 edition. Thisentails a quality loss concerning the coding of the profession (ISCO) (it is more difficult to distinguish skilled employees and other elementary occupations). However, the coding program used in 2014 has been improved and makes a clearer distinction between managers and professionals.

-           changes in the calibration processes

* improvement in the treatment of weights of influential units (Kokic and Bell)

* changes in the calibration margins : in 2014: NUTS1 and firm size x economic activity, whereas in 2010: firm size , economic activity x Ile-de-France/Other regions.

In the FPE survey:

-           changes in the calibration margins:

in 2014: age x sex; NUTS1 grouped into five zones; statutory category x status; ministries grouped together,

in 2010: status x category x sex; ministries x Ile de France/Other regions; age; statutory category.

-           other changes in treatment:

* The statistical unit used in the 2014 local unit table always refers to “local units” (more precisely to “establishments” with a SIRET identifier) whereas in 2010, concerning the state public services, only “pseudo” local units are available. The latter were constructed from the combination of the anynomised firm identifier, the location of the establishment and the economic activity but the SIRET identifier itself in the SIASP database was not enough reliable to be used to compute anonymised key identifiers.

* Local unit size. The local unit size was not supplied in 2010 (because of the “pseudo-SIRET” used). It has been supplied for FPE 2014 and computed in the comprehensive DADS/SIASP combination. However, 900 observations are missing.

* Changes in the “bonuses and allowances not paid at each pay period” variable. The NBI (new index bonus) is in 2014 included in monthly bonuses whereas in 2010 it was included bonuses and allowances not paid at each period of time. Some other bonuses such as the collective performance bonus and an indemnity covering expenses have been included in non-regular bonuses and allowances.

* Overtime and associated earnings. In 2010, the number of hours of overtime in the reference month provided to Eurostat was the number of hours of overtime for the year divided by 12, and did not account for the employee employment duration in the year. Overtime was, therefore, probably underestimated. This has been corrected in 2014. Moreover, overtime earnings provided in FPE2010 covered the whole year and were not reported to the reference month. This has also been corrected in 2014.

Finally, FPE 2010 used only overtime and related earnings that were exempted from employer social contributions. Since 2012, there are no more exemptions of employer social contributions for overtime, and supplementary variables have been added in the questionnaire to account for them. So changes between 2010 and 2014 concern above all teachers (supervising homework, supervising detention, annualised hours (HSA) over initial working hours...).

Not applicable.

The coherence analysis is presented in section 8.5 (comparison with external sources) and in section 8.6 (internal comparison).

The coherence analysis is presented in section 8.5 (comparison with external sources) and in section 8.6 (internal comparison).

The validations are based on comparisons of levels of, and changes in variables at the individual level (micro) and at some aggregate levels (earnings, employment duration, paid hours) between SES2010 and SES2014 and with external sources (DADS, SIASP, LFS, LCI, NA).

- Comparisons at the micro-level

Levels in the SES 2014 and changes between 2010 and 2014 of main variables are compared to those of “close” measures in the DADS and SIASP upon the same scope as the one of the survey. Eurostat concepts of earnings and earnings observed in the DADS do not completely overlap. The gross earnings observed in the DADS at date t include: gross wage and bonuses, benefits in kind, severance or early retirement payments above the minimum, and employee savings contributions at t-1. Earnings (B41) in the SES2014 survey include: gross wage and bonuses, benefits in kind, leaving or retirement bonuses and employee savings contributions at time t. The B42 and B43 variables do not include non-regular bonuses. Comparisons of levels of paid hours greatly depend on the imputation made for employees with annual working days contracts.

The main differences stressed are:

- B41: average change in the administrative databases of 7.2% vs 6.9% in the survey. Over-estimation in section D (Production and distribution of electricity, gas, steam and air conditioning) and under-estimation in section F (Construction) in the 2010 survey.

- B42: part-time employees: changes observed in the surveys smaller than what is observed in the administrative databases. This may be explained by the fact that the new method of adjustment relying on consistency.

- B32: the change in the methodology for the calculation of paid hours (imputed hours for employees with annual working days contracts) has an effect in sections D and K.

- B43: average change observed in the administrative databases + 6.3% vs 6.9% in the survey.

- Comparisons of aggregated data (macro-level)

SES2014 covers enterprises with 10 employees or more, excluding employees of private individual employers and of the agricultural industry. The data from the national accounts cover the entire population, including firms with less than 10 employees and private individual employers. The administrative databases (DADS and SIASP) are used here on the one hand with a scope close as possible to the survey scope and on the other hand with a scope as close as possible to that of the national accounts.

We first compare the gross annual earnings (B41) for the average number of employees in the year as required in regulation 698/2006. The biggest differences in levels between the national accounts (NA) and SES2014 occur for sections I (Accommodation and Food Services), K (Financial Activities) and R (Arts, Entertainment and Recreational Activities).

These differences can be explained by coverage differences between the survey and the NA: come from differences between the DADS and the national accounts (note that on K Financial Activities, there are no financial companies in the NA). The earnings in the NA also include those of self-employed. Changes between 2010 and 2014 observed in the survey are consistent with those observed in the DADS even when they are not directly consistent with those in the NA. One may stress again the large change in construction, which is due to an under-estimation of the level of earnings in SES2010). The changes in O and P result only from the extension of the coverage. The conclusions are the same when considering earnings for the reference month (B42).

 Sources: National accounts – Base 2010 ; DADS 2010 and 2014 ; SES 2010 and 2014

- Comparison with the “Labour Cost Index – Wages and Salaries”

The comparison of SES 2010/2014 changes with the annual Labour Cost Index – Wages and Salaries is restricted to sections B to N. The difference between the hourly wage change in the survey and the annual Labour Cost Index – Wages and Salaries changes between 2010 and 2014 is fairly low. The biggest differences appear in Construction (F), Accommodation and Food Services (I), and are explained by coverage differences: these sections contain a large proportion of local units with fewer than 10 employees.

Comparison of changes between 2010 and 2014 in the SES and the annual Labour Cost Index – Wages and Salaries.

Source : SES 2010 et 2014 ; annual labour cost (wages and salaries) 2010 et 2014

Scope: non-agricultural market sector excluding household services

SES 2014 : wage-earners of firms with at least 10 employees ;

LCI : wage-earners of firms with 1 employee or more

Structure comparison with the labour force survey 2014

The structures by sex (B21), age group (B22), occupation at ISCO single-digit level (B23), supervisory or managerial position (B24), highest completed level of education (B25), length of service in the firm (B26), part-time (B27), type of employment contract (B28) in the SES 2014 and the 2014 Labour Force Survey (LFS) were compared on a scope close to the one of the SES 2014 (all employees working for a firm (public or private) with 10 employees or more, including managers and directors who are employees, but excluding employees of private individual employers).

The structures are reasonably coherent by age, sex, length of service, type of employment contract. The proportion of managers and workers with supervisory positions (survey responses in both cases) is smaller in SES2014 than in the LFS (13% against 20%). The structures by occupation (ISCO at single-digit level) differ more, which may be explained by the fact that “main tasks and duties” are not recorded in SES2014 and because the socio-professional categories coming from the administrative databases may differ from those coming from a household survey. Finally, the proportion of part-time employees is slightly higher in the SES2014 than in the LFS, which may be explained by the worker/main job approach used in the LFS versus the job approach of SES 2014.

Sources : SES2014, LFS2014

Scope : wage-earners of firms with at least 10 employees

Eurostat requires to verify changes in a certain number of statistics between 2010 and 2014 by:

• Age group: below 30, 30-39, 40-49, 50-59, 60 and over;

• Sex;

• Economic activity: groups of Nace sections; BN, BF, GN, PS, BS_O

• Occupation: ISCO code at single-digit level (9 categories), then grouping together 1-5 (non-manual work) and 6-8 (manual work).

• Full-time/part-time

Any change exceeding 10% or deviating from the mean change is examined. Comparisons between surveys are carried out for comparable coverages: in this case, private companies + public bodies except Overseas Departments and except social security organisms, but including public employers’ local units (legal category 73) because there has been switching between legal categories (71 and 73) following the law on the autonomy of the universities, introduced between 2010 and 2014.

- changes in numbers of employees

The important changes involve:

-           ISCO 2 “professionals”, which is partly explained by the coverage extensions upon teachers, better representation of public researchers, improvement in occupation coding, but with a risk of coding too many people as “professionals” (see comparison with the Labour Force Survey).

-           ISCO 9 “elementary occupations”: risk of inaccurate coding because tasks and duties are not recorded.

-           Sections D and E (grouped in the calibration process): changes coherent overall but not these two sections taken separately.

-           Sections O and P: coverage extension.

- changes in numbers of employees in local units

In 2014, local units correspond to establishments both for ESS and FPE. In 2010, for FPE, only pseudo local units were provided. With comparable concepts between the two surveys, changes in the number of local units can be completely explained by coverage extensions and the effects of grouping together sections D and E in the calibration process.

- other variables

Eurostat requires the changes between SES2010 and SES2014 to be verified for the following statistics:

-           B41 (gross annual earnings): mean, median, Q10, Q90

-           B411 (annual bonuses not paid at each earnings period): overall mean, mean for employees with B441>0

-           B33 (annual days of holiday leave): mean

-           B42 (gross wage in the reference month): mean, median, Q10, Q90

-           B421 (overtime earnings): overall mean, mean of employees with B421 >0

-           B422 (shift work earnings): overall mean, mean of employees with B422 >0

-           B43 (average gross hourly earnings in the reference month= B42/B32): mean, median, Q10, Q90 and for each section of the NACE: mean

-           B32 (paid hours): overall mean

-           B321 (overtime hours): overall mean, mean of employees with B321 >0

-           Number of employees: weighted sum (see above)

-           Number of local units: weighted sum (see above)

-           Proportion of “low-wage workers” based on B42 (<50% of the median earnings)

We summarise the most noteworthy results here:

-           B41 and B43: changes distant from the average change for ISCO = 2 and ISCO =9 (related to the points already mentioned); in section D (production and distribution of electricity and gas), F (construction), N (administrative and support services).

-           B421 (overtime earnings) and B321 (overtime): these earnings and hours are better measured for teachers (thanks to the use of additional variables from the SIASP).

-           B32 (paid hours): changes distant from the average change for ISCO = 1, 4, 9, sections F, D, O, P, N, R linked to the new calculation of coherence for paid hours and with the changes in the ISCO coding.

Not applicable.

SES 2014 will be the subject of publications by the INSEE and by the Statistical Office of the Ministry for Labour (DARES) from the end of 2016 and in 2017.

The data are available and consultable on the Eurostat site.

The data are available and consultable on the Eurostat site.

The SES2014 anonymised individual data for France will be made available to researchers via the Eurostat data centre.

Several databases of individual data are available for statistical studies at the INSEE and for ministry statistics services, in particular each annual ESS survey and the complementary FPE survey.

In addition, related individual data for an extraction of variables at the employee level are made available to researchers via the Quételet centre.

The local units database is, moreover, available to researchers on request from the National Confidentiality Committee.

These databases will be available by the beginning of 2017.

Detailed tables of earnings structures will be available on the INSEE web site during 2017 (INSEE – résultats).

The ESS surveys are part of the Ecmoss device, which is the subject of a descriptive document on the INSEE site. The questionnaires from the last two surveys are also available on the INSEE site.

(https://www.insee.fr/fr/metadonnees/source/s1221)

The FPE survey is in presented on the INSEE website: https://www.insee.fr/fr/metadonnees/source/s1263. The questionnaire is also on line (documentation section).

Moreover, codebooks accompanying each database present also the survey (history and methodology) and, for each variable, state whether it has been the subject of adjustment.

The CNIS has delivered a positive opinion (avis d’opportunité). The Quality Label Committee for Official Statistics designated these surveys as being in the public interest, of high statistical quality and of an obligatory nature.

These documents are linked here and may be consulted on the CNIS site : http://www.cnis.fr/cms/Accueil/enquetes/Avis

Not applicable.

Not applicable.

In order to reduce the response burden for employers, the data from the Annual Declarations of Social Data (DADS) for private sector employees and from the Public Service Employees Information System (SIASP) for public-sector employees were used as much as possible. The surveys were thus restricted to questions for which the information was not already available from these administrative sources (or was likely to be of poor quality).

The time required to answer to an employee questionnaire is estimated to be around 20 to 25 minutes. For the FPE survey, the average time required to complete the questionnaire is estimated to be about 15 minutes.

The ESS and FPE surveys were the subject of a declaration to the CNIL (National Commission for Data Protection).

For more information about the ESS, see the Ministerial Decree dated 10 March 2006 relating to the creation of automated processing of individual information relating to a statistical survey of structure of earnings and the cost of labour, published in the Official Journal (NOR: ECOS0650011A).

As far as the FPE is concerned, see the Ministerial Decree dated 21 September 2011 relating to the creation of automated processing of individual information relating to a survey of State employees, complementary to the annual survey on labour cost and structure of earnings, published in the Official Journal (NOR: EFIS1125542A).

Individual data are anonymised.

The published results satisfy the rules for statistical secrecy.

Nothing to report.