Statistics Poland

Labour Market Department

Statistics Poland

Al. Niepodleglosci 208

PL-00-925 Warszawa

See below.

Polish labour cost survey collects information on labour costs which is presented by kind of activities and ownership sectors. The labour costs survey has a representative character and results are generalized on the whole population. The basic measures used in Polish labour costs statistics are: average labour cost per 1 employee, labour cost per 1 hour paid and labour cost per 1 hour worked. 

Polish Classification of Activities 2007 (based on NACE Rev.2) which was introduced by the Regulation of the Council of Ministers on 24 December 2007  (Journal of Laws No. 251, item 1885), in force since 1 January 2008, was applied in LCS 2016.

The survey covers NACE Rev.2 sections (A-S) and divisions. Data for division 94 are not included in Polish LCS (and other surveys like SES or JVS), because of significant difficulties in collecting these data.

Labour costs consist of a sum of gross wages and salaries as well as non-wage related expenditures incurred in order to acquire, maintain, requalify and improve the staff qualifications.

Labour costs comprise the following components:

– personal wages and salaries,
– annual extra wages and salaries for employees of budgetary sphere entities,
– wages and salaries from the contracts for mandate or contract work,
– fees,
– expenditures on improvement, training and retraining of personnel,
– expenditures on business trips,
– expenditures on occupational safety and health,
– retirement, pension and accident insurance contributions paid by employer altogether with the contributions to the Guaranteed Employment Benefit Fund and the Labour Fund,
– benefits from the establishment’s social fund,
– benefits in kind,
– other expenditures,
– payments from profits in enterprises and from the surplus in co-operatives.

Statistical units include mainly local units or enterprises in case of inability to use local units.

The survey covers units of the whole national economy with 10 and more employed persons.

Poland

Data as of 31 December 2016.

Not applicable.

See below.

The data were collected from selected units of the whole national economy with 10 and more employed persons, according to sections and divisions of Nace Rev. 2 and by ownership sectors.

Beginning from LCS 1996 the survey has been conducted every 4  years.

The random sample was drawn from over 241.8 thousand public and private sector units located all over the country (sampling frame). The number of surveyed units  covered about 20% random sample of the whole sampling frame.  In order to draw a sample, a scheme of stratified and proportional sampling was applied.

Since the 1 of January 2009 Poland  has been applying an electronic questionnaire to collect statistical data on labour costs.

The data are validated both at the collection stage (e-questionnaire controls) and before transmission to Eurostat (verification due to the arithmetic controls).

Data aggregates are calculated due to Eurostat’s requirements and Polish data users needs.

Not applicable.

See below.

The quality policy is based on continuous improvement of statistical processes in order to better  satisfy the users’ needs, reduce respondents’ burden and to  diminish the costs of statistical  production.

Activities focused on quality improvements are in line with  the following International and European standards:

• Fundamental Principles of Official Statistics
• European Statistics Code of Practice
• Quality Assurance Framework QAF (It is currently in the process of updating resulting from the revision of European Statistics Code of Practice)
• Regulation (EC) No 223/2009 of the European Parliament and of the Council of 11 March 2009 on European statistics (amended on May 2015).

The quality policy of Polish Official Statistics and the standard tools for measuring, assessing and monitoring the quality of statistical surveys are presented in  the study  Vademecum of Quality in official statistics. The study is available at:

http://bip.stat.gov.pl/files/gfx/bip/en/defaultstronaopisowa/55/1/1/poz_vademecum_zew_ang.pdf

From 23 to 27 February 2015 in Statistics Poland was organized peer review. The quality issues were important part of it. 

Information on quality in Polish statistics is available on Statistics Poland website at:

http://bip.stat.gov.pl/en/activity-of-official-statistics/quality-in-statistics/

Overall assessment of data quality, based on standard quality criteria is good, as it is presented in detail in this quality report.

See below.

1. Description of the users

Users of the LCS data can be divided into the following groups:

National users:

Government bodies such as the Ministry of Family, Labour and Social Policy, the Ministry of Finance, the Ministry of Entrepreneurship and Technology, The Ministry of National Education;

National Central Bank;

Research centres such as the Polish Academy of Science, universities, high schools;

Employers -  for example employers of: financial intermediation and real estate sector;

Trade unions;

Students;

Mass  media.

International users:

European  Union institutions – European Council, European Commission;

International organizations such as OECD, European Central Bank, International Labour Organization;

Foreign  employers;

Foreign research centres.

2. Description of users needs

Generally,  national users are interested in the basic measures of labour costs and their impact on the situation on labour market e.g.

the Ministry of Family, Labour and Social Policy examines the relation between labour costs measures and unemployment;

the Ministry of Finance examines labour costs in different types of activity;

the The Ministry of National Education examines labour demand in relation to the basic measures of labour costs;

Research centres  are interested in analysis of labour costs in comparison with work efficiency, besides they conduct international comparisons on the level of labour costs in European Union Member States  and  other countries;

Employers  make economic analyzes with applying labour costs measures  in order to conduct  their adequate employment and investment policy;

Students use labour costs data in their master thesis or Ph thesis.

There are 2 missing variables in the LCS dataset: D1112 and D4.

Data for division X94 are not available as well.

X94 has never been covered by Polish LCS.  This activity is very difficult to report and other surveys (like SES) also do not cover it.

D1112 has always been unavailable, because of the lack of appropriate information. 
There is a possibility that D1112 will be available for the LCS 2020, because since 2019 the new regulation on employee’s capital plans is coming into force, which will make possible to collect this information.

Regarding D4 Eurostat was informed that the definition of labour costs from Regulation 1737/2005 is applied in the Polish LCS so labour costs components include taxes (gross earnings), however D4 is not collected separately.
Nevertheless, Statistics Poland is planning to use other data sources, e.g. some administrative registers to cover D4 variable to be able to provide it for the next LCS edition.

2 variables are missing out of 34 mandatory variables, so the ratio is 94.1%.

See below.

Data obtained from sample surveys such as the survey on labour costs are biased with: sampling and non-sampling errors which determine accuracy of the survey. Thus, limitation and reduction of these errors significantly influences on improvement of data quality and correct interpretation of survey results.

Sampling errors are related to the sample size and sampling  schemes. Their nature consists in the fact that incomplete information concerning a phenomenon influences on lack of confidence regarding relevance of estimates obtained from a sample survey. Thus, results of a sample survey should be treated as only approximate estimation on a value of  an unknown  parameter of population.   Therefore, on one hand we should be aware of incomplete reliability of results (i.e. differences between values gained from a sample and actual values in population, possible to obtain only from a full survey), while on the other hand we should try to obtain maximum credibility of data through adequate sampling. 

 The collection of surveyed units is characterised by two variables: 

• kind of activity, i.e. NACE division or group according to section,
• ownership sector (public and private).

 198 sub-collections were created which altogether covered 241758 units.

It was assumed that the number of surveyed units should cover about 20% random sample drawn from the sampling frame, i.e. over 48.5 thousand of units. Surveyed sub-collections cover very differentiated numbers of units. In consequence, in 91 small sub-collections including less than 150 units full-scale survey was carried out, i.e. 2407 units were included in the collection of surveyed units. In the remaining 107 sub-collections appropriate units were drawn.

The following assumptions were adopted in surveyed units selection:

• the sample units size is fixed a priori jointly for 107 sub-collections and equals about 46 thousand of units;
• in each sub-collection first one stratum (the upper stratum) is distinguished which contains the largest units according to the adopted allocation and stratification criteria.  The number of employed from the Business Register was used as stratification and sample allocation variable. In the upper stratum of the sub-collection drawing is not applied but all units included in this stratum  are surveyed; number of units in the upper stratum equals in total 25 905 units;
• in particular sub-collections, after excluding the upper stratum, sampling in the remaining part is performed according to the stratified sampling scheme.

The LCS is a difficult and time consuming survey. The biggest difficulties were connected with the completing variables for NACE Rev.2  divisions: 72, 85, 86, 90, 93  because certain group of persons employed in these types of activity (e.g. researchers, teachers, doctors, artists) has other, specific working conditions regulations than other professional groups.

Evaluation of sampling errors in the LCS 2016 is carried out on the basis of the relative standard error. Standard error determines a range of variation of a sample mean estimator around a real mean in population (standard error square is called variance of estimated mean). 

Standard error is a measure of data precision. The lower is standard error the higher is precision and  vice  versa – the higher  is  standard  error the lower is  precision. 

Sampling errors are only available  for the basic measures of labour costs by sections according to NACE rev.2  and ownership sectors.

Data on sampling errors  by types of activity are presented in Annex 1  of this report.

Relative standard errors (Annex 1 ) were estimated for the basic indicators of  labour  costs.

The analysis of sampling errors shows that the sampling errors for the basic measures of labour costs for the  whole national economy were  the following:

The highest sampling errors by types of activity took place in:

professional, scientific and technical activities – attained the level of 32% for labour costs per 1 employee (monthly average) and for labour costs per 1 hour paid and worked;

agriculture, forestry and fishing – ranged from 14.10% for labour costs per 1 employee (monthly average) to 13.54% for labour costs per 1 hour worked;

other service activities – ranged  from 9.31% for labour costs per 1 employee (monthly average) to 9.03% for labour costs per 1 hour paid;

accommodation and catering – ranged  from 9.94% for labour costs per 1 employee (monthly average) to 10.33%  for  labour costs per 1 hour worked.

The smallest sampling errors by types of activity were in:

mining and quarrying – ranged from 0.26% for labour costs per 1 hour paid to 0.29% for labour costs per 1 hour worked; human health and social work activities  – ranged about 2.52% for all categories; manufacturing – ranged about 2.68% for all categories. 

Non-sampling errors, which are the second type of errors in sample surveys, are divided into:

• frame  errors;
• measurement and processing  errors;
• non-response errors;
• model assumption errors.

Generally,  frame errors are divided into  over-coverage and  under-coverage  errors.

Over-coverage  errors  relate  to units present in the frame and which, in fact, do not belong to the target population or to units not existing in practice (e.g. units that have not been contacted at all, units that are in scope but classified in the wrong sampling strata, duplication in the sampling frame).

In the LCS 2016 information on dead units and inactive units which are part of over-coverage are following:

• dead units  constituted about 1.0 % of the selected sample (the ratio of 510 dead units to the selected sample of  48597 units);
• inactive units constituted  about 0.7% of the selected sample (the ratio of 357  inactive units to the selected sample of   48597  units).

Under-coverage errors   refer to  units not included in the frame, but which should be (e.g. delays in birth registration, lost registration applications). For these units  no information is available.

As for  methods of limitation and reduction of frame errors, errors due to  lack of answers from the whole unit are eliminated mainly through updating addresses in a sampling frame  and methods of results weighting. Errors deriving from  lack of answers regarding items are limited through the grossing-up correction.

1.8% of the selected sample (or 3.6 % if units which have not been contacted at all are included).

Not applicable.

Measurement errors are divided into:  the survey instrument (questionnaire) errors, the respondent errors, the information system and the mode of data collection errors.

 As for the survey instrument- questionnaires errors, the questionnaire in the LCS 2016 is designed in such a way to eliminate these types of errors because the explanatory notes are placed on the e-questionnaire below questions to increase their clarity.

Generally, the errors regarding to unclear, illegible questions and explanations were reduced by staff of field statistical office during the data collection period. Staff of statistical office obtained detailed methodological instructions on how to deal with such errors.

As for the respondent errors,  these errors are connected with misunderstandings of methodological notes and misinterpretation. Respondents sometimes give incomplete answers in case of time consuming questions. These type of errors are eliminated during the control phase in field statistical office. If errors are caused by averse attitudes of  respondents,  a survey objective  is  explained once again, with emphasizing the role of respondents and clearance of any doubts concerning a survey. 

As for the errors of information system, these types of errors are connected with wrong figures that are inserted to the computer system from the paper forms. There are very rare, because of using e-questionnaire. Thanks to arithmetical and  logical control implemented directly in e-questionnaire these errors are identified very quickly, explained and corrected.

As for the mode of data collection errors,   the methods of data collection in the LCS 2016  were based on  electronic forms filled in with the use of electronic portal. All  doubts regarding variables were explained during the completing e-questionnaire (metadata displayed) and during  phone calls of respondents with the appropriated staff of field statistical office and  by the  e-mail contact.

 As for errors deriving from data compiling and processing, during the phase control the e-questionnaire is tested and improved .

Non-response errors are divided into unit non-response and item non-response.

Non-response errors due to missing answers of sampled units for the LCS 2016  amounted  to 25%, of which due to refusals – 21.0%. More detailed classification of non-response units is following:

non- response units consist of  11962  units  i.e.  24.6%  of  the  selected sample by  reasons given below:

refusals: 10227    i.e. 21.0%  of the selected sample;

over-coverage: 867    i.e.  1.8% of the selected sample of which:

• dead   units: 510    i.e.  1.1% of the selected sample i.e. 48597;
• inactive units: 357    i.e.  0.7%  of  the selected sample;

others (lack of contact): 868    i.e.  1.8%  of the selected sample.

Non response errors due to missing items are marginal and they are eliminated by appliance of grossing-up correction.

Not applicable.

No imputation applied.

There are not errors connected with benchmarking procedures (preliminary data before grossing up are available six months after the reference year - in June, final data after grossing up  are available eight months after the reference year – in August).

 The assumptions  regarding  the  grossing up procedure are following.

 The basic parameter estimated in the  labour costs survey  2016  is  a sum of values of the surveyed variable X. The parameter  is  estimated by the following formula:

 (1)

where:

M_ki – weight assigned to the i-th unit in the k-th sub-collection,

X_ki – value of variable X  in the  i-th unit of the k-th sub-collection.

Generalization of the results according to the formula (1) would be correct if the survey was complete in each of the surveyed sub-collections. In the  labour costs survey 2016  non-responses were observed. They were mainly due to refusals and the so-called lack of contact with the sampled units.  Therefore, in sub-collections where significant number of non-responses was observed, weights M_ki were adjusted with the use of information concerning the number of the employed persons recorded in the sample frame for the surveyed units, as well as those that were  not surveyed for various reasons. The corrections were made separately in 6 classes of unit size according to the number of the employed persons recorded in the sample frame. The following classes were determined: up to 19 employed persons, 20 – 49, 50 – 99, 100 – 199, 200 – 499 and 500 and more employed persons.          

In the k-th sub-collection weights were corrected according to the formula:

(2)

while:

(3)

where:

P_1kj – the number of the employed  persons  in the surveyed units in the j-th

class of the k-th sub-collection,

P_2kj – the number of the employed persons in the units that were not  surveyed

due to refusals in the j-th class of the k-th sub-collection,

P_3kj – the number of the employed  persons in the units that were not  surveyed

due to lack of contact  in  the j-th class of the k-th sub-collection,

P_4kj – the number of the employed  persons in the units that were not  surveyed

due to lack of activity, liquidation or  bankruptcy of the unit  in  the j-th class of

the k-th sub-collection.

The formulas (2) – (3)  indicate that  weights correction covered non-responses caused by refusals and  in proportion to the number of refusals  those that were due to lack of  contact  with a sampled unit.

Values of the variable X for the k‑th sub-collection estimated with the appliance of weights  M_k or W_kj, were then  aggregated, according to the needs, into NACE divisions and sections, ownership sectors and for the whole national economy. Moreover, quotients of variables X/Y  e.g.  labour costs per one employee  were also estimated on different aggregation levels. 

            The next assumptions regarding  labour costs survey are as follows:

• Labour costs survey cover calendar year figures. Where is necessary we ask the reporting units to supply calendar year estimates.

• Small enterprises, in this case enterprises with less than 10 employees, have not been accounted for.

• No combinations between survey data and data  from administrative source have been done. That is, all data come  from the survey results.

Not applied.

The transmission includes final data.
Revisions take place when data are corrected due to mistakes or inconsistencies.

As regard 2016 LCS transmission there were 3 data corrections, due to lack of variable D5 which was  finally completed, and because of  slight technical issues.

We collected  D5 (subsidies) separately in the LCS 2016 but didn’t send the variable data to Eurostat at first because the validation program would deduct them from labour costs (which are already  reported without subsidies). Eventually we allocated subsidies to the direct remuneration costs (D1, D11, D111, D1111, D11111) according to the Eurostat’s recommendation.

First revision took place on 27.07.2018 and the last one on 22.11.2018. There were in total 3 revisions of the LCS dataset.

See below.

As for drawing the sample for the LCS 2016:
Sample was prepared in February 2017;

As for preparing the electronic questionnaire/ offline application:
The application was ready  in March 2017;

As  for preparing  methodological  guideline for statistical regional office:
Methodological guideline was prepared in March 2017;

As for data- collection dates:
Data were collected via electronic questionnaire from 24.03.2017 to 18.04.2017;

As for key dates for the post-collection phase:
data were checked and processed in the field statistical office in May 2017;

Preliminary data before  the  grossing  procedure were available in June 2017;

Final data after the grossing procedure were available in August 2017;

As for key publication dates: 
publication on labour costs in the Polish-English version was available on 13 November 2017.

Link to the publication:

http://stat.gov.pl/en/topics/labour-market/working-employed-wages-and-salaries-cost-of-labour/labour-costs-in-the-national-economy-in-2016,6,4.html

Not applicable.

317 days.

Data on the LCS 2016 were  transmitted to Eurostat  on 8^th of June 2018 according to  Commission Regulation 1735/2005.

The LCS 2016 data were  transmitted to Eurostat 22 days before the scheduled time.

The final transmission including corrections took place on 22.11.2018.

See below.

Survey on labour costs in 2016 allowed analysis of basic indicators by voivodships. The data have been compared between 16 voivodships. There are no discrepancies and significant differences between the figures.

Detailed information on the comparison between the voivodships is included in publication “Labour costs in the national economy in 2016” on pages 25-28 and 135-143 (table 15). Link to the publication:

http://stat.gov.pl/en/topics/labour-market/working-employed-wages-and-salaries-cost-of-labour/labour-costs-in-the-national-economy-in-2016,6,4.html

Not available.

Beginning from LCS 1996 Poland has been conducting LCS every 4 years. In periods between the surveys, yearly assessments of basic ratios of labour costs are made. Thanks to such solution  we are able to conduct comparability over time with regard to LCS results. Definitions of statistical units, reference population and variables are based on Eurostat  recommendation and that is why the results of the LCS are comparable on international scale.

During the period 1996-2012 the coverage of units has been changed, namely the LCS data for years 1996-1999 covered units with more than 5 employed persons and since 2000 the LCS covers units with more than 9 employed persons .  There also had been changes in the past concerning distribution of social security contributions, namely up to 1998 all social security contributions were paid by employers, beginning from 1999 social security contributions  have been paid partly by employers and employees.

The comparison of  the average monthly labour cost per 1 employee  in years 2012-2016 in different kinds of activities is available in publication “Labour costs in national economy in 2016” on website:

http://stat.gov.pl/en/topics/labour-market/working-employed-wages-and-salaries-cost-of-labour/labour-costs-in-the-national-economy-in-2016,6,4.html

4 years.

It should be mentioned that annual labour costs estimations are carried out in between the LCS years.

1. The LCS coherence with the Labour Force Survey

See Annex "Coherence with LFS"

2. The LCS coherence with the Structural Business Statistics

See Annex "Coherence with SBS"

3. The LCS coherence with the Labour Cost Index

See Annex "Coherence with LCI"

The information is included in annexes in p.8.3.

See Annex "Coherence with National Accounts ".

Internal coherence is kept in the extent to which representative survey allows it. In each survey different units can be draw and it results in various data observed within the same survey or in a comparison with the previous surveys.

Information on precision is available in Annex 1 of this report.

Data are available in the form of tables, graphs, methodological description, main characteristics of basic indicators.

Not applicable.

Data are well documented in the form of:

Publications – the recent LCS publication (Polish-English version) consists of 144 pages, it covers methodological note, characteristics of basic measures of labour costs, information on sampling method, tables, graphs; the LCS publications are  disseminated every 4 years in form of paper, CD-ROM (previous ones) and Internet at:

http://stat.gov.pl/en/topics/labour-market/working-employed-wages-and-salaries-cost-of-labour/labour-costs-in-the-national-economy-in-2016,6,4.html

Information service of the Department of Education and Communication – distribution data on labour costs for national and external users;

Chapters on LCS in Yearbooks: Statistical Yearbook of the Republic of Poland, Concise Statistical Yearbook of Poland, Statistical Yearbook of the Regions – Poland,  Statistical Yearbook of Industry – Poland, Yearbook of Labour Statistics (published  every 2 years).

Currently there is no Polish online dabase with data from the LCS.

Not applicable.

Micro-data are not disseminated.

Not applicable.

Methodological chapter is included in the publication "Labour costs ..." described above in this quality report.

There is also publication "Methodological report. Statistics on labour market, wages and salaries" available on Statistics Poland website at:

http://stat.gov.pl/obszary-tematyczne/rynek-pracy/zasady-metodyczne-rocznik-pracy/zeszyt-metodologiczny-statystyka-rynku-pracy-i-wynagrodzen,1,2.html#

Information on quality management in Polish statistics is available on Statistics Poland website at: http://bip.stat.gov.pl/en/activity-of-official-statistics/quality-in-statistics/ The quality policy is specified in the mentioned study Vademecum of Quality in official statistics. The study is available at: http://bip.stat.gov.pl/files/gfx/bip/en/defaultstronaopisowa/55/1/1/poz_vademecum_zew_ang.pdf

Not applicable.

Not applicable.

The cost of the survey borne by Statistics Poland was 310000 PLN.

Respondent’s burden with preparing data and filling in the questionnaire amounted to 118544 hours in total.

See below.

According to the Polish Law on Public Statistics identifiable unit data collected in statistical surveys are subject to absolute protection. These data can be used only for studies, compilations and statistical analyzes and for the aim of preparing random sample of a statistical survey; making available or using these data for purposes other than those specified above is prohibited (statistical confidentiality).

Confidential  data in the transmitted dataset to Eurostat are flagged when the number of units is tiny for an individual record.

The report includes  Annex 1:

Annex1_Estimates on relative standard errors of estimators_LCS2016