Structural Earnings Survey 2006 of Spain Quality Report

Transcription

1 Structural Earnngs Survey 2006 of Span Qualty Report Labour Market Statstcs Drectorate Natonal Statstcal Insttute of Span December 2008

2 (I) Qualty report on the Structural Earnngs Survey 0. Introducton 1. Relevance 2. Accuracy 2.1. Samplng errors 2.2. Non-samplng errors overage errors Measurement and processng errors Non-response errors Model assumpton errors 3. Punctualty and tmelness 3.1. Punctualty 3.2. Tmelness 4. Accessblty and clarty 5. omparablty 5.1. Geographcal comparablty 5.2. omparablty over tme 6. oherence 0 Introducton 2

3 In Span, three Structural Earnngs Surveys were undertaken, all n collaboraton wth the Statstcal Offce of the European ommuntes (Eurostat). The frst was the Structural Earnngs Survey for the perod referrng to 1995, whch covered unts wth ten or more employees n the actvtes of ndustry, buldng, commerce, hotels and restaurants, transport, communcatons, fnance nsttutons and nsurance. The second, referred to 2002, broadened the coverage to nclude the actvtes outlned n sectons M, N and O of NAE Rev.1. The thrd and last survey, wth 2006 as a reference year, has as a man characterstc to nclude the small unts (those wth less than 10 employees) n the same actvtes than n The ommunty Regulatons used as the bases for producng the last survey were as follows: - ouncl Regulaton (E) No. 530/99 of 9 March 1999 concernng structural statstcs on earnngs and labour costs. - ommsson Regulaton (E) No. 1738/2005 of 21 October 2005 amendng Regulaton (E) No. 1916/2000 as regards the defnton and transmsson of nformaton on the structure of earnngs. - ommsson Regulaton (E) No. 698/2006 of 5 May 2006 Implementng ouncl Regulaton (E) No. 530/99 as regards qualty evaluaton of structural statstcs on labour costs and earnngs. The am of ths document s to be used to evaluate the qualty of the survey. The structure of ths report follows the content of ommsson Regulaton (E) No. 698/2006 of 5 May 2006 Implementng ouncl Regulaton (E) No. 530/99 as regards qualty evaluaton of structural statstcs on labour costs and earnngs. 1 Relevance The man users could be classfed n the followng groups: - Internatonal Organsatons: European Unon Insttutons, OED, Internatonal Monetary Fund, Internatonal Labour Organsaton, etc. - Publc Organsms: dfferent Mnstres such as the Mnstry of Economy, the Mnstry of Labour and Socal Affars, etc.; the Natonal Statstcal Insttute tself for several of ts unts, such as Natonal Accounts; the Bank of Span; Regonal Insttutons, etc. - Socal Insttutons such as trade unons, employers organsatons, poltcal partes,... - Research entres and Unverstes - The meda No survey has been carred out among users to know ther needs of nformaton and whether they are satsfed or not wth the publshed results. Ths may be accounted for by the lack of contact wth most users snce the remttance of results s often mpersonal, and by the fact that the nformaton s looked for n INTERNET. It s known only the opnon of users who have receved nformaton on request or who have asked for methodologcal detals. In general, these users are satsfed. Nevertheless they consder that the survey should offer more detaled breakdown of some varables (n partcular of regons, branch of actvty and wage components) and also nclude more varables related to the employee (famly stuaton, etc). 3

4 Moreover, the natonal publcaton was avalable on 5 November 2008 so, there s not much tme to know the users' opnon. 2 Accuracy 2.1 Samplng errors The estmators used for the survey are separate rato estmators, the number of employees n the regster beng used as an auxlary varable. The estmators for economc data of the employee j n the unt classfed n the actvty r, sze h and regon t are formed. The grossng up factors of frst and second stage are respectvely: N rth D = 1 F1 j = and n D rth = 1 B F 2 j = b where, D s the number of employees n the regster for the unt, B s the number of employees regstered durng the whole month of October 2006 n the Socal Securty and b s the number of employees n the sample. Thus: GH = X Y and GT = X Z are the hourly earnngs and the earnngs per employee n any cell of the table (by actvty, occupaton, sex and regons) Beng: X = j F * j F * 1 2 j X j Total earnngs (monthly or annual) Y = j F * j F * 1 2 j Y j Hours Z = j F j F 2 j 1 * Employees 4

5 j makes reference to the employees ncluded n the cell. The coeffcent s defned: ( ) ε ( X ) = * 100 where V X X V ( ) = ( X ) X V, h X = h j h, F F * * 1 j 2 j X j and (h s referred to the cross of varables regon, actvty and sze) N h ( N h nh ) V ( X ) = * h n h n = 1 h ( X R n h h 1 * D ) 2 N + n h h n h B * ( B b ) * S b = 1 2 beng - X b B = * b X j j= 1 ; where X j =0 f j h = 1 - R = h nh D n = 1 X b X j= 1 b j= S = b 1 j b X j 2 - s of gross earnngs n the reference month 5

6 Table 1. Numerator Denomnator Total employees , ,06 0,45 Full tme employees (FT). Total , ,51 0,48 FT Male , ,74 0,62 FT Female , ,77 0,74 Part tme employees , ,55 0,98 Table 2. Secton Numerator Denomnator Total , ,06 0, , ,56 1,35 D , ,66 0,46 E , ,69 1,49 F , ,23 1,77 G , ,73 1,34 H , ,04 1,37 I , ,88 2,67 J , ,24 1,31 K , ,05 1,31 M , ,39 1,28 N , ,42 0,90 O , ,18 1,53 Table 3. ISO Numerator Denomnator Total , ,06 0, , ,18 1, , ,58 0, , ,03 1, , ,02 0, , ,06 1, , ,90 5, , ,91 1, , ,26 0, , ,12 1,00 6

7 Table 4. AGE Numerator Denomnator Total , ,06 0,45 Under , ,34 1, , ,64 1, , ,73 0, , ,89 0, , ,59 1,10 60 and over , ,86 1,58 Table 5. NUTS1 Numerator Denomnator Total , ,06 0,45 ES , ,60 0,79 ES , ,77 0,65 ES , ,37 1,40 ES , ,99 0,68 ES , ,35 0,89 ES , ,20 0,90 ES , ,79 1,32 Table 6. ISED Numerator Denomnator Total , ,06 0, , ,01 1, , ,11 0, , ,83 1, , ,71 1, , ,66 1, , ,75 3,59 Table 7. SIZE Numerator Denomnator Total , ,06 0,45 E1_ , ,70 1,56 E10_ , ,60 0,65 E50_ , ,10 0,70 E250_ , ,60 0,92 E500_ , ,80 1,18 E , ,20 0,89 7

8 - s of average gross hourly earnngs n the reference month Table 1. Numerator Denomnator Total employees , ,76 0,44 Full tme employees (FT). Total , ,70 0,48 FT Male , ,28 0,62 FT Female , ,42 0,73 Part tme employees , ,06 1,02 Table 2. Secton Numerator Denomnator Total , ,76 0, , ,95 1,35 D , ,82 0,46 E 18263, ,82 1,48 F , ,66 1,74 G , ,27 1,29 H 84552, ,55 1,36 I , ,21 2,60 J 87746, ,58 1,31 K , ,88 1,31 M , ,99 1,41 N , ,88 0,94 O 79412, ,14 1,44 Table 3. ISO Numerator Denomnator Total , ,76 0, , ,65 1, , ,16 1, , ,23 1, , ,39 0, , ,57 1, , ,27 5, , ,75 1, , ,61 0, , ,13 0,95 8

9 Table 4. AGE Numerator Denomnator Total , ,76 0,44 Under , ,18 1, , ,38 1, , ,92 0, , ,94 0, , ,41 1,05 60 and over , ,93 1,89 Table 5. NUTS1 Numerator Denomnator Total , ,76 0,44 ES , ,68 0,78 ES , ,16 0,64 ES , ,65 1,37 ES , ,20 0,77 ES , ,24 0,85 ES , ,78 0,88 ES , ,05 1,33 Table 6. ISED Numerator Denomnator Total , ,76 0, , ,67 1, , ,08 0, , ,54 1, , ,12 1, , ,79 1, , ,57 3,60 Table 7. SIZE Numerator Denomnator Total , ,76 0,44 E1_ , ,54 1,44 E10_ , ,37 0,62 E50_ , ,56 0,66 E250_ , ,43 0,87 E500_ , ,81 1,09 E , ,04 0,83 9

10 The general rule to publsh a cell n a multdmensonal table used n the natonal publcaton s that, at least, 100 observatons support the estmaton. All fgures estmated wth less than 100 observatons have been erased; cells estmated wth a number of observatons between 100 to 500 have been marked to ndcate ths stuaton. 2.2 Non-samplng errors overage errors The framework of the survey s obtaned from the Socal Securty General Regster of ontrbutons Accounts. Employers that hre employees for the frst tme should request ther own regstraton as a company, at the Socal Securty General Treasury. They should do ths before commencng work actvtes. Regstraton s an admnstratve act by whch the Socal Securty General Treasury gves an dentfcaton and control number to the employer. Ths establshes what s known as the Account of Socal Securty ontrbuton. The lst of Accounts s used as a busness regster n all the Labour ost and Structure of Earnngs Surveys performed. The procedure for random selecton of unts corresponds to a stratfed samplng wth optmal allocaton, n whch the samplng unts are the accounts. The stratfcaton crteron s accomplshed usng three varables: Autonomous ommunty, economc actvty (NAE rev.1 from to O except L) and unt sze (n terms of number of employees). When the Regster s receved from the Socal Securty, a frst debuggng s made pror to the selecton of the sample, whch mples several stages: To elmnate economc actvtes regardng agrcultural actvtes, lvestock, fshery, publc admnstraton, defence, households wth domestc employees and extra-terrtoral organsms snce these are not part of the survey. To elmnate the unts that belong to the specal regme of Socal Securty sales agents, whose man compensaton conssts n commssons on sales and who, consequently cannot be surveyed ether. Afterwards, the sample s drawn and the lst of numbers of accounts of socal securty contrbutons selected s sent to the Socal Securty General Treasury agan. It provdes the lst of all the employees, dentfed by ther afflaton numbers, ncluded n these unts durng the reference year. Specfcally, the employees to be targeted for the SES are those employed n the observaton unt n the reference month. A smple random sample of employees s taken wthn each of the selected local unts accordng to the sze of the unt. In SES-2006 the number of employees selected was: - All employees n unts wth 1-4 employees - 4 employees n unts wth 5-9 employees - 5 employees n unts wth employees - 7 employees n unts wth employees - 10 employees n unts wth employees - 16 employees n unts wth employees - 22 employees n unts wth employees - 25 employees n unts wth 500 or more employees 10

11 - There were 30 local unts wth 50 employees selected due to ther specal locaton. The advantage of ths method s that the respondent does not choose the employees, because the employees to be ncluded n the questonnare are dentfed by ther afflaton number. The only dfference between the reference populaton and the study populaton s that the frst does not nclude the apprentces. The labour legslaton on apprentces n Span establshes very low labour costs (both wages and socal contrbutons). As a consequence, the number of apprentces s very small. Thus, at the end of October of 2006 the number of apprentces was from a total of employees n the actvtes ncluded n the survey (t represents 0,7% of the total). Moreover, the problem wth apprentces s that, due to ther partcular type of contract, the Socal Securty General Treasury regsters them, for control, n a dfferent afflaton fle, wth dfferent characterstcs, that make dffcult to use t jontly wth the general fle. On the other hand, because of t s such a small group, a random selecton does not assure to obtan representatve separate fgures for ths collectve. Ths fact makes necessary a great effort carryng out a specfc survey for the apprentces to assure the results. Ths effort s not corresponded wth the small fgures obtaned as was showed n the experence from the 2002 SES and 2004 LS. On the other hand, a sgnfcant proporton (one thrd approxmately) of apprentces estmated by the 2004 LS survey were, n practce, scholarshp employees, crcumstance not known before selectng the sample, and ther ncluson n the fnal fgures dstorts the apprentces fgures. onsequently the apprentces are not ncluded n the SES Once the questonnares are sent to the selected unts, the data collecton and debuggng reveal the errors n the surveyed unts. The sample was composed by selected unts unts were surveyed: unts answered and 2216 not. Data collecton showed that 725 unts were not located, 57 unts were nactve or closed down n 2006 and 47 unts were erroneously classfed Measurement errors and processng errors Before sendng the questonnares to the unts, the telephone numbers and addresses for the unts were checked and updated. The flled questonnares were gven back to the statstcal offce by mal, n the enclosed postage pad envelope, or electroncally, to whch purpose a regstraton and transmsson format on Internet was desgned. Ths tme was also possble to fll n the questonnare by Internet usng an dentfcaton number provded n the questonnare. Debuggng errors After recevng the questonnares, the statstcal offce recorded them, usng an ad hoc computer applcaton, whch at the same tme made a frst debuggng for the questonnare s nternal consstency. 11

12 Ths frst debuggng conssts n usng flters that allow to separate vald questonnares from those wth nconsstences to be revsed. The flters are of two knds: those detectng type I and type II errors. Type I errors: If they are not thoroughly corrected, the questonnare cannot be consdered as vald. Type II errors: They nvolve norms regardng the coherence of the data. The non-satsfacton of these norms does not necessarly mean that the questonnare s not vald, but the apparent ncoherence must be explaned. In case of doubt, a telephone call s made to the respondent to elucdate the queston. The questonnares are fltered a frst tme durng the recordng and a second tme by the team responsble for the results of the survey (ths team s dfferent from the recordng one). There are more than four hundred rules checked n each employee data. They assure: - not mssng data (partal non-response s not allowed) - coherence among ndvdual characterstcs: age, length of servce, level of educaton, type of contract, occupaton, and so on. - coherence among economc data: monthly and annual earnngs, between themselves, and both related to the hours pad, to the economc actvtes, occupatons, etc. - the codes assgned for the level of educaton and the occupaton exst n the classfcaton used (ISED-97 and ISO-88) and are coherent wth the varables, as economc actvty, age, etc. The varables level of educaton and occupaton were codfed at the tme of recordng the questonnares. Rules to assure that the code assgned exsts n the classfcaton were establshed. Moreover, the sample was dvded n portons. Random subsamples were selected from each porton and the codfcaton n t was revsed. If the errors n the codfcaton were hgher than the 3% of the total number of employees n the subsample, the whole porton was recodfed. Ths process was repeated untl ths percentage of errors was acheved. The processng, grossng up and tabulaton of the data have been programmed and supervsed by two dfferent teams. After the tabulaton, the results obtaned were analysed n order to know whether they were coherent wth the avalable short-term statstcs on labour and wage costs Non-response errors The followng tables show the response rates by economc actvty and Nuts. 12

13 Table 1. Response rate by secton of NAE Rev.1 Unts Employees Secton Sample collected Response rate Sample collected Response rate , ,5 D , ,3 E , ,3 F , ,9 G , ,0 H , ,5 I , ,4 J , ,1 K , ,8 M , ,2 N , ,6 O , ,1 Total , ,0 Table 2. Response rate by NUT1 Unts Employees Nuts Sample collected Response rate Sample collected Response rate ES , ,9 ES , ,1 ES , ,8 ES , ,0 ES , ,7 ES , ,8 ES , ,7 Total , ,0 As t s sad above, partal non-response s not allowed. When there was no response or an ncdence n the sample, the value of the analyss varables for each empty samplng unt or unt wthout nformaton was mputed usng the nformaton obtaned for the stratum to whch the unt belonged. Ths form of mputaton only requres replacng the rasng factors obtaned wth the selected sample wth the ones that result from the effectve sample. There s only one excepton n the tem non-response: the level of educaton of the employee. It has been the most dffcult varable to obtan, manly n large unts or groups of unts, wth a great number of employees selected n occupatons where the level of educaton s not an mportant requrement n the job (eg.: Major Group 9 of ISO-88). Most of ths knd of unts needed a lot of tme to answer the questonnare, because they had to ask to the employees ther level of educaton. At the end, ths tem was empty for employees (4,8% of the sample). The level of educaton of these employees was mputed usng the software applcaton IVEware (Imputaton and Varance estmaton Software). Ths software performs mputatons of mssng values usng the Sequental Regresson Imputaton 13

14 Method 1. Ths method has two man advantages: t takes nto account the structure of correlatons of the whole set of varables n the sample and t s bult on the SAS Macro Language, that s the software used for the rest of processes. The effect n the fnal results of the mputaton s neglgble due to the small number of mssng values mputed Model assumpton errors - to ensure that a representatve month s selected: The monthly questons have as reference October 2006 n the questonnare. It s not possble to answer for other month. Practces n the companes n Span suggest usng October because t s a month wthout seasonal payments and absences (lke hrstmas pay or summer holdays). On the other hand, October was the month used n the prevous SES, so t s the most sutable month to keep comparablty over tme. - to adjust the accountng or fscal year to the calendar year The accountng or fscal year concdes wth the calendar year n Span. - to ensure that NAE Rev.1 sectons are fully covered The regster used to select the sample has the economc actvty as a varable of classfcaton. The desgn of the sample takes nto account ths varable n the stratfcaton process jontly wth the unt sze and the regon. 3 Punctualty and tmelness 3.1 Punctualty Key data collecton dates: The feldwork took place between May - October The stages of the collecton perod are the followng: Remttance of the materal to the respondent unts. Ths frst stage took 3 days. In general terms, each malng contans the followng documents: - A questonnare that must be remtted n a delay not surpassng 20 days after t s receved. - A lst wth the number of Socal Securty of the employees selected n the unt. - A letter from the General Drector ndcatng the purpose of the survey and nformng on the laws that oblge to complete the questonnares and on those regardng Statstcal onfdentalty. - A postage pad envelope bearng the address where the respondent has to send the flled n questonnare. 1 Ths method s descrbed n the artcle "A multvarate technque for multply mputng mssng values usng a sequence of regresson models" by Raghunthan, Lepkowsk, Van Hoewyk and Solenberger (Survey Methodology, June 2001). 14

15 Locaton: The length of ths stage depends on the number of respondent unts to be located and lasted about one or two weeks. The work conssted n fndng the telephone numbers and/or the addresses of those unts for whch no contact telephone number was avalable or whose envelopes wth the documents were returned. The most frequent steps to locate a unt were the followng: - To phone or e-mal the nformaton servces of the telephone frms - To consult telephone drectores: Whte Pages and Yellow Pages (manual search or electronc search ) - To contact muncpaltes (Tax Department) - To contact Socal Securty Treasure - To contact enterprses of the same sector n the same muncpalty - To contact Tax Agences - To search INTERNET: Whte Pages, Yellow Pages, amerdata ontacts and clams: Ths stage s essental for a fluent and effcent collecton, to get a hgh percentage of success. At ths pont, the calls to enterprses were started and the questonnares clamed. The most useful tool for ths actvty was the telephone. Telephone contacts may occur n both drectons. To foster the respondents wllngness to call the NSI, they are provded, whenever possble (n some of the documents forwarded to them), wth a free telephone number. The calls are preferably answered by the ntervewer n charge of obtanng ther questonnares. If ths s not possble, any person tasked wth the collecton wll resolve the respondent s doubt or duly take the message (ndcatng the enterprse s Natonal Regster Number, ts address, name of the person who calls, contact telephone number, dentfcaton number n the survey and other comments). There s also a free fax number to receve questonnares and wrtten communcatons. Ths stage lasted approxmately two or three weeks and each ntervewer must contact the enterprses assgned to hm and request ther questonnare. lam wth acknowledgement of recept: All the respondent unts whch had not remtted ther flled n questonnare by the end of the above stage, receved by regstered mal and wth an acknowledgement of recept, a second questonnare wth the menton lam of compulsory statstcal data (document PS2) Locaton of non-found unts: At the end of the frst round of calls (to all the unts n the survey), t turned out that a percentage of them could not be touched. They all belonged to a specal queue of unts: the QUEUE OF NON TOUHED UNITS. Despte the mplementaton of all the avalable means, t was mpossble to touch some of them. However, the above mentoned telephone nformaton web pages contnued to be looked through. Stage followng the collecton of questonnares One of the ntervewer s tasks s the recordng and debuggng of all the ncomng questonnares. 15

16 The general rule s that the questonnares must be recorded at the latest from 3 to 5 days followng ther arrval, to facltate consultatons wth the enterprse as soon as possble after they were flled n. At the outset of the collecton perod, the locaton and frst contact wth respondent unts have the prorty over the recordng and debuggng of questonnares. After recordng the questonnares, they pass to the codfcaton team. The occupaton and the educaton level are codfed at ths pont. To elmnate the errors, all computer applcatons classfy the errors n two large blocks: type I or bg errors and type II or small errors. Type I errors are so mportant that they nvaldate the questonnare. Type II errors may arse from specfc crcumstances of the enterprse s actvty, from ts actvty durng the data reference perod or from any specfc event of the respondent unt. The frst debuggng should be carred out at the latest from 8 to 10 days after the recordng, that s, days after the questonnares are receved. The recordng and the frst debuggng stage fnshed by the end of October The second debuggng of all the unts was completed n March The tabulaton was prepared at the end of May and n June. The frst remttance of data to Eurostat was done at the 11 July 2007; several errors were found and corrected so that the frst of August the fnal verson was sent. Publcaton dates: On 5 November 2007, the detaled results were dssemnated. Ths publcaton s composed by a document wth the comment of the man results and large set of tables. The nformaton s avalable on INTERNET and on electronc support at request. 3.2 Tmelness The Structural Earnngs Surveys are publshed t+23 months after the reference year. 4 Accessblty and clarty The tables, the document on the results and the methodologcal document are avalable for free on the INE web ste. We are workng now n the desgn of a standard anonymsed fle usng a smlar methodology as presented n the LAMAS Workng Group on March Moreover, t s possble to prepare customsed anonymous survey fles studyng the varables requested and also, based on the basc statstcal operatons fles, crosses other than those publshed may be carred out between varables, accordng to the needs of the user. The release was sent to the man offcal users. The results are not remtted to the respondents. 5 omparablty 5.1 Spatal comparablty 16

17 There are no dfferences between natonal and European concepts regardng statstcal unts, defnton of varables and classfcatons. The only dfference n the reference populaton s that apprentces are not ncluded n SES 2006 as explaned above. Most of the effort made by the unt responsble for Labour ost Statstcs went on the detaled study of the varables contaned n ommsson Regulaton No. 1738/2005 and ts comparson wth labour laws and forms of retrbuton n force n Span n the year From ths comparatve study we obtaned a verson of the questonnare adapted to the realty of the country, whch allowed us to obtan the varables as defned n the above-mentoned regulaton. 5.2 omparablty over tme Snce the frst Structural Earnngs Survey was conducted the coverage of the followng surveys has been extended ncludng dfferent groups of unts. Thus, n frst SES 1995 unts wth ten or more employees n the actvtes of ndustry, buldng, commerce, hotels and restaurants, transport, communcatons, fnance nsttutons and nsurance were ncluded. The second, whch referred to the 2002, broadened the coverage to nclude the actvtes outlned n sectons M, N and O of NAE Rev.1. The thrd and the last survey, wth 2006 as a reference year, has as a man characterstc to nclude the small unts (those wth less than 10 employees) n the same actvtes than n As a consequence of the ncluson of the small unts n SES 2006, there s a decrease of the average earnngs compared wth the general SES 2002 results. It s necessary to elmnate de sze 1-9 employees from SES 2006 to compare homogeneous results wth SES oherence Accordng wth the Regulaton a comparson should be made between the varable gross annual earnngs n the reference year, expressed per employee, and the varable wages and salares, per employee, of the Natonal Accounts. The fgures avalable from Natonal Accounts correspond to the 2006 provsonal data about ompensaton of employees and to the full-tme equvalent employees from the Seres of Accounts Base year 2000 publshed n December Wages and salares are not avalable for Data on gross annual earnngs per employee from SES and compensaton per employee from Natonal Accounts are compared n the followng table: 17

18 Gross annual earnngs per employee n SES Euros ompensaton per employee from Natonal Accounts. Year 2006 (P). Euros Dfference n % Seccón 26003, ,1-28,3 D 23267, ,5-20,8 E 34340, ,3-28,0 F 18706, ,5-31,9 G 18765, ,1-15,5 H 14912, ,9-44,9 I 23710, ,2-19,3 J 39512, ,5-30,3 K 20110, ,8-41,6 M 22456, ,9-40,1 N 23957, ,3-31,5 O 19783, ,3-19,7 Total 21428, ,9-28,3 The fgures from Natonal Accounts are greater than those from SES n all sectons (between 20 and 40 %), because the varable compensaton of employees ncludes the employers socal contrbutons. Takng nto account that the employers socal contrbutons has a weght n the total labour costs that vares between 25%-35%, dependng on the economc actvty, the coherence between both sources seems to be acheved. 18