Quality and accuracy of business register and links to administrative sources. Applicability of the estimation model for enterprises below thresholds

Transcription

1 Austrian Report about the quality of the Structural Business Statistics according to the Commission regulation No. 1618/1999 for the reference year Quality indicators (article 5 of the regulation) a. Coefficients of variation In Austria since the reference year 2002 SBS for sections C-K of NACE Rev. 1.1 (except divisions 65 and 66, for which exclusively administrative sources from supervisory authorities have been used) is compiled by using a yearly cut off survey in combination with the use of administrative sources and model based estimation. Structural Business Statistics in Austria is comparable with a full census, because data for every single enterprise in the population is surveyed, estimated or available from administrative sources. The only restriction is that administrative data from the reporting year were not available for all enterprises at the time of calculation. Due to this strategy for compiling SBS results no coefficient of variation can be calculated. In general the quality of main data as well as the quality of regional data has improved. Nevertheless, different quality factors have to be taken into account for quality evaluation. Coverage of the SBS survey Quality and accuracy of business register and links to administrative sources Validity of estimation model Applicability of the estimation model for enterprises below thresholds Stability of variables across time b. Unit non-response The term Unit non response subsumes all enterprises above the defined thresholds which did not fulfil their obligation to report after different actions of demanding the information were taken from Statistics Austria (also including reminders as well as legal steps). In general, unit non response is imputed by using the estimation model for enterprises below thresholds. In contrast to the model based estimation for enterprises below thresholds higher turnover thresholds (maximum EUR 50 millions) were applied for calculation of model parameters for unit non response. If survey data of the previous year for the enterprises in question were available, the detailed breakdown of variables of these enterprises was taken into account. Information from STS in Industry and Construction (sections C to F of NACE Rev. 1.1) is also used for unit imputation. For large enterprises for which additional information was available these information (e.g. financial statements from statistics of public limited companies) was taken into consideration. For the reference year 2007 the unit non response rate was 5.0%. c. Item non-response Item non response refers to missing items of individual enterprises which did not fulfil their obligation to report despite intensive reminders. Because of differences in the structure of economic activities item non response is treated very individually. No general statistical imputation model (e.g. Nearest Neighbourhood) has been used. Item non response is imputed, if different main items for calculation of macroeconomic variables (e.g. purchases of goods for resale, purchases of services for resale and other operating expenditures) are not available. 1

2 The missing items where calculated as follows: Estimation based on extrapolated ratios (by using turnover from the reference year) of survey data of the previous year Activity based ratios Using data from STS concerning Industry and Construction Use of administrative sources (Social Security Authorities, Tax Authorities) Other information (Financial Statements if available) For items of less importance no imputation is made. Item non response is calculated according to the Commission Regulation on quality. For the number of employees administrative data from the Social Security Authorities is available exhaustively, hence this value is zero. 2. Survey Strategy (article 6 (1) of the regulation) In Austria SBS surveys are carried out since the reference year To introduce the ESS a full census was carried out for the reference year From 1997 to 2001 sample surveys with random sampling and free grossing up were applied. The concept from the reference year 2002 onwards refers to a cut-off survey in line with a model based estimation of enterprises below thresholds. a. Threshold values and sample percentages: Industry and Construction In principle account must be taken of all enterprises employing at least 20 persons. Moreover the turnover of all enterprises must be recorded with sufficient accuracy for each ÖNACE 2003 (the Austrian version of the NACE Rev. 1.1)-division. Therefore the survey method had to be designed to facilitate the collection of data from enterprises representing at least 90% of national turnover per NACE Rev. 1.1 division of Industry and Construction (degree of representativeness). NACE-divisions with a coverage less than 90% of national turnover require to include also enterprises step by step with 19, 18 persons employed until the 90% criterion is met. However, this strategy is restricted insofar, as enterprises with less than 10 persons employed must not be surveyed according to national legal acts; therefore the criteria of representation (90% of the total turnover of every single occupied ÖNACE 2003 division) cannot always be reached. Services sectors (Sections G-K of NACE Rev. 1.1, excluded Divisions 65 and 66) Section G (Wholesale and retail trade ), Group 633 (Activities of travel agencies and tour operators), Group 634 (activities of other transport agencies): yearly turnover of EUR 1.5 Mio. Other Service sectors: yearly turnover of EUR 750 THSD. Approximately units of Industry and Construction and around units of the service sector (sections C-K, excluded Divisions 65 and 66) were included in the primary survey. In total, enterprises above the threshold values (around 12,4% of the total population) are covering approximately 78% of employees and 89% of total turnover. The coverage rate for key data by divisions of NACE Rev. 1.1 can be seen from Annex to this quality report. The population for selection of enterprises for the SBS survey 2007 consists of all enterprises of the NACE-Sections C-K (except Divisions 65 and 66), which were active in the Business Register of Statistics Austria at the end of the year Annex 6 (Division 65) of the SBS-regulation is fulfilled by using data from the Austrian Central Bank, Annexes 5 and 7 (Division 66) by using data from the Austrian Financial Market Supervisory Authority. These areas were not treated within this quality report. 2

3 b. Administrative sources used The following administrative sources are used: Source Social Security Authorities Enterprises above thresholds In combination with the survey Enterprises below thresholds Single source, basis for model based estimation of other variables Information Number employees of Variables Employees from social security authorities can be defined according to SBS requirements Tax Authorities In combination with the survey Single source, basis for model based estimation of other variables Turnover from VAT declarations to tax authorities Fiscal turnover has to be taken instead of turnover from company accounts for enterprises below thresholds. Data from Social Security Authorities and Tax Authorities are transmitted to Statistics Austria on regular intervals and linked with the enterprises in the Business register on a single enterprise basis. c. Model based estimation for enterprises below thresholds Enterprises below thresholds were estimated on a single enterprise basis by using all enterprises of business register (sections C-K of ÖNACE 2003, except divisions 65 und 66) and information from administrative sources with regard to the following key data for the estimation of other variables: number of employees and total turnover. In Industry and Construction information from STS was also used. Using these key data a combined (robust) Regression model 1 was implemented for the calculation of main variables (personnel expenditures, purchases of goods and services ). For detailed variables, like breakdown of turnover or breakdown of purchases of goods and services, a ratio estimation method has been conducted. The model parameters were estimated in accordance to the primary collected enterprises of the relevant reference year and applied to the registered enterprises below the defined threshold values. For the model basis construction a step by step approach based on turnover sizes and economic activities has been used. Generally, the estimation of the model parameters was based on the most detailed ÖNACE 2003 classification level (nationally 6-digits). If there were not enough enterprises in the survey (15 enterprises for regression and 10 enterprises for ratio estimation) the size class was raised step by step by EUR 1 Mio., starting with a turnover threshold of EUR 999 THSD. If there are not enough enterprises with a turnover of less than EUR 5 Mio in a relevant ÖNACE 6-digit-level, the parameters were estimated on a higher ÖNACE aggregate level. Adaptations of the estimation model by expert rating for investments and for inhomogeneous economic activities were made to increase the quality of results. 3. Principal Activity (article 6 (2) of the regulation): Each statistical unit is classified regarding to a subclass (6-digit level) of ÖNACE ÖNACE 2003 is strictly compatible with NACE Rev. 1.1 on the 4-digit-level. 1 Ordinary-Least-Square-Regression (OLS) based on Least-Trimmed-Squares-Regression (FAST-LTS algorithm., Rousseeuw and Van Driessen, 1999). 3

4 The principal activity of a unit is the activity which contributes the largest share to the value added at factor cost. If information on value added is not available, turnover or employment is used as an approximation. If there is no main activity that exceeds the 50% rule, than the Top-down-Method is used. The main activity of newly born enterprises is taken from administrative sources and checked in the course of classification-notifications, for which Statistics Austria is obliged due to the Federal Statistics Act (so called Bundesstatistikgesetz 2000 ). The classification-notification contains information to all enterprises about their allocation to an economic activity of Austrian Statistical Classification of Economic Activities (ÖNACE 2003). The enterprises have the right to apply for a change of their ÖNACE 2003 code. In general the principal activity, but also secondary and ancillary activities are recorded in the Business Register Unternehmensregister of Statistics Austria. The principal activity can change whenever new information is available but in general the principal activity should be held stable in the Business register within a calendar year with the exception of the restructuring of an observation unit. Within a calendar year it should not be changed for the reason of continuity and coherence with short-term statistics. Enterprises which are in the cut-off survey for SBS in services are asked for their principal activity. In general for about 3% of enterprises the principal activity needs to be changed. The register is updated with activities changed in the course of surveys. In Industry and Construction information about principal and secondary activities of enterprises in sections C-F of ÖNACE 2003 are emanated from short term statistics. For enterprises below thresholds the ÖNACE-Code of Business Register is taken. Annex: Coverage rates of key data by divisions of NACE Rev

5 Annex Coverage rates of key data by divisions of NACE Rev. 1.1 for 2007 Number of enterprises Total Coverage of enterprises Number of employees Total Coverage of enterprises Total Turnover in EUR Coverage of enterprises Divisions of NACE Rev. 1.1 Number above thresholds Number above thresholds Value above thresholds in % in % in % Total Total ( without 65 and 66) C D

6 E F G H I J ~ ~ ~ ~ ~ ~ K