Dublin, april 2012 Role of Business Register in coordinated sampling Boro Nikić
Coordination of the samples-motivation The statistical system is under increasing pressure to reduce the burden for respondents Budget for the surveys is decreasing Centralized system of sample selection
Business Surveys at SORS –current practice(1) The main source is Business Register approx. 220000 legal units and 250000 local kind of activity units 15% of units in BR is inactive some activities and addresses of units in BR are wrong Some LKAU in BR are missing There is no data about number of employees and turnover
Business Surveys at SORS –current practice(2) At the beginning of the each year two “master” sampling frames are created: Sampling frame of Legal units Sampling frame of Local units Inactive units are excluded, some activities of units are changed
Business Surveys at SORS –current practice(3) Information about number of employees and turnover is added Sources: - Statistical Register of Employment - Tax data - Annual accounts
Business Surveys at SORS –current practice(4) Administrative sources: The Annual Accounting Records of companies, sole proprietors, legal entities of private law, associations and legal entities of public law, which are collected by the Agency of the Republic of Slovenia for Public Legal Records and Related Services (AJPES) The Value Added Tax Database from of the Tax Authority.
Business Surveys at SORS –current practice(5) The main goals: Obtain (as much as we get) precise results about population or domain characteristics At the same time reducing the burden of responding unit as much as we can Reducing the burden of the people which are involved in the Survey process .
Business Surveys at SORS –current practice(6) Independently carrying out in three ways: - as a probability sample survey (long term surveys) - as a Cut-off survey (mostly short-term surveys) - as a Administrative based Survey (SBS Survey)
Business Surveys at SORS –current practice(7) Usually stratified Sampling Design. Two strata variables: - Number of employees or turnover or both of them - Nace (activity) classification (2 digit Nace codes, sectors,…, depends on domains of interests)
Business Surveys at SORS –current practice(8) If we assume that sample size is already determined: - units of the biggest size are entirely included in the sample - proportional allocation (for small units) is used most often - optimal allocation is used when we have auxiliary information (in Slovenian case turnover, number of employees,..)
Business Surveys at SORS –current practice(9) Two most often selection methods: SRSWOR (simple random sample without replacement) selection method SYSTEMATIC selection method
Coordination of the samples (1) There is around 10 probability samples selected each year Due to the independently selected samples we don’t have control of burden of small units. Some of them might be selected in many samples. In 2011 we started to study possibilities to employ coordination of the samples
Coordination of the samples (2) For 8 surveys conducted in 2011 which based on probability selected samples simulation was made in order to see if it is possible to reduce burden of small units: common sampling frame was created each unit in frame received permanent random number (between 0 and 1)
Coordination of the samples (3) Sample size and its allocation among strata remained the same 8 starting points were chosen (0,0.125,0.25,0.375,…,0.875) For the first survey units with random numbers starting from 0 in all strata were selected, for the second one units with random numbers starting from 0.125 wee selected,… Only negative coordination of samples was desired
Coordination of the samples (4) 2011 Number of units included in samples distributed by method of selection Coordination Samples Independent Samples 1 6954 9168 2 2176 2738 3 2324 1765 4 1353 779 5 193 231 6 29 95 7 8 Total number of units included in samples 13029 14783
Coordination of the samples (5) Results shows that we can reduce the burden of units if we employ coordination of the sample Improving could be even bigger if we know sample sizes for all surveys in advance. In that case we can chose starting points which are not equidistance
Coordination of the samples (6) Some disadvantages or challenges : systematic selection method is not possible (only SRSWOR?) Is it possible to use coordination of samples over the years (bias problem)? Is it possible to include complicity of questionnaire in coordinated sampling? We need to know sampling designs beforehand
Thank you for your attention!