assessment of quality in the swedish cvts2 · assessment of quality in the swedish cvts2 by lennart...

75
Assessment of Quality in the Swedish CVTS2 by Lennart Forssén Statistics Sweden Department of Labour and Education Statistics SE – 701 89 ÖREBRO SWEDEN Tel: + 46 19 17 65 65 Fax: + 46 19 17 70 82 E-mail: [email protected]

Upload: trinhminh

Post on 19-Jun-2018

213 views

Category:

Documents


0 download

TRANSCRIPT

Assessment of Quality in the Swedish CVTS2

by

Lennart Forssén Statistics Sweden Department of Labour and Education Statistics SE – 701 89 ÖREBRO SWEDEN Tel: + 46 19 17 65 65 Fax: + 46 19 17 70 82 E-mail: [email protected]

2

Contents Page 1. Implementation of the survey and relevance of the concepts 3 2. Accuracy 5 2.1. Sampling errors 5 Sample selection 5 Calculation of the final weights 13 2.2. Nonsampling errors 16 2.2.1. Coverage errors 16 Sampling frame 16 Coverage errors in general 17 2.2.2. Measurement errors 19 Questionnaire and guidelines to the respondents 19 Results from the testing at the Measurement Laboratory 21 Data collection 24 Reactions of respondents 26 Observed problems in connection to the data collection 27 2.2.3. Processing errors 29 Editing 29 Testing and corrections 30 Imputations 31 2.2.4. Nonresponse errors 33 3. Timeliness 41 4. Accessibility and clarity 42 5. Comparability 42 6. Coherence 43 Comparisons with the Swedish Structural Business Statistics (SBS) 43 Comparisons with the Swedish Staff Training Survey 45 7. References 49 Annex 1. Time table 50 Annex 2. Questionnaire for data collection 51 Annex 3. Guidance to the enterprises 63 Annex 4. Sampling Plan 67

3

1. Implementation of the survey and relevance of the concepts Sweden was not a member state of EU in 1994 and therefore did not participate in CVTS1. Sweden however, participated as an observer in the working group for CVTS1. Already then it was decided to try to adjust the Swedish way of measuring CVT as much as possible to CVTS and thereby be able to make better international comparisons of statistics on CVT. The Swedish survey has therefore ever since repeatedly been adapted to the CVTS and is planned to continue that way in the future. Enterprise size and training cost are examples of variables not included in the primary Swedish surveys. Both are now included. For quality reasons the Swedish measuring period is six months instead of a calendar year as in CVTS. By using the same sample of individuals for both half years during a calendar year and by developing new statistical methods it is now possible to present calendar year estimates on CVT. The Swedish way of measuring CVT is, however, based on a survey to individuals participating in CVT and on a model for measuring costs of CVT, unlike the CVTS based on a survey to enterprises. The Swedish survey for measuring CVT therefore cannot replace the CVTS2 but can be used for comparisons of results obtained in CVTS2. Some comparisons of results from CVTS2 with results from the Swedish way of measuring are presented in Section 6. Interest for the Swedish participation in CVTS2 was far from overwhelming by users of statistics on CVT in Sweden. The reason for that was, as described above, Sweden has already since many years had regular surveys on CVT, based on surveys to individuals in the labour force. What is new with statistics from CVTS2 is mainly that information is collected from enterprises about policies concerning CVT, about contributions to collective funding arrangements and receipts from collective funds in different enterprises. The interest for internationally comparable statistics on CVT is quite large from many different users like ministries, education providers, enterprises and different labour market organisations. It is hoped by many users that results from CVTS2 can be used for that purpose. The interest for CVTS2 from different national users of statistics about CVT has, so far, mainly been focused on the fact that for the first time it will be possible to get information about CVT from an enterprise survey. The expectation of many users is, for the first time a possibility is given to compare CVT policies for enterprises in different branches of Industry and for enterprises of different size. At this moment it is too early to conclude anything about satisfaction of user’s needs and therefore also concerning the relevance of different concepts. Results from the Swedish CVTS2 will probably not be published until the end of 2001. The work with the implementation of the survey started in October/November 1999 with preliminary allocations of the planned sample with different assumptions for the expected nonresponse rate. In the middle of November 1999 we also delivered the received common CVTS2 questionnaire and guidelines to one of our experts on translations from English into Swedish. Both the questionnaire and the guidelines were then adjusted to better suit conditions in Sweden before we delivered this material together with drafts of introduction and reminder letters to out Measurement Laboratory for evaluation. Discussions also started at the end of 1999 with the department responsible for most of Statistics Sweden’s surveys to enterprises. We asked for advice on whom to address our questionnaires, their opinion about our different questions and the wording of our questions, the possibility to get correct answers etc. At the end of 1999 discussions also continued with our survey department within Statistics Sweden. We asked, for instance, what they thought about nonresponse rates, data

4

collection costs, best way of collecting the data, to what extent we could expect help from them, for how long period etc. In order to perform a survey with enterprises as respondents, contacts with involved employer organisations are mandatory in Sweden. In the case of a voluntary survey in the private labour market sector like the CVTS2, at least the Board of Swedish Industry and Commerce for better Regulation (NNR) must be consulted before starting the data collection. NNR has as one of its main tasks to safeguard the response burden on enterprises. In every survey where information is collected from enterprises, municipalities, county councils or state authorities it is always important to be able to inform the respondents that one has come to an agreement with NNR concerning the way in which the survey will be conducted. The reason for that is, if you do not come to an agreement with NNR on how to perform the survey you cannot mention in your questionnaire or introductory letters that you have an agreement with NNR. That really is something that can affect the response rate of your survey. In the case of CVTS2, NNR objected to several of the questions and meant that some questions were too difficult to understand or would lead to a too high response burden on the enterprises. The need for certain questions like for instance Questions A4b, A5a and A5b were questioned and should they said be excluded from future similar surveys. Both Questions A5a and A5b were also considered to be better connected to other questions like, for instance, if it had had an influence on the need for skills or CVT within the enterprise. NNR had a lot of different viewpoints and doubts concerning CVTS2. However, as the survey was not mandatory, they often pointed out that we should reconsider certain matters before they could come to an agreement with us. For instance they mentioned that we should inform the respondents in the questionnaire or in the guidelines that they could give estimated values, either in days or hours etc. They also wanted us to refer to different accounts according to the EU BAS 99 accounting system. NNR also quite vigorously questioned why we were asking the enterprises and not the individuals themselves about fields of training. They simply believed that it was impossible for the enterprises to answer such questions. In consideration to all the criticism, which we received from NNR on several questions in the CVTS2 questionnaire, it will probably be very difficult to accomplish renewed agreements with NNR in the future. Especially difficult will it be if future CVT-surveys will be mandatory for the enterprises and if at the same time the questionnaire will still contain more or less the same kind of questions as in CVTS2. In order to be able to take full advantage of a possible legal obligation in the future it is also probably necessary to give the enterprises a forewarning about that at least one and a half year in advance. The implementation of the survey is described more in detail below in different sections. Here I just want to mention that the work needed to be done with the survey has been much more time consuming than expected. The reasons for that are many but the main reason had in some way to do with difficulties for the enterprises to provide us with correct answers to many of the questions. In this report all references to question names and variables refer to, if not otherwise mentioned, the question and variable names mentioned in the European outline questionnaire or Code Book (EUROSTAT 2000a).

5

2. Accuracy 2.1. Sampling errors Sample selection The Swedish sample is based on a stratified random sample of enterprises in the 20 by 3 NACE x Size groups agreed on and according to the description in the document EUROSTAT/E3/99/CVTS017. The Swedish Business Register contains information of good quality, which made it easy to select the enterprises which should be included in the sampling frame for the CVTS2. The Swedish CVTS2 sample was allocated in such a way that we wanted the maximum length of half the confidence interval to be 0.10 with an assumption of an expected nonresponse rate of around 33% in each stratum. It later turned out we should have assumed a higher non-response rate, as it was impossible to reach a response rate of 67 %. The final sample selection of enterprises to the Swedish CVTS2 was done on the 9th of February 2000. The sampling frame used was the Swedish Business Register. This register is of good quality at any time of the year and both overcoverage and undercoverage are usually fairly small in this register. See Section 2.2. below for a description of what is included in the Swedish Business Register. A description of the nonresponse and the overcoverage registered in the Swedish CVTS2 can be found in Sections 2.2.4. and 2.2.1. respectively. In Tables 1 and 2 below is a description of the total number of enterprises respective the total number of sampled enterprises in all the different branches of Industry included in the CVTS2.

6

Table 1. Total number of enterprises by NACE- and size group in the sampling frame. NACE-group Size group Total 10-49 50-249 250- employees employees employees NACE 10-14 47 12 3 62 NACE 15-16 468 91 42 601 NACE 17-19 171 43 9 223 NACE 21-22 748 203 61 1012 NACE 23-26 518 197 55 770 NACE 27-28 1346 264 48 1658 NACE 29-33 1211 397 133 1741 NACE 34-35 207 110 39 356 NACE 20, 36-37 752 219 29 1000 NACE 40-41 175 73 18 266 NACE 45 2595 202 38 2835 NACE 50 777 139 15 931 NACE 51 3055 401 60 3516 NACE 52 2336 208 51 2595 NACE 55 1393 144 20 1557 NACE 60-63 1813 245 49 2107 NACE 64 38 18 15 71 NACE 65-66 161 102 30 293 NACE 67 107 24 4 135 NACE 70-74, 90-93 6029 986 163 7178 Total 23947 4078 882 28907 Table 2. Total number of sampled enterprises by NACE- and size group. NACE-group Size group Total 10-49 50-249 250- employees employees employees NACE 10-14 47 12 3 62 NACE 15-16 180 86 42 308 NACE 17-19 171 43 9 223 NACE 21-22 173 116 61 350 NACE 23-26 175 120 55 350 NACE 27-28 215 139 48 402 NACE 29-33 177 135 95 407 NACE 34-35 120 87 39 246 NACE 20, 36-37 189 126 29 344 NACE 40-41 106 70 18 194 NACE 45 223 126 38 387 NACE 50 153 97 15 265 NACE 51 185 136 60 381 NACE 52 177 113 51 341 NACE 55 183 101 20 304 NACE 60-63 207 130 49 386 NACE 64 38 18 15 71 NACE 65-66 105 84 30 219 NACE 67 86 24 4 114 NACE 70-74, 90-93 176 151 103 430 Total 3086 1914 784 5784

7

NACE-codes were taken from the Swedish Business Register at the time of the sample selection. As the quality in this register is considered to be high, it was not considered necessary to include Question A1 about principal economic activity of the enterprise in the Swedish CVTS2 questionnaire. We have therefore no differences between sampled NACE-group and observed NACE-group. The differences between the number of normal answers by NACE-group and by sampled size group respective observed size group is described in Table 3 below. Table 3. Number of enterprises by NACE-group, sampled and observed size group. NACE-group Sampled size group Total Observed size group 10-49 50-249 250- 10-49 50-249 250- employees employees employees employees employees employees NACE C, 10-14 27 5 2 34 27 5 2 NACE D, 15-16 81 36 30 147 77 40 30 NACE D, 17-19 81 23 8 112 81 24 7 NACE D, 21-22 82 47 39 168 79 49 40 NACE D, 23-26 86 73 36 195 89 69 37 NACE D, 27-28 106 77 32 215 113 70 32 NACE D, 29-33 83 70 64 217 86 68 63 NACE D, 34-35 60 51 28 139 61 50 28 NACE D, 20,36-37 83 70 20 173 86 67 20 NACE E, 40-41 60 52 8 120 59 51 10 NACE F, 45 99 66 24 189 105 60 24 NACE G, 50 52 45 11 108 51 44 13 NACE G, 51 83 56 41 180 93 48 39 NACE G, 52 43 42 31 116 46 40 30 NACE H, 55 58 41 9 108 61 39 8 NACE I, 60-63 75 60 29 164 82 49 33 NACE I, 64 13 10 14 37 13 8 16 NACE J, 65-66 46 36 15 97 48 36 13 NACE J, 67 33 10 2 45 32 9 4 NACE K+O 75 70 62 207 74 71 62 Total 1326 940 505 2771 1363 897 511 Question A2 about number of persons employed at the end of 1999 was included in the Swedish questionnaire. We therefore have observed some changes between sampled size group and the observed size group due to changes in the number of employees. An enterprise with less than 10 employees is not included in CVTS2 and is coded accordingly as overcoverage. For normal answers the number of enterprises by sampled size group and observed size group is described in Table 4 below.

8

Table 4. Number of enterprises by sampled and observed size group. Sampled size group Observed size group 10-49 50-249 250- Total employees employees employees 10-49 employees 1283 38 5 1326 50-249 employees 78 834 28 940 250- employees 2 25 478 505 Total 1363 897 511 2771 Below in Tables 5 to 14 are some key statistics with coefficient of variation broken down by observed NACE- and size group. The coefficient of variation is the ratio of the squared root of the variance of the estimator to the expected value in per cent. The coefficient of variation to an estimated value can be found to the right of the ± sign in each of Tables 5 to 14 below. Table 5. Total number of persons employed by observed NACE- and size group. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 employees 50-249 employees 250- employees NACE C, 10-14 852 ± 2 % 1344 ± 9 % 5722 ± 0 % 7917 ± 2 % NACE D, 15-16 8104 ± 4 % 10099 ± 5 % 41278 ± 5 % 59481 ± 4 % NACE D, 17-19 3213 ± 3 % 4130 ± 4 % 5383 ± 2 % 12727 ± 1 % NACE D, 21-22 12486 ± 4 % 23340 ± 5 % 45686 ± 4 % 81512 ± 2 % NACE D, 23-26 11002 ± 3 % 20671 ± 4 % 42989 ± 3 % 74663 ± 1 % NACE D, 27-28 26349 ± 3 % 23899 ± 6 % 49826 ± 2 % 100074 ± 1 % NACE D, 29-33 25285 ± 4 % 40377 ± 4 % 111362 ± 4 % 177024 ± 2 % NACE D, 34-35 4853 ± 4 % 11796 ± 5 % 73996 ± 3 % 90644 ± 2 % NACE D, 20,36-37 15044 ± 3 % 20125 ± 4 % 23255 ± 3 % 58424 ± 1 % NACE E, 40-41 3640 ± 4 % 7757 ± 5 % 13217 ± 3 % 24614 ± 1 % NACE F, 45 42275 ± 3 % 18400 ± 9 % 72539 ± 3 % 133214 ± 2 % NACE G, 50 12429 ± 5 % 14105 ± 7 % 15187 ± 13 % 41721 ± 4 % NACE G, 51 57146 ± 3 % 36971 ± 6 % 36019 ± 8 % 130135 ± 2 % NACE G, 52 35299 ± 9 % 29353 ± 31 % 77361 ± 3 % 142013 ± 6 % NACE H, 55 22082 ± 6 % 18077 ± 19 % 19066 ± 7 % 59226 ± 5 % NACE I, 60-63 32921 ± 4 % 20855 ± 9 % 76482 ± 5 % 130259 ± 3 % NACE I, 64 948 ± 11 % 1556 ± 15 % 94436 ± 1 % 96940 ± 0 % NACE J, 65-66 2955 ± 4 % 10952 ± 5 % 46659 ± 4 % 60566 ± 3 % NACE J, 67 1790 ± 5 % 1964 ± 10 % 13707 ± 62 % 17460 ± 49 % NACE K+O 86394 ± 7 % 115100 ± 10 % 230048 ± 24 % 431542 ± 12 % Total 405068 ± 2 % 430870 ± 4 % 1094217 ± 5 % 1930155 ± 3 %

9

Table 6. Total number of enterprises that provided CVT by observed NACE- and size

group. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 employees 50-249 employees 250- employees NACE C, 10-14 36 ± 5 % 12 ± 0 % 3 ± 0 % 51 ± 3 % NACE D, 15-16 316 ± 5 % 105 ± 6 % 42 ± 7 % 464 ± 3 % NACE D, 17-19 114 ± 4 % 42 ± 5 % 10 ± 4 % 165 ± 3 % NACE D, 21-22 552 ± 4 % 213 ± 7 % 67 ± 6 % 831 ± 3 % NACE D, 23-26 431 ± 4 % 186 ± 4 % 59 ± 5 % 676 ± 2 % NACE D, 27-28 1037 ± 5 % 256 ± 8 % 49 ± 6 % 1342 ± 3 % NACE D, 29-33 1085 ± 3 % 373 ± 5 % 135 ± 4 % 1593 ± 2 % NACE D, 34-35 158 ± 5 % 106 ± 4 % 41 ± 5 % 304 ± 2 % NACE D, 20,36-37 588 ± 5 % 203 ± 5 % 28 ± 7 % 819 ± 3 % NACE E, 40-41 154 ± 2 % 72 ± 5 % 22 ± 5 % 249 ± 0 % NACE F, 45 1914 ± 5 % 216 ± 13 % 40 ± 7 % 2170 ± 4 % NACE G, 50 637 ± 4 % 137 ± 7 % 26 ± 25 % 800 ± 3 % NACE G, 51 2511 ± 3 % 364 ± 10 % 62 ± 11 % 2938 ± 3 % NACE G, 52 1644 ± 6 % 272 ± 25 % 55 ± 9 % 1971 ± 3 % NACE H, 55 817 ± 7 % 194 ± 18 % 28 ± 7 % 1039 ± 5 % NACE I, 60-63 1213 ± 6 % 221 ± 13 % 61 ± 9 % 1495 ± 5 % NACE I, 64 26 ± 15 % 15 ± 14 % 16 ± 5 % 58 ± 6 % NACE J, 65-66 133 ± 2 % 89 ± 4 % 26 ± 7 % 248 ± 0 % NACE J, 67 87 ± 3 % 19 ± 6 % 9 ± 29 % 115 ± 0 % NACE K+O 3973 ± 5 % 1229 ± 12 % 298 ± 29 % 5501 ± 3 % Total 17428 ± 2 % 4324 ± 4 % 1077 ± 8 % 22828 ± 1 % Table 7. Ratio of enterprises providing CVT by observed NACE- and size group. Per cent. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 50-249 250- NACE C, 10-14 85 ± 5 % 100 ± 0 % 100 ± 0 % 89 ± 3 % NACE D, 15-16 82 ± 5 % 100 ± 0 % 97 ± 2 % 87 ± 3 % NACE D, 17-19 78 ± 4 % 95 ± 3 % 100 ± 0 % 83 ± 3 % NACE D, 21-22 89 ± 4 % 97 ± 2 % 100 ± 0 % 92 ± 3 % NACE D, 23-26 91 ± 3 % 99 ± 1 % 100 ± 0 % 93 ± 2 % NACE D, 27-28 82 ± 4 % 100 ± 0 % 100 ± 0 % 86 ± 3 % NACE D, 29-33 93 ± 3 % 96 ± 2 % 100 ± 0 % 95 ± 2 % NACE D, 34-35 82 ± 5 % 98 ± 2 % 100 ± 0 % 89 ± 2 % NACE D, 20,36-37 84 ± 5 % 100 ± 0 % 95 ± 3 % 88 ± 3 % NACE E, 40-41 100 ± 0 % 100 ± 0 % 100 ± 0 % 100 ± 0 % NACE F, 45 81 ± 5 % 100 ± 0 % 100 ± 0 % 83 ± 4 % NACE G, 50 95 ± 3 % 98 ± 2 % 100 ± 0 % 96 ± 3 % NACE G, 51 93 ± 3 % 100 ± 0 % 100 ± 0 % 94 ± 3 % NACE G, 52 93 ± 4 % 98 ± 2 % 100 ± 0 % 94 ± 3 % NACE H, 55 82 ± 6 % 100 ± 0 % 100 ± 0 % 85 ± 5 % NACE I, 60-63 80 ± 6 % 98 ± 2 % 100 ± 0 % 83 ± 5 % NACE I, 64 74 ± 14 % 100 ± 0 % 93 ± 2 % 85 ± 6 % NACE J, 65-66 100 ± 0 % 100 ± 0 % 100 ± 0 % 100 ± 0 % NACE J, 67 100 ± 0 % 100 ± 0 % 100 ± 0 % 100 ± 0 % NACE K+O 92 ± 4 % 100 ± 0 % 99 ± 1 % 94 ± 3 % Total 88 ± 1 % 99 ± 0 % 99 ± 0 % 91 ± 1 %

10

Table 8. Total number of enterprises that provided CVT- courses by observed NACE- and size group. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 50-249 250- NACE C, 10-14 31 ± 7 % 12 ± 0 % 3 ± 0 % 46 ± 5 % NACE D, 15-16 234 ± 9 % 100 ± 5 % 42 ± 7 % 377 ± 5 % NACE D, 17-19 90 ± 6 % 34 ± 9 % 10 ± 4 % 134 ± 4 % NACE D, 21-22 475 ± 6 % 208 ± 7 % 67 ± 6 % 749 ± 4 % NACE D, 23-26 379 ± 5 % 183 ± 4 % 59 ± 5 % 621 ± 3 % NACE D, 27-28 924 ± 6 % 253 ± 8 % 49 ± 6 % 1226 ± 4 % NACE D, 29-33 983 ± 5 % 373 ± 5 % 135 ± 4 % 1491 ± 3 % NACE D, 34-35 130 ± 7 % 104 ± 5 % 41 ± 5 % 274 ± 3 % NACE D, 20,36-37 400 ± 9 % 196 ± 6 % 28 ± 7 % 625 ± 6 % NACE E, 40-41 152 ± 3 % 72 ± 5 % 22 ± 5 % 246 ± 1 % NACE F, 45 1694 ± 6 % 210 ± 14 % 40 ± 7 % 1944 ± 5 % NACE G, 50 577 ± 6 % 134 ± 8 % 26 ± 25 % 736 ± 4 % NACE G, 51 2382 ± 4 % 356 ± 10 % 62 ± 11 % 2801 ± 3 % NACE G, 52 1444 ± 8 % 272 ± 25 % 55 ± 9 % 1771 ± 6 % NACE H, 55 638 ± 10 % 194 ± 18 % 28 ± 7 % 860 ± 7 % NACE I, 60-63 1066 ± 8 % 161 ± 8 % 61 ± 9 % 1288 ± 6 % NACE I, 64 23 ± 18 % 15 ± 14 % 16 ± 5 % 54 ± 6 % NACE J, 65-66 133 ± 2 % 89 ± 4 % 26 ± 7 % 248 ± 0 % NACE J, 67 82 ± 4 % 19 ± 6 % 9 ± 29 % 110 ± 3 % NACE K+O 3811 ± 6 % 1229 ± 12 % 298 ± 29 % 5338 ± 3 % Total 15648 ± 2 % 4215 ± 4 % 1077 ± 8 % 20939 ± 1 % Table 9. Ratio of enterprises providing CVT- courses by observed NACE- and size group. Per cent. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 50-249 250- NACE C, 10-14 74 ± 7 % 100 ± 0 % 100 ± 0 % 81 ± 5 % NACE D, 15-16 61 ± 8 % 95 ± 4 % 97 ± 2 % 71 ± 5 % NACE D, 17-19 62 ± 5 % 78 ± 8 % 100 ± 0 % 67 ± 4 % NACE D, 21-22 76 ± 6 % 95 ± 3 % 100 ± 0 % 83 ± 4 % NACE D, 23-26 80 ± 5 % 97 ± 2 % 100 ± 0 % 86 ± 3 % NACE D, 27-28 73 ± 5 % 99 ± 1 % 100 ± 0 % 78 ± 4 % NACE D, 29-33 85 ± 5 % 96 ± 2 % 100 ± 0 % 89 ± 3 % NACE D, 34-35 67 ± 7 % 96 ± 2 % 100 ± 0 % 80 ± 3 % NACE D, 20,36-37 57 ± 9 % 97 ± 2 % 95 ± 3 % 67 ± 6 % NACE E, 40-41 99 ± 1 % 100 ± 0 % 100 ± 0 % 99 ± 1 % NACE F, 45 72 ± 6 % 97 ± 2 % 100 ± 0 % 75 ± 5 % NACE G, 50 86 ± 6 % 96 ± 3 % 100 ± 0 % 88 ± 4 % NACE G, 51 88 ± 4 % 98 ± 2 % 100 ± 0 % 89 ± 3 % NACE G, 52 82 ± 7 % 98 ± 2 % 100 ± 0 % 85 ± 6 % NACE H, 55 64 ± 9 % 100 ± 0 % 100 ± 0 % 70 ± 7 % NACE I, 60-63 70 ± 7 % 72 ± 13 % 100 ± 0 % 71 ± 6 % NACE I, 64 64 ± 15 % 100 ± 0 % 93 ± 2 % 80 ± 6 % NACE J, 65-66 100 ± 0 % 100 ± 0 % 100 ± 0 % 100 ± 0 % NACE J, 67 94 ± 4 % 100 ± 0 % 100 ± 0 % 95 ± 3 % NACE K+O 88 ± 4 % 100 ± 0 % 99 ± 1 % 91 ± 3 % Total 79 ± 2 % 97 ± 1 % 99 ± 0 % 83 ± 1 %

11

Table 10. Total number of persons employed by trainers, by observed NACE- and size group. Coefficient of variation to the right of the ± sign.

NACE-group Number of employees Total 10-49 employees 50-249 employees 250- employees NACE C, 10-14 763 ± 6 % 1344 ± 9 % 5722 ± 0 % 7829 ± 2 % NACE D, 15-16 6924 ± 6 % 10099 ± 5 % 40910 ± 5 % 57933 ± 4 % NACE D, 17-19 2701 ± 4 % 4014 ± 4 % 5383 ± 2 % 12098 ± 1 % NACE D, 21-22 11563 ± 5 % 23011 ± 5 % 45686 ± 4 % 80261 ± 2 % NACE D, 23-26 10044 ± 4 % 20319 ± 4 % 42989 ± 3 % 73352 ± 2 % NACE D, 27-28 22298 ± 6 % 23899 ± 6 % 49826 ± 2 % 96022 ± 1 % NACE D, 29-33 24090 ± 4 % 39615 ± 4 % 111362 ± 4 % 175068 ± 2 % NACE D, 34-35 4264 ± 5 % 11626 ± 5 % 73996 ± 3 % 89886 ± 2 % NACE D, 20,36-37 12413 ± 7 % 20125 ± 4 % 22592 ± 3 % 55130 ± 2 % NACE E, 40-41 3640 ± 4 % 7757 ± 5 % 13217 ± 3 % 24614 ± 1 % NACE F, 45 35078 ± 6 % 18400 ± 9 % 72539 ± 3 % 126017 ± 2 % NACE G, 50 11914 ± 5 % 13926 ± 8 % 15187 ± 13 % 41027 ± 4 % NACE G, 51 54157 ± 4 % 36971 ± 6 % 36019 ± 8 % 127146 ± 2 % NACE G, 52 33913 ± 10 % 28997 ± 31 % 77361 ± 3 % 140271 ± 6 % NACE H, 55 19368 ± 8 % 18077 ± 19 % 19066 ± 7 % 56511 ± 6 % NACE I, 60-63 26941 ± 7 % 20571 ± 9 % 76482 ± 5 % 123994 ± 3 % NACE I, 64 748 ± 17 % 1556 ± 15 % 93689 ± 1 % 95992 ± 1 % NACE J, 65-66 2955 ± 4 % 10952 ± 5 % 46659 ± 4 % 60566 ± 3 % NACE J, 67 1790 ± 5 % 1964 ± 10 % 13707 ± 62 % 17460 ± 49 % NACE K+O 80189 ± 8 % 115100 ± 10 % 229383 ± 24 % 424672 ± 13 % Total 365754 ± 2 % 428322 ± 4 % 1091773 ± 5 % 1885849 ± 3 % Table 11. Total number of participants in CVT-courses by observed NACE- and size group. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 employees 50-249 employees 250- employees NACE C, 10-14 327 ± 11 % 1068 ± 10 % 4459 ± 0 % 5854 ± 2 % NACE D, 15-16 2703 ± 13 % 4052 ± 9 % 24069 ± 5 % 30824 ± 4 % NACE D, 17-19 1209 ± 9 % 1645 ± 12 % 3644 ± 4 % 6498 ± 4 % NACE D, 21-22 5930 ± 9 % 12362 ± 10 % 33127 ± 5 % 51420 ± 4 % NACE D, 23-26 4740 ± 10 % 13562 ± 7 % 32990 ± 4 % 51292 ± 3 % NACE D, 27-28 11345 ± 9 % 13734 ± 9 % 34743 ± 4 % 59822 ± 3 % NACE D, 29-33 12655 ± 9 % 21996 ± 7 % 77283 ± 7 % 111933 ± 5 % NACE D, 34-35 1625 ± 14 % 6238 ± 9 % 62388 ± 2 % 70252 ± 2 % NACE D, 20,36-37 4263 ± 15 % 7102 ± 9 % 8945 ± 16 % 20310 ± 8 % NACE E, 40-41 3030 ± 6 % 7020 ± 5 % 12469 ± 3 % 22519 ± 2 % NACE F, 45 19947 ± 10 % 9469 ± 10 % 44856 ± 6 % 74272 ± 4 % NACE G, 50 6504 ± 10 % 7233 ± 11 % 6928 ± 15 % 20665 ± 6 % NACE G, 51 33537 ± 7 % 18748 ± 12 % 19609 ± 4 % 71894 ± 4 % NACE G, 52 19002 ± 15 % 10501 ± 16 % 51625 ± 5 % 81128 ± 4 % NACE H, 55 7554 ± 16 % 7723 ± 13 % 11341 ± 5 % 26618 ± 5 % NACE I, 60-63 11589 ± 12 % 8176 ± 12 % 41867 ± 9 % 61632 ± 7 % NACE I, 64 318 ± 21 % 851 ± 22 % 63013 ± 1 % 64183 ± 1 % NACE J, 65-66 2121 ± 7 % 9620 ± 6 % 40757 ± 3 % 52498 ± 2 % NACE J, 67 1197 ± 8 % 1060 ± 17 % 9672 ± 62 % 11929 ± 50 % NACE K+O 55439 ± 11 % 66450 ± 11 % 151178 ± 30 % 273067 ± 16 % Total 205036 ± 4 % 228611 ± 4 % 734962 ± 6 % 1168609 ± 4 %

12

Table 12. Ratio of participants in CVT-courses to total number of employees by observed NACE- and size group. Per cent. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 50-249 250- NACE C, 10-14 38 ± 10 % 80 ± 8 % 78 ± 0 % 74 ± 1 % NACE D, 15-16 33 ± 12 % 40 ± 9 % 58 ± 3 % 52 ± 3 % NACE D, 17-19 38 ± 8 % 40 ± 12 % 68 ± 3 % 51 ± 4 % NACE D, 21-22 47 ± 8 % 53 ± 8 % 73 ± 3 % 63 ± 3 % NACE D, 23-26 43 ± 9 % 66 ± 5 % 77 ± 3 % 69 ± 2 % NACE D, 27-28 43 ± 9 % 57 ± 6 % 70 ± 3 % 60 ± 3 % NACE D, 29-33 50 ± 8 % 54 ± 6 % 69 ± 5 % 63 ± 4 % NACE D, 34-35 33 ± 13 % 53 ± 7 % 84 ± 2 % 78 ± 2 % NACE D, 20,36-37 28 ± 14 % 35 ± 9 % 38 ± 16 % 35 ± 8 % NACE E, 40-41 83 ± 3 % 90 ± 1 % 94 ± 2 % 91 ± 1 % NACE F, 45 47 ± 9 % 51 ± 6 % 62 ± 7 % 56 ± 5 % NACE G, 50 52 ± 9 % 51 ± 8 % 46 ± 10 % 50 ± 5 % NACE G, 51 59 ± 7 % 51 ± 11 % 54 ± 8 % 55 ± 5 % NACE G, 52 54 ± 12 % 36 ± 27 % 67 ± 5 % 57 ± 7 % NACE H, 55 34 ± 14 % 43 ± 14 % 59 ± 8 % 45 ± 6 % NACE I, 60-63 35 ± 11 % 39 ± 12 % 55 ± 6 % 47 ± 5 % NACE I, 64 34 ± 18 % 55 ± 13 % 67 ± 1 % 66 ± 1 % NACE J, 65-66 72 ± 5 % 88 ± 3 % 87 ± 4 % 87 ± 3 % NACE J, 67 67 ± 6 % 54 ± 12 % 71 ± 2 % 68 ± 2 % NACE K+O 64 ± 7 % 58 ± 8 % 66 ± 7 % 63 ± 5 % Total 51 ± 3 % 53 ± 3 % 67 ± 2 % 61 ± 2 % Table 13. Ratio of participants in CVT-courses to total number of employees in trainer enterprises by observed NACE- and size group. Per cent. Coefficient of variation to the right of the ± sign. NACE-group Number of employees Total 10-49 50-249 250- NACE C, 10-14 43 ± 10 % 80 ± 8 % 78 ± 0 % 75 ± 2 % NACE D, 15-16 39 ± 11 % 40 ± 9 % 59 ± 3 % 53 ± 3 % NACE D, 17-19 45 ± 7 % 41 ± 11 % 68 ± 3 % 54 ± 4 % NACE D, 21-22 51 ± 8 % 54 ± 8 % 73 ± 3 % 64 ± 3 % NACE D, 23-26 47 ± 9 % 67 ± 5 % 77 ± 3 % 70 ± 2 % NACE D, 27-28 51 ± 8 % 57 ± 6 % 70 ± 3 % 62 ± 3 % NACE D, 29-33 53 ± 8 % 56 ± 6 % 69 ± 5 % 64 ± 4 % NACE D, 34-35 38 ± 13 % 54 ± 7 % 84 ± 2 % 78 ± 2 % NACE D, 20,36-37 34 ± 13 % 35 ± 9 % 40 ± 16 % 37 ± 8 % NACE E, 40-41 83 ± 3 % 90 ± 1 % 94 ± 2 % 91 ± 1 % NACE F, 45 57 ± 8 % 51 ± 6 % 62 ± 7 % 59 ± 5 % NACE G, 50 55 ± 8 % 52 ± 8 % 46 ± 10 % 50 ± 5 % NACE G, 51 62 ± 6 % 51 ± 11 % 54 ± 8 % 57 ± 5 % NACE G, 52 56 ± 12 % 36 ± 27 % 67 ± 5 % 58 ± 7 % NACE H, 55 39 ± 13 % 43 ± 14 % 59 ± 8 % 47 ± 7 % NACE I, 60-63 43 ± 10 % 40 ± 12 % 55 ± 6 % 50 ± 5 % NACE I, 64 43 ± 16 % 55 ± 13 % 67 ± 1 % 67 ± 1 % NACE J, 65-66 72 ± 5 % 88 ± 3 % 87 ± 4 % 87 ± 3 % NACE J, 67 67 ± 6 % 54 ± 12 % 71 ± 2 % 68 ± 2 % NACE K+O 69 ± 6 % 58 ± 8 % 66 ± 7 % 64 ± 5 % Total 56 ± 3 % 53 ± 3 % 67 ± 2 % 62 ± 1 %

13

Table 14. Total direct costs for CVT courses by observed NACE- and size group. Coefficient of variation to the right of the ± sign. NACE -group Number of employees Total

10-49 50-249 250-

NACE C, 10-14 160453 ± 17 % 745332 ± 16 % 4491730 ± 0 % 5397515 ± 2 %

NACE D, 15-16 1803098 ± 21 % 2407581 ± 13 % 23817137 ± 6 % 28027816 ± 6 %

NACE D, 17-19 543011 ± 14 % 1444562 ± 18 % 1528711 ± 7 % 3516283 ± 8 %

NACE D, 21-22 3620270 ± 16 % 7979787 ± 12 % 30885119 ± 6 % 42485176 ± 5 %

NACE D, 23-26 3576398 ± 17 % 10565382 ± 11 % 25783425 ± 9 % 39925205 ± 6 %

NACE D, 27-28 4828525 ± 16 % 5093513 ± 17 % 17542062 ± 9 % 27464100 ± 7 %

NACE D, 29-33 8890925 ± 17 % 14880116 ± 12 % 94745295 ± 9 % 118516336 ± 7 %

NACE D, 34-35 1186583 ± 26 % 2915341 ± 12 % 42683157 ± 7 % 46785081 ± 7 %

NACE D, 20,36-37 1314094 ± 52 % 3296055 ± 10 % 13505379 ± 37 % 18115528 ± 28 %

NACE E, 40-41 3318094 ± 10 % 10893522 ± 6 % 36761958 ± 36 % 50973574 ± 26 %

NACE F, 45 7453421 ± 14 % 3186600 ± 12 % 37080702 ± 15 % 47720724 ± 12 %

NACE G, 50 3961075 ± 15 % 5014922 ± 15 % 7005319 ± 25 % 15981316 ± 12 %

NACE G, 51 27295058 ± 13 % 19425883 ± 22 % 24027637 ± 10 % 70748578 ± 8 %

NACE G, 52 8452308 ± 26 % 4591904 ± 21 % 17875067 ± 15 % 30919279 ± 11 %

NACE H, 55 3703129 ± 21 % 4400653 ± 23 % 4079621 ± 17 % 12183403 ± 11 %

NACE I, 60-63 5053952 ± 22 % 5198172 ± 16 % 47959691 ± 10 % 58211816 ± 8 %

NACE I, 64 725050 ± 28 % 1648265 ± 24 % 82700714 ± 1 % 85074029 ± 1 %

NACE J, 65-66 3304443 ± 12 % 13214652 ± 9 % 25818392 ± 9 % 42337487 ± 6 %

NACE J, 67 1781720 ± 16 % 1341812 ± 21 % 22082431 ± 54 % 25205963 ± 47 %

NACE K+O 61561290 ± 31 % 72768655 ± 15 % 168130156 ± 22 % 302460102 ± 14 %

Total 152532896 ± 13 % 191012711 ± 6 % 728503704 ± 6 % 1072049311 ± 5 %

Calculation of the final weights Preliminaries: In the following, let index h, h=1, 2, …, H, be used for the strata used in the sampling procedure, where H is the total number of strata in the survey, i.e. H=60.

In stratum h, the number of enterprises in the sampling frame is Nh. A sample sh of size nh is

taken from stratum h by simple random sampling and the response set rh contains $

hhh mmm +=′ responding enterprises, where mh is the number of enterprises that belong to

the population of interest, while $

hm is the number of overcoverage enterprises.

We want to estimate the total of a variable y of interest in the population or some subpopulation. For that we need to apply a weight wk to each enterprise k in the response set.

The weight should reflect the sampling design and also adjust for the nonresponse. When auxiliary information is available it should be included in the weight as well. The weight may then be constructed in the following way, kkkk gdw ×ν×= , where dk is the design weight,

νk is the nonresponse weight and gk is the impact if the auxiliary information.

Construction of the weights: The design weight dk is hh nN when enterprise k belongs to stratum h.

14

The nonresponse weight νk is hh mn ′ when enterprise k belongs to stratum h. The assumption

behind this weight is that all enterprises within stratum h respond with the same probability. When the assumption is not true we will get biased estimates. Unfortunately, there is no way to estimate the nonresponse bias or to test the model assumption from the observed data only. We try to reduce the nonresponse bias by using information about the number of enterprises and the total number of employees within each stratum. This information is available from the sampling frame.

From the Statistical Business Register (SBR) we know for each stratum h the number of enterprises, Mh, and the number of employees, ∑=

hU kxh xt , where Uh is the set of

enterprises in stratum h and xk is the number of employees in enterprise k according to SBR.

Note that an enterprise may have changed NACE and/or Size from the sampling occasion. The known totals are shown in Tables 1 and 2 below.

We use a linear model to describe the relationship between the two variables and the y-variable within each combination of NACE and Size.

The resulting weight gk, for a responding enterprise k that belongs to stratum h is defined by,

( ) ( ) ( )( )

∑ ∑ −′

−−+

h r hkh

h

hhhhkhh

hxx

m

NMxx

xxMM2~

ˆ~~1ˆ ,

where hM̂ is the estimated number of enterprises, hxhh Mtx = is the known mean value in

stratum h, hxhh Mtx ˆˆ~ = is the estimated mean value in stratum h and rh is the set of

responding enterprises in stratum h.

This is the result from a special case of the Generalised REGression (GREG) estimator, (see for example; Särndal, Swensson and Wretman (1992), Model Assisted Survey Sampling, New York:Springer-Verlag, pp 230-233).

All gk:s should have a value around 1, however, there is no guarantee that all gk:s are positive

for all samples. Negative, small positive and very large weights are undesirable and should be avoided. This was done by imposing the restriction on the gk:s to be between 0.5 and 3.

The final weight is ( ) khhk gmNw ′= . The ratio hh mN ′ is close to 1 in some strata that

might give wk<1 for some enterprises. This is not acceptable since each enterprise should at

least represent itself in the estimation process. In order to avoid this the additional restriction wk≥1 was imposed on the final weights.

The wk:s were calculated by the software CLAN97 developed at Statistics Sweden. This

software was also used for the calculation of the coefficient of variation for the desired key statistics.

15

Table 15. Number of enterprises in each NACE- and size group, Mh

NACE-group Size group Total 10-49 50-249 250- employees employees employees NACE 10-14 47 12 3 62 NACE 15-16 468 91 42 601 NACE 17-19 171 43 9 223 NACE 21-22 752 219 29 1000 NACE 23-26 748 203 61 1012 NACE 27-28 518 197 55 770 NACE 29-33 1346 264 48 1658 NACE 34-35 1211 397 133 1741 NACE 20, 36-37 207 110 39 356 NACE 40-41 175 73 18 266 NACE 45 2595 202 38 2835 NACE 50 777 139 15 931 NACE 51 3055 401 60 3516 NACE 52 2336 208 51 2595 NACE 55 1393 144 20 1557 NACE 60-63 1813 245 49 2107 NACE 64 38 18 15 71 NACE 65-66 161 102 30 293 NACE 67 107 24 4 135 NACE 70-74, 90-93 6029 986 163 7178 Total 23947 4078 882 28907 Table 16. Number of employees in each NACE- and size group, txh

NACE-group Size group Total 10-49 50-249 250- employees employees employees NACE 10-14 953 1217 5913 8083 NACE 15-16 9499 8937 40859 59295 NACE 17-19 3562 3880 6090 13532 NACE 21-22 15836 20565 22924 59325 NACE 23-26 14645 24395 47305 86345 NACE 27-28 11287 21673 45654 78614 NACE 29-33 26735 25480 51304 103519 NACE 34-35 25552 41269 119249 186070 NACE 20, 36-37 4639 11690 75338 91667 NACE 40-41 4245 8637 12718 25600 NACE 45 45982 17560 69116 132658 NACE 50 14054 13875 10672 38601 NACE 51 58314 38652 33342 130308 NACE 52 41879 18506 73996 134381 NACE 55 26621 12745 19045 58411 NACE 60-63 33968 23518 69792 127278 NACE 64 875 2069 86988 89932 NACE 65-66 3498 11351 47781 62630 NACE 67 1982 2536 1929 6447 NACE 70-74, 90-93 116574 98221 142086 356881 Total 460700 406776 982101 1849577

16

2.2. Nonsampling errors 2.2.1. Coverage errors Sampling frame Sweden has used the Business Register at Statistics Sweden as a sampling frame. This Register contains information on legal ( � enterprises ) and local units. A legal unit is a unit that has the juridical responsibility for its activities. In more than 99 % of all enterprises the legal unit is the same as the enterprise. In case of several legal units in an enterprise, one of them is chosen as the main legal unit and the other legal units are connected to the main legal unit. The Business Register contains all enterprises, authorities and organisations as well as their local units in Sweden that perform any economic activity irrespective whether they belong to the private or the public sector. It contains information about branch of economy, status codes, size class, number of employees, organisation number, legal form, institutional sector, category of ownership, local units, name of the company and firm for individuals, address, telephone number etc. The Business Register is continuously updated and is considered to be of good quality as a frame of enterprises at any time of the year. The National Tax Authority, Swedish Address Changes Ltd. and the Swedish Patent and Registration Office are the main providers of information to the Register, which is updated every two weeks. The National Tax Authority provides information about branch of economy for all new established enterprises. The information received from different data sources is also complemented each spring through a data collection survey addressed to all enterprises with more than one local unit asking for instance about number of employees. In the Register, ”employed” is defined as an individual having a wage/salary sum of more than one Basic Amount per calendar year (= more than SEK 36 400 or EURO 4 100 during the calendar year 1999 at an exchange rate of 1 EURO = 8.9 SEK). The definition of legal units in the Register is as follows: - All legal entities (excluding estates) - Physical persons meeting at least one of the following criteria: - registered for VAT - registered as employer - registered for business tax - having a registered firm - Estates being - registered for VAT and/or - registered as employer and/or - registered for business tax The definition of local units in the Register is as follows: A local unit is the same as the address, the building or the group of buildings close to each other, where the enterprise runs its activity. All active legal units in the Business Register have at least one local unit.

17

In the following conditions are described guiding the registration of the local units in the Business Register. A local unit, in principle, has to fulfil three conditions: 1) There has to be some activity 2) There has to be a place where the activity is going on (geographically defined unit =

address) 3) The activity must be ongoing for some period of time To register another local unit after the first one, there has to be at least one person working half time. If a legal unit has a number of geographically defined units, each of its units will be registered as a separate local unit. Coverage errors in general The Swedish Business Register has been used as a sampling frame in the Swedish CVTS2. This register is continuously updated and is considered to be a good sampling frame of enterprises at any time of the year. Accordingly there is no reason to expect major problems with overcoverage or undercoverage caused by inefficiencies in the sampling frame. It is usually a great turbulence among small enterprises. Quite often small companies go bankrupt or are taken over by other enterprises. That is one of the reasons why we have registered somewhat higher overcoverage for enterprises in the size group 10-49 employees than for enterprises with more than 50 employees (see Table 17 below). The mentioned turbulence in small companies may also to some extent have influenced the results for bigger enterprises. The main reason for that is when a small company is taken over by a bigger enterprise it may be difficult for the new enterprise to be able to tell all the education activities that took place for all the staff in the enterprise just taken over. We may therefore in some cases have underestimates or at least some uncertainties about the education activities involving staff from small enterprises merged during 1999 with other enterprises.

18

Table 17. Ratio of over coverage units to the total number of sampled enterprises by sampled NACE- and size group. Per cent. NACE-group Number of employees Group Total 10-49 50-249 250- NACE C, 10-14 6 0 0 5 NACE D, 15-16 7 0 0 4 NACE D, 17-19 8 0 0 6 NACE D, 21-22 4 3 0 3 NACE D, 23-26 7 0 0 3 NACE D, 27-28 5 0 0 2 NACE D, 29-33 4 0 2 2 NACE D, 34-35 2 0 0 1 NACE D, 20,36-37 4 0 0 2 NACE E, 40-41 6 3 0 4 NACE F, 45 4 2 0 3 NACE G, 50 5 0 0 3 NACE G, 51 6 0 0 3 NACE G, 52 7 2 0 4 NACE H, 55 10 2 5 7 NACE I, 60-63 7 1 2 4 NACE I, 64 3 0 0 1 NACE J, 65-66 10 7 0 8 NACE J, 67 7 4 0 6 NACE K+O 13 1 0 6 Group Total 6 1 0 4 Table 18. Number of observed enterprises by observed NACE- and size group. NACE-group Number of employees Group Total 10-49 50-249 250- NACE C, 10-14 42 12 3 57 NACE D, 15-16 384 105 44 533 NACE D, 17-19 145 44 10 199 NACE D, 21-22 622 218 67 906 NACE D, 23-26 476 189 59 724 NACE D, 27-28 1262 256 49 1567 NACE D, 29-33 1162 386 135 1683 NACE D, 34-35 193 108 41 341 NACE D, 20,36-37 702 203 30 934 NACE E, 40-41 154 72 22 249 NACE F, 45 2351 216 40 2607 NACE G, 50 668 140 26 834 NACE G, 51 2704 364 62 3131 NACE G, 52 1762 277 55 2094 NACE H, 55 1002 194 28 1224 NACE I, 60-63 1525 226 61 1811 NACE I, 64 35 15 18 68 NACE J, 65-66 133 89 26 248 NACE J, 67 87 19 9 115 NACE K+O 4339 1229 301 5869 Group Total 19748 4362 1083 25193

19

Table 19. Ratio of observed enterprises with the same sampled and observed NACE- and size group. Per cent. NACE-group Number of employees Group Total 10-49 50-249 250- NACE C, 10-14 100 100 100 100 NACE D, 15-16 100 87 93 97 NACE D, 17-19 99 94 100 98 NACE D, 21-22 99 87 87 95 NACE D, 23-26 97 94 90 96 NACE D, 27-28 97 88 89 96 NACE D, 29-33 97 93 96 96 NACE D, 34-35 98 94 91 96 NACE D, 20,36-37 98 94 93 97 NACE E, 40-41 99 90 83 95 NACE F, 45 99 80 91 97 NACE G, 50 99 90 52 96 NACE G, 51 98 90 89 97 NACE G, 52 99 62 88 94 NACE H, 55 98 60 100 92 NACE I, 60-63 97 82 75 95 NACE I, 64 94 88 86 90 NACE J, 65-66 96 92 100 95 NACE J, 67 100 100 34 95 NACE K+O 98 71 51 90 Group Total 98 81 78 94 Question A1 about principal economic activity of the enterprise was not included in the Swedish questionnaire. The table directly above therefore only give the ratio of enterprises where observed and sampled size group are the same. 2.2.2. Measuring errors Questionnaire and guidelines to the respondents Only minor difficulties were encountered in translating the common questionnaire into Swedish. In Questions C5a-C5i we had in some alternatives to specify some Swedish providers that should be included in different groups. No alterations were made in sequencing of the questions in the national questionnaire vis à vis the common questionnaire. However, in some cases introductory Yes/No questions were included in order to facilitate for the respondents to answer. No extra national questions were included in the Swedish questionnaire. Already from the start we decided to exclude some of the questions in the common questionnaire from the Swedish questionnaire. The reasons for that were many. For instance we thought that we could get at least as good information about the principal economic activity for the enterprise from our Business Register as was possible to get by asking the enterprises. Separation by sex of the variable A299tot into the variables A299m and A299f we also planned to do with the help of data from our Business Register and our Income Verification Register. Questions A298tot, A2AAm, A2AAf and A2AAtot were also excluded in our questionnaire. For the four last mentioned variables we also planned to use our

20

Business Register and our Income Verification Register instead, something that we however later did not do due to quality reasons. Questions C1 and D1 were not included in our questionnaire due to quality reasons and also to the fact that we already have access to high quality information even for more detailed occupation groups through our ordinary staff training surveys. Unfortunately that information is however not possible to transfer to the different enterprises in our CVTS2-file, as this information is based on a survey to gainfully employed individuals. In order to make it easier for the respondents we only asked for proportions and not for working hours in question C4a-C4k and C5a-C5i. We also did not ask for totals in Questions C2, C3 and C6 for the same reason. Instead we edited 7’s in at least one of the alternatives in the C2, C3 or C6-questions if we only received total values. We did this in order to later on in the imputation phase take this information into consideration. The used questionnaire in the Swedish survey can be found in Annex 2. One of our experts on the use of the English language translated both the common questionnaire and the guidelines into Swedish. Drafts in Swedish of the questionnaire, the guidelines and also different types of introductory and reminder letters were then constructed, checked, corrected and/or modified. All this material was then sent to our Measurement Laboratory for further testing and evaluation. The Measurement Laboratory performed quite extensive testing of all the different materials and presented their results in written form in a separate report. When possible most of all the suggestions presented in that report were later on taken into account when constructing the final versions of the questionnaire, guidelines, introductory letters, reminder letters and instructions to our interviewers or other kinds of personnel used in connection to the work with the Swedish CVTS2. A description of our Measurement Laboratory is given in Annex 4. Below is a summary of the way in which the testing at the Measurement Laboratory was performed and of the findings. The Measurement Laboratory first did some checks of received material, which led to some minor corrections of the wording. The material was also given to an expert for further checks and views. This expert is employed by Statistics Sweden but is not normally working for the Measurement Laboratory. After that testing was done with 6 enterprises of different sizes and in different branches of Industry. The aim of this testing was to discover common difficulties, misunderstandings, indistinct questions and/or guidelines. The testing with selected enterprises was done in the way that the respondents were asked to think-aloud when trying to answer the different questions. The result of the testing was so called ‘think-aloud protocols’ for each interviewed enterprise which was commented on by the interviewer and sometimes also by the leading investigator from the Measurement Laboratory. In retrospective and in view of all the difficulties that has occurred in the Swedish CVTS2, a pilot survey should perhaps also have been performed before starting the data collection. There was however no time for that, because then the start of the data collection would have been delayed. Thereby the received information would also have been influenced by sample information not correctly describing the measuring period decided on in CVTS2.

21

Results from the testing at the Measurement Laboratory In the report from the Measurement Laboratory a lot of valuable information on shortcomings in our questionnaire were given as well as on our introductory letters and also in our guidelines. Concerning our introductory letter we received some criticism for not describing the purpose of the survey well enough. What will our enterprise gain by participating in the survey? some enterprises wondered. In our introductory letter we had given examples of the questions that we thought could be difficult to answer and we had also unwisely mentioned that most of the other questions in general were quite simple to answer. Some enterprises reacted to that and meant that there hardly existed any simple questions in the questionnaire and that even some of the Yes/No questions were quite difficult to answer. Concerning the guidelines it was observed that some enterprises hardly read anything from our guidelines. Quote from one enterprise: I go to the guidelines when it is necessary to do so. Too many definitions and sometimes difficult to find relevant definitions to a certain problem were also some of the comments from the enterprises. In the report from the Measurement Laboratory a lot of general advises and warnings were also given. For instance advice was given on whom to address the questionnaire and on our planned way of data collection from enterprises of different sizes. In consideration of the many difficult questions we were also warned that both the total nonresponse rate and the partial nonresponse rate could be quite high. Adjustments according to the shortcomings described above were made in our introductory letters and in our guidelines. In order not to scare the respondents off directly we also reduced our preliminary guidelines to only four pages. We omitted all the definitions that we thought were quite obvious and some of the definitions were shortened. Most comments and criticism presented in the report from the Measurement Laboratory concerned the questionnaire, however, the wording of different questions. That was not so easy to correct as we did not want to deviate from the common CVTS2 questionnaire. Some of the comments presented in the report to the different questions are presented quite detailed below in order to give an understanding of all the difficulties that later on arose in different parts of the survey. Question A3 It is difficult to give an answer to this question. It is like a puzzle were you have to fit the pieces together. Due to the definitions given in the guidelines, available statistics has to be recalculated. It is necessary to pick out needed information from different systems. Question A4a This question is even more complicated than Question A3. Figures are available in the enterprises’ accounting system but the figures must be adjusted according to the definitions. Question A5b This question is difficult to answer for enterprises, like hotels that are delivering services.

22

Questions A5c and A5d Some kind of re-organisations is regularly taken place. It is difficult to know how big a re-organisation should be in A5c and A5d, respectively. Sometimes there is no clear limit between A5c and A5d. Question B1 Many enterprises mentioned that there is an almost ever-ongoing discussion about needed skills and future composition of the manpower. That is why it is quite easy to always tick the Yes-alternative. Question B2 Many enterprises have in principal routines for regular talks with all their employees concerning development and progress. However, in practise it is difficult to have such talks each year with all employees. Often it is done every other year or even only every third year or so. Accordingly it is quite difficult to find a suitable alternative to tick. The question works better for the managerial level and then it is the first alternative that will be ticked. One enterprise suggested that it should be possible to answer with a percentage share for different staff categories instead. To choose between the different alternatives in this question is like choosing between apples and pears when there is something else that you want. Questions B3a-B3g It was difficult for the enterprises to separate the alternative of recruiting unemployed people from the alternative of recruiting unskilled people. Question B4 What is a written training plan/programme? some of the enterprises asked. How about a catalogue describing available courses? As it is a clear risk that some enterprises will not take time to read the guidelines it is also a clear risk that course-catalogues will sometimes be treated as written plans and sometimes not. Questions B4ya-B4yg The different alternatives can be quite difficult. Where to tick if CVT-activities are taken place when needed? The CVT-activity may have aroused from talks with an employee about development and progress, sometimes ordered from above and sometimes from the employment role a person has etc. The need to further educate and train different staff categories is something that usually has to be budgeted. Where to tick in all these cases? It may be a lot of miscellaneous answers in the alternative “Other”. A spontaneous commentary by one of the enterprises: It would be dreadful if we had vocational training plans/programmes just to obtain certification or to comply with laws. The most important thing must always be to let all the employees grow and be able to improve in their different tasks. Questions B8a-B8g ‘Some employee’ – is it really enough with just one person to tick an alternative was a spontaneous commentary from one enterprise. As it is now it will quite easily be a “Yes” ticked for all the different alternatives whether that type of training has taken place or not. Question B9 The answer will usually be “Yes”. Someone wondered what the purpose was with this question. A ticked “No” should perhaps be followed up with other questions.

23

Questions B11a-B11c All the alternatives for this question are difficult to understand and also difficult to answer. Difficult to keep track of what one is comparing in order to give the right answer and not for instance answer ‘increased’ when it should have been ‘decreased’. Relative- or absolute comparison? What should be compared? Why not ask for proportions instead? Part C in general Always difficult with references to earlier questions. In this case two pages back to Question B8. It is difficult in most surveys to get the respondent to look back to earlier questions. A lot of information often not easily available is asked for in Part C. The question in many cases is whether the required information is possible to get at all. Many enterprises do not have accounting systems containing all the information asked for. C2nm and C2nf Quite easily it will be the number of employees that will be given in this question. Everyone has probably participated in some course. C2tm and C2tf Shear guesses. Even if the requested information may be noted in the time-accounts for an enterprise the noted figures usually are not correct. Several enterprises meant that it was extremely difficult to answer this question. A lot of the following questions in Part C where the answers are dependent on the answer given to this question will therefore also be difficult to give correct answers to. This question is especially difficult for big enterprises. C3i and C3e Very difficult question. One has to keep track of what is taken place internally and what is taken place externally. Even if there exists registers for such things, the registers are often not maintained enough. Thereby they can often not be used for giving a reasonable correct answer to this question. A lot of work is needed in order to be able to answer this question. A forewarning at least one year in advance is needed for us to be able to answer this question, one of the enterprises mentioned. C4a-C4k This question involves a lot of summing-up. The information is also quite often not available. No one will be able to correctly answer this question. The distribution of the answer on 11 different fields of training makes the question extremely difficult to answer. C5a-C5i It is possible to guess but there are too many alternatives. What is the difference between specialised training institutions and private training organisations. (Comment: Specialised training institutions are often private in Sweden). C6a-C6e, C6cn, C6cd and C6sub Here assistance from the economy department is necessary. Difficult questions as the demarkation between different alternatives are not the same as in our accounting system. Business trips are for instance not possible to separate from travels for education and training. C6, C6fa-C6fc, C6g, C6ga-C6gi Alternatives like “No” and “Don’t know” are missing in the questionnaire.

24

C7, C7Ya-C7yf, C7Na-C7nf Unclear to some enterprises whether the questions refer to both internal and external courses or not. C8a1-C8h1, C8a2-C8h2, C8a3-C8h3 One enterprise mentioned that to him the whole question seemed peculiar. He did not tick all the different alternatives. The different staff categories feel strange. Special training courses are sometimes interpreted as meaning by the company itself composed courses. Who belongs to the group “Ethnic minorities”? This alternative sounds exotic. Why not use the word immigrants? (Comment: In Sweden the only ethnic minority group except for immigrants during the last few decades are a few thousand Laplanders in the northern part of Sweden). Employed at risk of losing their job? How to form an opinion of that? Employed without formal qualifications? For work in our company or for any job? If we employ a person he or she has the needed formal qualifications. Difficult to distinguish between general and special training courses. Data collection Most of the data collection work in connection to CVTS2 has been carried out within Statistics Sweden by our own staff of interviewers. Except for a small group of centrally organised interviewers most of them have their own homes as their workplace. Almost all of the interviewers are well experienced and work full time. Statistics Sweden has a nationwide interviewer corps. Education and information to the interviewers was given in written form. Ad hoc information and assistance was given to the interviewers by e-mail or by telephone. A centrally organised group of interviewers and interviewers in the Gothenburg area were also given education and information directly in form of classroom teaching. Sweden performed face-to-face interviews with all enterprises with 250 or more employees according to the version of the Business Register used. This was due to cost reasons, to simplify the survey methodology and to take into account the fact that it would probably be possible to get data with higher quality by face-to-face interviews than with a mail survey. Almost all enterprises with 250 or more employees we thought would probably be ‘trainers’, which in itself can be seen as an important reason for face-to-face interviews for this size group. For all enterprises with less than 250 employees a mail survey was used. 784 gross sample units were included in the face-to-face part of the survey, while 5 000 were included in the mail survey part of the survey. In other words around 14 per cent of the total gross sample units were included in the face-to-face part of the survey. In our introduction letters to the enterprises in both the mail survey part and in the face-to-face part of the survey we informed the respondents that they could always contact us if they needed more information. Names, telephone numbers and e-mail addresses were given in our introductory letters to several persons that they could contact. In the face-to face part it was also possible for the respondents to contact the interviewer herself/himself. In the face-to-face part of the survey the interviewers involved besides our initial written instruction also received extra instructions by mail or by e-mail a couple of times. All our

25

interviewers also had the possibility to contact experts in our main office if they for instance needed clarifications or had come in contact with things not covered in our initial instructions. The mail survey part consisted of the following parts: - A mail survey to 5 000 enterprises with less than 250 employees

according to our Business Register. The material sent out consisted of a questionnaire, guidelines and an introductory letter. The distribution of the material started on the 24th of February 2000, initially planned to start on the 21st of February.

- Mail reminders to 4 309 enterprises on the 15th of March with a new

questionnaire, new guidelines and a reminder letter to all enterprises that as yet had not answered. From the start assumed to be around 3 400 enterprises to remind on the 6th of March with 21st of February as the starting date for distributing the first questionnaires.

- Telephone reminders or interviews were from the start planned with all

enterprises that had not answered before the 3rd of April Assumed initially to be around 2 600 enterprises but was instead around 3 250 enterprises on the 11th of April. Due to the much higher nonresponse rate than expected, we had to send out another mail reminder to 650 enterprises on the 11th of April. Telephone reminders with all the other enterprises that as yet had not answered started also on the 11th of April.

- The goal was initially to reach at least 60 % response rate in this part of

the survey. That was however impossible and in the end we reached around 49 per cent in total response rate including around 4 per cent overcoverage in this part of the survey.

We were able to consider all answers received from this part of the survey up until the beginning of October. The active data collection in this part of the survey, however, ended around the middle of July. Data collection from the end of May until the middle of July mainly consisted of contacts with enterprises in order to correct for errors in the received answers.

The face-to-face part of the survey consisted of the following parts: - This part of the survey consisted of all the 784 enterprises in the sample

with 250 or more employees according to our Business Register. An instruction to the interviewers, questionnaires, guidelines and introduction letters to the enterprises were distributed to our interviewers on the 24th of February. The face-to-face interview was then booked by telephone by the interviewer herself/himself. This was done in order to get in touch with persons at the enterprise responsible for CVT-issues.

- The first contacts with the enterprises were taken around the 3rd of March

followed by mailing the questionnaire, introduction letter and other information material to the contact person/persons given by the enterprise

26

in the first telephone contact with the enterprise. This was done in order to give the respondent time to prepare for the interview.

- Face-to-face interviews with the persons responsible for CVT- issues in different enterprises started around 8th of March. - Face-to-face interviews were also in a few cases performed with a few

bigger local units within an enterprise. The reason for that was that an enterprise at the ”central” level sometimes did not have access to knowledge enough to be able to answer all our questions. In some cases we had to use telephone interviews with the local units instead of face-to-face interviews. Contacts with local units were done for around 40 enterprises and with around 20 of these enterprises telephone contacts instead of face-to-face interviews were performed with the different local units.

- The goal was to reach at least 60 % response rate in this part of the

survey. Due to the importance of high quality data, especially from large enterprises, we did our utmost in trying to get as high response rate as possible in this part of the survey. In the end we therefore succeeded in reaching around 67 per cent in total response rate. The overcoverage in this part of the survey was just three enterprises out of the total of 784 enterprises in the sample.

Reactions of respondents Many enterprises had problems with Questions A3, A4a, A4b and most of the questions in Part C. Some of the enterprises pointed out that it would take them several hours just to answer some of the quantitative questions in Part A and Part C. They therefore often refused to answer especially as we did not have a legal obligation to lean on. Some respondents wanted better clarifications of what to include in Questions A4a and A4b. They wanted for instance references in the guidelines to all main account groups in the EU BAS 99 accounting system valid in Questions A4a and A4b. A lot of enterprises reacted to Question C8 and especially to the alternatives ‘women’ and ‘ethnic minorities’. Some enterprises have written, “Where are the men” all over that page of the questionnaire and refused to give us answers to the question. In Sweden there is only minor differences in the accessibility to CVT or the amount of CVT received by men and women. In fact women usually participate in CVT activities to a somewhat higher extent than men. People aged 50 years or more also participate in CVT activities to almost the same extent as employees in the age group 25- 49. One reason for the reactions to this question may therefore be that they could not see any sense with all or several of the alternatives. The partial nonresponse rate in Question C8 is around 24 %. At least one enterprise has made complaints about the survey to the Board of Swedish Industry and Commerce for Better Regulation (NNR). This authority is safeguarding the response burden on enterprises. Some enterprises suggested that we should ask the employees instead. Several of especially the bigger enterprises pointed out that we had to inform them at least a year and a half in advance that we would come and ask them this kind of questions. In that way they meant that they could adjust their accounting systems and thereby better be able to give us requested information. Some enterprises were unwilling to let the interviewers do face-to-face

27

interviews and they also could not see the meaning with that. On the other hand some enterprises also welcomed the interviewers and thought it to be both interesting and informative to meet our interviewers and discuss CVT issues. Observed problems in connection with the data collection The registered high unit nonresponse rates in many NACE and size groups coupled with frequent item nonresponse or incorrect information in the received answers may be taken as an indicator that the data quality is not so good in the Swedish CVTS2. Below some other problems are described that also were observed and have had an influence on the final data quality. In many cases the problems mentioned in this section were solved by corrections made after re-contacts with the enterprises. Due to the great number of observed inconsistencies and difficulties for the enterprises to give us correct information some of the problems mentioned below have influenced the final data quality. In a few cases the enterprise had another accounting period than the calendar year 1999. In these cases we hade to use information for their last budget year. For instance in these cases we had to accept figures for the period September 1998 - September 1999 instead of for the calendar year 1999 and the number of employees in Question A2 was given as the number of employees at the end of September 1999. As enterprises are the responding units in CVTS2, memory errors are probably not a major reason for measurement errors. The reason for that assumption is that enterprises have registers to consult. However, the information in available registers can be wrong or incomplete. For instance in quite many enterprises it was observed that the enterprise only kept record of the number of course participants and not of the number of persons participating in courses. It was therefore quite early during the data collection period noted that we would probably get an overestimate of the number of individuals in Question C2 compared to the number of course hours in the same question. That is something that later on was confirmed by our comparisons of results from CVTS2 with results from the Swedish Staff Training Survey 1999 (see below in Section 6). Quite reasonably we expected to get an overestimate of the same order of the number of men and women participating in courses due to the fact that many enterprises only keep record of course participants and not of persons participating in courses. We have however observed that the overestimate of women participating in courses compared with the number of course hours given is less for women than for men. One probable reason for that may be that men participate more in expensive external courses that the enterprises keep better records of than internal courses in which perhaps more women than men may have participated. In some cases the respondent could not give us figures for the whole enterprise. Instead we received answers for more or less than we wanted. In some of these cases it was possible to make corrections for that as we knew for what we had been given answers. However, in some cases the respondent may have given us data for only some parts of the enterprise or for more than we asked for without us knowing. The reason for that had to do with the fact that the respondent did not always have a clear understanding of what is included or not in the enterprise where he or she is employed. Our received answers may therefore in some cases not include all work places for an enterprise or we may have received answers for somewhat more than asked for.

28

For 13 respectively 2 enterprises where we know that we received answers for less respectively more than the enterprise asked for, we received information about how many employees it was employed by the enterprise asked for and for how many employees the received answers in different questions represented. In these cases we used this information in order to blow up or reduce the volume received in quantitative questions. For qualitative questions in these cases we tried the get answers that the respondent thought could represent a probable answer for the selected enterprise. The whole questionnaire Question A4b We had some problems with the fact that some enterprises had included business taxes in this question. What we had to do in some of these cases were to impute by using information of average indirect costs from other data sources for all the different NACE groups in CVTS2. Questions B8, C3, C5 and C6 The respondents often could not correctly differentiate between what should be considered as internal or external courses. In some cases the enterprises had considered a course as an external course when the training had taken place outside the company’s premises. Sometimes we were also given inconsistent answers concerning internal versus external courses in B8, C3, C5 and C6. Questions C2nm and C2nf A lot of enterprises did not give us the number of persons participating in courses. Instead we were given the number of male and female course participants without us knowing that. As an individual quite often may have participated in several courses during 1999 we therefore often have too high values in these questions. Questions C2tm, C2tf and C3 and to some extent C6 In several cases the respondents answered with number of hours per participant instead of number of hours for all participants in CVT courses. In some of these cases it was also almost impossible to know with certainty if the given figures were per participant or for all participants. In a few cases the respondents also answered with fees and payments in C6a per participant instead of for all participants. Questions C3, C4 and C6 Quite often we only received answers in these questions for external courses. Many enterprises simply do not keep track of costs for internal training. Questions C4 and C5 In several cases the enterprises were not able to give us the amount of hours devoted to different fields of training. Instead they only provided us with information about the different fields they had had courses in. The same problem we also had in a many cases in Question C5. In a few cases where we could not get the respondents to better specify their answers we

29

simply had to impute by setting the same proportion of training for all the specified alternatives. Questions C6a-C6e and C6sub Due to incomplete accounting procedures in many enterprises we have noticed that many companies have registered fees and payments to external training organisations, travel costs as well as sometimes also costs of premises in Question C6a. In some cases the sum of C6a and C6b have also been registered in C6b. Quite often costs of premises are not registered in the accounting system of an enterprise. A consequence of this is that the costs of premises that we have got in C6e are probably too low. Another consequence of this is that we have clear problems with partial nonresponse in C6e. It is often not possible to say when we have a nonresponse in C6e due to difficulties for the respondent to give us correct cost figures or whether lack of an answer in C6e is due to the fact that the company see access to premises for training as a free resource. It is therefore a clear risk that our imputations procedures may have influenced our results when it comes to costs of premises. Question C6dn In this question it was noted that in some enterprises quite a big share of the staff were involved as part time teachers in courses for their workmates. Questions C6f and C6g In very few cases the enterprises had made contributions to collective funding arrangements. Considerably many enterprises had received receipts or subsidies from different funds or sources. Often in both these questions the respondents were not able to give us the amount of contributed or received money. As a consequence of this our imputation procedures may have influenced our results for those variables. Question E1 A lot of enterprises had missed to tick this variable due to the fact that all the C-questions are between Question B11 and Question E1. Sweden did not include Question D1 in the questionnaire. Another reason for the relatively high partial nonresponse rate in Question E1 is due to the fact that we missed to include a jump instruction in our questionnaire for non-trainers after Question B11. 2.2.3. Processing errors Editing The editing phase of the survey started on the 1st of March 2000 and ended more or less in the middle of July 2000. We did use the codes 1 = ‘Yes’, 2 = ‘No’, 7’s = ‘Only totals specified’, 8’s = ‘Not applicable’ and 9’s = ‘No answer’ in the editing phase instead of 0 = ‘No’, 1 = ‘Yes’, 9’s + 8 = ‘Not applicable’ and 9’s = ‘No answer’ as mentioned in the Code Book. One reason for that was that in Sweden we usually use this kind of coding and we did not want to risk any confusion

30

in the editing phase. We believed that it was a clear risk for mistakes in that someone could by mistake take a zero code as a ‘No answer’. Most of the 8’s and 9’s were not edited in the editing phase but instead were left as zeros to later on be mechanically coded into 8’s and 9’s. 8’s and 9’s were only edited when we thought that it could be difficult to mechanically make a clear demarcation between these two alternatives. Scanning was planned in the beginning to be used for data entry. It was later on however decided that it was better to use just ordinary data computer editing with the software Rode 3.0. In connection to the data entry our staff was given short written instructions on how to edit if the respondents had made certain evident mistakes in their answers. In connection to the data entry, validity tests were made of the edited information. In the editing phase a variable field was filled with 9’s if we lacked an answer. This was done in order to simplify re-contacts with the enterprises and later on also to have a clear mark on every record and variable where we might have to impute. Relevant fields were filled with 7’s if we had received only total number of hours or total costs but not number of hours or costs in all the different alternatives to a question. This was done in order to be able to take into account that kind of information later on for instance in the imputation phase even though we had not included total variables in our questionnaire. Instructions on how to treat answers received only in the form of text strings were usually given orally as the personnel involved in the editing of the questionnaires were our own. Testing and corrections Testing, corrections and re-contacts with the enterprises with the help of personnel from other departments within Statistics Sweden took place more or less in the whole period from April to the middle of July. Further adjustments and corrections of our file by personnel from our own department took place until we delivered our file to EUROSTAT for the first time by the end of March 2001. Testing of registered data was mainly done in a test program based on the rules in the paper with the final checking rules (EUROSTAT (2000a)). The checking rules in our test program were first adapted to the conditions prevailing in Sweden. Only some minor uncertainties arose when we tried to transform certain test limits into suitable test limits for conditions in Sweden. Later on however we had to adjust a few test limits in order not to get too many correct answers marked as possible errors. One reason for that was also that we saw, at a relatively early stage, that we quite simply did not have economic means and personnel enough to deal with all the unforeseeable problems that we got e.g., with the uncommonly high nonresponse rate. Visual Basic 5.0 in combination with Sybase 11.5 was the software used in the test program routines. All possible information was given on the computer screen in order to assist the personnel involved in re-contacts with the respondents in order to correct for missing or incorrect answers. It was possible for the staff involved to make re-tests of individual records immediately after data corrections had been made. Our personnel involved in the work with testing and corrections were given written instructions, which were presented in a short course.

31

The different errors were classified into three different classes depending on how serious an error was considered to be. The most serious errors were marked with red, the next most serious errors with yellow and the least serious errors with green. This was done in order to make it easier for our personnel making the re-contacts with the enterprises. Hereby it was possible to see directly how important remaining errors were by immediately testing a corrected record while still being in telephone contact with a respondent. Around 20 per cent of all registered records were totally correct and around 80 per cent of the records contained one or more items with incorrect or missing data. Of all the records with registered missing or incorrect data around one third had missing or incorrect data in at least four of the most important questions. Our test program functioned very well. We had some problems, though, in getting personnel enough for the necessary re-contacts with the enterprises as we got much more incorrect records than we assumed in advance. We also had time problems with our re-contacts with the enterprises. We simply could not in many cases get in contact with the person or persons that could give us correct information. One reason was that the holiday season had already started for many companies. We also ran out of money due to all our problems with the high unit nonresponse rate, high partial nonresponse rate and all the incorrect information that we had received. Some testing and corrections were also later on done with the help of the software SPSS. This was done in connection to adjustments and calculations that we had to do in our file before we could start with our imputation procedures. Before our imputations we also did some further testing and corrections with Visual Basic 5.0 in combination with Sybase 11.5. One reason was that we had included some new variables in our file that we planned to use in our imputation program. We simply had to check if all the new variables in our file were correct and could be used for our imputations. In connection to that we therefore also created a fourth class of errors only for our new tests. After our imputations we also did some testing and a few manual corrections of our file with Visual Basic 5.0 in combination with Sybase 11.5. The testing program used in this connection was a somewhat reduced version of our initial testing program. Still further checks and a few minor corrections were later on made in SPSS in connection to comparisons made between frequencies in our file of all the different variable alternatives before and after our imputations. After our first delivery of our file in EUROSTAT-format at the end of March 2001 we also had to do a few corrections. The reason for that had mainly to with misunderstandings from our side when to have 9’s +8 or just 9’s in different variable fields. One mistake also depended on a difference between our SNI92 codes and the NACE-codes at four-digit level that we had not noticed. Imputations All the partial nonresponse items and with certainty incorrect answers still existing in our file after the end of all re-contacts with the respondents were imputed except for the variables B11a, B11b and B11c. The reason for not imputing the B11- variables was partly because we missed to include them in our imputation program. However, the main reason was that we had

32

observed a lot of problems with these questions. All the B11-questions were difficult to understand and also difficult to answer. It was difficult for the respondents to keep track of what he or she had to compare in order to give a correct answer and not, for instance, answer ‘increased’ when it should have been ‘decreased’. Many enterprises wondered whether we meant a relative- or an absolute comparison in these questions. As a consequence of all our observed indications of uncertainty with these questions it was decided not to impute partial nonresponse in these questions in order not to make the situation even worse. Before we started our mechanical imputations we had to create several variables and calculate a lot of matrixes for use in our imputation program. The thus created variables and matrixes had also to be checked and sometimes corrected. One special problem arose in this context due to the fact that we in order to make it easier for the respondents had asked only for proportions and not for working hours in Questions C4a-c4k and C5a-C5i. Before creating imputation tables for these variables we therefore had to convert the percentage figures into hours. A procedure that we later on also had to repeat a couple of times because otherwise we would have had to make other changes such as changing some tests in our checking program. This conversion of percentage values to hour and vice versa probably did not introduce any processing errors but we got a lot of extra work not expected in the beginning. In our imputation program we used SQL as the programming language and SYBASE 11.5 as a database. We used mainly three different methods in our imputation procedures: (a). Imputation of quantitative data more or less as described in Norman Davies imputation paper ((EUROSTAT (2000b). (b). Imputation of partial nonresponse in mostly, on other variables dependent qualitative data, by using weighted percentage distributions for different variables and different variable alternatives where it was possible to calculate the known weighted percentage distribution. (c). For more or less on other variables independent qualitative variables, where we had partial nonresponse, we used a method where we created two files, one with records containing no partial nonresponse in the variables that we had to impute in this context and one containing partial nonresponse in one or more of the same variables. Both files were then sorted by the same variable(s) (Stratum or size-group). Imputations in both (a), (b) and (c) were done per stratum (20 NACE by 3 SIZE-groups) if possible (i.e., with enough records with usable answers) and otherwise per size-group only. Which variable code (b) to impute or record to use (c) for the imputations were decided by what random number had been created in connection to every time we had to impute. The created random number was a number between 1-10 000 or between 1 and the number of records usable for imputation in a stratum or a size-group. If in (c) a certain record had been used for imputations, the next record in the file with known values in the same stratum or size group was used for imputation of the next record with partial nonresponse. In a few cases we ran out of records and we came back to the first record used for the imputations in that stratum or size group. If we had to use the same record with known information several times due to a lot of partial nonresponse in a stratum for that variable, we preferred to make the imputations by size-group instead of by stratum.

33

2.2.4. Nonresponse errors The unit nonresponse rate as well as the item nonresponse is quite high for different NACE-groups as well as for different size groups. See Tables 20-29 below. Practically all item nonresponse in the Swedish CVTS2-file has been imputed in such a way as to avoid distortion. Calibration techniques, as described in Section 2.1., have been used in order to compensate for unit nonresponse. Due to, in some cases, quite high number of unit nonresponse in some strata and also quite high item nonresponse for some variables it is a clear risk that nonresponse may have influenced the results. We have, however, in different phases of the survey made our utmost in trying to reduce these errors. The nonresponse rate, as well as the overcoverage rate, is highest for small enterprises, something that is quite evident by the tables below. Face-to-face interviews were performed with all enterprises with 250 or more employees. One consequence of that was that the interest to participate and the engagement from the respondents became higher than it would have been otherwise. As is evident from Table 20 the final nonresponse rate became 51 %, 50% and 35 %, respectively for enterprises with 10-49 employees, 50-249 employees and 250 or more employees. The data collection cost for the face-to-face interview part of the survey was, however, extremely high. The total sample of enterprises with 250 or more employees was just 784 enterprises compared to 5 000 enterprises with less than 250 employees. The data collection costs (excluding costs for re-contacts) was around EURO 85 000 for the postal survey part compared with around EURO 100 000 for the face-to-face part.

34

Table 20. Ratio of nonresponse enterprise to the total number of enterprises in the sample by sampled NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 36 58 33 40 NACE D, 15-16 48 58 29 48 NACE D, 17-19 44 47 11 43 NACE D, 21-22 52 41 31 47 NACE D, 23-26 46 59 36 49 NACE D, 27-28 46 39 35 42 NACE D, 29-33 47 45 31 44 NACE D, 34-35 51 48 33 46 NACE D, 20,36-37 46 41 28 41 NACE E, 40-41 38 23 56 34 NACE F, 45 52 46 37 48 NACE G, 50 61 54 27 57 NACE G, 51 49 59 32 50 NACE G, 52 68 61 39 62 NACE H, 55 58 57 50 57 NACE I, 60-63 57 53 39 53 NACE I, 64 63 44 7 46 NACE J, 65-66 46 50 50 48 NACE J, 67 55 54 50 54 NACE K+O 44 53 40 46 Group Total 51 50 35 48 Table 21. Ratio of item A3 nonresponse enterprises to the total number of enterprises by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 0 0 0 0 NACE D, 15-16 1 3 3 2 NACE D, 17-19 2 0 0 2 NACE D, 21-22 1 4 0 2 NACE D, 23-26 2 1 3 2 NACE D, 27-28 4 3 6 4 NACE D, 29-33 5 4 6 5 NACE D, 34-35 3 2 7 4 NACE D, 20,36-37 0 1 0 1 NACE E, 40-41 0 2 20 3 NACE F, 45 3 2 4 3 NACE G, 50 4 0 0 2 NACE G, 51 5 2 8 5 NACE G, 52 2 3 0 2 NACE H, 55 3 8 38 7 NACE I, 60-63 6 6 0 5 NACE I, 64 15 25 6 14 NACE J, 65-66 8 8 8 8 NACE J, 67 13 11 0 11 NACE K+O 5 3 6 5 Group Total 4 3 5 4

35

Table 22. Ratio of item A4 non-response enterprises to the total number of enterprises by observed NACE- and size group. Percent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 4 0 0 3 NACE D, 15-16 4 3 7 4 NACE D, 17-19 1 0 0 1 NACE D, 21-22 1 2 3 2 NACE D, 23-26 2 3 3 3 NACE D, 27-28 6 1 6 5 NACE D, 29-33 6 4 5 5 NACE D, 34-35 2 2 7 3 NACE D, 20,36-37 2 3 0 2 NACE E, 40-41 2 2 10 3 NACE F, 45 5 2 4 4 NACE G, 50 6 2 15 6 NACE G, 51 2 2 10 4 NACE G, 52 4 0 0 2 NACE H, 55 5 8 25 7 NACE I, 60-63 5 4 6 5 NACE I, 64 8 13 6 8 NACE J, 65-66 6 6 8 6 NACE J, 67 9 0 0 7 NACE K+O 8 3 3 5 Group Total 4 3 5 4 Table 23. Ratio of item B3 nonresponse enterprises to the total number of enterprises by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 0 40 0 6 NACE D, 15-16 6 8 0 5 NACE D, 17-19 9 8 0 8 NACE D, 21-22 10 8 3 8 NACE D, 23-26 4 12 0 6 NACE D, 27-28 12 10 3 10 NACE D, 29-33 6 4 0 4 NACE D, 34-35 5 10 0 6 NACE D, 20,36-37 14 7 10 11 NACE E, 40-41 3 2 0 3 NACE F, 45 13 3 0 8 NACE G, 50 4 11 8 7 NACE G, 51 9 10 0 7 NACE G, 52 17 3 3 9 NACE H, 55 3 15 0 7 NACE I, 60-63 5 2 3 4 NACE I, 64 8 0 0 3 NACE J, 65-66 0 0 0 0 NACE J, 67 9 0 0 7 NACE K+O 7 1 3 4 Group Total 8 7 2 6

36

Table 24. Ratio of item C2ntot nonresponse enterprises to the total number of enterprises providing CVT courses by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 5 0 0 4 NACE D, 15-16 4 10 31 13 NACE D, 17-19 4 5 0 4 NACE D, 21-22 8 13 18 12 NACE D, 23-26 0 12 14 7 NACE D, 27-28 1 4 9 4 NACE D, 29-33 1 18 13 10 NACE D, 34-35 0 8 18 8 NACE D, 20,36-37 6 3 5 4 NACE E, 40-41 3 16 0 8 NACE F, 45 1 2 4 2 NACE G, 50 7 12 15 10 NACE G, 51 4 15 26 12 NACE G, 52 3 3 7 4 NACE H, 55 8 3 0 5 NACE I, 60-63 2 7 15 7 NACE I, 64 33 25 20 25 NACE J, 65-66 13 8 8 10 NACE J, 67 3 33 25 12 NACE K+O 5 14 16 12 Group Total 4 10 14 8 Table 25. Ratio of item C2ttot nonresponse enterprises to the total number of enterprises providing CVT courses by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 15 0 0 11 NACE D, 15-16 10 18 34 19 NACE D, 17-19 12 5 0 9 NACE D, 21-22 11 26 33 21 NACE D, 23-26 11 24 30 20 NACE D, 27-28 6 10 13 9 NACE D, 29-33 12 23 21 18 NACE D, 34-35 8 17 21 15 NACE D, 20,36-37 10 11 21 12 NACE E, 40-41 10 31 10 19 NACE F, 45 6 9 4 7 NACE G, 50 16 21 31 20 NACE G, 51 12 26 36 21 NACE G, 52 11 8 13 10 NACE H, 55 18 13 25 16 NACE I, 60-63 2 12 30 12 NACE I, 64 56 50 33 44 NACE J, 65-66 33 19 31 28 NACE J, 67 13 56 25 23 NACE K+O 14 25 33 24 Group Total 12 19 25 17

37

Table 26. Ratio of item C3 nonresponse enterprises to the total number of enterprises providing CVT courses by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 15 0 0 11 NACE D, 15-16 10 18 34 19 NACE D, 17-19 12 5 0 9 NACE D, 21-22 11 26 33 21 NACE D, 23-26 11 24 30 20 NACE D, 27-28 6 10 13 9 NACE D, 29-33 12 23 21 18 NACE D, 34-35 8 17 21 15 NACE D, 20,36-37 10 11 21 12 NACE E, 40-41 10 31 10 19 NACE F, 45 6 9 4 7 NACE G, 50 16 21 31 20 NACE G, 51 12 26 36 21 NACE G, 52 11 8 13 10 NACE H, 55 18 13 25 16 NACE I, 60-63 2 12 30 12 NACE I, 64 56 50 33 44 NACE J, 65-66 33 19 31 28 NACE J, 67 13 56 25 23 NACE K+O 14 25 33 24 Group Total 12 19 25 17 Table 27. Ratio of item C6 nonresponse enterprises to the total number of enterprises providing CVT courses by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 25 40 0 26 NACE D, 15-16 23 33 59 35 NACE D, 17-19 31 58 57 40 NACE D, 21-22 19 34 53 33 NACE D, 23-26 23 34 59 35 NACE D, 27-28 24 23 44 27 NACE D, 29-33 22 33 57 36 NACE D, 34-35 15 33 64 34 NACE D, 20,36-37 32 28 58 34 NACE E, 40-41 17 37 50 29 NACE F, 45 17 26 33 23 NACE G, 50 36 43 23 37 NACE G, 51 17 43 54 33 NACE G, 52 26 26 40 30 NACE H, 55 33 31 25 31 NACE I, 60-63 16 34 55 31 NACE I, 64 22 50 40 38 NACE J, 65-66 33 25 46 32 NACE J, 67 10 44 75 23 NACE K+O 26 35 59 40 Group Total 23 33 52 33

38

Table 28. Ratio of item C7 nonresponse enterprises to the total number of enterprises providing CVT courses by observed NACE- and size group. Per cent. NACE-group Number of employees 10-49 50-249 250- Group Total NACE C, 10-14 0 40 0 7 NACE D, 15-16 8 10 0 7 NACE D, 17-19 12 11 0 10 NACE D, 21-22 15 11 3 10 NACE D, 23-26 3 13 0 6 NACE D, 27-28 12 9 3 9 NACE D, 29-33 5 5 0 3 NACE D, 34-35 5 8 0 5 NACE D, 20,36-37 10 9 5 9 NACE E, 40-41 3 0 0 2 NACE F, 45 16 5 0 9 NACE G, 50 2 14 8 8 NACE G, 51 11 11 3 9 NACE G, 52 13 5 0 7 NACE H, 55 0 18 0 8 NACE I, 60-63 9 2 3 5 NACE I, 64 11 0 0 3 NACE J, 65-66 2 3 0 2 NACE J, 67 13 0 0 9 NACE K+O 6 1 3 4 Group Total 8 8 2 7 Below is a list describing item nonresponse before imputations for all variables in the Swedish CVTS2 file. All the variables in the list are in the order according to the codebook for CVTS2. Ratios for the variables A1 to B11C are calculated in per cent of the total unweighted number of enterprises in the file delivered to EUROSTAT (2 771 enterprises). Ratios for the variables C1A to C4K and C6A to D1EE are calculated in per cent of the total unweighted number of enterprises that have provided CVT courses in the file delivered to EUROSTAT (2422 enterprises). Ratios for the variables C5A to C5I are calculated in per cent of the total unweighted number of enterprises that have provided external courses in the file delivered to EUROSTAT (2 253 enterprises). Ratios for the variables E1A to E1H are calculated in per cent of the total unweighted number of non-trainer enterprises in the file delivered to EUROSTAT (180 enterprises). Variables without any values were not included in the Swedish questionnaire. Variables with variable names and values in bold in Table 29 are described in more detail in the tables above.

39

Table 29. Item nonresponse for all variables. Variable Number of Ratio of Variable Number of Ratio of item non- item non- item non- item non- response response response response units in % units in % A1 0 0 B8F 5 0 A298TOT - - B8G 3 0 A299M 0 0 B9 195 7 A299F 0 0 B10A 313 11 A299TOT 0 0 B10B 463 17 A2AAM - - B11A 248 9 A2AAF - - B11B 250 9 A2AATOT - - B11C 287 10 A3 102 4 C1A - - A4A 108 4 C1B - - A4B 593 21 C1C - - A5A 204 7 C1D - - A5B 229 8 C2NM 202 8 A5C 170 6 C2NF 198 8 A5D 166 6 C2NTOT 199 8 B1 189 7 C2TM 431 18 B2 186 7 C2TF 417 17 B3 176 6 C2TTOT 414 17 B3A 11 0 C3I 506 21 B3B 11 0 C3E 547 23 B3C 11 0 C3TOT 414 17 B3D 11 0 C4A 403 17 B3E 11 0 C4B 430 18 B3F 11 0 C4C 425 18 B3G 11 0 C4D 444 18 B4 177 6 C4E 379 16 B4YA 7 0 C4F 431 18 B4YB 7 0 C4G 460 19 B4YC 7 0 C4H 423 17 B4YD 7 0 C4I 429 18 B4YE 7 0 C4J 378 16 B4YF 7 0 C4K 384 16 B4YG 7 0 C5A 583 26 B4NA 19 1 C5B 600 27 B4NB 19 1 C5C 610 27 B4NC 19 1 C5D 657 29 B4ND 19 1 C5E 651 29 B4NE 19 1 C5F 610 27 B4NF 19 1 C5G 601 27 B5 187 7 C5H 618 27 B6 185 7 C5I 572 25 B7A 206 7 C6A 822 34 B7B 23 1 C6AE 0 0 B8A 0 0 C6B 765 32 B8B 0 0 C6BE 0 0 B8C 3 0 C6C 234 10 B8D 3 0 C6CE 0 0 B8E 4 0 C6CN 221 9

40

Table 29. Item nonresponse for all variables (continued). Variable Number of Ratio of Variable Number of Ratio of item non- item non- item non- item non- response response response response units in % units in % C6D 991 41 C8D2 57 2 C6DE 0 0 C8E2 52 2 C6DN 924 38 C8F2 39 2 C6E 999 41 C8G2 45 2 C6EE 0 0 C8H2 124 5 C6SUB 730 30 C8A3 414 17 C6F 166 7 C8B3 129 5 C6FE 0 0 C8C3 132 5 C6FA 157 6 C8D3 57 2 C6FB 157 6 C8E3 52 2 C6FC 157 6 C8F3 39 2 C6G 249 10 C8G3 45 2 C6GEST 0 0 C8H3 124 5 C6GA 165 7 D1AA - - C6GB 165 7 D1AB - - C6GC 165 7 D1AC - - C6GD 165 7 D1AD - - C6GE 165 7 D1AE - - C6GF 165 7 D1BA - - C6GG 165 7 D1BB - - C6GH 165 7 D1BC - - C6GI 165 7 D1BD - - C6TOT 791 33 D1BE - - C7 161 7 D1CA - - C7YA 6 0 D1CB - - C7YB 6 0 D1CC - - C7YC 6 0 D1CD - - C7YD 6 0 D1CE - - C7YE 6 0 D1DA - - C7YF 6 0 D1DB - - C7NA 37 2 D1DC - - C7NB 37 2 D1DD - - C7NC 37 2 D1DE - - C7ND 37 2 D1EA - - C7NE 37 2 D1EB - - C7NF 37 2 D1EC - - C8A1 9 0 D1ED - - C8B1 322 13 D1EE - - C8C1 322 13 E1A 42 23 C8D1 323 13 E1B 42 23 C8E1 323 13 E1C 42 23 C8F1 323 13 E1D 42 23 C8G1 323 13 E1E 42 23 C8H1 322 13 E1F 42 23 C8A2 414 17 E1G 42 23 C8B2 129 5 E1H 42 23 C8C2 132 5

41

3. Timeliness The total inflow week by week is described in the table below. The final distribution of normal answers and over coverage is 2 989 normal answers and 218 overcoverage. The reason for the difference compared with the table below has to do with the fact that we later on had to change one of the enterprises from a normal answer to a nonresponse unit due to item nonresponse in almost the whole questionnaire. We also had to change 11 enterprises from normal answers to overcoverage due to the fact that these enterprises either had less than 10 employees at the end of 1999 or because these enterprises did not exist in the calendar year 1999 or belonged to NACE-groups not included in CVTS2. The dates given in Annex 1 describe more or less dates when active work started and ended in the different phases of the survey. Some units where however received even after those dates and thereby some minor activities also went on with for instance data collection even after the in Annex 1 given final dates. This is evident by the fact that we received some answers in week 33 long after all active work with the data collection had ended. Due to the high nonresponse rates we wanted to make use of every available answer possible. Table 30. Inflow of answers per week in the Swedish CVTS2. Week in Normal Over Total Accumulated Accumulated the year answers coverage answers total inflow total inflow 2000 % Week 09 262 15 277 277 4,8 Week 10 238 43 281 558 9,6 Week 11 190 10 200 758 13,1 Week 12 460 42 502 1260 21,8 Week 13 322 25 347 1607 27,8 Week 14 227 8 235 1842 31,8 Week 15 199 12 211 2053 35,5 Week 16 166 10 176 2229 38,5 Week 17 136 5 141 2370 41,0 Week 18 133 3 136 2506 43,3 Week 19 123 18 141 2647 45,8 Week 20 72 8 80 2727 47,1 Week 21 119 4 123 2850 49,3 Week 22 65 1 66 2916 50,4 Week 23 34 2 36 2952 51,0 Week 24 5 0 5 2957 51,1 Week 25 5 0 5 2962 51,2 Week 26 8 1 9 2971 51,4 Week 27 3 0 3 2974 51,4 Week 28 0 0 0 2974 51,4 Week 29 0 0 0 2974 51,4 Week 30 0 0 0 2974 51,4 Week 31 0 0 0 2974 51,4 Week 32 0 0 0 2974 51,4 Week 33 16 0 16 2990 51,7 Total 2783 207 2990 2990 51,7

42

Checking of the Swedish CVTS2 file to some extent went on all the time up until the signing of the confirmation form on the 21st of May 2001 although most of the checking ended on the 13th of February 2001. No CVTS2 statistics for Sweden has as yet been published and will probably not be so until the end of 2001. Dissemination and checking of the received indicative and control tables took somewhat more time than expected due to some uncertainties about what was meant by Eurostat to be included in different columns of the tables. We also had some minor problems with our comparisons of CVTS2 results with results from our ordinary staff training survey. The main reason was that we had to construct full year estimates for year 1999 in order to combine some of the results from our half-year estimates for both half-years in 1999. As the Swedish staff training survey include more or less the whole labour market we also had to make some selections in order to be able to get as close as possible to the population included in CVTS2 before we could make any comparisons of results. 4. Accessibility and clarity Results from the Swedish CVTS2 will probably not be published until the end of 2001. Enterprises and respondents, asking for results during the data collection period, were given some results from our ordinary Staff Training Survey. A four-page brochure containing some results for the period 1986-1998 was sent to a lot of enterprises. We also referred some enterprises to Statistics Sweden’s webb site for results from our Staff Training Survey. 5. Comparability There are only minor deviations between the Swedish questionnaire and definitions compared with the European qustionnaire and definitions except for some questions and definitions not included in the Swedish questionnaire and guidelines to the enterprises. No deviations exist between the European concepts and the concepts used in the Swedish CVTS2. In Questions C5a-C5i we had in some alternatives to specify some Swedish providers that should be included in different groups. No alteration were made in sequencing of the questions in the national questionnaire vis à vis the common questionnaire. However, in some cases introductory Yes/No questions were included in order to make it easier for the respondents to answer. No additional national questions were included in the Swedish questionnaire. We decided already from the start to exclude some of the questions in the common questionnaire from the Swedish questionnaire. There were many reasons for that. For instance we thought that we could get at least as good information about the principal economic activity for the enterprise from our Business Register as was possible to get by asking the enterprises. Separation by sex of the variable A299tot into the variables A299m and A299f we also planned to do with the help of data from our Business Register and our Income Verification Register. Questions A298tot, A2AAm, A2AAf and A2AAtot were also not included in our questionnaire. For the four last mentioned variables we also planned to use our Business Register and our Income Verification Register instead, something that we, due to quality reasons, did not carry out.

43

Questions C1 and D1 were not included in our questionnaire due to quality reasons and also to the fact that we already have access to high quality information even for more detailed occupation groups through our ordinary staff training surveys. Unfortunately that information is however impossible to transfer to the different enterprises in our CVTS2-file as this information is based on a survey to gainfully employed individuals. To facilitate for the respondents we only asked for proportions and not for working hours in Questions C4a-c4k and C5a-C5i. This was something that quite unexpectedly later on led to a lot of extra work. The reason was that we simply had to convert percentage into hours and vice versa several times during different stages in the data processing. For instance, before and after performed imputations and before constructing the final file for delivery to Eurostat. We did not ask for totals in Questions C2, C3 and C6 for the same reason. Instead we edited 7’s in at least one of the alternatives in the C2, C3 or C6-questions if we only received total values. We did this in order to, later on, in the imputation phase take this information into consideration. The deviations between the Swedish survey and the recommendations in different papers from Eurostat described above we think is something that only influence the comparability in the sence that it is impossible to compare results with other countries for questions not asked or included in the Swedish file. The used questionnaire in the Swedish survey can be found in Annex 2. A more complete description of why we made the above mentioned deviations from the by Eurostat recommended approach are given in Section 2.2.2. 6. Coherence Comparisons with the Swedish Structural Business Statistics (SBS) There are quite conciderable differences in some strata between the number of enployed according to SBS and CVTS2, which is evident by the values presented in Tables 31-33 below. The main reason for the differences has do do with the fact that some enterprises has changed from the smallest size group to the biggest. A few enterprises also have changed from the biggest size group to the smallest. Most enterprises in the smallest size group have quite big weights and most of the enterprises in the biggest size group have quite small weights. These both facts together explain most of the differences registered in Tables 31-33. Note that it has been impossible to get SBS data for the NACE groups 65-66 and 67 to the Table 31 below.

44

Table 31. Number of employees according to SBS by NACE- och size group. NACE-group Number of employees Group Total 10-49 50-249 250- NACE C, 10-14 1048 1466 5913 8427 NACE D, 15-16 9890 9986 42653 62529 NACE D, 17-19 3718 3959 6090 13767 NACE D, 21-22 15213 24740 49639 89592 NACE D, 23-26 11904 22737 48565 83206 NACE D, 27-28 28829 26240 53196 108265 NACE D, 29-33 27615 41729 126259 195603 NACE D, 34-35 5117 11560 74781 91458 NACE D, 20,36-37 16797 21308 24360 62465 NACE E, 40-41 4558 8874 13520 26952 NACE F, 45 52096 19632 57863 129591 NACE G, 50 15088 15850 11676 42614 NACE G, 51 63823 42337 35968 142128 NACE G, 52 45734 19796 74538 140068 NACE H, 55 29673 13577 20244 63494 NACE I, 60-63 38576 26786 89377 154739 NACE I, 64 1130 2250 87342 90722 NACE J, 65-66 - - - - NACE J, 67 - - - - NACE K+O 104487 92333 127371 324191 Group Total - - - - Table 32. Number of employees according to CVTS2 by NACE- och size group. NACE-group Number of employees Group Total 10-49 50-249 250- NACE C, 10-14 852 1344 5722 7917 NACE D, 15-16 8104 10099 41278 59481 NACE D, 17-19 3213 4130 5383 12727 NACE D, 21-22 12486 23340 45686 81512 NACE D, 23-26 11002 20671 42989 74663 NACE D, 27-28 26349 23899 49826 100074 NACE D, 29-33 25285 40377 111362 177024 NACE D, 34-35 4853 11796 73996 90644 NACE D, 20,36-37 15044 20125 23255 58424 NACE E, 40-41 3640 7757 13217 24614 NACE F, 45 42275 18400 72539 133214 NACE G, 50 12429 14105 15187 41721 NACE G, 51 57146 36971 36019 130135 NACE G, 52 35299 29353 77361 142013 NACE H, 55 22082 18348 19066 59497 NACE I, 60-63 32921 20855 76482 130259 NACE I, 64 948 1556 94436 96940 NACE J, 65-66 2955 10952 46659 60566 NACE J, 67 1790 1964 13707 17460 NACE K+O 86394 115100 230048 431542 Group Total 405068 431141 1094217 1930426

45

Table 33. Percentage difference between the values in Table 31 and in Table 32. NACE-group Number of employees Group Total 10-49 50-249 250- NACE C, 10-14 19 8 3 6 NACE D, 15-16 18 -1 3 5 NACE D, 17-19 14 -4 12 8 NACE D, 21-22 18 6 8 9 NACE D, 23-26 8 9 11 10 NACE D, 27-28 9 9 6 8 NACE D, 29-33 8 3 12 9 NACE D, 34-35 5 -2 1 1 NACE D, 20,36-37 10 6 5 6 NACE E, 40-41 20 13 2 9 NACE F, 45 19 6 -25 -3 NACE G, 50 18 11 -30 2 NACE G, 51 10 13 0 8 NACE G, 52 23 -48 -4 -1 NACE H, 55 26 -35 6 6 NACE I, 60-63 15 22 14 16 NACE I, 64 16 31 -8 -7 NACE J, 65-66 - - - - NACE J, 67 - - - - NACE K+O 17 -25 -81 -33 Group Total - - - - There are quite big differences in a couple of strata in Table 33. The reason has mainly to do with differences between sampled and observed size group for some enterprises. The most serious effect is caused by the enterprises that have changed from the smallest size group to the biggest size group. The reason is that the weights often are in the order of 1-3 in the biggest size group while the weights in the smallest size group sometimes can exceed 50. In the stratum with the biggest difference in Table 33, for instance, just by treating two enprises as if they had belonged to the biggest size group from the beginning, the difference would only be 24 per cent instead of 81 per cent. From a statistical viewpoint it can sometimes be somewhat dubious to make the kind of corrections of the weights just described. The differences observed in Table 33 will of course have effects on total estimates in some of the strata, but will probably only have marginal effects on totals presented for NACE- and size groups separately. On quota estimates the effects will probably be marginal even if results are presented for NACE- and size groups combined. For the above described reasons, Statistics Sweden desided not to adjust the weights due to the observed differences in Table 33. Comparisons with the Swedish Staff Training Survey It is evident by Tables 34 and 35 below that the number of individuals participating in CVT courses is higher in CVTS2 than in the Swedish Staff Training Survey. The main reason is probably that several enterprises answered with number of course participants instead of with number of individuals participating in CVT courses. Many enterprises simply did not keep records of the number of individuals participating in CVT courses, something that we for

46

instance observed in connection to the data collection. Note also that the figures in Table 35 imply that this problem seems to be a concern for big enterprises only. Note that figures marked with .. in Tables 34 and 36 below are values based on answers from fewer than 20 individuals in the Swedish Staff Training Survey and thereby considered to be too uncertain to publish. Table 34. Comparisons of number of individuals participating in CVT- courses in CVTS2 and The Swedish Staff Training Survey. NACE-group CVTS2 Results from the Swedish Staff Training Survey in per cent of comparable CVTS2 results Men Women Total Men Women Total NACE C, 10-14 5291 563 5854 .. .. .. NACE D, 15-16 19326 11498 30824 131 47 100 NACE D, 17-19 3741 2756 6498 .. .. .. NACE D, 21-22 37184 14236 51420 79 89 82 NACE D, 23-26 33627 17665 51292 70 74 72 NACE D, 27-28 47442 12380 59822 67 .. 62 NACE D, 29-33 84944 26989 111933 84 110 90 NACE D, 34-35 57000 13251 70252 68 .. 69 NACE D, 20,36-37 15679 4631 20310 78 .. 77 NACE E, 40-41 17416 5103 22519 64 .. 76 NACE F, 45 67192 7080 74272 83 .. 86 NACE G, 50 16956 3709 20665 155 .. 146 NACE G, 51 49384 22509 71894 124 125 124 NACE G, 52 23101 58027 81128 89 70 75 NACE H, 55 10786 15832 26618 100 84 90 NACE I, 60-63 40242 21390 61632 99 97 99 NACE I, 64 40771 23412 64183 80 66 75 NACE J, 65-66 22612 29886 52498 110 89 98 NACE J, 67 6182 5747 11929 .. .. .. NACE K+O 152660 120407 273067 81 69 76 Group Total 751537 417072 1168609 86 79 83 Table 35. Comparisons of number of individuals participating in CVT- courses in CVTS2 and The Swedish Staff Training Survey. Size group CVTS2 Results from the Swedish Staff Training Survey in per cent of comparable CVTS2 results Men Women Total Men Women Total 10-49 employees 134509 70526 205036 108 92 102 50-249 employees 150300 78311 228611 109 123 114 250- employees 466727 268235 734962 72 63 69 Group Total 751537 417072 1168609 86 79 83

47

Costs for CVT are measured in the Swedish Staff Training Survey with the help of volume data from the survey to individuals, average salaries and wages for different groups on the Swedish labour market and a cost model for eight different types of costs. Below is a comparison of costs and education hours for CVT-courses in CVTS2 and in the Swedish Staff Training Survey for the CVTS2-population. It is evident that the Swedish cost model give total costs for all NACE groups comparable to the total cost in CVTS2 if consideration is taken to the higher measured number of education hours that one gets by asking individuals rather than enterprises. Not surprisingly comparable total labour costs are lower and total direct costs are higher in the Swedish cost model than in CVTS2. One reason for that, except for the bigger volume one gets by asking individuals rather than enterprises, is that in surveys to enterprises one often miss a lot of internal courses due to the fact that enterprises do not always keep as good record of internal education as they do of external courses. There is reason to assume that on average personell participating in internal courses have lower labour costs than those participating in external courses recorded by the enterprises. Note that the labour costs for course hours are probably also in general underestimated in the CVTS2. According to the Swedish Staff Training Survey it is known that staff with high wages and salaries in general participate more in CVT activities than staff with lower wages and salaries. This fact combined with the fact that labour costs for course hours in CVTS2 are based on average labour costs for all the personell within each enterprise is the explanation for this. There are several reasons for the fluctuation in the figures for different NACE-groups in the table below. Uncertainties both in the CVTS2 and in the Swedish cost model are probably one of the reasons. Problems with both high unit and item nonresponse as well as the many measuring problems in CVTS2 described in this paper are reasons for that. Another reason for the quite big differences in different NACE groups is the fact that the Swedish cost model probably has not taken into account all necessary information in order to get high quality estimates of CVT-costs. The Swedish cost model is also not allocated to show results for different NACE groups. The cost model is designed to show results for individuals in different age groups, of different sex, with different attained education levels and in different labour market sectors like municipalities, counties, government employed, self-employed and the rest of the private sector as a whole. When it comes to hours in CVT-courses there is reason to assume that the Swedish Staff Training Survey give better estimates than CVTS2. In other words there is strong indications that the number of course hours in the Swedish CVTS2 on average is an underestimate in the order of 55%. According to Table 37 below the underestimate in the number of course hours in CVTS2 seems to be bigger for small and medium sized enterprises than for big ones. In Table 38 comparisons has been made for all the size groups that is possible to make comparisons for. Please observe that these intervals are not always the intervals used in the CVTS2. The figures in Table 38 indicate that in general it is quite a big underestimate in CVTS2 in the number of course hours for all enterprises except perhaps for enterprises with 500 or more employees. Note that figures marked with .. in Table 36 below are values based on answers from fewer than 20 individuals in the Swedish Staff Training Survey and thereby considered to be too uncertain to publish.

48

Table 36. Comparisons of costs and education hours for CVT-courses in CVTS2 and the

Swedish Staff Training Survey. Thousand EUROs and per cent. NACE -group CVTS2 Results from the Swedish Staff Training

Survey in per cent of comparable CVTS2 results Labour Labour costs of Direct Total Number costs of Direct Total Number

course par- costs of costs of of course course par- costs of costs of of course

ticipants courses courses hours ticipants courses courses hours NACE C, 10-14 6360 5398 11757 199545 .. .. .. .. NACE D, 15-16 15812 28028 43840 699319 275 240 253 288 NACE D, 17-19 3867 3516 7383 184275 .. .. .. .. NACE D, 20,36-37 11832 18116 29948 545868 206 153 174 213 NACE D, 21-22 43156 42485 85641 1686393 158 389 273 203 NACE D, 23-26 30394 39925 70320 1243463 163 218 194 190 NACE D, 27-28 37816 27464 65280 1785913 57 112 80 62 NACE D, 29, 30-33 100099 118516 218615 4039310 130 242 191 148 NACE D, 34-35 96574 46785 143359 3565894 75 296 147 100 NACE E, 40-41 25389 50974 76363 932186 70 42 52 77 NACE F, 45 45361 47721 93082 1763065 144 347 248 184 NACE G, 50 9648 15981 25629 485832 231 272 257 234 NACE G, 51 52332 70749 123080 1897815 220 227 224 274 NACE G, 52 32387 30919 63306 1640806 84 152 117 89 NACE H, 55 9473 12183 21657 519652 166 577 397 191 NACE I, 60-63 42407 58212 100619 1628321 141 117 127 182 NACE I, 64 40490 85074 125564 2031177 119 91 100 115 NACE J, 65-66 28792 42337 71129 1027380 236 171 197 284 NACE J, 67 11910 25206 37116 644596 .. .. .. .. NACE K, 70-74,90-93 262232 302460 564692 9171340 125 152 140 155 Total 906331 1072049 1978380 35692149 131 187 162 155 Table 37. Comparisons of costs and education hours for CVT-courses in CVTS2 and the

Swedish Staff Training Survey. Thousand EUROs and per cent. Size group CVTS2 Results from the Swedish Staff Training

Survey in per cent of comparable CVTS2 results Labour Labour costs of Direct Total Number costs of Direct Total Number

course par- costs of costs of of course course par- costs of costs of of course

ticipants courses courses hours ticipants courses courses hours 10-49 employees 145020 152533 297553 6136568 159 250 206 174 50-249 employees 149914 191013 340927 5901226 182 246 218 212 250- employees 611397 728504 1339900 23654356 112 159 137 136 Total 906331 1072049 1978380 35692149 131 187 162 155

49

Table 38. Comparisons of costs and education hours for CVT-courses in CVTS2 and the Swedish Staff Training Survey. Thousand EUROs and per cent.

Size group CVTS2 Results from the Swedish Staff Training

Survey in per cent of comparable CVTS2 results Labour Labour costs of Direct Total Number costs of Direct Total Number

course par- costs of costs of of course course par- costs of costs of of course

ticipants courses courses hours ticipants courses courses hours 10-49 employees 145020 152533 297553 6136568 159 250 206 174 50-99 employees 54946 72700 127647 2334170 204 265 239 222 100-249 employees 94968 118313 213280 3567056 170 235 206 206 250-499 employees 76579 99613 176191 3033032 172 290 239 220 500-999 employees 168124 198739 366863 5839797 72 145 112 96 1000- employees 366694 430152 796846 14781527 118 135 127 135 Total 906331 1072049 1978380 35692149 131 187 162 155 7. References Documents in Swedish Statistics Sweden (2001a). Dokument på svenska om CVTS2 AM/UA PM 2001:4 Documents in English Särndal, Swensson and Wretman (1992). Model Assisted Survey Sampling, New York:Springer-Verlag, pp 230-233. EUROSTAT (1999a). Sampling DOCUS/E3/99/CVTS2017 EUROSTAT (2000a). Eurostat Working Papers. Continuing Vocational Survey (CVTS2). Population and Social Conditions 3/2000/E/Nº17 EUROSTAT (2000b). Imputations in CVTS2 (1. Revision) EUROSTAT/E3/2000/CVTS10

50

Annex 1. Time table

Production process Date

(dd/mm/yyy)

Start End

Data collection 24/02/2000 14/07/2000

Distribution of questionnaires 24/02/2000 11/04/2000

Reminders and follow-up 15/03/2000 14/07/2000

Face-to-face interviews 08/03/2000 26/05/2000

Data checking and editing 01/03/2000 14/07/2000

Further validation and imputation 27/03/2000 13/02/2001

Nonresponse survey - -

Estimations 01/11/2000 08/11/2000

Data transmission to Eurostat 22/03/2001 11/04/2001

Dissemination of national results 24/04/2001 21/05/2001

51

Annex 2. The Swedish questionnaire FÖRETAGSUTBILDNING

Företagsutbildning i Sverige 1999 Statistiska centralbyrån (SCB) utför på uppdrag av EU, regering och riksdag en undersökning om personalutbildning. Syftet med denna undersökning är att erhålla information om den utbildning som företagen erbjuder sin personal. Personalutbildning definieras som utbildningsåtgärder eller utbildningsaktiviteter och avser i första hand förvärv av nya yrkeskunskaper eller utveckling och förbättring av redan befintliga yrkeskunskaper. För att räknas som personalutbildning ska utbildningen finansieras helt eller delvis av företaget. Undersökningen inbegriper inte grundläggande yrkesutbildning av det slag som ges t.ex. till lärlingar och praktikanter. Anställda inkluderar: - egna företagare - personer som arbetar regelbundet inom företaget - oavlönade arbetande familjemedlemmar - personer som tillhör företaget och som är betalda av företaget men arbetar på annan plats (t.ex. försäljare, varubud, reparations- och undershållsgrupper) - deltidsarbetande och säsongsarbetare - personer som är frånvarande under en kortare period (t.ex vid sjukledighet eller betald ledighet) Följande ingår inte: - personer som genomför reparationer eller underhåll vid företaget på uppdrag av andra företag - arbetskraft som tillhandahålls och avlönas av andra företag - personer som deltar i allmän värnplikt - lärlingar och andra med ett utbildningskontrakt som deltar i grundläggande yrkesutbildning men som inte täcks av ett arbetskontrakt

52

A UPPGIFTER OM FÖRETAGET A1 Totala antalet anställda i företaget 1999-12-31? Se sid 1! Exklusive lärlingar och praktikanter. .............................. Personer A2 Totala antalet arbetade timmar av anställda i företaget under 1999? Se definition 2! Exklusive lärlingar och praktikanter. .............................. Timmar A3.1 Totala arbetskostnaden för anställda i företaget under 1999? Se definition 3 och EU BAS 99 kontogrupperna 70-76! Exklusive lärlingar och praktikanter. .............................. Kronor A3.2 Hur stor andel av dessa arbetskostnader utgjordes av indirekta kostnader? Se EU BAS 99 kontogrupperna 74-76! Indirekta kostnader består av sociala avgifter, pensionsavgifter, skatter etc. ............. % av totala arbetskostnaden A4.1 Införde företaget under 1999 några tekniskt sett helt nya eller väsentligt förbättrade produkter/tjänster? Se definition 4! 1 Ja 2 Nej

53

A4.2 Införde företaget under 1999 några tekniskt sett helt nya eller väsentligt förbättrade metoder för att producera sina produkter eller tjänster? Se definition 5! 1 Ja 2 Nej A5.1 Blev företaget föremål för sammanslagningar, uppköp eller rekonstruktioner under 1999? 1 Ja 2 Nej A5.2 Påverkades företaget av några andra större organisatoriska förändringar under 1999? Till exempel förändringar i ledningsprocesser, kvalitetskontroll 1 Ja 2 Nej

54

B FÖRETAGETS UTBILDNINGSPOLICY B1 Gjorde företaget under 1999 någon utvärdering av framtida arbetskraftsbehov och/eller kompetensbehov? D.v.s. genom att utvärdera de förändringar som väntas behövas i yrkes- och/eller kompetenssammansättningen bland de anställda. 1 Ja 2 Nej B2 Gjorde företaget under 1999 någon utvärdering av enskilda anställdas kompetens- och utbildningsbehov? 1 Ja, men bara för chefer och arbetsledare 2 Ja, men bara för andra kategorier av anställda 3 Ja, för alla kategorier av anställda 4 Nej B3 Har företaget under den senaste treårs-perioden (1997 - 1999) behövt skaffa eller utveckla ny kompetens? 1 Ja 2 Nej → Gå till fråga B5 B4 Vilka var de tre viktigaste sätten på vilka företaget skaffade eller utvecklade den kompetens som behövdes? Ange högst tre kryss! 1 Redan anställda lärde sig genom att få praktisk erfarenhet i arbetet 1 Redan anställda fick utbildning 1 Nyanställda lärlingar/praktikanter fick grundläggande utbildning 1 Arbetslösa rekryterades och utbildades 1 Förvärvsarbetande som saknade erforderlig utbildning rekryterades och utbildades 1 Personer med erforderlig kompetens anställdes 1 Annat sätt, ange vilket ...............................................................................

55

B5 Hade företaget under 1999 en skriftlig kollektiv yrkesutbildningsplan/program i vilken personalutbildning ingick? 1 Ja 2 Nej → Gå till fråga B7 B6 Vilken var anledningen till att det fanns en kollektiv utbildningsplan? Flera alternativ kan anges! 1 Det fanns en varaktig överenskommelse mellan ledning och anställda 1 Företaget ville få certifiering/auktorisation (t.ex. ISO 9000-certifiering) 1 För att engagera alla ledningsnivåer i företaget till att göra program/planer kända 1 För att efterkomma lagar eller kollektiva avtal 1 I syfte att få ekonomiskt stöd från EU 1 I syfte att få ekonomiskt stöd från andra externa källor t.ex. branschfonder 1 Annat skäl, ange vilket ................................................................................ → Fortsätt med fråga B8! B7 Varför saknade företaget en kollektiv utbildningsplan? Flera alternativ kan anges! 1 Ansågs inte nödvändigt (t.ex. med hänsyn till företagets storlek) 1 Varje anställd bär själv ansvaret för sin yrkesmässiga fortbildning och bör genomföra den på sin fritid 1 Personalutbildning ges bara när anställda eller deras arbetsledare framför önskemål om utbildning 1 Tidsbrist 1 Sakkunskaper som krävs för att utveckla en plan saknas 1 Annan anledning, ange vilken ...........................................................................

56

B8 Hade företaget en särskild utbildningsbudget under 1999, i vilken medel avsattes för personalutbildning? 1 Ja 2 Nej B9 Har företaget ett internt utbildningscentrum som används helt eller delvis för personalutbildning? 1 Ja 2 Nej B10.1 Hade företaget under 1999 någon överenskommelse med några anställda eller deras företrädare som täckte personalutbildning? 1 Ja 2 Nej → Gå till fråga B11 B10.2 Var det ett formellt avtal mellan parterna på arbetsmarknaden? 1 Ja 2 Nej B11 Deltog någon anställd i någon av följande aktiviteter för personalutbildning under 1999? Se definition 6-11! JA NEJ KURSER 1 2 a) Kurser som utformats och skötts av företaget självt (internt anordnade kurser) b) Kurser som utformats och skötts av organisationer som inte ingår i företaget (externt anordnade kurser) ANNAN FORM AV PERSONALUTBILDNING c) Perioder av utbildning, undervisning eller praktik direkt på arbetsplatsen eller i arbetssituationen d) Inlärning genom arbetsrotation, utbyten eller till- fälligt upphållande av annans tjänst e) Deltagande i studie-/kvalitetscirklar f) Självstudier g) Undervisning vid konferenser, workshops, föreläsningar och seminarier

57

B12 Tillhandahöll företaget någon form av personalutbildning som beskrivs i fråga B11 under 1997 eller 1998? 1 Ja 2 Nej B13 Kommer företaget under de närmaste två åren (2000 och 2001) att tillhanda- hålla någon form av personalutbildning som beskrivs i fråga B11? Ja Troligt- Nej vis 1 2 3 a) Personalutbildning i form av kurser b) Annan form av personalutbildning B14 Har det skett någon förändring från början av 1997 till 1999 när det gäller ... Ökat Minskat Oför- Ej till- ändrat lämplig 1 2 3 8 a) Tiden ägnad åt kurser jämfört med tiden ägnad åt andra utbildningsformer? b) Tiden ägnad åt kurser under betald arbetstid jämfört med tiden under de anställdas fritid? c) Den anställdes egen utgift i samband med yrkesmässig fortbildning jämfört med det som betalas av företaget?

58

C PERSONALUTBILDNING I FORM AV KURSER Fråga C1 - C12 avser vanliga kurser som anordnades internt eller externt under 1999. C1 Hur många av företagets anställda deltog i personalutbildningskurser under 1999? Anställda räknas bara en gång, se definition 12! ......................... Män ......................... Kvinnor C2 Hur många kurstimmar har företagets personal fått på betald arbetstid? Se definition 13! ......................... Timmar (Män) ......................... Timmar (Kvinnor) C3 Hur har den betalda arbetstiden för personalutbildningskurser fördelats mellan internt och externt anordnade kurser? Se definition 6! ......................... Timmar (Internt anordnade kurser) ......................... Timmar (Externt anordnade kurser) C4 Inom vilka utbildningsområden/ämnen har kurser getts? Fördela det totala antalet kurstimmar procentuellt på utbildningsområdena! ........ % Språk ........ % Försäljning, marknadsföring, handel, inköp, lager/godshantering, distribution ........ % Redovisning, ekonomi, finansiell verksamhet, bank och försäkring ........ % Ledning och administration, organisationsteori, löner, personalutveckling ........ % Kontorsarbete, sekreterararbete ........ % Personliga färdigheter/utveckling, jämställdhet, fackligt arbete, kvalitetssäkring arbetsliv, gruppsamverkan, certifiering, om företaget/organisationen ........ % Datorteknik/användning, IT, systemering, programmering ........ % Teknik och tillverkning, drift/underhåll av automatiserade system, utveckling ........ % Miljöskydd, arbetarskydd, arbetsplats- och miljösäkerhet, ergonomi, stress ........ % Tjänster, transport, trafik, skydd av person/ägodelar, hotell, restaurang, turism ........ % Övrig utbildning

59

C5.1 Har någon extern kursanordnare anlitats av företaget under 1999? 1 Ja 2 Nej → Gå till fråga C6 C5.2 Vilka externa kursanordnare har anlitats av företaget? Fördela det totala antalet externa kurstimmar procentuellt på olika kursanordnare! ........ % Leverantörer av utrustning ........ % Grund- och gymnasieskolor, komvux, folkhögskola ........ % Högskolor/universitet ........ % Specialiserade utbildningsinstitutioner (t.ex. AmuGRUPPEN AB, AMI) ........ % Övriga privata utbildningsanordnare (t.ex. Utbildningscentrum, Utbildningsstaden, Industriskolan, Lärdata, Antonia, konsulter) ........ % Moderbolag, koncernbolag ........ % Fackföreningar, intresseorganisationer, studieförbund ........ % Handelskammare, branschorganisationer och arbetsgivarorganisationer ........ % Annan anordnare (t.ex. stat, kommun, landsting) C6 Hur stora kostnader har företaget haft för anställdas personalutbildningskurser? Se definition 14-16. Se även EU BAS 99 kontogrupp 7610 för a). Uppskattade värden kan anges för b) till e). a) Kursavgifter för externa kurser och kostnader för .................... kr externa lärare på interna kurser b) Resekostnader och traktamenten för kursdeltagare .................... kr c) Arbetskostnader för interna lärare och övrig personal .................... kr som är helt sysselsatta med personalutbildning → Antal personer helt sysselsatta med personalutbildn. .................... personer d) Arbetskostnader för interna lärare och övrig personal .................... kr som är delvis sysselsatta med personalutbildning → Antal personer delvis sysselsatta med personalutbildn. ................... personer e) Lokalkostnader och kostnader för utrustning och material ................... kr

60

C7.1 Har företaget lämnat bidrag till fonder för personalutbildningskurser? 1 Ja 2 Nej → Gå till fråga C8.1 C7.2 Hur stora var lämnade bidrag? .............................. kr C7.3 Vart gick bidragen? Flera alternativ kan anges! 1 Regionala branschfonder 1 Nationella fonder 1 Andra fonder, ange vilka ............................................................................. C8.1 Har företaget tagit emot bidrag/ersättningar för personalutbildningskurser? 1 Ja 2 Nej → Gå till fråga C9 C8.2 Hur stora var mottagna bidrag/ersättningar? .............................. kr C8.3 Varifrån kom bidragen? Flera alternativ kan anges! 1 Regionala branschfonder 1 Statliga rabatter på utgifter 1 Nationella fonder 1 Skattelättnader 1 EU-källor 1 Privata stiftelser 1 Statliga bidrag 1 Royalty 1 Ersättning för externa deltagare på företagets kurser

61

C9 Har företaget formella rutiner för utvärdering av personalutbildningskurser? 1 Ja 2 Nej → Gå till fråga C11 C10 På vilket sätt har utvärderingen utförts? Ange högst tre kryss! 1 Genom att mäta hur nöjda deltagarna är 1 Genom test för att bekräfta att deltagarna har skaffat sig nya färdigheter 1 Genom en formell validering eller certifiering 1 Genom att mäta om nya färdigheter används i arbetet 1 Genom att använda indikatorer (produktionstider, leveranstider, materialutnyttjande etc.) 1 Andra metoder, ange vilka .......................................................................................... C11 Varför har inte någon utvärdering utförts? Ange det viktigaste skälet! Endast ett kryss. 1 Av kostnadsskäl 2 På grund av tidsbrist 3 Det är svårt att få pålitliga/godtagbara resultat 4 Utvärdering prioriteras inte 5 Utvärdering uppskattas inte av personalen 6 Annan anledning, ange vilken .................................................................................

62

C12 Vilka av följande personalgrupper fanns i företaget under 1999 och i vilken mån deltog dessa i personalutbildningskurser? Personalgruppen deltog under 1999 i Personal- Allmänna ”Skräddar- Ej gruppen kurser sydda” kurser deltagit fanns för denna i kurs 1999 personalgrupp 1 1 2 3

Kvinnor →

Personer 25 år eller yngre →

Personer 50 år eller äldre →

Funktionshindrade →

Etniska minoriteter →

Anställda som riskerar att bli arbetslösa →

Anställda som saknar formella kvalifikationer →

Deltidsanställda → D ANLEDNINGAR TILL ATT INGEN PERSONALUTBILDNING GAVS FRÅGAN SKA BESVARAS OM INGEN PERSONALUTBILDNING GAVS UNDER 1999! D1 Vilka var de viktigaste anledningarna till att företaget inte tillhandahöll någon form av personalutbildning under 1999? Ange högst tre kryss! 1 De anställda har redan de yrkeskunskaper som företaget behöver 1 Företaget rekryterade personer med de yrkeskunskaper som behövdes 1 Kostnaderna för personalutbildning är för höga för företaget 1 De anställda har för mycket att göra, de hinner inte med utbildning 1 Företaget har nyligen gjort en satsning på utbildning, ingen ny behövdes 1999 1 Det är svårt att bedöma vilka behov som företaget har 1 Grundläggande utbildning är tillräcklig för att utveckla kompetens som behövs 1 Annan anledning, ange vilken ....................................................................

63

Annex 3. Guidelines to the enterprises

Företagsutbildning i Sverige 1999

Definitioner 1 Personalutbildning Utbildningsåtgärder eller andra utbildningsaktiviteter som

företagen själva finansierar, helt eller delvis, för sina anställda. 2 Totala antalet arbetade Här avses det totala antalet timmar som alla anställda förutom timmar lärlingar och praktikanter faktiskt har arbetat under 1999. Räkna med: - arbetat tid under normala arbetsperioder - arbetad tid utöver normala arbetsperioder vanligtvis mot högre ersättning (övertid)

- tid som tillbringats på arbetsplatsen i beredskap eller utan att utföra arbete (men för vilken ersättning utgår)

- tid som motsvarar kortare rastperioder (t.ex. kafferaster) Räkna inte med :

tid för betald tjänstledighet, betalda helgdagar, betald sjukledighet, betalda måltidsraster och tid för resor mellan hemmet och arbetsplatsen.

3 Totala arbetskostnaden Beräkningen av den totala arbetskostnaden avser alla utgifter

för anställda som arbetsgivaren har för att kunna ha anställda. Den ska täcka: direkta arbetskostnader som - direkta löner - andra lönetillägg och ersättningar - ersättningar för dagar där inget arbete utförs - naturaförmåner indirekta arbetskostnader som - obligatoriska sociala avgifter och familjebidrag - icke obligatoriska betalningar - andra sociala avgifter - kostnader för yrkesutbildning (brutto) - skatter, med avdrag för bidrag

64

4 Tekniskt sett helt En tekniskt ny produkt är en produkt med tekniska egenskaper

nya eller väsentligt som skiljer sig avsevärt från egenskaper eller användningar hos

förbättrade tidigare framställda produkter. Sådana nyskapande produkter kan produkter/tjänster utgå från radikalt nya tekniker, baseras på en ny kombination av

befintliga från radikalt nya tekniker, baseras på en ny kombination av befintliga tekniker eller härstämma från användning av nya kunskaper.

En tekniskt förbättrad produkt är en befintlig produkt vars prestanda har avsevärt höjts eller uppgraderats. En enkel produkt kan förbättras (i form av bättre prestanda eller till lägre kostnad) genom användning av komponenter eller material med högre prestanda. Eller så kan en komplicerad produkt som består av ett antal integrerade tekniska delsystem förbättras genom delförändringar i något av delsystemen.

En ny eller förbättrad tjänst anses vara tekniskt nyskapande när dess egenskaper och användningssätt är antingen helt nya eller

avsevärt förbättrade kvalitetsmässigt, i form av prestanda eller teknologi.

5 Tekniskt sett helt nya Införande av en ny eller väsentligt förbättrad tjänst/produkt, eller väsentligt produktions- eller leveransmetod kan kräva att en radikalt ny förbättrade metoder teknik/ny kombination av befintliga tekniker används. Tekniken för produktionen av ifråga kräver ofta nya eller förbättrade maskiner, utrustningar produkter eller tjänster eller programvaror. De nya kunskaperna i fråga kan vara ett resultat av forskning eller kan härröra från anskaffandet eller användningen av särskilda färdigheter eller särskild kompetens. 6 KURSER för Kurser som är planerade uteslutande för att ge utbildning och som personalutbildning äger rum på annat håll än på den direkta arbetsplatsen, t.ex. i

särskild utbild-ningslokal. Undervisningen ska vara lärarhandledd under en tidsperiod som bestämts i förväg av kursanordnaren.

Internt anordnade kurser är sådana som utvecklas och genomförs

av företagets självt, fast de kan förläggas till annan lokal utanför företaget.

Externt anordnade kurser är sådana som utvecklas och genomförs av en organisation som inte ingår i företaget även om den genomförs i företagets lokaler. 7 Perioder av utbildning Denna form av utbildning kännetecknas av planerade perioder av

i arbetssituationen utbildning, undervisning eller praktik med användning av normala

arbetsverktyg, direkt på arbetsplatsen eller i den direkta arbets- situationen.

65

8 Arbetsrotation, Arbetsrotation och utbyten med andra företag innebär utbildning utbyten med andra bara om dessa åtgärder planeras i förväg i det uttryckliga syftet att företag utveckla eller förbättra färdigheten hos de berörda arbetstagarna. Normala omplaceringar av arbetstagare från en uppgift till en annan som inte ingår i ett planerat utvecklingsprogram ska inte tas upp. 9 Studiecirklar/ Studiecirklar är grupper av anställda som träffas regelbundet för

kvalitetscirklar i första hand utöka sina kunskaper om de krav som ställs av

arbetsorganisationen/arbetsplatsen. De utgör en form av individuella studier i gruppmiljö. Kvalitetscirklar är arbetsgrupper som organiseras för att diskutera och söka lösningar på problem inom produktionen och på arbetsplatsen. Deltagarna måste ingå i företagets planerings och

kontrollprocedurer. Kvalitetscirklarna samordnas av en ordförande. 10 Självstudier Självstudier kan vara öppna kurser och distansstudier, video/ljudband, korrespondenskurser, datorbaserade metoder eller användning av ett inlärningsresurscentrum. För att räknas som självstudier ska den som genomgår utbildningen själv ordna med tid och plats för studierna. 11 Undervisning på Deltagande i konferenser, workshops, branschmässor och konferenser, föreläsningar/seminarier räknas som utbildningsaktivitet bara när workshops, den anställde deltar främst i utbildningssyfte. föreläsningar 12 Totalt antal En kursdeltagare är en person som har deltagit i en eller flera deltagare kurser för personalutbildning vid något tillfälle under 1999. Varje person ska räknas bara en gång, oavsett hur många gånger hon eller han har deltagit i en kurs. Om en anställd har deltagit i två externt anordnade kurser och en internt anordnad kurs, ska hon/ han räknas som en kursdeltagare. 13 Betald arbetstid Det sammanlagda antalet timmar av personalutbildning avser den totala tid som alla deltagarna sammanlagt har tillbringat i kurser för personal-utbildning under 1999. Om kurser bara delvis aged rum under 1999 ska enbart den del av tiden som inföll under 1999 tas med. Antal timmar ska bara täcka själva utbildningstiden. Eventuella perioder av normalt arbete mellan utbildningstider samt restid etc. tas ej med.

66

Om en utbildning pågår i t.ex. en vecka, ska bara den tid som faktiskt ägnades åt kurs/kursuppgifter ingå. Om en person deltar i en kurs en dag i veckan i flera veckor, så ska dessa enstaka dagar räknas samman. Dagarna av arbete mellan utbildningsdagarna ska inte ingå. Endast tid inom de anställdas arbetstid som ägnats åt utbildning ska räknas (dvs. tid som de anställda tillbringar med utbildning när de normalt skulle arbeta eller tid som företaget betalar ersättning för). 14 Arbetskostnader Arbetskostnader för interna lärare, personal vid utbildnings- för interna lärare centrum och annan personal som är uteslutande sysselsatta med och övrig personal personalutbildning. Här ska tas upp total arbetskostnad för: som är helt sysselsatta med personal- - Interna lärare och personal vid utbildningscentrum

utbildning - Ledare och andra höga chefer som är engagerade i utbildnings-policy

- Handledare och utbildningsledare/handläggare - Kontors- och administrativt stöd för dessa verksamheter 15 Arbetskostnader Arbetskostnader för interna lärare, personal vid utbildnings- för interna lärare centrum och annan personal som är delvis sysselsatta med och övrig personal personalutbildning. och övrig personal Här ska kostnaden för tid som ägnats åt personalutbildning tas som är delvis upp för: sysselsatta med personalutbildning - Interna lärare och personal vid utbildningscentrum

- Ledare och andra höga chefer som är engagerade i utbildnings- policy - Handledare och utbildningsledare/handläggare - Kontors- och administrativt stöd för dessa verksamheter 16 Lokalkostnader Följande kostnader ingår:

och kostnader för - Kostnader för att driva ett utbildningscentrum (exklusive arbets-

utrustning/material kostnader för personalen) eller andra lokaler som används för kurser för personalutbildning - Kostnader för utrustning och material som anskaffas direkt för

kurser för personalutbildning. Här ingår både kapitalvaror och för- brukningsvaror.Vid kapitalvaror ska bara det årliga avskrivnings- värdet för 1999 tas med (kan även avse utrustningsköp tidigare år). Förbrukningsvaror måste behandlas som ”löpande utgifter” och ska därför tas upp till inköpspriset.

67

Annex 4. Sampling Plan

Sampling Plan in CVTS2 for Sweden

Introduction In this sampling plan for CVTS2 in Sweden, we have tried to give a complete picture of how CVTS2 will be carried out in Sweden. References are often given to the Survey Guidelines (EUROSTAT/E3/99/CVTS27). Sampling frame Sweden will use the Business Register at Statistics Sweden as a sampling frame. This Register contains information on legal ( � enterprises ) and local units. A legal unit is a unit that has the juridical responsibility for its activities. In more than 99 % of all enterprises the legal unit is the same as the enterprise. In case of several legal units in an enterprise, one of them is chosen as the main legal unit and the other legal units are connected to the main legal unit. The Business Register contains all enterprises, authorities and organisations as well as their local units in Sweden that carry on any economic activity irrespective of if they belong to the private or public sector. It contains information about branch of economy, status codes, size class, number of employees, organisation number, legal form, institutional sector, category of ownership, local units, name of the company and firm for individuals, address, telephone number etc. The Business Register is continuously being updated and is considered to be of good quality as a frame of enterprises at any time of the year. The National Tax Authority, Swedish Addresschanges Ltd. and the Swedish Patent and Registration Office are the main providers of information to the Register, which is updated every two weeks. The National Tax Authority provides information about branch of economy for all new established enterprises. The information received from different data sources is also complemented each spring through a data collection survey addressed to all enterprises with more than one local unit asking for instance about number of employees. In the Register, ”employed” is defined as an individual having a wage/salary sum of more than one Basic Amount per calendar year (= more than SEK 36 400 or EURO 4 100 during the calendar year 1999 at an exchange rate of 1 EURO = 8.9 SEK).

68

The definition of legal units in the Register is as follows: - All legal entities (excluding estates) - Physical persons meeting at least one of the following criteria: - registered for VAT - registered as employer - registered for business tax - having a registered firm - Estates being - registered for VAT and/or - registered as employer and/or - registered for business tax The definition of local units in the Register is as follows: A local unit is the same as the address, the building or the group of buildings close to each other, where the enterprise runs its activity. All active legal units in the Business Register have at least one local unit. In the following is described conditions guiding the registration of the local units in the Business Register. A local unit, in principle, has to fulfil three conditions: 1) there has to be some activity 2) there has to be a place where the activity is going on (geographically defined unit = address) 3) the activity must be ongoing for some period of time To register another local unit after the first one, there has to be at least one person working on half time. If a legal unit has a number of geographically defined units, each of its units will be registered as a separate local unit. Sample selection The survey is planned to be a stratified random sample of enterprises in the 20 by 3 NACE x Size groups agreed on and according to the description in the document EUROSTAT/E3/99/CVTS017. The Swedish Business Register contains information of good quality that makes it easy to select the enterprises that should be included in the sampling frame for the CVTS2.

69

Regional aspects The survey will cover the whole of Sweden and the survey will be conducted in the same way in all regions. No stratification will be done by geographical regions. Scope of the sample The gross sample will comprise around 6 000 enterprises according to the description given in the document EUROSTAT/E3/99/CVTS017. When allocating the sample for CVTS2 the most current version of the Business Register will be used. New counting and allocation of the number of enterprises in each stratum (20x3, NACE by Size group) both in the Business Register and in the sample will be carried out. In the sampling situation it will be assumed that the maximum length of half the confidence interval will equal to 0.1. A response rate of 60 % will also be assumed.

Questionnaire Statistics Sweden will follow the common questionnaire. However, we will not ask questions where we already have access to information of satisfactory quality from other data sources. According to the document EUROSTAT/E3/99/CVTS25 this implies the following. Question 1.10 will not be asked as we already have access to high quality enough information from our Business Register that we can use instead. Question 1.11 will be asked for the total number of employed at the end of 1999 only. Information on number of employees and the employees’ distribution by gender is possible to get information on with high enough quality from our Business Register and our Income Verification Register. Apprentices and other trainees are rare in Sweden. For that reason it would not be possible to get answers of high enough quality if we tried to ask the enterprises about it. For instance the total number of given journeyman certificates and master certificates in Sweden during 1997 were only 915 and 147 respectively. The reason for including of the question about the total number of employees is due to Netherlands experiences in CVTS1. They found that it was important to include this question in order to be able to check whether given figures by the enterprises entailed the whole company and not for instance only a local unit. Question 2.11 will not be asked. One reason is because it is very difficult for the enterprise to give us an accurate enough answer. High quality information even for more detailed occupation groups

70

is accessible in Sweden for this question from our ordinary staff training survey. However, since our staff training statistics is based on a sample survey of individuals it is impossible to transfer this information to the CVTS2-files for each enterprise. In order to ease the burden on the enterprises we will in some questions ask for total number of hours and ask the respondents only for proportions in different sub groups. After the data collection we will thereby be able to transform the proportions into hours. All other questions will be asked in the manner given in the final questionnaire. The reference period will be the calendar year 1999 as stated in the Survey Guidelines. Statistics Sweden will use its own Measurement Laboratory for testing the questions in the questionnaire. The Measurement Laboratory consists of four to five experts in different fields of statistics who when needed also recruit extra help from other experts and experienced interviewers within Statistics Sweden.

Measurement Laboratory The Measurement Laboratory normally rely on two methods to test questionnaires, Follow-up interviews with probes and Think-aloud Protocols. 1. Follow-up interviews with probes Eight to twelve test persons are recruited. They constitute mostly some kind of quota sample. The test persons fill in the entire questionnaire before being interviewed. The test is planned and analysed by a test leader. Specially trained interviewers monitor each test interview. The interviewer notices the comments and reactions of the respondent and makes notes when the test person has problems. Afterwards one asks general (prepared) probes - and special probes based on what the interviewer observes when the test person hesitates before answering a question, changes the answer or has spontaneous comments to a question. Structured general probes can sometimes be designed in co-operation with the client. This is the way our five or six specially trained ”lab” interviewers work. Test interviews are sometimes made in the locations of Statistics Sweden and sometimes in the office or home of the test person. The ”lab” interviewers report by:

71

• returning the filled out questionnaires with her/his own comments • filling out a special form, detailing the information collected • returning the tapes of the interviews • writing a very brief report about the interviewer’s opinion, the major problems and suggestions for solutions. The test leader summarises the findings and writes a report for the client. It includes the observations and remarks by the test person in full detail.

When testing a questionnaire for a mail survey for firms, questions are often asked while the test person fills in the questionnaire. Definitions are discussed with the respondent. When the respondent gets into difficulties with certain questions, the problem is followed up right away and the respondent is asked how they usually do.

2. Think-aloud Protocols

The procedure is mainly the same as in Follow-up interviews with Probes but the ”think-aloud” process is added. The lab interviewers are not trained for this so these test are performed by the central staff at the Measurement Laboratory. First the test leader demonstrates the idea, then the test person is given a short training and then finally the ”real” questionnaire is tested. The respondents think-aloud while they are answering the questions. Usually the respondent is allowed to go through the whole questionnaire without interruptions. Afterwards the test leader asks the probes e.g., probes like ”Tell me what you were thinking”, ”Tell me why you said that?” 3. Sample selection The tests are often done with relatively small samples. It is seldom random samples. Instead many different types of persons as possible are chosen, for example both men and women, people of different ages, with different educational levels, different occupations and so on. The test leader and the lab interviewers select test persons. The laboratory does not advertise for test persons or ”recycle” them. It sometimes takes a lot of screening to find test persons with the right type of knowledge or experience. So far the laboratory has not had any problems finding people willing to assist. The test persons are not paid but offered lottery tickets as incentives. Even people working in firms are often willing to give us one hour. The test person is informed that their help is needed to improve the questionnaire. It seems that they think it is interesting to tell us about their special problems to fill in the questionnaires and hope it will be easier to fill them in the next time. It happens, but not very often, that persons in small companies say that they do not have the time to spare.

72

4. Reporting A base form is used. One person makes the report, reads the protocols from the interviewers, listen to the tapes if necessary and compiles the information. To get an overview and make it easy to compare different tests a special form is used to summarise the test. It covers one page and gives the customer a simple overview of the test conditions. Data collection Sweden will use Statistics Sweden’s own staff of interviewers1. Except for a small group of centrally organised interviewers most of them have their own homes as their workplace. Almost all of the interviewers are quite experienced and work full time. Statistics Sweden has interviewers in every county and in almost all municipalities. Education and information to the interviewers will be in written form. Ad hoc information and assistance will be given to the interviewers by e-mail or by telephone. The centrally organised group of interviewers will also be given education and information directly in form of classroom teaching. Sweden will perform face-to-face interviews with all enterprises with 250 or more employees according to the version of the Business Register that will be used. This is due to cost reasons, to simplify the survey methodology and to take into account the fact that we probably will be able to get data with higher quality by face-to-face interviews than with a mail survey. Almost all enterprises with 250 or more employees will probably be ‘trainers’ which in itself can be seen as an important reason for face-to-face interviews for this size group. For all enterprises with less than 250 employees a mail survey will be used. Around 800 gross sample units is estimated to be in the face-to-face part of the survey, while around 5 200 will be included in the mail survey part of the survey. This means that it will be face-to-face interviews with around 14 per cent of the total gross sample of enterprises. _____________ 1.) All work with CVTS2 will be carried out within Statistics Sweden. No sub-contracting will be done.

73

The mail survey part will consist of the following parts: - A mail survey to around 5 200 enterprises with a questionnaire. - Mail reminders including a new questionnaire to all enterprises

that as yet have not answered. Assumed to be around 3 400 enterprises. Statistics Sweden believes that the nonresponse rate will be high especially in the mail survey part, mainly due to the fact that the questionnaire is too long and too complex.

- Telephone interviews with all enterprises that have not answered

after the reminder. Assumed to be around 2 600 enterprises. Statistics Sweden’s own CATI1-system will probably be used.

- The goal is to reach at least 60 % in response rate in this part of

the survey. We will try to reach 60 % but it will be difficult with such a long and complex questionnaire.

The face-to-face part of the survey will consist of the following parts: - The face-to-face interview is booked on telephone. This is done in

order to get in touch with persons at the enterprise responsible for CVT-issues.

- Distribution of the questionnaire, introduction letter and other

information material to the contact person/persons given by the enterprise in the first telephone contact with the enterprise. This is done in order to give the respondent time to prepare for the interview.

- Face-to-face interview with the person/persons responsible for

CVT-issues in the enterprise. - If necessary face-to-face interviews will also be performed with

local units in a sub-sample. If it is believed that an enterprise at the ”central” level does not know the total volume of participation in staff training supplementary interviews will be done with a sub-sample of local units.

- The goal is to reach at least 60 % in response rate in this part of

the survey. Due to the importance of high quality data especially from large enterprises we will do our utmost to try to get an even higher response rate in this part of the survey.

____________ 1). CATI = Computer Aided Telephone Interviews

74

Field work The field work will be carried out in Spring 2000 in accordance with the rules given in the Survey Guidelines. Editing and control Data checking will be done after data capture and variable tests, both validity and inconsistency checks between different variables will be done. Statistics Sweden’s own CATI-system will probably be used in the mail survey part with the enterprises that will be interviewed by telephone. It will thereby for these sample units be possible to have editing checks in direct connection to the capture of data. Nonresponse Unit nonresponse As described above concerning data collection, Statistics Sweden will by different reminder procedures try to reach as high response rate as possible in the planned different parts of the survey. No imputation procedures will be used when information still will be totally lacking for a sample unit. Item nonresponse Sometimes it will be necessary to improve on inconsistent or lacking item data. We will in these cases use as much as possible direct methods like calling back to improve on received data. As a last resort, however, we will use different methods for imputation of item nonresponse according to the rules given in the Survey Guidelines. Estimation procedures Estimation and weighting will be done according to the rules given in the Survey Guidelines. The same weight will be used for all variables connected to an enterprise. Standard errors will be calculated, at least for the most important variables, and given in the technical report. Statistics Sweden will use the software CLAN for those calculations. Delivery to Eurostat Transmission of data to Eurostat will be done according to the time-table given in the Survey Guidelines. Sweden does not foresee any problems to follow the time table. Data will be delivered as a micro dataset containing all the information needed for the estimation. A

75

technical report of the Survey will be delivered to Eurostat in connection with the data delivery. Main problems in CVTS2 Statistics Sweden does not foresee any serious problems concerning the time table or informatics. There will be two main problems in CVTS2 according to our experiences: 1. Measurements errors. Some of the questions cannot be answered

correctly by the enterprises, because they do not know the answers.

2. Nonresponse rate. The nonresponse rate will be too high if the

questionnaire is not simplified. The questions on occupation for instance should be excluded from the survey.

3. In a future CVTS3 there must be a regulation in order to make it

possible to put a legal obligation on the enterprises to answer the questionnaires.