INNOVATION SURVEY RESULTS Statistical Bulletin RĪGA, 2003 ANNEX Quality Report on the Survey of Innovative Activities Sampling frame For the formation of the sample the Business Register of the Central Statistical Bureau of the Republic of Latvia was used. The Business Register holds information on all those enterprises and organisations in Latvia that are economically active regardless whether they operate within the private or public sector. The register information contains a unique identification code, the name of the enterprise, its main kind of economic activity (last year and during the last reporting period), net turnover in the last two years, common statutory capital, the code of activity status, etc. The Register is continuously updated. For the building of the sample the code of activity status is of great importance since it allows selecting only the economically active enterprises, which represent less than 50% of all registered enterprises. The Enterprise Register of the Republic of Latvia and the State Revenue Service are the main source of information for the CSB Business Register. The statistical Business Register is also updated by means of a special register survey as well as by drawing information from various CSB enterprise surveys. All newly created enterprises are surveyed after a period of six months since the day of their registration. Sample design The sample of the Latvian innovation survey was created as a stratified simple random sample. All enterprises that employed not less than 10 persons were divided into 10 strata (typological groups) on the basis of three indicators characterising the enterprise: the NACE group (the code describing the main line of economic activity of enterprise), the size group of enterprise and the location (registration) of the enterprise (in Riga or the rest of Latvia). According to the methodological recommendations CIS3, the survey covered all those enterprises in Latvia that employed at least 10 employees and whose main activity was characterised by codes 10 - 41, 51, 60 - 67, 72 - 73, 742 or 743. The total number of such enterprises is comparatively small (only 4398 enterprises) and in total 278 strata were formed. Strata were formed according to the description given in the Eurostat document CIS3: 39 NACE groups were established as well as 5 size groups depending on the number of employees (10 - 19; 20 - 49; 50 - 99; 100 - 249; 250 and more) and 2 territorial groups. Information on the number of enterprises within the population broken down by NACE groups and size groups are given in Table 1 and Table 2, respectively). Table 1. Sampled enterprises by economic activity, 1999-2001 NACE 10 - 14 15 - 16 17 - 19 20 - 22 23 - 26 27 - 28 29 - 33 34 - 35 36 - 37 40 - 41 51 60 - 63 64 65 - 66 67 72 73 742 743 Total Number of enterprises 27 285 188 412 121 102 115 43 110 82 411 334 31 61 15 51 34 49 20 2491 % 1,08 11,44 7,55 16,54 4,86 4,09 4,62 1,73 4,42 3,29 16,50 13,41 1,24 2,45 0,60 2,05 1,36 1,97 0,80 100 Table 2. Enterprises by size group at the moment of sample formation Number of employees 10 - 19 20 - 49 50 - 99 100 - 249 ≥250 Total Number of enterprises total in the sample 2039 733 1398 797 505 505 299 299 157 157 4398 2491 Responded Did not respond 570 731 469 296 157 2223 163 66 36 3 0 268 Sampling fraction (%) 36 57 100 100 100 57 2 Estimation procedures For the estimation of the population and domain totals the Horvitz-Thompson estimators were used. The weighting coefficients were calculated as a ratio between the total number of enterprises and the number of responding enterprises in each stratum. The sample of the 3rd, 4th and 5th size group included all enterprises, however, not all enterprises responded. Thus in some strata within the 3rd and 4th size groups the weighting coefficients of the enterprises exceed 1. Sampling errors of estimates Variance and the coefficient of variation of an estimate were calculated following the Horvitz-Thompson formulas using SPSS (version 8.02) and SUDAAN (version 7.5.2) software. The Taylor linearisation option was used for the SUDAAN software. The estimated coefficient of variation of the Boolean variable, which shows the presence of an innovative activity in the enterprise, is 3,41%. The estimated coefficient of variation of the amount of resources invested in innovation, is 2,84%. Frame imperfections Overcoverage Of the 2491 sampled enterprises, 2290 enterprises submitted returns although after data collection it turned out that in 67 enterprises the number of employees was lower than 10 so they should not have participated in the survey. The number of the actually valid questionnaires was 2223. For this reason the overcoverage is 2,9%. Classification errors True misclassification occurred in 10 enterprises where some stratification characteristics were not determined correctly: for 7 enterprises classification errors were made both concerning the territory and the NACE code group, for 2 enterprises – only the territory and for 1 enterprise – only NACE code group. True misclassification in this survey is therefore 0,45%. Pseudo-misclassification arises with variables of high variability, as the size group of employment. The larger the number of group, the greater the difficulty to maintain the register up-to-date and the higher the probability of pseudomisclassification occurring to this variable. There are 188 cases (8,5%) of such enterprises. Undercoverage Undercoverage implies a situation when for some reasons a part of the surveyed enterprises is not included in the sample frame. All economically active enterprises that employed not less than 10 persons were included in the sample frame. Two main reasons explaining undercoverage may be identified. Firstly, an economically active enterprise in the business register may be misclassified as economically inactive thus as a consequence it is not included in the sample frame. Secondly, the number of employees of an economically active enterprise may be 3 indicated as lower than 10 while the actual number of employees in this enterprise is higher than 10. The size of undercoverage related to the first reason was estimated during the register survey that was conducted by the CSB of Latvia. This survey shows that among the enterprises that in the business register are classed as economically inactive, less than 7% are in reality economically active. It should be emphasised that in the majority of cases the number of employees in enterprises that are erroneously classified as inactive ranges from 0 to 5. It is only very rarely that the number of employees in such enterprises was higher than 10 therefore the influence on the survey of undercoverage associated with the first reason should be ignored. It is comparatively difficult to estimate the undercoverage in relation to the above-mentioned second reason. Non-response Table 3. Response rate NACE 10 - 14 15 - 37 50 - 52 60 - 64 65 - 67 70 - 74 Total 10 - 49 90,00 85,49 85,42 86,44 71,79 81,42 85,03 Number of employees 50 - 249 ≥250 100,00 100,00 95,55 100,00 96,61 100,00 97,09 100,00 82,61 100,00 77,50 100,00 94,68 100,00 During the data collection stage no efforts were spared to attain a high response rate. Special attention was paid to the acquisition of data from the large (with more than 250 employees) and medium-sized (50 - 249 employees) enterprises. The total response rate was 82,9%. The response rate for the large enterprises was 100%, for the medium-sized it was 94,68% and for the small enterprises (10 - 49 employees) – 85,03%. 4