Early estimate of Swiss business structure statistics (STATENT-Flash)

Each year, the Structural Business Statistics (STATENT) provide comprehensive and detailed information on the structure of the Swiss economy, focusing on the structural and geographic characteristics of the production system at the level of businesses and establishments and in terms of jobs and jobs in full-time equivalents. The data are widely used for both economic analyses (e.g. market analyses) and spatial analyses (e.g. spatial planning). However, STATENT is published 20 months after the end of the reference year t to which it relates. Data for the reference year 2022, for example, will be available in August 2024. This lengthy time lag is due to the fact that the master data are available approximately 12 months after the reference year and production requires the integration of numerous data sources. This is a prerequisite for guaranteeing the completeness and accuracy of STATENT. 

As part of its innovation strategy, the FSO develops new products, for example by making full use of the information that is already available, thus avoiding the necessity of new surveys. The current project is innovative as it aims to shorten considerably the time needed to make information available compared with the conventional STATENT. It concentrates on the most relevant variables for users in the short-term, i.e. the number of businesses and jobs broken down by size class, economic section (FSO NOGA 50) and canton. 

To produce these results outside the usual STATENT production cycle, it is necessary to use estimates based on prediction methods, which enable missing or incomplete information to be calculated. Ultimately, the aim is to publish short-term structural results (known as STATENT-Flash) 8 months after the end of the reference year. This means that STATENT-Flash data for the reference year 2023 will be available in August 2024. Concurrently with the conventional STATENT, the FSO can thus offer an overall view of the economy's structure as a whole.

The conventional STATENT population is composed of legal units with at least one job in December of the year t. This job must have an annual income above CHF 2300 that is subject to OASI contributions.

Because of the time the OASI funds need to provide their data to the FSO, this population cannot be identified immediately. Two options have been identified to produce the STATENT-Flash.

The first is to predict the legal units that will be present, giving us an estimated population for which jobs are subsequently predicted. There are two models in this option: one for the presence of a legal unit and another for the entity's jobs. This approach is said to be "individual" as modelling takes place at legal unit level. It has yet to be investigated and no results are available for the time being.

The second option tests modelling directly at the level of the most requested aggregations. Here too, there are two models: one for the number of jobs and one for the number of legal units.

The economic sections in the primary sector (NOGA 01 to 03) and the public administration (NOGA 84) are excluded from the aggregated modelling and the results shown here. The number of entities and jobs in these sections are not sourced from the OASI but mainly from administrative sources. These sources are available at very short notice and there is therefore no great need for them in the modelling. The sections will be processed using the individual approach and results will be available in future publications.

Aggregated modelling is based on the development of explanatory variables. In addition to STATENT, these variables come from provisional data from OASI funds, the Business and Enterprise Register (BER) and the Job Statistics (JOBSTAT). The quality of the estimate relies heavily on the variables' ability to explain the number of jobs and legal units. Different modellings were tested and the results assessed on the basis of various criteria (e.g. using root mean squared forecast error (RMSFE)).

Above all, the results shown are used to test the procedure chosen and to assess the extent to which the results correspond to users' needs. As the estimates are experimental, caution is advised when interpreting results. In particular, the following limitations should be considered. It was not possible to produce a confidence interval meaning that the uncertainty of the estimates has not been determined. In addition, not all model assumptions have been tested and biases may remain. Tests on the years 2019 to 2021 have shown that there are sometimes considerable differences between the prediction of the aggregated method presented here and the actual STATENT. It is worth mentioning that these years were particular as Switzerland was facing the challenges of COVID at the time. These initial experimental results indicate what needs to be done so that the estimates meet the high standards usually applied by the Federal Statistical Office.

The tables show, for both models used (one for jobs and one for legal units), the results aggregated by size.

The predicted number of businesses was 605 778 in 2025. 0.3% of businesses had more than 250 jobs (large enterprises). The canton of Zurich was the only one with more than 100 000 businesses. At the other extremity, the canton of Appenzell Innerrhoden had 1326 businesses. "Human health activities" (NOGA 86) had the largest number of businesses (72 150), followed by "other service activities" (NOGA 94–96: 60 674). Only 237 businesses were active in "mining and quarrying".

The predicted number of jobs exceeded 5 million in 2025. Almost 36% of jobs were in large enterprises. Like the number of businesses, the cantons of Zurich and Bern had the most jobs, with 1 085 597 and 624 404 respectively. "Human health activities" (NOGA 86) had the most jobs (480 478), followed by "retail trade" (NOGA 47: 335 181).

STATENT provides information on the whole of the Swiss economy. In the future, STATENT-Flash should also provide results for the primary sector sections (NOGA 01 to 03) and the public administration.

Given that the aggregated approach is currently unable to produce sufficiently robust and accurate results, instead of trying to improve it, we have decided to focus on the individual approach. This approach offers the advantage of incorporating a larger variety of data sources (and therefore of explanatory variables), which should potentially lead to improved estimates. The models developed in the future will always be compared with the ones shown here, in the hope that the results of these experimental statistics will be improved with the help of future work.

Email: statent@bfs.admin.ch 
Tel.: +41 58 463 62 66 

We welcome any questions, ideas and comments on the STATENT-Flash. Please send them to the information service of the Economic Structure and Analyses Section (WSA).