Is training apprentices worthwhile? Company perspectives on costs, benefits and training quality 2025

Alexander Gehret, Andreas Kuhn, Jürg Schweri IS TRAINING APPRENTICES WORTHWHILE? COMPANY PERSPECTIVES ON COSTS, BENEFITS AND TRAINING QUALITY 2025 SFUVET Swiss excellence in vocational education and training SWISS FEDERAL UNIVERSITY FOR VOCATIONAL EDUCATION AND TRAINING EXECUTIVE SUMMARY 2 INTRODUCTION 3 1 WHY CONSIDER COSTS AND BENEFITS IN VOCATIONAL EDUCATION AND TRAINING? 4 1.1 Strategic considerations for training companies 4 1.2 Cost–benefit studies generate insights for steering the system 5 2 THE COST–BENEFIT MODEL 7 2.1 Gross costs 7 2.2 Productive output 9 2.3 Net benefit 10 2.4 Benefit of continued employment (recruitment opportunity returns) 11 3 SURVEY PROCESS 13 3.1 Revision of the survey framework 13 3.2 Pretest 15 3.3 Sampling 15 3.4 Field phase 17 3.5 Responses in the two subsamples 17 3.6 Composition of the two net samples 18 3.7 Plausibility checks 21 3.8 Approach to missing variable values 22 3.9 Sample weighting 22 4 RESULTS 23 4.1 Average costs and benefits of initial vocational education and training 23 4.2 Costs and benefits of the two-year initial vocational education and training leading to an FVETC 25 4.3 Costs and benefits of the three-year initial vocational education and training leading to an FVETD 28 4.4 Costs and benefits of the four-year initial vocational education and training leading to an FVETD 30 4.5 Results by company size 33 4.6 Results by sector 34 4.7 Results by major region 35 4.8 Results by occupation 36 4.9 Findings on the benefits of continued employment 39 4.10 Extrapolation of training costs, productive output and net benefit 39 4.11 Comparison with Germany 40 4.12 Conclusion 41 5 MULTIVARIATE ANALYSES 42 5.1 Analysis of the 2022/23 survey 42 5.2 Multivariate comparison with the survey for the 2016/17 apprenticeship year 44 CONTENTS Imprint Editor: Swiss Federal University for Vocational Education and Training SFUVET Suggested citation: Gehret, A., Kuhn, A. & Schweri, J. (2025). Is training apprentices worthwhile? Company perspectives on costs, benefits and training quality 2025. Zollikofen: Swiss Federal University for Vocational Education and Training SFUVET. November 2025. All rights reserved. Reproduction or use of content only with permission from SFUVET Version: 1.0 Contact: Swiss Federal University for Vocational Education and Training SFUVET, Jürg Schweri and Alexander Gehret, Kirchlindachstrasse 79 CH-3052 Zollikofen, +41 58 458 27 82 and +41 58 458 27 88, [email protected], www.sfuvet.swiss/cost-benefit 6 TRAINING PLANS AND THE VOCATIONAL BACCALAUREATE 46 6.1 VET ordinances and training plans 46 6.2 Federal Vocational Baccalaureate 1 from a company perspective 48 7 IN-HOUSE TRAINING QUALITY 50 7.1 What is the definition of in-house training quality? 50 7.2 How do training companies assess their training quality? 50 7.3 Five training types of training companies 53 7.4 How do the companies in the different training types differ? 55 7.5 Why is training quality important? 56 8 REASONS TO TRAIN (OR NOT) 57 8.1 Companies that do not provide apprenticeship training 58 8.2 Comparison of reasons for and against training 58 REFERENCES 62 2 Company perspectives on costs, benefits and training quality 2025 EXECUTIVE SUMMARY comprehensive occupational skills among apprentices that go beyond the needs of individual companies. Conversely, 19 per cent of companies report that they provide additional qualifications not included in the training plan. – The Federal Vocational Baccalaureate Type 1 taken alongside the apprenticeship (BM 1) results in slightly lower productive output for companies due to apprentices’ increased absences. This reduces the net benefit for companies by around CHF 3,000, but it remains positive. – Training companies were asked eight questions about the quality of their training and, based on their responses, were grouped into five training types. Overall, more than three quarters of training companies belong to the three types that provide satisfactory to very good training quality. A quarter of companies provide excellent training quality. A third of companies offer average quality training. Nineteen per cent offer training on a more ad hoc basis with little planning, but still achieve average to good quality results. The fourth group, described as ‘authoritarian’ in their training approach, makes up 8 per cent: these give apprentices little scope to find autonomous solutions and pay less attention to avoiding potential overload. Almost 15 per cent of companies rate their training quality as below average across all dimensions. – For training companies, the strongest motivation for offering training is the sense that it is a collective responsibility of the business community and therefore a contribution to society. At the same time, economic motives have gained in importance compared with the study conducted six years earlier: the second most common reason cited is the wish to train skilled workers who can be retained in the company in the long term. Apparently, the shortage of skilled labour is strengthening this motive for offering training. Companies also place greater importance on being able to make use of apprentices as part of their workforce during the training period. – The online survey produced a dataset with information from 6,655 training companies and 3,655 non-training companies. The response rate among companies randomly selected from the Federal Statistical Office’s business register was 39 per cent for training companies and 31 per cent for non-training companies. – Across all apprenticeships, the average net benefit in the training year 2022/23 was just over CHF 4,500 per apprenticeship year and place. A majority of apprenticeships (around 71 per cent) therefore achieved a net benefit for the training company. The net benefit is defined as the value of apprentices’ productive work during the training period (‘productive output’) minus all training costs (‘gross costs’). – This positive average result is consistent with companies’ own assessments, with 80 per cent reporting that they are fairly or very satisfied with the cost–benefit balance of training their apprentices. – The results for the twoand three-year apprenticeships do not differ significantly from those of the previous cost–benefit survey in the 2016/17 training year. The average net benefit per apprenticeship is CHF 9,630 for the Federal Vocational Education and Training Certificate (FVETC) and CHF 13,940 for the three-year Federal Vocational Education and Training Diploma (FVETD). The net benefit of CHF 17,510 for four-year FVETD apprenticeships is significantly higher than in 2016/17. This is partly due to apprenticeships that switched from a threeto a four-year duration. In addition, the survey questionnaire has been simplified and improved since the last survey, which may also have contributed to the change in results. – Companies generally find that the VET ordinances (BiVo) and training plans (BiPla) correspond well to their needs. On average, companies consider 82 per cent of the training content defined in the BiPla to be relevant for their operations. This is a high proportion given that these documents are intended to ensure Company perspectives on costs, benefits and training quality 2025 3 Initial vocational education and training (IVET) is based on companies’ willingness to train young people – and also adults – in an apprenticeship. For this willingness to be maintained, companies must see IVET as worthwhile. The State Secretariat for Education, Research and Innovation (SERI) commissioned this study to provide an upto-date picture of company costs and benefits and the factors influencing them. The Swiss Federal University for Vocational Education and Training (SFUVET) was awarded the contract to conduct the survey through a WTO public procurement procedure. This report outlines the concept, planning and field phase of the survey, as well as the data processing and main findings. The fifth survey on the costs and benefits of initial vocational education and training builds on a series of earlier surveys. The basic methodology of the previous cost– benefit surveys has remained largely unchanged since the first survey in the 2000/2001 training year (Schweri et al., 2003). To reduce the burden on participating companies and improve data quality, the online questionnaire for the current survey was completely overhauled. Nevertheless, the results of the various surveys described in Chapter 1 remain largely comparable. With an increased response rate, this study is based on data from more than 6,600 training companies. To make the results more accessible, findings on individual apprenticeship occupations are provided in dedicated chapters on the SFUVET website. https://sfuvet.swiss/cost-benefit Since the report replicates many of the core results of the previous survey from the 2016/17 training year (Gehret et al., 2019) using newly collected data for 2022/23, it does not attempt to reformulate any arguments that have remained unchanged. This makes it easier for readers to compare the results of the two reports. At the same time, the current report also presents many entirely new findings, including Chapter 7 on in-house training quality. The authors would like to thank all participating companies and institutions involved in the survey, in particular Jürg Bieri and Philipp Theiler from SERI, and Alain Chassot, Natalia Dorontsova, Laurent Inversin and Frédéric von Kessel from the Federal Statistical Office (FSO). They are also grateful to Samuel Mühlemann, Harald Pfeifer, Gudrun Schönefeld and Felix Wenzelmann for their valuable input in revising the questionnaire. Last but not least, the authors would like to thank the SFUVET staff involved in the study, especially Yolanda Hübsch, Ivana Lovric, Tania Pichierri, Filippo Pusterla, Lea Rathlef and Matilde Wenger, as well as assistants Anaclet Deguenon, Esperanza Marx, Lorenzo Peduzzi, Lea Raas and Julia Stadler. INTRODUCTION 4 Company perspectives on costs, benefits and training quality 2025 Information on the costs and benefits of IVET from the perspective of companies is relevant for different target groups: firstly, for the companies themselves, and secondly for all stakeholders in the VET system who depend on companies’ voluntary participation in training. The following section therefore starts by outlining the relationship between training strategies and costs and benefits at company level. It then describes how cost–benefit data can contribute to steering the VET system. The available literature on labour market and education economics analysing what motivates companies to train apprentices is extensive. This chapter provides only a brief introduction and does not attempt a comprehensive review. For German-language overviews, see Wolter (2008) and Strupler & Wolter (2012); for English-language overviews, see Mühlemann (2019) and Mühlemann & Wolter (2014). 1.1 Strategic considerations for training companies From a business perspective, apprenticeship training should be worthwhile for companies. A range of cost and benefit factors come into play, but these are not always easy to identify or quantify, since training is rarely presented in full in company accounts. With the help of the cost model developed for this survey (see Chapter 2), companies can carry out their own cost–benefit calculation. The survey results also show companies the average costs and benefits observed among other companies in Switzerland, including in comparison with companies in the same sector or offering the same apprenticeship. This allows them to benchmark their own training figures against the survey averages and better assess the cost-efficiency of their training. But the issue is not simply whether a company does ‘better’ or ‘worse’ by showing lower or higher net costs. Cost–benefit results reflect differences in the training strategies and organisation chosen by each company. Depending on the objectives a company pursues with training, the cost–benefit balance will vary. The crucial question is whether the way in which training is organised enables companies to meet their business goals while at the same time providing apprentices with high-quality training and sound preparation for their careers. Cost–benefit analyses give training managers an opportunity to reflect on and improve their training practices accordingly. Companies must constantly decide which employees with which qualifications they need to produce their goods or services efficiently. In Switzerland, most companies employ skilled workers with either a Federal Vocational Education and Training Diploma (FVETD) or a Federal Vocational Education and Training Certificate (FVETC). They must decide whether to recruit skilled workers with IVET from the external labour market or (also) to train them in-house. This decision is complex because many relevant factors are uncertain. Apprenticeships last two to four years, and apprenticeship contracts cannot easily be terminated. Hiring apprentices, therefore, represents a multi-year commitment for companies – one that carries risks, since future product demand, staffing requirements and apprentice performance cannot be predicted with certainty. Companies must therefore weigh the expected costs and benefits of training against these risks. Costs include the staff costs arising from training and supervising apprentices, apprentice wages, and all training-related material costs and fees. These costs are offset by various benefits. During their training, apprentices work productively in the company and can perform tasks that would otherwise have to be carried out by other employees. These may be skilled workers but also semi-skilled or unskilled employees. Since apprentice wages are lower than those of these categories of employees, the productive output of apprentices alone can generate sufficient benefits to cover all training costs. Another benefit is that apprentices can be employed as skilled workers after completing their apprenticeship. This benefits the company because the apprentices are already well integrated into the company, whereas skilled workers from the external labour market must first be recruited, hired and subsequently trained. Keeping apprentices on allows companies to avoid these costs. In addition, training companies are able to assess the work quality and productivity of their apprentices 1 WHY CONSIDER COSTS AND BENEFITS IN VOCATIONAL EDUCATION AND TRAINING? Company perspectives on costs, benefits and training quality 2025 5 and can selectively retain those who have performed particularly well. This involves less risk than recruiting skilled workers from the labour market. Skilled workers trained in-house also often show greater loyalty to the company, which reduces turnover costs (Euwals & Winkelmann 2004). From an economic theory perspective, the benefit of retaining apprentices lies in the fact that those trained in-house provide companies with higher productivity, but compared with skilled workers from the external labour market, this is not – or only partly – reflected in higher wages. Training companies therefore earn a ‘rent’ on their former apprentices. The relevant literature (e.g. Acemoglu & Pischke, 1999) examines the conditions under which companies have more or fewer opportunities to earn such rents on the labour market. Apprenticeship training also brings other advantages that benefit companies directly. Not only can it boost their reputation with customers and (potential) employees, but it can even contribute to knowledge transfer and innovation (Backes-Gellner & Lehnert, 2023). Apprentices may pass on ideas they have picked up in inter-company courses or at vocational school, whether from teachers or classmates. Training also keeps companies more closely connected to their trade association or professional organisation, strengthening communication across the sector and encouraging knowledge exchange (Rupietta & Backes-Gellner, 2018). A fundamental distinction between training strategies is based on the primary economic motive for training. When a company trains apprentices primarily to secure its future skilled workforce, this is referred to as an ‘investment-oriented’ training strategy. The company seeks to train its future skilled workers as closely as possible to its own needs and is therefore often prepared to bear net costs during the apprenticeship. These are recouped by keeping apprentices on and benefiting from their subsequent productivity. In contrast, under a ‘production-oriented’ training strategy (Lindley, 1975), the company seeks to cover all training costs during the apprenticeship through the productive work of apprentices. Companies following this strategy are not necessarily dependent on retaining apprentices: even if apprentices are recruited by other companies after completing their training, the apprenticeship has still been worthwhile. The situation is different for companies pursuing an investment-oriented strategy, as these depend on retaining enough apprentices to meet their skilled labour needs and avoid the recruitment costs of hiring from the external labour market. If many apprentices do not remain with the company because they are recruited elsewhere – or because they continue their education with a Federal Vocational Baccalaureate or a university of applied sciences degree – the company cannot recoup its investment. This is likely to be an important reason why large companies frequently establish regional training alliances (Walther et al., 2005) to involve other companies in apprenticeship training in their area. Companies can use a variety of levers to foster the development of apprentices’ skills during their apprenticeship, and these also shape the cost–benefit balance. Particularly important and costly are the training hours provided by training managers in the company. Infrastructure for training also plays a role. Both aspects require good planning and targeted use. Yet even during the apprenticeship, these should not be seen purely as cost factors: if apprentices build their skills as a result, they can also be deployed more productively – both during the apprenticeship and, where applicable, afterwards. At the same time, the type of productive labour itself contributes to their skills development. By the end of the apprenticeship at the latest, apprentices should be competent in the tasks they will later perform as skilled workers. It therefore makes sense for firms that apprentices not only become familiar with these productive tasks during the apprenticeship, but also gain some practical experience of them. For companies, deploying apprentices in this way also helps reduce the workload of skilled workers. The cost–benefit surveys allow these levers in training companies to be analysed in detail. In this way, companies can learn how the different elements affect costs and benefits in training companies (Chapter 2). The average values of the main indicators are also presented for each apprenticeship occupation in dedicated chapters (available online at https://sfuvet.swiss/cost-benefit). 6 Company perspectives on costs, benefits and training quality 2025 1.2 Cost–benefit studies generate insights for steering the system For VET partners, policymakers and researchers, cost– benefit surveys make it easier to understand and take into account companies’ training motives. When VET ordinances for IVET programmes are revised, the number of lessons and days in vocational school or even the total duration of the apprenticeship may, for example, be changed. Such changes affect both the training effort required of companies and the benefit they derive from the productive work of apprentices. The actual changes in companies’ costs and benefits therefore depend on how training is organised in the respective apprenticeship occupation, for instance on how training effort and productive work are distributed across the apprenticeship years. Research using cost–benefit data from Germany and Switzerland has fundamentally changed the understanding of how the dual apprenticeship system works in these two countries. In Germany, companies recorded substantial net costs at the end of the apprenticeship period in the cost–benefit surveys conducted by the Federal Institute for Vocational Education and Training (BIBB) from the outset (e.g. von Bardeleben et al., 1995). International labour market economics research therefore compared Germany with the USA and emphasised the importance of labour market regulations in Germany for the dual system. Various interventions in the labour market (dismissal protection, collective agreements, strong trade unions, works councils, etc.) limit the mobility of German skilled workers compared with the relatively lightly regulated US labour market. This allows training companies to continue employing their former apprentices at wages somewhat below their productivity, generating a rent that offsets the net costs of their training. From this perspective, labour market regulation provided the basis for the existence of the German dual system: Germany is characterised by a ‘low quits/high training’ equilibrium – a situation in which apprentices seldom leave the company after training (low quits) and in which a great deal of training takes place (high training). The USA, on the other hand, has a ‘high quits/low training’ equilibrium, with high labour mobility but little company-based training (Acemoglu & Pischke, 1998). When the results of the 2000 survey for Switzerland were published (Schweri et al., 2003), they showed that companies achieved a net benefit on average. Compared with the German results, this came as a major surprise to both the study authors and their colleagues at BIBB. However, it explains how Switzerland is able to sustain a ‘high quits/high training’ equilibrium. Since the Swiss labour market is far less regulated than the German one, apprentice mobility after training is higher than in Germany. Companies can therefore rely less than in Germany on covering their training costs by retaining apprentices. Instead, most companies ensure that they cover their gross costs during the apprenticeship through apprentices’ productive work (Mühlemann et al., 2010). Switzerland therefore represents a ‘third way’ of vocational education and training compared with the USA, which has no regulated VET system, and Germany, which combines VET with a regulated labour market. These findings were confirmed in another German study (Jansen et al., 2015), which showed that deregulation of the labour market since 2003 had led German companies to make greater productive use of apprentices, thereby reducing the net costs of training. Cost–benefit studies over recent decades have therefore sharpened our understanding of how the dual systems in Germany and Switzerland function, as well as of their similarities and differences. Company perspectives on costs, benefits and training quality 2025 7 The cost–benefit model is used to calculate the costs and benefits of training companies based on the information they provide. The aim is to compare the situation of training companies with a hypothetical situation in which the same companies had no apprentices. How much extra time do companies spend supervising apprentices? What tasks do apprentices perform that add value to the company? For certain cost elements, companies can provide figures directly in monetary terms. These feed straight into the cost–benefit model – for instance, apprentice wages or material expenses. Other data, such as the time spent by trainers and the productive hours contributed by apprentices, must first be converted into monetary costs and benefits. The net benefit of training apprentices for a company corresponds to total productive output minus gross costs. The result is a positive net benefit when productive output exceeds gross costs, and net costs when the opposite is the case. In principle, costs and benefits for a training company can vary between individual apprentices. Some apprentices require less guidance and become productive more quickly. To reduce the effort required by companies, however, this survey did not ask about every individual apprentice, but about the average of all apprentices in the same apprenticeship year. The net benefit calculated in the cost–benefit model therefore always refers to the average apprentice in a given year and may vary across the course of the apprenticeship within a company. 2 THE COST–BENEFIT MODEL The costs of training apprentices naturally arise for a company only during the apprenticeship period. By contrast, companies also benefit after the apprenticeship if they retain trained apprentices and thereby reduce the need to recruit staff from the labour market. This benefit from retention – known as the ‘recruitment opportunity returns’ – is also included in the cost–benefit model. Other benefit elements, such as reputational gains from training activity, are not included. The cost–benefit model described below essentially corresponds to the one used in previous Swiss studies (Schweri et al., 2003; Mühlemann et al., 2007b; Strupler & Wolter, 2012; Gehret et al., 2019) and German studies (Beicht et al., 2004; Schönfeld et al., 2016; Wenzelmann & Schönfeld, 2022) on the costs and benefits of apprenticeship training. Since the questionnaire was comprehensively revised for this survey, the cost–benefit model had to be slightly adjusted in a few elements. Overall, however, comparability with earlier Swiss and with earlier and more recent German studies (see Chapter 4.11) is largely ensured. The text explicitly highlights any results that may have been influenced by the revised survey framework. 2.1 Gross costs Apprentice wage costs One of the largest cost items for companies is apprentice wages. These consist mainly of the apprentices’ monthly gross wage, but also include non-wage labour costs, any regular and irregular special payments, and the thirteenth-month salary. 14 Company perspectives on costs, benefits and training quality 2025 time spent at vocational school…’), since the context of each question was clearer and respondents could follow a logical thread. The issue with this, however, is that this kind of structure requires the questionnaire to be completed in sequence. For this reason, unlike in 2016/17, it was no longer possible to skip entire chapters. Simpler language The 2016/17 survey questionnaire was based on those used in earlier surveys in Germany and Switzerland. For reasons of comparability, much of the terminology and sentence structure was retained; however, the analysis of previous surveys described above indicated that simplifying the language would make the questionnaire easier to complete and improve data quality. Wherever possible, technical jargon was replaced by clearer, more self-explanatory terms. For instance, rather than asking about ‘training facilities’ – a term often confused with the company itself – and distinguishing these from ‘dedicated training rooms for company-based instruction’, companies are now asked whether they have ‘specific training rooms’ that are ‘used exclusively for training and continuing education’. This approach makes it possible to capture the information needed for the calculations without having to rely on a common understanding of terminology or lengthy explanations. An important example of the linguistic revision is the terminology for the two categories of tasks that apprentices perform in the workplace (see Chapter 2). Instead of asking about ‘productive output I: simple tasks’ and ‘productive output II: difficult tasks’, the questionnaire now consistently refers to ‘unskilled tasks’ and ‘skilled tasks’. The terms ‘simple’ and ‘difficult’ depend on the perspective of the respondent: someone with extensive work experience is likely to regard a large part of apprentices’ tasks as ‘simple’ and only a small part as ‘difficult’, whereas someone with little experience would probably see things differently. By contrast, ‘skilled’ and ‘unskilled tasks’ are far less dependent on respondents’ perspective. ‘Unskilled tasks’ are activities that can also be carried out by unskilled persons, while ‘skilled tasks’, whether simple or difficult, are typically performed by qualified skilled workers. These terms also align with the economic concept that the contribution of apprentices corresponds to the costs a company would face if these tasks were performed by other categories of employees. Simplifying the language and terminology reduces respondents’ cognitive load. This in turn should lessen measurement errors – meaning the gap between what happens in practice and what is reported in the questionnaire – and improve data quality. The downside is that reducing measurement error inevitably comes at the expense of strict comparability with earlier question wording. As a result, differences in average values between surveys may not reflect real changes in companies, but rather the outcome of lower measurement error. Even so, this trade-off is considered worthwhile, as it yields a more accurate picture of companies’ costs and benefits. Greater questionnaire customisation By moving fully to an online format, the questionnaire can now be tailored more closely to each company and adapted in real time. Certain questions are asked only of companies or occupations with specific characteristics, thereby reducing reading effort and the potential for misunderstandings. For example, questions on the basic training year and seasonal block courses are only asked for occupations in which such courses actually exist. Other questions (e.g. on ‘full-time trainers’) are only displayed once a company exceeds a minimum number of apprentices. Information already provided is sometimes repeated later in the questionnaire to remind respondents and help them give consistent answers. Adjustments that were introduced in 2016/17 and proved effective have been retained in the new questionnaire. For example, participants can correct conspicuous values or confirm them and leave explanatory comments in real time. Companies that were mistakenly contacted as non-training companies are redirected to the questionnaire for training companies (and vice versa). All of these adjustments have also been carried over to the non-training company questionnaire, wherever the same information is collected. Company perspectives on costs, benefits and training quality 2025 15 3.2 Pretest The questionnaire was trialled in a series of cognitive pretests conducted on site in selected companies and refined step by step. Participants completed it on their computer at work while a member of the survey team observed. They were asked to voice their thoughts as they went along, and both sides had the opportunity to ask questions. The tests resulted in the following main adjustments: – Less is more: many additional explanations were not read and led to uncertainty or misunderstandings. Some of these texts were therefore removed, while others were moved to footnotes at the bottom of the relevant question instead of appearing in the main text. – Ability to ‘park’ certain questions: in some cases, respondents did not have the required information to hand and needed to consult colleagues. For these questions, an option was introduced to park them, continue with the rest of the questionnaire, and return to them later. – Some participants were unable to answer the questions on skilled worker recruitment because they had no insight into this process. This part of the questionnaire can now be forwarded to another person via an input mask. – Further linguistic adjustments. The outcomes of the revision and the pretest were discussed in half-day expert workshops with Harald Pfeifer, Gudrun Schönefeld and Felix Wenzelmann from BIBB, as well as Samuel Mühlemann from Ludwig Maximilian University of Munich, and were then submitted to the contracting authority. Once the revision had been approved, M.I.S. Trend created an online version of the questionnaire, which the project team then tested extensively in terms of both functionality and content. At the same time, the texts were translated into French and Italian and incorporated into the online questionnaire by M.I.S. Trend. These versions too were tested in a final step by the project team. 3.3 Sampling Essentially, separate stratified random samples were drawn for training and non-training companies, and the companies selected were invited by post to take part in the survey (see Chapter 3.4). The sampling approach followed the procedure used in the previous survey (Gehret et al., 2019; Kuhn & Schweri, 2019). The key parameters of the sampling from the two sub-populations are outlined below, with further details on the procedure provided in the accompanying methods report (Kuhn et al., 2025). Defining the target population The first step was to apply a small set of common restrictions to both sub-populations of companies. Specifically, businesses from the primary sector (agriculture), private households and extraterritorial organisations or bodies were excluded from the target population. In addition, the target population was limited to companies with at least two employees. The subsequent procedure and the data basis for drawing the sample differ between the two sub-populations and are, therefore, discussed separately below. Training companies The sample of training companies was drawn from an existing dataset linking the Vocational Education and Training Statistics (SBG) with the Business and Enterprise Register (BUR). Company information from the BUR was combined with the population of all active apprenticeship contracts in a given year. For the purpose of drawing the sample, the Federal Statistical Office (FSO) granted the project team access to an anonymised version of this data. The linked dataset – referred to hereafter as SBG-BUR – contained around 186,200 dual apprenticeship contracts for the 2022/23 training year. After applying the restrictions on the population (e.g. excluding agricultural companies and companies with fewer than two employees), the relevant number of apprenticeship contracts was reduced to around 176,200. 16 Company perspectives on costs, benefits and training quality 2025 It is important to note that 'company' always refers to the local unit of an enterprise, since the training of apprentices often takes place locally, i.e. in the local unit (branch). The specific linkage of SBG and BUR data has the major advantage of addressing two structural features in the data that would otherwise make sampling much more complicated. Firstly, only a relatively small proportion of all companies in Switzerland train apprentices in any given apprenticeship year (FSO, 2024). At the same time, there are large differences in how frequently the various apprenticeship occupations are offered (Kuhn, 2024). This creates a challenge, since a simple random draw would mainly result in non-training companies and the more common occupations being included in the sample. In the SBG–BUR data, however, training companies can already be identified before the actual sampling begins. This makes it possible to draw a sample in which training companies are over-represented compared with non-training companies. Secondly, the data also shows– again before sampling – which companies offer which apprenticeship occupations. This information makes it possible to identify companies that train in less common occupations and assign them a higher probability of selection. This in turn ensures that less common occupations are also represented in the survey in sufficient numbers. In a preparatory step, each company that trained in more than one apprenticeship occupation was assigned a single occupation. This was necessary because companies invited to take part in the survey were generally only able to provide information for one. The assignment of a unique occupation was carried out using probabilities, based on the frequency of the apprenticeships offered within the company. In other words, the more common an apprenticeship occupation was within a company, the more likely it was to be chosen; less common occupations were selected with proportionately lower probability. The next step was to draw a stratified random sample by apprenticeship occupation and company type, meaning samples were taken from groups of companies (strata). This had the advantage of allowing certain occupations to be deliberately over-represented in the sample. In this step, an attempt was made to meet several sampling objectives at once. Firstly, separate evaluations were to be possible for at least 25 different apprenticeship occupations, including five FVETC programmes, which required a sufficient number of companies providing information for each of these occupations. Secondly, companies offering apprenticeships with a longer training duration were given a higher probability of selection, since information must be available for all apprenticeship years. Thirdly, stratification was used to give more weight to companies with a large number of apprenticeship contracts, in order to cover as many active contracts as possible through the survey. Finally, companies belonging to enterprises with a very large number of establishments (e.g. major retail chains with many branches) were under-represented in the sample, as they are otherwise unlikely to participate. Based on the previous survey, we assumed a response rate of around 27.4 per cent for training companies (Gehret et al., 2019; Kuhn & Schweri, 2019). To ensure a net sample of at least 4,500 training companies, the gross sample was therefore set at 17,400 companies, with an additional safety margin. This corresponds to around 32 per cent of all training companies in the relevant initial population. Non-training companies A separate sample was also drawn from the population of non-training companies. Unlike the training company sample, this draw was carried out by the FSO. The first step was to mark the training companies in the BUR using the SBG–BUR linkage. The linked companies represented the population of training companies, while the complementary set constituted the population of non-training companies. From this population, a stratified random sample was then drawn by company size class, with larger companies given a higher probability of selection. Based on the previous survey, the gross sample for non-training companies was set at 12,600, with the aim of achieving a net sample of at least 3,500 companies. In the 2016/17 apprenticeship year, the response rate Company perspectives on costs, benefits and training quality 2025 17 among non-training companies was around 28.8 per cent; here too, an additional reserve of companies was added. The gross sample size corresponds to around 4.6per cent of all non-training companies in the target population. Training rate in the target population Using the population datasets from which the two subsamples were drawn, it is also possible to estimate the aggregated training rate – that is, the percentage of companies that trained at least one apprentice during the 2022/23 apprenticeship year. In the defined target population, a total of 53,828 training companies and 273,481 non-training companies were identified, giving 327,309 companies in total. This results in an aggregated training rate of around 16.5 per cent for the 2022/23 apprenticeship year. In other words, around one in six companies in Switzerland trained at least one apprentice during that year. In the gross sample as a whole, training companies are significantly over-represented, accounting for 58 per cent of the companies contacted. This reflects the central aim of the study, which is to provide information primarily on the costs and productive outputs of apprentice training from the perspective of training companies. 3.4 Field phase To ensure the survey covered an apprenticeship year no longer significantly affected by COVID, the field phase was postponed from 2022 to 2023. Companies in the gross samples were invited to take part from the first week of May onwards. Each company received an invitation letter in the relevant national language – or in German for companies in the Romansh-speaking area – sent to the business address recorded in the Business and Enterprise Register. The letter included the login details, basic information on the survey and the support service contact details, together with a short information brochure. The mailing was supported by a free support service, both by telephone and email, for the companies. This service was available Monday to Friday from 8 a.m. to 6p.m. in three languages. Companies could use it to ask questions about the survey in general and about the questionnaire in particular. An FAQ website was also maintained and updated on an ongoing basis. Both the hotline and the email contact option were used extensively, particularly after invitation and reminder letters were dispatched. As in earlier surveys, response rates rose sharply after the two reminders were sent out. The entire field phase proceeded smoothly without any notable technical problems. Direct contacts and training centres Unlike in previous surveys (Strupler & Wolter, 2012; Gehret et al., 2019), large companies and the main training centres and associations were not treated separately at the invitation stage. After the end of the primary field phase, these organisations were contacted later where necessary. Depending on the allocation of responsibilities within the organisation and after a preliminary phone call, this took the form of a multi-stage written survey. As shown in Table 4 in Chapter 3.6, large companies are well represented in the net sample and account for a higher proportion than in the previous survey (see Gehret et al., 2019, p. 22, Table 4). 3.5 Responses in the two subsamples In the end, only a portion of the companies contacted could be included in the analysis. The main reasons were incomplete or incorrect addresses, companies that had already ceased trading by the time of the survey, a lack of willingness to participate in the survey, and too many missing or unclear answers despite companies’ active participation. 18 Company perspectives on costs, benefits and training quality 2025 Table 3 documents the responses in the two subsamples of our survey. The gross sample comprised 17,400 training companies and 12,600 non-training companies. As in the previous survey, the address data from the SBG– BUR proved relatively reliable, since only a small number of companies could not be contacted by post due to incorrect addresses. Of the training companies in the gross sample, invitations could not be delivered to only 301 companies, corresponding to around 1.7 per cent. Among non-training companies (drawn directly from the BUR), on the other hand, a significantly larger share of around 5.4 per cent could not be reached by post. More significant are drop-outs due to refusal to participate and cases where companies left a comparatively large or crucial part of the questionnaire blank or filled it in incorrectly, making these responses unusable for data analysis. Around 61 per cent of training companies at least opened the questionnaire, and 39.1 per cent of responses were usable. Among non-training companies, around 35 per cent opened the questionnaire and around 31 per cent of responses were usable. In addition, 999 non-training companies – around 27 per cent of the usable responses – reported that they could not train apprentices because they were not economically independent entities or were very small, highly specialised branch offices. These companies are therefore not included in the subsequent analyses. In absolute terms, this leaves information from 6,686training companies and 2,656 non-training companies for the analyses that follow. 3.6 Composition of the two net samples In the next step, the composition of the two net samples is described with reference to several key characteristics. As the following two tables focus on unweighted companies, they also reflect the partly disproportionate stratification of the sampling plan. For example, there are marked differences between language regions in the importance of initial vocational education and training, resulting in a higher share of training companies in German-speaking Switzerland (Aepli et al., 2021a; Kuhn et al., 2022). Since training companies were over-represented in the gross sample, this also leads to a disproportionately large share of companies from German-speaking Switzerland. Training companies Non-training companies Number Percentage Number Percentage Number of invitations 17,400 100.00 12,600 100.00 Undeliverable invitations 301 1.73 679 5.39 Number of invitations (minus postal returns) 17,099 100.00 11,921 100.00 Questionnaire accessed 10,465 61.20 4,155 34.85 Questionnaire completed 7,379 43.15 3,655 30.66 Usable responses 6,686 39.10 3,655 30.66 Too specialised for training   999 27.27 Table 3: Responses in the two gross samples Company perspectives on costs, benefits and training quality 2025 19 Table 4 first focuses on the net sample of 6,686 training companies. Around 29.6 per cent of these companies have up to 9 employees, 40.4 per cent between 10 and 49, 10.8 per cent between 50 and 99, and the remaining 19.1 per cent have 100 or more. Broken down by sector, about 14 per cent of companies belong to the construction and 14 per cent to the manufacturing sector, a majority of 46 per cent are in the services sector and around 26 per cent in public administration and non-profit organisations. Finally, the vast majority of training companies – around 76.6 per cent – are located in German-speaking Switzerland (including regions where Romansh is the predominant language), 19.7 per cent are from French-speaking Switzerland, and 3.8 per cent from Ticino and other predominantly Italian-speaking regions. The lower part of the table shows how many questionnaires were completed for each apprenticeship occupation. For occupations with at least 40 observations per apprenticeship year (14 for FVETC apprenticeships) from a minimum of 65 companies (25 for FVETC), separate evaluations are available on the SFUVET website (see https://sfuvet.swiss/cost-benefit). Across all apprenticeships, about 7 per cent of companies provided information on a two-year FVETC apprenticeship, 61 per cent on a three-year FVETD apprenticeship, and 32 per cent on a four-year FVETD apprenticeship. The distribution across specific apprenticeship occupations further shows that data from at least 30 different training companies is available for a total of 30 occupations (including five FVETC apprenticeships). The smallest group comprises 30 companies (Logistics Assistant, FVETC), while the largest comprises 750 companies (Commercial Employee, FVETD). Characteristic Category Number Percentage Company size class 1–9 employees 1,982 29.64 10–49 employees 2,704 40.44 50–99 employees 723 10.81 100+ employees 1,277 19.10 Sector Manufacturing 928 13.88 Construction 935 13.98 Services 3,054 45.68 Administration, non-profit organisations 1,769 26.46 Language region German 5,121 76.59 French 1,314 19.65 Italian 251 3.75 Duration 2 years 456 6.82 3 years 4,082 61.05 4 years 2,148 32.13 Table 4: Composition of the net sample of training companies 20 Company perspectives on costs, benefits and training quality 2025 Characteristic Category Number Percentage Apprenticeship  Commercial Employee, FVETD 750 11.22 Draughtsman/Draughtswoman, FVETD 444 6.64 Retail Professional, FVETD 422 6.31 Certified Social Care Worker, FVETD 397 5.94 Joiner, FVETD 202 3.02 Automotive Technician, FVETD 193 2.89 Dental Assistant, FVETD 185 2.77 Healthcare Assistant, FVETD 178 2.66 Medical Secretary and Assistant, FVETD 169 2.53 Electrician, FVETD 166 2.48 Carpenter, FVETD 140 2.09 Pharmacy Assistant, FVETD 140 2.09 Chef, FVETD 131 1.96 Automotive Mechatronics Technician, FVETD 117 1.75 Painter, FVETD 116 1.73 Building and Grounds Custodian, FVETD 114 1.71 Logistics Professional, FVETD 114 1.71 Hairdresser, FVETD 110 1.65 Information Technologist, FVETD 107 1.60 Gardener, FVETD 104 1.56 Polymechanic, FVETD 98 1.47 Plumber, FVETD 95 1.42 Bricklayer, FVETD 93 1.39 Automation Technician, FVETD 85 1.27 Building Services Assistant, FVETC 78 1.17 Construction Electrician, FVETD 67 1.00 Retail Assistant, FVETC 64 0.96 Office Assistant, FVETC 52 0.78 Health and Social Care Worker, FVETC 48 0.72 Logistics Assistant, FVETC 30 0.45 Other four-year FVETD 789 11.80 Other three-year FVETD 704 10.53 Other FVETC 184 2.75 (6,686 companies, unweighted) Company perspectives on costs, benefits and training quality 2025 21 In addition, Table 5 shows the composition of the net sample of 2,656 non-training companies. A large share of these companies – 67.6 per cent – have fewer than 10employees, 20.3 per cent have between 10 and 49 employees, 6.8 per cent have between 50 and 99 employees, and 5.3 per cent have 100 or more employees. By sector, the majority – 63.5 per cent – are in services, followed by public administration and non-profit organisations with 19.7 per cent, manufacturing with 8.8 per cent, and construction with around 8.0 per cent. In terms of language regions, 66.4 per cent are based in German-speaking Switzerland, 26.4 per cent in French-speaking Switzerland, and 7.2 per cent in mainly Italian-speaking regions. 3.7 Plausibility checks To guarantee the quality of the data, the survey responses went through several stages of plausibility checking. First, during data entry in the browser, responses were checked against predefined thresholds. If these thresholds were exceeded or not met, respondents were alerted to the unusual values. They could then correct the values directly in the browser or confirm and comment on them. These comments were used in subsequent plausibility checks to better understand anomalies or, in some cases, to adjust values retrospectively. Compared with the previous survey, the quality of responses was generally better, with fewer implausible answers. Even so, many cases still required a follow-up call. To identify these cases and the companies concerned, the data was analysed once again against threshold values. If responses contained implausible or missing entries, the relevant companies were contacted by telephone, where possible, about the affected parts of the questionnaire. Characteristic Category Number Percentage Company size class 1–9 employees 1,796 67.62 10–49 employees 539 20.29 50–99 employees 180 6.78 100+ employees 141 5.31 Sector Manufacturing 233 8.77 Construction 213 8.02 Services 1,687 63.52 Administration, non-profit organisations 523 19.69 Language region German 1,764 66.42 French 700 26.36 Italian 192 7.23 (2,656 companies, unweighted) Table 5: Composition of the net sample of non-training companies 22 Company perspectives on costs, benefits and training quality 2025 3.8 Approach to missing variable values Consequences of missing variable values Calculating the costs and benefits of company-based training (see Chapter 2) requires a company to answer all questions relating to the variables included in the model. In practice, however, many companies left one or more of these questions unanswered. The difficulty here is that a missing value for one variable leads to missing values at all subsequent levels of calculation or analysis. This applies even in the extreme case where only a single variable required for a calculation step is missing. Imputation of missing values For this reason, missing values for variables required in the cost–benefit calculation were imputed using statistical methods (Dong & Peng, 2013; Little & Rubin, 2019). Imputation means that missing values in individual variables are replaced with plausible valid values. Put simply, the basic principle of imputation is that a missing entry for one company is substituted with a valid entry from other, as far as possible comparable, companies. In practice, this process is often highly complex, since missing values may occur simultaneously in several variables within the same company, and the patterns of missing values vary between companies. Modern imputation methods operate iteratively – that is, missing values are gradually replaced over several repeated runs of a complex algorithm. This makes it possible to correct practically any pattern of missing data (see van Buuren, 2007, 2018). Building on the imputation process used in the previous survey, this time a multiple imputation method was chosen. This means that missing values were replaced not with a single value, but with 35 different values. The data analyses were then essentially carried out 35 times, and the results of these estimates averaged. This approach offers two advantages: first, it accounts for the additional uncertainty caused by imputation in the calculation of variances and standard errors; second, multiple imputation is less influenced by outliers, which can occur when imputing single values. The trade-off, however, is that this process significantly increases computing requirements and makes subsequent analyses more complex. 3.9 Sample weighting As explained in Chapter 3.3, the sampling process deliberately selected a disproportionate number of training companies compared with non-training companies, as well as less frequently chosen apprenticeship occupations compared with more popular occupations. This altered the composition of the gross sample compared with the overall population of companies. Sample weights serve first and foremost to correct for this in the analysis. They are also used to adjust for differences in realised response rates across the various strata of the sample. The sample weights were determined in two successive steps, each based on the strata used in the sampling design – namely apprenticeship occupation and company type for training companies, and company size class for non-training companies. In the first step, the initial or design weight is determined. This is defined as the inverse of the probability of a company from the population being included in the gross sample. This weight thus enables inference from the gross sample to the population, compensating for the partly disproportionate selection probabilities in the sampling process. In the second step, the response rate within each stratum is determined. This generates a second weight that corrects for an incomplete response. This second weight is used to draw inferences from the net sample to the gross sample, thereby balancing differences in response rates across strata. The final sample weight is the product of these two weights and thus corrects simultaneously for both the disproportionate sampling design and the differences in response rates. Further information on the sample weights is provided in the accompanying methods report (Kuhn, Gehret & Schweri, 2025). Company perspectives on costs, benefits and training quality 2025 23 This chapter presents the results of the cost–benefit survey. The first step is to examine the gross costs, productive output and net benefit of initial vocational education and training from a company perspective. These are considered first in aggregate, and then differentiated by training duration (two-year apprenticeships leading to an FVETC and threeand four-year apprenticeships leading to an FVETD), as well as by company size, sector and major region. The chapter then summarises the net benefit of the most frequent apprenticeship occupations. It also illustrates how companies can generate additional benefits by employing apprentices once their training is complete. The chapter concludes by extrapolating the results to Switzerland as a whole and comparing them with those of the latest German cost–benefit study. 4.1 Average costs and benefits of initial vocational education and training As outlined in Chapter 2 in the discussion of the cost-benefit model, companies incur gross costs as a result of training. These consist of apprentice wages, staff costs, material and equipment costs, and additional costs. In the 2022/23 training year, the average gross cost across all apprenticeships was CHF 27,990 per apprenticeship year. At the same time, companies benefit from the work provided by apprentices, who can be assigned to either unskilled or skilled tasks. The monetary value of this is captured in the productive output, which averaged CHF32,530 per apprenticeship year across all apprenticeships. Since productive output exceeded gross costs, companies achieved an average net benefit of CHF 4,540 per apprenticeship per year. This fundamental finding – that training apprentices generates a net benefit on average across all apprenticeships – confirms the results of previous Swiss surveys in 2016/17 (Gehret et al., 2019), 2009/10 (Strupler & Wolter, 2012) and 2008/09 (Fuhrer & Schweri, 2010). The slightly higher net benefit compared with 2016/17 is due to a small decrease in gross costs, while productive output remained constant. Personnel costs rose slightly, but companies on average estimated material and equipment costs somewhat lower. Although the total productive output has changed little compared with 2016/17, its composition has shifted fundamentally. Whereas unskilled tasks previously accounted for the larger share, it is now skilled tasks that bring greater benefits to companies. This change was very likely caused by the revision of the survey framework. In assessing apprentices’ activities in the workplace, the terminology was deliberately revised and standardised. As explained in Chapter 2, the categories ‘simple tasks’ and ‘difficult tasks’ were no longer used, as this distinction left too much room for interpretation. Instead, the terms ‘unskilled tasks’ and ‘skilled tasks’ were adopted consistently, both of which are directly related to the company’s work and production processes. From a pedagogical perspective, too, the key question is whether apprentices are able to acquire and practise the skills needed to perform the tasks expected of them as skilled workers. It is therefore encouraging that skilled tasks now outweigh unskilled tasks. When interpreting the findings, it is important to bear in mind that the net benefit reflects the cost–benefit balance at the end of the apprenticeship (or for a specific year of training) and does not include certain benefit elements, such as the gains from employing apprentices after their training. An average positive net benefit does not mean, however, that all companies achieve a net benefit from training apprentices by the end of the apprenticeship. Figure 1 illustrates this by showing the distribution of net benefit per apprenticeship place. Apprenticeships above the zero-franc threshold represent cases with a net benefit, while those below it represent cases with net costs. In fewer than 30 per cent of apprenticeships, net costs are incurred. By contrast, the majority – around 71 per cent– result in a smaller or larger net benefit. 4 RESULTS 30 Company perspectives on costs, benefits and training quality 2025 in the way apprentices’ time is distributed in the workplace. While companies assessed the share of unproductive time similarly to 2009/10 and 2016/17, the division between unskilled and skilled tasks differs in all training years. In contrast to 2016/17 and also 2008/09, companies now estimate the share of skilled tasks significantly higher in every apprenticeship year. As with the two-year FVETC programmes (see Chapter 4.2), this change is likely due to adjustments in the survey framework. For example, the estimate that apprentices spend 66 per cent of their time on skilled tasks in the final year, compared with 45 per cent in 2016/17, also seems realistic for the three-year FVETD programmes. 0 20 40 60 80 100 Percentage Apprenticeship year 1 Apprenticeship year 2 Apprenticeship year 3 Skilled tasks Unskilled tasks Other/exercises 4.4 Costs and benefits of the fouryear initial vocational education and training leading to an FVETD In the 2022/23 training year, according to FSO data (and the population definition in Chapter 3.3), almost 51,300apprentices were trained over the course of the four-year IVET in 79 occupations leading to a Federal VET Diploma (FVETD). Among these apprentices, 17 per cent were women and 4.5 per cent were adult apprentices (aged over 25). Fewer than 7 per cent of the apprenticeship places were shortened training programmes. The following analysis is based on 58 different occupations in 2,148 training companies, with a total of 4,773apprentices across different years. In this sample, the Federal Vocational Baccalaureate (BM 1) rate was almost 14per cent. Figure 6: Time spent by apprentices at the company workplace in three-year FVETD apprenticeships Company perspectives on costs, benefits and training quality 2025 31 The total net benefit that a training company achieves on average per apprenticeship in four-year IVET programmes (see Table 10) is around CHF 17,510 – significantly higher than in the 2016/17 survey. The 95 per cent confidence interval (‘+/–’) is plus/minus CHF 4,060, i.e. between CHF 13,450 and CHF 21,570. 0 10 20 30 40 In CHF thousands Apprenticeship year 1 Apprenticeship year 2 Apprenticeship year 3 Apprenticeship year 4 Apprentice wage costs Personnel costs Material/equipment costs Additional costs Figure 7: Breakdown of gross costs in four-year FVETD apprenticeships Apprenticeship year 1234Total Gross costs 26,300 27,090 29,850 32,670 115,910 +/- 700 670 660 720 1,960 Productive output 21,650 27,160 37,600 46,990 133,410 +/- 950 940 1,130 1,390 3,590 Net benefit -4,650 70 7,750 14,330 17,510 +/- 1,150 1,120 1,320 1,600 4,060 Table 10: Gross costs, productive output and net benefit per four-year FVETD apprenticeship (in CHF) 32 Company perspectives on costs, benefits and training quality 2025 Gross costs For gross costs and their components (see Figure 7), the typical increase in costs per apprenticeship year is evident as the training duration rises. The reason lies in rising apprentice wages (see also Table 11), while personnel costs decline slightly. Companies estimate the average weekly time spent on training apprentices at 6.6hours in the first year, 5.6 hours in the second, 4.9hours in the third and 4.3 hours in the fourth. Material and equipment costs and additional costs are on average higher than in twoor three-year IVET programmes, mainly because many of the four-year IVET programmes are in technical occupations. When looking at the monthly median wage of apprentices (see Table 11), it is striking that it is lower in the first two years than in two-year IVET programmes and lower in the first three years than in three-year programmes. This pattern, also observed in earlier surveys, is likely related to the fact that apprentices in the more demanding four-year programmes take longer to reach a productivity level comparable with that of skilled workers. In the final year, however, wages rise to CHF 1,350. As in the shorter programmes, wages show some variation: in the final year, 25 per cent of apprentices earn CHF1,450 or more per month, while another 25 per cent earn CHF1,200 or less. Apprenticeship year 1234 25th percentile 550 750 950 1,200 Median wage 650 800 1,050 1,350 75th percentile 700 930 1,200 1,450 Table 11: Monthly gross wage of apprentices in four-year FVETD apprenticeships (in CHF) 0 20 40 60 80 100 Percentage Skilled tasks Unskilled tasks Other/exercises Apprenticeship year 1 Apprenticeship year 2 Apprenticeship year 3 Apprenticeship year 4 Figure 8: Time spent by apprentices at the company workplace in four-year FVETD apprenticeships Company perspectives on costs, benefits and training quality 2025 33 Productive output Given that four-year FVETD programmes tend to cover more demanding occupations, it is hardly surprising that first-year apprentices spend about a quarter of their time in the company on unproductive activities such as training exercises (Figure 8). This share decreases as training progresses. In the final year, apprentices spend on average more than 87 per cent of their time in the company on productive activities, most of which involve tasks that would otherwise be performed by qualified skilled workers. The share of such skilled tasks is already 44 per cent in the first year and rises to 74 per cent in the final year. Apprentices’ performance level in these skilled tasks increases from 31 per cent to 46 per cent, 63per cent and finally 78 per cent in the last year. Comparison with the previous survey Compared with the 2016/17 training year, net benefit rises more steeply over the four years of training and is almost CHF 10,000 higher in total. This significant difference is explained by lower gross costs combined with higher productive output. As with the twoand threeyear programmes, the lower gross costs are largely due to reduced material and equipment costs and slightly lower additional costs. It is possible that these costs have, in fact, fallen, but it is also possible that they appear lower because the revised questionnaire described them in clearer and simpler terms. Another reason for the increase in net benefit since 2016/17 lies in the changed composition of four-year apprenticeships. The new four-year training programmes (before: three years) in the occupations of Carpenter FVETD, Plumber FVETD, Heating Installer FVETD and Sheet-Metal Worker FVETD, which have only been introduced after the last survey, raise the average net benefit. For the three building-services occupations, however, this applies only to the first three years of training, since the newly introduced fourth year took place for the first time in the year after the survey. As with the twoand three-year programmes, the breakdown of productive output in four-year apprenticeships has shifted strongly towards skilled tasks. This reflects the revised way in which workplace activities were assessed, and is linked to clearer question wording in the questionnaire. So far, the results have been presented by training duration, whereas the following chapters describe the findings by company size, sector and major region. The perspective changes here: instead of analysing individual years of training (first year, second year, etc.), the costs and benefits of an average year of training are examined. This allows two-, threeand four-year apprenticeships to be considered as a whole. 4.5 Results by company size In the 2022/23 training year, according to FSO data from the SBG-BUR, most apprentices were trained in small companies with 10 to 49 employees (38 per cent of apprenticeships) and in large companies with 100 or more employees (27 per cent). Very small companies with up to 9 employees also trained a substantial share at over 20 per cent, while medium-sized companies with 50 to 99 employees accounted for almost 13 per cent. Only a small number of training companies in the BUR did not report their company size. Figure 9 shows gross costs, productive output and net benefit by company size. The results show that, on average, companies of all sizes achieve a net benefit. In contrast to the 2009/10 and 2016/17 surveys, medium-sized and large companies no longer record a marked drop in net benefit. Productive output is now also very similar across the different company sizes. 34 Company perspectives on costs, benefits and training quality 2025 4.6 Results by sector In the 2022/23 training year, around 63 per cent of apprenticeships were in the service sector, just under 16per cent in manufacturing, almost 14 per cent in construction and nearly 8 per cent in public administration or non-profit organisations (SBG-BUR). As shown in Figure 10 – and in contrast to the results by company size– there are larger differences between sectors in average net benefit. All sectors achieve a clear net benefit, but construction stands out with a significantly higher figure. Although gross costs in construction are slightly higher than in public administration and non-profit organisations and also in services, its productive output is clearly greater than in the other sectors. The higher productive output in construction is mainly driven by the greater number of days apprentices spend in the company. On average, apprentices in construction spend 7 to 13 more days per year in the workplace. They also spend less time on unproductive activities such as training exercises. This effect is reinforced by the skilled-worker wages used to value productive skilled tasks, which are higher in construction than in public administration/non-profit organisations and in the service sector. Higher skilled-worker and unskilled-worker wages in the calculation of productive output also explain why apprenticeships in manufacturing show higher productive output than those in public administration/non-profits and in services. 0 10 20 30 40 <10 10–49 50–99 >99 In CHF thousands Number of employees within the company Gross costs Productive output Net benefit Figure 9: Gross costs, productive output and net benefit per average apprenticeship year, by company size Company perspectives on costs, benefits and training quality 2025 35 4.7 Results by major region In the 2022/23 training year, apprenticeship places were spread across all major regions. Espace Mittelland accounted for more than 22 per cent of all contracts, Zurich 19 per cent, Eastern Switzerland 16 per cent, the Lake Geneva region 14 per cent, Central Switzerland 11 per cent and Ticino just over 3 per cent. Compared with the sector results, the differences between the various major regions are smaller (see Figure 11). Once again, all groups show a positive net benefit, but the difference is statistically significant only between the Lake Geneva region and Ticino. The higher net benefit in the Lake Geneva region is explained by higher productive output, especially from skilled tasks, despite slightly higher gross costs. Productive output in the Lake Geneva region is also higher than in Espace Mittelland and in Northern and Eastern Switzerland, owing to slightly higher skilled-worker wages, which are used to calculate the value of skilled tasks. As in 2016/17, the cost–benefit balance in Ticino is marked by lower productive output. This reflects the impact of lower wages for skilled and unskilled workers on the calculation of output: when wages are lower, the value of the work performed by apprentices is reduced. Compared with 2016/17, it is also notable that productive output in the Zurich region has converged with that of the other regions (except Ticino), leading to a more favourable cost–benefit balance. Figure 10: Gross costs, productive output and net benefit per average apprenticeship year, by sector In CHF thousands Services Manufacturing Construction Administrative and non-profit Gross costs Productive output Net benefit 0 10 20 30 40 36 Company perspectives on costs, benefits and training quality 2025 4.8 Results by occupation The following figures show the average net benefit for the most common FVETC occupations (Figure 12) and FVETD occupations (Figure 13 and Figure 14). Detailed analyses of these occupations are available online at the following link: https://sfuvet.swiss/cost-benefit The occupational titles used correspond to those in force in the 2022/23 training year. For example, the German title ‘Fachfrau/Fachmann Apotheke EFZ’ also covers apprentices in the occupation previously known as ‘Pharma-Assistent/-in EFZ’ who were in their second or third year of training at that time. In the two-year IVET programmes, Logistics Assistant FVETC and Building Services Assistant FVETC can now be reported separately, while the case numbers for Automotive Assistant FVETC and Kitchen Employee FVETC are no longer sufficient for individual analysis. As with other FVETC occupations with too few cases (fewer than 25 participating companies or fewer than 14 observations per apprenticeship year), these are now grouped under ‘Other FVETC’. Overall, the average net benefit of FVETC occupations is higher than in 2016/17 (see Chapter 4.2). What stands out is the relatively high average net benefit in the Logistics Assistant FVETC occupation, driven by comparatively high productive output alongside slightly below-average gross costs. By contrast, the drop in net benefit for Office Assistant FVETC compared with 2016/17 is striking, as it reflects a marked decline in productive output. Figure 11: Gross costs, productive output and net benefit per average apprenticeship year, by major region 0 10 20 30 40 In CHF thousands Lake Geneva Region Espace Mittelland Northwestern Switzerland Zurich Eastern Switzerland Central Switzerland Ticino Gross costs Productive output Net benefit Company perspectives on costs, benefits and training quality 2025 37 Figure 12: Average net benefit of the most common FVETC apprenticeship occupations 010 20 30 Net benefit in CHF thousands Health and Social Care Worker, FVETC Building Services Assistant, FVETC Office Assistant, FVETC Other FVETC Retail Assistant, FVETC Logistics Assistant, FVETC Figure 13: Average net benefit of the most common three-year FVETD apprenticeship occupations Net benefit in CHF thousands Bricklayer, FVETD Chef, FVETD Healthcare Assistant, FVETD Plumber, FVETD Other three-year FVETD Hairdresser, FVETD Automotive Technician, FVETD Retail Professional, FVETD Commercial Employee, FVETD Construction Electrician, FVETD Medical Secretary and Assistant, FVETD Painter, FVETD Certified Social Care Worker, FVETD Building and Grounds Custodian, FVETD Pharmacy Assistant, FVETD Logistics Professional, FVETD Gardener, FVETD Dental Assistant, FVETD 010 20 30 40 38 Company perspectives on costs, benefits and training quality 2025 In the three-year FVETD programmes, fewer apprenticeship occupations are reported than in 2016/17 because case numbers in some occupations were too low. Occupations no longer reported separately include Hospitality Services Professional FVETD, Restaurant Service Employee FVETD and Specialist in Hotel Housekeeping FVETD. These and other three-year FVETD occupations with low case numbers (fewer than 65 participating companies or fewer than 40 observations per apprenticeship year) have been grouped under ‘Other three-year FVETD’. With the most recent reform of the VET ordinance, the Plumber FVETD programme was extended from three to four years. Since there were no apprentices in their fourth year in 2022/23, this occupation is still shown as a three-year programme in all analyses. The actual net benefit may change once the fourth year is included, as the final apprenticeship year usually increases net benefit. By contrast, the Carpenter FVETD occupation was observed across all training years and is now listed under the four-year apprenticeships. Compared with 2016/17, net benefit has also risen in the three-year programmes. None of the occupations analysed now records a significantly negative net benefit. Most show a modest to substantial net benefit. The Plumber apprenticeship appears at first glance to contradict this positive trend, showing a marked decline in net benefit. However, since only the first three apprenticeship years were observed here, it is not possible to estimate how much the net benefit will increase in the final year of apprenticeship. The trend towards higher net benefit is also evident in the four-year FVETD programmes. This is particularly clear in the Polymechanic FVETD apprenticeship, which in earlier surveys (2016/17 and 2009/10) was still shown as an occupation with significant net costs. Gross costs still outweigh productive output in this case, but to a lesser extent than before. A similar pattern can be seen in the Information Technologist FVETD and Automotive Mechatronics Technician FVETD apprenticeships. For the latter, this is the first time since the initial survey in 2000/01 (see Schweri et al., 2003; Mühlemann et al., 2007b; Strupler & Wolter, 2012; Gehret et al., 2019) that a positive net benefit has been estimated. Fouryear FVETD apprenticeship occupations with low case numbers (fewer than 65 participating companies or fewer than 40 observations per apprenticeship year) have been grouped under ‘Other four-year FVETD’. Figure 14: Average net benefit of the most common four-year FVETD apprenticeship occupations Net benefit in CHF thousands Polymechanic, FVETD Automation Technician, FVETD Information Technologist, FVETD Other four-year FVETD Automotive Mechatronics Tech., FVETD Draughtsman/-woman, FVETD Joiner, FVETD Carpenter, FVETD Electrician, FVETD -20 020 40 60 Company perspectives on costs, benefits and training quality 2025 39 4.9 Findings on the benefits of continued employment The previous chapters focused on the costs and benefits that arise during training. As explained in Chapter 2, companies can derive additional benefits by keeping apprentices on after their training. This is because they avoid some of the recruitment and induction costs that would otherwise arise when hiring skilled workers from the external labour market. In this way, training companies achieve an average additional benefit of around CHF 13,490 per apprenticeship across all company sizes. Table 12 lists the average recruitment and induction costs and the average additional benefit from retention (recruitment opportunity returns) by company size. The results show that larger companies achieve greater gains. This is consistent with the observations from 2009/10 and 2016/17, although the differences are now once again more pronounced as in the 2009/10 survey. One reason is that, as in 2009/10, recruitment and induction costs also rise clearly with company size. 4.10 Extrapolation of training costs, productive output and net benefit Table 13 shows the extrapolation of training costs, productive output and net benefit for all ongoing apprenticeship contracts for the 2022/23 training year. This is calculated by multiplying the relevant average estimates by the number of places in the corresponding apprenticeship year. The number of ongoing apprenticeship contracts can be determined from the SBG-BUR data, which form the basis for the sample of training companies, as outlined in Chapter 3.3. For the target population of training companies defined there, a total of 176,183 ongoing dual apprenticeship contracts was recorded for the 2022/23 training year. Since some companies were excluded from the target population (e.g. agricultural companies), the extrapolated costs and benefits for Switzerland are likely to be somewhat underestimated. Total costs and benefits per apprenticeship year, in CHF millions Gross costs 4,930 +/- 71 Productive output 5,731 +/- 123 Net benefit 800 +/- 129 Note: The extrapolations refer to a total of 176,183 apprenticeship contracts in the 2022/23 training year. The corresponding 95 per cent confidence interval (‘+/–’) is shown in the second line in each case. Table 13: Extrapolation of costs and benefits for the 2022/23 apprenticeship year Based on the average gross costs in the net sample, the extrapolated total amounts to around CHF 4,930 million in training costs (gross costs). This compares with a total of around CHF 5,731 million in productive output generated by apprentices during their training, resulting in an extrapolated positive net benefit of around CHF 800 million for the 2022/23 training year. The 95 per cent confidence interval (‘+/-’) is plus/minus CHF 129 million, i.e. Number of employees within the company <10 10–49 50–99 >99 Recruitment and induction costs 14,130 17,880 20,500 22,130 Benefit of continued employment 8,550 11,190 12,400 17,860 Table 12: Average recruitment and induction costs per skilled worker and benefit of continued employment, by company size 46 Company perspectives on costs, benefits and training quality 2025 The education system includes regulations that directly guide companies in how they train apprentices, and others that shape the training context more indirectly. A central example of prescriptive regulations are VET ordinances and training plans, which are also binding for companies. By contrast, the Federal Vocational Baccalaureate Type 1 (BM1), which accompanies the apprenticeship, is an educational pathway aimed primarily at apprentices, though it also affects companies’ training environments. This chapter discusses the significance of training plans and the Federal Vocational Baccalaureate for training companies. 6.1 VET ordinances and training plans VET ordinances (“Bildungsverordnung” BiVo) and training plans (“Bildungsplan” BiPla) set out the tasks and skills that training companies are expected to teach apprentices in a given occupation. They ensure that apprentices acquire comprehensive occupational skills that extend beyond the needs of any single company. This requires a certain degree of standardisation in the training content of apprenticeships. But this also means that BiVo and BiPla may not always match the realities of every company offering training in the same occupation. For example, the day-to-day work of an information technologist in an industrial company is likely to differ from that of an information technologist in a service company, yet trainers in both are bound by the same BiVo and BiPla. This raises the question of how well companies feel the BiVo and BiPla relevant to them meet their needs. Assessment of BiVo and BiPla by the companies For the second time after 2016/17, companies were asked about the BiVo and BiPla relevant to them. The first question asked whether the BiVo and BiPla meet the requirements of their company.2 Figure 15 summarises the responses from the 6,562 training companies that took part. In most cases, the BiVo and BiPla appear to meet company requirements. For 55 per cent of companies this was the case, and for a further 13 per cent it was even ‘fully the case’. For 27 per cent, the BiVo and BiPla only partly met company requirements. For 6 per cent, they did not or did not at all meet company requirements. These assessments have remained largely unchanged since the 2016/17 survey. In a second question, companies were asked to indicate the share of training content that is specified in the official training plan but not required in their company.3 According to Figure 16, companies consider 82 per cent of the training content defined in the BiPla to be rele6 TRAINING PLANS AND THE VOCATIONAL BACCALAUREATE 0% 20 % 40% 60% 80% 100% Yes, in all respects Yes, for the most part Neutral Not really Not at all 2 The question was actually phrased as follows: “With regard to the VET ordinance of SERI and the official training plan of the professional/trade organisation for the selected occupation, does the following statement apply to your company? ‘The VET ordinance and training plan meet the requirements of our company.’” 3 The question was actually phrased as follows: “What proportion of the training content specified in the official training plan for the selected occupation would you estimate is not required in your company?” Figure 15: Do the VET ordinances (BiVo) and training plans (BiPla) meet the company’s requirements? Company perspectives on costs, benefits and training quality 2025 47 vant for their operations. Although this figure is one percentage point lower than in 2016/17, it still appears very high given that training plans are intended to ensure comprehensive occupational skills that extend beyond the needs of individual companies. Nevertheless, the average responses differ across apprenticeship occupations. The confidence interval, shown by the fine lines, indicates the degree of statistical uncertainty. For the occupations shown in the lower part of the figure, the proportion of unnecessary training content is estimated to be below average, while for those in the upper part, a greater share of content in the training plan is deemed unnecessary for the company. Health and Social Care Worker, FVETC Healthcare Assistant, FVETD Pharmacy Assistant, FVETD Dental Assistant, FVETD Plumber, FVETD Certified Social Care Worker, FVETD Automotive Technician, FVETD Construction Electrician, FVETD Gardener, FVETD Painter, FVETD Other FVETC Electrician, FVETD Building and Grounds Custodian, FVETD Bricklayer, FVETD Carpenter, FVETD Other three-year FVETD Office Assistant, FVETC Medical Secretary and Assistant, FVETD Chef, FVETD Draughtsman/Draughtswoman, FVETD Other four-year FVETD Retail Assistant, FVETC Joiner, FVETD Automotive Mechatronics Technician, FVETD Polymechanic, FVETD Hairdresser, FVETD Commercial Employee, FVETD Retail Professional, FVETD Building Services Assistant, FVETC Automation Technician, FVETD Logistics Assistant, FVETC Information Technologist, FVETD Logistics Professional, FVETD Average 0% 5% 10% 15% 20% 25% 30% Figure 16: Proportion of content not required in training plan (BiPla) 48 Company perspectives on costs, benefits and training quality 2025 Companies also provide their apprentices with additional qualifications that are not covered by the BiPla. This may be the case, for example, when such qualifications are relevant to internal company processes. Nineteen per cent of companies report that they teach their apprentices additional qualifications beyond the BiPla – a significantly higher proportion than in the 2016/17 survey (12 per cent).4 Using data from the 2016/17 survey, Schweri et al. (2021) and Aepli et al. (2021b) examined whether a mismatch between the skills prescribed in the training plan and those required in the workplace results in costs for companies. According to the authors, this is indeed the case: both types of mismatch – skills not required or additional qualifications taught – are associated with higher gross training costs and a lower productive contribution from apprentices. These effects are statistically significant but not very large: e.g., for every ten percentage points of training content not required, the net benefit per apprenticeship year is around CHF 630 lower. The key finding from these analyses is that broader training plans are associated with costs for companies. Very broad skills profiles, such as those that would result from a significant reduction in the number of apprenticeship occupations, could therefore entail substantial costs for companies and lower their willingness to provide training. 6.2 Federal Vocational Baccalaureate1 from a company perspective The number of apprentices able to pursue a Federal Vocational Baccalaureate Type 1 which accompanies the apprenticeship depends in part on whether companies are willing to support or actively promote this option. A key factor here is the extent to which reduced time spent in the workplace – due to the higher number of school lessons – affects the cost–benefit balance. As Trede et al. (2020, pp. 17–20) show, the cost–benefit balance differed in the 2016/17 training year between apprenticeship places with and without BM1. Places with BM1 displayed a lower net benefit during training because apprentices spent less time in the company. Since BM1 apprentices are unevenly distributed across occupations, sectors and companies, these differences cannot be measured directly in the data but must be adjusted for using multivariate regression models (see Trede et al., 2020). Figure 17 presents the results of this estimation, based on 11,388 observations per apprenticeship year from 5,519 companies for the 2022/23 training year. The cost values were calculated for an average apprenticeship year in two hypothetical companies: one training no BM1 apprentices and the other training only BM1 apprentices. As in the 2016/17 training year, the average net benefit per training year in 2022/23 is lower for apprentices enrolled in BM1, once differences in company characteristics are controlled for. The reduction in net benefit is just under CHF 3,000, somewhat smaller than in 2016/17 (just under CHF 4,000), but statistically clearly significant. Nevertheless, companies training BM1 apprentices also achieve a positive net benefit on average already during the training period. The difference between apprenticeship places with and without BM1 is due to the lower productive output of the apprentices – not to higher gross costs. Companies fulfil their training responsibility and do not compensate for lower attendance by cutting back on their training input. Instead, apprentices are less frequently assigned to productive tasks, particularly unskilled activities. 4 The question was actually phrased as follows: “In your company, were apprentices in the 2022/23 training year taught additional qualifications beyond the content specified in the official training plan for the selected occupation?” Company perspectives on costs, benefits and training quality 2025 49 These results do not take into account that BM1 enables companies to attract particularly capable apprentices. If they succeed in retaining them as skilled workers, the company gains an additional benefit – especially where recruiting and onboarding skilled workers through the external labour market is a demanding process. 0 10 20 30 40 In CHF thousands No BM1 apprentices Only BM1 apprentices Gross costs Productive output Net benefit Figure 17: Gross costs, productive output and net benefit per average apprenticeship year with or without BM1 apprentices 50 Company perspectives on costs, benefits and training quality 2025 Since apprentices spend the majority of the training period in their host company, these play a key role in their skills development. It is therefore unsurprising that the quality of company training has been the subject of research for many years (Schwander, 1949; Häfeli et al., 1981). Apprentices (and their parents) have a strong interest in receiving optimal support from the company, not least to reduce the risk of terminating the apprenticeship contract prematurely or failing the final examinations. As part of this latest cost–benefit survey, companies were asked about their training processes, i.e. how they work with apprentices and allocate tasks to them. This chapter evaluates those responses, providing an overview of training quality in Swiss companies. 7.1 What is the definition of in-house training quality? Despite the obvious importance of training companies for vocational education and training, in-house training quality is difficult to define and measure (Berger et al., 2023). Quality models often distinguish between input, process and output quality (Gonon, 2008). From an economic perspective, output quality is of particular significance: good training means that apprentices master their occupation and are able to act competently in all professional situations. This definition of quality, however, does not specify how companies can achieve a high level of output quality. Research in work and organisational psychology, as well as in vocational and business education, has identified company processes that make employees in general and apprentices in particular both satisfied and productive (Hackman & Oldham, 1975; Morgeson & Humphrey, 2006; Böhn & Deutscher, 2021). In a company survey, Negrini et al. (2016) used questionnaire items developed in this literature. Using data from 335 training companies for the occupations of cook and painter, they demonstrated a significant correlation between their quality items and the number of premature apprenticeship contract terminations. Measuring training quality This cost–benefit survey used eight items as measures of process quality in training companies, most of them taken from Negrini et al. (2016). Respondents were asked to rate the following items: 1. Planning: I plan the training units in advance. 2. Autonomous solutions: I allow the apprentices to find autonomous solutions when working on an assignment. 3. Variety: I ensure that the apprentices are given a variety of tasks. 4. Feedback: I give the apprentices regular feedback on their performance. 5. Insight: The apprentices gain insights into all phases of the work and production process. 6. Responsibility: I assign the apprentices tasks that carry real responsibility. 7. Listening: I make an effort to listen carefully to the apprentices. 8. Optimal workload: I make sure the apprentices are not overwhelmed. In addition to these eight items, further information that influences training quality was collected from training companies. This notably includes data on the time resources available to company training managers and on the available infrastructure. These variables are understood as measures of input quality and are later used to examine whether differences between companies in the training processes described above can be attributed to differing levels of resources. 7.2 How do training companies assess their training quality? The survey prompted companies to indicate how accurately they felt the eight statements above applied to their own training, using a scale from one (not at all) to seven (in all respects). On the whole, companies rated themselves very positively, as shown in Figure 18: the average responses for all eight items lie between 5.1 and 6.2. What matters for the analyses that follow, however, is not the generally 7 IN-HOUSE TRAINING QUALITY Company perspectives on costs, benefits and training quality 2025 51 high averages, but the variation between companies. It can be assumed that higher self-assessments also correspond to a higher level of training quality. In addition to surveying companies, Negrini et al. (2016) also asked their apprentices about perceived training quality and found that, on average, apprentices gave lower ratings than their company. Nevertheless, the responses of apNeutral Yes, to some extent Yes, for the most part Yes, in all respects I make an effort to listen carefully to the apprentices. I assign the apprentices tasks that carry real responsibility. I allow the apprentices to find autonomous solutions when working on an assignment. I make sure the apprentices are not overwhelmed. The apprentices gain insights into all phases of the work and production process. I give the apprentices regular feedback on their performance. I plan the training content in advance. I ensure that the apprentices are given a variety of tasks. 4 5 6 7 Figure 18: Statements on training quality, as rated by companies prentices and companies correlated strongly, meaning that in companies with higher self-ratings, apprentices also gave higher assessments, and vice versa. This suggests that companies’ self-assessments do indeed reflect differences in training quality – at least as seen by their apprentices. 52 Company perspectives on costs, benefits and training quality 2025 Strong training companies stand out because they maintain good processes in all eight areas. It can therefore be assumed that the eight items are positively correlated. In other words, if a company scores above average on one item, it is more likely to score above average on the others too. This assumption is confirmed in Figure 19, which shows the pairwise correlation coefficients of all items. Theoretically, correlation coefficients can range between minus one and plus one, but in this case all correlations are above zero, albeit consistently below 0.6. The higher the coefficient, the bluer the field. Only the two items ‘Feedback’ and ‘Listening’ show a correlation of more than 0.5. A clear link between feedback and listening seems plausible, as both are likely to occur in conversations between trainers and apprentices. Companies that take time more frequently for such conversations are likely to score higher on both items. 0.32 0.32 0.33 0.41 0.33 0.47 0.21 0.26 0.42 0.41 0.21 0.38 0.44 0.41 0.47 0.31 0.33 0.46 0.53 0.40 0.46 0.34 0.35 0.34 0.38 0.29 0.36 0.49 Autonomous solutions Variety Feedback Insight Responsibility Listening Optimal workload Autonomous solutions Planning Variety Feedback Responsibility Insight Listening Figure 19: Correlations between individual items on training quality Company perspectives on costs, benefits and training quality 2025 53 7.3 Five training types of training companies The eight items make it possible to group companies according to their training practices. A cluster analysis assigns companies to training types (clusters) so that companies within a type are as similar as possible across the eight items, while differing as much as possible from companies in other types. Five training types can be identified, each characterised below by a short label and a description of their training practices. These labels were intentionally chosen to be clear and somewhat overstated so that their characteristics can be easily understood. 1. TOP: Companies in this type achieve consistently very high training quality. 2. OKAY: Companies in this type show satisfactory – i.e. average – training quality across all items. 3. SPONTANEOUS: Companies in this type provide satisfactory to good training quality across all items, but training is planned much less frequently in advance. 4. AUTHORITARIAN: Companies in this type rarely allow apprentices to find autonomous solutions. They also plan training less in advance and pay less attention to preventing apprentices from feeling overwhelmed. On the remaining items, however, they generally score slightly higher than the OKAY group. 5. BASIC: Companies in this type are characterised by consistently below-average training quality. The five types and the associated companies are presented in more detail in Table 17. 54 Company perspectives on costs, benefits and training quality 2025 TOP 24.7% SPONTANEOUS 18.6% OKAY 34.4% AUTHORITARIAN 7.8% BASIC 14.5% Training quality Planning 6.2 3.45.7 5.0 4.0 Autonomous solutions 6.4 5.75.8 3.3 4.6 Variety 6.6 5.75.6 5.8 4.6 Feedback 6.7 5.75.8 6.0 4.5 Insight 6.7 6.25.9 6.3 4.8 Responsibility 6.5 6.05.7 5.6 4.5 Listening 6.8 6.26.1 6.3 4.9 Optimal workload 6.5 5.45.6 5.1 4.5 Company characteristics Training requirements Trainers have less time Poorer training infrastructure Company digitalisation More impacted by digitalisation Less impacted by digitalisation Company size More very small companies (<10 employees) Fewer very small companies (<10 employees) Language region (reference: German) – French + Italian + French + Italian – French – Italian – French – Italian + French Trainers have more time Better training infrastructure Trainers have less time Table 17: Types of training Company perspectives on costs, benefits and training quality 2025 55 The upper half of the table shows the average values of companies for all eight questionnaire items on training quality. The main findings have already been described above in the descriptions of the individual types. The coloured bars indicate the share of all training companies in the sample assigned to each of the five types. Overall, more than three quarters of training companies belong to the three types that provide satisfactory to very good training quality (TOP, OKAY, SPONTANEOUS). This is an encouraging result. It should be noted, however, that the allocation of individual companies to the five types also depends on the exact specifications of the cluster analysis. The reported shares of the five training types should therefore be regarded as approximate values. The uneven distribution of roughly three quarters to one quarter is due to the fact that many companies reported high values, resulting in a concentration of responses at the upper end of the scale. Training quality is therefore not symmetrically distributed, as in a Gaussian normal distribution, but skewed to the right. This asymmetry suggests that most training companies strive to provide good training, with companies offering poorer training being less common. At the same time, the pattern may also reflect overestimation by companies themselves– as suggested by Negrini et al. (2016), who found apprentices tended to give lower ratings of training quality. Future studies on training quality should therefore aim to verify the distribution of training quality using representative data not based solely on company self-assessments. As Chapter 7.5 will show, however, company self-assessments are closely linked to the subsequent training success of their apprentices and therefore, despite some open questions, constitute a valid measure of training quality. 7.4 How do the companies in the different training types differ? The training companies in the sample were assigned to five training types solely based on their assessment of training processes. The lower part of Table 17 shows which other characteristics distinguish the companies in the different types – in other words, which types of companies are overor under-represented in each group. These figures are based on a multivariate regression analysis. This also means that the reported effects – such as the effect of time resources for training or of digitalisation – cannot be attributed to the separately controlled effects of company size, language region, economic sector and occupation. A decisive factor is whether companies provide trainers with sufficient time for training activities and good infrastructure. Companies that do so are more frequently found in the TOP group. Where time is scarce and/or infrastructure is suboptimal, a company is more likely to fall into the group with the lowest training quality. Such companies are also more often found in the group with spontaneous, less planned training practices. Apparently, trainers in this group manage to achieve a satisfactory to good level of training (process) quality despite unfavourable conditions and limited planning. The significance of time resources is not entirely surprising, as they are regarded in the literature as part of input quality. The strong association with process quality is empirically confirmed in the present analysis and also supports the findings of Lamamra et al. (2019) and Wenger and Lamamra (2023), which emphasise the importance of sufficient time for in-company trainers. Companies more affected by digitalisation are more likely to be in the TOP group and less likely to be in the BASIC group. It may be assumed that companies more exposed to digitalisation generally have better processes, which also translate into higher training quality. Very small companies with fewer than ten employees tend to be among those that plan training less, while being less often represented in the OKAY group. A strong effect is also evident in the language regions: companies from French-speaking Switzerland are less often represented in the first three types and more frequently in the type with authoritarian training as well as in the type with below-average training quality (BASIC). It remains unclear to what extent the lower popularity and weaker anchoring of dual vocational education and training in French-speaking Switzerland play a role 62 Company perspectives on costs, benefits and training quality 2025 Acemoglu, D., & Pischke, J. (1998). Why do firms train? Theory and evidence. Quarterly Journal of Economics, 113, 79–119. Acemoglu, D., & Pischke, J. (1999). The Structure of Wages and Investment in General Training. Journal of Political Economy, 107(3), 539–572. https://doi. org/10.1086/250071 Aepli, M., Kuhn, A., & Schweri, J. (2021a). Culture, norms, and the provision of training by employers: Evidence from the Swiss language border. Labour Economics, 73, 1–21. https://doi.org/10.5281/zenodo.5749913 Aepli, M., Schweri, J., & Kuhn, A. (2021b). Bildungspläne sichern breite Ausbildung – verursachen aber auch Kosten für die Lehrbetriebe. EHB-Studie zur Passung von Bildungsplänen und betrieblichen Bedürfnissen. Transfer. Berufsbildung in Forschung und Praxis, 6(3). Af Burén, P., & Schweri, J. (2024). Firms’ training processes and their apprentices’ education success. Swiss Leading House – Economics of Education, Firm Behaviour, Training Policies, Working Paper No. 225. Backes-Gellner, U., & Lehnert, P. (2023). Berufliche Bildung als Innovationstreiber: Ein lange vernachlässigtes Forschungsfeld. Perspektiven der Wirtschaftspolitik, 24(1), 85–97. Bardeleben, R. von, Beicht, U., & Fehér, K. (1995). Betriebliche Kosten und Nutzen der Ausbildung: Repräsentative Ergebnisse aus Industrie, Handel und Handwerk. In German Federal Institute for Vocational Education and Training BIBB (Eds.), Berichte zur beruflichen Bildung, issue 187. Bielefeld: Bertelsmann. Beicht, U., Walden, G., & Herget, H. (2004). Kosten und Nutzen der betrieblichen Berufsausbildung in Deutschland. Bielefeld: Bertelsmann. Berger, J.-L., Wenger, M., & Sauli, F. (2023). Engagement en formation professionnelle initiale duale et perceptions de la qualité de la formation. Didactique, 4(2), 33–64. Böhn, S., & Deutscher, V. (2021). Development and validation of a learning quality inventory for in-company training in VET (VET-LQI). Vocations and Learning, 14, 23–53. Bonin, H., Fries, J., Hillerich-Sigg, A., Maier, M.F., & Walter, T. (2014). Begleitforschung "Auswirkungen des Ausbildungsbonus auf den Ausbildungsmarkt und die öffentlichen Haushalte". Abschlussbericht. Berlin: German Federal Ministry of Labour and Social Affairs (BMAS). Dionisius, R., Muehlemann, S., Pfeifer, H., Walden, G., Wenzelmann, F., & Wolter, S. C. (2009). Costs and benefits of apprenticeship training. A comparison of Germany and Switzerland. Applied Economics Quarterly, 1, 7–37. Dong, Y., & Peng, CY.J. (2013). Principled missing data methods for researchers. SpringerPlus, 2, 222. https://doi.org/10.1186/2193-1801-2-222 Dornmayr, H., Graf, L., Kroll, S., Markowitsch, J., Neff, D., & Neumann, J. (2025). Indikatoren zur dualen Berufsbildung 2024. Germany – Austria – Switzerland. Research report. Bonn – Vienna – Zollikofen: Austrian Federal Ministry for Labour and Economy. Euwals, R., & Winkelmann, R. (2004). Training intensity and first labor market outcomes of apprenticeship graduates. International Journal of Manpower, 25(5), 447–462. Federal Statistical Office (FSO) (2024). Aktuell. Berufliche Grundbildung: Lernorte der Lernenden. Neuchâtel: Federal Statistical Office (FSO). Fuhrer, M., & Schweri, J. (2010). Kosten und Nutzen von zweijährigen beruflichen Grundbildungen. Schlussbericht. Zollikofen: Swiss Federal Institute for Vocational Education and Training. REFERENCES Company perspectives on costs, benefits and training quality 2025 63 Gehret, A., Aepli, M., Kuhn, A., & Schweri, J. (2019). Lohnt sich die Lehrlingsausbildung für die Betriebe? Resultate der vierten Kosten-Nutzen-Erhebung. Zollikofen: Swiss Federal Institute for Vocational Education and Training. Gonon, P. (2008). Qualitätssicherung in der beruflichen Bildung. Eine Bilanz. In E. Klieme & R. Tippelt (Eds.), Qualitätssicherung im Bildungswesen (p. 96–107). Beltz. https://www.pedocs.de/frontdoor.php?- source_opus=7271 Hackman, J.R., & Oldham, G.R. (1975). Development of the Job Diagnostic Survey. Journal of Applied Psychology, 60, 159–170. https://dx.doi.org/10.1037/ h0076546 Häfeli, K., Frischknecht, E., & Stoll, F. (1981). Schweizer Lehrlinge zwischen Ausbildung und Produktion. Forschungsergebnis einer Lehrlingsuntersuchung. Muri: Cosmos Verlag. Jansen, A., Strupler Leiser, M., Wenzelmann, F., & Wolter, S.C. (2015). Labour market deregulation and apprenticeship training: A comparison of German and Swiss employers. European Journal of Industrial Relations, 21(4), 353–368. Kriesi, I., Bonoli, L., Grønning, M., Hänni, M., Neumann, J., & Schweri, J. (2022). Spannungsfelder in der Berufsbildung international und in der Schweiz – Entwicklungen, Herausforderungen, Potenziale. OBS SFUVET Trend Report 5. Zollikofen: Swiss Federal University for Vocational Education and Training SFUVET. Kriesi, I., Neumann, J., Schweri, J., Griga, D., Kuhn, A., Schmid, E. und Baumeler, C. (2016). "Bleiben? Gehen? Neu beginnen? Häufigkeit, Ursachen und Folgen von Lehrvertragsauflösungen". Trend Report 1 of the Swiss Observatory for Vocational Education and Training. Zollikofen: Swiss Federal Institute for Vocational Education and Training. Kuhn, A., Gehret, A., & Schweri, J. (2025). Methoden bericht zur fünften Kosten-Nutzen-Erhebung der beruflichen Grundbildung. Zollikofen: Swiss Federal University for Vocational Education and Training. Kuhn, A. (2024). Eine Flotte aus Jollen und Barken, Kuttern und Tankern. Transfer. Berufsbildung in Forschung und Praxis, 9(1). Kuhn, A., & Schweri, J. (2019). Methodenbericht zur vierten Kosten-Nutzen-Erhebung der beruflichen Grundbildung. Zollikofen: Swiss Federal Institute for Vocational Education and Learning. Kuhn, A., Schweri, J., & Wolter, S. C. (2022). Local norms describing the role of the state and the private provision of training. European Journal of Political Economy, 75. https://doi.org/10.1016/j.ejpoleco.2022.102226 Lamamra, N., Duc, B., & Besozzi, R. (2019). Im Herzen der dualen Berufsbildung: Berufsbildnerinnen und Berufsbildner. Forschungsergebnisse und Massnahmen für die Akteurinnen und Akteure der Berufsbildung. Zollikofen: Swiss Federal Institute for Vocational Education and Training. Lindley, R. M. (1975). The demand for apprentice recruits by the engineering industry, 1951–71. Scottish Journal of Political Economy, 22(1), 1–24. Little, R.J.A., & Rubin, D.B. (2019). Statistical Analysis with Missing Data. 3rd edition. Hoboken, NJ: Wiley. Morgeson, F. P., & Humphrey, S. E. (2006). The Work Design Questionnaire (WDQ): Developing and validating a comprehensive measure for assessing job design and the nature of work. Journal of Applied Psychology, 91(6), 1321–1339. https://doi.org/10.1037/00219010.91.6.1321 Mühlemann, S. (2013). Der Einfluss der Internationalisierung auf die arbeitsmarktorientierte Bildung in der Schweiz. Economics of Education Working Paper Series No. 92, University of Zurich. 64 Company perspectives on costs, benefits and training quality 2025 Mühlemann, S. (2019). Measuring Performance in Vocational Education and Training and the Employer’s Decision to Invest in Workplace Training. In D. Guile & L. Unwin (Eds.), The Wiley Handbook of Vocational Education and Training (p. 187–206). https://doi. org/10.1002/9781119098713.ch10 Mühlemann, S., Pfeifer, H., Walden, G., Wenzelmann, F., & Wolter, S.C. (2010). The financing of apprenticeship training in the light of labor market regulations. Labour Economics, 17(5), 799–809. Mühlemann, S., Schweri, J., Winkelmann, R., & Wolter, S.C. (2007a). An empirical analysis of the decision to train apprentices. LABOUR – Review of Labour Economics and Industrial Relations, 21(3), 419–441. Mühlemann, S., & Wolter, S.C. (2014). Return on investment of apprenticeship systems for enterprises: Evidence from cost–benefit analyses. IZA Journal of Labor Policy, 3(1), 1–22. doi.org/10.1186/2193-90043-25 Mühlemann, S., Wolter, S.C., Fuhrer, M., & Wüest, A. (2007b). Lehrlingsausbildung – ökonomisch betrachtet. Chur: Rüegger. Müller, B., & Schweri, J. (2012). Die Betriebe in der dua len Berufsbildung: Entwicklungen 1985 bis 2008. Neuchâtel: Federal Statistical Office Negrini, L., Forsblom, L., Gurtner, J. L., & Schumann, S. (2016). Is there a relationship between training quality and premature contract terminations in VET? Vocations and Learning, 9(3), 361–378. Rupietta, C., & Backes-Gellner, U. (2018). How firms’ participation in apprenticeship training fosters knowledge diffusion and innovation. Journal of Business Economics. doi.org/10.1007/s11573-018-0924-6 Schönfeld, G., Jansen, A., Wenzelmann, F., & Pfeifer, H. (2016). Kosten und Nutzen der dualen Ausbildung aus Sicht der Betriebe. Ergebnisse der fünften BIBB-Kosten-Nutzen-Erhebung. Bielefeld: Bertelsmann. Schwander, A. (1949). Die Ausbildung des Lehrlings im Betrieb. Verlag der deutschschweizerischen Lehrlingsämter-Konferenz. Schweri, J., Aepli, M., & Kuhn, A. (2021). The Costs of Standardized Apprenticeship Curricula for Training Firms. Empirical Research in Vocational Education and Training, 13(16). Schweri, J., & af Burén, P. (2024). Abwechslungsreiche Aufgaben im Lehrbetrieb machen Lernende erfolgreicher. Transfer. Berufsbildung in Forschung und Praxis, 9(12). Schweri, J., Mühlemann, S., Pescio, Y., Walther, B., Wolter, S.C., & Zürcher, L. (2003). Kosten und Nutzen der Lehrlingsausbildung aus der Sicht Schweizer Betriebe. Chur/Zurich: Rüegger. Strupler, M., & Wolter, S.C. (2012). Die duale Lehre: eine Erfolgsgeschichte auch für die Betriebe: Ergebnisse der dritten Kosten-Nutzen-Erhebung der Lehrlings ausbildung aus der Sicht der Betriebe. Zurich: Rüegger. Swiss National Bank (2025). Foreign exchange rates – Year. https://data.snb.ch/en/topics/ziredev/cube/ devkua (last accessed: 27 May 2025). Trede, I., Hänni, M., Leumann, S., Neumann, J., Gehret, A., Schweri, J., & Kriesi, I. (2020). Berufsmaturität. Bildungsverläufe, Herausforderungen und Potenziale. OBS SFIVET Trend Report 4. Zollikofen: Swiss Federal Institute for Vocational Education and Training. van Buuren, S. (2007). Multiple imputation of discrete and continuous data by fully conditional specification. Statistical Methods in Medical Research, 16(3), 219–242. van Buuren, S. (2018). Flexible Imputation of Missing Data. Second Edition. Boca Raton, FL: CRC Press. Walther, B., Schweri, J., & Wolter, S.C. (2005). Shall I train your apprentice? An empirical investigation of outsourcing of apprenticeship training in Switzerland. Education + Training, 47(4), 251–269. Company perspectives on costs, benefits and training quality 2025 65 Wenger, M., & Lamamra, N. (2023). Les besoins et préférences en matière de formation continue des personnes formatrices d’apprenti-e-s. Analyses de l’enquête en ligne – Rapport final. Renens: Swiss Federal University for Vocational Education and Training SFUVET. Wenzelmann, F., Schönfeld, G., Linckh, C., & Pfeifer, H. (2025). Eigene Ausbildung oder externe Fachkräftegewinnung – mit welchen Kosten müssen Betriebe rechnen? Ergebnisse der BIBB-Kosten-Nutzen-Erhebung 2022/2023. BIBB Report 2. Bonn. Wolter, S.C. (2008). Ausbildungskosten und -nutzen und die betriebliche Nachfrage nach Lehrlingen. Perspektiven der Wirtschaftspolitik, 9 (Special Issue), 90–108. 66 Company perspectives on costs, benefits and training quality 2025 Figure 1: Distribution of net benefit 24 Figure 2: Satisfaction with the cost–benefit balance of company-based training 24 Figure 3: Breakdown of gross costs in FVETC apprenticeships 26 Figure 4: Time spent by apprentices at the company workplace in FVETC apprenticeships 27 Figure 5: Breakdown of gross costs in three-year FVETD apprenticeships 29 Figure 6: Time spent by apprentices at the company workplace in three-year FVETD apprenticeships 30 Figure 7: Breakdown of gross costs in four-year FVETD apprenticeships 31 Figure 8: Time spent by apprentices at the company workplace in four-year FVETD apprenticeships 32 Figure 9: Gross costs, productive output and net benefit per average apprenticeship year, by company size 34 Figure 10: Gross costs, productive output and net benefit per average apprenticeship year, by sector 35 Figure 11: Gross costs, productive output and net benefit per average apprenticeship year, by major region 36 Figure 12: Average net benefit of the most common FVETC apprenticeship occupations 37 Figure 13: Average net benefit of the most common three-year FVETD apprenticeship occupations 37 Figure 14: Average net benefit of the most common four-year FVETD apprenticeship occupations 38 Figure 15: Do the VET ordinances (BiVo) and training plans (BiPla) meet the company’s requirements? 46 Figure 16: Proportion of content not required in training plan (BiPla) 47 Figure 17: Gross costs, productive output and net benefit per average apprenticeship year with or without BM1 apprentices 49 Figure 18: Statements on training quality, as rated by companies 51 Figure 19: Correlations between individual items on training quality 52 Figure 20: Reasons for not training apprentices in-house 57 Figure 21: Reasons for training apprentices in-house 59 Figure 22: Assessment of measures to boost training activity 60 LIST OF FIGURES Company perspectives on costs, benefits and training quality 2025 67 Table 1: Overview of gross costs 8 Table 2: Overview of productive output 9 Table 3: Responses in the two gross samples 18 Table 4: Composition of the net sample of training companies 19 Table 5: Composition of the net sample of non-training companies 21 Table 6: Gross costs, productive output and net benefit per FVETC apprenticeship (in CHF) 25 Table 7: Monthly gross wage of apprentices in FVETC apprenticeships (in CHF) 26 Table 8: Gross costs, productive output and net benefit per three-year FVETD apprenticeship (in CHF) 28 Table 9: Monthly gross wage of apprentices in three-year FVETD apprenticeships (in CHF) 28 Table 10: Gross costs, productive output and net benefit per four-year FVETD apprenticeship (in CHF) 31 Table 11: Monthly gross wage of apprentices in four-year FVETD apprenticeships (in CHF) 32 Table 12: Average recruitment and induction costs per skilled worker and benefit of continued employment, by company size 39 Table 13: Extrapolation of costs and benefits for the 2022/23 apprenticeship year 39 Table 14: Comparison of the costs and benefits of vocational training in Switzerland and Germany 40 Table 15: Multivariate analysis of costs and benefits 2022/23 43 Table 16: Multivariate analysis of costs and benefits 2016/17 in comparison to 2022/23 45 Table 17: Types of training 54 Table 18: Determinants of training participation 61 TABLES Swiss Federal University for Vocational Education and Training SFUVET Kirchlindachstrasse 79 CH-3052 Zollikofen +41 58 458 27 00 www.sfuvet.swiss [email protected]