How do we measure the semiconductor-intenstity of firms?
This is known as the firm’s degree of ‘semiconductor-intensity’, and in order to determine this for each firm, a standard approach is required that takes firms’ heterogeneity into account. Statistics Netherlands (CBS) combines two types of data in order to do this: data on the goods that are exported by firms, and data on the role that firms play in the wider supply chain.
From data on exports to an initial estimate
The first step is based on data on the international trade in goods. For each firm, an assessment is made to determine which goods are exported and to what extent these appear on a list of semiconductor-related products. This calculation looks only at domestically produced exports. Re-exports are not considered, because the transit of products manufactured elsewhere is not counted as part of the Netherlands’ own semiconductor sector. The list of semiconductor products is also dynamic, not set in stone: an initial draft, based on work done by the OECD (2019), was used by the European Commission’s Joint Research Centre (JRC) (2025) to produce 74 GN8 product codes, and then expanded to 127 GN8 codes following validation by JRC experts and input from the sector itself, in order to provide better coverage of raw materials and critical inputs (see European Commission: Joint Research Centre (2025), Table A1.1). This analysis uses one specific version of this list, which was applied to all the years surveyed.For certain goods, the link with semiconductors is clear: they appear exactly as listed and are counted in full as semiconductor activity. For other goods, the situation is more complex, because the standard classification of goods is less detailed than the semiconductor list: for example, one product code may cover a whole family of electronic components, only some of which actually relate to semiconductors. In such cases, an estimate is made as to what share of exports in that family of components come from known semiconductor companies. That share is used as a measure of how closely the product is related to the semiconductor sector. For goods that are even further removed from the core list, a similar estimate is made with one additional precautionary step: the firm itself is excluded from the calculation, so that a firm cannot confirm its own estimate. In this way, each firm receives an initial approximate estimate of its semiconductor-intensity based on its own trading patterns, along with an indication of how reliable that estimate is: the closer the goods are to the core list, the more reliable the figure.
What happens if the trade figures do not provide enough certainty?
The goods exported need to be sufficiently representative of the firm as a whole. If this is not the case, and if exports account for only a small share of a firm’s total turnover (less than 5 percent) for example, the trade figures will provide too narrow a picture of the firm’s overall economic activity and the trade-based estimation method is not used. There is one exception: firms whose exports, although limited, consist almost entirely of goods on the semiconductor list and are spread across several product categories. These firms are included in the figures based on their own trade data. To prevent firms that are close to this threshold from moving back and forth between the two methods from one year to the next, the threshold for remaining in one route (via this exemption) is slightly lower than the threshold for entering it.Even for companies that are assessed on the basis of their trade data, there is sometimes little data to go on - trade figures on only one or two types of goods, for instance. An estimate that is based solely on such data could fluctuate significantly from year to year, even with no real change in their business operations. To prevent this, Statistics Netherlands (CBS) combines the trade-based estimate with a more general estimate that is based on the typical semiconductor-related turnover of firms that play a similar role in the supply chain (e.g. chip manufacturers, chip designers, equipment suppliers or service providers). These role-based estimates are based on the professional judgement of RVO experts regarding how semiconductor-intensive a particular area of activity typically is.
Both of these sources are weighed up against each other: the more robust and unambiguous a firm’s trade figures are, the greater the weight they carry in the final figure. The more uncertain or less detailed the trade figures are, the more the result will reflect the average for that type of firm in the sector. For firms whose goods exports are almost all clearly included on the semiconductor list, the result will closely match the firm’s own trade data. For a firm with a more limited volume of trade or firms that are difficult to classify, the nature of the role in the wider supply chain is given greater weight.
Firms with multiple roles, and firms that are not involved in exports
Some firms operate in several different ways; for example, they act both as equipment suppliers and as service providers. In such cases, an average is taken of the characteristics associated with these different roles. For firms that have no goods exports, such as certain research or design firms that primarily supply knowledge and services rather than physical goods, only the role-based estimate is used with no adjustment based on trade data. The margin around the result is based on the variation observed among organisations that fulfil the same role but do have trade data available.This entire procedure was carried out separately for each year, covering the period from 2019 to 2025 inclusive. This means that changes in trade patterns or in the group of firms classified as part of the sector are reflected in the figures each year, rather than assuming a fixed share based on previous data.
Checks and final result
Before the figures were used, checks were carried out to ensure that the method was consistent. For example, it was examined whether firms whose goods exports corresponded less closely with the semiconductor list did in fact show a stronger tendency to correspond with the characterisation of their role, and whether firms whose trade figures relied on just a single type of goods were assigned a wider margin of error than firms with more trade data available. In both cases the checks were affirmative, which suggests that the balance between the firm’s own trade figures and the general knowledge of the sector had been applied consistently.The procedure provides an estimate of the semiconductor-intensity for each firm and each year, together with a range indicating the margin of error for that estimate. Results that were based mainly on the core semiconductor list or are based on a large amount of trade data have a smaller margin of error. Figures that are largely based on the experts’ characterisation of the firm and its role have a wider margin. The estimates for all the firms, when totalled and broken down according to the relevant categories, form the basis for the figures presented in this news release.