This Annex presents the set of principles for developing the semiconductors production database, based on the discussions of the Semiconductor Informal Exchange Network on a semiconductor taxonomy (OECD, 2024[1]). These principles reflect a prioritisation to help attain the objectives outlined in Section 1.
Trust. The taxonomy, any data sharing activities and resulting databases developed in the context of the Network should build on co-operation and trust among governments and stakeholders in the semiconductor industry to ensure that data and related analyses meet shared goals. Respecting constraints associated with sharing granular industry data would be particularly important to building trust with stakeholders.
Availability. The taxonomy should reflect available data. Areas for which data are not available are not included. Areas for which data availability is very limited are identified as such and, as in any database, there may be numerous missing values. Data analyses are only possible insofar as data are available. Any data collection efforts should complement and rely on, rather than duplicate, existing data efforts by governments, industry and other stakeholders.
Comparability. Comparison across classifications used in different economies and jurisdictions should be as simple and easy as possible. A taxonomy should also be consistent over time, thus allowing for analysis that can build on historical data.
Tractability. Semiconductor manufacturing processes are highly complex. The taxonomy should strive to simplify as much as possible to obtain a tractable taxonomy, noting that this may lead to nuanced differences in semiconductor manufacturing processes and products being obscured.
Adaptability. The taxonomy should be able to adapt as semiconductor production processes and products evolve, keeping abreast of innovation and limiting the needs for revisions of the taxonomy, but without prejudice to the above principle of comparability (through time).1
Granularity. Allow for the highest level of disaggregation possible with available data. More granular data offers additional information and allows for more detailed analysis, noting that a higher level of granularity might increase complexity. Input from the Network during its second meeting on 19-20 September 2023 highlighted the importance of ensuring that policymakers can draw meaningful insights from the data. This suggests that availability and tractability should be prioritised over granularity. Granularity should not hinder efforts to build trust and ensure the longstanding engagement of stakeholders. Granularity can be increased in further versions of the taxonomy, if needed and upon availability of the corresponding data and sufficient resources.
These principles may be reviewed in the future to ensure that they remain applicable in practice and well-suited for the objectives of the SIEN – e.g. granularity might become more relevant with the development of the resulting semiconductor production database. The relative importance of the different principles may also change depending on how critical different chips are – e.g. when analysing more critical/vulnerable and non-substitutable chips it might be worth considering putting more weight on granularity and adaptability.