Version 1.0 of the OECD MAGIC database covers 482 consolidated industrial groups – including both publicly listed and non-listed companies – over the period 2005‑22, where applicable and depending on data availability. These firms were selected because they are the largest ones globally by sales revenue, output, or capacity in each of the 14 industrial sectors listed in Table A B.1. For steel, this includes 47 of the largest steel firms worldwide, which collectively accounted for about half of global steelmaking capacity and about 65% of global revenue for the steel sector in 2022.
The firms covered in the OECD MAGIC database are based in several different jurisdictions, with the resulting geographical split depending on these jurisdictions’ relative strength in global manufacturing. The home jurisdiction of the companies is taken to be either the location of their headquarters or their main place of business (e.g. measured by tangible assets or employees) in cases where the location of their headquarters corresponds to inexistent or insignificant volumes of manufacturing activity. This helps mitigate the issue of multinationals choosing to locate their headquarters in low-tax jurisdictions. In the case of the steel firms covered here, the sample comprises 13 firms based in the OECD and 34 based outside the OECD. These firms represent respectively 36% and 62% of total steelmaking production capacity for 2022 in OECD Member countries and partner economies respectively.
A consequence of looking at what firms obtain – as opposed to looking at what governments provide – is that the OECD MAGIC database does not distinguish between subsidies based on their policy objectives or design characteristics (e.g. eligibility criteria). Thus, the data do not make a distinction between one-off crisis measures and longstanding industrial policies seeking to affect firm behaviour durably. That said, in showing the persistence of various forms of support over time the database can shed light on the role of government support in shaping the market in a given sector. Moreover, while the MAGIC database does not differentiate between measures targeting R&D and those targeting investment in new industrial capacity, these issues have been explored in related sectoral studies by the OECD, for example in the semiconductor sectors (OECD, 2019[25]) and the rolling-stock (OECD, 2023[26]).
The above caveats notwithstanding, the OECD MAGIC database has the important advantage of capturing all support received by firms, irrespective of which jurisdiction provides it. This avoids selective coverage of policies caused by lack of government transparency and by the existence of numerous subsidy programmes at sub-national levels. It also helps capture the support provided through state enterprises such as state banks (OECD, 2024[16]), which might otherwise fall outside the perimeter of national budget reporting.
As of version 1.0, the OECD MAGIC database covers three types of subsidies, namely government grants, corporate income tax concessions and below-market borrowing:
Annual values for government grants are obtained directly from primary sources, which are generally corporate disclosures, at times complemented by government sources (all the while avoiding double counting). It has to be noticed that recorded grant amounts do not always represent the actual money flow provided to a company on a given year, as explained below in Box 8.
Estimates of corporate income tax concessions in the OECD MAGIC database do not refer to the tax revenue foregone by governments (i.e. tax expenditures) but to the tax savings of companies due to particular provisions of the tax code of the jurisdictions in which these companies operate. The estimates are primarily based on corporate disclosures and involve some amount of interpretation, judgment, and estimation by the OECD, such as where only certain subsidiaries of a group are eligible to a preferential rate of corporate income tax. Tax concessions are generally less internationally comparable than government grants owing to international differences in statutory tax rates.
Below-market borrowings are estimated by the OECD by comparing actual interest rates that firms are charged on their borrowings against hypothetical benchmark interest rates that would normally prevail in the market based on borrowers’ financial profile. The resulting differences are multiplied by the amount of debt to arrive at the volume of below-market borrowings for each year. Because below-market borrowings generally involve state banks offering loans at below-market rates, they tend to be much more common in jurisdictions in which the state plays a large role in the financial system.
While the MAGIC database represents a very significant improvement in the availability of data on industrial subsidies, it does not exhaust all possible ways in which governments can subsidise industrial companies. The full spectrum of possible support measures can include, for example, the effect of differential treatment in relation to regulatory measures or support resulting from export restrictions on upstream inputs. OECD estimates of the monetary value of below-market energy inputs exist for several firms and sectors but do not yet have sufficient coverage to warrant inclusion in the MAGIC database (OECD, 2023[27]). Estimates of the value of land acquired or rented by firms at below market prices are, for their part, non-existent due to a lack of sufficient data and methodological difficulties. However, anecdotal evidence does indicate that below-market land may confer significant benefits to certain industrial firms. These examples suggest that there is still considerable room for further improvements in the measurement of industrial subsidies and analysis thereof.
Monetary variables in the OECD MAGIC database are expressed in millions of current USD. They are converted from firms’ original reporting currencies into current USD using nominal, period average exchanges rates from the OECD and the IMF for every year covered by the OECD MAGIC database. Since many firms in the database are multinationals operating across several jurisdictions, the choice is made not to correct for inflation given the complications involved in choosing adequate deflators. This implies that changes in subsidy amounts (where not expressed in % of revenue) can reflect policy changes but also differences in inflation rates and exchange rate movements.
While the OECD makes its best efforts to cover the entire period 2005-22, there are several companies in the MAGIC database for which year coverage is shorter than 18 years. In most cases, this results from firms having entered a sector covered after 2005 (late entry), firms having left a sector covered before 2022 (early exit), or firms having merged with or having been acquired by another entity. In fewer cases, this does concern data that are genuinely missing. Overall, time coverage is generally good for the period 2010-22 but can be more limited in certain sectors (e.g. aluminium, shipbuilding, and solar cells and modules) over the period 2005-09.
Since firms were selected to cover as much of world steel production as possible (as opposed to be representative of each individual jurisdiction), the data certainly covers large steel players more than the smaller steel firms. It may or may not lead to higher level of government support than those witnessed in a “typical”, median steel firm. On one side, subsidies received by the large steel firms included in the sample may over-represent the subsidies received by smaller firms, since the fixed costs of applying for subsidies can be more easily borne by larger firms, and larger firms may be better positioned for lobbying government officials to obtain subsidies. On the other side, non-transparent steel firms for which the annual reports were not accessible through desk research were not included in the sample for obvious reasons, and those may be large recipients of government support. All in all, it is still necessary to exercise caution when assigning the results of the present study to a “typical” firm, especially if the firm is of a small size, as the results are probably not authoritative in covering all steel producing firms in any of the major steel-producing economies represented in the sample.
