STARTING SITUATION
SAS Willink issued two ten-year bonds convertible into shares at an interest rate of 8% per annum, subscribed by two French venture capital funds and a UK company.
ADMINISTRATIVE CONTROL
The tax authorities questioned the deductibility of the financial charges incurred by this transaction, insofar as the fraction of interest paid to the funds and the British company exceeded the rate stipulated in article 39-I-3° of the CGI. In addition, the tax authorities considered that the interest due on the differential constituted a gift to the British company.
To justify its position, the company subsequently produced a rate study using Riskcalc software. This study is based on a model calculating the probability of default in the short and long term, and then associates implicit scoring. A search for comparable transactions on the open market was then carried out using the SetP Capital IQ database.
DECISION OF THE CAA DE PARIS
In rejecting the company's argument, the CAA dismissed the results obtained from the Riskcalc scoring tool, finding that it was a statistical model based on historical quantitative data from companies that were not representative of the market, since defaulting companies were over-represented; that it only took into account a dozen or so financial data items supplied by the company itself; and that there was no evidence that the risk rating obtained using this tool would adequately take into account all factors recognized as predictive, and in particular the specific characteristics of the business sector concerned.
DECISION BY THE CONSEIL D'ÉTAT
The Conseil d'Etat considered that the Court, which had followed the tax authorities, had erred in law by rejecting the probative value of the Riskcalc tool used, and by giving preference to the rating agencies. In rejecting the probative value of the risk estimate obtained from the Riskcalc scoring tool, without investigating whether elements relating to the use of this tool in the case in question or derived from other elements of comparison led to calling it into question, the Court vitiated its ruling with an error of law.
OUR ANALYSIS
IMPOSSIBLE PROOF" PUT TO THE TEST BY CASE LAW
Article 212-I, which leads to the application of the reference rate referred to in article 39-1-3 of the CGI, has already undergone a clear shift under the impetus of the tax judge. For the record, when a loan is granted by an affiliated company, interest is deductible up to the limit of that calculated on the basis of a rate defined in article 39-1-3° of the CGI. However, the borrowing company benefits from a mechanism of proof to the contrary: it may deduct interest calculated on the basis of a higher rate if it demonstrates that the latter corresponds to the rate it could have obtained from independent financial institutions or organizations under similar conditions (article 212-I-a of the CGI). The administration's strict reading of the texts had led to a mechanism of "impossible proof" according to practitioners. However, the tax judge has tempered the rigidity induced by administrative practice by validating the fact that (i) alternative searches for comparables may not be contemporaneous with the financial transactions (e.g.: TA Paris, June 7, 2018, Paul Ka); the credit rating systems employed are deemed to sufficiently reflect the intrinsic elements of the economic markets concerned (e.g.: CE, Dec. 29, 2021, Apex Tool); or (iii) the rate may be determined by reference to the bond market (CE opinion, Jul. 10, 2019, SAS Wheelabrator Group).
HOW IS CREDIT RATING DEFINED?
Determining the borrower's credit rating is the first step in the analysis, insofar as this indicator reveals the company's overall solvency. On this point, the Conseil d'Etat formally acknowledges that ratings obtained using "RiskCalc" type tools are admittedly more approximate than a credit rating that might be carried out by a rating agency. However, it pragmatically admits that "the use of such a rating is not necessarily appropriate, given its cost, in an intra-group transaction". On the other hand, contrary to what the Paris CAA ruled, this tool does take into account the company's sector of activity, which must be entered by the user. In fact, the resulting ratings are based on data taken from the company's accounts, without the latter being able to modify the parameters used by the application. In this case, the Minister did not dispute the overall robustness of the application. The reference to such a tool can therefore be considered sufficiently reliable to justify a company's risk profile. Taxpayers therefore have two ways of establishing their credit rating: using a specialized computer tool (e.g. RiskCalc), or referring directly to a rating agency. It should be noted that the use of the RiskCalc tool has already been validated in the Studialis ruling (CAA Paris, Oct. 22, 2020).
REBALANCING THE BURDEN OF PROOF
The burden of proof under article 212-1 of the CGI does indeed shift to the taxpayer. However, the tax judge has gradually repositioned the administration in the system of proof, making it responsible for demonstrating the inaccuracy of the taxpayer's analysis. In so doing, the Conseil d'Etat laid the foundations for a 3-step waltz:
If the rate referred to in article 39-1-3 is not applied, the taxpayer bears the full burden of proof.
The latter can nevertheless carry out alternative economic analyses using specialized statistical and digital tools.
These alternative analyses are deemed reliable, unless the administration can demonstrate that the approach is flawed. For example, the tax authorities may provide detailed criticism of the use made of the RiskCalc tool in the case in point, or evidence that the resulting valuation was flawed in the particular case. It can also demonstrate that reference to the bond market does not constitute a viable reference when such loans constitute, in each case under consideration, an unrealistic alternative to a traditional intra-group loan.