TABLE A.12 Contingent liabilities (VaR) per product and at portfolio level in France

The “Loss" in the portfolio, used to estimate the impact in the agricultural sector and the contingent appear to potentially overestimate losses, compared to the thresholds of 20% or 30% below average that are adopted in CAP ARM programs, it should be considered that the estimation carried out in this analysis is based on less variable national level yield references, and that nation-wide yield shortfalls below the aver- age may be compatible with farm level losses of 20% or 30%. This is a feature of the current analysis and must be addressed in more depth in the future. As a first next step, it would be recommended to perform a sub-national analysis based on Eurostat data at NUTs2 level (Nomenclature des unités territoriales statistiques). The Value-at-Risk (VaR) is used to estimate Contingent Liability (CL), and it quantifies a Loss that might be exceeded with a given annual probability. For instance, based on the actuarial analysis, there is a 1% probability (1-in-100) that losses at portfolio level exceed EUR 7.3 billion, or 10.6% of the reference value. Error! Reference source not found. shows the Contingent Liabilities for each product due to produc- tion losses, as measured by the Value at Risk (VaR). The diversification effect, which arises from the non-perfect correlation among products, significantly reduces Contingent Liabilities. The portfolio CL without diversification is calculated as the sum of the VaR of each crop, a highly conservative estimate that implicitly assumes all products are perfectly positively cor- related. In contrast, the portfolio VaRs were estimated considering the diversification effect among products. When crops are not perfectly correlated, adverse events affecting one product are less likely to impact others to the same extent, resulting in lower overall risk for the portfolio. For example, at a 1% exceedance probability EUR 7.3 billion, and therefore the diversification benefit is EUR 8.3 billion. This pattern holds across all exceed-