These two issues were the most highly rated across Europe, Asia, North America and South Africa.
Weaknesses in data quality and availability
The challenge of data quality is probably linked to the fact that, historically, insurance companies have not been strong in the area of data management. A history of mergers and takeovers has left much of the industry burdened with out-of-date and patched-up legacy IT systems. Extracting quality data from such systems and being able to report it in a useful way for EC modeling constitutes a significant challenge. In Europe, companies have been forced to address this problem as part of their Solvency II compliance, largely through significant investment in data warehouses and other IT infrastructure.
Another factor is the unavailability of data. EC modeling requires a significant amount of historic data to understand how portfolios have performed in the past. For example, to understand the risks in equity markets, firms need to understand what has happened in equity markets over the past 100 years. This is a challenging but achievable task, given the availability of data dating from the early 20th century until now. But when it comes to companies’ own internal data, such as the behavior of their portfolios or their customers over a similar length of time, the data often does not exist, which can compromise the inputs to the capital models. The lack of data can lead to a heavy reliance on expert judgment to fill the gap. But expert judgment is by definition subjective and can materially affect results. This can cause some concern about the objectiveness of the model and how much management can truly rely on its results.
Complexity of computations
EC calculations are complex and can take time to produce accurately. Many companies struggle with implementing EC modeling frameworks that produce accurate information quickly enough to allow them to respond in a meaningful way.
Companies are responding to these challenges by introducing simplifications and approximations into their models. While these are an inevitable part of implementing the model, it is vital that companies understand the limitations of these adjustments and set appropriate tolerance limits around their application.