Identical monthly calibrations can lead to different prices for longer-term claims due to hidden dependencies.

Financial models try to match the prices of stock market options across several future dates. Standard methods glue together separate one-step models. This stitching forces a shortcut: the model assumes tomorrow depends only on today, wiping out older market history. The researchers created a global framework that fits all dates at once without this shortcut. They found that exact local and global solutions match in basic building blocks, yet stitched models form a strict subset of all valid path models. Because stitched models throw away history, models that match market prices equally well can still assign different prices to multi-period payouts. To solve this, the researchers built a customized descent algorithm. Tests show their scheme keeps required targets about 25 times tighter than an older cyclic method and keeps pricing errors below 0.70 volatility points.

Why it matters

Traders and risk managers rely on accurate pricing of multi-period financial claims. This work shows that standard methods might miss hidden dependencies, leading to incorrect pricing and potential financial losses.

Questions answered