At work this morning I found myself thinking about calibration. Lately, some people have tried to debunk the idea of calibrating financial models, but I still feel that calibration is one of the key ideas of financial modeling.
Financial models are models of relative value; in essence they tell you how to value something when you know the price of something else. Calibration is about satisfying that constraint: building a model to value one thing and making sure that it values known things correctly.
Of course, you can take anything too far, and the people who prefer mathematical style over content will do that. You can calibrate any old model, and that doesn’t make it right. Calibration alone isn’t enough. A model has to have appropriate dynamics. But once you’ve built in realistic dynamics, you need some kind of calibration to determine the parameters.
If the dynamics is badly wrong, calibration is of dubious help. You can’t make a silk purse etc. But if the dynamics is not too far off, then calibration is critical. You can’t make any purse at all without some stitching.