Quant ideas: Do we need realistic models?

It is often assumed that realistic models are a prerequisite for successful risk management, but this is not necessarily true. While using any model blindly can be extremely dangerous, simple Black-Scholes or Ornstein-Uhlenbeck models are more than sufficient as long as they are properly used

In practice, realistic models are rarely used in risk management

It is a major and unspoken assumption in virtually any academic treatment of commodities that realistic models are a prerequisite for successful risk management. Yet, in practice, realistic models are rarely used. Models that are grossly simplified or even downright wrong are, as a rule, the mainstay of many a risk management operation. This presents a puzzle. If the realistic models are really a prerequisite, then why is the practice so different?

Obviously, part of the answer flows from the