The dog and the frisbee

Jump to: navigation, search

Authors of the Article

  • Andrew G. Haldane, Executive Director, Financial Stability, Bank of England
  • Vasileios Madouros, Economist, Bank of England

Publication Year



  • Regulators and risk managers try to build exhaustive models to capture risk. However the complexity of the models actually decreases their efficiency while increasing their costs.
  • Complexity hurts models because: gathering information is costly in terms of cash and time; weighting factors is near-impossible (e.g. 1/N is a very efficient portfolio); We only have small samples available, which makes estimates prone to large errors.
  • Empirical studies show that simple models routinely outperform complex ones: risk-based capital ratios (using Basel's III rules) have a lower explanatory power for bankruptcy of large banks than their vanilla leverage ratios. SImilarly, estimating volatility with MA (moving average) leads to less violation of VaR than more complex models such as GARCH (Generalized autoregressive conditional heteroscedasticity) or EWMA (exponentially weighted moving average).
  • Simpler rules and financial regulations would be more efficient, harder to game and less costly to implement.


"So what is the secret of the dog’s success? The answer, as in many other areas of complex decision making, is simple. Or rather, it is to keep it simple. Studies have shown that the Frisbee‑catching dog follows the simplest of rules of thumb: run at a speed so that the angle of gaze to the Frisbee remains roughly constant."
"So what is the secret of the watchdogs’ failure? The answer is simple. Or rather, it is complexity."
"When sample sizes are small, simpler models are unambiguously superior. With highly imperfect information, adding model complexity simply increases prediction errors."
"Complexity of models or portfolios generates robustness problems when understanding a complex financial system over plausible sample sizes. More than that, simplicity rather than complexity may be better capable of solving these robustness problems."