When considering market based early warning signals, the million dollar question how often we get false positives (type I errors). Despite Paul Samuelson’s famous quip that the “stock market has predicted nine out of the last five recessions” (Samuelson 1966), research suggests that there is predictive value in significant price changes. Jeremy Siegel rebutted Samuelson with a study showing that “… 38 of the 41 measured recessions since 1802 have been preceded by and 8 % decline in the stock returns index. There have been twelve “false alarms” using this
criterion___ Despite these faulty signals there is a significant gain to stock investors
from being able to predict turning points in the business cycle over all time periods.” (Siegel 1991)
The broad persistence of momentum in stock markets globally (Fama and French 2012) is further evidence of social (as opposed to instant) diffusion of information. Didier Sornette and his Financial Crisis Observatory have a growing track record of bubble forecasting in various asset classes.11 However, further careful and extensive backtesting should be conducted to confirm that VaR outliers have predictive value. Such backtests might measure conditional returns after VaR outliers in different market regimes. These backtests should help provide insight about what proportion of price change can be attributed to random noise versus signal. Given that noise is likely to be a Gaussian distribution, we should expect that the market’s actual fat-tailed distribution to be least be partly attributed to the social diffusion of information. Note that according to RiskMetrics backtesting research, the Gaussian distribution fits markets well until about 95 % confidence (e. g., 1.65 standard deviations). After that, the accuracy of the Gaussian drops significantly, and a Student t distribution with 5 degrees of freedom is significantly more accurate in volatility forecasting from 1 day to 1 year (Zumbach 2007). 
From a practical perspective, the monitoring and discussion of outlier signals should be part of investment discipline. At a minimum, VaR outliers should trigger a formal discussion, as is the practice at active traders such as J. P. Morgan and Goldman Sachs. Given that outlier signals are often ambiguous, expertise and judgment is required to connect the dots. Therefore, outlier risk management should not be formulaic, but rather based on a discipline of rigorous social intelligence.