“…Bliss and Panigirtzoglou (2002) and Bondarenko (2003) compare several competing procedures and conclude that nonparametric methods base on either the smoothed (spline) implied volatility smile and the positive convolution approximation seem to dominate the two-lognormal approach and other parametric techniques when estimating RNDs. Moreover, Anagnou, Bedendo, Hodges and Tompkins (2005) for the UK option market, Craig, Glatzer, Keller and Scheicher (2003) for the German stock option data, Bliss and Panigirtzoglou (2004) for the US and the UK option data, and Alonso, Blanco and Rubio (2005) for the Spanish option prices conclude that the RND is not an unbiased estimator of actual probability density function. This may not be surprising given the risk-neutrality embedded in these estimates.…”