“…Many researchers use the posterior mean, see e.g. Kim and Nelson (1999), Elerian, Chib, and Shephard (2001), Kim and Piger (2002), Kim, Morley, and Nelson (2005), Pesaran, Pettenuzzo, and Timmermann (2006), Johnson and Sakoulis (2008), Liu and Maheu (2008), Maheu and Gordon (2008), Fruhwirth-Schnatter and Wagner (2008), Nakajima and Omori (2009), and few, such as Paroli and Spezia (2008), use the mode. To the best of our knowledge, no study has clearly documented the sensitivity of the marginal likelihood value obtained by Chibs' method to the value of the parameter chosen for its computation in the case of Markov-switching models.…”