Identifying Speculative Bubbles with an Infinite Hidden Markov Model
Identifying Speculative Bubbles with an Infinite Hidden Markov Model
Identifying Speculative Bubbles with an Infinite Hidden Markov Model
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
(money growth) is consistent <strong>with</strong> the explosive dynamics of the exch<strong>an</strong>ge rate <strong>an</strong>d consumerprice. However, we also find bubbles in the exch<strong>an</strong>ge rate emerge in April 1987, October 1987<strong>an</strong>d September 1988, which have some locally explosive dynamics that c<strong>an</strong> not be explainedby the money base. This is further supported by the posterior me<strong>an</strong> of β st (i.e. E(β st |Y ))displayed in Figure 3.0.0 0.2 0.4 0.6 0.8money baseP(βt>0 | Data,IHMM)P(βt>0 | Data,MS2)0.2 0.4 0.6 0.80.0 0.2 0.4 0.6 0.8 1.0 1.20.0 0.2 0.4 0.6 0.8 1.0exch<strong>an</strong>ge rateP(βt>0 | Data,IHMM)P(βt>0 | Data,MS2)priceP(βt>0 | Data,IHMM)P(βt>0 | Data,MS2)0.2 0.4 0.6 0.8 1.0 0.5 0.6 0.7 0.8 0.91984−01 1985−01 1986−01 1987−01 1988−01 1989−01Figure 2: The growth rate <strong>an</strong>d the posterior probabilities of β stexch<strong>an</strong>ge rate <strong>an</strong>d consumer price> 0 for the money base,The posterior probability <strong>an</strong>d posterior me<strong>an</strong> of the money base from the iHMM (dashedline) are identical to those from the MS2 model. Therefore, it is reasonable to have a two-regimespecification for the money base. On the other h<strong>an</strong>d, the iHMM shows distinct probability <strong>an</strong>dme<strong>an</strong> patterns for the exch<strong>an</strong>ge rate <strong>an</strong>d consumer price.For the exch<strong>an</strong>ge rate, the posterior probability of β st > 0 implied by the iHMM in Figure 2is almost a mirror image of that by the MS2 model except in 1989. Any plunge of P (β st > 0 | Y )of the iHMM is associated <strong>with</strong> a decrease of the exch<strong>an</strong>ge rate, which is consistent <strong>with</strong> Ev<strong>an</strong>s’s(1991) bubble collapse regime. Ironically, for the MS2 model, 4 out of 5 spikes in Figure 2correspond to decreasing of the exch<strong>an</strong>ge rate, which is counter intuitive. This is caused by the15