The second part of the research uses a Markov switching model with time-varying transition probabilities to capture the changes in inflation and their determining factors. This model was developed through the evolution of several previous studies and is considered to be relevant to the research at hand. The Markov Switching model used is the result of similar studies by Bleaney (1997) and Blix (1999) that used a switching VAR model to obtain time-varying probabilities of inflation processes. A suitable model was examined in Dropsy and Grand (2000) that asked a similar question to the one being explored in this research using a similar data set. The Markov switching model used by them sufficiently describes the data set being used in this research model.

The data for this research is derived from country data among 26 countries, both developing and industrialized. All of the countries examined have changed their exchange rate policy several times to one of those discussed previously. Floating policies are the most interesting as far as this research is concerned. The period being examined in this research is from 1982 to 1998, which is the period when many of the changes took place.

At some time during this period the countries chosen had a floating policy. This period was isolated and selected for the regression analysis. The period before the after the changes were closely examined. Several countries changed back to a pegged exchange rate after several years of floating rates. One such example is Argentina, which changed back to a pegged rate after runaway inflation sent the economy spiraling (U.S. BLS, 2001).

The Markov model used for the analysis accounts for these changes using shifts from high inflation to low inflation. This model was preferred because it allowed the shifts to move in either direction, therefore they could sufficiently explain the data set being examined in this study with little modification. The results of this model help to measure the effectiveness of inflation target policy using currency rates as a means of control. Countries were able to apply various exchange rate models in an attempt to control inflation. Therefore, exchange rate policies act as the independent variable in this study. Inflation is reactionary to exchange rates, therefore will serve as the dependent variable.

The results of the Markov Switching model used for this analysis serve as a measurement of the degree of effectiveness of inflation target policy (Dropsy and Grand, 2004). As reported by Dropsy and Grand, this model has certain limitations that may hinder its effectiveness as an assessment tool. For instance, the model uses fixed transition probabilities from one policy form to another over the entire period of time being considered. Using a fixed approach to exchange rate has an averaging effect over time. The market responds to shocks and these may be lost due to this averaging effect.

The monetary market is dynamic and constantly changing. However, incorporating each and every change over a period of time becomes cumbersome. In addition, such detailed analysis makes it difficult to sort out trends from single events. This has been reported as a limitation of the Markov switching model. However, for the purposes of this research it may be an advantage rather than a hindrance. If a spike is found, then that particular set of data can be re-examined to determine if the change fits the category being examined in this analysis. Although the model does not typically account for shocks, it can still be used to examine them more closely. The Markov switching model as used in Dropsy and Grand (2004) is an excellent tool for examining the data set used in this analysis.

Dropsy and Grand overcame the obstacles by using an estimation of the changes from high inflation policies to low inflation policies using their transition probabilities. This model developed by Dropsy and Grand is a result of the expansion of several earlier models. It is based on the Markov-switching model with fixed transition probabilities developed by Hamilton (1989). This model, as used in Dropsy and Grand (2004) was expanded by allowing the transition probabilities to change over time using fluctuations of an information variable (Filardo, 1994, 1998).

An examination of these works further explains how these changes help to account for shocks. The changes introduced by Filardo are the key to making the model an effective instrument in the current study. The model used as the basis for this regression was originally used to measure the effects of policy changes in Morocco and Tunisia (Dropsy and Grand, 2004). The...

The economic profile of Morocco and Tunisia make them excellent testing grounds for the Markov-switching model using a similar data set to that being used in this study. Dropsy and Grand received reliable results using this model and it can be expected that it will perform as effectively on data from the countries used in this research due to the similarities in data.
The Model

An explanation of the Markov-switching model can be found in Dropsy and Grand (2004). However, it is important to explain how the various components of the model relate to the current research model and the interpretation of the data. The following explains how the model was applied to the data set in this research.

The first portion the model assumes that inflation can be derived using the following equation:

It the equation used by Dropsy and Grand ? is the inflation rate. The equation states that the statistical mean (?) and the variance (?) of inflation depend on the state of the economy (ts). In this modes a high inflation period is indicated as ts=0 and a low one is represented by ts=1. According to the authors of the model a high inflation period is characterized by much more volatility than the low inflation one. This agrees with the information obtained in the theoretical development of the model. This means that inflation variance can vary with the state of the economy.

Hamilton (1989) defined the following transition probability matrix:

Hamilton defined the transition probabilities as follows:

According to Hamilton's model, the probability of being in either one of the two states depends solely on the state in period t-1. A key limitation of Hamilton's model, as noted in Dropsy and Grand (2004) is that this matrix makes the probability of being in a particular state constant over time. This is where Filardo (1994, 1998) improved on the model. He allows transition probabilities to change over time using an indicator variable zt as follows:

"In this revision, p is the probability of staying in a high inflation regime, q is the probability of remaining in a low inflation regime and Zt = {zt, zt-1,...} is the set of exogenous variables considered to predict the future course of inflation," (Dropsy and Grand, 2004).

Filardo's revision to Hamilton's model estimated the transition probabilities (p) and the parameters of the equation concurrently. If 0-0, 1, = = I i, then the transition probabilities are fixed as in Hamilton's model (Dropsy and Grand, 2004).

This is the revision that allows model to be used on the current data set. This revision allows us to consider the factors that influence the high inflation period. As one will recall, one of the key limitations of the regression study is that even though one event closely follows another it does not mean that causality can be assumed. This revision of Hamilton's model allows us to better determine causality and the effects of influences other than the dependent variable.

Empirical Results

Using data from Calvo and Reihart (2001) 25 period of floating monetary policy were found among the 26 countries considered for the study. Inflation data was obtained from a compilation of the International Monetary Fund (IMF, 2002).

The primary goal of the research was to determine if a relationship exists between exchange rates and inflation. In order to do this we had to isolate the type of exchange rate policy that is considered to be the least volatile and the most reflective of the country being examined.

Not all of the countries considered in the final analysis were found to have floating policies, or at least during the period in question. For instance, France, Greece, Germany, Egypt, Columbia and Chile have never used a floating exchange rate policy (Calvo and Reinhart, 2001). For those that did have a floating policy for some time they later switched to a managed or limited floating policy. A small number switched back to a pegged policy after a period of floating policy, but most adopted a managed floating policy after the floating policy period ended (Calvo and Reinhart, 2001).

Using the regression model, it was found that floating policies resulted in high periods of high inflation that began shortly after the adoption of the floating monetary policy. Both exchange rate volatility and inflation were found to increase after adoption of a floating exchange rate policy. Average inflation rates for the high periods…