Similar to the cited research, this work deals primarily with the question of monetary independence under fixed and flexible regimes. It also tries to account for individual countries’ ability to exercise an independent monetary policy while other countries may be unable to. Out of the original sample of 30 middle-income countries studied in the thesis, only 22 had short-run interest rate series that were suitable to carry out this analysis and only 19 countries showed sufficient variation of interest rates over long enough periods of time to allow a proper analysis to be
carried out.22 Exchange Rate classification was taken from the ‘natural’
periodisation of Reinhart and Rogoff (2002). Two broad categories were compared; fixed and flexible. The total number of country-regime episodes is 34; however, this encompasses a larger number of episodes since slight changes in intermediate regimes were lumped in one episode to lengthen the time series and increase the power of the test. Fixed regimes encompass regimes of hard and soft pegs and flexible regimes encompass regimes with intermediate flexibility, namely moving pegs, moving bands and managed floating and only one episode of freely floating (South Africa from 1999 to 2004). The episodes examined were constructed according to the change in the exchange rate regime in place and excluding any periods where a currency crisis occurred. Also freely falling (hyperinflation) episodes according to Reinhart and Rogoff’s classification were eliminated. This process of elimination ensures that the episodes in question correspond to periods of normality where monetary independence can be assessed with greater certainty. A complete list of episodes is provided in the appendix to this chapter. The data set was obtained from the IFS database and individual central bank websites where needed.
The main objective of this research is to assess monetary independence in middle income developing countries. In previous research, it has been assumed that a
22 The set of 22 countries where short-term interest data was available included Argentina, Bolivia,
Brazil, Chile, Colombia, Egypt, Guatemala, Lebanon, Malaysia, Mauritius, Mexico, Morocco, Paraguay, Philippines, South Africa, Sri Lanka, Thailand, Tunisia, Turkey, Uruguay and Venezuela. Three countries were eliminated: Tunisia, Turkey and Uruguay due to the lack of variation in the case of Tunisia and lack of sufficient observations after eliminating hyperinflation episodes in the case of Turkey and Uruguay. The appendix provides details on the interest rate series used for each country.
strong response to world interest rates signifies a lack of monetary independence. Here this understanding of monetary independence is modified to take into account the ability of a country to respond to domestic objectives and targets, namely inflation and the output gap, as well as responding to world interest rates (the US interest rate). In other words, this work accepts that in a world of interdependent financial markets it is impossible to eliminate the impact of world interest rates on domestic monetary policy, and tries to assess whether there is still scope for countries to respond to domestic variables under different exchange rate regimes despite increased global and regional integration. Given that developing countries are moving towards increasing rather than decreasing integration into the world economy, this understanding of monetary independence becomes more accurate. The domestic variables used to assess monetary independence are the inflation differential and the output gap differential. The inflation differential is calculated as the difference between domestic inflation and US inflation. The output gap differential was constructed using quarterly GDP data where available. The output gap was calculated for each country by regressing GDP on a linear and quadratic trend function and the output gap was obtained as the residuals of this regression. The output gap series was standardised by dividing the quarterly observations by the series’ own standard deviation. The monthly output gap was then extrapolated from the standardised quarterly series. The differential standardised output gap between the domestic economy and the US economy is the variable used in the analysis to assess the response of monetary policy to the real economy. The reason for this procedure was that output gaps in developing countries are typically larger in both directions compared with those in developed countries such as the US; standardisation makes it possible to capture the relationship between the two business cycles which could have been masked by the larger magnitude of variation in developing countries had the series not been scaled using the standard deviations. The analysis is conducted on individual country-exchange rate episodes using an Error Correction Model (ERM) and applying OLS to the following equation:
t t t us t j t us t j t c a r a r a r a i a y u r = + Δ + + + + + Δ 1 2 −1 3 −1 4 −1 5 −1 , where j t
r is domestic interest rate, us
t
r is US interest rate, i , y and u refer to the
US interest rate is used as a proxy for world interest rates because all the countries in the sample have close ties with the US and their exchange rates were linked to
the USD according to Reinhart and Rogoff’s classification.23 The coefficient a
2
provides the speed of adjustment to the long-run relation between domestic interest rates and the other variables. The long-run coefficients can be derived by dividing
ai by a2, where i=3, 4, and 5, to obtain the long run parameters for the
responsiveness of domestic monetary policy to foreign interest rates, inflation and output gap differentials respectively. Statistical inference was possible by running Wald coefficient tests on the long run parameters.
The above equation can be understood as integrating into the monetary independence literature the essential insights of the literature on Taylor rules (e.g. Taylor, 1993, 1999; Clarida, Gali and Gertler, 1998), in which the domestic interest rate is regarded as responding to domestic inflation and the domestic output gap together with, in the case of smaller and more open economies, the interest rate in the US or Germany.
The original theoretical formulation of Taylor rules focused on the response of monetary policy to domestic variables of inflation and output gap only. Later it discussed the possibility of formulating a central bank response function that also takes account of the exchange rate or a foreign interest rate in open economy models. Taylor (2001) explains that although a closed-economy monetary policy rule does not include a response to the exchange rate, it does in fact imply such response through the expected impact of an appreciation (depreciation) on both inflation and output. An appreciation in the current period may signal the need for easing monetary policy because it would lower inflation and output in the future. A forward-looking monetary policy rule will respond to the expected decline in output and possibly inflation in the future by cutting interest rates today. Empirical literature estimating Taylor rules for various countries also suggests that central banks are heavily influenced by world interest rates in formulating their reaction functions. Clarida, Gali and Gertler (1998) found that in their sample of European countries (UK, France and Italy), German monetary policy was found to be a
23 Morocco is the only country in the sample that had its exchange rate linked to the FF and the Euro
according to the Rienhart and Rogoff’s classification but when FF and/or Euro are used in the regression the coefficient was very small and statistically insignificant. The USD is used instead.
significant constraint on operating monetary policy in those countries even before Exchange Rate Mechanism (ERM) became a genuinely hard fixed exchange rate regime, as they maintained interest rates that are much higher than warranted by their domestic conditions. The paper estimated reaction functions for each of the three countries and found that adding the German interest rate to the estimated Taylor rule improved the specification considerably and generally halved the coefficients on inflation. The paper found that a one percentage point increase in the German interest rate induced an increase of 60, 59 and 114 basis points in interest
rate in the UK, France and Italy respectively. The paper concludes that “monetary
policy in these countries appears to have boiled down to fighting inflation by
following the Bundesbank” p. 1058.
More recently, Adam, Cobham and Girardin (2005) estimated reaction functions for the UK to examine the impact of institutional and monetary framework constraints on the conduct of monetary policy over the three periods 1985-1990, 1992-1997 and 1997-2003. They find that when adding the US and German interest rates, the domestic variables become insignificant. When estimating two version of the reaction function; domestic and international, the international model unambiguously dominated the domestic model in the pre-ERM period (1985-1990) and neither domestic inflation nor the output gap enter the estimated function significantly. Upon exit from the ERM (1992-1997), the domestic model became more recognizable and domestic variables became positive and significant; however, the estimated inflation coefficient was less than unity and the UK monetary policy was still strongly influenced by the German interest rate. It was only after 1997 that a well-defined domestic reaction function dominated the international one.
While the present work is not concerned with estimating Taylor rules for the central banks in the sample, linking the literature on monetary independence to the Taylor rule insights provides a more comprehensive and complete picture of the formulation of monetary policy in small open developing countries.
The regression results were complemented by measuring the correlation between domestic and US business cycles using the standardised output gap variable mentioned earlier. The correlation of the business cycles is of course related to the standardised output gap differential: as the degree of synchronisation between
business cycles increases, the magnitude of the output gap differential is expected to be smaller. This may also affect the size and the statistical significance of its coefficient in the regression.
The magnitude and sign of the correlation coefficient between the two output gaps provide an additional insight into the degree of monetary independence and the ability of a country in practice to exploit its flexible exchange rate to target domestic variables. For example, if there is a high degree of correlation between domestic and US business cycles, then there is little value in attempting to pursue a monetary policy that is at odds with the movement in the US interest rates. Thus, having an independent monetary policy when the business cycles are strongly related is of little value in practice. In that case, what appears in the regression analysis to be a strong impact of US interest rates on domestic interest rates may reflect the synchronisation of business cycles, rather than an absence of monetary independence.
Central bank independence was also considered explicitly when analysing the degree of monetary independence. The episodes in this sample were divided into three groups of high, medium and low CBI according to the information obtained about the individual countries from various sources (see detailed classification and methodology in Chapter 1). Two main sources were used to arrive at this classification of the cases: Mahadeva and Sterne (2000) reporting the results from the Bank of England survey of central banks in 94 industrialised and developing countries, Jacome (2001) which surveys legal and actual central bank independence in Latin America and the returned responses to a survey administered to the countries in this sample. On the basis of the numerical values assigned to the degree of CBI in these sources, a 3-teir classification of high, medium and low CBI was possible. For almost all the countries in this work, information about CBI was available in at least two surveys, and there were no conflicts between their assessments of CBI. This allowed for a clear cut classification of the episodes in this chapter.