1

A Preliminary Evaluation of Monetary Policy Rules for Russia

November 2003

Preliminary Draft. Please do not quote

Akram Esanov Christian Merkl

Lúcio Vinhas de Souza1

Abstract The paper reviews the recent conduct of monetary policy and the central bank’s rule-based behaviour in Russia. Using several policy rules, we test whether central banks in Russia react to changes in inflation, output gap and the exchange rate in a consistent and predictable manner. Our results indicate that during the period of 1993-2002 the Bank of Russia has used monetary aggregates as a main policy instrument in conducting monetary policy

Key words: monetary policy rules; exchange rate; central bank; Russia

JEL Classification: E52, F33, F41, E61

1Kiel Institute for World Economics (IfW). Christian Merkl is a student at the IfW’s ASP (Advanced Studies Program), Akram Esanov is both a Research Assistant at the IfW and a ASP student, and Lúcio Vinhas de Souza is an Economist at the IfW. Akram Esanov and Lúcio Vinhas de Souza were financed by the USAID/IRIS Project n° 220/001.0-03-337. Corresponding author: Lúcio Vinhas de Souza, Kiel Institute for World Economics (IfW), Düsternbrooker Weg 120, 24105 Kiel , Germany, phone +(49)(431) 8814-205, fax +(49)(431) 8814-202, email: desouza@ifw.uni-kiel.de. Websites: http://www.uni-kiel.de/ifw/staff/desouza.htm and http://www.tinbergen.nl/~phare/Partners/Souza.html.

2

1. Introduction

The effective development of a financial sector implies the need for institutions,

policy-setting, regulatory and supervisory ones. Arguably, the most important of such

institutions is the monetary authority, the central bank. Therefore, the development of the

Russian financial system, and its importance for long-term growth and stability, cannot

be seen separately from the development and effectiveness of its monetary authority,

from its capacity in setting and pursuing a monetary policy. Consequently, our aim in this

wok is to study the different tools and rules, and their effectiveness, that the Central Bank

of Russia has at its disposal.

The last ten years have witnessed an upsurge in research on monetary policy rules

evaluation motivated by the seminal paper of Taylor (1993). An evidence of this is the

enormous volume of recent papers and conferences on the topic. Taylor’s original paper

shows that the interest rate setting behavior of the Federal Reserve System in the US can

be characterized by a simple reaction function where short-run nominal interest rate used

as an instrument by the Fed changes in response to deviations in inflation and output gap.

Following this study, a great number of researchers have investigated the Fed’s

behavior using a simple Taylor rule or with some variations, like including lags of short-

term interest rate or output deviation. Overall, in case of the US or other developed

countries, the Taylor like rule explained very well central banks behavior indicating that

in developed countries most of the time central banks reacted in a way to stabilize

deviations either from a target level inflation or output gap.

3

However, in case of developing countries and emerging markets, let alone data-

scarce transitional economies, findings of monetary policy rule evaluations were

inconsistent and results often changed from study to study depending upon time span and

model specification. These conflicting findings are explained by several facts, such as

that in developing and emerging countries, financial markets are not mature, some of

these economies adopt fixed exchange rate regimes and that the objectives of the

monetary authority in those economies differ from that of the industrial countries.

Despite these discouraging observations, over the past few years a number of studies has

investigated monetary policy rules in emerging markets, finding that even with all

shortcomings, central banks in emerging markets follow a rule-like policy in conducting

monetary policy and, with appropriate modifications, a simple Taylor rule can describe

well the short-term behavior of interest rates. Because of model specification difficulties

and problems associated with collecting reliable data, very little research has been done

on monetary policy rule in transitional economies. This is one of the first attempts to fill

this existing gap. This paper examines the conduct of monetary policy in Russia during

the period of 1993-2002.

The Russian economy has experienced sharp swings since the collapse of the Soviet

Union in 1991. The start of transition from planned economy to a market economy

triggered a severe drop in output and the economy did not fully recover until 1997.

However, early signs of recovery were dashed by the August 1998 default and following

financial crises. Since 1999 the Russian economy has been experiencing a robust growth

due to high oil prices and trade surplus, generating a real exchange rate appreciation.

4

Against this background we would like to investigate Russian central bank’s

behavior over the past ten years. Particular attention will be given to the evaluation of

monetary policy rules for Russia. The rest of the paper is organized as follows. Section 2

briefly reviews the recent literature on monetary policy rule. Section 3 describes

evolution of monetary policy instruments and regime changes in the Russian central bank

in a chronological order. Section 4 specifies different empirical models to be used in

evaluating monetary policy rule while section 5 presents results of empirical estimation.

Finally, section 6 draws some conclusions.

2. Review of the Literature

Since the famous paper by Taylor (1993), much of the research on monetary policy

rule has focused on developed countries. The widely used simple instrument rule is called

a Taylor rule, where interest rate responds only to the inflation and output gaps according

to

* *0 1 2( ) ( ),t t t ti y yβ β π π β= + − + − (1)

where ti is the instrument rate in period t, 0β is a constant, *tπ π− is the “inflation

gap”, where tπ is the rate of inflation and * 0π ≥ is a given inflation target, *t ty y− is the

output gap, where ty is log of output and *ty is log of potential output, and the coefficients

1β and 2β are positive2.

2 More specifically, the condition that determines a unique non-explosive equilibrium, corresponding to the optimal commitment, is 1β >1 and 0< 2β <1 (see Woodford, 1999).

5

Since then, substantial research in this area has attempted to examine this simple

instrument rule and contributed many important insights. Particularly, researchers show

interest in evaluating the conduct of monetary policy in developed countries using a

simple Taylor rule. The general finding of this literature is that variants of the Taylor rule

perform quite well in explaining central bank behavior in industrial countries and

estimations are robust, in the sense that results do not substantially change across

different model settings. Furthermore, there is a widely accepted agreement among

researchers that no central bank adopts a simple rule as a formal policy rule in conducting

monetary policy although different modifications of the Taylor rule have been used as a

guideline in a decision-making process.

Application of a rule-like policy in emerging markets and transitional economies

raises the question as to usefulness of rule-based monetary policy, and in a greater extent,

about the effectiveness of monetary policy, given the specific market structure of

emerging economies. Again Taylor (2000) addresses this question and argues that policy

rules have many of the same advantages in emerging markets that they have in developed

economies. As to the choice of policy instrument in conducting monetary policy in

emerging economies, he states that given the specific nature of markets in emerging

economies, the policy instrument could be not only short-term interest rate but also the

monetary base or some other monetary aggregate. Furthermore, he stresses the

importance of exchange rates in monetary policy rule setting in emerging economies and

argues that inclusion of exchange rate in the central bank reaction function does not

contradict to the objective of central bank since in emerging economies sometimes

6

exchange rate stabilization is a precondition for output stabilization and in bringing down

inflation to a targeted level.

Regarding usefulness of monetary policy in emerging economies Taylor (2001)

argues that even though effect of monetary policy on real variables through the financial

markets is limited because of the less developed nature of these markets, still monetary

policy could have significant impacts through changes in wages and property prices. So,

a predictable behavior of central banks in emerging economies considerably improves the

transmission and effectiveness of monetary policy.

In line with this theoretical argument over the past few years the monetary policy

regime in emerging economies has shifted towards adopting a rule like policy. For

instance, Mohanty and Klau (2003) indicate that out of 13 leading emerging economies in

their study, only two had not adopted inflation targeting (IT), a related type of rule-based

policy3. Given the fact that inflation targeting leads to a more systematic response by the

central bank to inflation, the interest rate setting process in these economies has been

guided by a rule-like policy. The main conclusion of their study is that in emerging

economies central banks, most of the time, change short-term interest rate in response to

deviations in inflation and exchange rate movements. They also note that, although price

stabilization remains a main objective of central banks in emerging countries, other

objectives such as output stabilization, stability of the exchange rate and in few cases,

stability of asset prices and current account deficit, can be highlighted as a central bank’s

objective.

3The countries in their study are: Brazil, Chile, Czech Republic, Hungary, India, Korea, Mexico, Peru, the Philippines, Poland, South Africa, Taiwan and Thailand. Of this set, only India and Taiwan had not introduced a inflation targeting regime by late 2003.

7

3. Development of Monetary Policy Instruments in Russia

The dissolution of the USSR at the end of 1991 did not immediately lead to the

establishment of a truly Russian monetary authority (the Bank of Russia), capable of

conducting an independent and effective monetary policy4, as, until mid 1993, some of

the former republics of the Soviet Union still used the ruble, the Russian national

currency, and central banks of those republics conducted their own credit policy

simultaneously with the Bank of Russia. Only after 1993 the Bank of Russia started to

conduct its own independent monetary policy, although the scope of the policy was

limited by the need to finance a huge budget deficit, mainly caused by a dramatic decline

in output (see Figure 3). This loose monetary stance continued until the mid of 1995,

when the Russian economy started showing signs of stabilization and a new law on the

Bank of Russia was passed, providing some degree of legal independence to the Bank of

Russia in conducting monetary policy5.

These positive developments have allowed the Bank of Russia to adopt a tighter

monetary policy and to introduce a pegged exchange rate regime with a crawling band

against the US dollar from July 1995 onwards. As a result of these measures inflation

slowed down (see Figure 2). Furthermore, because of favorable developments in local

securities market, direct credit to the government significantly decreased and the Bank of

4The Central Bank of the Russian Federation (Bank of Russia) was founded on 13 July 1990, based on the Russian Republic Bank of the State Bank of the USSR. On 2 December 1990, the Supreme Soviet of the RSFSR passed the Law “On the Central Bank of the RSFSR (Bank of Russia)”, which declared the Bank of Russia a legal entity and the main bank of the Russian Federation 5Nevertheless, still today the Bank of Russia maintains some functions not traditionally seen as belonging to a central bank: for instance, in spite of being a baking supervisor and regulator, the CBR has a majority stake in the largest Russian bank (and state owned bank), Sberbank Rossii, who has 23% of all banking assets, 70% of household deposits, 20% of corporate deposits and 21,000 branches across Russia, and, until late 2002, also had participation in the second the second largest state owned bank, the VTB. Also, acting as an agent for the Ministry of Finance, it set up and manages the government securities market, known as the GKO market.

8

Russia started to conduct monetary policy through indirect instruments, such as interest

rates and reserve requirements. However, the start of Asian crisis of 1997 spread a

negative shock throughout emerging markets. This unfavorable external shock decreased

investment confidence in Russia and caused capital outflows, forcing the Bank of Russia

to defend the band. Although during the exchange market interventions in November

1997 the Bank of Russia lost over $6 billion of its liquid reserves, which was equal two

thirds of total reserves at that time, the exchange band was successfully defended for a

while.

Despite these efforts of the Bank of Russia, due to the severe August 1998 financial

crisis, the government was forced to default its domestic debt obligations. The ruble was

devaluated and the exchange rate band was abandoned leading to adoption of a floating

regime. One consequence of the sharp depreciation was a rapid acceleration in inflation.

Although ruble-denominated debt was restructured, investor confidence kept declining

because of an increase in political uncertainty and private capital outflows. Under such a

situation, the Bank of Russia, fulfilling its role as a lender of last resort, attempted to

preserve the payment system, which came to a halt during this period, by injecting

liquidity into banking system through a reduction of reserve requirements and extending

large amount of new credits. However, base money declined significantly in real terms

reflecting the sharp decline in output and increased use of non-monetary forms of

payment.

9

Figure 1. Real and Nominal Effective Exchange Rates (1995=100)

20

40

60

80

100

120

140

160

180

94 95 96 97 98 99 00 01 02

REER NEER

Figure 2. Refinancing Rate and CPI (Quarterly Based, in percent)

0

10

20

30

40

50

60

70

80

90

94 95 96 97 98 99 00 01 02

REFINANCINGQUARTERLYBAS QUARTERLYCPI

10

Figure 3. GDP Index and M1 Index (1993=1)

0 . 3

0 . 4

0 . 5

0 . 6

0 . 7

0 . 8

0 . 9

1 . 0

1 . 1

9 3 9 4 9 5 9 6 9 7 9 8 9 9 0 0 0 1 0 2

G D P I N D E X M 1 I N D E X

From 1999 onwards, one of the main objectives of the Bank of Russia was to bring

inflation down to 30 %, while keeping output decline in the range of 1-3 percent, within

the framework of a dirty float of the Ruble. To achieve this objective, monetary policy

was tightened by reducing net credit to banking system. Because of this measure,

inflation fell sharply and the exchange rate depreciation slowed. Furthermore, fiscal

performance significantly improved due to the approval of a new package of fiscal

measures and improvements in revenue collection. The external environment

strengthened as world energy prices increased, generating trade surplus and capital

inflows, causing the exchange rate to become one of the main targets of monetary policy

(see Figure 1).

According to the Bank of Russia the main objective of the monetary policy in 2000

was to reduce inflation to 18 % and to achieve an annual growth rate of GDP 1.5 percent.

However, the continuing strength of the balance of payments and the Bank of Russia’s

11

reluctance to permit a real appreciation of the ruble has placed increasing pressure on

monetary policy. Given this continued favorable economic situation in recent years, the

Bank of Russia has placed more weight on the exchange rate stability, while accepting

the inflationary consequences of such a decision. This policy of the Bank of Russia has

slowed the real appreciation of the ruble and reduced inflation, even though the pace of

disinflation has been slower than the one formally targeted by the authorities6.

4. Specification of the Empirical Model

The recent literature on monetary policy rule primarily distinguishes two types of

instrument rules: interest rate based instrument rule and monetary based instrument rule,

referred to as the Taylor rule and the McCallum (see McCallum, 1988) rule,

respectively7. The key difference in these rules involves the choice of an instrument in

central bank’s reaction function in response to changes in macroeconomic conditions.

The Taylor rule, as specified in section 2, uses a short-term nominal interest rate as an

instrument and is widely used in monetary policy estimations because of its simplicity

and realistic nature. The McCallum rule uses the growth rate of monetary base as an

instrument and can be expressed as follows:

)(5.0 1**

−∆−∆+∆−∆=∆ ttt xxvxb , (2)

where tb∆ is rate of growth of the monetary base, percent per year, *x∆ is the target rate

of growth of nominal GDP, percent per year, tv∆ rate of growth of base velocity, percent

6 The policy relevance of such concerns with real appreciation are somewhat doubtful, as is unclear that the Russian Ruble in above its long run equilibrium value, or merely recovering from an undershooting (see IMF, 2003). 7 Razzak (2001) shows that the McCallum and Taylor rules are co-integrated.

12

per year, averaged over the previous four years, in the original McCallum estimation,,

and x∆ is rate of growth of nominal GDP, percent per year. In this rule the target value of

nominal GDP growth is calculated as the sum of the target inflation rate and the long-run

average rate of growth of real GDP.

Originally the both rules were designed to be used in evaluation of the monetary

policy in large industrial countries, and many observers expressed concerns regarding the

effectiveness of this basic policy rules for evaluating the conduct of monetary policy in

emerging economies. This concern raises the question as to what kind of modification are

needed to fit better the realities of emerging economies with underdeveloped financial

markets, dependence on primary commodity exports, sharp swings in productivity and

relative prices, and high exposure to volatile capital flows.

To address adequately this question researchers use modified versions of these

instrument rules. One general consensus in this regard is that monetary policymakers in

emerging economies are more concerned about exchange rate movements than those in

mature economies, among other reasons due to the degree of exchange rate pass-through

to prices, and this feature has been incorporated into central bank’s reaction function8.

Some researchers, such as Ball (1998) even suggest that, in an open economy, the central

bank could use weighted average of nominal interest rate and exchange rate as an

instrument. However, this approach has not been popular among empirical researchers

because of uncertainties involved in determining weights.

Since 1991 the Russian economy has experienced both sharp fluctuations in main

macroeconomic variables and deep structural changes. Given this unstable nature of

8 Detken and Gaspar (2003), show that a monetary authority that cares about price deviations will also care about exchange rate developments, even without formally targeting those.

13

economic environment in Russia, the task of estimating a monetary policy rule is

complicated and one may a priori assume that no single policy rule would fully capture

all aspects of central bank behavior during this period. However, close observation of

monetary policy decision-making behavior of the Bank of Russia indicates that it is worth

to analyze monetary policy in Russia in the context of a rule based reaction function.

We start our analysis with the estimation of a modified version of the famous

Taylor rule. In his seminal work, Taylor (1993) proposed the following, now well-known,

policy rule to describe the Fed’s behavior in setting the short term interest rates:

0.5 0.5( 2) 2i yπ π= + + − + , (3)

where i is the short term interest rate, p is the inflation over the four previous quarters, y

is the percent deviation of real GDP from a target. The inflation target and the

equilibrium real interest rate are set at 2 and assumed as constant over time. The “policy

maker” is here assumed to care equally about deviations of inflation and output from

target. In this formulation Taylor used a linear trend to approximate potential output.

This simple equation cannot be estimated in the original form in the case of Russia,

since a relatively stable long-run average inflation does not exist. The only way to

estimate equation 3 is to assume that there is a constant intercept and estimate the

coefficients by running a simple regression without specifying the other parameters of the

model. Our initial results, where the output gap was calculated by using a linear de-

trending approach generated unsatisfactory results, so we recalculated the output gap

making use of a different approach, namely the traditional Hodrick-Prescott (HP) filter9.

9The HP filter is the usual method to “smooth-out” the estimate of the long-term trend component of a series. It is a two-sided linear filter that computes a smoothed series, denoted by os, of a original series,

14

In a later work, Taylor (2001) suggested that market conditions in emerging

economies may require modifications of the original policy rule, which had been

designed to describe monetary policy behavior in developed countries. In this regard, one

possible modification of the original Taylor rule could be inclusion of one measure of the

exchange rate into the equation, since for many emerging economies exchange rate

fluctuations are of a great concern, because, as indicated above, of their vulnerability to

either external shocks, high exchange rate pass-through, generating the “fear of floating”

phenomenon (Calvo and Reinhard 2000). Given these considerations, we estimate the

following modified open economy Taylor rule:

0 1 2 3 4 1 5 1t t t t t ti y xr xr i uβ β π β β β β− −= + + + + + + , (4)

where txr is the growth of the real effective exchange rate, tu is a white noise error term

and t-1 indicates the past values of the variables. The remaining variables are the same as

in the equation 3. The expected signs of the parameters are as follows: 0β , 2β , 5β are

positive (but 5β #1), 2 5/(1 ) 1β β− > , 3 0β < , and 4β is undefined (and β1,>1).

Money based rules

As discussed in section 2, the refinancing rate (short-term interest rate) has not been

the most important instrument in conducting monetary policy in Russia due to several

factors such as underdeveloped nature of capital markets and limited role of the banking

denoted by yhp below, by minimizing the variance of y around os, subject to a “penalty parameter” that constrains the second difference of os. That is, the HP filter chooses ost as to minimize

( ) ( ) ( )[ ]∑ −−−∑ −−

==−++

1

2

2

1

2

11,T

t

T

tosososososy tttttthp λ

The penalty parameter λ controls the smoothness of the series. The larger the λ used, the smoother the series ost will be. The HP filter has known limitations on short samples and also on its estimated end sample.

15

sector in supply of funds. Thus, the Bank of Russia has been placing an equal importance

on other instruments such as non-market based interventions (specially at the beginning

of the transition period), exchange rate interventions or the change of minimum reserve

requirements in conducting monetary policy.

Furthermore, before the 1990s, monetary aggregates have often been referred in

policy analysis as an important monetary policy instrument10, as the uncertainty in

measuring real interest rates and big shocks to investment or net exports would make

monetary aggregates a preferred instrument. This may still be the case in Russia, and

especially at the beginning of the nineties (See Taylor, 2000).

Despite the problems associated with the direct control of the M1 monetary

aggregate and significant fluctuations in monetary velocity, we use this aggregate as a

policy instrument in conducting monetary policy in Russia. We are also aware of the fact

that some existing studies attempt to explain inflation dynamics by the growth of

monetary aggregates (see e.g. Korhonen and Pesonnen, 1998, Dabrowski et al. 2003)

using those as an explanatory variable. However, our Granger causality tests indicate that

at least in the short-run there (up to seven months) is only a Granger causality from prices

to M1 and not the other way around.

It is widely accepted that the time series data usually suffer from some level of

autocorrelation, and if this is not corrected the estimation results cannot be treated as

reliable. To correct for the autocorrelation problems, we will use differences rather than

levels and add several lags, according to information criteria and statistical significances

of the coefficients.

10Perhaps the most traditional of those “monetary targeters” was the German Central Bank, the Deutsche Bundesbank. Nevertheless, several works (see, for instance, Clarida and Gerter, 1996) put into question the exclusive –or even main- reliance of the Bundesbank on monetary aggregates even during its “golden age”.

16

A Hybrid Rule

Ball (1998) argues that interest rate based Taylor rules are inefficient unless they

are modified. He stresses that monetary policy affects the economy through exchange

rate as well as interest rate channels. Ball sets up a simple model with an open economy

IS curve, Phillips curve and a link between interest and exchange rate. Rearranging terms

yields the following optimal policy rule:

1(1 ) ( )t t t t twi w xr y xrα β π δ −+ − = + + , (5)

where w is a weight that depends on the calibration of the model and δ is the effect of a

one percent exchange rate appreciation on inflation, α and β also depend on

calibrations of the model. The calibration parameters we use will be based on the work of

Ball (1998). We use different weights and check their effect on the estimated coefficients.

Finally, to address the econometric problem caused by several possible structural

breaks in the Russian economy during the period 1993-2002, we use different

combinations of dummy variables.

17

5. Empirical results

Data and Methodology

Data for Russia have to be cautiously treated. The availability is limited and

phenomena such as dollarization and the barter economy may lead to a somewhat biased

picture. Some authors (see e.g. Falcetti et al. (2000)) also believe that the decline in

output was overestimated during the first years of the transition period11. In our empirical

estimations we use monthly data covering the time span 1993-2002. The period has been

chosen for data availability reasons. Alternatively, in several occasions we use quarterly

data to check the robustness of our results. The sources of the data are the International

Monetary Fund’s International Financial Statistics database, the web site of the Bank of

Russia, the monthly database of the Vienna Institute for International Economic Studies

(WIIW), and the Russian European Centre for Economic Policy (RECEP). For our

purpose we need the data on the short-term interest rate (refinancing rate), the consumer

price inflation, the monetary aggregate M1, the output gap, different exchange rate

measures (dollar exchange rate, nominal effective exchange rate, and real effective

exchange rate), the labor share (for an alternative proxy of the output gap), and the

budget deficit. We use output numbers from RECEP and WIIW (industrial production

numbers) and deflated them by monthly CPI because of the lack of a monthly GDP

deflator.

11For instance, Åslund (2001), estimates that, for an official figure of just 60.2% of the Russian 1989 GDP in 1995, the actual figure, after taking into account, among other things, illegal and under-reported activities, was an amazing 94%, or, in other terms, a mere marginal GDP loss.

18

Results for the Taylor Rule

The estimation results of the Taylor rule suggest that the Bank of Russia reacts to

each point increase of yearly inflation rate with a mere 0.3 point increase of the

refinancing rate. The estimated coefficient is significant at the 1 percent significance

level. However, the estimated coefficient of the output gap shows a wrong sign (negative)

and is not significant (for results see Table 1 in the Annex). When we use monthly or

quarterly data instead of annual inflation, the estimated coefficient rises more than

proportionally, which may indicate a very strong reaction to short-term inflation

pressures. The data fits better the model when we use year-to-year inflation numbers. The

Granger causality tests indicate that CPI Granger-causes the refinancing rate if we have

less than six lags. With more than six lags the causality relationship goes into the other

direction.

When we deflate the refinancing rate by the monthly CPI data and add dummies for

the three months before the financial crisis in 1998 (May, June, July) and one month

afterwards, we obtain the theoretically expected results (see table 2 in the Annex).

Almost all of the estimated coefficients show the expected signs and are significant at a 1

percent significance level. The only exception is the intercept term. However, these

highly significant results suffer from a strong first-order and higher-order-autocorrelation

problem.

When we estimate the open economy version of the Taylor rule , the estimated

coefficient of inflation shows the expected sign across the different equation

specifications. The estimated coefficient of the output gap shows neither the right sign,

nor is significant (for results see Table 2 in the Annex). Other proxies of the output gap,

19

such as the real unit labor cost suggested by Gali and Gertler (1999), also show

unsatisfactory results. The estimated coefficients of the real effective exchange rate and

the nominal effective exchange rate are only significant under certain model

specifications and most of the times have the wrong sign. However, when we use the

nominal dollar-ruble exchange rate, the estimated coefficient shows the right sign.

The estimated coefficient of the lagged interest rate is equal to 0.9 (very close,

therefore, to the value necessary for a non-explosive equilibrium), and this value remains

stable over the different model specifications, indicating that the interest rate in a new

period is about 90% of the old interest rate, plus the effect of the other independent

variables.

Overall, the estimation results suggest that a simple Taylor rule and its

modifications do not describe well the interest rate setting behavior of the Bank of

Russia. The results are not robust, in the sense that they depend on the model

specification and suffer from some statistical problems such as autocorrelation.

Results of Money Based Models

Because of data availability problem for the M1 series, some missing points have

been recovered by using the M2 series, since those two series are highly correlated. We

deflated the approximated M1 series with the monthly consumer price index. We expect

that signs of the estimated coefficients will be reversed because a decrease in the M1

means monetary contraction and a decrease in the interest rate a monetary expansion.

As the regression results indicate, in general the simple money based policy rule or

McCallum rule, performs much better in explaining the behavior of the Bank of Russia

compared with a simple interest rate based rule or the simple Taylor rule (see Tables 1

20

and 2 in the Annex). The estimated coefficients show the right signs, but both the output

gap and budget deficit measures are insignificant, even though these results no longer

suffer from a first-order autocorrelation problem.

When we replace the refinancing rate in our previous open economy rule by the

monetary aggregate M1, the results shows the expected signs, and the coefficients are

significant. However, the M1 series is non-stationary and this casts some doubt to the

robustness of the results. When we correct this problem by differencing, the results

mostly remain unchanged, except for the output gap (and also when replacing it by

nominal or real GDP). These results allow us to conclude that the Bank of Russia has

been targeting monetary aggregates in its policy decisions. At times of high inflation

pressure, the Bank of Russia responded by limiting the M1 growth, while at times of

exchange rate appreciation the policy response was an expansionary monetary policy.

Moreover, these results are not sensitive to the model specification and there are no major

statistical problems.

Given the absence of explicit inflation targeting in Russia we estimate in equation 6

a gap model as defined in Mohanty and Klau (2003). The advantage of this model is that

it allows us to use an HP measure of trend inflation instead of a targeted level, and I

given by

)6()1(log()()()())1(log(

15

1143210

tt

tttt

uMdxrtrendxrxrtrendxryCPItrendCPIMD

+++−+−++−+=

−

−−

ββββββ

where CPItrend is the Hodrick-Prescott (HP) filter of the inflation rate, xrtrend is a log

of the HP-filter of the exchange rate change. We add another lag of CPI to control for

autocorrelation problems. In addition, we include seasonal dummies for December and

January and a dummy for the period before 1995.

21

The regression results indicate that the Bank of Russia reacted to above trend

inflation with a contraction in M1. If the dollar exchange rate in the current period is

higher than the HP-trend, the Bank of Russia also responded with a reduction in M1. All

estimated coefficients are significant and exhibit the right signs. However, if we

substitute monthly inflation data with quarterly inflation data, the results are no longer

significant, indicating that the Bank of Russia is more concerned with short-term inflation

pressures.

Estimation Results for the Ball Rule

The estimation results for the open economy Ball model are mixed and unstable

(see Table 3, Annex).

Testing responses during different time periods

The Russian economy has experienced different shocks during different time

periods. It would be interesting to see whether the Bank of Russia has responded

differently during different time periods. Since the money based model performs better,

we will test for different policies during different time periods in this model12. First of all,

we separate the period before and after 1995, as Chow breakpoint tests indicate a

structural break at this time. We use for this purpose the equation of the following type:

)7(***)1(log(

*****inf**inf*))1(log(

109817

1654

3210

tt

ttt

ttt

udummydecdummyjandummydecMdangeratedollarexchangeratedollarexchdummyangeratedollarexch

ydummyM

+++++++

++++=∆

−

−

βββββββ

ββββ

12Estimating the interest rate rule with a similarly varying timeframe do not substantially change the previous results.

22

where dummy is a dummy variable that is 1 for the period before 1995 and zero

otherwise.

The estimation results clearly suggest that the Bank of Russia conducted a different

monetary policy before and after 1995. The estimated coefficients indicate that before

1995 the Bank of Russia was more concerned with reducing inflation13, while after 1995

priorities have shifted towards exchange rate stabilization. These findings are consistent

with the official announcements of the Bank of Russia. The results also remain

unchanged when we used the quarterly data, instead of monthly.

We obtain a similar result when we use a dummy variable for the crawling peg

period, from October 1994 through August 1998. As one would expect, the commitment

to react to exchange rate changes was greater during that period. During the high inflation

period, the Bank of Russia attached a greater priority to inflation, while at times of

relatively low inflation the main concern was exchange rate stabilization.

13Of course, average inflation pre-1995 was also substantially greater than post-1995.

23

6. Concluding Remarks

This paper examined the conduct of monetary policy in Russia during the period of

1993-2002. We estimated three sets of monetary policy rules, the Taylor rule, the

McCallum rule and the Ball rule, using the monthly and quarterly data. The regression

results indicate that a simple Taylor rule and its different variations, where the short-term

interest rate is used a policy instrument, describe poorly the interest rate setting behavior

of the Bank of Russia. The results are unstable, changing with the model specification

and the time span covered. In addition, the regular statistical tests indicate the existence

of some econometric problems, further complicating the interpretation of the results.

The McCallum rule, where the policy instrument is a monetary aggregate, fits

better the data14. This is in sharp contrast with the recent experience of other advanced

emerging markets, were interest rate rules produce a very good description of the policy

setting behavior of the monetary authority (see, for instance, Mohanty and Klau (2003),

Torreas Garcia (2003), Minella at al. (2003)). The estimated coefficients are significant

and remain unchanged across different equation specifications. The results indicate that

during the period of 1993-2002, the Bank of Russia has used monetary aggregates as the

main policy instrument in conducting monetary policy. Furthermore, the results presented

here also suggest that before 1995, the Bank of Russia was more concerned with inflation

reduction, while after 1995 the primary objective was exchange rate stabilization. The

estimation results of the hybrid rule or Ball rule, where weighted average of the interest

14One must remember that, even today, the CBR main actual instrument of monetary policy are deposit auctions.

24

rate and the exchange rate is used as a policy instrument, draw a mixed picture.

Depending on the choice of the weights, results change and most of the time the

estimated coefficients are insignificant.

The results on our estimations, of course, are backward looking, in the sense that

the represent the relationships that existed so far in the data. As the experience of other

advanced emerging markets show, the promotion of forward looking behavior among

Russian economic agents, aided by the development of stronger institutions -especially

by the strengthening of the credibility of the Bank of Russia, plus the deepening

of Russia's financial markets, shall, in time, enable the implementation of a successful

interest-rate policy rule, coupled with inflation targeting and a floating exchange rate

regime.

25

References:

Åslund, Anders. The Myth of Output Collapse after Communism. Carnegie Endowment

Working Papers Series, n° 18, March 2001.

Ball, Laurence (1998), Policy Rules for Open Economies, NBER Working Paper No.

6760.

Calvo, Guilermo and Reinhart, Carmen (2000), Fear of floating, NBER Working

Paper No. 7993.

Clarida, Richard and Gertler, Mark (1996). How the Bundesbank Conducts Monetary

Policy. NBER Working Papers No, 5581.

Dabrowski, Marek, Paczynski, Wojciech, and Rawdanowicz, Lukasz (2002), Inflation

and Monetary Policy in Russia: Transition Experience and Future Recommendations,

CASE Studies and Analyses.

Detken,Carsten and Gaspar, Vítor (2003). Maintaining price stability under free-

floating: a fearless way out of the corner?, ECB Working Papers series No. 241.

Falcetti, Elisabetta, Raiser, Martin, and Sanfey, Peter (2000), Defying the odds: initial

conditions, reforms and growth in the first decade of transition, EBRD Working Paper

No. 55.

Gali, Jordi, and Gertler, Mark (1999), Inflation Dynamics: A Structural Econometric

Analysis, Journal of Monetary Economicvs, 44(2), October 1999, 195 – 222.

IMF (2003). Russian Federation: Selected Issues. Country Report No. 03/146.

McCallum, Bennett (1988), Robustness properties of a rule for monetary policy,

Carnegie-Rochester Conference Series on Public Society 29 (1988), 173 – 204, North-

Holland.

26

Minella, André, Springer de Freitas, Paulo, Goldfajn, Ilan Kfoury Muinhos,

Marcelo (2003). Inflation Targeting in Brazil: Constructing Credibility under Exchange

Rate Volatility. Working Paper presented at the Kiel Workshop for Monetary Policy and

Macroeconomic Stabilization in Latin America, Kiel Institute for World Economics, 11 –

12 September 2003.

Mohanty, Madhusudan, and Klau, Marc (2003), Monetary policy rules in emerging

market economies, Issues and Evidence. Working Paper presented at the Kiel Workshop

for Monetary Policy and Macroeconomic Stabilization in Latin America, Kiel Institute

for World Economics, 11 – 12 September 2003.

Pesonen, Hanna, and Iikka, Korhonen (1998), The Short and Variable Lags of Russian

Monetary Policy, Review of Economies in Transition, 1998, No.4, Bank of Finland.

Razzak, W. (2001).Is the Taylor rule really different from the McCallum rule? Reserve

Bank of New Zealand Discussion Paper 2001/07

Taylor, John (1993), Discretion versus policy rules in practice, Carnegie-Rochester

Conference on Public Policy 39, 195 – 214, North Holland.

Taylor, John (2000), Using Monetary Policy Rules in Emerging Market Economies,

Revised paper presented at the 75th anniversary conference at the Banco de Mexico.

Taylor, John (2001), The role of the exchange rate in monetary-policy rules, American

Economic Review Papers and Proceedings 91, 263 – 267.

Torres Garcia, Alberto (2003). Monetary policy and interest rates: evidence from

Mexico. Working Paper presented at the Kiel Workshop for Monetary Policy and

Macroeconomic Stabilization in Latin America, Kiel Institute for World Economics, 11 –

12 September 2003.

Woodford, Michael (1999). Pitfalls of Forward-Looking Monetary Policy. Mimeo,

Stanford University.

27

Appendix 1

Table 1: Dependent Variable is Refinancing Rate, Monthly Data 1993:01-2002:12

Independent variable

Traditional Taylor rule (with yearly inflation)

Traditional Taylor rule (with monthly inflation)

Open economy rule (with yearly inflation)

Open economy rule (with monthly inflation)

Intercept 45.65 (4.64) ***

57.37 (6.10) ***

3.00 (2.61)

1.60 (2.55)

Year-to-year consumer price inflation

0.31 (0.03) ***

0.04 (0.02) *

Monthly consumer price inflation

4.95 (0.68) ***

1.31 (0.48) ***

Output gap -0.40 (1.07)

-370.60 (133.60) ***

0.32 (0.57)

0.05 (0.50)

Growth in real effective exchange rate

-29.96 (35.33)

54.92 (42.37)

Growth in real effective exchange rate (-1)

-5.95 (31.73)

-7.42 (31.08)

Interest rate (-1) 0.90 (0.04)***

0.90 (0.03) ***

R squared 0.62 0.34 0.94 0.94 Adjusted R squared 0.61 0.33 0.94 0.94 Durbin Watson statistics

0.17 0.34 2.39 2.52

Breusch-Godfrey test*

*** *** No rejection No rejection

Notes: -The Breusch-Godfrey serial correlation LM-test (with no autocorrelation as a null hypothesis) was conducted for twelve lags. (-1) indicates a first lag. -Standard errors are in parentheses. The asterisks indicate levels of significance a 10 (*), 5(**) or 1 (***) percent level.

28

Table 2: Dependent Variable is Growth Rate of M1, Monthly Data 1993:01-2002:12

Independent variable Difference model

Gap model* Full Model

Intercept 0.0183 (0.0044) ***

0.0109 (0.0047) **

0.0123 (0.0045) ***

Quarter-to-quarter-inflation

-0.0027 (0.0006) ***

-0.0026 (0.0007) ***

Quarter-to-quarter-inflation (-1)

0.0016 (0.0006) ***

0.0012 (0.0006) *

Monthly inflation -0.0017 (0.0015)

Dummy (for period before 1995) * monthly inflation

-0.0050 (0.0020) **

Monthly inflation (-1) -0.0013 (0.0007) *

Output gap -0.0016 (0.0011)

-0.0012 (0.0009)

0.0010 (0.0009)

Growth in real effective exchange rate

0.4244 (0.0719) ***

Growth in real effective exchange rate (-1)

-0.2330 (0.0664) ***

Growth in bilateral dollar exchange rate

-0.2567 (0.0536) ***

-0.2920 (0.0846) ***

Dummy for period before 1995 growth in bilateral dollar exchange rate /1

0.2701 (0.1262) **

Growth in bilateral dollar exchange rate (-1)

0.2917 (0.0541) ***

0.1107 (0.0480) **

Growth rate of M1(-1) 0.2211 (0.0685) ***

0.1716 (0.0697) **

0.2821 (0.0666) ***

Seasonal dummy for January

-0.1239 (0.0129) ***

-0.1114 (0.0137) ***

-0.1294 (0.0130) ***

Seasonal dummy for December

0.0814 (0.0106) ***

0.0843 (0.0112) ***

0.0885 (0.0107) ***

Slope dummy for August 1998

-0.0757 (0.0331) **

Dummy for the period before May 1998

-0.0156 (0.0065) **

Dummy for period before 1995

0.0431 (0.0231) *

R squared 0.78 0.74 0.76 Adjusted R squared 0.76 0.72 0.74 Durbin Watson statistics 1.97 1.67 1.97 Breusch-Godfrey test No rejection No rejection No rejection Note: /1 In this case we deduct the HP-trend from quarter-to-quarter inflation and the growth in the dollar exchange rate. Standard errors are in parentheses. The asterisks indicate levels of significance a 10 (*), 5(**) or 1 (***) percent level.

29

Table 3: Dependent Variable is a Weighted Average of Interest Rate and Real Effective Exchange Rate, Monthly Data 1993:01-2002:12

Independent variable Exchange rate weight=1 Exchange rate weight=0.5 Intercept -5.3370

(3.0818) * 29.5188 (18.7053)

Output gap 0.0143 (0.1569)

-1.2445 (1.1052)

Output gap (-1) 0.3811 (0.1620) **

-0.2921 (1.3301)

Output gap (-2) 0.3619 (0.1630) **

-1.0559 (1.1005)

Monthly interest rate + 0.5 * (real effective exchange rate (-1))

2.3246 (0.2588) ***

0.8054 (0.4665) *

Monthly interest rate (-1) + 0.5 * (real effective exchange rate (-2))

-0.2791 (0.2624)

0.2628 (0.4603)

R squared 0.97 0.20 Adjusted R squared 0.96 0.16 Durbin Watson test statistics 1.87 0.16 Breusch-Godfrey test No rejection possible *** Notes: - The Breusch-Godfrey serial correlation LM-test (with no autocorrelation as a null hypothesis) was conducted for twelve lags. - (-1) indicates a first lag. - Standard errors are in parentheses. The asterisks indicate levels of significance at 10 (*), 5(**) or 1 (***) percent level.