1

Exchange rate shocks and monetary policy in Russia

Brigitte Granville and Sushanta Mallick Centre for Globalisation Research

School of Business and Management Queen Mary, University of London

Mile End Road, London E1 4NS, UK

Abstract

The failure of the Russian central bank to control inflation is striking. Russian

monetary policy has failed persistently to achieve sustained low inflation, both in

absolute terms and relative to the peer group of countries similarly exiting from

Soviet-style central planning. The question raised in this paper is why? What kind of

monetary policy has been followed by the central bank during the period 1995 to

2007? Our contribution is to search for a possible transmission channel between the

real interest rate, inflation rate, exchange rate, output growth and foreign reserve

growth, after having controlled for the effect of oil price inflation. Using a vector

autoregressive model in error-correction form and using sign restrictions, we show

that the monetary authorities’ failure to abate double-digit inflation appears to be

driven by the policy of exchange rate targeting, as reflected in our identified exchange

rate shocks.

I. Introduction

In recent years, for most central banks around the world, monetary policy – whether

expressed in terms of interest rates or growth of monetary aggregates – has been

increasingly geared towards the achievement of price stability and low inflation. Very

high or hyperinflation seems to have disappeared, at least since the mid 1990s.

2

Average inflation in developing countries (including many countries of Eastern

Europe and Latin America with histories of high inflation) has declined from triple-

digit figures in the late 1980s to low single-digit figures by the end of 2001 (Domaç

and Yücel, 2005). Wynne and Kersting (2007: 2) report that inflation in the last three

decades has dropped dramatically, averaging just 5.8 percent in developing countries

and 2 percent in industrial countries since 2000. It is in this context of relatively low

global inflation that the double-digit Russian inflation stands out. In 2007, the

consumer price index increased by 11.9%, which is significantly higher than 9.0% in

2006. The government’s inflation forecast at the beginning of that year (2007) was

8.0%. In a global environment where inflation outcomes have dramatically improved

independent of the monetary framework for achieving price stability – that is, both in

countries that have adopted inflation targets as well as in countries that have not (Ball

and Sheridan, 2005) – the failure of the Russian central bank to control inflation is

striking.

This paper aims to identify the true strategy pursued by the Bank of Russia:

targets are announced for both the exchange and the inflation rates (respectively, a

managed float (Keller and Richardson, 2003) and a twelve-month CPI target)

reflecting therefore a problematic mix (Vdovichenko and Voronina, 2006). These

targets seem to be often missed with no consequence for the monetary authority.

Stabilizing the exchange rate has become a major monetary policy goal in a number

of CIS countries, namely Belarus, Kazakhstan, Russia and Ukraine, (Bauer and Herz,

2007).

In the absence of well-developed financial markets (Kutan and Brada, 2000),

capital inflows causes serious problems for domestic monetary policies. With rising

capital inflows, the central bank is more likely to buy excess foreign exchange supply

3

on the market, leading to monetary expansion and inflation. On the other hand, if the

central bank does not intervene in the foreign exchange market, the result could be

nominal appreciation of the local currency which may lead to deterioration of the

current account combined with reduced profitability. At the same time, repatriated

proceeds from the export of oil and gas are quite successfully sterilized by means of

oil price-related variations in the marginal rates of taxes specific to the oil sector, the

proceeds of which flow into the stabilization fund.

Against this background, we search for a possible relation between the real

interest rate, inflation rate, exchange rate, reserve growth, output growth (proxied by

growth in industrial production) and oil prices to indicate what type of monetary

policy has been followed by the Bank of Russia during the period 1995 to 2007. We

test whether the real interest rate is reacting to policy driven exchange rate shocks. If

this is the case, then it can be demonstrated that the monetary policy stance of the

central bank has not been primarily designed for price stability. Our focus is informed

by the argument that inflation persistence is shown to come from economic agents’

limited information about the central bank’s policy objectives (Erceg and Levin,

2003). Thus in this paper, by imposing sign restrictions on impulse responses, we

identify exchange rate shocks and then examine the impulse response functions.

The structure of this paper is as follows. Section 2 gives a snapshot of the story

of the Russian inflation and derives an inflation equation. A monetary policy rule in

an open economy is formulated in order to test empirically the reaction of the real

interest rate to exchange rate changes, inflation, and output and reserve growth.

Section 3 then presents the data, methodology and empirical results. Using both a

vector error correction (VEC) and Uhlig (2005)’s ‘pure sign restriction’ modelling

strategy, this paper suggests that the authorities have targeted the exchange rate –

4

specifically, accumulating foreign reserves in an attempt to prevent nominal

appreciation, but the net effect of what might be termed this weak rouble policy has

become a rise in inflation. Section 4 concludes.

II. Modelling the inflation Dynamics

We have written extensively on the repeated episodes of high inflation in post-

communist Russia (Granville (1995), Ferguson and Granville (2000), Granville

(2001), Granville and Mallick (2006), explaining the constraints – namely the budget

deficit (until 2000 when the fiscal adjustment was made) and the exchange rate –

faced by the central bank in its conduct of monetary policy. While the literature is

extensive, naming a few include Choudhry (1998), Cotarelli and Doyle (1999)

Esanov, Merkl and de Souza (2005), Kutan and Brada (2005), Oomes and Ohnsorge

(2005), Papazoglou and Pentecost (2004), Starr (2005) and Vdovivhenko and

Voronina (2006), our aim in this section is not to offer a review of this literature but

rather a brief summary of our work aimed primarily at explaining why we see the

Russian authorities’ preoccupation with the exchange rate as hampering monetary

policy in its task of achieving low inflation.

A. A Snapshot of the Inflation Dynamics

In the early period 1992-93, an effective monetary policy framework was lacking due

not only to the challenge of establishing new institutions and regulations, but

especially also to the difficulty of overcoming the legacy of central planning where

budget and credit financing were indistinguishable. The absence of domestic financial

markets (the T-bills market was introduced in May 1993) and access to international

5

capital meant that the government budget deficits (consistently over 5% of GDP year

after year) were financed borrowing from the central bank. This ‘monetary financing’

led to very high inflation. The average monthly inflation rate reached 41 percent in

1992 due to the price jump of over 200 percent in January 1992 following price

liberalization (that average monthly figure would be 18 percent without the January

1992 observation), and 21 percent in 1993 (Granville, 2001:100). In our empirical

analysis, we decided to omit 1992 and 1993 in the hope that by omitting these first

years of transition, the data will produce a less noisy measure to extract the results of

the central bank’s performance.1

The hard inflationary times of this early period eventually forced the

government into a stabilization effort, culminating in the signing of a fully-fledged

IMF programme on 26 March 1995, and in April that year, the independence of the

central bank was increased allowing for the institutional separation of monetary and

fiscal policy.2 The central bank stopped officially acting as the banker of the

government and state firms.3

It is during this period that the exchange rate started to act as the other major

constraint for monetary policy – especially after T-bill issuance replaced seigniorage

as the deficit financing mechanism. From early 1995, there was a shift in the

financing of the budget deficit away from central bank credits in favour of short-term

1 Although our sample starts from 1995, in calculating growth rates of different variables, 1994

observations have been used as the initial values.

2 The “Central Bank Law” enacted in April 1995.

3 http://www.cbr.ru/eng/today/history/central_bank.asp: “On April 26, 1995, the Bank of Russia

stopped extending loans to finance the federal budget deficit and centralised loans to individual

industries and sectors of the economy.”

6

treasury bills (or ‘GKOs’).4 The curbing in the growth of domestic credits brought a

sharp appreciation of the domestic currency – the rouble – in April-May (Granville,

2001:108). This led to the adoption in July 1995 of an exchange rate target as the

nominal anchor for the price stabilisation programme. This stabilisation drive resulted

in lower average monthly inflation compared to the previous sub-period (1.85%

compared with 13.18%). But less than three years later, in August 1998, Russia was

thrown back into a profound financial crisis. Why did this happen?

The answer is that substituting bond financing for monetary financing of

stubbornly high budget deficits was like papering over the cracks. Just as Sargent and

Wallace (1981) said would happen, it was ultimately doomed, with the catalyst of

doom proving to be the exchange rate target. In Basdevant and Hall (2002) exchange

rate expectations are shown to have played a key role in the origin of the August 1998

financial crisis.

Between May 1995 and July 1998, the central bank was demonstrating its

commitment to holding the exchange rate band by raising the refinance rate to

whatever level necessary for attracting capital inflows. The refinance rate acted as an

effective cap on the T-bill yield and so signalled the level at which the central bank

would support the price. The calculation was that capital inflows would be attracted

by the real returns available from the combination of high nominal yields on rouble-

denominated debt and the promise of a stable exchange rate. Unfortunately, the Asian

financial crisis eroded risk appetite, and investors in government papers (T-bills) were

deterred by higher interest rates in face of low or zero growth prospects, which

created doubts that the government could afford debt service costs reaching 40 per

cent of federal expenditures in May 1998. The ex-post real interest rate reached

4 1995 Federal Budget Law.

7

unsustainable levels as in May 1998 when the refinance rate was raised from 42

percent to 150 percent.

In this environment, all demand, domestic and foreign, for new issues of rouble

T-bills disappeared. The government could no longer pay debt with debt. Redemption

of weekly maturities averaging about Rbs 9 billion ($1 billion) had to be financed out

of general taxation. The problem was exacerbated by a fall in oil prices and a rise in

imports which led to the first current-account deficit since 1993, while putting further

pressure on fiscal revenues. The IMF-led rescue package agreed in mid-July 1998,

although worth some $22.6 billion, failed to reassure investors and therefore to bring

interest rates low enough to achieve a major expenditure reduction. The only possible

rescue now was drastic cuts in non-debt-service expenditure – that is, ordinary public

spending. The Finance Minister at the time, Mikhail Zadornov, warned legislators that

the pain of making such cuts would be nothing compared to the pain which would

follow a financial collapse. Zadornov’s warning went unheeded, and a sovereign

default duly occurred on 17 August 1998, soon followed by the abandonment of the

exchange rate peg (Granville, 2001: 120). That triggered a chaotic 60% devaluation,

in turn producing a resurgence of inflation.

The 1998 crisis proved a positive turning point leading by 2000 to a federal

budget surplus. The fiscal situation improved due to a radical change in

macroeconomic policy, forced – ironically enough – on the new communist-backed

government by the loss of credit-worthiness after the default. Fiscal surpluses have

replaced fiscal deficits removing one constraint for monetary policy. Fiscal surpluses

have been achieved thanks to a combination of fiscal discipline (in a radical departure

from previous practice) and high oil prices – which, at the same time, have generated

serious problems for monetary and exchange rate policy. Faced with a balance of

8

payments surplus, the monetary authorities have been pulled between the goal of

reducing inflation and the goal of restraining the real appreciation of the rouble.

Starting in 2000, the control of inflation took second place to a policy expedient

deemed necessary to protect jobs and promote output. This expedient was an

exchange rate target. The stability of the real exchange rate is one of the objectives

stated by the central bank in its monetary programme.5 This stated goal appears to be

aimed mainly at protecting output and jobs. Vdovichenko and Voronina (2006),

estimating a monetary rule for the period 1999 to 2003, show that slowing the real

exchange rate appreciation to shelter domestic producers and employment from

import competition has been given priority over reducing inflation.

During the period 2000-2007, Russia achieved annual real GDP growth rates of

around 7 percent, current account surpluses of 10 percent, and fiscal surpluses of 4

percent on average. In an environment of oil-driven balance of payments surpluses,

exchange rate targeting meant artificially restraining the natural appreciation of the

rouble to preserve competitiveness – not only in the most obvious area in this context,

which is the natural resource exporting sectors which account for the bulk of Russian

GDP and over half of the revenues of the federal budget, but also and perhaps more

especially in the import-substituting industries which are major employers. This

‘weak rouble’ policy has been pursued in the form of massive rouble interventions in

the foreign exchange market. The result has been higher inflation.

B. The Inflation Equation6

5 http://www.cbr.ru/eng/today/publications_reports/on_2005e.pdf: 3.

6 Variable definitions are reported in Appendix 1.

9

We consider price formation in the context of an open economy, therefore the

exchange rate plays an important role in both demand and supply disturbances.7 The

improvement in Russia’s external liquidity position has been spectacular in the recent

years. The substantial reserves accumulation reflected a combination of significant

trade surpluses and foreign investment inflows. That also resulted in the relative

strength of the domestic demand, and money supply growth has been much higher

than nominal GDP. Thus we need to derive a reaction function by the monetary

authority to changes in international reserves. The first relationship is the flow supply

of money ( s

tM ) that comes from the central bank’s balance sheet (where tR and

tDA are respectively international reserves and domestic assets) as given by equation:

[1] s

t t tM R + DA∆ = ∆ ∆

The second relationship defines the flow demand for money. The level of real output

(y), real interest rate (r) and exchange rate (e) determine the money demand (Md)

growth. The money demand can thus be shown via an open economy LM function as

follows:

[2] t

d

t t t tM y r eθ β δ ε∆ = ∆ − − ∆ +

where δβθ ,, are all positive. The real interest rate equals )( ei π− , where πe is

expected inflation and i is nominal interest rate. And εt is a demand disturbance. As

real interest rate is a linear combination of the two variables, it can be stationary in

line with the other first-differenced variables and hence it can be left in levels. If we

were to link equation (2) to the Cagan model of money demand under hyperinflation,

7 Starr (2005: 446) also assumes the economies of the CIS to be open.

10

then the parameter β will be of interest, as it would capture the effect of the expected

inflation rate.8

The third key relationship in this model is the assumption of money market flow

equilibrium, which continuously holds:

[3] d sM M∆ = ∆

The capital flow reaction function can thus be derived as:

[4] t t t t t tDA y r e Rθ β δ ε∆ = ∆ − − ∆ − ∆ +

In equilibrium, equation 4 shows that there is an inverse relationship between

changes in international reserves and domestic assets under a fixed exchange rate

regime, which means the central bank has to sterilize the capital inflows to keep the

money supply constant. But, with a flexible exchange rate regime, currency

fluctuations are likely to influence the conduct of monetary policy. Domestic residents

may also hold foreign currency for transactions or precautionary purposes in the

presence of domestic inflation meaning that domestic currency depreciation may lead

to a decline in real money balances encouraging currency substitution (Papazoglou

and Pentecost, 2004). Money demand in dollarized economies often appears to be

highly unstable, making it difficult to forecast and control inflation (Oomes and

Ohnsorge, 2005). This situation where real money balances are influenced by

expected inflation is partly in line with a Cagan style relation under conditions of

hyper inflation (see for example Taylor (1991) and Frankel and Taylor (1993) who

also include currency depreciation in the estimation of the money demand function for

high inflation countries). Choudhry (1998) found that the rate of change of the

exchange rate needs to be included in the demand function for M2 to obtain a

8 See Taylor, 1991; Phylaktis and Taylor, 1993, for more details on Cagan model.

11

stationary long-run relationship. Thus under a flexible exchange rate system, we need

to assess the relative impact of changes in foreign assets on the monetary side in

response to exchange rate shocks. In fact, as the economy grows, there could be

secular growth in monetary expansion.

The aggregate supply equation can be formulated following an open-economy

Phillips curve:

[5] O

t t t t t ty e DAπ λ φ ω ηπ ν= + ∆ + ∆ + +

where the coefficients λ and φ are greater than zero. υt is a supply disturbance. πO is

oil price inflation. The inclusion of DA in this equation suggests that if the exchange

rate does not change in response to changes in capital flows, higher money growth

can contribute to generating inflation. The changes in oil prices also affect the macro-

economy, as oil and gas exports account for big part of the total exports in the Russian

economy.9 We include the oil price in order to take account of external oil price

shocks as a source of business cycles.

Substituting equation [4] in equation [5], we get:

[6] ( ) ( ) ( )O

t t t t t t t ty r e Rπ λ ωθ ωβ φ ωδ ω ηπ ωε ν= + ∆ − + − ∆ − ∆ + + +

This equation suggests that as the real interest rate increases, inflation goes down;

when the currency depreciates, inflation rises; when international reserves go up,

inflation can be lowered on the back of allowing a currency appreciation, following

the surge in capital inflows. If the coefficient associated with the change in reserves

comes out negative, then the main factor that could be driving inflation in Russia is

exchange rate depreciation.

9 The gas price in Russia’s European export market is set in line with crude oil price movements.

12

The Bank of Russia determines the interest rate, lowering it to discourage

currency appreciation, leading to higher inflation via monetary expansion, hence

establishing a trade-off between the exchange rate and inflation. The exchange rate

can be included as a determining factor in the central bank’s reaction function, as

suggested by Taylor (2001). The main issue in the literature on flexible inflation

targeting is to what extent the real disturbances preventing the stabilization of both

inflation and the output gap justify the “departure from complete and (immediate)

stabilization of inflation” (Giannoni and Woodford, 2005: 96). Given the rate of

inflation as in [6], the monetary authority can adopt a monetary policy rule. In an

open economy context, a modified version of the closed-economy Taylor rule can be

written for the unobservable equilibrium real interest rate as follows10:

[7] *( )t t t tr by c d eπ π= + − + ∆

Where b, c and d are parameters, measuring the magnitude of the response of the

monetary policy to output, inflation, and changes in exchange rate, respectively. In a

high-inflation open economy, it is important to monitor the real interest rate as

negative real rates are harmful for economic activity.

This reaction function suggests first that as inflation increases the real interest

rate increases, and secondly as the real interest rate increases, money growth declines

from equation [2], and hence inflation gets stabilised from equation [6]. Also as the

exchange rate is defined as the domestic currency price of a foreign currency,

exchange rate depreciation can produce an increase in the real interest rate. As

discussed earlier, the central bank has focused more on exchange rate targeting rather

than inflation targeting as an instrument of monetary control, so the inclusion of both

10 See, for example, Molodtsova, Nikolsko-Rzhevskyy and Papell (2008).

13

exchange rate and inflation in a monetary policy rule is justified in order to

disentangle their effect on the real interest rate.

III. Empirical Results

A. Data

As explained in section II, we start our empirical analysis in 1995, as this period

corresponds more truly to a stronger resolve both institutionally and de facto by the

monetary authorities to reduce inflation. Table 1 gives a summary statistics of Russian

inflation since 1995 where four periods are identified. All the results (statistics and

graphs) are computed in EViews and RATS econometric softwares.

This paper extends Granville and Mallick (2006) where an optimal monetary

rule was derived by minimising a loss function, but using monthly data for the period

February 1992 to February 2005, without considering real output and oil price effects

on the macro economy and taking the refinance rate as the proxy for the interest rate.

In this paper, we use the overnight inter-bank lending rate rather than the refinancing

rate. Vdovichenko and Voronina (2006) explain that the refinancing rate may not be

the best proxy for the interest rate and instead they use the interbank market interest

rate on one day credit and the Bank of Russia interest rate on overnight deposit. This

is a better proxy for the interest rate, because banks in Russia over big part of the

period under analysis could not borrow at the refinancing rate from the Bank of

Russia. As a result, the refinancing rate may not be viewed as an equilibrium market

rate. Also we look at the real interest rate behaviour, as we have included a proxy for

economic activity.

14

As the time series considered in our theoretical review of interrelationships are

the real interest rate, the inflation rate, the rate of change of the exchange rate, output

growth, oil price inflation and reserves growth, we include these same variables in our

empirical exercise. In this context, it is worth mentioning Starr (2005), who uses

quarterly data to show how monetary policy affects real economic activity in Belarus,

Kazakhstan, Russia and Ukraine and finds mixed evidence but with greater effects for

Russia, where a significant effect of interest rates on output is found. In our paper,

data at a monthly frequency are used between January 1995 and December 2007.

Monthly observations on the seasonally unadjusted consumer price index, CPI, were

compiled from Datastream and were de-seasonalised using the US Census Bureau’s

X12 seasonal adjustment method in Eviews. Inflation rates (INFt) at an annual rate are

calculated using this adjusted data.11 The ruble/US$ exchange rate and the overnight

money market interest rate (inter-bank lending rate) are also taken from Datastream.

This interest rate is likely to respond to the changes in macroeconomic variables. We

calculate the real interest rate (RIRt) using the Fisher identity. Money supply here

refers to the broad monetary aggregate (M2), and money growth refers to the annual

rate of growth of M2 12 Reserves data (LIRA) refer to gross international reserves

(official reserve assets in US$), compiled from the Bank of Russia’s statistical

11 Our inflation rate is calculated as follows: inflation=((cpi_sa/cpi_sa (-12))- 1)*100; To calculate the

consumer price index, Goskomstat, the Russian State Statistics agency collect prices of 400

representative goods and services from 350 cities including every capital city of the 89 regions each

month. Monthly price changes are then aggregated for each of the 89 regions, where the weights are

based on the structure of household expenditures for the region in previous year. See Gibson et al.

(2004), page 3.

12 For more details on the money supply definition and its changes over time, see Granville and Mallick

(2006).

15

resources. The real activity variable is proxied by the industrial production index (IP).

We have used the growth in IP to reflect the real output growth in the economy and it

is available at a monthly frequency.

B. The Results

A necessary but not sufficient condition for any possible cointegration is that each of

the variables should be integrated of the same order (more than zero) or that both

series should contain a deterministic trend (Granger, 1986). The test results are

reported in Table 2. The standard Augmented Dickey-Fuller unit root tests for

inflation, IP growth, and reserves growth reject the unit root null hypothesis at 5%

level of significance, while DF-GLS and KPSS tests for inflation and real interest rate

– which outperform the ADF tests – indicate the presence of non-stationarity. DF-

GLS test suggests that the real interest rate is I(1). The inflation time series being

integrated indicates the persistence of inflation. Hence it is important to know the

source of such persistence in inflation.

More importantly, the sample period appears to have numerous structural breaks

in the policy of the Bank of Russia: the period of fast disinflation in the early part of

the sample altered by a period of relatively low (but still two digit) and stable inflation

and gradual real appreciation of the ruble before the crisis of August 1998 followed

by a period of fast growth with a low real exchange rate and since 2000 a period of

increased emphasis on inflation reduction and a declared switch from the use of the

exchange rate as a nominal anchor to monetary targets. We therefore test for possible

regime switches/structural breaks in the sample using the Zivot-Andrews test (see

Figure 2), which clearly reflects a major break in 1998.

16

Further, the estimation of unrestricted VAR models with integrated variables is

problematic, as applying the conventional statistical theory is no longer appropriate.

Given that the time series involved are either I(1) or I(0), we can apply Johansen’s

cointegration test to find whether there is any cointegrated relation between the

variables. Thus we use the reduced form theoretical relations in section 2 to identify

the cointegrated relations and derive a vector error correction (VEC) model, which can

be used to examine the impact of exchange rate shocks.

Figure 1 plots the real interest rate (RIR), inflation (INF), the rate of change of

nominal exchange rate (DEXR), the growth of industrial production (GRIP), reserves

growth (DLIRA) and oil price inflation (OINF). A visual inspection of Figure 1

indicates that the variables appear to have structurally changed. Such apparent

changes in the time series process can result from events such as a financial crisis or

significant government policy shifts. In the case of Russia, the respective instances of

these events were the August 1998 default and the change from a money anchor to an

exchange rate anchor for the stabilisation programme. Due to the currency crisis in the

late 1990s creating breaks in the data, we have added three dummy variables with the

aim to capture changes in the exchange rate volatility for the outliers in 1998 (May

and September) and in 1999 September, as exogenous variables in the VAR to

overcome the normality problem in the data generating process. Similar dummies

have been used in Esanov, Merkl and Vinhas de Souza (2005).

Thus we estimate a VAR with 4 lags as found to be optimal by the Schwarz

information criterion, which in general outperforms other lag selection test criteria

suggesting five lags (see Table 3). Within such a VAR, we test whether there is any

cointegrating relationship between the five variables (with oil price inflation

considered exogenously along with three dummy variables) for the full sample using

17

the Johansen (1988) maximum likelihood method. We find the existence of two

cointegrating relations using a linear model (with intercept and trend). The presence

of two cointegrating vectors has also been checked with alternate assumptions. The

trace statistics (LR test) indicate two cointegrating equations at 1% significance level.

The fact that the null r=1 is rejected at a 1% significance level means that there exists

two meaningful long-run relations between these variables (see Table 4 and Figure 3).

The Trace test indicates the existence of two cointegrating relations and we have

identified the first relation to be an interest rate rule and the second vector to be an

inflation equation by imposing five restrictions. Here we discuss the restrictions that

are imposed to identify the cointegration vectors. Denoting the two long-run

equilibrium relationships by β1 = [β11, β12, β13, β14, β15]′ and β2 = [β21, β22, β23, β24, β25]′

respectively, we impose the following restrictions on the vector of variables [RIR, INF,

DEXR, GRIP, DLIRA] for identification: β1 = [1, β12, 0, β14, 0]′ and β2 = [0, 1, β23, β24,

β25]′. The restrictions on β1 indicate an equation for RIR. Theoretically the second vector

must be interpreted as an inflation equation in order to have a complete model in which

exchange rate generates inflation, which in turn influences RIR. The over-identifying

restriction is the coefficient associated with international reserves: β15=0, which can be

included in the second vector, but may not make good sense in the first vector. So the

hypothesis that international reserves can enter the second cointegrating relation, but

not the first one, cannot be rejected. The likelihood ratio (LR) test statistic for testing

the binding β restrictions (rank =2) is distributed as χ2(1) = 2.013 [0.156], which is

accepted. The LR test for these restrictions identifies both the cointegrating vectors.

The normalized cointegrating equations obtained from Johansen test in Table 4 can be

written as:

18

[8] (16.71) (7.645) (6.206)

(7.127) (8.608) (3.842) (4.952)

0.366 1.32 0.513 0.003

0.007 0.593 0.035 0.254 0.004

t t t t

t t t t t

RIR INF LEXR T

INF LEXR GRIP LIRA T

= − + − ∆ −

= + ∆ − − ∆ +

where T refers to a time trend. Figures in the parentheses under the estimated

normalized coefficients are t-values. All the variables included in the normalised

cointegrating relations are statistically significant, exchange rate changes have

significantly influenced inflation, thus influencing real interest rate in the first long-

run relation. The signs of the coefficients turned out to be as per the apriori

expectation in equations [6 and 7]. The above two long-run relationships can be

summarized as follows:

First, an increase (decrease) in inflation by 1 percent is associated with an

increase (decrease) in the real interest rate by about up to 1.32 percent in the sample

period. It should be noted that the relatively high coefficient of inflation partly reflects

the fact that inflation and nominal interest rates have been highly volatile during the

first part of the sample period.

Second, an increase (depreciation) in exchange rate by 1 percent is associated

with a reduction in the real interest rate of about 0.51 percent. Also as the exchange

rate depreciates, there is 0.59 percent increase in inflation from the second long-run

relation. Considering the two relations together, the real interest rate increases by 0.27

percent in response to a 1 per cent currency depreciation.

Third, from the second cointegrating relation, there is a trade-off between output

growth and inflation, which implies that an increase (decrease) in output growth of 1

percent is associated with a decline (increase) of the real interest rate of 0.05 percent

from the first relation. The negative coefficient associated with reserves growth

suggests that inflation should decline. But the persistent inflation indicates that

inflation has been driven by changes in the exchange rate towards depreciation.

19

Finally, the inclusion of a time trend in both relations suggests that the real

interest rate has significantly declined over the sample period, inflation appears to

have significantly increased over time, most likely due to the authorities’ resistance to

exchange rate appreciation, which would otherwise have occurred following the surge

in foreign exchange inflows.

Given these long-run relations, we need to formulate a short-run model to

examine the dominant effect of a shock to these variables. Further, as there are four

lagged variables for each series, the summary responses can be more insightful, and

we address this by employing generalised impulse response (GIR) analysis with the

VEC model. The GIR functions are not sensitive with respect to the ordering of

variables as in the Choleski decomposition (Pesaran and Shin, 1998). The impulse

responses following a shock to each of the endogenous variables are presented in

Figure 4. The results suggest that the exchange rate is the key driving factor for

inflationary pressure; and changes in the exchange rate contribute more to inflation,

which in turn adds to the variability of the real interest rate in the long-run.

Besides this traditional identification strategy, to further validate the results, we

adopt Uhlig (2005)’s new sign restriction method, which is robust to non-stationarity

of series including breaks. Also the zero long-run restrictions may appear very

stringent, as they are based on statistical properties of the data rather than serious

theoretical consideration. Within this framework, we identify three types of

underlying disturbances, namely an oil price, exchange rate and monetary policy (real

interest rate) shock. Given that oil prices can capture supply shocks, having obtained

the parameter estimates of the reduced form VAR, we impose three sign restrictions

as follows:

20

1. The oil price inflation does not decrease (>0) in response to its own

positive shock.

2. The changes in exchange rate will not decline (>0) in response to its

own positive shock, i.e. exchange rate depreciation.

3. The real interest rate does not decrease (>0) in response to a positive

exchange rate shock, as monetary policy will tighten to back up the

exchange rate. As in Figure 1, the real interest rate always remains

positive. This has also been observed in the cointegrating relations, as

discussed earlier.

These restrictions seem reasonable in the light of the observed pattern in the Russian

data. As the domestic currency depreciates, inflation could increase, while exported

goods become cheaper relative to imported goods, thus increasing demand for

domestic goods and thus higher industrial production. But we do not pre-judge this

outcome, as we would like this to be revealed from the impulse response functions.

Also exchange rate changes can either exhibit depreciation or appreciation. We

examine both types of exchange rate changes to establish any asymmetric behaviour

in responses. These restrictions help derive respective impulse vectors, which are

defined as innovations to the VAR system in response to a unit shock in each

disturbance. We keep those impulse vectors whose impulse response functions satisfy

the sign restrictions and discard the others.

We undertake the exercise with both exchange rate depreciation and

appreciation, as the Russian central bank has been struggling to find a balance

between inflation and nominal appreciation of the rouble. As in Figures 5 and 6, we

find that inflation responds positively to rouble depreciation, but as foreign reserves

growth has been significantly high, it puts pressure on the central bank to allow the

21

currency to appreciate. Thus we examine what would happen if the currency

appreciation shock occurs (see figure 6). As this has not been practised by the central

bank, the response does not appear to mirror the depreciation case. Rather the

response remains mixed across the board. The exchange rate shock does not obey the

restriction in the long-run, and it starts to depreciate after 7 months. This clearly

suggests that Russian monetary authorities have been following exchange rate

targeting to keep it depreciated that in turn keeps inflation persistent. We also check

for the robustness of our result by replacing international reserves with broad money

(see figures 7 and 8). The responses still remain similar, as in the recent years

international reserves make up the big part of the monetary expansion. Besides, there

seems to be a positive real effect of exchange rate shocks in the long-run as found in

the impulse responses for GRIP (see figure 5). This comes out clearly with

depreciation, but not with appreciation. Thus the central bank remains reluctant to let

the currency appreciate thus depreciation fuels inflation.

Overall, the exchange rate channel has played an important role in the conduct

of monetary policy in Russia. Since 2000, the Russian central bank made an attempt

both to delay (or even reverse) the real appreciation of the ruble by actively buying

foreign reserves in the open market, and to reduce inflation from double digit levels.

Clearly, these two conflicting goals, without an effective ability to sterilize the

inflows, can directly lead to inflationary pressures in the economy. In this paper, we

measured empirically the implicit policy that the Russian monetary authority has

pursued – real appreciation deceleration versus reduction in inflation – finding that

exchange rate shocks are crucial determinants of inflationary pressure.

22

IV. Conclusion

In the long-run, we find that while inflation and exchange rate changes have

significantly influenced the real interest rate, it is exchange rate changes rather than

reserves and output growth which have had the stronger long-run impact on inflation.

This is in contrast to studies including the period before 1995, where monetary

aggregates were the key factor determining inflation as for example in Esanov, Merkl

and Vinhas de Souza (2005). In addition, the finding that real interest rates react

positively to inflation suggests that the ‘Fisher effect’ does not seem to hold in the

long-run in a high-inflation transition economy. For the Fisher effect to hold, the real

interest rate should not change in response to changes in inflation, and, consequently

the inflation rate can have a one-to-one positive effect on the nominal interest rate.

This result further strengthens our finding that the exchange rate changes have been

the driving forces behind interest rate determination, confirming that the central

bank’s policy objective was to prioritise resistance to real exchange rate appreciation

over price stability. Within a high oil price environment, the monetary authorities

seem to have prioritised stabilizing the exchange rate in search of short run output

gains over bringing down .inflation. This explains in part why inflation has become

entrenched at a relatively high level. This policy of balancing the inflation target with

a contradictory exchange rate target seems futile, since the resulting higher inflation

in any case drives up the real trade-weighted value of the currency, producing exactly

the loss of competitiveness which the authorities want to avoid.

23

References

Ball, L. and N. Sheridan (2005). Does Inflation Targeting Matter? The Inflation-

Targeting Debate. B. Bernanke and M. Woodford. Chicago, University of

Chicago Press: 249-82.

Bauer, C., and Herz, B., (2007) Credibility of CIS Exchange Rate Policies – a

Technical Trader's View, Emerging Markets Review, 8(1): 50-66.

Basdevant, O. and S. G. Hall (2002). "The 1998 Russian Crisis: Could the Exchange

Rate Volatility have Predicted it?" Journal of Policy Modeling 24(2): 151-168.

Central Bank of the Russian Federation (2004). “Guidelines for the State Monetary

Policy in 2005” Approved by the Bank of Russia Board of Directors on

November 15, http://www.cbr.ru/eng/today/publications_reports/on_2005e.pdf

Choudhry, T. (1998). "Another Visit to the Cagan Model of Money Demand: The

latest Russian Experience." Journal of International Money and Finance 17(2):

355-376.

Cotarelli, C. and P. Doyle (1999). "Disinflation in Transition 1993-1997." IMF

Occasional Paper, No. 179.

Domaç, I., and E.M. Yücel (2005). What Triggers Inflation in Emerging Market

Economies? Review of World Economics, 141 (1): 141 – 164.

Erceg, C. and A.Levin (2003). “Imperfect credibility and inflation persistence”,

Journal of Monetary Economics, 50(4): 915-44.

Esanov, A., C. Merkl and L Vinhas de Souza (2005). “Monetary Policy Rules for

Russia”, Journal of Comparative Economics, 33 (3): 484-499.

Ferguson, N. and B. Granville (2000). "Weimar on the Volga: Causes and

Consequences of Inflation in 1990s Russia Compared with 1920s Germany."

Journal of Economic History 60(4): 1061-1087.

24

Frankel, J. and M. P. Taylor (1993). "Money Demand and Inflation in Yugoslavia,

1980-89." Journal of Macroeconomics 15: 455-481.

Giannoni, M. P. and M. Woodford (2005). Optimal Inflation-Targeting Rules. The

Inflation-targeting debate. B. Bernanke and M. Woodford. Chicago and London,

National Bureau of Economic Research, Studies in Business Cycles, The

University of Chicago Press. 32: 93-172.

Gibson, J., S. Stillman and T. Le (2004). "CPI bias and Real Living Standards in

Russia during the Transition." William Davidson Institute Working Paper

684(May).

Granger, C. W. J. (1986). "Developments in the study of cointegrated economic

variables." Oxford Bulletin of Economics and Statistics 48: 213-228.

Granville, B. (1995). The Success of Russian Economic Reforms published by RIIA

distributed by Brookings, 1995.

Granville, B. (2001). The Problem of Monetary Stabilisation. Russia's Post-

Communist Economy. B. Granville and P. Oppenheimer. Oxford, Oxford

University Press, 93-130.

Granville, B. and S. Mallick (2006). "Does Inflation or Currency Depreciation Drive

Monetary Policy in Russia?" Research in International Business and Finance,

20 (2): 163-179.

Keller, P. M. and T. Richardson (2003). "Nominal Anchors in the CIS." IMF Working

Paper WP/03/179.

Kutan, A.M. and J.C. Brada (2000) “The Evolution of Monetary Policy in Transition

Economies”, Review, The Federal Reserve Bank of St. Louis, 82 (2),

(March/April 2000), pp. 31-40.

25

Molodtsova, T., A. Nikolsko-Rzhevskyy and D.H. Papell (2008). Taylor rules with

real-time data: A tale of two countries and one exchange rate. Journal of

Monetary Economics, 55(1): S63-S79.

Oomes, N., Ohnsorge, F. (2005). Money demand and inflation in dollarized

economies: The case of Russia, Journal of Comparative Economics, 33 (3):

462-483.

Papazoglou, C. and E. Pentecost (2004). "The Dynamic Adjustment of a Transition

Economy in the Early Stages of Transformation." Journal of Macroeconomics

26(3): 547-561.

Pesaran, H.H. and Y. Shin (1998). ‘Generalized Impulse Response Analysis in Linear

Multivariate Models’, Economics Letters, 58, 17-29.

Phylaktis, K. and Taylor, M.P (1993). ‘Money Demand, the Cagan Model and the

Inflation Tax: Some Latin American Evidence’, The Review of Economics and

Statistics, 75 (1), 32-37.

Starr, M. A. (2005). "Does money matter in the CIS? Effects of monetary policy on

output and prices." Journal of Comparative Economics 33: 441-461.

Taylor, J. B. (1993). “Discretion versus policy rules in practice.” Carnegie-Rochester

Conference Series on Public Policy 39: 195-214.

Taylor, J. B. (2001). "The Role of the exchange Rate in monetary policy rules." The

American Economic Review, Papers and Proceedings 91(2): 263-267.

Taylor, M. P. (1991). "The Hyperinflation model of money demand revisited."

Journal of Money, Credit, and Banking 23: 327-351.

Uhlig, H. (2005). What are the effects of monetary policy on output? Results from an

agnostic identification procedure, Journal of Monetary Economics, 52 (2): 381-

419.

26

Vdovichenko, A.G. and V.G. Voronina (2006). "Monetary policy rules and their

application in Russia." Research in International Business and Finance 20 (2):

145-162.

Wynne, M. A. and E. K. Kersting (2007). "Openness and Inflation." Staff Papers

Federal Reserve Bank of Dallas 2(April).

27

TABLE 1-Russian Inflation, 1995-2007: Summary Statistics

(annual percentage change in CPI)

Sample: Jan. 1995-July 1995 Aug. 1995-July 1998 Aug. 1998-Dec. 1999 Jan. 2000-Dec. 2007

Mean 219.7 49.5 81.2 13.3

Median 221.2 19.0 83.9 12.5

Maximum 225.7 222.0 125.7 25.9

Minimum 209.7 5.8 9.3 7.1

Std. Dev. 5.8 60.5 34.9 4.2

Observations 7 36 17 96

Source: calculated with data from Datastream

TABLE 2-Testing for stationarity

ADF(τ) ADF(µ) DFGLS KPSS

INFt -3.934** -3.757** -2.190 0.107 RIRt -5.208** -4.581** -1.127 0.248**

∆LEXRt -2.696 -2.505 -2.159 0.063

GRIPt -4.011** -3.742** -3.319* 0.092 OINFt -3.238 -3.122* -3.199* 0.047

∆LIRAt -2.548 -1.681 -2.929 0.075

5% -3.442 -2.881 -2.989 0.146

Notes: ADF(τ) and ADF(µ) are tests of the unit root null hypothesis, containing a constant and a trend, a constant and no deterministic components, respectively. Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test assumes that the null-hypothesis is stationary. DFGLS and KPSS tests also include a constant and a

trend; * and ** denote 5% and 1% levels of significance respectively. INFt, RIRt, ∆LEXRt, GRIPt,

OINFt and ∆LIRAt are inflation, real interest rate, change in exchange rate, growth of industrial production, oil price inflation and change in international reserves respectively.

TABLE 3 -VAR Lag Order Selection Criteria

Lag LogL LR FPE AIC SC HQ

0 -216.4847 NA 2.83e-05 3.7151 4.2666 3.9392

1 679.2332 1653.633 4.31e-11 -9.6805 -8.5776 -9.2324

2 745.5728 117.3701 2.29e-11 -10.3165 -8.6621 -9.6443

3 789.7993 74.8447 1.71e-11 -10.6123 -8.4065 -9.7160

4 868.9578 127.8714 7.54e-12 -11.4455 -8.6883* -10.3251*

Notes: Sample: 1995M01 2007M12; Included observations: 130; Endogenous variables: RIR INF

GRIP, ∆LEXR ∆LIRA; Exogenous variables: C, DUM, DUM1, DUM2, OINF; * indicates lag order selected by the criterion; LR: sequential modified LR test statistic (each test at 5% level); FPE: Final prediction error; AIC: Akaike information criterion; SC: Schwarz information criterion; HQ: Hannan-Quinn information criterion.

TABLE 4: Unrestricted Cointegration Rank Test (Trace)

Hypothesized Trace 0.05

No. of CE(s) Eigenvalue Statistic Critical Value Prob.**

None * 0.713 252.514 88.804 0.000

At most 1 * 0.271 85.105 63.876 0.000

At most 2 0.169 42.749 42.915 0.052

At most 3 0.122 17.871 25.872 0.353

At most 4 0.004 0.490 12.518 1.000

Notes: Linear deterministic trend is included. Trace test indicates 2 cointegrating eqn(s) at the 1% level; * denotes rejection of the hypothesis at the 0.05 level; **MacKinnon-Haug-Michelis (1999) p-values

28

FIGURE 1 - Real Interest rate (RIR), Inflation (INF), rate of change of nominal

exchange rate (DEXRATE), Growth of industrial production (GRIP), Oil price

inflation (OINF) and international reserve growth (DLIRA), January 1995 to

December 2007

0.0

0.4

0.8

1.2

1.6

2.0

95 96 97 98 99 00 01 02 03 04 05 06 07

RIR

0.0

0.2

0.4

0.6

0.8

1.0

1.2

95 96 97 98 99 00 01 02 03 04 05 06 07

INF

-.2

.0

.2

.4

.6

.8

95 96 97 98 99 00 01 02 03 04 05 06 07

DEXRATE

-0.8

-0.4

0.0

0.4

0.8

1.2

95 96 97 98 99 00 01 02 03 04 05 06 07

DLIRA

-20

-10

0

10

20

30

95 96 97 98 99 00 01 02 03 04 05 06 07

GRIP

-0.8

-0.4

0.0

0.4

0.8

1.2

95 96 97 98 99 00 01 02 03 04 05 06 07

OINF

Source: Calculated using data from Datastream

29

Figure 2: Zivot-Andrews Structural break test for RIR

1997 1998 1999 2000 2001 2002 2003 2004 2005-5.6

-5.2

-4.8

-4.4

-4.0

-3.6

-3.2

-2.8

Figure 3: Estimated cointegrating relations

-1.2

-0.8

-0.4

0.0

0.4

0.8

95 96 97 98 99 00 01 02 03 04 05 06

Cointegrating relation 1

-1.2

-0.8

-0.4

0.0

0.4

0.8

1.2

95 96 97 98 99 00 01 02 03 04 05 06

Cointegrating relation 2

Figure 4: Impulse responses from the vector error correction model

-.05

-.04

-.03

-.02

-.01

.00

.01

.02

.03

5 10 15 20 25 30 35

INF

GRIP

DEXR

DLIRA

Response of RIR to Generalized One

S.D. Innovations

-.08

-.06

-.04

-.02

.00

.02

.04

5 10 15 20 25 30 35

INF

GRIP

DEXR

DLIRA

Response of INF to Generalized One

S.D. Innovations

30

Figure 5: Impact of exchange rate depreciation, oil price inflation and RIR, on CPI Inflation

IRA and IP growth

Shocks in RIR, exchange rate depreciation and Oil inflation

Impulse Responses w ith Pure-Sign Approach

Impulse Responses for rir

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.01

0.00

0.01

0.02

0.03

0.04

0.05

0.06

Impulse Responses for inf

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

0.00

0.01

0.01

0.01

0.02

0.03

0.03

0.04

Impulse Responses for dlexr

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.01

0.00

0.01

0.02

0.02

0.03

0.04

0.05

0.06

0.06

Impulse Responses for grip

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-1.60

-1.20

-0.80

-0.40

-0.00

0.40

0.80

1.20

Impulse Responses for dlira

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.07

-0.05

-0.02

0.00

0.03

0.05

Impulse Responses for doil

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.01

0.00

0.01

0.02

0.04

0.05

0.06

0.07

0.08

Figure 6: Impact of exchange rate appreciation, oil price inflation and RIR, on CPI Inflation

IRA and IP growth

Shocks in RIR, exchange rate appreciation and Oil inflation

Impulse Responses w ith Pure-Sign Approach

Impulse Responses for rir

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.02

-0.01

0.00

0.01

0.02

0.03

0.04

Impulse Responses for inf

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.04

-0.03

-0.02

-0.01

-0.01

-0.00

0.01

0.01

0.02

Impulse Responses for dlexr

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.06

-0.05

-0.03

-0.02

0.00

0.02

0.03

0.05

Impulse Responses for grip

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-1.50

-1.00

-0.50

0.00

0.50

1.00

1.50

2.00

Impulse Responses for dlira

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.07

-0.05

-0.02

0.00

0.03

Impulse Responses for doil

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.04

-0.02

0.00

0.02

0.04

0.06

0.08

0.10

31

Figure 7: Impact of exchange rate depreciation, oil price inflation and RIR, on CPI Inflation,

M2 and IP growth

Shocks in RIR, exchange rate depreciation and Oil inflation

Impulse Responses w ith Pure-Sign Approach

Impulse Responses for rir

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0 .01

0 .00

0 .01

0 .02

0 .03

0 .04

0 .05

0 .06

Impulse Responses for inf

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0 .01

-0 .01

0 .00

0 .01

0 .01

0 .01

0 .02

0 .03

0 .03

0 .04

Impulse Responses for dlexr

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0 .01

0 .00

0 .01

0 .02

0 .02

0 .03

0 .04

0 .05

0 .06

0 .06

Impulse Responses for grip

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-1.50

-1.00

-0.50

0.00

0.50

1.00

Impulse Responses for dlm2

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.02

-0.01

0.00

0.01

0.02

0.03

Impulse Responses for doil

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.01

0.00

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Figure 8 - Response to Exchange rate appreciation, RIR and oil price inflation

Shocks in RIR, exchange rate appreciation and Oil inflation

Impulse Responses w ith Pure-Sign Approach

Impulse Responses for rir

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0 .02

-0 .01

0 .00

0 .01

0 .02

0 .02

0 .03

0 .04

0 .05

Impulse Responses for inf

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0 .04

-0 .02

-0 .01

-0 .00

0 .01

0 .02

Impulse Responses for dlexr

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0 .05

-0 .04

-0 .02

0 .00

0 .02

0 .04

0 .05

Impulse Responses for grip

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-2.00

-1.50

-1.00

-0.50

0.00

0.50

1.00

Impulse Responses for dlm2

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.04

-0.04

-0.03

-0.03

-0.02

-0.01

-0.01

-0.00

0.00

0.01

Impulse Responses for doil

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

-0.05

-0.03

0.00

0.03

0.05

0.08

0.10

32

Appendix 1: Variable Definitions

Variables Definitions Variables Definitions

Theoretical Empirical

supply of money M2

R international reserves

IRA international reserves

DA domestic assets

y level of real output IP industrial production index

GRIP growth of industrial production

r real interest rate RIR real interest rate

e exchange rate EXR nominal exchange rate

Md demand of money

parameters

expected inflation INF inflation rate

OINF oil price inflation

i nominal interest rate

l coefficient

f coefficient

b parameter

c parameter

d parameter

s

tM

δβθ ,,

εt demand disturbance

υt supply disturbance

πO oil price inflation

πe