Sri Lanka: Selected Issues
Author: International Monetary Fund
Publication Date:
04 Nov 1997
eISBN:
9781451823370
Language:
English
Keywords:
ISCR; CR; liberalization; export; import; inflation rate; exchange rate; banking system; treasury bill rate; GDP deflator; Demand for money; Monetary aggregates; Inflation
Download PDF (2.7 MB)

This Selected Issues paper on Sri Lanka provides background information on economic developments and on selected policy issues facing Sri Lanka. The main economic developments in 1996 and the first quarter of 1997 are discussed. The paper highlights that in 1996, a severe drought, power shortages, and an escalation in the military conflict contributed to a sharp deterioration in the economic situation. With the end of the drought and power shortages, and a rise in investor confidence, macroeconomic conditions in 1997 were more favorable.

Abstract

This Selected Issues paper on Sri Lanka provides background information on economic developments and on selected policy issues facing Sri Lanka. The main economic developments in 1996 and the first quarter of 1997 are discussed. The paper highlights that in 1996, a severe drought, power shortages, and an escalation in the military conflict contributed to a sharp deterioration in the economic situation. With the end of the drought and power shortages, and a rise in investor confidence, macroeconomic conditions in 1997 were more favorable.

IV. Money Demand and Monetary Policy in Sri Lanka31

A. Introduction

124. Estimates of money demand can play an important role in guiding monetary policy. However, the stability of these empirical estimates, especially in the short-run, is critical for their operational usefulness. Most industrial countries that used to rely solely on various monetary aggregates for intermediate targets have, over the last decade, adopted other approaches. This change was prompted by increased instability in the demand for money following the financial liberalization and innovation of the 1980s which rendered monetary targeting a much less reliable predictor of inflation and output. Popular among the new approaches is inflation targeting, where inflation forecasts replace monetary aggregates as the intermediate target. Less formalized is the eclectic, multiple indicator approach followed by some central banks, where varying degrees of emphasis are given to monetary aggregates, interest and exchange rates, asset prices, inflation and real activity indicators in guiding monetary policy.

125. Central banks in developing countries still place considerable importance on monetary aggregates. To some extent, this emphasis may be the result of the “financial programming” approach to policy formulation which tends to stress monetary and credit aggregates as critical variables, affecting inflation, the balance of payments and real activity. It may also be because this approach is institutionally less demanding than some of the alternatives. Yet, as development of the financial sector advances, developing countries could also be faced with instability in money demand as has been the case in industrial countries, on account of both the process of monetization and increasingly sophisticated financial instruments.

126. The issue of instability in money demand is of particular relevance in Sri Lanka, where, although inflation has rarely exceeded the high teens, monetary policy has not succeeded in permanently bringing down inflation. The question of whether this outcome was a result of instability in money demand, which rendered monetary aggregates an inadequate guide to the conduct of policy, has clear implications for the approach to formulating monetary policy in the future.

127. The paper is organized as follows: Section B contains a brief description of the major structural changes that have taken place in Sri Lanka since the late 1970s and presents some stylized facts about the main monetary aggregates, activity variables, prices and interest rates. Section C discusses issues related to the specification of the money demand equation, the estimation methodology, and the main results. These results suggest that:

  • There is only weak evidence of stability of broad money demand over the past two decades (with estimated income elasticities of about 1.1 to 1.3), while the demand for narrow money is highly unstable.

  • Parameter instability in both the long-run and short-run dynamic relationships, as well as a weakening of the fit of the short-run equation in the early 1990s, suggest that changes are under way that are likely to render even broad money targeting progressively more difficult.

  • Although interest rates do not appear to have a long-run impact on broad money demand, they do affect short-run money holding behavior, suggesting their potential as both instruments and intermediate targets of monetary policy.

128. Section D describes the monetary policy framework in Sri Lanka, in terms of objectives, operating and intermediate targets and instruments. Section E contains a description of the implications for the conduct of monetary policy of these results, and a discussion of the use of alternative monetary policy frameworks and intermediate targets.

B. Stylized Facts

129. Sri Lanka’s economy has undergone major structural changes over the past two-three decades. The entire structure of the economy was reoriented in 1977, a year which marked the start of a period of significant economic and financial liberalization The most important changes were made in the exchange and trade systems—quantitative restrictions were largely replaced by tariffs under an open general licensing system; the exchange rate was unified; many controls on capital transactions were lifted; the rupee was devalued by more than 45 percent and a managed float was adopted with a view to making the exchange rate an active policy instrument. In addition, most price controls were lifted and prices became market-determined; universal food subsidies were replaced by a means-tested food stamp scheme; a number of measures were implemented to encourage foreign investment, including generous tax incentives and the establishment of an export processing zone.

130. Of special relevance to the issues in this paper were the changes that were implemented in the financial sector. Deregulation in the financial sector began with the easing of administrative controls on interest rates in 1977, followed by the lifting of entry restrictions into the banking system, including by branches of foreign banks. This was accompanied by a foreign currency banking scheme under which commercial banks would be authorized to operate foreign currency banking units in the country. Although interest rates were no longer administered, they remained subject to government intervention, and became market determined only in 1991. Even so, the two state-owned banks continue to dominate the markets for deposits and loans, affecting the setting of interest rates.

131. The evolution of the main variables of interest in this chapter are depicted in Charts IV.1 to IV.8. Chart IV.1 shows the evolution of conventionally defined narrow (Ml) and broad money (M2) aggregates since 1960. Movements in an augmented measure of broad money (M2+) which includes foreign currency deposits are also shown. With the liberalization of the financial sector, it is clear that the behavior of monetary aggregates exhibited a dramatic change since the early 1980s, characterized by much higher growth rates (a steeper slope) and relatively greater volatility (Chart IV.6). Also, foreign currency deposits as a share of M2 have risen over this period—from 20 percent in the early 1980s to 34 percent in 1996.

CHART IV.1
CHART IV.1

SRI LANKA: MONETARY AGGREGATES, 1960–96

(In billions of rupees, end of period)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.1/ Including foreign currency deposits.

132. Chart IV.2 illustrates the behavior of prices—measured both by the CPI and the GDP deflator—before and after the liberalization episode of 1977 and again suggests that a sharp structural break occurred during the late 1970s, concurrent with price liberalization. Although inflation in Sri Lanka has rarely exceeded the teens—averaging 12–13 percent since 1977—it has not remained below 5 percent for more than one year (Chart IV.7).

CHART IV.2
CHART IV.2

SRI LANKA: PRICE DEVELOPMENTS, 1960–96

(1990=100)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.

133. The behavior of interest rates on three-month treasury bills is depicted in Chart IV.3, and shows that the initial interest rate liberalization in 1977, which consisted of the raising of interest rate ceilings, resulted in several step adjustments in treasury bill rates. Beginning in 1980, treasury bill interest rates exhibit wider movements (in both directions), suggesting their progressively greater sensitivity to market forces.

CHART IV.3
CHART IV.3

SRI LANKA: INTERNET RATES, 1960–96 1/

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.1/ Three-month treasury bill rate, in percent per annum, end of period.

134. Chart IV.4 contains information on the real GDP and the sectoral decomposition of GDP. Unlike the other series discussed above, there does not appear to be a clearly discernible change in the behavior prior to and after the late 1970s. As can be seen from Chart IV.8, apart from a short-lived pick-up in output growth rates in the aftermath of the liberalization and reforms in the late 1970s, growth rates have fluctuated around the same historical average levels over the entire sample period. There has been some sectoral diversification, particularly a declining trend in the share of the agricultural sector in total GDP—albeit with some fluctuations—but this process has been fairly gradual.

CHART IV.4
CHART IV.4

SRI LANKA: INDICATORS OF ECONOMIC ACTIVITY, 1960–96

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.

135. Finally, the ratio of monetary aggregates to GDP are shown in Chart IV.5. The ratio of narrow money to GDP has halved from about 20 percent in the early 1960 to about 10 percent in 1996, while exhibiting sizeable fluctuations during this period. These fluctuations reflect not only portfolio shifts in response to macroeconomic and structural conditions, but also idiosyncratic effects such as those related to the escalation of the civil war which may have encouraged the public to increase their demand for liquid narrow money balances. Over the same period, the ratio of broad money to GDP has risen by about 10 percentage points from 23 to 33 percent, implying an increase in quasi-money by about 20 percent of GDP. This increase in monetization has generally mirrored the increase in the role of the nonagricultural sector, and, by definition, implies a steady decline in the velocity of broad money—a trend that is particularly evident since the late 1970s. Finally, the steady increase in the ratio of augmented broad money to GDP argues for paying more attention to this series as a more representative measure of money balances.32

CHART IV.5
CHART IV.5

SRI LANKA: TRENDS IN MONETIZATION, 1960–96

(In percent of GDP)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.

C. Issues in the Specification and Estimation of Money Demand

Model specification

136. To capture the many roles played by money in an economic system—facilitating transactions and providing a store of value—the money demand equation is usually written as:

M/P = f(Y/P, Io, Ia)

where, M represents the monetary aggregate in question, P the price level, Y a nominal scale variable, usually income or GDP, I0 is the “own” rate of return on the aggregate, and Ia is a nominal interest rate representing the opportunity cost of holding money. Alternatively, the money demand function can be written in log-linear form as:

ln (M/P) = α + β ln(Y/P) + γ Io + δIa

137. With the a priori assumption of the absence of money illusion, typically the restriction is imposed that money is homogenous of degree 1 in prices, and money demand functions are estimated in real terms. This restriction is empirically tested prior to estimation.

138. Most studies assume that aggregate income or output is an adequate proxy for the size of transactions demand. However, recent studies (e.g., Arrau et al., 1991) point out that the appropriate scale variables and related income elasticities of money demand may be different across different sectors of the economy. In the case of Sri Lanka, it is possible that the agricultural sector may have a different demand for money than the nonagricultural sector. For this reason, the share of nonagricultural income in GDP (Yna/Y) is included as an explanatory variable in the money demand function.33

139. Another point that is made by Arrau et al. (1991) is that most models of money demand in developing countries tend to perform poorly because they do not explicitly model the process of financial innovation and deepening. Arrau et al. attempt to explicitly examine the role of financial reforms and innovations in developing countries by including a stochastic trend that proxies shifts in transactions technology. In the specification used here, if indeed the agricultural sector tends to be less monetized than the rest of the economy, inclusion of the share of the agricultural sector in the economy should help to overcome some of these biases.

140. Finally, there is the question of which interest rates to use for I0 and Ia. Given the lack of variability in deposit rates, it is unlikely that their inclusion as measures of I0 will yield statistically meaningful results. In this paper, the three-month treasury bill interest rate is used to measure Ia. In addition, the inflation rate (π), measured either by the rate of change in the CPI or in the GDP deflator is also included as another opportunity cost variable. One reason for considering an alternative to the CPI is the severe measurement problems in the CPI in Sri Lanka.34 However, in order for the results derived here to have some operational relevance, the CPI is also used.

141. The specification that is used in the estimation is therefore given by:

ln (M/P) = α + β ln (Y/P) + βλ ln(Yna/Y) + γ π + δ Ia

If λ = 0, the specification resembles a conventional money demand function, and if λ=1 so that the coefficients on the second and third terms in the above equation are equal, it implies that real nonagricultural income is a better approximation of transactions demand for money. Values of λ between 0 and 1 reflect partial monetization of the agricultural sector.

Estimation procedure and main results

142. Results of previous studies. Before turning to the discussion of the main results of the present analysis, a brief review of the results of earlier studies is in order. Table IV.1 summarizes the main results of these studies. An early study of monetary policy in Asia is contained in Aghevli et al. (1979). Using a partial adjustment specification of money demand, they found that the income elasticity of narrow money demand is 1.08, while that of broad money is 1.48. Money demand was also found to be sensitive to the inflation rate. The period covered by their study 1950–1977, however, preceded the major structural reforms that began in 1977.

Table IV.1.

Results of Previous Money Demand Studies

article image

143. A more recent multi-country study was conducted by Tseng and Corker (1991) using data from the post-liberalization period 1978 to 1989 for Sri Lanka. Using a cointegration and error-correction framework as proposed by Engle and Granger (1987), they find that there is no evidence of a cointegrating relationship in the demand for real narrow money but that there is weak evidence of cointegration in the demand for real broad money. The estimated income elasticities are 0.92 for narrow money and 1.22 for broad money.

144. Rose (1993) re-examined the issue using data covering the period 1950 to 1991 and finds that a significant structural shift took place in money demand in Sri Lanka after the implementation of the 1977 reforms. A stable relationship between the broad money aggregates (both including and excluding foreign currency deposits) and real activity and prices is found for the post-liberalization period. The income elasticity of demand for broad money is found to be about 1.15. By contrast, the demand for narrow money which was relatively stable during the 1950–76 period, breaks down in the period since 1977. This finding is consistent with that of Tseng and Corker that no cointegrating relationship exists for narrow money and its determinants.

145. Data. In the present analysis, annual data on narrow money (M1), broad money (M2), and augmented broad money, including foreign currency deposits (M2+) are used as the three monetary aggregates. The scale variable is proxied by GDP at market prices in constant 1982 prices. The opportunity cost variable is the three-month treasury bill rate. In addition, two measures of inflation are used, one based on the CPI and the other on the GDP deflator. The ratio of nonagricultural output to GDP is used to proxy any difference in money demand behavior between the agricultural and nonagricultural sectors of the economy. The available annual data span the period 1960–96 for all the variables of interest, with the exception of M2+, for which data are only available from 1980.35

146. Estimation strategy. The general estimation strategy is to examine cointegration between the theoretical determinants of money demand as evidence of stability in the demand functions for money. (Annex I contains a more detailed outline of the strategy and the various approaches that are used in the estimation.) Once such a relationship is found, the short-run dynamic behavior of money demand is examined through the estimation of an error-correction model. Parameter stability and out-of-sample forecasts are examined to assess the performance of the estimated money demand equations.

147. Are the individual time series stationary? First, the order of integration of all the series of interest is examined using unit root tests, which, generally speaking, involve testing the hypothesis that α = 1 in a regression of the type yt = αyt-1 + εr. A variety of tests have been developed to test for unit roots in economic time series, of which the most widely used are Dickey-Fuller and augmented Dickey-Fuller tests. These tests are conducted using the following regression:

Δyt = constant + δt + βyt-1 + γ1 Δyt-1 + … + γn Δyt-n + εt

where the null hypothesis of interest is that β=0 or that the process yt has a unit root. The lagged dependent variables are added essentially to ensure that the residuals are white noise. The constant term should be present if the series is believed to have a non-zero mean and the linear trend term should be present if the series is believed to have a discernible trend over time. The results of these tests conducted using data from 1960–96 are shown in Table IV.2 and suggest that the null hypothesis that the variables in levels have a unit root cannot be rejected, while the null hypothesis that their first differences have a unit root is rejected. That is, all the variables of interest in this analysis are found to be I(1).36

Table IV.2.

Unit Root Tests 1/

(ADF tests statistics, regressions include a constant, trend and one lag of the dependent variable)

article image

Critical values of the test statistic are -3.516 at the 5 percent significance level and -4.184 at the 1 percent significance level. Values below these critical values (in absolute terms) imply non-rejection of the null hypothesis that there is a unit root in the series being tested. Values above these critical values imply rejection of the null hypothesis that there is a unit root. ** denotes significance at the 1 percent level, and * stands for significance at the 5 percent level.

148. Unit root tests were also conducted for these variables over the period 1978–96, but the shortness of the sample period greatly reduces the power of the tests, which were unable to reject the hypothesis that a unit root was present even in the first differenced series. This is not a surprising result given that in small samples, these tests are known to be biased toward finding unit roots. Nevertheless, in the rest of the paper, we proceed on the assumption that the variables of interest are indeed I(1) based on the tests for the whole sample period and on Charts IV.6 to IV.8, from which it would appear that the variables of interest are likely to be stationary in their first differences. Furthermore, in all cases where the hypothesis of unit roots in the first differenced series could not be rejected, the coefficient being tested was less than 0.5, arguably closer to 0 than to 1.

CHART IV.6
CHART IV.6

SRI LANKA: GROWTH IN MONETARY AGGREGATES, 1960–96

(In percent)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.1/ Including foreign currency deposits.
CHART IV.7
CHART IV.7

SRI LANKA: INFLATION TRENDS, 1960–96

(Annual average percent change)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.
CHART IV.8
CHART IV.8

SRI LANKA: OUTPUT GROWTH, 1960–96

(In percent)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.

149. Is money homogenous of degree 1 in prices? Prior to estimating real money demand equations, the hypothesis of homogeneity in prices was tested.37 When the entire sample from 1960 to 1996 is used, the data support the hypothesis of price homogeneity. Specifically, two first-order autoregressive distributed lag equations are estimated with narrow money and broad money as the two dependent variables, and real GDP, prices (measured by the GDP deflator), and the interest rate on three-month treasury bills as the right-hand side variables. The static long-run solution is then derived for the money equation and the results are shown in Table IV.3. The coefficient on the price variable in these equations suggests that long-run price homogeneity does hold over the longer sample period. However, these results do not hold up when the sample is restricted to the period since 1978. Among other things, this suggests that structural changes in the economy have altered the relationship between nominal money balances and prices. Because the results of these tests are inconclusive, the demand for both real and nominal money balances is examined in the cointegration analysis below.

Table IV.3.

Testing for Price Homogeneity

(Annual data, t-statistics in parentheses)

article image

150. Is there evidence of cointegration? We turn next to an examination of cointegration between the variables of interest. Finding a cointegrating relationship would suggest the existence of a stable demand for the particular monetary aggregate. As a first step, the relevant data are plotted in Charts IV.9 to IV.11. Chart IV.9 depicts movements in M1, M2 and M2+, together with those in the CPI and the GDP deflator, Chart IV.10 plots movements in each monetary aggregate and real GDP and Chart IV.11 shows movements in inverse velocity and the treasury bill rates. These charts all indicate that there are close co-movements amongst the traditional determinants of money demand identified in the literature, and suggest that they may be cointegrated.

CHART IV.9
CHART IV.9

SRI LANKA: MONEY AND PRICES, 1960–96

(In billions of rupees; log scale)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.
CHART IV.10
CHART IV.10

SRI LANKA: MONEY AND OUTPUT, 1960–96

(In billions of rupees; log scale)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.
CHART IV.11
CHART IV.11

SRI LANKA: VELOCITY AND INTEREST RATES, 1960–96

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.1/ Velocity is defined as the ratio of nominal GDP to the average of beginning- and end-period money stocks.

151. Next, the Johansen multivariate analysis of cointegration was applied for the period since 1978. The tests for the existence of a cointegrating relationship were rejected both for nominal and real M1, and for nominal and real M2+ balances. No further analyses of these variables was therefore attempted. There was, however, weak evidence of a stable demand for real broad money balances, but not for nominal broad money balances.38 However, in view of the short sample (19 observations since 1978) and the fact that the Johansen procedure requires estimation of a vector autoregression (VAR), the number of degrees of freedom is too small, even with the shortest lag structure (namely, one lag each of all the variables in the system), for meaningful statistical inference.39

152. The next step, therefore, was to use a single-equation two-step estimation procedure as originally proposed by Engle and Granger. For this exercise, two measures of real broad money balances are used—one using the CPI and the second using the GDP deflator as the deflator for nominal broad money balances. The results of this procedure are shown in Table IV.4. In the final analysis, the choice between the two will depend on the relative performance of each equation in out-of-sample forecasts.

Table IV.4.

Real Broad Money Demand 1/

(Annual data from 1978 to 1996)

article image

t-statistics based on heteroskedastic-consistent standard errors are shown in parenthesis beneath the coefficient estimates. Figures in parentheses next to the diagnostic test statistics refer to the probability of accepting the null hypothesis in each case. ** denotes significance at the 1 percent level and * denotes significance at the 5 percent level.

153. The ratio of nonagricultural sector to total GDP proved not to be significant in any specification and therefore was dropped from the cointegrating relationship. Likewise, the treasury bill rate was insignificant and was also dropped. The final first stage equation that was estimated implies an income elasticity of real broad money ranging from 1.12 to 1.28.40 However, most of the regression diagnostics shown in the table suggest that the equations are relatively poorly specified. Recursive estimates of the parameters of the long-run money demand functions were also computed by first estimating the equation over a truncated sample and then re-estimating it adding one observation at a time. If the estimated parameters are reasonably stable, then the confidence intervals around the point estimates should narrow as more and more observations are added. The results of this recursive estimation are shown in Charts IV.12 and IV.13. They suggest that the income elasticity of money demand from either specification above is reasonably stable and accurately estimated, but that the coefficients on the two inflation variables are less accurately estimated as the confidence bands do not narrow noticeably as more observations are added. The presence of a structural break in 1991—when interest rates moved closer to being market determined—was tested using a Chow test. No strong evidence was found of a structural break in 1991.

CHART IV.12
CHART IV.12

SRI LANKA: RECURSIVE PARAMETER ESTIMATES, 1988–96

(Demand for M2/CPI)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

CHART IV.13
CHART IV.13

SRI LANKA: RECURSIVE PARAMETER ESTIMATES, 1988–96

(Demand for M2/GDP deflator)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

154. Short-run dynamics. The next step was to examine the short-run dynamic behavior of money demand in response to changes in the theoretical determinants of money demand. The specification search began with an unrestricted autoregressive distributed lag model taking the following form:41

Δ(M2/P)t = α + ΣiβiΔ(RGDP)t-i + ΣiγiΔ(INFL)t-i + Σiδi(TBRATE)t-i + φECt-1 + ε1

155. The specification of the short-run dynamic model as an error-correction equation provides an additional test of the existence of a stable long-run relationship. A significant coefficient between 0 and -1 on ECt-1 (the term that measures the deviation of the variable in any given time period from its equilibrium value) suggests that there exists a long-run equilibrium to which the process tends to return.

156. A general-to-specific (or testing-down) model selection technique was used to obtain the second-stage error correction model. The final specifications are given by:

Δ(logM2/CPI)t=α+βΔlogRGDPt−1+γΔlog+vtCPIt+δTBRATEt−1+ηECCPIt−1[A]

and

Δ(logM2/DEFL)t=α+βΔlogRGDPt−1+γΔlogDEFLt++μECDEFLt−1+ξtδ1TBRATEt+δ2TBRATEt−1[B]

157. The results of this procedure are shown in Table IV.5. In all but one case, the coefficients have the expected sign. Specifically, in both equations, lagged real GDP growth has a positive effect on growth in real money balances, while the current period change in the inflation rate has a negative effect. Lagged values of the opportunity cost variable—the treasury bill rate—have the expected negative effect on real broad money balances. However, in the equation B, where money balances are deflated using the GDP deflator, the contemporaneous value of the treasury bill rate enters the equation with a counterintuitive positive and significant coefficient, implying that an increase in today’s treasury bill rate would be associated with an acceleration of real broad money growth. The contemporaneous interest rate term does not enter equation A, where money balances are deflated using the CPI. In both cases, the coefficient on the error correction term is about -0.4 suggesting that two thirds of the adjustment to the equilibrium path of money demand is completed in about 2 years. Overall, the equations had satisfactory statistical properties—with the possible exception of first order serial correlation in equation B suggesting misspecification—and passed formal tests of parameter constancy over the estimation period.

Table IV.5.

Short-Run Dynamics of Money Demand 1/

(Annual data, 1978–96)

article image

Absolute values of t-statistics based on heteroskedastic-consistent standard errors are shown in parenthesis next to the coefficient estimates. Figures in parentheses next to the diagnostic test statistics refer to the probability of accepting the null hypothesis in each case.

** denotes significance at the 1 percent level and * denotes significance at the 5 percent level.

158. Chart IV.14 plots actual and fitted values of the change in the log of the two measures of real money balances given by equations A and B above. They suggest that the fitted equations capture the turning points in the growth in real monetary aggregates quite well until the early 1990s. But, notwithstanding the results of the Chow test that does not detect a structural break in 1991, the relatively weaker performance of these equations in the early 1990s suggests that there may be changes on-going in the financial sector that could weaken the estimated relationship. This would strengthen the case (as discussed in the next two sections of the paper) for broadening the array of indicators that should be examined when evaluating and formulating monetary policy.

CHART IV.14
CHART IV.14

SRI LANKA: ACTUAL AND FITTED REAL MONEY BALANCES, 1979–96

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

159. Out-of-sample forecasts are also estimated and the forecast performance of the two equations is shown in Chart IV.15. The forecasting performance of these equations is evaluated in two ways. First, the models’ ability to detect movements in real broad money growth in the early 1990s is examined. The equation for real money balances deflated by the CPI tends to perform slightly better in tracking the turning points in the series. Second, Table IV.6 reports several objective criteria to evaluate forecast performance and from these, it can be seen that the equation based on the CPI outperforms the one based on the GDP deflator.

Table IV.6.

Comparison of Forecast Performance

article image
CHART IV.15
CHART IV.15

SRI LANKA: REAL MONEY BALANCES: OUT-OF-SAMPLE FORECASTS, 1990–96

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

160. Summary of results. Overall, the various estimation strategies lead to the conclusion that there is only weak evidence of a stable demand for broad money in Sri Lanka. At the same time, however, there is no evidence of highly unstable money demand. Thus, the lack of durable inflation control in Sri Lanka over the past two decades would appear to be more the result of overly expansionary monetary policies fueled by fiscal imbalances, rather than to the monetary policy being complicated by volatile and unpredictable money demand. Notably, paremeter instability in the dynamic relationships and a weakening of the fit in the short-term equation in the 1990s suggest that targeting broad money is likely to become increasingly difficult because of changes taking place in the economy.

161. Stable and predictable relationships could not be found for the narrow money aggregate and the measure of broad money augmented to include foreign currency deposits. While it can be expected that in the process of economic and financial development, the narrow money aggregate would become increasingly irrelevant as a predictor of inflation or real activity, the same cannot be said of the augmented broad money aggregate. Foreign currency deposits are a rapidly growing component of the total deposit base in Sri Lanka and as such, from a forward-looking policy standpoint, the behavior of augmented broad money including foreign currency deposits should be kept under review.

162. Treasury bill interest rates were found not to enter the long-run cointegrating relationship—in part because they have not been entirely sensitive to market forces for the entire sample period—but were found to be significant in explaining the short-run behavior of money demand. This suggests the potential for treasury bill rates to be used as both an instrument and an intermediate target of monetary policy.

D. The Conduct of Monetary Policy in Practice

163. Although it may seem that monetary aggregates play an overriding role in the conduct of monetary policy by the Central Bank of Sri Lanka (CBSL), in actual practice the approach is much more eclectic. In September of every year, the CBSL prepares a comprehensive pre-budget report that includes a monetary survey for the following year, intended as an input into the budget formulation process. The forecasted monetary survey is consistent with the official forecasts for output and the balance of payments, the inflation objective, and the likely budget deficit. The forecasts for broad money are based on the simplifying assumption that income velocity will be broadly stable over the coming year (i.e., an income elasticity of one), which one could argue is conservative given the trend decline in velocity that has occurred. Adjustments are made if the residual room for growth of credit to the private sector seems unreasonable in light of its historical relationship to investment and economic growth. Following the decisions on the budget, the broad money forecasts are revised, if necessary.

164. The CBSL does not publicly announce its monetary and inflation targets. The most explicit formulations of an official monetary target are usually contained in the Budget Speech. For example, the 1997 Budget Speech notes “the urgent need to reduce inflation on a sustainable basis to a single digit level” and expresses the expectation that the Colombo CPI will “show a moderate inflation below 10 percent” by the end of 1997.42 The Ministry of Finance also includes summary indicators of monetary and inflation targets in its publication “The Budget at a Glance”. Nonetheless, it is fair to say that the government’s precise inflation objective is not given much prominence.

165. The forecasted monetary survey is reviewed by the Monetary Board, the Bank’s highest decision making body.43 Throughout the year, the Monetary Board is informed regularly on monetary developments. If monthly inflation exceeds 10 percent, or monetary aggregates grow more than 15 percent in one month, the Monetary Board is formally required, under the Monetary Law Act, to submit a detailed report to the Minister of Finance.

166. For the conduct of monetary policy, the CBSL relies heavily on open-market type operations in treasury bills by means of which it controls the supply of reserves to the banking system (Box IV.1). At the same time it creates a large demand for reserves through its policy of high statutory reserve requirements.44 Interest rate determination is free of direct controls but the government continues to exert a strong influence, through the treasury bill auctions and through the National Savings Bank and the state-owned commercial banks.45 From the forecasted monetary survey, the CBSL derives a monthly reserve money program that takes account of seasonal patterns in government cash management, the balance of payments and currency circulation. Next, this is interpolated into weekly targets which guide the Bank’s operations in the primary market for treasury bills. The CBSL affects the level of reserve money through the amounts of treasury bills its takes into its own portfolio at the weekly auctions, and the conditions it sets for its “secondary window” where these bills are subsequently offered for sale.

167. The main sources of reserve money creation are the accumulation of foreign reserves and the extension of credit to government (Table IV.7). Large injections from these sources have been responsible for high rates of reserve money growth in recent years, feeding into excessive broad money growth and high rates of inflation. To insulate monetary policy better against developments in the balance of payments, the CBSL would have to be prepared to let the exchange rate float more freely. As to its financing of the budget deficit, the CBSL has only limited flexibility and independence.

Table IV.7.

Sri Lanka: Contributions to Reserve Money Growth, 1992–97

article image
Source: Central Bank of Sri Lanka.

Central Bank of Sri Lanka-Transactions in Treasury bills

Primary market:

On behalf of the Government, the CBSL organizes weekly auctions of treasury bills (and, since 1997, of treasury bonds). The discount bills have maturities of 3, 6, and 12 months, and are offered solely to 23 “accredited primary dealers,” after first having satisfied the requests from the provident funds and the National Savings Bank on a noncompetitive basis.

CBSL secondary window:

Despite the fact that primary dealers are obliged to “make the market” in treasury bills by giving two-way quotes on request, most secondary trading is with the CBSL. Through its secondary window, the CBSL stands ready to sell (“discount”) and buy (“rediscount”) treasury bills. The transactions are at a penalty interest rates which are usually set once a week in reference to the relevant rates in the primary market, with a varying spread varied depending on whether the CBSL wants to encourage or discourage transactions (e.g., it was widened from 1.5 in 1995 to 3 percent in March 1996 to discourage the banks from offering them for rediscount, and then narrowed again to 1.8 percent in December 1996).

Repurchase facilities:

To be able to inject and absorb liquidity for any maturity and still have full collateral, the CBSL introduced repurchase and reverse repurchase agreements.

Repurchase agreements (selling treasury bills with the simultaneous agreement to repurchase them) were introduced in October 1993 in order to stabilize the lower end of the call market. Giving banks an opportunity to invest surplus funds, it effectively provides a floor to interest rates. Access is at the discretion of the banks but limited by the size of the CBSL’s portfolio of treasury bills; the CBSL cannot for legal reasons do repos in its own securities. Repos are currently available with maturities up to 91 days, but in practice are mostly overnight.

In November 1995, when interbank call market rates were excessively high at times, the CBSL introduced the telephonic auctioning of reverse repurchase agreements (purchasing treasury bills with the agreement to resell) to inject liquidity. Shortly thereafter, banking system liquidity unproved and the facility was suspended.

Proposed reforms:

The CBSL is planning reforms that would encourage the development of a secondary market outside of its own window. It is also considering abolishing the standing repo facility in favor of frequent open market operations, with repos and reverse repos, aimed a stabilizing the call market rate withing a target range.

168. The CBSL provides credit to the government through two channels. First, it grants the government an interest free overdraft facility of up to 10 percent of estimated revenues of the current year. In practice, the government tends to always use this facility up the limit, and although these loans are called “provisional advances,” they are in fact not repaid. Secondly, due to the expansion of the market for treasury bills, the CBSL holdings of bills are now much reduced.46 Nevertheless, the CBSL is still regularly forced to purchase unsold bills in the auctions whenever the cut-off interest rate is set below market clearing levels. The only protection for the CBSL in this respect is the subceiling on treasury bills that Parliament imposes as part of the public borrowing program at the time the budget is approved.47

169. The CBSL has established reasonably firm control over another potentially serious source of reserve money, namely lending to the banking system. Critical in this regard was the phasing out of the various short and long-term refinance facilities, to banks and other entities, that were part of the old, developmentally-oriented system of directed credits.48 The CBSL lender-of-last-resort facility, available freely to banks albeit at the penal “bank rate,” has been made so odious that the banks are reluctant to use it.49 While the repo facility, through which the banks can freely deposit surplus funds, is another weakness in control, by varying the interest rate the CBSL can limit access. Finally, since the CBSL has the option of issuing its own securities, it can in principle absorb any amount of reserve money, even after it has exhausted its portfolio of treasury bills.

170. In practice, the CBSL applies a great deal of judgment in deciding whether to offset deviations from the forecasted path for reserve money. If it decides not to, the reserve money path for the remainder of the year is rebased. For instance, the CBSL responded to surges in capital inflows over the period 1991–95 by regularly allowing reserve money to grow much faster than its forecasted path, and in 1966, it allowed reserve money to grow much slower, this time in response to an apparent downward shift in credit supply. In deciding whether to offset deviations or not, the Bank looks at a number of indicators, such as interest rates (call market, 3 months treasury bills, weighted average prime lending rate); private sector credit growth (comprehensively available with a 4–6 weeks lag, but in approximate form on a weekly basis); real economy indicators; imports; exports; the exchange rate; and inflation.

E. Alternative Monetary Frameworks

171. Before discussing alternative monetary frameworks, it should be emphasized that monetary policy has only limited effectiveness unless supported by fiscal policy. Reducing inflation in Sri Lanka to the low single digits while minimizing output loss will depend critically upon a further reduction in the fiscal deficit. Fiscal consolidation would allow the phasing out of CBSL financing of the deficit, thus helping to lower the growth of monetary aggregates. Furthermore, deficit reduction would also lower the government’s recourse to the financial system other than the CBSL, thus expanding the room for credit growth to the private sector. Without fiscal consolidation, disinflation policies inevitably put greater upward pressure on interest rates.

172. It is clear from the empirical analysis that the money demand estimates are not sufficiently stable for the CBSL to derive precise monetary targets.50 Data limitations are an obvious problem, and it is therefore not possible to state with certainty that the underlying relationship is indeed unstable. In any event, the implication is that it is not feasible for the CBSL to frame its monetary policy around broad money as an intermediate target. The attraction of such an intermediate target would have been twofold: it would have furnished the CBSL with a simple contemporaneous indicator to guide its monetary policy actions, and it would have furnished the public a clear yardstick by which to judge the CBSL’s performance. Monetary policy credibility would benefit greatly from the transparency and accountability inherent in such a framework. The result would have been greater monetary policy effectiveness and reduced costs of disinflation. The challenge is for the CBSL to develop an alternative framework that approximates these features.

173. The obvious alternative to a monetary intermediate target is some kind of fixed exchange rate arrangement, but this would not be the preferred approach for Sri Lanka. Presently the exchange rate of the Sri Lanka rupee floats within a sliding band set by the CBSL. In operational terms, the CBSL sets its middle rate on a daily basis, and is prepared to intervene to keep the market rate within 1 percent on either side of this rate. The daily setting of the CBSL rate is generally in line with market pressures, and also takes into account various indicators of competitiveness, while preventing sharp variations. This flexible arrangement seems to have served Sri Lanka well. Moreover, the experience of other countries in the region shows the advantages of exchange rate flexibility to insulate domestic monetary policy in the presence of large capital flows as Sri Lanka hopes to attract in the future. In addition, Sri Lanka’s inflation rate (about 8 percent on a 12-month basis in May 1997) may already be out of the relatively high range where countries typically benefit the most from fixed exchange rate arrangements.

174. Another alternative is inflation targeting, an approach that is rapidly gaining followers among central banks around the world. Pioneered by New Zealand in 1989, the approach has been adopted in Canada, Sweden, and the United Kingdom, among others. 51 The precise modalities differ between countries, but New Zealand can serve as a model of the approach in its most rigorous form. Inflation targeting in New Zealand encompasses three features: (i) the central bank’s sole monetary objective is price stability and it has instrument independence to go with that; (ii) the central bank governor is subject to a performance contract that requires inflation to be kept within a narrow, precisely defined target range; and (iii) monetary policy is based on inflation forecasts. The inflation forecasts relative to the target are in a sense the contemporaneous intermediate target, analogous to broad money, that guide monetary policy. The central bank of New Zealand publishes quarterly inflation forecasts for two to three years in the Monetary Policy Statements it is required to file twice a year with Parliament, and updated forecasts in its Economic Projections published in the intervening quarters.52 53

175. In the inflation targeting approach, much hinges on the quality of the inflation forecasts, the understanding of the monetary transmission process, and the effectiveness of the monetary instruments. A central bank can aim for a fairly narrow inflation target range (or, an target point with a narrow confidence interval) provided it can forecast inflation accurately, knows with some precision how and with what lags monetary conditions affect inflation, and has effective and efficient instruments to vary these conditions. The advantage of a narrow inflation target is that it gives a strong signal to inflationary expectations during the disinflation phase, and once price stability is achieved, that it minimizes the costs associated with variations in the inflation rate. On the other hand, if these conditions are not met, a narrow target would lead to frequent breaches, which would undermine the credibility of the inflation targeting framework.

176. It is unreasonable to expect central banks in a developing country such as Sri Lanka to meet these conditions to the same extent as central banks in industrial countries. In developing countries, accurate inflation forecasts will be hard to produce given the weak statistical database. Moreover, monetary transmission in those countries is much less predictable due to the underdeveloped and weak state of financial markets; and market-based monetary instruments are often still in their infancy. Finally, central banks in developing countries generally do not have the independence necessary to allow them to employ their monetary instruments solely in the pursuit of price stability. Hence, a rigorous approach to inflation targeting would not be feasible. Nonetheless, the gradual adoption of elements of the inflation targeting approach would seem a natural extension of the present eclectic approach to monetary policy by the CBSL.

177. Advances can be made on several fronts simultaneously. The CBSL could start by becoming more transparent about its medium-term inflation target, and the progress it has made towards its achievement. More frequent, policy-oriented publications or public pronouncements could be a vehicle. This would also help in fostering a constituency for low inflation, thereby improving the acceptance of the stance of monetary policy. Over time, it will be critical to grant the CBSL a greater degree of independence vis-a-vis the government. This would involve a sharpening of the statutory mandate to give clear priority to price stability, and changes in the composition of the Monetary Board to emphasize the CBSL’s independence from the government. Independence also means separation of debt management and monetary policy. The market for government securities has already developed to the point where the government can dispense with borrowing from the CBSL. In that case, the CBSL would be buying (and selling) treasury bills only for monetary policy purposes, and would be able to exert full control over short-term interest rates.

178. Equally important are improvements to the various indicators that are critical to the conduct of monetary policy. Three categories of indicators can be distinguished; those that measure inflation; those that are helpful in forecasting inflation; and those that are needed to assess the stance of monetary policy in achieving the inflation objective.

179. The foremost candidate for improvement is the inflation measure. Inflation is presently measured with the Colombo Consumer Price Index (CCPI), which is based on an outdated basket (from 1952) which is dominated by staple foods (61 percent weight). The food component is responsible for high variability, due to periodic droughts, and strong seasonality, around the two harvest periods; regular changes in administered prices add further noise (Chart IV.16). Consequently, the CBSL tends to focus on the annual average rate of change in this index but covering data as far back as 24 months, this is a poor indicator of current inflationary trends. A revision of the CCPI basket, based on a new household survey, is in progress. Nonetheless, the weight for food is likely to continue to be high, and thus the current problems in using the CCPI for monetary policy will remain. Therefore, in conjunction with the CCPI revision, the construction of an indicator of “underlying inflation” would need to be explored. To best capture the trend in the general price level, reflecting the balance between aggregate supply and demand in the economy, this measure should seek to identify and remove temporary variations in prices.

CHART IV.16
CHART IV.16

SRI LANKA: COLOMBO CPI INFLATION, 1994–97

(In percent)

Citation: IMF Staff Country Reports 1997, 095; 10.5089/9781451823370.002.A004

Source: Data provided by the Sri Lanka authorities.

180. Monetary aggregates, while not sufficiently stable to serve as intermediate targets, nonetheless can have a predictable relationship to future inflation and should be carefully researched for any such information. In addition there are many other indicators that may shed light on future inflation. Sri Lanka already avails over a wide array of them, and the challenge is to formally incorporate them into a regular analysis of inflationary developments. In some cases, improvements can be made. For instance, an industrial production index would usefully supplement the available recent information on the state of the economy, and would make up somewhat for the absence of quarterly GDP data. As the private sector is given more information to assess the implications of monetary and fiscal policies for inflation, asset prices (foreign exchange, bonds, equity, real estate) will become useful indicators of inflationary expectations. A technique widely used in developed countries, but possibly also with applicability in developing countries, is the output gap approach which assumes that inflation is determined by either the gap between potential and actual output or by the change in that gap.54

181. Finally, indicators need to be developed that can help in assessing the stance of monetary policy. The obvious choice for a benchmark interest rate is the three-month treasury bill rate, as this is a risk free security, auctioned under more or less free conditions, and for which there is a reasonably deep market. Unfortunately, the rate is subject to substantial variations, largely due to government cash management, and therefore hard to interpret. Notably, in the beginning of the year, when the limit on provisional advances is raised, the government uses that as an opportunity to substitute this interest free credit for treasury bills, and the rate shows a steep decline. A more transparent interest rate could be combined with the exchange rate into a monetary conditions index which, because of the close connection between the two rates in a small open economy, may be better indicator of the monetary stance then either of them separately.55

ANNEX

Estimation methodology

The most common conventional specification of money demand functions was the partial adjustment model, which would then be estimated using OLS. However, as argued by several authors (e.g., Granger and Newbold, 1974, Phillips, 1986, and Stock and Watson, 1988), OLS estimators are not consistent in the presence of nonstationary time series. Thus, although OLS estimation results have high R2s and significant t-statistics, inference based on the traditional test statistics may not be correct—the so-called spurious regression problem. These findings suggested that stationarity of economic time series should be tested for and stationarity induced before performing regression analysis. Econometricians (notably Box and Jenkins, 1970, and Granger and Newbold, 1974) suggested that the relevant time series should be made stationary by differencing, usually once or twice, and regressions should be run using these differenced time series. This method, however, suffers from the problem that information about the long-run relationship between the variables is discarded, since in the long run, differenced values of these variables are zero.

Outline of the Engle-Granger approach

The development of the concept of cointegration suggested a way to improve this method. Granger (1981) first showed that there is a way to link integrated processes and the long-run steady-state equilibrium.56 Simply put, the idea behind the concept of cointegration is that even though level variables are individually I (1), special linear combinations of these I (1) variables can be I (0).57 In this case, the long-run components of these series cancel each other out to produce a stationary series and are said to be cointegrated. The vector of coefficients of the linear combination is called the cointegrating vector. Initially, studies of cointegration used single equation estimation techniques such as the one suggested by Engle and Granger in which OLS is used to estimate the parameters of the cointegrating vector from non-stationary variables. Stock (1987) and Stock and Watson (1988) show that when a set of variables is cointegrated, the OLS estimator of the cointegrating vector provides a “superconsistent” estimator of the true vector, in the sense that the estimators converge to the true parameters at a much faster rate than if they were not cointegrated.

The major drawback of the Engle-Granger methodology is that, in a multivariate model, there can exist more than one cointegrating relationship between the variables because it is theoretically possible to have several equilibrium relationships between a set of variables. To overcome this shortcoming of the Engle-Granger methodology, a maximum likelihood estimation procedure was developed by Johansen (1988) to estimate the cointegrating relationship between a set of variables, and has gained a great deal of popularity in recent years in time series studies. The Johansen technique fully captures the long-run relationships among the variables, provides a test of how many, if any, cointegrating vectors there exist between the variables of interest, provides estimates of all possible cointegrating vectors, and permits testing of hypothesis on these cointegrating vectors.

Outline of the Johansen approach

A brief discussion of the Johansen approach is as follows: Consider an unrestricted k dimensional vector autoregression (VAR), which includes j lags, given by:

Zt = ρ + Θ1 Zt-1 + Θ2 Zt-2 + … + Θj Zt-j + ζt

where, ζ1, … ζk are i.i.d N(0,Σ). On the assumption that all the variables in Zt are nonstationary and integrated of order 1, ΔZt will be stationary. The model above can therefore be rewritten as

ΔZt = Σi Γj-1 i=1 Δ Zt-i + Π Zt-j + εt

where Γi = -I + Θ1 + … + Θi

and Π = -(I + Θ1 - … - Θj)

In the long run, ΔZt is equal to 0; thus the equation ΠZ=0 incorporates the long-run relationship between the variables. Assume there are r cointegrating vectors among these variables, where 0 ≤ r ≤ k. In this case, Johansen shows that the matrix can be decomposed into two k by r matrices Π=αβ’. The matrix β contains the parameters of the cointegrating vector such that β’ Zt is stationary. The matrix a contains the weights with which the vectors enter the equations in the system. They can also be seen as adjustment coefficients measuring the feedback effects of lagged disequilibrium in the cointegrating relationship onto the variables in the VAR system.

Johansen (1988) has shown how to calculate a maximum likelihood estimator for α and β in multivariate models. He also presents two likelihood ratio tests of the hypothesis that there are almost r cointegrating relationships amongst the variables in the VAR. One test is based on the maximal eigenvalue of the stochastic matrix Π to test the null hypothesis that the number of cointegrating vectors is less than or equal to r against the alternative of r+1 vectors. The second test is based on the trace of the stochastic matrix II and tests the null hypothesis against the alternative that there are at least r+1 cointegrating vectors.

Error correction mechanisms

Another important implication (the so-called Granger Representation Theorem) of finding a cointegrating relationship between a set of variables is that it implies that there is some adjustment process—an error-correction mechanism—that prevents the deviations from the long-run relationship from becoming progressively larger. It is this mechanism that describes the short-run dynamics of the behavior of money demand out of equilibrium. The most common approach for estimating the error-correction model has been to treat it as the next step after estimating the cointegrating relationship. Thus, the long-run relationship is first estimated using either the single equation (OLS) or the multivariate (maximum-likelihood) approaches. The residual from this step is tested for stationarity—an essential condition for cointegration between the variables of interest. If it passes this test, a lagged value of the residual from the long-run equation can be included in the equation that attempts to describe the short-run dynamics of the system. Indeed, finding a statistically significant coefficient between -1 and 0 on the lagged residual provides another test of cointegration or stability of the long-run relationship, because it implies that there is a valid error-correction mechanism that brings the variable of interest back to its long-run equilibrium path.

References

  • Aghevli, Bijan B., Mohsin S. Khan, P.R. Narvekar, and Brock K. Short, “Monetary Policy in Selected Asian Countries,” IMF Staff Papers, Vol. 26, December 1979, pp. 775–824.

    • Search Google Scholar
    • Export Citation

  • Arrau, Patricio, Jose De Gregorio, Carmen Reinhart, and Peter Wickham, “The Demand for Money in Developing Countries: Assessing the Role of Financial Innovation,” IMF Working Paper, WP/91/45, May 1991

    • Search Google Scholar
    • Export Citation

  • Bernanke, Ben S., and Frederic Mishkin, “Inflation Targeting: A New Framework for Monetary Policy,” Journal of Economic Perspectives, Vol. 11, Number 2, Spring 1997, pp. 97–116.

    • Search Google Scholar
    • Export Citation

  • Boughton, James M. And George S. Tavlas: “What Have We Learned About Estimating the Demand for Money: A Multicountry Evaluation of Some New approaches,” IMF Working Paper, WP/91/16, February 1991

    • Search Google Scholar
    • Export Citation

  • Coe, David, and C. John McDermott, “Does the Gap Model Work in Asia?, Staff Papers, International Monetary Fund, Vol. 44 No. 1, March 1997, pp. 59–80.

    • Search Google Scholar
    • Export Citation

  • Cuthbertson, Keith, Stephen G. Hall and Mark P. Taylor: “Applied Econometric Techniques”, The University of Michigan Press, Ann Arbor, 1992

    • Search Google Scholar
    • Export Citation

  • Dekle, Robert, and Mahmood Pradhan, “Financial Liberalization and Money Demand in ASEAN Countries: Implications for Monetary Policy,” IMF Working Paper WP/97/36 International Monetary Fund, March 1997

    • Search Google Scholar
    • Export Citation

  • Engle, Robert F., and Clive W.J. Granger, “Cointegration and Error Correction: Representation, Estimation and Testing,” Econometrica Vol. 55, pp. 251–276, 1987

    • Search Google Scholar
    • Export Citation

  • Ericsson, Neil R., and Sunil Sharma, “Broad Money Demand and Financial Liberalization in Greece,” IMF Working Paper WP/96/62, International Monetary Fund, June 1996.

    • Search Google Scholar
    • Export Citation

  • Freedman, C., “The Role of Monetary Conditions and the Monetary Conditions Index in the Conduct of Policy,” Bank of Canada Review, Autumn 1995

    • Search Google Scholar
    • Export Citation

  • Johansen, Soren: “Statistical Analysis of Cointegrating Vectors,” Journal of Economic Dynamics and Control, Vol. 12, 1988, pp. 231–254

    • Search Google Scholar
    • Export Citation

  • MacKinnon, J.G., “Critical Values for Cointegration Tests,” in Long-Run Economic Relations—Readings in Cointegration, edited by Robert Engle and Clive Granger, Oxford University Press, New York, 1991

    • Search Google Scholar
    • Export Citation

  • Nelson, Charles and C. Plosser: “Trends and Random Walks in Macroeconomic Time Series: Some Evidence and Implications,” Journal of Monetary Economics, Vol. 10, (1982) pp. 139–62.

    • Search Google Scholar
    • Export Citation

  • Parker, Karen A: “The Demand for Money in India”, in India—Background Papers, IMF Staff Country Report No. 95/87, September 1995.

  • Rose, Marjorie B: “The Demand for Money and Financial Liberalization in Sri Lanka, Unpublished mimeograph, International Monetary Fund, February 1993.

    • Search Google Scholar
    • Export Citation

  • Stock, James H., and M.W. Watson: “Testing for Common Trends,” Journal of the American Statistical Association, Vol. 83, 1988.

  • Treichel, Volker, Broad Money Demand and Monetary Policy in Tunisia, IMF Working Paper WP/97/22, International Monetary Fund, March 1997.

    • Search Google Scholar
    • Export Citation

  • Tseng, Wanda, and Robert Corker: “Financial Liberalization, Money Demand and Monetary Policy in Asian Countries,” Occasional Paper No. 84, International Monetary Fund (Washington), July 1991.

    • Search Google Scholar
    • Export Citation
31

Prepared by Kalpana Kochhar and Anton Op de Beke, with assistance from Siddhartha Choudhury.

32

The Central Bank of Sri Lanka has recognized the growing importance of foreign currency deposits and are revising the monetary survey to include such deposits.

33

The specification here follows the approach taken in Parker (1995) to estimate the demand for money in India.

34

These deficiencies are documented in detail in SM/95/94—Sri Lanka: Background Papers and include being based on an outdated consumption basket and consumption weights, which results in a heavy weight on staple food items and in turn, makes the CPI unduly susceptible to seasonal factors.

35

The use of quarterly data, which are available for almost all the variables of interest, is precluded by the lack of a quarterly activity variable. While the use of an interpolated quarterly series is an option, it was not used here because first, it is not clear that the gain in statistical power from a large number of degrees of freedom would offset the potential biases that could be introduced by the use of constructed data and second, the operational relevance is not clear as the authorities do not have a quarterly real activity variable when formulating their monetary policy.

36

Critical values of the test statistics are from MacKinnon (1991).

37

Averages of beginning- and end-period stocks of all monetary aggregates were used in all the statistical analysis.

38

The trace test indicated the presence of a unique cointegrating vector in the systems using real broad money, while the eigenvalue test did not detect the presence of even one cointegrating vector.

39

Consideration was given to using the entire sample and just inserting a dummy variable for the post-1978 period. However, since the reforms of the late 1970 involved many fundamental changes to the determination of key prices in the economy, they are likely to have altered the underlying behavioral relationships in a more complex way than can be captured by a simple intercept dummy.

40

The coefficient estimates and standard errors have been “corrected” using the Engle-Yoo procedure or the so-called third step. Two disadvantages of the first-stage static cointegrating regression are (a) that it gives consistent but inefficient estimates of the cointegrating vector and (b) that the distribution of the estimators tends to be non-normal so that inference cannot be drawn about the significance of parameters. The Engle-Yoo third step of the regression provides a correction to the estimates which makes them asymptotically equivalent to the full-information maximum-likelihood estimators obtained using the Johansen procedure and provides a set of standard errors which permit a valid calculation of standard ‘t’ tests. For a fuller discussion of this computationally tractable technique, see Cuthbertson, Hall and Taylor (1992).

41

Error-correction models, apart from the ease of estimation, are intuitively appealing since they can be viewed as generalizations of earlier approaches to money demand estimation, including the partial adjustment model, the inventory adjustment model, etc. (See Boughton and Tavlas, 1991), and indeed from any specified autoregressive distributed lag model. In addition, the ECM formulation is appealing in that it immediately provides the parameter describing the rate of dynamic adjustment toward equilibrium. In so doing, it provides an indirect test of the existence of a stable long-run relationship.

42

See Budget Speech pp. 47 and 50. In the CBSL 1996 Annual Report, the Bank’s main publication, no reference is made to an explicit inflation target.

43

The Monetary Board is composed of three people: the Governor, who chairs; the Secretary to the Minister of Finance; and one member appointed by the President as a private sector representative. It meets every two weeks.

44

As of March 1997, reserve requirements are 12 percent for both demand and time deposits; they are unremunerated. The CBSL no longer treats the reserve requirement ratio as an active instrument but intends to gradually reduce the level so as to reduce this implicit tax on financial intermediation and reduce the spread between deposit and lending rates.

45

Until 1995, the NSB was subject to government directives in the rate it offered on its savings accounts, and this rate effectively placed a floor under deposit rates. The SCBs with their great need for provisioning against non-performing loans and large administrative overheads, are forced to charge commensurately high rates for their loans. Due to their dominating market share, this had tended to place a floor under lending rates.

46

End-1996 the CBSL held 14 percent of all treasury bills outstanding, against 70 percent end-1988.

47

This ceiling does not cover the recently introduced treasury bonds, but then the CBSL is not under any pressure to purchase those.

48

The outstanding stocks are gradually being repaid.

49

Borrowing has to be collateralized with government or private sector securities. The facility has not been accessed for 15 years and the rate has been unchanged at 17 percent since 1991.

50

Dekle and Pradhan (1997) reached a similar negative conclusion for Indonesia, Malaysia, Singapore, and Thailand. On the other hand, Ericsson and Sharma (1996) and Treichel (1997) do find stable money demand functions, for Greece and Tunisia, respectively.

51

See Bernanke and Mishkin (1997) for a discussion which concludes that inflation targeting, when construed as a framework rather than a rigid rule, has advantages, including more transparent and coherent policy making, increased accountability, and greater attention to long-run considerations in day-to-day policy debates and decisions.

52

See the Reserve Bank of New Zealand web site at http://www.rbnz.govt.nz.

53

Accountability is further ensured by the special reports the Minister of Finance can request in case of breaches of the target. Although legally the Governor can be dismissed in case of a breach, this is not what happened when in 1995 and 1996 inflation marginally exceeded the 0–2 percent target range. Instead, the Minister judged that the Governor had acted professionally and on the basis of the best information available to him.

54

Coe and McDermott (1997) conclude that the gap model works well in almost all Asian countries they studied.

55

Such an index was first used by the Bank of Canada, and is now also published by central banks in Sweden, Norway and New Zealand. For a description, see Freedman (1995)

56

Nelson and Plosser (1982) have shown that most economic time series are integrated of order 1, that is, their first differences are integrated of order zero, or are stationary.

57

The shorthand notation 1(d) refers to the order of integration of a particular time series. Specifically, d refers to the number of times the series must be differenced in order to produce a stationary time series.

Same Series

Other IMF Content

Other Publishers

Cover IMF Staff Country Reports
Sri Lanka: Selected Issues
Author: International Monetary Fund
Volume/Issue:
Volume 1997: Issue 095
Publisher:
International Monetary Fund
ISBN:
9781451823370
ISSN:
1934-7685
Pages:
180
DOI:
https://doi.org/10.5089/9781451823370.002

SRI LANKA: MONETARY AGGREGATES, 1960–96

(In billions of rupees, end of period)

SRI LANKA: PRICE DEVELOPMENTS, 1960–96

(1990=100)

SRI LANKA: INTERNET RATES, 1960–96 1/

SRI LANKA: INDICATORS OF ECONOMIC ACTIVITY, 1960–96

SRI LANKA: TRENDS IN MONETIZATION, 1960–96

(In percent of GDP)

SRI LANKA: GROWTH IN MONETARY AGGREGATES, 1960–96

(In percent)

SRI LANKA: INFLATION TRENDS, 1960–96

(Annual average percent change)

SRI LANKA: OUTPUT GROWTH, 1960–96

(In percent)

SRI LANKA: MONEY AND PRICES, 1960–96

(In billions of rupees; log scale)

SRI LANKA: MONEY AND OUTPUT, 1960–96

(In billions of rupees; log scale)

SRI LANKA: VELOCITY AND INTEREST RATES, 1960–96

SRI LANKA: RECURSIVE PARAMETER ESTIMATES, 1988–96

(Demand for M2/CPI)

SRI LANKA: RECURSIVE PARAMETER ESTIMATES, 1988–96

(Demand for M2/GDP deflator)

SRI LANKA: ACTUAL AND FITTED REAL MONEY BALANCES, 1979–96

SRI LANKA: REAL MONEY BALANCES: OUT-OF-SAMPLE FORECASTS, 1990–96

SRI LANKA: COLOMBO CPI INFLATION, 1994–97

(In percent)