Can money endogeneity be generalized? Empirical evidence from Caribbean economies

Can Money Endogeneity be Generalized? Empirical Evidence from Caribbean Economies KEVIN GREENIDGE, ROLAND CRAIGWELL, AND DARRIN DOWNES*

This study investigates the direction of causality between the money supply and reserve money for some selected Caribbean countries using recently developed techniques of causality tests. Thefindings suggest that neither money endogeneity nor money exogeneity can be generalized to all small, open economies. The causal patterns may differ according to whether the monetary arrangements of the countries follow either afixed orflexible exchange rate regime. Moreover, this study highlights the fact that a mixed bag of policies must be implemented by respective central banks to maintain macroeconomic stability. (JEL E50)

Introduction

In developed economies, monetary theorists assume a stable money multiplier relation, expressed as the ratio of an endogenous money supply and an exogenous monetary base. Under this proposition, the monetary authority exerts a high degree of control over the monetary base, and by predicting the value of the money multiplier, it manipulates the monetary base to achieve a desired level of the money supply. This theoretical view has been questioned for some time and has recently become the focal point of debate, especially among practitioners [Goodhart, 1984, 1994]. In light of this, Howells and Hussein [ 1998] investigates whether broad money is endogenous in the G7 countries (the U.S. United Kingdom, France, Germany, Japan, Italy, and Canada). Their empirical findings reveal that it is endogenous in all cases, that is, there is unidirectional causality from loans to deposits. In the context of small, open economies, not only is the stability of the money multiplier in question, but more fundamentally, it is argued that the monetary base is endogenous (see McClean [1981, 1985]). In recent studies, McClean [1997, 1998] examined the causal relationship between the narrow money supply (M1) and the monetary base for the small, open economy of Barbados. His empirics suggest that narrow money has a causal predominance over base money. This study builds on the efforts of the previously mentioned authors. In particular, it extends the research of McClean [1997, 1998] by providing evidence of the direction of causality between the money supply and reserve money for 10 fixed exchange rates and three flexible exchange rates for small, open Caribbean economies. The aim here is to compare these findings with those of Howells and Hussein [1998] on developed countries and to comment on the generalization of the endogeneity results to small, open economies. A"

*Central Bank of Barbados--West Indies. 253

254

IAER: MAY 2001, VOL. 7, NO. 2

Granger causality test, which is based on cointegration and error correction representations, is employed using a consistent set of quarterly data. This paper is presented as follows. The second section outlines the econometric methodology and describes the data sources, the third section interPrets the empirical results, and the fourth section concludes. Econometric Methodology and Data

A review of the plots of the raw data did not discern any significant seasonal patterns. Hence, as a first step, the data series for each country is tested for unit roots using the familiar augmented Dickey-Fuller (ADF) and Phillips-Perron tests. Next, the Johansen cointegration procedure was performed on the variables of interest. If the variables were found to be cointegrated, then a vector error correction model (VECM) was estimated. On the other hand, if no long-run relation existed between the variables, a vector autoregression (VAR) in first difference form was estimated. The Johansen approach is sensitive to the choice of lag length. Therefore, the optimum lag length was carefully determined by the Schwarz-Bayesian information criterion. The next step in the econometric analysis is to use the VECM and VAR to determine the direction of causality. The VECM suggests that Granger causality can be expressed in terms of the long run and the short run while the VAR is only concerned with Granger causality in the short run. For example, in the VAR, if the money supply is hypothesized to Granger cause the monetary base, then the coefficients on successive lagged values of the change in the money supply must be significant in forecasting future values of the change in the monetary base. In the VECM, not only is this short-run (change) effect possible but if the error (correction) term is significant and negative, this suggests Granger causality in at least one direction in the long run. In all cases, the zero restriction on the coefficients of the lagged change in the independent variables are performed using the Wald coefficient test. This study utilizes quarterly data which were taken from the statistical database of the International Monetary Fund. For Barbados, St. Lucia, St.Vincent, the Grenadines, and Grenada, the data set spanned the sample period of 1960:1 to 1998:3. In the case of Trinidad/ Tobago, Jamaica, and Guyana, their respective data sets covered those periods when their monetary arrangements were under a fixed exchange rate regime. This ensures consistency in the interPretation of the empirical results. The estimation periods of these countries were 1960:1 to 1993:1, 1960:1 to 1990:2, and 1960:1 to 1990:4 for Trinidad/Tobago, Jamaica, and Guyana, respectively. For curiosity's sake, these models were also estimated up to the 1998:3 period when the flexible exchange rate regime was in effect. The logs of the narrow money supply (LNM), broad money (LBM), and reserve money (base money) (LRM) were calculated and used as the main variables in the empirical analysis. Narrow money is currency plus demand deposits while broad money is narrow money plus saving and time deposits. All estimations were performed using the Eviews 3.1 [????] statistical software.

GREENIDGE ET AL.: MONEY ENDOGENEITY

255

Empirical Results

Since the underlying results for broad money and narrow money are similar, this discussion will focus mainly on the narrow money concept. Table 1 presents the results of the unit root tests. It reports that except for the LRM and the LNM for Grenada which are I(0), the LNM and the LBM are I(1). The Phillips-Perron tests and the plots of the correlograms of the variables in levels and first differences also confirm these findings) TABLE 1 ADF Unit Root Tests: 1960:1 to 1998:3

LRM

LNM

LBM

Levels

First Differences

Levels

First Differences

Levels

Antigua/Barbuda

-1.87

-8.88

-1.00

-8.21

-0.66

-8.743

Barbados

-0.52

-8.67

-0.92

-12.35

-1.65

-9.318

Dominica

-2.27

-8.16

-2.91

-7.33

-3.71

-6.706

Grenada

-5.21

-10.28

-3.41

-11.43

-2.36

-9.627

Guyana

-1.69

-7.82

-3.33

-3.78

-1.59

-6.06

Jamaica

-1.73

-8.24

-2.72

-3.90

-1.01

-7.52

St. Kitts/Nevis

-2.68

-10.66

-2.91

-10.72

-2.57

-9.73

St. Lucia

-1.38

-10.48

-1.38

-10.45

-0.73

-8.68

St. Vincent/ Grenadines

-1.29

-5.59

-3.22

-6.22

-2.53

-10.77

Trinidad/Tobago

-2.18

- 10.63

-2.14

-6.32

- 1.24

-3.45

Countries

First Differences

Notes: The ADF unit root test equationincludesboth drift and trend components. Source: For criticalvalues,see McKinnon[1991].

Starting with a maximum lag of 12 on narrow money and base money, the SchwarzBayesian information criterion suggested a VAR of eight distributed lags for St. Kitts/Nevis, four lags for Grenada, Guyana, and Trinidad/Tobago, three for Dominica, two for Antigua/Barbuda and Barbados, and one for Jamaica, St. Lucia, and St. Vincent. The Johansen test results indicated that the null hypothesis of no cointegrating equation was rejected at the 5 percent level of significance for only Antigua/Barbuda and St. Kitts/Nevis. Therefore, VECMs were estimated for these countries (see Table 2). In contrast, causality models between LRM and LNM for the other territories were estimated as a VAR in first difference form (see Table 3).

256

IAER: MAY 2001, VOL. 7, NO. 2 TABLE 2 VECM Estimates

Explanatory Variables

Antigua/Barbuda: L N M

St. Kitts/Nevis: L N M

0.01

-0.01

--

-0.30

0.15

-0.07

ALM Coefficient on the Error Correction ALNM 1] A L R M ALRM Coefficient on the Error Correction

-0.38

Y, A L N M

0.85

--

ALRM

-0.60

-0.31

N u m b e r o f Logs

8

1

Notes: Only the sums of the significant logged terms are reported.

TABLE 3 VAR Estimates

Barbados Explanatory Variables

Guyana*

Jamaica** St. Vincent/ Grenadines

Dominica

St, Lucia

Trinidad/ Tobago***

ALM ~] A L N M

-0.27

--

0.27

2] A L R M

--

--

0.13

ALNM

0.14

0.19

--

Y~A L R M

-0.61

-0.27

0.77 4

--

-0.28

-0.29

-0.24

0.42

0.50

--

-0.29

0.19

--

-0.32

--

-0.06

--

1

1

4

-0.07

ALRM

Number of Lags

2

3

--

1

Notes: *, **, and ***denotes the estimation periods are 1976:4 to 1990:4, 1976: l to 1990:2, and 1976:4 to 1993:1, respectively. Table 4 indicates that overall, the results were mixed. In the short run, the evidence suggests that reserve m o n e y causes narrow m o n e y in Antigua/Barbuda, Barbados, and

GREENIDGE ET AL.: MONEY ENDOGENEITY

257

Dominica and vice versa in St. Kitts/Nevis and Guyana. Noncausality exists for St. Lucia, St. Vincent, and Grenada while there is two-way causality for Jamaica and Trinidad/Tobago. In the long run (Table 2), the appropriate normalization that gave a stable vector for Antigua/Barbuda was via LRM while for St. Kitts, it was through LNM. Hence, the negative and significant coefficient on the error correction term implies that in the long run, money is endogenously determined in St. Kitts/Nevis but is exogenous in Antigua/Barbuda. Both results are consistent with what was found in the short run. TABLE 4 Wald Coefficient Test

Countries

Does LRM cause LNM (LBM)?

Does LNM (LBM) cause LRM?

Antigua/Barbuda

0.03

(0.83)

0.19

(0.01)

Barbados

0.06

(0.00)

0.41

(0.16)

Dominica

0.03

(0.03)

0.69

(0.84)

Grenada

0.35

(0.62)

0.39

(0.84)

Guyana

0.63

(0.12)

0.01

(0.39)

Jamaica

0.04

(0.13)

0.08

(0.13)

St. Kitts/Nevis

0.24

(0.12)

0.03

(0.00)

St. Vincent/Grenadines

0.26

(0.09)

0.15

(0.77)

Trinidad/Tobago

0.01

(0.02)

0.006

(0.005)

Notes: The p-valuesare in parentheses.The null hypothesisis no causal link.

Finally, utilizing the period of flexible exchange rate seems to have changed the results dramatically. For Guyana, instead of one-way causality for narrow money to reserve money, there is now also a causal link from reserve money to narrow money. For Trinidad/Tobago, the short-run bivariable causality result is lost, but due to cointegration between the two variables, causality is present in the long run from reserve money to narrow money. The finding for Jamaica is now short-run causality from reserve money to narrow money with no feedback. Conclusions

This paper has applied causality tests to determine the relationship between money (narrow and broad) and reserve money for 10 Caribbean countries. The empirical evidence indicates that the definition of money supply is irrelevant to the causal relation in small, open economies. However causal patterns may differ according to whether the monetary arrangements of the countries follow either a fixed or flexible exchange rate regime.

258

IAER: MAY 2001, VOL. 7, NO. 2

Moreover, the study highlights the fact that a mixed bag of policies must be implemented by respective central banks to maintain macroeconomic stability. These preliminary results are not in accordance with the findings of McClean [1997, 1998] and asserts that neither money endogeneity nor money exogeneity can be generalized to all small, open economies. The results are also in contrast to the endogeneity result of Howells and Hussein [ 1998] for the OECD countries.

Footnotes

1.

The results are available from the authors upon request.

References

Goodhart, Charles A. E. Monetary Theory and Practice, London, United Kingdom: Macmillan, 1984. "What Should Central Banks Do? What Should be Their Macroeconomic Objectives and Operations?," Economic Journal, 104, 427, November 1994, pp. 1424-36. Howells, Peter; Hussein, Khaled. "The Endogeneity of Money: Evidence from the G7," Scottish Journal of Political Economy, 45, 3, August 1998, pp. 329-40. McClean, Wendell A. "The Stability and Predictive Efficiency of the Traditional Money Multiplier in the Jamaica Context: A Comment on Bourne's Analysis," Social and Economic Series, 30, 3, September 1981, pp. 137-43. k . "Further Comment on the Conventional Money Multiplier," Social and Economic Studies, 34, 3, September 1985, pp. 85-106. . "Monetary Dynamics in Barbados: The Evidence from Cointegration Analysis and Error Correction Modeling," in D. Worrell; R. Craigwell, eds., Macroeconomics and Finance in the Caribbean: Quantitative Analyses, St. Augustine, Trinidad: Caribbean Center for Monetary Studies, 1997, pp. 85-106. . "The M1-M0 Relation in Barbados: Some Evidence on Causality," working paper, Central Bank of Barbados, 1998, pp. 115-9. McKinnon, J. G. "Critical Values for Cointegration Tests," in R. F. Engle; C. W. J. Granger, eds., Long-Run Economic Relationships." Readings in Cointegration, Oxford, United Kingdom: Oxford University Press, 1991.