Chapter

VI Exchange Rate Developments and Their Relationship to Inflation

Author(s):
Jose Martelino, S. Erbas, Adnan Mazarei, Sena Eken, and Paul Cashin
Published Date:
December 1995
Share
  • ShareShare
Show Summary Details

This section briefly reviews developments in the exchange rate of the Lebanese pound and provides an empirical study of the relationship between movements in the exchange rate and inflation. The evidence suggests that since the mid-1970s, the volatility of the Lebanese pound exchange rate has increased substantially, while the relationship between exchange rate movements and inflation has become more significant.

Evolution of the Exchange Rate

Lebanon, unlike most countries, has had a flexible exchange rate system since 1948. The monetary authorities have at times intervened in the foreign exchange market to stabilize short-term exchange rate fluctuations, but the exchange rate has largely been determined in an interbank system.53

In the early 1970s, Lebanon had a strong macroeco-nomic position. The economy in general registered relatively high rates of growth, low inflation, and overall balance of payments surpluses. This performance reflected small fiscal deficits and a liberal economic system. In these macroeconomic circumstances, coupled with a stable political environment, the LL/USS exchange rate registered an appreciating trend with limited volatility (Table 11, Chart 12). Specifically, during 1957–74, the Lebanese pound appreciated by about 35 percent against the U.S. dollar, and the maximum monthly increase and decrease in the LL/USS exchange rate were both limited to about 6 percent.

Table 11Exchange Rate Indices and Inflation
YearLL/SDR (End of period)LL/US$ (End of period)LL/US$ (Period overage)NEER1 (1989= 100)REER2 (1989 = 100)CPI Inflation (In percent, period overage)
19483.603.603.60
19493.253.253.25
19503.743.743.74
19513.803.803.809.52
19523.573.573,57
19533.193.193.19-4.35
19543.243.243.24-4.35
19553.253.253.25
19563.203.203.204.76
19573.163.163.189.09
19583.163.163.184.17
19593.163.163.164.00
19603.153.153.173.85
19613.023.023.08
19623.063.063.01-3.70
19633.083.083.103.85
19643.083.083.0716,639.41133.503.70
19653.073.073.0716,656.54132.603.57
19663.173.173,1316,361.04129.14
19673.133.133.2016,007.39127.803.45
19683.183.183.1616,540.14127.06
19693.253.253.2516,129.29124.666.67
19703.253.253.2716,051.88118.29
19713.433.163.2315,963.94113.19
19723.273.013.0515,739.47110.856.25
19733.032.512.6116,926.00115.565.88
19742.822.302.3319,523.16128.8711.11
19752.842.432.3019,103.22122.1310.00
19763.402.932.8716,909.59125.5327.27
19773.643.003.0715,526.27124.8819.64
19783.913.012.9614,576.45120.9410.45
19794.293.263.2412,755.10120.3522.97
19604.653.653.4412,051.00125.6424.18
19815.384.634.3111,371.48127.6419.47
19824.203.814.7411,614.54142.3618.52
19835.755.494.5313,240.38163.236.88
19848.718.896.5110,207.38140.3818.13
198519.8818.1016.424,201.6793.6669.31
1986106.4287.0038.371,681.3671.6495.32
1987645.49455.00224.60304.8874.46487.13
1988713.22530.00409.23116.5970.55154.97
1989663.65505.00496.69100.00100.0072.20
19901,197.88842.00695.0966.57107.2868.82
19911,257.35879.00928.2349.01114.2951.46
19922,527.251,838.001,712.7928.69142.11119.99
19932,350.161,711.001,741.3627.41170.0829.11
Source: International Monetary Fund, International Financial Statistics.

Nominal effective exchange rate. Decline implies depreciation.

Real effective exchange rate. Decline implies depreciation.

Source: International Monetary Fund, International Financial Statistics.

Nominal effective exchange rate. Decline implies depreciation.

Real effective exchange rate. Decline implies depreciation.

In the mid-1970s, with the onset of the civil war, the budgetary situation began to deteriorate and led to rising credit creation. While the overall balance of payments remained in surplus until the early 1980s, the uncertain political environment and the pickup in inflation led to increased currency substitution in private portfolios and to speculative behavior. Reflecting these developments, in the period 1975–82, the LL/US$ and nominal effective exchange rates depreciated steadily, and exchange rate volatility increased significantly. During 1975–82, the Lebanese pound depreciated by about 50 percent against the U.S. dollar, while the nominal effective exchange rate depreciated by about 40 percent. Moreover, the maximum increase and decrease in the monthly LL/USS rate rose to 8 percent and 11 percent, respectively.

During the second half of the 1980s, the macroeconomic situation deteriorated significantly as the domestic conflict escalated. These developments contributed to an acceleration in inflation, a weakening in the external position and a rapid dollarization of the economy, and put significant downward pressure on the Lebanese pound.54 Meanwhile, with shifting expectations, the volatility of the exchange rate escalated sharply: during 1983–90, the Lebanese pound depreciated by close to 100 percent vis-a-vis the U.S. dollar and in nominal effective terms. The maximum increase and decrease in the monthly LL/USS rate rose to 43 percent and 20 percent, respectively.

After the adoption of the Tai'f Agreement, which provided a framework for settling the domestic conflict, the Government embarked upon a stabilization plan in 1991. The authorities, while not officially fixing the exchange rate, in effect attempted to link the Lebanese pound to the U.S. dollar without any formal commitment to maintaining a fixed exchange rate. Significant fiscal adjustment took place in 1991, with the fiscal deficit as a percentage of total expenditures declining to 56 percent from 84 percent in 1990. However, a retroactive and sizable wage increase in December 1991 undermined the public's confidence in the Government's fiscal position, which led to a run on treasury bills. The flight from government paper, capital outflows, and a shift in private sector portfolios out of the Lebanese pound and into foreign currencies led to intense pressure on the exchange rate in the first half of LL 879 per US$1 in January 1992 to LL 2,527.8 per US$1 by September 1992, and inflation soared.

In the last quarter of 1992, the authorities adopted a package of stabilization measures designed to reduce the fiscal deficit and credit creation and to re-establish financial stability. The consequent return of confidence, attractive interest rates on Lebanese government securities, and favorable prospects in the real estate and construction sectors all led to a surge in capital inflows, which put upward pressure on the exchange rate.55 In the circumstances, the authorities intervened in the foreign exchange market to moderate the appreciation of the Lebanese pound and to build up further official foreign exchange reserves. Despite this intervention, the Lebanese pound appreciated by about 30 percent against the U.S. dollar between October 1992 and March 1994, thereby reversing the sharp depreciation that took place in the third quarter of 1992. Not surprisingly, starting in 1993, there was a sharp decline in exchange rate volatility.

Chart 12Volatility of the Lebanese Pound/U.S. Dollar Exchange Rate

(In percent)

Sources: Bank of Lebanon; and International Monetary Fund, international Financial Statistics.

1 The coefficient of variation is based on 12-month moving averages and standard deviations, which are backward-looking. For example, the observation for January 1980 is based on data from February 1979 to January 1980.

The sizable movements in inflation and the exchange rate have led to significant swings in the real effective exchange rate of the Lebanese pound, with the most important developments occurring in the 1980s. During 1964–81, in spite of large movements in the LL/US$ exchange rate and in the nominal effective exchange rate, the real effective exchange rate remained relatively stable; in 1981 it was only about 4 percent more depreciated than its level of 1964 (Chart 13).56 However, the real effective exchange rate depreciated by 57 percent between 1983 and 1988, although it subsequently appreciated by about 70 percent between 1989 and 1993.57 In spite of these very large swings, the real effective exchange rate in 1993 was only about 32 percent more appreciated than its level of 1974. During 1993, the Lebanese pound appreciated more or less steadily against the U.S. dollar, but domestic prices did not follow pari passu the movements in the exchange rate. Consequently, the real effective rate appreciated sharply (Chart 14).

Chart 13Real and Nominal Effective Exchange Rates, 1964–93

(1980=100)

Source: International Monetary Fund, International Financial Statistics.

However, inferences about Lebanon's external competitiveness based on movements in the real effective exchange rate reported above should be drawn with caution. In addition to the limitations of indicators based on aggregate consumer price indices for assessing the profitability related to the production of traded goods, other indicators would present a somewhat different picture of Lebanon's external competitiveness.58 For example, the available information indicates that real labor costs in Lebanon have declined dramatically since 1974 (Table 12); during 1974–93, real minimum wages in terms of the Lebanese pound and the U.S. dollar are estimated to have declined by 86 percent and 43 percent, respectively.

Table 12Legal Minimum Wages
YearMinimum Wage in Lebanese Pounds (Annual earnings.)Real Minimum Wage in Lebanese Pounds (Constant 1974)Real Minimum Wage in Lebanese Pounds (1974 = 100)Minimum Wage in U.S. DollarsMinimum Wage in U.S. Dollars (1974=100)
19743,3003,3001001,418100
19753.7203,3851031,621114
19774,9802,948891,623114
19808,1002,842862,358166
19819,6002,822862,225157
198211,1002,751832,347166
198313,2003,051922,915206
198415,1202,971902,322164
198518,0002,087631,09677
198632,4001,9235384460
198781,0008192536025
1988210,0008322551336
1989540,0001,243381,08777
1990540,0007362276954
1991900,0008102597068
19921,416.0005791882758
19931,416.0004491481357
Source: Bank of Lebanon.
Source: Bank of Lebanon.

Chart 14Real and Nominal Effective Exchange Rates, 1989–93

(1980=100)

Source: International Monetary Fund, Information Notice System.

The dearth of comparable data on wages, unit costs, and other indicators of international competitiveness precludes a systematic assessment of Lebanon's international competitiveness in relation to other countries in the Middle East region. Nevertheless, it is useful to compare movements in the Lebanese nominal and real effective exchange rates with those for a number of countries in the region—Egypt, Israel, Jordan, Syria, and Turkey (Charts 15 and 16).59 Taking 1980 as the base year, while Lebanon's nominal effective exchange rate has depreciated significantly, its real effective exchange rate has appreciated by more than those of nearly all the other countries in the sample.

Chart 15Lebanon and Selected Other Countries: Nominal Effective Exchange Rates

(1980=100)

Source: International Monetary Fund. Information Notice System.

Chart 16Lebanon and Selected Other Countries: Real Effective Exchange Rates

(1980=100)

Source: International Monetary Fund, Information Notice System.

Long-Term Relationship Between Exchange Rate Movements and Inflation

Movements in the annual CPI and the LL/US$ exchange rate during the periods 1951–80 and 1975–93 are captured in Charts 17 and 18, respectively. While there was some co-movement between inflation and exchange rate movements during 1951–74, the nature of the relationship became tight and more contemporaneous after the start of the civil war. With increased political tensions, economic imbalances, and dollarization of the Lebanese economy, price-setting behavior became more closely affected by developments in the exchange rate. However, as shown in Chart 19, this relationship weakened in 1991. In particular, since the beginning of 1993, exchange rate policy has been geared to maintaining a stable nominal rate, and movements in the price level have not affected the exchange rate.

To obtain statistical insight into the relationship between inflation and nominal exchange rate movements, a simple regression equation relating the differential between Lebanese and U.S. inflation rates to changes in the LL/US$ exchange rate was estimated.60 This equation was estimated for the periods before (1951–74) and after (1975–1993) the start of the civil war (Table 13). The results indicate that the relationship between the Lebanese-U.S. inflation differential and exchange rate movements was positive but not statistically significant before the civil war but has been both positive and strongly significant since 1975. The relationship also proved to be positive and significant for the entire period of 1951–93.

Table 13Regressions of the Differential Between Lebanese and U.S. Inflation Rates on Changes in the LL/US$ Exchange Rate1
Time Period
Independent VariableOverall sample 1951–93Before civil war 1951–74Since civil war 1975–93
Intercept0.050.000.09
(1.14)(0.03)(1.01)
Change in LL/US$0.820.150.83
exchange rate(15.47)(1.00)(10.27)
Adjusted R20.930.010.90
DW statistic1.371.921.29
Source: IMF staff calculations.

All equations were estimated using annual CPI data. The estimation method was the exact maximum likelihood method with an AR(I) error specification. The numbers in parentheses are t-statistics.

Source: IMF staff calculations.

All equations were estimated using annual CPI data. The estimation method was the exact maximum likelihood method with an AR(I) error specification. The numbers in parentheses are t-statistics.

Chart 17Price Level and Exchange Rates, 1951–80

(Annual percent changes)

Sources: Bank of Lebanon; and International Monetary Fund, International Financial Statistics.

Chart 18Price Level and Exchange Rates, 1975–93

(Annual percent changes)

Sources: Bank of Lebanon; and International Monetary Fund, International Financial Statistics.

For a more rigorous approach, annual data were used to examine the long-term relationship between exchange rate movements and inflation during 1951–93, by searching for possible changes in the relationship between exchange rate movements and inflation.61 Before the Johansen cointegration test can be applied, it must be determined whether the CP1 and the LL/US$ exchange rate series are nonstationary. After establishing that the annual period average series for the Lebanese CPI and the LL/US$ exchange rate were nonstationary by the Dickey-Fuller and the augmented Dickey-Fuller tests (Table 14),62 the Johansen method was used to examine the existence of a long-term relationship between inflation and changes in the exchange rate.63

Table 14Dickey-Fuller Unit Root Tests, 1951–93
Variable2
Test Statistic1ep
DF0.641.55
ADF(I)-0.91-0.33
ADF(2)-1.01-0.36
ADF(3)-0.68-0.13
Source: IMF staff calculations.

DF and ADF are respectively the Dickey-Fuller and the augmented Dickey Fuller test statistics for testing the null hypothesis of stationarity. The numbers in parentheses are tag lengths. The critical value for the 5 percent level of significance is -3.52.

e = log (LL/US$ exchange rate); p = log (CPI).

Source: IMF staff calculations.

DF and ADF are respectively the Dickey-Fuller and the augmented Dickey Fuller test statistics for testing the null hypothesis of stationarity. The numbers in parentheses are tag lengths. The critical value for the 5 percent level of significance is -3.52.

e = log (LL/US$ exchange rate); p = log (CPI).

Chart 19Price Level and Exchange Rates, 1989–941

(Monthly percent changes)

Sources: Bank of Lebanon; and International Monetary Fund, International Financial Statistics.

1 Covers January 1989-March 1994; monthly CPI data for Lebanon were not available prior to 1989.

The test results for 1951–74 indicated that there was no cointegration between these variables, which may be due to the fact that appreciations in the exchange rate are not passed through to prices as fully as depreciations. The appreciation of the exchange rate during 1951–74 may partly explain the lack of cointegration between exchange rate movements and inflation during this period. However, when applied to the period 1975–93, the data pointed to the existence of a cointegrated relationship between inflation and exchange rate movements at the 5 percent level of significance (Table 15), The results indicated a sizable and positive relationship between exchange rate movements and inflation since the outbreak of the civil war. The absence of cointegration before 1975 and its presence thereafter indicate a closer relationship between exchange rate movements and inflation since the start of the civil war. These results, however, should be interpreted cautiously in view of the small size of the sample used and the structural changes (a significant increase in the mean and trend of the CPI and exchange rate data) in the time series for inflation and the exchange rate.

Table 15Johansen Cointegration Test, 1975–931
Cointegration likelihood ratio test based on maximal eigenvalue of the stochastic matrix
Hypothesis2
NullAlternativeTest StatisticCritical Value3
r = 0r = 125.5214.07
r ≤ 1r = 20.453.76
Cointegration likelihood ratio test based on the trace of the stochastic matrix
Hypothesis2
NullAlternativeTest StatisticCritical Value3
r = 0r ≥ 025.9715.41
r ≤ 1r = 20.453.76
Estimated cointegration on vector (coefficients normalized on p)4
p-1.00
e1.21
Source: IMF staff calculations.

Sample: 1975–93: maximum lag of one year in VAR.

r represents the number of cointegrating vectors.

Critical values are for the 5 percent level of significance.

e = log (LL/US$ exchahge rate): p = log (CPI).

Source: IMF staff calculations.

Sample: 1975–93: maximum lag of one year in VAR.

r represents the number of cointegrating vectors.

Critical values are for the 5 percent level of significance.

e = log (LL/US$ exchahge rate): p = log (CPI).

Subsequently, the possibility of a change in the long-term relationship between inflation and exchange rate movements following the stabilization measures after October 1992 was examined using monthly data. For this purpose, the Johansen procedure was used to test for the existence of a cointegrated relationship between inflation and exchange rate movements for the period starting in January 1989. To test the stationarity of the CPI and the LL/US$ exchange rate series, Dickey-Fuller and augmented Dickey-Fuller tests were again conducted.64 The results of these tests indicated that the time series for the CPI and the exchange rate were again nonstationary (Table 16).65

Table 16Dickey-Fuller Unit Root Tests, 1989–921
Variable3
Test Statistic2epCritical Value4
DF-2.37-1-78-3.51
ADF(1)-3.00-2.63-3.51
ADF(2)-1.94-1.74-3.51
ADF(3)-2.28-1.84-3.51
ADF(4)-2.89-1.94-3.52
ADF(5)-2.82-2.61-3.52
ADF(6)-3.44-3.09-3.52
ADF(7)-3.93-3.75-3.52
ADF(8)-4.24-3.41-3.53
ADF(9)-4.48-3.96-3.53
ADF(10)-3.09-3.35-3.53
ADF(11)-2.87-3.03-3.54
ADF(12)-2.08-2.65-3.54
Source: IMF staff calculations.

Sample: January 1989-December 1992.

DF and ADF are. respectively, the Dickey-Fuller and the augmented Dickey-Fuller test statistics for testing the null hypothesis of stationarity. The numbers in parentheses are lag lengths.

e = log (LL/US$ exchange rate); p = log (CPI).

Critical values are for the 5 percent level of significance

Source: IMF staff calculations.

Sample: January 1989-December 1992.

DF and ADF are. respectively, the Dickey-Fuller and the augmented Dickey-Fuller test statistics for testing the null hypothesis of stationarity. The numbers in parentheses are lag lengths.

e = log (LL/US$ exchange rate); p = log (CPI).

Critical values are for the 5 percent level of significance

The cointegration test was conducted for the period January 1989-December 1992, and the existence of a long-term relationship between inflation and exchange rate movements could not be rejected (Table 17).66 The evidence presented suggests the existence of a close, long-run relationship between inflation and exchange rate movements in Lebanon from January 1989 until the end of 1992. Since the initiation of the stabilization program, the nominal exchange rate has gradually appreciated, but the earlier close relationship between inflation and exchange rate movements has not been present. This could reflect the fact that there had been little exchange rate volatility during this period and that appreciations in the exchange rate are not passed through to prices as fully as depreciations.

Table 17Johansen Cointegration Test, 1989–921
Cointegration likelihood ratio test based on maximal eigenvalue of the stochastic matrix
Hypothesis2
NullAlternativeTest StatisticCritical Value3
r = 0r = 119.4614.06
r ≤ 1r = 20.373.76
Cointegration likelihood ratio test based on the trace of the stochastic matrix
Hypothesis2
NullAlternativeTest StatisticCritical Value3
r = 0r ≥ 019.8215.41
r ≤ 1r = 20.373.76
Estimated cointegration on vector (coefficients normalized on p)4
p-1.00
e1.41
Source: IMF staff calculations.

Sample: January 1989–December 1992, maximum lag of six months in VAR.

r represents the number of cointegrating vectors.

Critical values are for the 5 percent level of significance.

e = log (LL/US$ exchahge rate): p = log (CPI).

Source: IMF staff calculations.

Sample: January 1989–December 1992, maximum lag of six months in VAR.

r represents the number of cointegrating vectors.

Critical values are for the 5 percent level of significance.

e = log (LL/US$ exchahge rate): p = log (CPI).

Short-Term Dynamics of Inflation and the Exchange Rate and the Two-Way Relationship Between Them

Although the above analysis was not intended to be a comprehensive analysis of the dynamics or determinants of inflation in Lebanon, it provides some evidence with regard to the close relationship between Lebanese inflation and exchange rate movements. Yet this does not imply a one-way influence from the exchange rate to prices. To examine the possibility of feedback from inflation to exchange rate movements, a number of error-correction equations were estimated using monthly data for January 1989-December 1992. The existence of a long-run relationship (cointegration) between prices and the exchange rate, as shown above, indicates that the short-term relationship among these variables can be captured through an error-correction representation. For this purpose, the two-step method of Engle and Granger was used. The estimated error-correction equation was of the following form:

where the error-correction term ECt-1 represents the difference between the actual value of the CPI and the value predicted for the previous period from the cointegration relationship, and εt is the error term.67 All variables are in natural logarithms. Estimation of the above equation for the period January 1989-December 1992 by OLS yielded the following results:

The numbers in parentheses are t-statistics. The results point to a significant impact from exchange rate movements on the rate of inflation in the same month. Moreover, the negative coefficient for ECt-1 indicates that departures of the inflation rate in any one period from the magnitude implied by the long-term relationship between the price level and the exchange rate will be corrected in the following period. Consequently, it is possible in the case of Lebanon that sharp movements in the exchange rate, which may be due to political events and capital flows, could trigger inflationary episodes.

The above results assumed that movements in the exchange rate were determined exogenously. To examine the possibility of weak exogeneity of the price level with respect to the exchange rate, the error-correction equation was re-estimated after switching the place of Άpt and Άet.68 An answer to this question is also important for the statistical interpretation of the above error-correction framework; unless the exchange rate is exogenous with respect to the price level, the coefficients of equation (1) would be biased and inefficient. A sufficient condition for weak exogeneity is that the coefficient δ2 in the following equation be equal to zero:

where ut is the error term. Estimation of equation (2) yielded a statistically significant nonzero value for δ2:

Thus, the possibility of feedback from the price level to the exchange rate could not be rejected, indicating that the exchange rate and the price level are determined simultaneously in Lebanon.

For discussions of the evolution of the exchange rate system and exchange rate developments in Lebanon, see Badrud-Din (1984), Chami (1992), Makdisi (1987), Saϊdi (1981, 1984b), and Spitäller (1980).

The foreign exchange reserves of the BDL declined from $1,051 million in 1985 to $462 million in 1986 and to $336 million in 1987. The ratio of pound-denominated deposits to total deposits in the banking system declined from 65 percent in 1985 to about 8 percent in 1987.

Reliable data on capital flows for Lebanon are not available. Nevertheless, some inference about the size of such flows may be drawn from changes in the foreign exchange reserves of the BDL. which in 1993 rose by about 50 percent to $2.2 billion.

In Charts 13-16, a decline in the exchange rate indices indicates a depreciation.

Indeed, any judgment on the degree of appreciation of the real effective exchange rate depends on the choice of the base year. For example, using 1983 (the year after the Israeli incursion) as the reference year would indicate that by 1993 the real effective exchange rate had appreciated by only 4 percent. In addition, it should be noted that the CPI data for Lebanon are not official and may not be highly reliable.

These limitations include the absence of perfect substitutability of traded goods; the shortcomings of aggregate price indices in reflecting factor rewards, owing in part to different patterns of productivity growth in various countries; price controls and subsidies; and noninclusion of intermediate goods in consumer price indices. On the limitations of CPI-based measures of the real exchange rate, see Lipschitz and McDonald (1991), and Marsh and Tokarick (1994).

All exchange rate indices were computed using consumer prices.

The estimated regression equation was as follows:

INFDIFF, = α0 + α1e, + u1,

where INFDIFF, is the differential between Lebanese and U.S. inflation rates, et is the LL/US$ exchange rate, and ut is an error term. All variables are in logarithms.

See Johansen (1988), and Johansen and Juselius (1990) for details of the Johansen test. The Johansen cointegration test was used here instead of the Phillips-Ouliaris test because this section has been largely confined to examining the correlation between inflation and exchange rate movements, rather than testing a specific, single-equation econometric model as was done for the money demand function in Section V.

See Dickey and Fuller (1979, 1981). Dickey-Fuller tests for stationarity are used in this section as they are univariate versions of the Johansen procedure.

The cointegration test was applied using alternative lag lengths in the vector autoregression (VAR).

However, the results tended to be mixed, as the null hypothesis of a unit root was rejected for both the exchange rate and the CPI for lag lengths of seven, eight, and nine months, which might be due to the parameter instability of the Dickey-Fuller test, or to the small size of the sample used.

Initially, unit root tests were conducted for the entire period January 1989-March 1994. Although the null hypothesis of a unit root in the exchange rate and CPI series could not be rejected, a closer inspection indicated that there was a structural break in these time series after the start of the stabilization program of 1992. Over the period January 1993-March 1994, the exchange rate and CPI series were in fact stationary. Consequently, the results of the unit root tests are reported only through the end of 1992, and the cointegration tests are limited to the period January 1989-December 1992.

The existence of cointegration over January 1989-December 1992 was also verified using the two-step method of Engle and Grangcr(1987).

Owing to a lack of degrees of freedom, only the most parsimonious error-correction equation is employed,

As explained by McDermott and Wong (1990), for purposes of drawing inference and model formulation, the appropriate concept of exogeneity is weak exogeneity and not strict exogeneity (Granger non-causality). In addition. Toda and Phillips (1993) argue that Granger non-causality tests have serious limitations when applied to cointegrated systems. Hence, Granger non-causality tests were not applied in this section, and tests of weak exogeneity were applied to the error-correct ion model. For general discussions of exogeneity, especially in the context of cointegration. also see Banerjee and others (1993), and Ericsson (1991).

    Other Resources Citing This Publication