Download DETERMINANTS OF HIGH INFLATION IN AN LDC:

DETERMINANTS OF INFLATION AND ITS INSTABILITY
A Case Study of a Less Developed Economy
1970-1994
by
A-M.M. Abdel-Rahman*
* Department of Economics, King Saud University, P.O.Box 2459; Riyadh 11451, Saudi Arabia.
2
Abstract
.
During the past three decades inflation in Sudan accelerated and was highly
variable. Various factors contributed to this acceleration and variability. Predominant
among these is the monetary financing of a persistent budget deficit in the country. Other
real, expectational and structural factors also contributed to the incidence of this high
inflation episode witnessed in the country. This paper investigates the sources of
instability in both the mean and variance of inflation. Account of the shifting means is
undertaken by shift dummies whereas account of the variable variance is undertaken via
Autoregressive Conditional Heteroscedasticity(ARCH) and Error Correction(EC)
estimation mechanisms. The analysis exploits recent results on cointegration given the
obvious nonstationarity of the data within the high inflation context. In due process, we
also trace the nominal, real, structural and expectational determinants of the Sudan
inflation.
3
DETERMINANTS OF INFLATION AND ITS INSTABILITY
A Case Study of a Less Developed Economy
A-M. M. Abdel-Rahman
Introduction:
During the past three decades, the inflation rate accelerated considerably in Sudan.
Conventionally, theoretical treatments of the determinants of inflation usually focus on
the relevant monetary and real factors. To these, expectational effects are sometimes
added. These are especially powerful during high inflation episodes and may even swamp
other effects if the inflationary regime shifts into a hyper gear. Moreover, in Less
Developed Countries(LDCs) structural factors may also add further impetus to the
process.
This paper attempts to study the inflation process in Sudan through the period
1970:1 to 1994:2. Section 1 of the paper discusses the variables, the data and the models.
Section 2 starts a more intensive study on the problem by conducting firstly unit root tests
on the individual series to determine the stationarity characteristics of the individual
series. Section 3 then undertakes the analysis which begins with some preliminary
estimates on some simplified conventional inflation functions. Instability is then explored
and incorporated in the used functional formulations. This is done firstly via the inclusion
of shift and interactive slope dummies to account for possible structural breaks in the
means of the process within the framework of a ‘general to specific’ methodology and
second via the allowance for Autoregressive Conditional Heteroscedasticity(ARCH)
effects to model simultaneously the observed instability in the means and conditional
variances. Section 4 then concentrates on short-run and long-run considerations using
Error Correction(EC) and cointegration methodologies. A final section then concludes the
study.
4
1. The Variables, Data and the Models:
To gauge the issues we used data on money, prices and output. The consumer
price index(CPI) was used for prices(P), M1 was used for money(M) and real Gross
Domestic Product(GDP) was used for output(Y). The variables were initially measured in
logarithmic form to obtain log prices(p), log money(m) and log output(y) respectively
over the period 1970:1-1994:2. Annualized rates of growth were then computed to
 ) and output growth( y ). Annual data
provide for the rate of inflation( p ), money growth( m
on real output were obtained from the World Book and interpolated via a Lagrangian
procedure to obtain quarterly figures1.
Using data on the Consumer Price Index(CPI)2 we computed the inflation rate on
an annual quarter-to-quarter basis. Data were obtained from the International Monetary
Fund- International Financial Statistics(IMF-IFS). Inspection of the inflation data in the
Sudan over the sample period reveals an obvious upward trend in the inflation rate as
witnessed by Fig(1) below. There is also a noticeable drift in the rate and an increase in
its volatility especially toward the latter parts of the sample period.
100
80
60
40
20
0
-20
72 74 76 78 80 82 84 86 88 90 92 94
Fig(1): Annualized inflation in Sudan; 1970:1 1994:2
1
Quarterly data are specially useful in the context of LDC inflations, since as noted by Ryan and
Milne(1994) for example, often the structure of the economy in developing countries is characterized by
price controls which are frequently adjusted by the government and the impact of depreciations, changes in
commodity prices(Oil) can have important effects on the quarterly - as contrasted to the annual - pattern of
inflation.
2
A shortcoming of the use of the CPI is that it doesnot show the extent of price controlled items in the
index; see for example Domowitz and Elbadawi(1987).
5
To further examine the trend and variance instability of inflation we computed the
bi-decadeal means, medians and standard deviations of the inflation rate over the sample
period. These are shown in Table(1) below:
Table(1)
Mean Inflations and Variability
1971:1-1990:3
period
Mean
Median
Standard
Deviation
1971:1-1974:4
12.220
13.357
10.263
1975:1-1979:4
16.289
15.878
9.643
1980:1-1984:4
24.713
24.764
7.036
1985:1-1989:4
35.893
37.240
16.532
1990:1-1994:2
70.939
73.877
13.777
Again the obvious trend is noticeable from the bi-decadeal means and medians and the
variability is noticeable from the fluctuating standard deviations.
2. Stationarity:
In this section we investigate the time series properties of the variables by
conducting unit root tests to gauge the orders of integration of the used variables and to
resolve the issue of trend stationarity versus difference stationarity characterizations.
Augmented Dickey-Fuller(DF) and Phillips-Perron(PP) tests were used on the various
variables to test for the order of integration. Results for data in rates form are:
Table(2)
Stationarity of rates
Test
p

m
y
ADF
-1.999
-3.408
-2.642
PP
-2.917
-2.751
-2.935
6
All variables are nonstationary, hence we proceeded to check for results in
difference form. Table(3) shows results for the respective variables in log-difference
growth rate form:
Table(3)
Stationarity of differences in rates
Test
 p

m
 y
ADF
-5.061
-6.218
-3.652
PP
-8.955
-13.779
-5.421
The results show that at the 5% level all the rate series are I(1). Thus we arrive at
the conclusion that the rates variables are nonstationary and I(1) whereas their firstdifference form are stationary and I(0). This reveals that the estimation should be
undertaken with first differences not growth rates
3. Conventional Estimates, Stability and Structural Breaks:
In this section we estimate the determinants of inflation and its variability. The
model of inflation used for the analysis assumes variants of the Autoregressive
Distributed Lag(ADL) form:
 t   (L)y t  (L)z t   t
(L)p t    (L)m
where (L), (L),  (L) and (L) are appropriate lag polynomials, z is a vector of some
exogenous determinants which may include structural variables besides the monetary,
output and expectational explanatories. This is a more general formulation than that used,
for example, by Bairam(1988,1990), Darrat(1985,1986) ... inter alia. The equation
provides a suitable testing cite to discriminate between the Monetarists, Keynesian and
Structuralists theories of inflation. As far as the monetary factor is concerned, in
Keynesian models output responds to money variations whilst in monetarists
interpretations the response to money is only transitory. The rational expectations(RE)
school holds that only unanticipated changes in money can affect output and for the real
7
business cycle(RBC) theory money has no impact whatsoever on output. Most theories,
however, agree that changes in the money supply affect the price level. To accept the
Monetarist interpretation, the s, which measure the response of inflation to the state of
the economy should be insignificant while the s which measure of the response of
inflation to the monetary stimuli should be positive and significant. For the Keynesian
interpretation to be valid, the reverse should hold true. The presence of y may also reflect
supply side effects whereby the increased supply of goods serves to reduce inflationary
pressures and in that case the s would be significant and negative. The s would be
highly significant in the case of high and hyperinflationary situations where there is
usually a strong inflationary inertia. The lagged dependent variables will also measure the
effect of expectations in such an environment and in a simple adaptive formulation the s
may measure the speed of inflation. For structuralists interpretations3 we expect the s to
be significant. We also expect them to be positive for the case of adverse shocks and
negative for the case of favorable shocks.
We begin with a simple inflation function to investigate the role of the key
determinants and to determine the sources of instability. Annualized nominal money
 ) and output growth( y ) were used in some preliminary regressions for the
growth( m
inflation rate( p ). A first regression was obtained as:
t
p t  2.668 0.730 y t  0.998 m
( 0.698 )
R
2
 0.559
R
2
LM (1)  54.691
( 0.000)
ARCH (1)  23.455
( 0.000)
Chow F
 8.879
79:4 ( 0.000)
3
 0.549
( 3.329 )
( 9 .379 )
  15.960
LM (1  4 )  56.545
( 0.000)
ARCH (1  4 )  23.267
( 0.000)
Chow F
 5.506
92:1 ( 0.001)
F  57.675
( 0.000)
d  0.500
RESET  F  1.249
( 0.267)
HESC  21.466
( 0.001)
AIC  5.572
SC  5.653
See for example, Kirkpatrick and Nixson(1976) and Johnson(1986). Structural factors, e.g. the effect of a
major devaluation as was witnessed in 1979, 1982 and 1984 immediately increases the domestic prices of
imports. Most Sudanese imports are intermediate and capital goods hence the effect was an increase in the
final cost of home manufactured goods and increased prices plus an ensuing wage-price spiral.
8
where figures in parentheses beneath the estimated coefficients are the t-ratios. In
addition to the conventional test statistics, we also report on various LM tests on serial
correlation which are chi-square distributed with their p-values in parentheses. HESC is
Breusch-Pagan LM test on heteroscedasticity. ARCH is Engle’s test for autoregressive
conditional heteroscedasticity, Chow F is Chow breakpoint test at the subscripted time
period where the period 1979:4 for example heralded the onset of a major stabilization
and structural adjustment program, AIC is Akaike Information Criterion and SC is
Schwartz Criterion. R 2 and d were low which could be an indication of a possible
misspecified relationship. There was ample evidence on serial correlation and ARCH
effects. Apart from that, we could initially infer the expected results that nominal money
contributed to positive inflations in an almost one-to-one basis whereas real output served
to depress inflation in an indication of possible supply factors at work i.e. inflation is
related in a negative way to growth. Adding inflationary inertia to the picture resulted in
the following estimated equation:
 t  0.904 p t 1
p t  2.773 0.256 y t  0.055 m
(1.490 )
R
2
 0.898
R
2
LM (1)  0.838
( 0.360)
ARCH (1)  0.144
( 0.703)
Chow F
 1.345
79:4 ( 0.260)
( 2 .358 )
 0.895
( 0.734 )
  7.656
LM (1  4 )  16.765
( 0.002 )
ARCH (1  4 )  3179
.
( 0.528)
Chow F
 2.579
92:1 ( 0.043)
(17 .491)
F  262.581
( 0.000)
d  1828
.
RESET  F  1.261
( 0.264 )
HESC  15.895
( 0.069 )
AIC  4.113
SC  4.222
The introduction of the lagged dependent variable which accounts for the inflationary
momentum and perhaps the expectations feeding it swamped the effects of the other
variables both theoretically and economically yielding higher R 2 and d statistics. It also
removed the ARCH effect present in the static formulation.
Comparing the two formulations, we note that the first simple equation gives
importance to monetary influences whereas the second reflects the predominance of
inflationary momentum and expectations. The sample period was a period of high and
9
persistent government budget deficits and an increasing tendency to finance these deficits
 has an obvious
by resort to money creation - seigniorage. The rate of money growth m
relationship with seigniorage and the evidence is thus consistent with the hypothesis that
deficits - seigniorage - raises money and inflation and lowers rates of economic growth.
To explain the significant ARCH effect we note that erratic exchange rates policies and
monetary accommodation of exogenous shocks added considerable noise to the economic
environment over the period by increasing the variance of money growth shocks. This in
turn increased the variance of inflation - an ARCH effect - which generally has a
significant negative relationship with economic growth.
3.1. Instability in Mean Inflation:
Inspection of the stability or otherwise of inflation was done using the Cumulative
Sum(CUSUM) of the residuals which revealed that a number of significant structural
breaks may have occurred in the periods 1975:3, 1979:4, 1982:4, 1985:2, 1988:3 and
1992:1. The Chow breakpoint tests at these time periods were respectively 2.451, 6.245,
9.596, 4.889, 16.528 and 5.506 which were significant respectively at levels 0.070, 0.001,
0.000, 0.006, 0.000 and 0.002. In particular, the break in 1975:3 may have been due to the
overheating of the economy because of massive capital inflows from the oil-rich Arab
countries in the form of foreign direct investment and assistance. This same period
coincided with the highest growth rate registered in the sample period. The break in
1979:4 could be attributed to the onset of the effects of the massive readjustment program
adopted in the third quarter of 1979 which included sweeping trade and exchange rate
liberalization and devaluation measures; that of 1985:2 could be traced to the
intensification of the civil war in the southern part of the country and the collapse of the
then ruling military regime. This precipitated a decline in the inflation rate probably due
to expectational effects. The break in 1988:3 was due to wide scarcities because of
massive famines and floodings throughout the country which occurred early August 1988.
The break in 1992:1 signaled the start of another broad and wide-ranging economic
liberalization program
10
To account for these breaks we accordingly defined a set of dummy variables to
cater for shifts in means and slopes over these time periods thus representing the various
structural factors which affected the Sudanese economy during the period of study and
reestimated the basic regression in rates form. A reduction process similar in spirit to that
of the “general to specific methodology” then ensued whereby shift dummies for 1979:4,
1982:4 and 1988:3 proved insignificant and were hence subsequently discarded. Results
were:
 t  0.676 D82  m
t
p t  10.045 55.459 D85  0.529 y t  1040
.
D85  y t  0.356 m
( 3.565)
( 4 .170 )
( 6.730 )
( 3.123)
( 3.354 )
( 4 .599 )
 t  1023
 t  0.268 D92  m
t
 2.144 D85  m
.
D88  m
( 7 .272 )
R
2
 0.893
(11.109 )
R
2
 0.883
  8.124
( 3.246 )
F  88.933
d  0.882
( 0.000 )
LM (1)  30.565
( 0.000)
ARCH (1  4 )  22.536
( 0.000)
LM (1  4 )  36.978
( 0.000)
HESC  53164
.
( 0.000)
ARCH (1)  17.127
( 0.000)
AIC  4.280
SC  4.524
All coefficients were significant and showed the required responses except that for
 which was in the opposite direction to what was expected.
the interactive D85 and m
The overthrowal of the then ruling military regime resulted in a change of expectations
and the subsequent transitory and civilian regimes showed a higher level of monetary
discipline. The effect of that was a declining inflation rate. The other results again show a
similar pattern to those previously obtained. Nominal money, output and the rest of the
structural factors were significant determinants of inflation in the country over the period
in reflection of the fact that both demand side and supply side factors were at work in the
economy. However, a remaining deficiency of the equation is the low d and LM statistics
plus the significant ARCH and HESC effects which could be an indication of still present
misspecification. These are generally thought to be in non-regression directions as
evidenced by the various statistics.
11
3.2. Variance instability and ARCH effects:
Here we proceed to provide empirical measures of the uncertainty of inflation.
This is done by investigating the variance of inflation through the employment of ARCH
effects. Since the ARCH was significant in the above formulation that would be taken as
an indication of an increase in inflation uncertainty. Hence we proceed next to incorporate
the instability in the variance and the resulting inflation uncertainty, i.e. the ARCH effect
into the estimation process and then to treat any possibly remaining nonstationarities via
application of cointegration methodology. We analyze the annualized rate of inflation
between 1970:1-1994:2 within an ARCH context. A first trial with an ARCH(1) process
using a BHHH - outer product of the gradient procedure - resulted in the following
results:
t
p t  10138
.  0.534 y t  0.348 m
( 4 .185)
( 4 .419 )
( 4 .416 )
  27.034  1048
.
 2t 1
2
t
( 2 .200 )
LM (1)  0.011
( 0.915)
( 2 .823)
LM (1  4)  0.783
( 0.941)
Q 36  29.239
The LM tests on autocorrelation now show the absence of the problem and the coefficient
of the lagged squared errors being significant at level 0.006 in the ARCH process. Again
a similar pattern to that obtained in the original static formulation was obtained. The
output effect remained negative and significant whereas the nominal money effect was
positive and significant but with a much reduced impact of 0.348 magnitude. To combine
the mean and variance effects we reintroduced the shift dummies into the formulation
where an ARCH model was also estimated for the dummy variables model and the results
obtained were:
12
 t  0.678 D82  m
t
p t  11367
.  54.130 D85  0.477 y t  2.070 D85  y t  0.288 m
( 5.755)
( 4 .941)
( 9 .402 )
( 8.455)
( 3.922 )
( 4 .642 )
 t  1063
 t  0.234 D92  m
t
 2.102 D85  m
.
D88  m
( 8.655)
(14 .909 )
( 3.345)
 2t  18.519  0.767  2t 1
( 2 .633)
R
2
 0.876
R
2
 0.862
( 2 .790 )
  8.846
F  58.871
d  0.852
( 0.000 )
ARCH (1)  0.328
( 0.567 )
Despite the overall significance of the equation and ARCH effect, there was still evidence
of an autocorrelation problem as judged by the still low d statistics. To tackle that
problem we proceeded to estimate the model in EC form in the next section.
4. Error Corrections:
Since the variables are stationary, we proceeded to estimate a dynamic short-run
ECM. The estimated ECM assumed the form:
  0118
 p  0.634  0.836  y  0135
.
m
.
e 4
( 0.802 )
R
2
 0.115
SC  4.182
( 3.539 )
  7.490
LM (1)  0.233
( 0.629 )
ARCH (1  4)  0.948
( 0.918)
(1.504 )
F  4.838
( 0.004 )
( 2.168)
d  2.096
LM (1  4)  22.788
( 0.000)
HESC  5.543
( 0.785)
AIC  4.071
ARCH (1)  0.0005
( 0.983)
RESET F  4.542
( 0.036)
Statistically, the ECM for inflation passes the usual tests for serial correlation and
stability. The coefficient on the fourth order disequilibrium ECM is 0.118 in absolute
value implying that 11.8% of any disequilibrium in any one year is compensated for
during the following year. This reflects a stable ECM which eventually converges to its
long-run path. The coefficient on the income growth variable was -0.836 and represents
the short-run elasticity. The coefficient on the money growth variable was 0.135 and was
significant at 13.6% level. It represents the short-run elasticity. The long-run elasticities
could be derived from the disequilibrium ECM but we preferred to estimate it directly via
the application of the Johansen cointegration procedure in the following sections.
13
All in all, the results show significant negative short-run effects of the income
variable and some positive effects of money on inflation. The previous disequilibrium
from the long-run relationships was also significant.
To account for the structural factors in the ECM we estimated the following
specification:
 t  0.858 D82   m
t
p t  0.869  0.603  y t  1107
.
D85   y t  0.068  m
(1.188 )
( 2 .188 )
( 2 .213)
( 0.395)
( 2 .121)
 t  0.723 D88   m
 t  0.277 e t 1
 1295
.
D85   m
( 2 .352 )
R
2
 0.266
( 2 .823)
(1.673)
R
2
 0.204
  7.103
F
4.252
d  2.120
( 0.000 )
LM (1)  0.348
( 0.555)
ARCH (1)  0.171
( 0.679 )
The results confirm the depressing effect output has on inflation throughout the
estimation period and regardless of the incidence of breaks. Money, however, was
insignificant before 1982 but its effect increased sharply since then as evidenced by the
coefficients on the 1982 and 1988 breaks. The coefficients on money during the 1985
period was negative as obtained previously and similar considerations to those noted
above would explain its negative behavior in that period. As far as the short-run
disequilibrium is concerned, the results show that 22.7% of any disequilibrium will be
removed within the time period. Adjustments to errors thus are large and rapid during the
advent of high inflation.
5. Cointegration:
This section sequentially tests for cointegration of the hypothesized relationships
in order to address the issue of the presence of long-run equilibria between the variables.
Allowing for a previously noted linear deterministic trend in the inflation data we proceed
to undertake cointegration analysis via the application of a Vector Autoregression(VAR)
of order 2. Results are presented in table(4) below:
14
Table(4)
Johansen Cointegration*
Eigenvalue
LR
5%
no. of CE(s).
0.216
35.633
29.68
None**
0.134
13.442
15.41
At most 1
0.003
0.314
3.76
At most 2
* LR is the Likelihood Ratio Test. no. of CE(s) is the number of cointegrating equations. ** denotes
rejection of the hypothesis at 5% significance level.
The result of the LR test again indicates the presence of one cointegrating
equation at the specified significance level. The normalized cointegrating equation was
estimated as:

p  21734
.
 0.246 y  1695
.
m
( 0.494 )
( 0.259 )
Loglikelihood  707.005
where figures in brackets are standard errors. According to these estimates, real output
growth (supply) was not particularly significant since it possessed a t-value of 0.497
magnitude. The coefficient on the money growth variable was significantly different from
zero but not significantly different from one at 5% level. This is an indication of the fact
that over the long run the main determinant of inflation in Sudan was money growth, i.e.
long-run homogeneity was holding.
We also reestimated the extended function within a cointegration framework and
results obtained firstly indicated the presence of one cointegrating vector as given in
table(5) below:
15
Table(5)
Johansen Cointegration*
The Extended Function
Eigenvalue
LR
5%
no. of CE(s).
0.576
155.363
124.24
None**
0.346
77.312
94.15
At most 1
0.159
38.707
68.52
At most 2
0.123
22.935
47.21
At most 3
0.091
10.969
29.68
At most 4
0.024
2.306
15.41
At most 5
0.001
0.125
3.76
At most 6
* LR is the Likelihood Ratio Test. no. of CE(s) is the number of cointegrating equations. ** denotes
rejection of the hypothesis at 5% significance level.
The result of the LR test again indicates the presence of one cointegrating
equation at the specified significance level. The normalized cointegrating equation was
estimated after some reduction procedures as:
 t  0100
t
p t  12.813  0.333 y t  4.454 D85  y t  0.374 m
.
D82  m
( 0.146 )
( 0.548 )
( 0.118 )
( 0.079 )
 t  0198
t
 0.474 D88  m
.
D92  m
( 0.090 )
( 0.081)
Loglikelihood  1434.495
where figures in brackets are standard errors. According to these estimates, real output
growth (supply) was significant before and after its 1985 break. The coefficients on the
money growth variable were also significantly different from zero before and after the
successive monetary breaks. They were also not significantly different from one at 5%
level after the conclusion of the breaks in 1992. This is an indication of the fact that over
the long run the main determinant of inflation in Sudan has become money growth, i.e.
long-run homogeneity was holding at the end of the inflationary era.
16
Conclusion:
This paper attempted to study the determinants of the inflation process in Sudan
during the past two decades. Inflation was high, accelerating and variable during that time
period. These characteristics were incorporated in the estimation procedure via
allowances for structural regime shifts and ARCH effects within some conventional
formulations directed towards discrimination between alternative inflation theories within
the context of a LDC. Results obtained generally validate the conjecture that the inflation
witnessed in Sudan is really a hybrid of monetary demand, structural supply and
expectational origins.
While the problem was predominantly one of a fiscal deficit, other factors played
significant roles e.g. inflationary inertia and the various structural factors which impacted
the economy during the period of study. In particular, the widesweeping stabilization and
structural adjustment programs of 1979 and after unleashed forces that thwarted the
authorities ability to harness inflation. The budget deficits were not kept within limits the
economy could absorb and in turn they ignited new inertia to the already existing
inflationary forces. This coupled with the altogether different phenomenon which
emerged since 1992 where inflationary finance through seigniorage became a formal
policy tool of the government. The result is witnessed in the present quasi hyperinflation
effecting the economy of Sudan.
17
References:
Bairam, E. ‘The Variability of Inflation: A New Approach and Some New Empirical
Evidence’ Economic Letters 28(1988)327-29.
-
‘Money and Inflation: The case of Western Developed Countries 1960-1980’
Applied Economics 22(1990)863-70.
Darrat, A. F. ‘The Monetarist versus the Neo-Keynesian Views on Inflation. Some Italian
Evidence’ Economic Letters 19(1985)237-41.
-
‘Money, Inflation and Causality in North Africa’ Journal of Macroeconomics
8(1986)87-103.
Darrat, A. F. and A. C. Arize ‘Domestic and International Sources of Inflation in
Developing Countries: Some Evidence from the Monetary Approach’
International Economic Journal 5(1990)55.69.
Dhakal, D. and M. Kandil ‘The Inflationary Experiences of Six Developing Countries in
Asia: An Investigation of Underlying Determinants’ Applied Economics
25(1993)413-25.
Domowitz, I. and I. Elbadawi ‘An Error Correction Approach to Money Demand. The
case of Sudan’ Journal of Development Economics 26(1987)257-75.
Dornbusch, R. and S. Fischer ‘Stopping Hyperinflations, Past and Present’
Weltwirtschaftliches Archiv (1986)1-4.
Johnson, O. E. G. ‘On Growth and Inflation in Developing Countries’ IMF Staff Papers
31(1984)636-60.
Kirkpatrick, C. H. And F. I. Nixson ‘The Origins of Inflation in Less Developed
Countries: A Selective Review’ in M. Parkin and G. Zis(eds.) “Inflation in Open
Economies” Manchester University Press, Manchester, 1976.
Montiel, P. ‘Empirical Analysis of High Inflation Episode in Argentina, Brazil and Israel’
IMF working papers 88/68(1988).
Ryan, T. C. I. And W. J. Milne ‘Analyzing Inflation in Developing Countries: An
Econometric Study with Application to Kenya’ Journal of Development Studies
31,1(1994)134-156.
Saini, K. ‘The Monetarist Explanation of Inflation: The Experiences of Six Asian
18
Countries’ World Development 10(1982)871-84.
Siklos, P. L. ‘The Money Growth-Inflation Relationship under Hyperinflation: An
Illustration from Hungary’s Postwar Experience’ Applied Economics
23(1991)1453-60.
Vogel, R. C. ‘The Dynamics of Inflation in Latin America, 1950-1969’ American
Economic Review 64(1974)102-14.