The relationship between inflation and inflation uncertainty in vietnam over the period 1995 2010

UNIVERSITY OF ECONOMICS

INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY

THE HAGUE

VIETNAM

THE NETHERLANDS

VIETNAM - NETHERLANDS
PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

THE RELATIONSHIP BETWEEN INFLATION
AND INFLATION UNCERTAINTY IN VIETNAM
OVER THE PERIOD 1995-2010

BY

NGUYỄN VĂN DŨNG

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

HO CHI MINH CITY, NOVEMBER, 2011


UNIVERSITY OF ECONOMICS

INSTITUTE OF SOCIAL STUDIES

HO CHI MINH CITY

THE HAGUE

VIETNAM

THE NETHERLANDS

VIETNAM - NETHERLANDS
PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

THE RELATIONSHIP BETWEEN INFLATION
AND INFLATION UNCERTAINTY IN VIETNAM
OVER THE PERIOD 1995-2010
A thesis submitted in partial fulfilment of the requirements for the degree of
MASTER OF ARTS IN DEVELOPMENT ECONOMICS

By

NGUYỄN VĂN DŨNG
Academic Supervisor:

DR. TỪ VĂN BÌNH
HO CHI MINH CITY, NOVEMBER, 2011


TABLE OF CONTENTS
List of Tables

iv

List of Figures

iv

List of Acronyms

v

Acknowledgement

vi

Abstract

vii

Chapter 1: Introduction

1

1.1 Background of the Study

1

1.2 Problem Statement

2

1.3 Research Objectives

3

1.4 Research Questions

3

1.5 Research Hypotheses

4

1.6 Justification of the Study

4

1.7 Scope of the Study

4

1.8 Organization of the Study

4

Chapter 2: Literature Review

6

2.1 Inflation Uncertainty

6

2.2 Theories about the inflation-inflation uncertainty relationship

7

2.3 Approaches to estimate inflation uncertainty

9

2.4

Methods to test the causal relationship between inflation and inflation

uncertainty

12

2.5 Empirical studies about the inflation-inflation uncertainty relationship

13

2.6 Conceptual Framework

24

2.7 Chapter Summary

26

Chapter 3: Research Methodology and Data

27

3.1 Research Methodology

27

3.1.1 Descriptive statistics

27

3.1.2 Unit root testing

27

3.1.3 Diagnostic tests for serial correlation, heteroskedasticity and ARCH effects
29
ii


3.1.4 Measuring inflation uncertainty

30

3.1.5 Granger causality tests

31

3.1.6 Analytical Framework

32

3.2 Data

33

3.3 Chapter Summary

33

Chapter 4: Findings and Discussion

34

4.1 Descriptive statistics

34

4.2 Unit root testing

35

4.3 OLS estimation of AR(p) model of inflation

36

4.4 Measuring inflation uncertainty

39

4.5 Granger causality tests

46

4.6 Comparison with previous studies

48

4.7 Chapter Summary

49

Chapter 5: Conclusion and Policy Implications

51

5.1 Conclusion

51

5.2 Policy Implications

51

5.3 Limitation and Further Studies

55

5.3.1 Limitation

55

5.3.2 Further Studies

55

References

57

iii


List of Tables
Table 2.1: Early empirical studies about the inflation-inflation uncertainty relationship
...................................................................................................................................... 19
Table 2.2: Empirical studies about the inflation-inflation uncertainty relationship using
the simultaneous estimation approach .......................................................................... 20
Table 2.3: Empirical studies about the inflation-inflation uncertainty relationship using
the two-step approach with symmetric GARCH models to estimate inflation
uncertainty .................................................................................................................... 21
Table 2.4: Empirical studies about the inflation-inflation uncertainty relationship using
the two-step approach with the extensions of GARCH model to capture the
asymmetric inflation uncertainty .................................................................................. 22
Table 4.1: Unit root tests .............................................................................................. 35
Table 4.2: Lag selection of AR(p) process ................................................................... 36
Table 4.3: OLS estimation of AR(13) model ............................................................... 37
Table 4.4: AR(13)-(GARCH(1,1), TARCH(1,1), PARCH(1,1), EGARCH(1,1))
models ........................................................................................................................... 39

Table 4.5: Granger Causality Tests .............................................................................. 46

List of Figures
Figure 4.1: Descriptive statistics of inflation in Vietnam 1995-2010 .......................... 34
Figure 4.2: Average rates of inflation by month in the period 1995-2010 (%) ............ 34
Figure 4.3: Inflation uncertainty of the AR(13)-EGARCH(1,1) model ....................... 45
Figure 4.4: Inflation and inflation uncertainty over the period 1995-2010 .................. 46

iv


List of Acronyms
AIC: Akaike Information Criterion
AR: Autoregressive
ARCH: Autoregressive Conditional Heteroskedasticity
EGARCH: Exponential Generalized Autoregressive Conditional Heteroskedasticity
GARCH: Generalized Autoregressive Conditional Heteroskedasticity
HQ: Hannan-Quinn criterion
OLS: Ordinary least squares
PARCH: Power Autoregressive Conditional Heteroskedasticity
SBV: State Bank of Vietnam
SC: Schwartz criterion
TARCH: Threshold Autoregressive Conditional Heteroskedasticity
UK: the United Kingdom
US: the United States of America

v


Acknowledgement

I would like to express my sincere gratitude to my supervisor Dr. Tu Van Binh
who gave me valuable guidelines, comments, suggestions, and inspiration for the
successful completion of this study. Besides, his friendly and inspiring approach has
given me a great deal of encouragements to overcome difficulties in the whole
research process.
I am also thankful to all lecturers and program administrators in the Vietnam –
The Netherlands Program for M.A. in Development Economics. They gave me
wonderful knowledge and help me kindly during the course.
Last but not least, I would like to express my appreciation to my family, my
friends who have given me a lot of support when I pursue my studies at the program.

vi


Abstract
The study investigates the causal relationship between inflation and inflation
uncertainty in Vietnam over the period 1995-2010 using the two-step procedure.
Inflation uncertainty is modeled by both symmetric model (GARCH(1,1)) and
asymmetric models (TARCH(1,1), PARCH(1,1), EGARCH(1,1)). The results indicate
that there exists an asymmetric impact of inflation shocks on inflation uncertainty,
implying that a positive inflation shock induces higher inflation uncertainty, while a
negative inflation shock induces lower inflation uncertainty. Based on information
criteria including AIC, SC, and HQ as well as diagnostic tests, AR(13)-EGARCH(1,1)
is considered the best model to model inflation uncertainty. Then Granger causality
tests are employed to test the causality between inflation and inflation uncertainty
(estimated by the AR(13)-EGARCH(1,1) model). The results support that a rise in
inflation significantly leads to more inflation uncertainty, which confirms the
Friedman-Ball hypothesis; and increasing inflation uncertainty leads to more inflation,
confirming the Cukierman-Meltzer hypothesis, which indicates an “opportunistic”
State Bank of Vietnam. The policy implication is that the monetary authorities have to

keep inflation low, stable and predictable to eliminate the negative impact of inflation
uncertainty.

Key words: inflation, inflation uncertainty, relationship, GARCH models,
Granger causality tests

vii


Chapter 1: Introduction
1.1 Background of the Study
Vietnam experienced hyperinflation in the late 1980s (approximately
300%/year) and early 1990s (approximately 50%/year) due to bad weather, weak
financial system, and especially poor governance of the authority. The year 1995
marked an important turning point when hyperinflation was completely controlled
and Vietnam began its deep international integration (i.e. formally normalized
diplomatic relations with the US and became the full member of ASEAN).
The years after 1995 witnessed the 1997-1998 Asian Financial Crisis and its
consequences to the world prices and aggregate demand. Because of the negative
consequences of the crisis, both demand for Vietnamese goods and domestic
demand declined. This period was marked by low inflation with mild deflation in
2000 (-0.5%) despite rapid monetary and credit growth (approximately 3040%/year) and VND’s sharp devaluation (approximately 36%) in the period 19972003 (Nguyen & Nguyen, 2010).
After the period of stably low inflation, inflation began increasing sharply
with 9.5% in 2004. When the negative impact of the Asian crisis declined, demand
began to rise. Demand increase and the rise of salary in both the public and FDI
sector in 2003 pushed the prices to rise. Additionally, supply shocks (due to bird flu
and bad weather) contributed to the price increase. The government considered
supply shocks mainly responsible for inflation. Food prices increased by 15.5%
compared with the general inflation rate of 9.5% and inflation of non-food products
was 5.2% in 2004 (Nguyen & Nguyen, 2010).

For the fear of increasing inflation, State Bank of Vietnam (SBV)
implemented tightening monetary policy, making interest rates increase slightly,
and fixed the exchange rate since 2004. The rigid management of exchange rate
until late-2008 did not stabilize inflation as in the period 2000-2003. Inflation, after

1


declining slightly in 2006, increased sharply to 12.6% in 2007 and up to 20% in
2008 (Nguyen & Nguyen, 2010).
There are many causes of high inflation in the period 2007-2008, which
include the sharp increase of minimum wage rate, sharp rise of international
commodity prices, lax and inflexible monetary policy, rigid exchange rate
management, and the opening of Vietnam to the world economy since Vietnam’s
join the WTO in late-2006, leading to indirect investment flows of foreign countries
into Vietnam, pushed stock prices and asset prices increase dramatically. To
stabilize the exchange rate, SBV had to pump a large amount of money into the
economy (approximately VND 145 thousand billion), contributing to more severe
inflation. In fact, the increase of money supply and credit growth in the economy in
the last decade was very strong, especially in 2007 when M2 increased by 47% and
credit growth increased by 54% (Nguyen & Nguyen, 2010).
Impacts of the 2007-2008 Global Financial Crisis made inflation slow down
in Vietnam since late-2009. The decrease of international prices and total demand
made Vietnam reverse alarming trend of increased inflation since 2008. However,
for fear of recessionary impacts, the government introduced the stimulus package
since the second quarter of 2009, which made money supply and credit soar. The
early months of 2010 saw relatively stable inflation rates. However, inflation
increased sharply since September 2010 and reached 11.75% for the whole year.
In general, the period 1995-2010 is the relatively stable inflation time
compared with the hyperinflation period in the late 1980s-early 1990s. Inflation is

rather low in a decade from 1996 to 2006. However, high inflation has returned to
the Vietnamese economy since 2007 with two-digit inflation rate, which poses
many threats to the economy.

1.2 Problem Statement
Inflation is a worldwide problem that causes a negative impact on every
economy in the world. In the Vietnamese case, the country incurred relatively high
2


level of inflation during a long period 1995-2010, on average 7%/year, which is
more persistent and more volatile than that of other countries in the Southeast Asia
region. Inflation may result in many serious consequences. Among them, inflation
uncertainty is regarded as one of the most significant consequences of inflation.
Inflation uncertainty is defined as a situation in which future price levels are
unpredictable and the general public does not know whether inflation will increase
or decrease in the future. As a result, it affects negatively consumers and producers’
decisions about saving and investment in the future, which leads to the loss of
economic well-being (Golob, 1994). As a result, understanding the nature of the
relationship between inflation and inflation uncertainty is essential for improving
current policies to control inflation and stabilize the macro economy.

1.3 Research Objectives
The specific objectives of the study are to
(i) examine whether there is an asymmetric impact of inflation shocks on
inflation uncertainty in Vietnam
(ii) test whether inflation causes inflation uncertainty in Vietnam
(iii) test whether inflation uncertainty causes inflation in Vietnam
(iv) offer some policy implications to better control inflation and inflation
uncertainty

1.4 Research Questions
Main question:
(i) What is the relationship between inflation and inflation uncertainty in
Vietnam over the period 1995-2010?
Sub-questions:
(i) Is there an asymmetric impact of inflation shocks on inflation uncertainty?
(ii) Does inflation cause inflation uncertainty?
(iii) Does inflation uncertainty cause inflation?

3


1.5 Research Hypotheses
The null hypotheses are as follows:
H0: Inflation does not cause inflation uncertainty.
H0: Inflation uncertainty does not cause inflation.

1.6 Justification of the Study
This study makes a major contribution in two aspects as follows.
First, there have been many empirical studies on the relationship between
inflation and inflation uncertainty. However, most of the studies mainly focus on
developed countries with relatively low inflation rates (Jiranyakul & Opiela, 2010).
There are still few studies about this issue for developing countries with relatively
high inflation rates. In the case of Vietnam, there are still no published studies about
this topic. Therefore, this paper contributes to the literature as one of the first
comprehensive analysis about this issue in the Vietnamese case.
Second, the evidence from this study is informative and useful for monetary
authorities so that they can understand the nature of the inflation-inflation
uncertainty relationship in Vietnam during the past sixteen years empirically. As a

result, they have reliable foundation to propose and implement policies to control
inflation better.

1.7 Scope of the Study
The study will investigate the inflation-inflation uncertainty relationship in
Vietnam during the period 1995-2010.

1.8 Organization of the Study
The remaining of the paper is structured as follows: Chapter 2 gives a review
of definition, consequences, and methods to model inflation uncertainty. In
addition, theories, empirical studies about the inflation-inflation uncertainty
relationship and methodology to test this relationship are also presented. Chapter 3
4


presents the research methodology. Chapter 4 reports the findings and discussion.
Chapter 5 presents the conclusion, suggests some practical policy implications, and
discusses the limitation and direction for further studies.

5


Chapter 2: Literature Review
2.1 Inflation Uncertainty
Definition
It is important to understand clearly the concept “inflation uncertainty”
before going further with the study. Inflation uncertainty is defined as “a situation in
which future prices are unpredictable and general public does not know whether
inflation will increase or decrease in the future. In simple words future inflation rate
is fickle to public” (Asghar et al., 2011, p. 86).

Economic consequences of inflation uncertainty
According to Golob (1994), inflation uncertainty affects the economy in two
ways: ex-ante and ex-post effects.
Ex-ante effects refer to the situation in which economic agents make
economic decisions different from their decisions in the case of no inflation
uncertainty. There are three channels through which ex-ante effects transmit to the
economy. Firstly, inflation uncertainty makes long-term interest rates increase in the
financial markets. Secondly, inflation uncertainty makes other economic variables
(future wages, future rents, tax rates) which are significant for the decisions of
businesses become uncertain. Lastly, inflation uncertainty induces businesses to
spend large resources to avoid the risk of future inflation.
Ex-post effects happen when inflation in reality is different from the
expected one. As a result, unexpected inflation causes a transfer of wealth among
different sides when they use nominal money for the settlement of payments in the
contract. In particular, if inflation rises more than anticipated, the contract payments
will be less than expected in terms of their real value.
Why rising inflation may lead to an increase in inflation uncertainty
It is widely acknowledged that the Friedman-Ball hypothesis holds for all
countries with different rates of inflation (discussed deeper in the empirical study
6


section). The explanation for this relationship is that monetary policies may cause
inflation uncertainty due to uncertainty about the timing and short-run effects of
these policies on inflation.
First, there is uncertainty about the timing of controlling-inflation policies
due to short-run tradeoffs when implementing tightening monetary policy. Although
the central bank always aims at the long-run goal of controlling inflation, it may
fear of the short-run downturn of the economy. If inflation happens in the period of
poor economic growth, it is uncertain which priority is: fighting inflation or

supporting economic growth. As a result, there is rising uncertainty about the time
when the central bank implements contractionary monetary policy to lower inflation
(Golob, 1994).
Second, as pointed out by Holland (1993), in the case of inflation, there is
still uncertainty about the effect of monetary policy on inflation even though the
tightening monetary policy is guaranteed. Specifically, it takes time for the policy to
transmit its impacts to the banking sector, then to the real economy, and finally to
inflation. Furthermore, it is difficult to estimate the extent of the price level’s
response to the monetary policy. Hence, the complicatedness of forecasting the
speed and extent of the price level’s response to the monetary policy brings about
inflation uncertainty (as cited in Golob, 1994).

2.2 Theories about the inflation-inflation uncertainty relationship
Friedman (1977) proposes a framework for the relationship between inflation
and inflation uncertainty. He forms an “informal two-part argument about the real
effects of inflation” (as cited in Asghar et al., 2011, p. 88). In the first part, when
inflation increases, there will be political pressure from the public and voters that
force policy makers to reduce inflation. However, the policy makers may not
implement contractionary monetary policy to lower inflation because of the fear of
recessionary impacts. As a consequence, the future monetary policy is
unpredictable, which causes inflation uncertainty to increase in the future. In the
7


second part, Friedman argues that this increasing inflation uncertainty influences the
economic growth of the country negatively (as cited in Thornton, 2007; Jiranyakul
& Opiela, 2010; Asghar et al., 2011).
Ball (1992) presents the first part of Friedman’s argument formally in the
framework where the general public faces the asymmetric information about policy
makers. He categorizes two types of policymakers: tough and soft. Tough

policymakers are willing to disinflate. However, the soft type is afraid of the
recessionary effects of disinflation. The public do not know exactly which type of
policymakers will be in power, which causes uncertainty about future inflation.
Hence, the hypothesis presented by Friedman (1977) and Ball (1992) is named
Friedman-Ball hypothesis in economic literature (as cited in Kontonikas, 2004;
Nazar & Mojtaba, 2010).
On the other hand, the inflation-inflation uncertainty nexus is considered in
the opposite direction by Cukierman and Meltzer (1986). They argue that an
increase in inflation uncertainty leads to rising inflation. The reason is that the
monetary authority may take advantage of this inflation uncertainty to make
inflation surprise (expansionary monetary policy) to stimulate economic growth.
This type of central bank is regarded as an “opportunistic” one (as cited in Fountas,
Ioannidis & Karanasos, 2004; Thornton, 2007).
Holland (1995), in contrast to Cukierman and Meltzer (1986), argues that
rising inflation uncertainty leads to lower inflation in the future. The reason is that
the central bank would like to minimize the welfare cost of more inflation
uncertainty; thus it carries out tightening monetary policy to lower inflation
uncertainty. This type of central bank is considered a “stabilizing” one (as cited in
Thornton, 2007 Erkam & Cavusoglu, 2008).
For decades, there have been many empirical studies to examine whether
rising inflation causes increasing inflation uncertainty (Friedman-Ball hypothesis),
rising inflation uncertainty causes increasing inflation (Cukierman-Meltzer
hypothesis), or rising inflation uncertainty causes lower inflation (Holland
8


hypothesis). These studies use different methods and focus on different countries
with different sample periods, frequency of data sets. The following part reviews
the methods to estimate inflation uncertainty from simple to more complex ones.

2.3 Approaches to estimate inflation uncertainty
Engle (1982) is the first economist who develops Autoregressive Conditional
Heteroskedasticity (ARCH) model to estimate inflation uncertainty. He employs the
standard inflation model as an Autoregressive - AR(p). The ARCH model is
proposed to model and forecast the conditional variance. The conditional variance is
estimated on a time-varying basis. In the ARCH model, the variance equation is a
function of past squared forecast errors. This variance can be used as a proxy for
inflation uncertainty (as cited in Nazar & Mojtaba, 2010; Asghar et al., 2011).
Mean inflation equation
p

 t   0    i t  i   t
i 1

(AR(p) process) where  t is assumed to be ~ N (0,  2 )

Variance equation
q

ht   0    j t2 j
j 1

(1)

Where   0 ,  j  0 so that the conditional variance is positive
The disadvantage of the ARCH model widely pointed out in empirical
studies is that it shows long lag processes of the squared forecast errors (Bollerslev,
1986). To model the persistent effect better, economists have developed some
extensions of the ARCH framework. Bollerslev (1986) introduces the Generalized
Autoregressive Conditional Heteroskedasticity (GARCH) model, in particular the

GARCH(1,1) model. In the GARCH(1,1) model, the conditional variance is a
function of past value of the forecast error and its own lagged value.
GARCH (1,1)
ht  0  1 t21   ht 1

(2)

9


Where α0 > 0 1  0,   0 so that the conditional variance is non-negative
α1+β <1 for a variance stationary model
However, Brunner and Hess (1993) argue that the ARCH and GARCH
models impose symmetric restriction on the response of inflation uncertainty to
inflation shocks, which seems to be “inconsistent with the Friedman’s hypothesis”.
Specifically, according to Friedman (1977), the inflation-uncertainty nexus is
defined as “the higher the rate, the more variable it is likely to be”, which means
that a rise in inflation causes more inflation uncertainty, while a fall in inflation
leads to less uncertainty. Meanwhile, the conventional ARCH or GARCH models
put the ε2 in the conditional variance equation, implying that positive shocks to
inflation cause inflation uncertainty at the same extent as negative shocks; thus they
bias the Friedman’s hypothesis (Rizvi & Naqvi, 2008;. Jiranyakul & Opiela, 2010).
Therefore, some variations of the GARCH model are proposed to model the
asymmetry characteristics of the conditional variance; three most popular ones are
TARCH, PARCH and EGARCH models.
Glosten, Jagannathan, and Runkle (1993) and Zakoian (1994) propose the
threshold ARCH (TARCH) model which can capture the asymmetric effect of
positive and negative shocks on volatility. The specification of the TARCH(1,1)
model is as follows
TARCH(1,1)

ht  0  1 t21   ht 1   dt 1 t21

(3)

where
dt 1  1 , if  t 1  0
d t 1  0 , if  t 1  0

In equation (3), γ is considered the asymmetric or leverage term. If γ = 0, the
model becomes the conventional GARCH model. In the case   0 , the effect of
negative inflation shocks on uncertainty is represented in α1+γ; the effect of positive
inflation shocks on uncertainty is represented in α1. Thus, if γ is significant and
10


negative, positive inflation shocks have a larger impact on inflation uncertainty than
negative ones (as cited in Caporale & Caporale, 2002; Kontonikas, 2004).
Another model to capture the asymmetric effect of positive and negative
inflation shocks on volatility is the asymmetric power ARCH (PARCH) model
proposed by Ding et al. (1993). The specification of the PARCH(1,1) model is as
follows.
PARCH(1,1)
( ht )   0  1 (  t 1   t 1 )   ( ht 1 )

(4)

In equation (4), γ denotes the asymmetric impact of inflation shocks on
inflation uncertainty. If γ is negative, positive shocks raise the inflation uncertainty
more than negative ones, and vice versa (Daal, Naka, & Sanchez, 2005). δ
represents the power transformation parameter.

Nelson (1991) proposes the EGARCH model which is preferable in
investigating the asymmetric inflation-inflation uncertainty relationship. The
EGARCH(1,1) model is formulated as follows:
EGARCH(1,1)
log ht   0  1

 t 1
ht 1



 t 1
ht 1

  log ht 1

(5)

According to Nelson, the EGARCH process constructs the conditional
variance in logarithm (loght); thus even if the parameters are negative, ht in (5) is
still positive. Therefore, this specification overcome the requirement of artificially
non-negative parameters in the GARCH model (as cited in Rivzi, 2008; Jiranyakul
& Opiela, 2010).
As shown in (5), γ is considered the asymmetric or leverage term. If   0 ,
the impact of inflation shocks on inflation uncertainty is asymmetric. Specifically,
when  is positive, a positive inflation shock leads to more inflation uncertainty
and vice versa (Asghar et al., 2011). This explanation is also identical to Friedman’s
argument of inflation-inflation uncertainty nexus (Brunner & Hess, 1993).
11

2.4 Methods to test the causal relationship between inflation and
inflation uncertainty
There are commonly two approaches to test the causal relationship between
inflation and inflation uncertainty in the literature. The first one is the two-step
approach. The first step is to estimate the conditional variance by ARCH, GARCH
or variations of GARCH models, which is used latter as the proxy for inflation
uncertainty. Then Granger causality tests are performed to examine how the two
variables related to each other. There is a popular use of this approach in examining
the inflation-inflation uncertainty relationship. Some major studies include Fountas,
Ioannidis & Karanasos (2004), Daal, Naka & Sanchez (2005), Thornton (2007),
Jiranyakul & Opiela (2010), Asghar et al. (2011), etc.
However, Pagan (1984) points out the problem with this two-step approach
is that generated variables in the first step are employed as the regressors in the
second step. Consequently, the standard errors in the Granger tests are biased,
suggesting the biasness of Granger tests’ results (as cited in Caporale & Kontonikas,
2009; Baharumshah, Hasanov & Fountas, 2011).
The second approach is the simultaneous estimation approach: estimating
inflation and inflation uncertainty simultaneously using a bivariate GARCH-inmean (GARCH-M) framework. Specifically, we will integrate the conditional
variance in the mean inflation equation and the inflation in the conditional variance
equation. Some major studies using this approach include Kontonikas (2004),
Thornton (2008), Keskek & Orhan (2010), etc.
This approach is considered more efficient than the two-step approach
because it helps eliminate the issue of generated repressors (Pagan, 1984).
Nevertheless, the disadvantage of this approach is that it does not permit the
checking of impacts of lagged inflation on inflation uncertainty or vice versa, which
is supposed to last for several periods. Consequently, this limits the possibility of

12

finding the causality between the two variables (Fountas et al., 2004, Caporale &
Kontonikas, 2009, Jiranyakul & Opiela, 2010).
Because the aim of this study is to investigate the causality between these
two variables, the two-step approach seems to be more suitable for this purpose.

2.5 Empirical studies about the inflation-inflation uncertainty
relationship
The early empirical studies about the link between inflation and inflation
uncertainty include the ones by Engle (1982), Engle (1983), and Bollerslev (1986).
Engle is the first researcher to employ the ARCH model to estimate inflation
uncertainty. In Engle (1982), based on UK CPI from the second quarter of 1958 to
the second quarter of 1977 and the ARCH model, a significant ARCH effect is
found. Moreover, inflation uncertainty rises substantially in the 1970s. However, no
evidence is found for the inflation and inflation uncertainty relationship.
Following this method of study, Engle (1983) examines the US CPI in 19471979 using the similar ARCH framework. A significant ARCH effect is found with
inflation uncertainty being higher in 1970s than in 1960s. Nevertheless the inflation
uncertainty in these periods is relatively low compared with the 1940-1950 period.
Additionally, the inflation-inflation uncertainty nexus is still not found.
Following Engle (1982), Engle (1983), Bollerslev (1986) applies the
improvement of the ARCH model - GARCH model, in particular GARCH(1,1)
model to estimate inflation uncertainty based on US CPI and US GNP deflator
1948Q2 - 1983Q4. He also finds a significant ARCH effect. Even though the
GARCH model is better at modeling US inflation data than the ARCH one, there is
still no evidence for the inflation-inflation uncertainty relationship.
Brunner and Hess (1993) point out two reasons why previous studies by
Engle (1982), Engle (1982), and Bollerslev (1986) cannot find the inflationinflation uncertainty relationship. First, the degree of inflation is not included in the
conditional variance model in these studies. Second, they use symmetric ARCH and
13

GARCH models which do not conform to Friedman-Ball hypothesis. Although no
evidence is found for the inflation-inflation uncertainty relationship, these studies
set the fundamental foundation in terms of methodology for subsequent studies.
Subsequent studies mainly use the two above-mentioned econometric
approaches (section 2.4) to examine the inflation-inflation uncertainty relationship.
As for the simultaneous estimation approach: estimating inflation and
inflation uncertainty simultaneously using a bivariate GARCH-in-mean (GARCHM) framework, in their major work, Caporale and McKiernan (1997) apply the
GARCH-M framework in US inflation data after the World War II from 1947M1 to
1994M8, they intend to examine whether there exists a positive association between
the degree of inflation and its variability. The findings indicate that the magnitude
of inflation is positively related to its variability, affirming the hypothesis of
Friedman-Ball.
Another study employing the GARCH-M models is Fountas (2001). Using
UK inflation data from 1885 to 1998, he also finds strong confirmation of the
Friedman-Ball hypothesis.
Kontonikas (2004), employing British data from 1972 to 2002 and GARCHM, threshold GARCH-M, and component GARCH-M models, provides evidence to
support the Friedman-Ball hypothesis. Furthermore, it is found that the
implementation of an explicit inflation targeting helps to lower inflation constancy
and long-run inflation uncertainty.
Thornton (2006) uses the GARCH-M framework to examine the connection
between inflation and uncertainty in South Africa during 1957M1-2005M9. The
result shows a positive and significant link between inflation and uncertainty in
South Africa during this period, which is in agreement with the Friedman's
hypothesis. Following the similar methodology, using the GARCH(1,1)-M model,
Thornton (2008) claims that the Friedman hypothesis holds for Argentina during
1810-2005.

14

Keskek and Orhan (2010), employing Turkish inflation series during
1984M1-2005M10 and different categories of GARCH-M models, find that
increasing inflation leads to more inflation uncertainty. Additionally, Holland
hypothesis is confirmed because the monetary authority shows strong determination
of stabilizing the economy. Similar to Kontonikas (2004), there is evidence that
explicit inflation targeting policies help to lower inflation constancy and inflation
uncertainty.
The second approach: two-step approach has been widely used in a great
number of empirical studies for its advantages which have been mentioned above.
In this study, I categorize the studies applying the two-step approach into two
groups: the first one using ARCH or GARCH model with symmetric inflation
uncertainty, the second one using some variations of the conventional GARCH
model with asymmetric inflation uncertainty. The reason for this distinction is the
importance of modeling the asymmetric effects of inflation shocks on inflation
uncertainty which has been explained in the previous section.
The following section presents a brief summary of major studies using
conventional ARCH and GACRH models with symmetric inflation uncertainty.
Grier and Perry (1998) use the GARCH model to measure the inflation
uncertainty in G7 countries from 1948 to 1993. The Granger causality test is
employed to test the causal relationship between inflation and inflation uncertainty.
There is evidence for Friedman-Ball hypothesis in all the countries. However,
evidence for Cukierman-Meltzer hypothesis is mixed. An increase in inflation
uncertainty causes higher inflation in Japan and France, which is consistent with
Cukierman-Meltzer hypothesis. Meanwhile, rising inflation uncertainty leads to
lower inflation in US, UK and Germany, which supports the Holland hypothesis.
Thornton (2007) employs the GARCH (q,v) model to generate the inflation
uncertainty estimates of twelve emerging countries with different time periods
ranging from the 1950s to 2000s. Friedman-Ball hypothesis is supported in all
twelve countries based on Granger test results. Evidence for Cukierman-Meltzer

15


hypothesis is found in Hungary, Indonesia, and Korea where increasing inflation
uncertainty causes average inflation rates to rise. In the meanwhile, the Holland
hypothesis is supported in Colombia, Israel, Mexico, and Turkey.
In the following section, the symmetric restriction in the ARCH and GARCH
models is treated with some variations of the conventional GARCH model to
capture the asymmetric inflation uncertainty in the following major studies.
Fountas et al. (2004) investigate this relationship in six European countries
(France, Germany, Italy, the Netherlands, Spain, and the UK) from 1960 to 1999 by
both the two-step approach and the simultaneous estimation approach in an
EGARCH-in-mean (EGARCH-M) framework. For the first approach, the Granger
tests show that the Friedman-Ball hypothesis is supported in all countries apart from
Germany. Nevertheless, apart from the UK, inflation uncertainty is found not to lead
to negative output impacts in other countries. It indicates that the common European
monetary policy may cause asymmetric real effects (output) through the channel of
inflation uncertainty. As for the relationship going from uncertainty to inflation, the
results are less robust. The Cukierman-Meltzer hypothesis is confirmed in Italy,
Spain, and France. Meanwhile, the Holland hypothesis holds in Germany and the
Netherlands. The second approach also confirms the Friedman-Ball hypothesis in
all countries except the Netherlands and Germany, showing the consistency of the
two approaches. As for the relationship in the opposite direction, significant impacts
of inflation uncertainty on inflation are not found. This result is reasonable because
it takes time for the effects of inflation uncertainty to materialize in inflation; thus it
is difficult to fairly examine this relationship in a contemporary model like
EGARCH-M.
Daal, Naka, and Sanchez (2005), using the power GARCH model, analyze
the inflation-inflation uncertainty relationship in both developed (G7 countries) and
emerging countries (Asian, Latin America, Middle East). They conclude inflation

Granger-causes inflation uncertainty in most both developed and emerging

16


countries, firmly confirming the Friedman–Ball hypothesis. However, there is
mixed evidence for causality running in the opposite direction.
Erkam and Cavusoglu (2008) study the inflation-inflation uncertainty
relationship in seven transitional countries (Ukraine (1997M1-2007M5), Azerbaijan
(1996M1-2004M9), Armenia (1996M1-2007M3), Georgia (1997M1-2006M2),
Kazakhstan (1997M1-2007M5), Kyrgyz Republic (1997M1-2007M4), and Russia
(1997M1-2007M4)) using various forms of ARCH and GARCH models. In order to
test the causal linkage, both Granger causality tests and Holmes-Hutton tests are
employed. The evidence for the Friedman-Ball hypothesis is found in Ukraine,
Azerbaijan, Russia. The Cukierman-Meltzer hypothesis is supported in Russia and
Kyrgyz Republic. The Holland hypothesis holds in Azerbaijan, which indicates the
effective monetary stabilization policy of the central bank.
As for Asian countries, Payne (2009) employs the ARIMA–GARCH(1,1)
model to examine the inflation-inflation uncertainty relationship in Thailand during
the period 1965M1-2007M3. The results indicate that the execution of inflation
targeting in Thailand since 2000 lowers the impacts of inflationary shocks on
inflation volatility persistence. Similar to previous studies, the Granger causality
tests confirm the Friedman-Ball hypothesis. Nonetheless, a rise in inflation
uncertainty leads to a fall in inflation, supporting the Holland hypothesis.
Another study applying the two-step framework to test this relationship for
Asian countries is Jiranyakul and Opiela (2010). Using the data of monthly CPI in
ASEAN-5 countries (Malaysia, Indonesia, the Philippines, Singapore, and
Thailand) during the period 1970M01–2007M12 and the AR(p)-EGARCH(1,1)
model, it is concluded that there is a significant support for both the Friedman-Ball
and Cukierman-Meltzer hypothesis in all countries. For Thailand’s case, the

Cukierman-Meltzer hypothesis is supported, not the the Holland hypothesis (as in
Payne, 2009) irrespective of the similar time span. It is the difference in the method
to model inflation uncertainty (ARIMA–GARCH model in Payne (2009) compared

17