Decision (March 2020) 47(1):79–90

https://doi.org/10.1007/s40622-020-00237-7

RESEARCH ARTICLE

An empirical study of trade openness and inflation in India

Megha Chhabra . Qamar Alam

Published online: 1 April 2020

Ó Indian Institute of Management Calcutta 2020

Abstract The nexus between inflation and trade Keywords Inflation  Romer’s hypothesis  Open
openness has been one of the major concerns among economy  Time series data
the researchers and policy makers in both developed
and developing economies. The exact relationship JEL Classification F41  E31  C22
between inflation and openness is still in ambiguity.
Various studies have been done for different countries
and regional groups using different methodologies to
Introduction
measure the relationship between openness and infla-
tion, but all remained in vain. Although there are
Trade is the way to success that helps countries to
different views regarding this relationship, most of the
grow with the speed of lightning. It is one of the best
empirical studies supported the Romer’s hypothesis
choices to be opted by economies to eradicate the
(Q J Econ 108(4):869–903, 1993) of an inverse
socio-economic problems like poverty, unemploy-
association between openness and inflation in the
ment, etc. (World Bank). Openness towards trade
economy. The present study has made an attempt to
generally reflects the integration of an economy with
examine the factors influencing the price level,
other economies, commonly done by lowering the
especially trade openness, in India from 1974–1975
import tariffs. However, a simultaneous increase in
to 2015–2016. For empirical investigation, the study
non-tariff barriers has also been observed restricting
has employed autoregressive distributed lag (ARDL)
the flow of trade.
model bounds testing approach to co-integration
The trade at the global level offers the diversifica-
(Pesaran et al., J Appl Econ 16(3):289–326, 2001).
tion of products available at the lower prices and
The obtained results revealed the presence of positive
provides an opportunity to deviate resources towards
relationship between inflation and trade openness in
the production of the goods in which the country is
India, which negates the Romer’s hypothesis.
highly efficient. It lends a helping hand for the
economies to develop economically and socially,
M. Chhabra and India is one of those economies. Enormous
Department of Economics and Finance, Birla Institute of
reforms carried out in various sectors of the economy
Information and Technology, Pilani, India
in the 1990s resulted in better integration of India with
Q. Alam (&) other world economies. A significant rise in global
Department of Economics, Banasthali Vidyapith, finance and trade activities has been observed in India
Rajasthan 304022, India
since the start of the liberalization process in the 1980s
e-mail: alamqamar58@gmail.com

123
80 Decision (March 2020) 47(1):79–90

that picked up pace in the early 1990s. On average, the 2002). Many economists have believed that the
trade of India as percentage of GDP has increased growing competitiveness reduces the inflation rate
from 11.53% in the period of 1961–1991 to 35.19% in (Rogoff 1985), which has been observed in case of
1991–2015. In the same period where the trade many countries. Various studies have been conducted
openness has increased, the inflation rate in India has to measure the relationship between openness and
come down from an average of 8.1–6.5% (World inflation for different countries and regional groups
Bank). This raises an important question whether two using different methodologies and proxy variables, but
of the events are related and affect each other. the exact relationship between inflation and openness
Commentators globally analyse the effect of open- is still in ambiguity. The empirical results and findings
ness on inflation with the help of the Phillips curve of the relationship differ, and sometimes, they are
also. Recently, it has been observed that globalization contradictory. The findings on this subject from the
flattens the Phillips curve, which means the inflation in literature can be categorized under five parameters:
home country is affected more by the change in other openness positively affects inflation (Triffin and
economies and less by the changes in output produced Grubel 1962; Mahmoudzadeh and Shadabi 2012;
domestically. Contrary to this, Romer (1993) claimed Munir and Kiani 2011; Sepehrivand and Azizi
a steeper Phillips curve in open economies which leads 2016), openness affects inflation negatively (Romer
to lesser inflation rates. However, there are many 1993; Temple 2002; Gruben and Mcleod 2004;
pieces of evidence supporting the negative relation- Bowdler and Nunziata 2006; Sachsida and Mendonça
ship, but the evidence for positive relationship also 2006; Kurihara 2013), openness affects economy only
cannot be ignored. in the short run (Bleaney 1999; Gruben and Mcleod
In the present scenario, most developing and 2004; Nasser et al. 2009), openness affects economy
emerging economies trade globally, which leads to only in the long run (Lane 1997; Sachsida and
concern about the rapid inflation acting as a serious Mendonça 2006; Munir and Kiani 2011), and there
obstacle in the economic development process. In the is no regular relationship (Iyoha 1973; Wu and Lin
backdrop of this, the present study focuses on the trade 2008; Aliyev and Gasimov 2014). The current section
openness as the tool to control inflation while exam- of the study analyses the available pieces of evidence
ining the relationship between them over the period of related to Romer’s hypothesis stating the inverse
1974–2015. The paper tries to analyse the existence of relationship between trade openness and inflation.
Romer’s hypothesis in India that states negative
relationship between the trade openness and price Romer’s hypothesis
levels in the economy. The paper further briefly
explains theoretical and empirical literature review Romer’s (1993) hypothesis reflects that the more the
and mentions about the Romer’s hypothesis in the open economies, the less the inflation rate under
second section. The third and fourth sections cover the discretionary policy making. Romer pointed out that if
hypotheses tested and the dataset and methodology no monetary policy was made in advance, the nation
employed. The empirical estimations have been would experience a high rate of inflation. In the
summarized in the fifth section, followed by the absence of precommitment in monetary policy, an
conclusion in the sixth section. unexpected monetary expansion would lead to real
exchange rate depreciation that may result in higher
inflation in more open economies. The hypothesis
Review of literature demonstrates that the benefits of surprise monetary
expansion (i.e. extent of inflation) will be less in more
Fixing the inflation rate of the economy with the open economies and vice versa.
simultaneous objective of economic growth has Theoretically, Romer found a larger effect of
always been an important part of monetary policy. monetary expansion on domestic inflation and CPI
Rapid inflation has been a major point of concern for inflation and a smaller effect on output in more open
the policy makers as it can affect the economic growth economies. He considered that the change in respec-
due to the uncertainty and is considered as an tive variables (domestic inflation, CPI inflation and
economic puzzle, which needs to be solved (Temple output) does not depend on the degree of openness.

123
Decision (March 2020) 47(1):79–90 81

The effect of monetary expansion on the real exchange openness to trade has been reviewed, whereas evi-
rate is also considered as independent of openness. dence for existence of respective relationship has been
However, a higher degree of openness leads to higher studied in ‘‘Empirical studies’’ section.
welfare cost of depreciation of real exchange rate in
more open economies. The reason is the higher degree Theoretical studies
of openness that allows import of more goods.
If monetary expansion takes place in the presence The literature on inflation in the developing economies
of sticky prices in the economy, the incremental has been heavily commented. But despite many years
change in domestic output will be greater than the of research, the relationship of inflation with other
change in foreign output. It will create a need for the variables of the economy has remained in argument.
relative fall in prices of domestic goods to encourage The monetarist Milton Friedman rightly said ‘‘In-
the purchase of domestic goods. Such a relative fall in flation is always and everywhere a monetary phe-
domestic prices creates real depreciation in the nomenon’’ (1963, p17). The economists at the Chicago
economy that creates two channel effects. First, the School assumed a stable relation between the money
prices of foreign goods are part of the CPI inflation; stock and inflation. They favoured a rule-based
thus, depreciation directly causes the increase in CPI monetary policy to control the inflation as fiscal
inflation. Second, it also increases the cost of the deficits lead to change in money supply, causing
domestic firms. For instance, if there are flexible inflation. Accordingly, a reduction in the fiscal deficit
nominal wages in the market, then increased inflation of the country will bring down its inflation rates.
rates may cause demand for high wages. The incre- In contrast to above, Kydland and Prescott (1977)
ment in the prices of flexible-price firms can be quite classically demonstrated the existence of an incentive
larger than what has been done without depreciation, to the policy makers to create surprise inflation under
which raises the inflation rate in the economy. imperfect competition, in case monetary expansion
Therefore, in a more open economy, the higher the can affect real output. But the expansion can be done
fraction of foreign goods in CPI inflation, the larger up to the limit to the price and wage-setters’ expec-
the impact on CPI inflation and domestic inflation will tations. This implied that the absence of prior
be. Consequently, monetary expansion gets trans- commitments in monetary policy may lead to ineffi-
formed into higher prices and higher disproportionate ciently high inflation with output being at a suboptimal
output. level. Barro and Gordon (1983) supported this view,
In other scenarios where the size of the home who stated that the Kydland and Prescott’s model
country is smaller than that of the foreign country, the presents significant literature for the phenomena.
impact of monetary expansion in the foreign output is Taylor (1983), on the other extreme, stated that the
smaller. It is because a little appreciation in the existence of some institutions and mechanisms in the
currency of the foreign country creates a very little economy phases out the tendency of policy makers for
impact on the prices of foreign firms and hence little systematic attempts, which causes surprise inflation.
impact on output of foreign goods. Therefore, an Rogoff (1985) noted that the surprise monetary
increase in openness may result in reduced impact of expansion leads to the depreciation of the real
monetary expansion at home and foreign output. exchange rate of the economy that may cause the
This throws light on the channels through which reduced incentives for the expansion. Coordination of
trade openness affects inflation which includes the international policies of two countries will be unde-
effect of non-sticky prices, sticky prices and varying sirable in the absence of any prior commitment in
size of country. The model robustly states that high monetary policy as the coordinated policies of two
degree of openness worsens the trade-off between countries will raise the inflation. The policies coordi-
output and inflation and consequently lowers the nated reduce the incentive to expand as surprise
equilibrium inflation rate. monetary expansion cannot cause the currency of one
The literature review for the same has been divided country depreciating against others. Therefore, there is
into two sections that include the theoretical studies negative relationship between openness and inflation.
and empirical studies. In ‘‘Theoretical studies’’ sec- Romer (1993) tested the central element of these
tion, the theoretical linkage between inflation and models, that is, the absence of any prior commitment

123
82 Decision (March 2020) 47(1):79–90

in monetary policy provides incentives to inflate. to examine the relationship and revealed the existence
Romer’s (1993) hypothesis reflects that more open of positive relationship, contradictory to the Romer’s
economies have less inflation rate under discretionary results. In support to this, Mahmoudzadeh and Shadabi
policy making. It points out that in case no monetary (2012) verified the linkage and found a positive effect
policy was made a priori, the nation would experience of trade, freedom of employment and operation
a high rate of inflation as the unexpected monetary productivity using generalized method of moments
expansion would lead to real exchange rate depreci- (GMM) and trade freedom index by Heritage Foun-
ation that may result in higher inflation in more open dation Index.
economies. The benefits of surprise monetary expan- Sikdar et al. (2013) favoured the outward-looking
sion will be less in more open economies; however, policy to trade for the long run as it creates the
the benefits will be more in less open economy. opportunities to grab the possible advantages of
international trade and capital flows. The results also
Empirical studies indicated that any shock in short-run inflation will
gradually get adjusted in the long run.
Triffin and Grubel (1962) examined the hypothesis for Ada et al. (2014) investigated the relationship
five countries in the European Economic Community. between trade openness and inflation using VECM
They found that cheaper availability of goods and approach and verified the negative link between
services acts as a safety means towards the pressure of variables. Similarly, Haq et al. (2014) examined the
rising prices in the economy. Using Barron–Gordon impact of economic growth, supply of money and
model, Romer (1993) confirmed the existence of a trade openness on inflation for the small open
significant negative relationship between trade open- economies. The study showed that excess money
ness and inflation in the developing countries only and supply will reduce the incentive for government to opt
mentioned that the political stability and independence expansionary monetary policy and thus curb the
of central banks may weaken the relationship. How- inflation in small open economies. Salimifar et al.
ever, Terra, (1998) challenging Romer’s empirical (2015) conducted a time series study using ARDL
findings, demonstrated that the impact on inflation method. It mentioned that the higher degree of
varies with the change in the indebtedness level of the openness raises the competition among domestic
country. The study found validity of hypothesis for producers in the international market with high-
severely indebted countries in the debt crisis period. quality products. Also, the study found any increase
Nasser et al. (2009) rejected Terra’s argument. It in inflation can be controlled by reducing trade barriers
argued that the negative relationship also exists for the and improving relations with other countries. Sepehri-
countries having pre-debt crisis. Temple (2002) pro- vand and Azizi (2016), considering the influence of
vided a little support to Terra (1998) by establishing a monetary policy on the international market, investi-
link between the Phillips curve and trade openness. gated the effect of the globalization, which showed
The study claimed a more inclined Philips curve for that the high influence of monetary policy leads to
more open economies. swings in consumption demand for domestic goods.
Through dynamic panel estimation, Gruben and The lack of competition and currency fluctuations
McLeod (2004) revealed the existence of Romer’s leads to positive impact on inflation and does not
hypothesis particularly for the economies that have confirm Romer’s hypothesis.
adopted floating exchange rate system in the 1990s. The literature review demonstrated for various
However, Wu and Lin (2008) assessed that robustness countries and group of countries suggests that there is
of results depends whether the analysis is carried out no clear vision of trade openness as a means to
with and without constraints in the model. mitigate inflation. Therefore, the present study demon-
Using the Phillips curve, Eijffinger and Qian (2010) strates the impact of increasing trade openness on
examined the hypothesis for the inconsistency of time inflation and tries to solve the problem of inflation, if
period for the long run. The inflation is affected affected adversely by the trade integrity in Indian
through the channels other than the Phillips curve and context. Moreover, it also tries to verify the validity of
found to be different for different countries. Later, Romer’s hypothesis for India.
Munir and Kiani (2011) used a single equation model

123
Decision (March 2020) 47(1):79–90 83

Hypotheses of the study

where WPIAt shows the annual variation in WPI, FDt
means central government fiscal deficit, ERt means
The main and null hypothesis to be tested under the
rupee to dollar exchange rate, TPGDPt is trade as
study is that trade openness does not affect the
percentage of GDP and ITOTt stands for income terms
inflation rate in India. The testing of the sub-hypothe-
of trade, for the time period t. The noise error term of
ses includes the following:
time period t in the model is denoted by et. Eq. (1)
H1 Fiscal deficit does affect the WPI in India. shows the positive sign expected for the variables FD,
TPGDP and ITOT. There is a different rationale
H1 Trade to GDP ratio does affect the WPI in India.
behind the selection of respective variables to analyse
H1 Exchange rate does affect the WPI in India. the relationship between trade openness and inflation.
First, an improved term of trade under flexible
H1 Income terms of trade do affect the WPI in India.
exchange rate system creates a deflationary pressure
when exchange rate response turns higher than the
suggested threshold limit of exchange rate and there-
Dataset and methodology
fore is expected to show a positive sign (Gruen and
Shuetrim 1994; Gruen and Dwyer1995). Second, the
This study tries to examine the relationship between
ER is expected to show the negative sign in interaction
trade openness and inflation for India. The analysis
with WPIA as it may improve the inflationary
incorporates the annual time series data for a period of
conditions via pass-through, i.e. exchange rate acts
42 years, i.e. from 1974–1975 to 2015–2016. The
as a shock absorber. The decreasing pass-through
secondary data sources have been used to fetch data.
affects the changes to the inflation (Edwards 2006).
These include various yearly issues of Reserve Bank
Third, the fiscal deficit raises the prices in the market
of India (RBI) report, namely ‘‘Handbook of Statistics
and shows positive association with inflation. The
on the Indian Economy’’. Moreover, the data are also
process of seigniorage boosts prices in the economy
obtained from the online available sources of Interna-
with the rise in deficits (Sargent and Wallace 1981).
tional Monetary Fund (IMF). The variables used are in
The verification for the existence of Romer’s Model
the form of their log values. The variables used to
expects these to come true for proving trade openness
examine the relationship between trade openness and
as an effective tool to control inflation. In order to
inflation are listed in Table 1.
examine the hypothesis, the study employs the lately
popularized technique of autoregressive distributive
Model specification
lag (ARDL) model of co-integration.
The time series data considered for accomplishing the
ARDL co-integration
study objectives have been collected to carry out the
analysis. The variables used in the basic model are as
In applied econometrics, the Granger (1981) and
follows:
Engle and Granger (1987) autoregressive distributed
WPIAt ¼ b0 þ b1 FDt  b2 ERt þ b3 TPGDPt lag (ARDL) co-integration technique or bound test of
þ b4 ITOTt þ et ð1Þ co-integration (Pesaran et al. 2001), and Johansen and
Juselius (1990) co-integration techniques have
become the solution to determine the long-run

Table 1 Variables used in Variables Measurement unit of proxy variables Type of variables
the study. Source: Author’s
computation Central government fiscal deficit In Rs. billion Independent
Exchange rate Rupee to US dollar Independent
Trade to GDP ratio In percentage Independent
Income terms of trade In percentage Independent
Annual variation in WPI In percentage Dependent

123
84 Decision (March 2020) 47(1):79–90

relationship between time series that are non-station- shows the more stability towards the long-run equi-
ary, in addition to reparameterizing them to the error librium. The equation for the model with lagged
correction model (ECM). The reparameterized result variables can be specified as follows:
gives the short-run dynamics and long-run relationship
of the underlying variables. However, the study WPIAt ¼ b0 þ ai DWPIAt1 þ ci DFDt1 þ di DERt1
incorporates the use of ARDL co-integration tech- þ xi DTPGDPt1 þ si DITOTt1
þ kECMt1 þ et
nique. Autoregressive technique means along with the
exogenous variables, endogenous variable is also ð3Þ
explained by its own lag. ARDL model empirically The coefficients of ECMt-1, i.e. k, should be
analyses the long-run and short-run relationships and negative and statistically significant to balance the
dynamic interactions between the variables. The differences between the exogenous and endogenous
traditional ARDL model (earlier Pesaran et al. 2001) variables. The hypothesis to be tested under ARDL
did not support the variables integrated at the order one bound test approach is as follows:
(I(1)) because including them makes the model more
complicated. Later, the ARDL model developed by H0 : b1 ¼ b2 ¼ b3 ¼ b4 ¼ b5 ¼ 0
Pesaran et al. (2001) supports the variables integrated ði:e:there exists no long ð4Þ
at order one as well.  run relationship between the variablesÞ
In order to investigate the existence of long-run
relationship between variables, the bound F-statistic H 0 : ai ¼ c i ¼ di ¼ x i ¼ s i ¼ 0
has been employed. It computes the value for every
ði:e: there exists no short
variable assumed as either exogenous or endogenous.
This leads to the testing of ARDL hypothesis for the  run relationship between the variablesÞ
long-run relationship between underlying variables. ð5Þ
The equation for the model is as follows:
The acceptance of null hypothesis means the model
WPIAt ¼ b0 þ b1 WPIAt1 þ b2 FDt1 þ b3 ERt1 shows the absence of long-run and short-run relation-
þ b4 TPGDPt1 þ b5 ITOTt1 ships of exogenous variables for the correction of
Xq X q
inflation in the economy. The acceptance of alternate
þ ai DWPIAt1 þ ci DFDt1 hypothesis is fruitful along with showing the presence
i¼1 i¼1
X
q X
q of Romer’s hypothesis.
þ di DERt1 þ xi DTPGDPt1
i¼1 i¼1
Xq
þ si DITOTt1 þ et Empirical estimation and results
i¼1
ð2Þ This study analyses the data for the period of 42 years,
i.e. from 1974–1975 to 2015–2016. It can be explained
where WPIA stands for the annual variation in WPI, in a better way through its statistical features. The
FD is central government fiscal deficit, ER is rupee to features such as mean value, standard deviation and
dollar exchange rate, TPGDP is trade as percentage of standard error are shown in Table 2.
GDP, ITOT stands for income terms of trade and et is The first column of Table 2 provides information
the error term. The initial part of the equation with b0, about the variables incorporated for the analysis. The
b1, b2, b3, b4 and b5 represents the long-run coeffi- second column depicts the number of observations
cients of the model. However, the other half of the included for the respective variables. The total number
equation with ai, ci, di, xi and si, represents short-run of observations for the study of each variable is 42.
coefficients of the model. The short-run dynamics of The minimum and maximum limits of the variables
the model can be estimated using the ECT model with considering the collected data are shown in the third
the values of lagged variables. The significant ECT and fourth columns of the table. The further columns
of the table indicate the mean, standard deviation,
standard error and the variance of the variables. From

123
Decision (March 2020) 47(1):79–90 85

Table 2 Summary Variables N Min Max Mean Standard deviation Standard error Variance
statistics. Source: Author’s
computation FD 42 23.02 5327.91 1345.06 1691.51 261.01 2,861,215
ER 42 7.91 65.46 30.62 18.16 2.80 330.06
TPGDP 42 12.35 55.79 27.36 14.92 2.30 222.74
ITOT 42 10.99 279 95.95 82.16 12.67 6750.31
WPIA 42 - 3.65 25.2 6.94 5.07 0.78 25.75

the table, it can be inferred that the average change in except WPIA. The variables integrated at the first
inflation prevailed in the economy for the study period difference (I(1)) through ADF test exhibit the rejection
(from 1974–75 to 2015–16) is approximately 6.94% of the null hypothesis, i.e. all variables are stationary at
with the positive change of 3.65 (i.e. minimum value) the first difference. With regard to Phillips–Perron
and with a negative change of 25.2% rise in the test, all the variables excluding WPIA are backing the
inflation. It shows the radical shifts in the inflationary existence of unit root at level with high probability
conditions in Indian economy. In terms of standard values. The mixture of order of integration among the
deviation, the dataset values of WPIA show less variables can be observed from the results of Phillips–
variation from the mean. The values of fiscal deficit Perron test, with WPIA integrated at level (I(0)) and
vary mostly out of the average fiscal deficit (as inferred FD, ER, TPGDP and ITOT integrated at the first
from the standard deviation and variance of the FD). difference (I(1)). In general, it can be inferred that the
variables are integrated at either I(0) or I(1). The
Results of unit root test analysis has been continued with the detection of co-
integration among the variables used. To do so, the
The sustainability of the ARDL model is the concoc- long-run bound co-integration test has been employed.
tion of the results gained from the analysis of variables In order to proceed, the prior need is the selection of
integrated at level (I(0)) or at the first difference (I(1)). the optimal lag length for the model. The appropriate
The presence of integrated variables at second differ- lag length has been selected on the basis of comparing
ence is the flaw to the model. In order to examine the lag length models with high prediction power and
Romer’s hypothesis for India, in the initial stride, the less predicting errors. The preferred model under the
stationarity of the variables has been tested. The study for the lag length selection is the Akaike
methods of unit root, i.e. augmented Dickey–Fuller information criteria (AIC).
(ADF) test and Phillips–Perron test, have been
employed to check the stationarity. The results for Results of the ARDL model
the stationarity are shown in Table 3 with t statistics
and probability values. In the initial steps of bound test co-integration, the
The ADF unit root results show acceptance of null long-run relationship between the variables from
hypothesis at level (I(0)) for all the variables, i.e. the Eqs. (1) and (2) is tested from the F-statistic (which
presence of unit root with a high probability value is responsive with the lag length computed from AIC).

Table 3 Results of unit Variables ADF unit root test Phillips–Perron test
root test. Source: Author’s
computation At level At first difference At level At first difference
t stat Prob t stat Prob t stat Prob t stat Prob

FD - 1.5814 0.4828 - 5.9695 0.0000 - 1.6556 0.4457 - 7.0940 0.0000

ER - 0.7293 0.8279 - 4.3968 0.0011 - 0.7400 0.8251 - 4.3851 0.0012
TPGDP - 0.7385 0.8255 - 5.1127 0.0001 - 0.7987 0.8089 - 5.1707 0.0001
ITOT - 1.2806 0.6295 - 5.8807 0.0000 - 1.2992 0.6209 - 5.8092 0.0000
WPIA - 3.7058 0.0076 - 8.6006 0.0000 - 4.2950 0.0015 - 8.6052 0.0000

123
86 Decision (March 2020) 47(1):79–90

The F-statistics examine the hinge of null hypothesis Table 5 ARDL (3, 4, 0, 4, 1) model long-run coefficient
where the lagged values of the equation may have their results. Source: Author’s computation
coefficient zero. The computed value of F-statistic for Variables Coefficient Standard error t statistic Prob
the model is shown in Table 4 taking WPIA as a
dependent variable. FD 0.2408 0.2930 0.8221 0.0420
The bound test results from the F-statistics on the ER 0.4411 0.4653 0.9480 0.0353
basis of ARDL (3, 4, 0, 4, 1) model confirm the TPGDP 1.6936 0.3374 5.0188 0.0001
existence of long-run relationship among the vari- ITOT - 1.7341 0.3645 - 4.7445 0.0001
ables. The F-statistic value obtained from the testing is
10.19731, which is greater than the upper bound limit
of 3.01, 3.48 and 4.44 at the different significance
levels of 10%, 5% and 1%, respectively. The results leads to 1.69% change in the WPIA. The obtained sign
reveal that the coefficients of the variables in Eq. (1) for the variable is the same as the expected sign for
have values other than zero. The rejection of the null verifying Romer’s hypothesis. The high dependency
hypothesis from Eq. (2) signals towards the effect of of the people on the traded goods and services and raw
lagged variables in the long run. The convergence material may contribute to the escalated prices in the
among the variables approaching towards long-run economy. The analysis for the relationship between
equilibrium is done by estimating the coefficients of ITOT and WPIA reveals the negative relationship,
the variables confirming long-run relationship. The which is not as per the expected sign and not apt for the
table signifies coefficient and the direction of study. The highly significant variable in the relation-
relationship. ship shows the highest respective coefficient value
The long-run coefficients estimated in the ARDL with the change of 1.73%, i.e. more than 1% change in
(3, 4, 0, 4, 1) model from Table 5 show that the FD WPIA due to 1% change in the ITOT. This means an
positively affects the WPIA in the long run. Ceteris increase in purchasing power in terms of exports may
paribus, the results show that 1% change in the FD lead to decline in the prices. In mathematical terms,
may lead to 0.24% change in the WPIA. The relationship between the lagged variables can be
significant outcome of FD towards WPIA is as per denoted as follows:
the prior direction of the effect assumed. The increas-
ECTt1 ¼ WPIAt  ð0:2409FDt þ 0:4412 ERt
ing fiscal deficit is raising the WPIA which is ð6Þ
favourable as per the expected outcomes of the study. þ1:6936TPGDPt  1:7341ITOTt Þ
Similarly, the change of 1% in ER may result in 0.44% where ECT is the error correction term indicating
change in WPIA, which shows the significant and speed of adjustment for the lag of variable from the
positive association of the ER with the endogenous long-run equilibrium in the relationship, if get devi-
variable which is not as per a priori expectations. It ated. The results for the coefficients working in the
indicates that rise in the ER will contribute to the short-run dynamics can be obtained by applying error
increments in WPIA. The purchasing power of people correction model (ECM) along with ARDL (3, 4, 0, 4,
decreases with the rise in the exchange rate in the 1) model.
economy. The TPGDP with the positive sign shows Table 6 depicts the results for testing Eqs. (2) and
significant relationship and movement of the variables (3). It shows the short-run coefficients of the variables
in the same direction. The significant impact of and the direction of relationship. The outcomes in the
TPGDP can be observed where 1% change in TPGDP short run unveil that a 1% change in FD may result in

Table 4 ARDL bound testing analysis. Source: Author’s computation

Test statistic Level of significance (%) Lower critical bound Upper critical bound Value Inference

F-statistic 10 1.9 3.01 10.19731 Variables are co-integrated

5 2.26 3.48
1 3.07 4.44

123
Decision (March 2020) 47(1):79–90 87

Table 6 ARDL (3, 4, 0, 4, Variables Coefficient Standard error t statistic Prob

1) model ECM results.
Source: Author’s DWPIA 0.4113 0.1521 1.8352 0.0021
computation
DWPIA (-1) - 0.4148 0.1338 - 3.1134 0.0295
DWPIA (-2) - 0.5109 0.1405 - 3.6355 0.0220
DFD 0.3837 0.2224 1.7255 0.0409
DFD (-1) 0.0015 0.0019 0.7742 0.4483
DFD (-2) 0.0019 0.0025 0.7495 0.4627
DFD (-3) 0.0081 0.0022 3.6419 0.0017
DTPGDP 0.0865 0.2265 0.3819 0.7219
DTPGDP (-1) - 0.4822 0.2935 - 1.6425 0.1169
DTPGDP (-2) - 0.8958 0.2853 - 3.1395 0.0054
DITOT 0.0120 0.0378 0.3169 0.7547
ECTt-1 - 0.5971 0.0766 - 7.7909 0.0014
R squared = 0.8209 SER = 0.0924
SER standard error of Akaike information criterion = - 1.6675 Schwarz criterion = - 1.1451
regression

0.38% change in the WPIA. The positive and signif- Diagnostics and stability tests
icant effect of fiscal deficit on the prices can be
observed, which is steady as compared to results found The accuracy for the ARDL model can be identified
in the long run. However, TPGDP exhibits the positive with the help of diagnostic checks and stability checks.
but insignificant relationship with the first difference The specification of the model is very important for
of WPIA. The change of 1% in TPGDP leads to 0.08% analysing the relationship between the variables that
change in WPIA. The results from ECM show that includes carrying out diagnostic tests that are con-
ITOT also has insignificant positive relationship with cerned with the problem of autocorrelation,
the WPIA. This shows that results in short run are heteroscedasticity, normality and the stability tests.
inconsistent with the outcomes found in the long run.
In mathematical terms, the ARDL function for the Test for autocorrelation
inflation can be denoted as follows:
The Breusch–Godfrey test is used to assess the
DWPIA ¼ 0:5971 þ 0:2409FDt1 þ 0:4411ERt1
existence of autocorrelation. The presence of autocor-
þ 1:6936TPGDPt1  1:7341ITOTt1
relation does not create biasness but affects the
þ 0:4113DWPIAt1 þ 0:3837DFDt1
efficiency of the model, i.e. disturbs the BLUE in the
þ 0:6342DTPGDPt1 þ 1:9059DITOTt1
estimation. The autocorrelation is tested with the null
ð7Þ hypothesis of no autocorrelation and alternate hypoth-
Table 6 also unfolds the error correction speed of esis for the presence of autocorrelation. Table 7 shows
the model which is estimated to be 0.5971 with the the results drawn from testing autocorrelation.
negative sign showing the existence of negative The probable value of F-statistic and Chi-square
relationship. In case of any deviation occurring from shows the values greater than 0.05 (i.e. 5%) level of
equilibrium, it may converge back to the long-run significance. The absence of autocorrelation can be
equilibrium by up to 59.71% in a year (can be referred
to as moderate speed of adjustment). This also reveals Table 7 Results of Breusch–Godfrey correlation LM test (up
the rejection of null hypothesis for the short-run to lag 1). Source: Author’s computation
analysis [i.e. Eq. (5)], depicting that no short-run F-statistic 1.1067 Prob. F 0.3510
relationship exists between the variables. The ARDL Obs. R squared 3.8605 Prob. Chi-square 0.1451
model with OLS fits well by 82.09% shown by
coefficient of determination.

123
88 Decision (March 2020) 47(1):79–90

inferred by the values obtained from Breusch–Godfrey Test for stability

test. The acceptance of null hypothesis shows the
independence of error term from each other. The stability between the short-run and long-run
coefficients can be analysed with the help of CUSUM
Test for heteroscedasticity (cumulative sum of recursive residuals) and
CUSUMSQ (cumulative sum of squared recursive
Heteroscedasticity is simply the violation of one of the residuals) tests of stability. The variables considered
assumptions of OLS, i.e. constant variance for unlike for the study via ARDL model under the time series
values of exogenous variables. The presence of analysis can be affected by the worldwide shocks that
heteroscedasticity disturbs the robustness of the OLS make it more important to examine the stability of the
estimation via biased value of the standard error. The model. The stability is confirmed graphically when the
consistency and efficiency remain unaffected by the line drawn of CUSUM and CUSUMSQ test remains
heteroscedasticity. Table 8 shows the results com- within the band created at 95% confidence interval.
puted by applying the Breusch–Pragan–Godfrey test The values within the band reflect the stability among
for finding the presence of heteroscedasticity. short-run and long-run coefficients in the model.
The null hypothesis of homoscedasticity can be Figure 1 exhibits the outcomes drawn from CUSUM
accepted as per the inferences drawn from the obtained test of stability. The stability of the model from CUSUM
values. The variances of the residuals for undertaken test is clearly visible as values remain within the limit. It
variables are quite constant in the model. The BLUE is can be also observed that the values predominantly lie
not disturbed, and results obtained are unbiased. near the lower limit. Similarly, the stability can be
evidenced by the results drawn from applying
Test for normality CUSUMSQ test shown in Fig. 2. The line drawn from
the values of CUSUMSQ remaining between the bands
The normality is purely not an assumption of the OLS of 5% level of significance is showing the swings in the
but an important perspective to look at for statistically pattern. In both tests, the value of residuals does not
efficient results. The non-normality disturbs the values deviate from the mean values.
of the statistical estimates of coefficient as well as the
confidence intervals as per the required assumptions.
The results for the Jarque–Bera test (employed to test Conclusion
the normality) are mentioned in Table 9.
The results from Table 9 show the presence of The relationship between trade openness and inflation
normality by accepting the null hypothesis for the cites different results for the short-run and long-run
testing. The proper bell-shaped curve in the normality
exists for the estimates in the model. 15

10

Table 8 Results of Breusch–Pragan–Godfrey heteroscedas-

5
ticity test. Source: Author’s computation
F-statistic 1.1621 Prob. F 0.3704 0

Obs. R squared 17.8261 Prob. Chi-square 0.3342

-5

-10

-15
Table 9 Results of Jarque–Bera normality test. Source: 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6
-9 -9 -9 -9 -0 -0 -0 -0 -0 -0 -0 -0 -0 -0 -1 -1 -1 -1 -1 -1 -1
95 996 997 998 999 000 001 002 003 004 005 006 007 008 009 010 011 012 013 014 015
Author’s computation 19 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

Jarque–Bera 0.0548 Prob 0.0051 CUSUM Lower Limit Upper Limit

Fig. 1 Results of CUSUM stability test. Source: Author’s

computation

123
Decision (March 2020) 47(1):79–90 89

1.4 for the analysis, but due to the unavailability of data,

1.2 the study does not employ them.
1.0

0.8

0.6
References
0.4

0.2 Ada OE, Oyeronke A, Odunayo AJ, Okoruwa VO, Obi-Egbedi

0.0 O (2014) Trade openness and inflation in Nigerian econ-
omy: a vector error correction model (VECM) approach.
-0.2
Res J Finance Account 5(21):74–85
-0.4 Aliyev K, Gasimov I (2014) Openness-inflation nexus in south
caucasus economies. Munich Personal RePEc Archive.
19 -96
1 9 - 97
19 -98
19 -99
20 -00
20 -01
20 -02
20 -03
20 -04
20 -05
2 0 -06
20 -07
20 -08
20 -09
20 -10

20 -13
20 -14
20 -15
20 - 1 1
20 -12

6
-1
95
96
97
98

00

03
04
05
06

08

10
11
12
13
14
99

01
02

07

15
09

https://mpra.ub.uni-muenchen.de/62761/. Accessed 23
19

CUSUMSQ Lower Bound Limit Upper Bound Limit Nov 2017

Barro RJ, Gordon DB (1983) Rules, discretion and reputation in
Fig. 2 Results of CUSUMSQ stability test a model of monetary policy. J Monetary Econ
12(1):101–121
Bleaney MF (1999) The disappearing openness inflation rela-
dynamics. The existence of relationship between the tionship: a cross country analysis of inflation rates. In: IMF
respective variables of trade openness and inflation working paper 161
has been evidenced by the F statistic value which is Bowdler C, Nunziata L (2006) Trade openness and inflation
episodes in the OECD. J Money Credit Bank
higher than the upper bound value. In the short-run 38(2):553–563
dynamics, the indicators are showing the expected Edwards S (2006) The relationship between exchange rates and
results that are required to verify the existence of inflation targeting revisited. National Bureau of Economic
Romer’s hypothesis in India. Highly significant pos- Research, Cambridge
Eijffinger SCW, Qian Z (2010) Globalization and the output-
itive impact of the fiscal deficits, trade as percentage of inflation tradeoff: new time series evidence, vol 27. Center
GDP and income terms of trade has been observed on and European Banking Center, Tilburg University,
the inflation rate in India in the short run. The Netherlands, pp 1–24
exchange rate shows no impact over the price changes Engle R, Granger G (1987) Cointegration and error correction:
representation, estimation and testing. Econometrica
in the short run. In order to prove the hypothesis to be 55:251–276
true for the long run as well, the indicators show the Friedman M, Schwartz A (1963) A monetary history of the
vague results. The significant relationship of trade as United States. Princeton University Press, New Jersey
percentage of GDP ratio and income terms of trade Granger CWJ (1981) Some properties of time series data and
their use in econometric model specification. J Econ
with the price level can be observed. The relationship 28:121–130
between fiscal deficit and exchange rate is quite less Gruen D, Dwyer J (1995) Are terms of trade rises inflationary?
significant with the rate of inflation. However, the Reserve Bank of Australia, Sydney. Research discussion
fiscal deficit and trade as percentage of GDP fulfil the paper no. 9508
Gruen D, Shuetrim G (1994) Internationalization and the macro-
requirement for the existence of Romer’s hypothesis economy. In: International integration of the Australian
in the long run. But the existence of Romer’s Economy. Proceedings of a conference, Reserve Bank of
hypothesis in the long run cannot be verified due to Australia, Sydney, pp 309–363
the contrary direction of relationship between Gruben WC, Mcleod D (2004) The openness-inflation puzzle
revisited. Appl Econ Lett 11(8):465–468
exchange rate and income terms of trade for India. Haq IU, Alotaish M, MS, Kumara NGS, Otamorudov S (2014)
However, the use of different proxy variables for Revisiting Romer’s hypothesis: time series evidence from
the same indicators in the same country may provide mall open economy. Pak J Appl Econ 24(1):1–15
different results for the economy at different periods of https://unctadstat.unctad.org
https://www.worldbank.org
time. The results may differ on the basis of using Iyoha MA (1973) Inflation and openness in less developed
different techniques for empirical investigation. The economies: a cross-country analysis. Econ Dev Cult
presence of other factors may also be the reason for the Change 22(1):31–38
decline in inflation in India, other than trade openness. Johansen S, Juselius K (1990) Maximum likelihood estimation
and inference on cointegration-with applications to the
There are other proxy variables also that can be used demand for money. Oxford Bull Econ Stat 52(2):169–210

123
90 Decision (March 2020) 47(1):79–90

Kurihara Y (2013) International trade openness and inflation in Sargent T, Wallace N (1981) Some unpleasant monetary arith-
Asia. Res World Econ 4(1):70–75 metic. Federal Reserve Bank Minneap Q Rev 5(3):1–17
Kydland E, Prescott E (1977) Rules rather than discretion: the Sepehrivand A, Azizi J (2016) The effect of trade openness on
inconsistency of optimal plans. J Polit Econ 85:473–492 inflation in d-8 member countries with an emphasis on
Lane PR (1997) Inflation in open economics. J Int Econ Romer theory. Asian J Econ Model 4(4):162–167
42:447–462 Sikdar A, Kundu N, Khan ZS (2013) Trade openness and
Mahmoudzadeh M, Shadabi L (2012) Inflation and trade free- inflation: a test of Romer hypothesis for Bangladesh.
dom: an empirical analysis. World Appl Sci J Munich Personal RePEcArchive, pp 85–96
18(2):286–291 Taylor JB (1983) Comments. J Monetary Econ 12:123–125
Munir S, Kiani AK (2011) Relationship between trade openness Temple J (2002) Openness, inflation and the phillips curve: a
and inflation: empirical evidences from Pakistan. Pak Dev puzzle. J Money Credit Bank 34:450–468
Rev 50:853 Terra CT (1998) Openness and inflation: a new assessment. Q J
Nasser OMA., Sachsida A, and Mendonça MJC (2009) The Econ 113(2):641–652
openness-inflation puzzle: panel data evidence. Int Res J Triffin R, Grubel H (1962) The adjustment mechanism to dif-
Finance Econ 169–181. https://www.eurojournals.com/ ferential rates of monetary expansion among the countries
finance.htm. Accessed 23 Nov 2017 of the European economic community. Rev Econ Stat
Pesaran MH et al (2001) Bounds testing approaches to the 44:486–491
analysis of level relationships. J Appl Econ 16(3):289–326 World Bank (2002) Globalization, growth, and poverty: build-
Rogoff K (1985) The optimal degree of commitment to an ing an inclusive world economy. Oxford University Press,
intermediate monetary target. Q J Econ 100:1169–1189 Washington, DC
Romer D (1993) Openness and inflation; theory and evidence. Wu CS, Lin JL (2008) The relationship between openness and
Q J Econ 108(4):869–903 inflation in Asian4 (NIE’s) and the G7. National Bureau of
Sachsida A, Mendonça MJC (2006) Inflation and trade openness Economic Research, pp 109–137. https://www.nber.org/
revised: an analysis using panel data. Instituto de Pesquisa books/ito_08-1. Accessed 23 Nov 2017
Economics Aplicada, pp 1–12
Salimifar M, Razmi MJ, Taghizadegan Z (2015) A survey of the Publisher’s Note Springer Nature remains neutral with regard
effect of trade openness size on inflation rate in iran using to jurisdictional claims in published maps and institutional
ARDL. Theor Appl Econ 22(3):143–154 affiliations.

123