You are on page 1of 19

Journal of Quantitative Economics

https://doi.org/10.1007/s40953-018-0140-9

ORIGINAL ARTICLE

Welfare Cost of Inflation: Evidence from India

Irfan Ahmad Shah1 · Manmohan Lal Agarwal1 · Srikanta Kundu1

© The Indian Econometric Society 2018

Abstract
In this paper, we estimate the welfare cost of inflation (WCI) to understand how costly
anticipated inflation is in India. The WCI is estimated both in partial and general equi-
librium framework using consumer surplus and compensating variation approaches.
Based on the quarterly data from 1996–97Q1 to 2014–15Q2 , we found that the WCI
at an inflation rate of 10% is ~ 0.53% of GDP. This implies that reducing inflation
from 10 to 0% may result in an output gain of 0.53% of GDP. The study also shows
that the WCI is an increasing function of rate of inflation and inflation elasticity of
money demand. The inflation elasticity is estimated through double-log and semi-log
specifications of money demand using Fisher and Seater (Am Econ Rev 402–415,
1993) long-horizon regression approach. It is found that the specification of money
demand influences the magnitude of WCI.

Keywords Welfare cost of inflation · Inflation elasticity · Consumer surplus ·


Compensating variation approach · Money demand

JEL Classifications E310 · D110 · E410

Introduction

Inflation has always been a matter of concern and so have been its costs. These costs
arise either through anticipated or unanticipated inflation. However, the costs due to
unanticipated inflation also known as the redistributive and disruptive costs, have been
given greater attention than costs due to anticipated inflation. This is because the costs

B Irfan Ahmad Shah


shahirfan914@gmail.com
Manmohan Lal Agarwal
manmohan@cds.ac.in
Srikanta Kundu
kundu.srikanta@gmail.com

1 Centre for Development Studies, Thiruvananthapuram, Kerala, India

123
Journal of Quantitative Economics

due to anticipated inflation are usually considered insignificant compared to the costs
due to unanticipated inflation. However, this general belief/argument was challenged
by Bailey (1956) who argued that the costs due to anticipated inflation such as menu
costs, shoe-leather or storage costs may be insignificant at an individual level but they
inflict a significant cost at an aggregate level. These costs were termed as the welfare
cost of inflation (WCI).
The pioneering work in analyzing the WCI has been carried out by Bailey (1956)
and Friedman (1969). Bailey (1956) estimated the WCI through consumer surplus
approach. According to this approach, as the nominal interest increases due to increase
in the level of inflation, the demand for cash balance decreases.1 This fall in the demand
for cash reduces the area under the money demand curve which according to Bailey
(1956) represents the consumers’ surplus. So, with rising inflation there is fall in
consumers’ surplus. A part of this consumer surplus goes to the government in the
form of seigniorage revenue (which is rebated back to the economy in the form of
non-distortionary lump-sum transfers). While the rest of the area, which is actually
the dead weight lost, measures the WCI. This method came to be known as the welfare
triangle approach of estimating WCI.
This approach has provided a baseline in estimating WCI, however, it is restricted to
the partial equilibrium framework only. The costs that are considered in this approach
are only due to distortions in the money demand function, ignoring the other costs of
inflation. Such costs were analysed by Lucas (1994, 2000) using a general equilib-
rium framework. Lucas (2000) used Sidrauski (1967) model in which money yields
direct utility to give direct numerical estimates of WCI in partial as well as in general
equilibrium framework. Lucas (2000) method of estimating WCI came to be known
as the compensating variation method. This method provides an upper bound to the
WCI, estimated by partial equilibrium framework suggested by Bailey’s (1956) (see
Izadkhasti et al. 2013 for details).
We follow Bailey (1956), Lucas (2000) and Izadkhasti et al. (2013) to estimate the
WCI in partial as well as in a general equilibrium framework in India. Unlike many
other studies,2 we estimated WCI in India using inflation measured through consumer
price index (CPI-IW) rather than nominal interest rates. It is because money demand
in developing countries is considered more responsive to inflation rate than interest
rate changes and the data on interest rate are usually unavailable, (Yavari and Serletis
2011; Izadkhasti et al. 2013). Both double-log and semi-log specifications of money
demand are used to estimate the WCI. We broadly found that the WCI depends on
the specification of money demand, inflation elasticity and the equilibrium framework
(partial or general) used for the estimation. At 10 percent rate of inflation, the WCI
in India, is broadly found to be 0.53 percent of GDP. This implies that a reduction of
inflation from 10 percent to zero in an economy like India may increase the GDP by
12.2 billion dollars.
The study is organized as follows. Introduction of the study is followed by a brief
review of literature on WCI in “Literature Review” section. “Theoretical Under-
1 People will try to economize their cash balance either by buying more of physical goods or by changing
the procedure of their payment.
2 Kumar (2014), Uwilingye and Gupta (2009) and Serletis and Yavari (2004, 2005, 2007) estimated the
WCI in different countries using nominal interest rate.

123
Journal of Quantitative Economics

standing of WCI” section discusses theoretical understanding of WCI in partial and


in general equilibrium framework. In “Data and Methodology” section, we discuss
description of the data used in the study and empirical methodology. “Results” sec-
tion discusses the results of empirical analysis. Conclusion of the study is given in
“Conclusion” section.

Literature Review

Economists have widely debated the issue of WCI. Although, there is general agree-
ment on the welfare costs, researchers have different opinions regarding its magnitude.
The early research on the WCI including Bailey (1956), Friedman (1969), Fischer
(1981), Feldstein (1997) and Lucas (1981, 1994, 2000) assuming certain elasticities
of money demand, emphasized that the WCI is significant and leads to a substantial
loss of the welfare in the economy. However, recently Gupta and Uwilingiye (2008),
Uwilingye and Gupta (2009), Serletis and Yavari (2004, 2005, 2007) and Yavari and
Serletis (2011) found that the value of the WCI is low compared to the earlier studies.
One important reason for this is attributed to the low value of the elasticity of the
money demand which is estimated from the data rather than assumed. Using different
econometric methods, a plethora of studies estimating the WCI across different coun-
tries and across different time periods concluded that the WCI depends upon the size
of the country (Serletis and Yavari 2007), the methods used for estimating the WCI
(Uwilingye and Gupta 2009) and value of the interest elasticity of money demand (Fis-
cher 1981; Serletis and Yavari 2004). While many studies have been carried out across
different countries, there are few studies estimating WCI for India. A recent study by
Kumar (2014) using consumers’ surplus approach, suggested by Bailey (1956), esti-
mated the WCI in case of India using a time varying framework. This study observed
that in the recent times, the WCI in India has increased. However, this approach of
estimating the WCI is restricted to partial equilibrium analysis only. Studies such as
those of Dotsey and Ireland (1996), Zee (2001) argued that inflation affects welfare
of people through various interactions and it would be better to evaluate these costs
in a general equilibrium framework rather than in a partial equilibrium one. The cur-
rent study considers the partial equilibrium analysis before extending it to a general
equilibrium framework.

Theoretical Understanding of WCI

In this section we discuss the theoretical foundations of WCI in partial and general
equilibrium framework. “Theoretical Foundation of WCI in a Partial Equilibrium”
section discusses the theoretical foundations of WCI in partial equilibrium framework
while “Theoretical Understanding of WCI in a General Equilibrium Framework” sec-
tion discusses it in the general equilibrium framework.

123
Journal of Quantitative Economics

Fig. 1 The Welfare Triangle


L

a
r1
L
r0
X1 X0

Theoretical Foundation of WCI in a Partial Equilibrium

Bailey (1956) did pioneering work in analyzing WCI which seems to have been over-
looked. According to Bailey (1956) WCI is a tax on holding non-interest bearing cash
and to differentiate it from the redistributive and disruptive costs of inflation, Bailey
made the following assumptions: (a) the rate of inflation is publicly announced and
followed by the monetary authority, (b) all the adjustments in terms of writing con-
tracts etc. are carried out such that the real values remain constant, (c) Bank deposits
are not used as money.
The demand for money, being an inverse function of the rate of interest,3 decreases
as the rate of interest increases due to rise in the inflation rate.4 Consider the Fig. 1,
wherein the real cash balance is shown on the X-axis and nominal interest rate on
the Y-axis. The LL curve represents liquidity preference curve and the area under it
represents the consumers’ surplus. The liquidity preference curve depicts the marginal
productivity of cash balance or in utility terms, the marginal rate of substitution of real
goods for cash. At nominal interest rate r0 which is consistent with the stable price level,
the demand for real cash balance is X0 ; if the nominal interest rate increases to r1 due
to increase in inflation equivalent to r1 − r0 , the demand for real cash balance decreases
to X1 . Cash becomes less productive and demand for physical goods increases. This
increase in demand for physical goods will increase the price level at once at a rate
equal to the ratio of real cash balances (X0 /X1 ). After this, prices will rise steadily at
the rate at which the government is adding money supply so that the real cash balances
remain constant at X1 (Bailey 1956). This fall in the real cash balance from X0 to X1
reduces the consumer surplus by an amount equal to the area r1 aX0 r0 . A part of this
area r1 aX1 r0 goes to the government in the form of seigniorage revenue while the rest
of the area X1 aX0 is dead-weight loss which represents the WCI. Figure 1 below gives
a clear depiction of WCI:
However, estimating WCI using consumers’ surplus approach may not completely
capture inflationary distortions and thus result in underestimation of true costs of

3 This argument has been well developed by Baumol (1952) and Tobin (1956) in the inventory theory
approach wherein transaction demand for money being interest elastic is well explored.
4 This is in accordance with the Fischer equation, i  r + π, where ‘i’ is the money rate of interest and ‘π’
being the inflation rate.

123
Journal of Quantitative Economics

inflation. Lucas (2000) using Sidrauski (1967) model extended this approach to analyse
the WCI in a general equilibrium framework.
In the next section, we first derive both partial and the general equilibrium frame-
work before using them to estimate the WCI in Indian context. The money demand
function is given as

M
 L(π , Y ),
P

where M denotes money supply, P is the general price level, Y represents level of real
income and π represents rate of inflation.
Bailey (1956) used semi-log specification of money demand for analyzing WCI
while Lucas (2000) preferred double-log specification. We incorporated both speci-
fications to estimate the WCI in case of India. The double-log specifications of the
money demand function are given by5 :

M1
m  Aπ −η (1)
PY

and in its logarithmic form,6 it can be represented as;

ln m  ln A − η ln π (a)

where A > 0 is a constant, η (> 0) represents the absolute inflation-elasticity of money


demand. In a similar way, the semi-log specification of the money demand function is
given by:

m  Be−ξ π (2)

with its logarithmic form:

ln m  ln B − ξπ, (b)

where ‘e’ is exponential term, B > 0 is a constant and ξ (> 0) represents the absolute
semi-log inflation elasticity of money demand.
Let m(π ) and ψ(x) be the required money demand and the inverse money demand
function respectively. The WCI as discussed by Lucas (2000) in terms of consumer
surplus approach proposed by Bailey (1956) can be written as,
 m(0)  π
W (π )  ψ(x)d x  m(x)d x − π .m(π ), (3)
m(π ) 0

5 The income elasticity is assumed to be unitary. Restricting it to unity, rules out the possibility of economies
of scale due to money holding.
6 Logarithmic transformation is used for the sake of convenience to interpret the parameters in terms of
elasticities.

123
Journal of Quantitative Economics

where W (π ) is the dead-weight loss and represents the WCI in a partial equilibrium
framework.
The right hand side of Eq. (3) represents the area under the money demand curve
from zero inflation rate to a specified inflation rate π.7 Using this framework Lucas
(2000) derived an exact welfare cost function which can be used to determine the exact
value of the WCI in double-log and semi-log specification of money demand.
The welfare cost function for double-log specification is given as:
η
W (π )  A π 1−η (4)
1−η

and for semi-log specification:


B 
W (π)  1 − (1 + ξ π)e−ξ π (5)
ξ

Equations (4) and (5) can directly be used to estimate the exact value of WCI in a
partial equilibrium framework.

Theoretical Understanding of WCI in a General Equilibrium Framework

We followed Lucas (2000) and Izadkhasti et al. (2013) wherein Sidrauski (1967) frame-
work is used to estimates the WCI in a general equilibrium framework. In Sidrauski
(1967) model, the real money balance along with consumption directly enters the
utility function.8 In this model, the household derives utility from the consumption
of a non-storable good c and the real money balance z  M/P where M is the money
supply and P is the price level. The economy is populated with infinitely lived families
with zero rate of population growth. The utility is assumed to be strictly concave with
continuous first and second order derivatives.9 Both the commodities c and z are not
inferior goods.10 The total welfare of the household depends upon the time path of ct
and zt and is represented by the utility function U(ct , zt ). Where ct is the flow of real
consumption per unit of time and zt is the flow of services per unit of time derived
from real cash balance. Both the variables are in per capita terms and U(ct , zt ) is the
instantaneous utility function. The total welfare associated with the given time path
as shown by Izadkhasti et al. (2013) is given by:

 
π
7 The integral ∫ m(x)d x represents the total consumer surplus lost by the consumer due to increase
0
in inflation and π .m(π ) represents part of the consumer surplus that the government gets as seigniorage
revenue. The net of these two is the WCI.
8 McCallum and Goodfriend (1987) proposed another framework of the general equilibrium model wherein
the real money balance enters the utility function as transaction technology which reduces the cost of
transactions. However, such estimates of welfare cost obtained from McCallum and Goodfriend (1987) to
that of the Sidrauski’s (1967) model were not found out to be significantly different (Lucas 2000). Thus,
we restricted our model to Sidrauski’s (1967) framework for the estimation of the WCI.
9 These condition implies that u < 0, u < 0, j  u u – u2 > 0.
cc zz cc zz zc
10 For goods not to be inferior, the condition required is J  u – u z/u < 0 and J  u u /u – u
1 zz cz c 2 cc z c cz
< 0.

123
Journal of Quantitative Economics

∞
   
Max W  U ct, z t e−ρt dt , uc , uz > 0, ucc , uzz < 0,
0

where ρ  discount factor and e−ρt is the continuous time discounting factor with
ρ > 0 representing the time preference. W represents the total welfare which can be
maximized by increasing the consumption of c and z. Izadkhasti et al. (2013) further
showed that the wealth held by the household can either be in the form of money or in
capital. Thus, the budget constraint faced by the household in maximizing its utility
is given by:

ct  wt + rt kt − k̇t − ż t − πt z t + x (6)

where kt is capital asset, πt is the inflation rate, wt is real wage rate and rt is rate of
interest. The dot on the top of the variable represents its rate of change with respect
to time, π t zt is the amount of inflation tax and x represents government transfers. The
behavior of the economic agent is subject to two constraints; stock constraint (at ) and
the flow constraint. The stock constraint requires the total endowments to be allocated
between capital and real cash balance i.e.,
at  kt + zt (7)

and the flow constraint

wt + rt kt − k̇t − ż t − ct − πt z t + x  0 (8)

From these two Eqs. (7) and (8) we have another equation as

ȧt  wt + rt kt − ct − πt z t + x. (9)

Taking all these equations into consideration a Hamiltonian function is formed such
that the household maximizes its utility function subject to the given flow and the stock
constraints as follows:
∞
H  {u(ct , z t ) + λt (wt + rt kt − ct − πt z t + x − ȧt ) + qt (at − kt − z t )}e−ρt dt,
0
(10)

where λt is the costate variable11 and qt is the multiplier. The Hamiltonian function
measures the instantaneous total economic contribution made by the control vari-
ables towards the integral objective function and λt can be interpreted as the marginal
(imputed) value or the shadow price of the state variable.12 From Eq. (10), the first
order derivatives with respect to c, z, k and a are respectively as follows:
11 The costate variable can be interpreted as the Lagrange multiplier associated with the state equation.
12 Shadow price gives the small change in the objective function arising from a small change in the
constraint.

123
Journal of Quantitative Economics

∂H
 0, u c (ct , z t )  λt , (11)
∂c
∂H
 0, u z (ct , z t )  λt πt + qt , (12)
∂z
∂H qt
 0, λt rt  qt , rt  , (13)
∂k λt
λ̇t − ρλ  −qt or
qt  ρλ − λ̇t . (14)

For boundary conditions, the transversality condition13 ruling out the Ponzi condi-
tion, gives the upper bound as,

lim e−ρt λt at  0 (15)


t→∞

From Eq. (11), (12) and (13) the quantity demanded of consumption and real cash
balance can be solved as an implicit price of consumption (λ), the implicit interest
rate (r) and the inflation rate (π) (Sidrauski 1967). In the steady state as argued by
Izadkhasti et al. (2013) ż  λ̇  ȧ  0 and the first order condition imply that
u z (ct , z t )
 π + r  i, . (16)
u c (ct , z t )

where i is the nominal interest rate,u z (ct , z t ) is the marginal utility derived from the
real money balance z and u c (ct , z t ) is the marginal utility derived from the consump-
tion of non-storable good c. The ratio of these two marginal utilities represents the
marginal rate of substitution between consumption and the real money balance which
is equal to the nominal interest rate and interpreted as the price of the money services,
(Izadkhasti et al. 2013). This nominal interest is actually the opportunity cost of hold-
ing non-interest bearing money. When nominal interest rate increases due to rise in
inflation, people actually spend more time in economizing their cash-holdings rather
than utilizing it productively. This inefficient way of utilizing cash leads to WCI and
can be reduced by reducing the nominal interest rate from a positive rate to zero (Lucas
2000). This welfare cost is actually the percentage income compensation needed to
leave a household indifferent between the nominal interest rate i and 0. Thus, the
solution to the welfare function is given as;

U [(1 + w (i))Y, m(i) Y]  U[Y, m(0)Y], (17)

where w(i) is the welfare cost at nominal interest rate i and equals to zero when i is zero,
m(i) and m(0) are the required money demand functions when the nominal interest
rate is i and zero respectively. The utility is maximized when the nominal interest
rate equals to zero also known as the Friedman’s (1969) optimal rule. At this point,

13 The transversality condition means that the value of the household’s per capita assets must approach
zero as time approaches infinity.

123
Journal of Quantitative Economics

money demand held m(0) is maximum and one losses nothing by holding non-interest
bearing money. If the interest rate increases from 0 to i the money demand declines
M1
from m(0) to m(i) and the WCI increases from w(0) to w(i).14 Since m  PY  Yz
M1
with z  P , the equilibrium condition is given by c  Y and the WCI is estimated
using Lucas (2000) approach wherein the homothetic current-period utility function
u is given as;
1 z

u(c, z)  [cϕ ]1−σ , 0 < σ , σ  1, (18)


1−σ c

where c is the consumption good, z is real money balance and ϕ is a function. From
this we can obtain the welfare cost functions for double and semi-log specifications
in general equilibrium as15 :

η
η−1
w(π )  −1 + 1 − Aπ 1−η , (19)

where w(π) represents the WCI under double-log specification in general equilibrium
framework while as WCI under semi-log specification is given by
 
1
w  (π )  ξ Be−ξ π π + w(π ) (20)
ξ

From Eq. (4) and (5) in the partial equilibrium and Eqs. (19) and (20) in the general
equilibrium framework, the exact WCI can directly be estimated if the values of
inflation elasticity of money demand and the value of constants such as A and B are
known. We estimate these values from the data.

Data and Methodology

Data

The data for this study is taken on quarterly basis from 1996-97Q1 to 2014-15Q2 . The
lower limit of the sample size is restricted to 1996-97Q1 due to the unavailability of
nominal GDP on a quarterly basis before 1996. The variables we have used in the
study are money supply (M1− ), nominal income (PY) and Inflation (π). The rationale
for choosing narrow money (M1− ) as money supply is that it is relatively more liquid
compared to other forms of money and WCI is more prominent when M1 is used as
money supply (Cooley and Hansen 1989, Lucas 1981). Even other forms of money
supply such as high powered money has been used to analyse WCI (Fischer 1981).
So, choosing M1 as money supply may be a better measure to analyse the distortions
of inflation on non-interest bearing money. GDP at factor cost at current prices is
used for nominal income and inflation rate is calculated from the consumer price

14 For all i > 0, m (0) > m (i) and w (0) < w (i).
15 Derivation of these equations is given in “Appendix 1”.

123
Journal of Quantitative Economics

index-industrial worker (CPI-IW). As we are estimating the welfare cost of inflation


from the consumer’s side, it will be appropriate to use consumer price index rather
than the Wholesale Price Index. Since, there is no single consumer price index16 that
fully captures the consumer inflation, we use the consumer price index for industrial
workers (CPI-IW), which is considered as a relatively better index among the series
of CPI indices. All the time series variables are seasonally adjusted to eliminate the
seasonal fluctuations and a common base year 2004-05 is used for the analysis. All
the data series are taken from the Handbook of statistics, RBI.

Empirical Methodology

The empirical methodology is same for both the partial as well as the general equilib-
rium framework. All we need to do is to estimate the value of inflation elasticity and
the value of constants A and B. Following standard practices of time series analysis,
we start by analysing univariate characteristics of the data by testing stationarity of
the time series variables. We performed two main unit root tests, viz. the Augmented-
Dickey–Fuller (ADF) test and the Phillips–Perron (P–P) test with null hypothesis that
the time series posses a unit root and is nonstationary.
From the results of both test as shown in Tables 3 and 4 in “Appendix 2”, it is
observed that all the variables follow an autoregressive process with a unit root under
both the ADF and P–P tests. All the variables, however, after first difference are
found out to be stationary. As the variables are integrated of order one, it paved the
way to estimate the interest elasticity of money demand using long-horizon approach
proposed by Fisher and Seater’s (1993). In order to use long-horizon approach, the
basic requirement is that the data should at least be integrated of order one and of
the same order of integration. This condition is satisfied in our case.17 We did not
specifically check for cointegration because for this approach cointegration is neither
necessary nor sufficient condition (Uwilingye and Gupta 2009).
The basics of the Fisher and Seater (1993) approach can be described by taking a
bivariate autoregressive representation of money demand and inflation as follows:

αmm (L) <m> m t  αmπ (L) <π > πt + εtm , (21)


<π > <m>
απ π (L) πt  απ m (L) m t + εtπ , (22)

where αmm0  α 0  1, andα 0 andα 0 are not restricted,  1− L, where L is the


ππ mπ πm
lag operator, m is the money–income ratio, π is the inflation rate and < x > representing
the order of integration of x. Suppose x is integrated of order j or I (j) and j can be 0, 1,
2……, then < x>  j and < x >  < x> − 1. Further, the vectors (2m π 
t , εt ) are assumed
to follow normal distribution with zero mean and covariance Σε , the elements of which
are var (2m π m π
t ), var (2t ) and cov (εt , εt ) (Uwilingye and Gupta 2009).

16 The RBI has recently formulated a new consumer price index CPI (combined) that is relatively broad,
consisting of CPI-RL and CPI-UE but the data for this series is available only from 2011 onwards and
cannot be used for any historical analysis.
17 All the variables such as inflation (π), log of inflation (ln (π)) and log of the real money balance (ln (m))
follow an autoregressive process with a unit root.

123
Journal of Quantitative Economics

From Fisher and Seater (1993) approach, the long-run elasticity of money demand
for < m>  <π >  1 (as in our case as well) is given by long-run derivative of m with
respect to permanent change in π . This long run derivative, LRDm,π is defined as:
∂m t+k π
∂εt
LRDm,π  lim if lim ∂m t+k ∂επ  0. (23)
k→∞ ∂πt+k t
∂εtπ
k→∞

The long-run derivative is the ratio of two sequences. The numerator represents
the effect through time of an exogenous inflation disturbance on variable m and the
sequence in the denominator measures the effect of the same inflation disturbance
on inflation itself. Thus, the LRDm,π expresses the ultimate effect of an exogenous
inflation rate disturbance on m, relative to that disturbance’s ultimate effect on π,
(Yavari and Serletis 2011).
If limk→∞ ∂πt+k ∂επ  0, then there is no permanent change in the inflation rate
t
and LRDm,π is not defined. Since for our case where < m>  <π >  1, the LRDm,π is
defined and is equal to
θmπ (1)
LRDm,π  ,
θπ π (1)
∞ j
where θmπ (1)  ∞ θmπ (1)
j
j1 θmπ and θπ π (1)  j1 θπ π . The coefficient θππ (1) is the
long-run value of the impulse response of m with respect to π which also means that
the LRDm,π is interpreted as the long-run elasticity of m with respect to π, (Yavari
and Serletis 2011).
Under the Fisher and Seater (1993) approach where Cov (εtm , εtπ )  0 and π is
assumed to be exogenous in the long-run, the LRDm, π is given by lim k→∞ bk , where
bk is the coefficient from the regression:
⎡ ⎤ ⎡ ⎤
k k
⎣ <m> m t− j ⎦  ak + bk ⎣ <π > πt− j ⎦ + ekt (24)
j0 j0

For < m>  <π >  1, the consistent estimates of bk can be derived through ordinary
least square (OLS) estimation from the regression as follows:
 
m t − m t−k−1  ak + bk πt − πt−k−1 + ekt , (25)

where k  1, 2,…, K.

123
Journal of Quantitative Economics

Table 1 Results of ADF test

S. no. Series ADF p values Results

1 Inflation (π) − 2.74 0.22 Non-stationary


D (π) − 6.93*** 0.000 Stationary
2 ln (π) − 3.47* 0.0502 Non-stationary
D ln (π) − 9.07*** 0.000 stationary
3 ln (m) − 0.29 0.987 Non-stationary
D ln (m) − 8.00*** 0.000 Stationary
D(−) represents first difference of a time series
*(**) [***] indicates statistical significance at 10(5)[1] percent level

Table 2 Results of P–P Test

S. no. Series P–P p value Results

1 Inflation (π) − 2.96 0.148 Non-stationary


D (π) − 7.46*** 0.000 Stationary
2 ln (π) − 3.51** 0.0455 Non-stationary
D ln (π) − 10.74*** 0.000 Stationary
3 ln (m) − 0.28 0.989 Non-stationary
D ln (m) − 7.99*** 0.000 Stationary
D(−) represents first difference of a time series
*(**)[***] indicates statistical significance at 10(5)[1] percent level

Results

Using Eq. (25) and taking the value of ‘k’ as 30 for long-run (Uwilingye and Gupta
2009; Serletis and Yavari 2004, 2005), the inflation elasticity of money demand viz.
bk for both double-log and semi-log specification is given below in Tables 1 and 2,
respectively.
From Tables 1 and 2, it is observed that (in absolute terms) the inflation elasticity
of money demand in case of double-log model (η) is 0.064 while in the case of semi-
log model inflation elasticity of money demand (ξ) is 1.14. The double-log inflation
elasticity of money demand η represents the proportionate change in m due to the
proportionate change in π while the semi-log inflation elasticity of money demand
represents the proportionate change in m due to the absolute change in π .
As the value of η and ξ is obtained, the value of A and B can be estimated through
Eqs. (1) and (2). Following Lucas (2000), Uwilingye and Gupta (2009) and Ireland
(2009) the values of A and B are selected in such a way that they match the geometric
means of the data such as:


A , (c)
(π̄ )−η

B  −ξπ̄ , (d)
(e )

123
Journal of Quantitative Economics

9.0000

Welfare cost of Inflation (% of GDP)


8.0000
W(π) paral (DL) W(π) Paral (SL)
7.0000
6.0000
5.0000
4.0000
3.0000
2.0000
1.0000
0.0000
0 5 10 15 20 25 30 35 40 45 50 55 60
Inflaon Rate(%)

Fig. 2 WCI in partial equilibrium framework

where m̄ and π̄ represents the geometric means of m and π respectively. The values of
m̄ and π̄ are calculated from the data and are equal to 0.775 and 0.0593 respectively
which gives A  0.6467 and B  0.829.
Once the values of η, ξ, A and B are obtained, the values of WCI can be estimated in
case of partial as well as in general equilibrium framework. In the partial equilibrium
framework, the value of the WCI in both double-log and semi-log specifications can
be estimated using Eq. (4) and (5), respectively and is shown in Fig. 2.
The exact value of the WCI under partial equilibrium framework is given in Table 5
(“Appendix 3”). At low rate of inflation (up to 10%), the WCI under double-log
specification is greater than the WCI under semi-log specification. However, as the
level of inflation increases the WCI under semi-log specification dominates the WCI
under double-log specification. At an inflation rate of 10%, the WCI under double-log
specification is 0.51% of GDP while under the semi-log specification the same rate of
inflation produces a welfare loss of 0.43% of GDP. This implies that reducing inflation
in India from an inflation rate of 10–0% under double-log specification will lead to a
welfare gain of 0.51% of GDP.18 However, reducing the same level of inflation under
semi-log specification will only increase the welfare by 0.43% of GDP. In case the
initial level of inflation is high (above 10%), the welfare gain achieved by reducing
inflation under semi-log specification is more than the welfare gain under double-
log specification. This is explained as: in reducing inflation from an inflation rate of
30%, the welfare gain under semi-log specification is 3.4% of GDP while the welfare
gain under similar situation is only 1.43% of GDP under double-log specification. So,
policymakers before evaluating the costs and benefits of inflation should have a clear
understanding about the specification of money demand. It is not only the initial level
of inflation but also the specification of money demand that significantly affects WCI
in the economy.
Although this analysis of WCI in a partial equilibrium framework provides a mea-
sure of how costly anticipated inflation is in Indian context. It, however, represents
only a part of inflationary distortions in the money demand function. A better pic-
ture of these inflationary distortions can be seen in a general equilibrium framework

18 We define welfare in terms of the percentage change in the GDP of the country. The WCI being 0.51
percent means that with the reduction in the inflation, there is an increases in the output (GDP) by 0.51.
The increase in welfare is synonymously used with the increase in the GDP.

123
Journal of Quantitative Economics

0.6000

Welfare cost of Inflation(% of GDP)


W(π) paral (DL) W(π)General (DL)
0.5000

0.4000

0.3000

0.2000

0.1000

0.0000
0 1 2 3 4 5 6 7 8 9 10 11
Inflaon Rate (%)

Fig. 3 WCI in general and partial equilibrium under double-log specification

0.6000
Welfare Cost of Inflation (% of GDP)

W(π) Paral (SL) W(π) General (SL)


0.5000

0.4000

0.3000

0.2000

0.1000

0.0000
0 1 2 3 4 5 6 7 8 9 10 11
Inflaon Rate (%)

Fig. 4 WCI in general and partial equilibrium under semi-log specification

described in Figs. 3 and 4. Figure 3 compares the magnitude of WCI in general equi-
librium framework with that of the partial equilibrium framework under a double-log
specification and Fig. 4 compare the same WCI under semi-log specification.
From these figures we observe that at low rate of inflation the WCI under partial
and general equilibrium framework is almost same. However, as the rate of inflation
increases, the magnitude of WCI under general equilibrium framework diverges away
from the magnitude of WCI under partial equilibrium framework. This implies that
partial equilibrium analysis of WCI understates the true WCI. The exact estimates
of WCI for both partial and general equilibrium frameworks are given in Table 6
(“Appendix 4”). This table can be used for comparing the estimates of WCI under the
two frameworks for both the money demand specifications. From the table, we observe
that the WCI at 10% rate of inflation for both the money demand specifications is 0.53%
of GDP under the general equilibrium framework. This cost may look insignificant at
an individual level but for an economy like India with a GDP of 2.3 trillion dollars,
the aggregate loss will be equivalent to 12.2 billion dollars. These costs, with a large
section of population living in abject poverty, can no way be ignored.

123
Journal of Quantitative Economics

Conclusion

We estimated the WCI in partial as well as in general equilibrium framework, using


double-log and semi-log specification of money demand. We estimated the WCI in
partial equilibrium using consumer surplus approach suggested by Bailey (1956). We,
further, used Lucas’s (2000) compensating variation approach to estimate WCI in a
general equilibrium framework. Based on our analysis we found that the estimates of
WCI depends upon the rate of inflation, the inflation elasticity of money demand and
the specification of the money demand. Further, we estimated the WCI at different rates
of inflation. We found that at an inflation rate of 10%, the WCI in partial equilibrium
framework is 0.51 and 0.438% of GDP under double-log and semi-log specifications
respectively. However, under the general equilibrium framework, the WCI was slightly
higher around 0.53% of GDP, at same inflation rate for both the specifications. This
shows that reducing inflation from 10 to 0% in Indian may lead to an increase in its
GDP by 0.53%, which is around 12.2 billion dollars. Also we observed that, as the
rate of inflation increases the gap between the estimates of WCI in partial and general
equilibrium framework increases. This reflects the fact that partial equilibrium analysis
of WCI underestimates the true distortions of inflation.
These observations may be helpful for policymakers in a way to understand how
anticipated inflation can reduce the welfare of the people. It may also help policy-
makers to devise appropriate policy tools so as to minimize the WCI which in itself
can be a yard stick to assess the efficacy of monetary policy. Without having an accu-
rate assessment of the costs of inflation, setting mere targets may not be helpful for
policymakers.
The limitation of our study is that we did take into account producer’s perspective
(producer’s surplus) while analysing WCI. Different models such as the overlapping
generation model used by Bullard and Russell (2004) can be of further importance.19

Appendix 1

Here, derive WCI in a general equilibrium framework using Lucas (2000) approach
as:

U [(1 + w (i)) Y, m(i) Y]  U[Y, m(0)Y], (26)


1 z

u(c, z)  [cϕ ]1−σ , 0 < σ , σ  1, (27)


1−σ c

where c is the consumption good, z is real money balance and ϕ is a function. Using
Eq. (26) we have:
 
m(i)
(1 + w(i))ϕ  ϕ(m(0)), for i  0, (28)
1 + w(i)

19 Bullard and Russell (2004) found that the WCI at an inflation rate of 10 percent in the US is around 11.2
percent of GDP.

123
Journal of Quantitative Economics

and from Lucas (2000) the nominal interest rate is given by;

ϕ  (m)
 i. (29)
ϕ(m) − mϕ  (m)

Equation (29) can be transformed into a differential equation with ψ(m) as the
inverse money demand function as:
 
ψ(m)
ϕ  (m)  ϕ(m). (30)
1 + mψ(m)

Differentiating (28) with respective to (26) we get

w  (i)ϕ(m)  −ϕ  (m)[m  (i) − m.w  (i)] (31)

m(i)
Using the value of Eq. (30) in Eq. (31) with m  1+w(i) we have
 
m(i)
w  (i)  −ψ m  (i), with w (0)  0 (32)
1 + w(i)

Solving Eq. (32) and using the money demand function m(π )  Aπ −η we have


η
η−1
w(π )  −1 + 1 − Aπ 1−η , (33)

which represents the WCI under double-log specification in general equilibrium frame-
work. For semi-log specification we have
 
1
w  (π )  e−ξ π π + w(π ) (34)
ξ

where w’(π ) represents the first order derivative of the WCI. It becomes difficult to
solve Eq. (34) manually in its derivative form, so we used ‘Mathematica software20 ’
for getting the solution.

Appendix 2

The unit-root test is carried-out for inflation (π), the natural log of inflation (ln (π)) and
the natural log of the real money balance (ln (m)). The null hypothesis tests whether
the time series posses a unit root and is non-stationary. The results of both the tests
are given in Tables 3 and 4.

20 We used DSolve routine in Mathematica, version 10.4 for Eq. (2.20)

123
Journal of Quantitative Economics

Table 3 Double-log inflation elasticity of money demand

Coefficients Standard error t Stat P value

Intercept 0.12 0.018 6.635 5.4145E − 08


bk  η − 0.064 0.015 − 4.294 0.0001

Table 4 Semi-log inflation elasticity of money demand

Coefficients Standard error t Stat P value

Intercept 0.121 0.0186 6.512 8.09E − 08


bk  ξ − 1.14 0.0028 − 4.11 0.000184

Table 5 WCI estimate under double-log and semi-log specification

Inflation (%) W(π) Partial (DL) W(π) Partial (SL)

1 0.0594 0.0047
2 0.1136 0.0186
3 0.1661 0.0416
4 0.2174 0.0733
5 0.2679 0.1137
6 0.3177 0.1625
7 0.367 0.2196
8 0.4159 0.2846
9 0.4643 0.3575
10 0.5125 0.4381
15 0.749 0.9494
20 0.9804 1.626
30 1.433 3.3969

Appendix 3

In Table 5, we show the estimates of WCI in a partial equilibrium framework under


both the double-log(DL) and semi-log (SL) specifications of money demand.

Appendix 4

Table 6 represents the WCI in partial and general equilibrium frameworks using both
the double-log (DL) and semi-log (SL) specifications of money demand. The first
column represents the inflation rate next two columns represents the WCI under DL
specification while the last two columns represents the WCI under SL specification.

123
Journal of Quantitative Economics

Table 6 Double-log WCI in partial and general equilibrium

Inflation (%) W(π) partial (DL) W(π)General (DL) W(π) Partial (SL) W(π) General
(SL)

1 0.0594 0.05966 0.0047 0.0052


2 0.1136 0.11463 0.0186 0.0198
3 0.1661 0.16824 0.0416 0.0447
4 0.2174 0.22114 0.0733 0.0792
5 0.2679 0.27361 0.1137 0.1248
6 0.3177 0.32584 0.1625 0.18
7 0.367 0.37793 0.2196 0.25
8 0.4159 0.42997 0.2846 0.32
9 0.4643 0.48201 0.3575 0.42
10 0.5125 0.5341 0.4381 0.53

References
Bailey, M.J. 1956. The welfare cost of inflationary finance. Journal of political Economy 64 (2): 93–110.
Baumol WJ (1952) The transactions demand for cash: an inventory theoretic approach. The Quarterly
Journal of Economics 66 (4): 545–556.
Bullard J, Russell SH (2004) How costly is sustained low inflation for the US economy? Federal Reserve
Bank of St Louis Review 86 (May/June 2004).
Cooley TF, Hansen GD (1989) The inflation tax in a real business cycle model. The American Economic
Review 79 (4): 733–748.
Dotsey, M., and P. Ireland. 1996. The welfare cost of inflation in general equilibrium. Journal of Monetary
Economics 37 (1): 29–47.
Feldstein MS (1997) The costs and benefits of going from low inflation to price stability. In Reducing
inflation: Motivation and strategy, eds. Romer, C., and D. Romer, 123–166. Chicago: NBER.
Fischer, S. 1981. Towards an understanding of the costs of inflation: Ii. Carnegie-Rochester Conference
Series on Public Policy 15: 5–41.
Fisher ME, Seater JJ (1993) Long-run neutrality and superneutrality in an ARIMA framework. The American
Economic Review 83 (3): 402–415.
Friedman M (1969) The optimal quantity of money. In The Optimal Quantity of Money and Other Essays,
1–50. Chicago: Aldine Publishing Company
Gupta, R., and J. Uwilingiye. 2008. Measuring the welfare cost of inflation in South Africa. South African
Journal of Economics 76 (1): 16–25.
Ireland, P.N. 2009. On the welfare cost of inflation and the recent behavior of money demand. American
Economic Review 99 (3): 1040–1052.
Izadkhasti, H., S. Samadi, and R.D. Isfahani. 2013. The welfare cost of inflation in consumer surplus and
compensating variation method: Case study of iran. International Journal of Academic Research in
Business and Social Sciences 3 (8): 250.
Kumar, S. 2014. The varying interest elasticity and the cost of inflation in India. Applied Economics Letters
21 (7): 497–500.
Lucas, J.R.E. 1981. Discussion of: Stanley fischer, “towards an understanding of the costs of inflation: Ii”.
Carnegie-Rochester Conference Series on Public Policy 15: 43–52.
Lucas, J.R.E. 2000. Inflation and welfare. Econometrica 68 (2): 247–274.
Lucas, R.E. 1994. On the welfare cost of inflation. Stanford: Stanford University.
McCallum BT, Goodfriend M (1987) Money: Theoretical analysis of the demand for money, working paper,
2157, National Bureau of Economic Research.
Serletis, A., and K. Yavari. 2004. The welfare cost of inflation in Canada and the United States. Economics
Letters 84 (2): 199–204.

123
Journal of Quantitative Economics

Serletis, A., and K. Yavari. 2005. The welfare cost of inflation in Italy. Applied Economics Letters 12 (3):
165–168.
Serletis, A., and K. Yavari. 2007. On the welfare cost of inflation in Europe. Applied Economics Letters 14
(2): 111–113.
Sidrauski, M. 1967. Rational choice and patterns of growth in a monetary economy. The American Economic
Review 57 (2): 534–544.
Tobin, J. 1956. The interest-elasticity of transactions demand for cash. The Review of Economics and
Statistics 38 (3): 241–247.
Uwilingye, J., and R. Gupta. 2009. Measuring the welfare cost of inflation in South Africa: a reconsideration:
inflation. South African Journal of Economic and Management Sciences 12 (2): 137–146.
Yavari, K., and A. Serletis. 2011. Inflation and welfare in latin america. Open Economies Review 22 (1):
39–52.
Zee, H.H. 2001. Welfare cost of (low) inflation: a general equilibrium perspective. FinanzArchiv: Public
Finance Analysis 57 (4): 376–393.

123

You might also like