tightening of monetary policy to curb inflation. Estimation results based on the Spline
regression though provide primary evidence in favor of threshold inflation but further
examination of it in Sarel (1996) sense rejects presence of any threshold level for India.
This result find congruence with study Singh and Kalirajan (2003) as even the estimation
results using Hansen (1997) method of threshold estimation shows very high level of
threshold inflation. Rest of the paper has been schemed as follows: Methodological issue,
model formation and data description has been dealt in section 2. Empirical result and
discussion on result has been put in section 3 and finally, section 4 presents conclusion of
the study.
2. Model Specification, Methodology and Data Description
Spline regression is a restricted form of regression which is used to estimate the model
when the model behaves differently after certain level of some variable value, called
threshold point for that variable. In order to estimate the threshold level of inflation in
inflation-growth relationship in India, present study develops the empirical model based on
the framework developed by Khan and Senhadji (2001). Further, empirical growth
literature has been followed to identify the control variables to be used. Empirical works of
growth literature identifies investments (public and private both), education, population
growth rate, terms of trade and government expenditure as major determinants of growth
(see Barro 1991; Sala-i-Martin 1997; and Romer 1993). We follow this conventional
variable to estimate the growth and inflation relation even at country level study. Following
the seminal work of Friedman’s (1977) and Levi and Makin (1980) inflation volatility has
also been introduced in the growth model.
ΔYt=α0+β1INFt+β2INFVOLt+∑n βiXt+Ut (1)
i=1
Equation (1) estimates the simple inflation growth relationship based on the variables
identified. Equation (1) is modified with introduction of D(INFt -Π*)to estimate the
inflation growth relation in light of threshold inflation (see equation 2).
ΔYt=α0+β1INFt+β2INFVOLt+β3D(INFt-Π*)+∑n βiXt+Ut (2)
i=1
where,
ΔYt = Growth rate of real output and growth rate of real per capita output, calculated as
ΔYt=((Yt-Yt-1)/ Yt-1)*100 . For estimation with quarterly data, ΔYt is growth rate
of real output only;
INFt = Inflation, calculated as INFt = ((WPIt-WPIt-1)/WPIt-1)*100;
INFVOLt = Volatility of inflation calculated as five point moving average of inflation;
Π* = Threshold level of inflation;
D = Dummy variable; D= 1 if INFt > Π* and D = 0 if INFt < Π* ;