Xt = Set of control variables identified from growth literature, which includes—
government expenditure as percentage of GDP (GFCE), private and public
investment as percentage of GDP (INVPRI and INVPUB respectively) interest
rate (RBR) and population growth rate POPGR. Set of control variables for
estimation with quarterly data includes government expenditure as percentage of
GDP (GFCE), investment as percentage of GDP only;
Ut = Error term follows N (0, σ2).
Now equation (2) will have set of series of extra inflation (INFt -Π*) depending upon the
value of Π* ; 4 to 8.75 percent is the range of Π* in the present study. In case when
(INFt -Π*) < 0 (inflation is low) then β1 will be the estimate of inflation and it will be (β1
+ β2), i.e. sum of the coefficient of lagged inflation and extra inflation. The structural break
or threshold occurs at the value of (Π*), for which the R-square is maximum. It is
generally argued that the coefficient of D(INFt -Π*)indicates the difference in the
inflation effect on growth between the two sides of the structural break and its t-statistic
value tests whether or not the structural break is significant. As mentioned in Singh and
Kalirajan (2003), at some value of Π* , the log likelihood of the regression is maximized
(and the sum of the squared residuals is minimized) but the value of Π* at which the sum of
the coefficients of (INFt -Π*)and INFt significantly change sign may be less than at Π* .
Therefore, analysis of the entire set of coefficients is equally important in deciding about
the target inflation level, which is attempted in this paper.
Study uses annual data covering the period of 1971 to 2009 and quarterly data for the
period 1996:1 to 2009:3 to comprehend the threshold level of inflation on growth for India.
The data has been culled from Handbook of Statistics on Indian economy (2010) published
by RBI. The growth rate of GDP is based on constant prices at 1999-2000 as base year;
inflation is estimated by using wholesale price index (WPI), which is also based on 1993-
94 prices.
3. Empirical Results
Test of stationarity result based on ADF unit root test reveal mixed results, where some
variable are stationary and some are integrated of order one i.e. some variable are I(0) and
some are I(1). For sake of brevity we are not providing here the results of unit root test. For
analysis, threshold level of inflation, variables are used in difference form which ever
variable is I(1) to avoid problem of spurious regression.
To have general understanding of the kind of relationship inflation and output growth have
in Indian case, first a linear regression is run involving GDP growth rate, inflation, inflation
volatility and other control variables. This regression is started with lag of two for all
variables; result of most contemporaneous estimation is put in Table 1. The coefficient of
inflation and inflation volatility come out to be negative and significant. This negative
coefficient of inflation clearly suggest existence of long run negative relation between
inflation and output growth and negative coefficient of inflation volatility indicates that