Abstract:
The core inflation systematically underestimates headlines inflation
when food inflation is persistent. This study examines the statistical
properties of dynamic behaviour of food inflation, non-food inflation
and headline inflation. The data used in this study consist, monthly price
indices for all variables from 2003 to 2010. They were collected from
Central Bank, Sri Lanka. The empirical analyses are done using
descriptive statistics, Confidence ellipse, autocorrelation function,
Kernel density function, Box-plot, GARCH model, Granger causality
test, co-integration test. The empirical results showed that food,
nonfood, overall CPI price series in Sri Lankan are nonlinear, non-stationary
series with stochastic trend, 1(1). Standard deviation of each
inflation rate distribution dominates the mean value. Nonfood inflation
has fatter tails, has very high kurtosis (k=l 1.44). Food prices are
relatively persistent than non food prices. Food price inflation is not
only more volatile (SD=1.63>0.81) but also on average higher
(0.99>0.66) than non-food inflation. The overall shape of the confidence
ellipse indicates they are highly and positively correlated. The co-integration
analysis shows that food and nonfood inflation series have
long run relationship (co-integrated -EG test). Granger causality test
shows that food inflation granger cause non-food inflation (P=0.0031).
In Sri Lanka, food inflation runs higher than nonfood inflation. The
difference between food and nonfood inflation is statistically significant
(P=0.04). The empirical results of this study have important implications
for policy makers. A policy focus on measures of core Inflation that
exclude food prices can mispecify inflation, leading to higher
inflationary expectations, downward bias to forecasts of future inflation
and lags in policy responses
