Annex 3
Estimation of Threshold Inflation
Estimates of Threshold Inflation from Past Empirical Studies
Study |
Period |
Threshold Inflation
(Per cent) |
Methodology |
Chakravarty Committee Report (1985)# |
|
4 |
|
Rangarajan (1998)* |
|
6 |
Macro Econometric Model |
Kannan and Joshi (1998) |
1981-96 |
6-7 |
|
Vasudevan, Bhoi and Dhal (1998) |
1961-98 |
5-7 |
Correlation/regression |
Samantaraya and Prasad (2001) |
1970-99 |
6.5 |
|
Report on Currency and Finance (2001) |
1970-2000 |
5 |
Sarel’s Spline Method |
Singh and Kalirajan (2003) |
1971-98 |
No Threshold |
Spline regression |
Bhanumurthy and Alex (2010)** |
1975-2005 |
5 - 5.5 |
Spline regression |
Singh, Prakash (2010) |
1970-2009 |
6 |
Spline regression |
RBI Annual Report 2010-11 |
|
4 - 6 |
Spline regression, non-linear least squares and Logistic Smooth Transition Regression (LSTR) model. |
Pattanaik and Nadhanael (2013) |
1972-2011 |
6 |
Spline regression, non-linear approach, vector auto regression (VAR) |
IMF (2012) |
1996-2012 |
5-6 |
|
Mohanty et al (2011) |
1996-2011 |
4-5.5 |
Spline regression, non-linear least squares and Logistic Smooth Transition Regression (LSTR) model. |
Subbarao (2013) |
1996-2012 |
4.4-5.7 |
Spline regression, non-linear least squares and Logistic Smooth Transition Regression (LSTR) model. |
# cited as accepted rate of rise in prices
* Rangarajan(1996) observed that the objective of policy should be to keep inflation rate around 6 per cent.
**Using monthly data for January 2000 to April 2007, they suggested 4-4.5 per cent as the threshold. |
1. Univariate approach
A Logistic Smooth Transition Regression (LSTR) model proposed by Teräsvirta (19941; 19982) is used to
estimate the inflation threshold (Espinoza et al. (2010)3)
The model specification is as proposed by McAleer & Medeiros (2008)4 which employs a quasi maximum
likelihood (QML) estimator of smooth transition regression with multiple regimes.5
Quarterly data from 1996-97 to 2012-13 is used in the analysis. Apart from inflation and lagged values of
GDP growth, a control variable capturing world GDP growth is also used. GDP growth for OECD countries is
used as proxy for world GDP data. The impact of domestic factors is controlled using GDP lags.
2. Multivariate Approach
A Threshold Vector Auto Regression (TVAR) is a non-linear multivariate system of equations. TVARs
approximate the non-linear relationship by several regime-dependent formulations which are linear. Each
regime is defined in terms of threshold values and coefficients of the VAR system are specific to each
regime. The system of equations that is estimated for the reduced-form VAR with one threshold is given by
The chart indicates that in the case of both WPI and CPI-Combined, inflation above threshold reduces output
growth.
Cross-country Threshold Inflation Rates |
Country |
Threshold Inflation (Per cent) |
Country |
Threshold Inflation (Per cent) |
Armenia |
4.5 |
New Zealand |
3 |
China |
2.5 |
Nigeria |
11.2-12 |
Ghana |
11 |
South Africa |
4 |
Indonesia |
8.5- 11 |
USA |
2.5 |
Mexico |
9 |
India |
4-6 |
Source: Compiled from different empirical studies. |
|