SITEMAP FAQs GLOSSARY OTHER  LINKS CONTACT  US  
  ABOUT  US NOTIFICATIONS PUBLICATIONS DATABASE SPEECHES PRESS  RELEASES TENDERS FORMS EVENTS APPLICATION TRACKING SYSTEM  
Search
This Section
Entire Website
Home >> Publications - View Publications
 
Note : To obtain an aligned printout please download the (555 kb) version to your machine and then use respective software to print the story.
Date : 10 Oct 2011
Measures of Core Inflation in India – An Empirical Evaluation

Press Release

Janak Raj & Sangita Misra*

Abstract

This study attempts to analyse six exclusion-based measures of core inflation in India based on new series of WPI (2004-05=100). These are: WPI excluding food; WPI excluding fuel; WPI excluding food and fuel; non-food manufacturing; WPI excluding fuel and basic metals and metal products; and WPI excluding fuel, metal group and non-food primary articles. These measures were tested for volatility, unbiasedness and tracking the trend and predictive power. While WPI excluding food and WPI excluding food and fuel did not perform well in terms of volatility, the remaining four measures broadly satisfied the conditions relating to volatility, unbiasedness and tracking the trend and predictability of future inflation. A key property of core measure is that it should not revert to headline inflation. It was found that of the above four measures, all except non-food manufacturing revert back to headline inflation indicating that supply side shocks spill over to these core measures of inflation. This finding was further corroborated by granger causality and inflation persistence tests. Thus, non-food manufacturing, which the Reserve Bank uses as a measure of demand side pressures, is the only measure which satisfies all the properties of a core measure. A core measure of inflation is not an end in itself, but rather a means to achieve low and stable inflation by serving as a short-term operational guide for monetary policy.

Key Words: Core inflation.
JEL Classification: E31.

Introduction

Monetary policy works by influencing aggregate demand in the economy. However, at times an economy faces supply side shocks leading to large variations in relative prices. In such cases, headline inflation behaviour on its own could be misleading for policy purpose. Thus, a proper diagnosis of the price changes as to which price changes are transitory and which are permanent is critical. This is mainly because monetary policy operates with a lag. If increase in the price index is due to temporary shocks that could very soon reverse themselves, it may not require any monetary policy action. To deal with such situations, many central banks use measures of core inflation that are designed to filter the transitory price movements. Core inflation by eliminating the volatile components from the headline helps in identifying the underlying trend in headline inflation and is believed to predict future inflation better.

The Indian economy, like many other emerging market economies, has often been subject to some supply shock or the other. The Indian economy is currently experiencing high and persistent inflation. Inflation at above 8 per cent over last one and a half years has been one of the most persistent in the post-reform period. Although inflation prevailing now is generalised, it was triggered by supply side shocks, which in some form or the other have continued to persist, albeit with different intensity.

Price stability as defined by ‘low and stable inflation’ being one of the key objectives, the Reserve Bank monitors a range of price indices both at aggregate and disaggregated levels (RBI, 2006). Changes in the wholesale price inflation (WPI) is taken as the headline inflation for policy articulation and within the WPI, non-food manufactured products inflation is considered the core inflation (Mohanty, 2011). However, several other central banks normally use headline inflation, excluding food and energy, as measures of core inflation. An attempt, therefore, has been made in this paper to identify as to how non-food manufacturing inflation performs vis-a-vis some other measures of core inflation, normally used by several other central banks.

This paper is organised into five Sections. Section II explains the concept of core inflation. It also discusses briefly various techniques that are employed to measure core inflation. Section III briefly reviews the literature on core inflation, both in the international and domestic context. Section IV tests various exclusion-based measures of core inflation for India. Section V sums up the discussions.

II. Core Inflation: Concept and Measurement

Concept

The term core inflation was coined by Eckstein (1981) who defined it as 'the trend increase of the cost of factors of production' that 'originates in the long-term expectations of inflation in the minds of households and businesses, in the contractual arrangements which sustain the wage-price momentum, and in the tax system'. The concept of core inflation became popular in the 1970s during periods of high inflation and now normally refers to that component of inflation that is likely to persist for a long period, say, for several years and, therefore, useful for near-term and medium-term inflation forecasting (Blender 1997; and Byran and Ceechetti 1994). Most core measures are based on the concept that total inflation can be separated into two components: the core part, representing the underlying trend of inflation as shaped by the pressure of aggregate demand against capacity, and the non-core part, which reflects price movements caused by temporary shocks or relative price changes (Lafleche, 2006).

Chairman Bernanke of the US Fed in its report to the Congress (July 2007) emphasised that core measures were motivated by a desire to track and predict persistent inflation: “… food and energy prices tend to be quite volatile, so that, looking forward core inflation (which excludes food and energy prices) may be a better gauge than overall (headline) inflation of underlying inflation trends”. That is, by extracting underlying inflation trend, core inflation is able to predict future headline inflation better. If core inflation remains stable, then a surge in headline inflation is less likely to result in increase in inflation expectations, unless supply shocks get built into price expectations.

Although used by several central banks, core measure of inflation lack theoretical underpinnings. The headline measure of inflation is based on the theory of cost of living which provides a coherent framework for the evaluation of measures of headline inflation (Wynne, 1999). The choice of a basket and weights depends on the purpose for which the index is to be used for. The consumer price index represents the cost of a basket of goods and services consumed by a typical household. Most countries, therefore, use CPI as a measure of headline inflation.

Implicit in the discussion of core inflation is the idea that this type of inflation is fundamentally different to changes in the cost of living. Some analysts have linked core inflation to the measure of price change most closely related to monetary policy. By this reasoning, inflation is a monetary phenomenon in the long run, so core inflation should measure the component of price change related to monetary phenomena (Bryan and Cecchetti 1994; Wynne 1997, 1999). Because sustained relative price movements result from shifts in the relative demand for goods or changes in supply, not from monetary policy, core inflation should exclude relative prices changes. Defined in this way, core inflation is the measure over which monetary policy has the most influence (Roger; Shiratsuka; Cuitler; Hogan, Johnson, and Lafleche).

That is, there is a well-defined concept of monetary inflation that ought to be of concern to monetary policymakers and that this type of inflation, being conceptually different to the cost of living, is not adequately captured by the standard price statistics. Thus, it is argued that central banks ought to target a price index whose rate of increase corresponds to the inflation that generates the costs that central banks are seeking to avoid by focusing on an inflation control objective (Wynne, 1999).

Measures of core inflation, however, could not serve as a basis for inflation measurement that could possibly replace the theory of the cost of living. Measures of core inflation are thus no substitute for headline inflation. Focusing on core inflation does not mean that the central bank should not be concerned about inflation in the components excluded from this measure (e.g. food, energy etc), which represent a significant proportion of the consumer basket. Core inflation is simply a convenient guide to help the central bank achieve its objective of controlling total inflation. Most countries use measures of core inflation in addition to headline measures of inflation and not as a substitute. Of the 23 inflation targeting countries, only five countries target core inflation (Table 1). In the recent period, there have also been some countries such as Korea which have moved away from targeting core inflation to headline inflation.

Choice of a Measure

While the idea of core inflation is intuitively appealing, its practical policy usefulness has at times been questioned. There is no unique way of compiling core inflation and there is no generally accepted and intuitively plausible criterion to assess the policy usefulness of competing core inflation measures directly (----).

Notwithstanding the fact that there is no consensus on an appropriate measure of core inflation, literature has generally classified such measures into two broad categories: (i) statistical measures/order statistics [such as trimmed mean, weighted median, moving averages, filtered series, exponentially smoothened series and structural vector auto regression (SVAR), among others] and (ii) exclusion-based measures, i.e., by excluding some highly volatile elements from the headline such as food and fuel. Although statistical measures such as trimmed mean and median have better statistical properties of core inflation, such pure statistical measures are difficult for the public to understand and hence difficult to effectively communicate for the central banks [Wynne (1999), Clark (2001)]. The trimmed mean, for example, removes a different set of components each month, with the excluded set comprising a percentage of a distribution. More sophisticated core inflation measures such as SVAR are even more difficult to explain to the general public.

Table 1: Inflation Targeting Countries : Inflation Target

Country

Year of Introduction

Inflation Target

Brazil

1999

CPI

Chile

1991

CPI

Columbia

1999

CPI

Mexico

1999

CPI

Peru

2002

CPI

Indonesia

2000

CPI

Korea

1998

CPI

Philippines

2002

CPI

Thailand

2000

Core CPI

Czech Republic

1998

CPI

Hungary

2001

CPI

Poland

1998

CPI

Israel

1992

CPI

South Africa

2000

CPI

Turkey

2006

CPI

Australia

1994

CPI

Canada

1991

CPI-X

Iceland

 

CPI

New Zealand

1990

CPI-X

Sweden

1993

CPI Ex interest and Indirect tax

Norway

2001

CPI ex tax and energy

Switzerland

2000

CPI

United Kingdom

1992

CPI

Source: IMF
Note: Core CPI for respective countries explained in Table 2.

On the contrary, exclusion-based core measures have been the preferred choice of policymakers essentially because of their simplicity. They are easy to communicate to public when compared with the pure statistical measures of core inflation. The criticism that such measures often face is that completely removing the volatile items is a very crude methodology and has the potential risk of a permanent loss of significant information. Exclusion-based measures, although desirable from simplicity point of view, very often do not satisfy economic criteria. Core inflation measures based on statistical smoothing techniques provide smoother core inflation series, although they too are not supported very often by economic theory. However, most countries that use core inflation either as their inflation target or as an official core measure employ the exclusion method (Table 2). In doing so, the objective of policymakers is to keep the core inflation measure simple, well understandable and effectively communicate to the public on inflation trends and policy decisions.

Table 2: Official Core Inflation Measures: Cross-Country Practices

Core Inflation Targeting Countries

Canada

CPIX that excludes 8 most volatile components like fruits, vegetables, gasoline, natural gas, fuel oil, mortgage interest costs, intercity transportation and tobacco products

Sweden

CPI excluding interest and indirect tax

Norway

CPI excluding tax and energy

New Zealand

CPI excluding interest charges

Thailand

Core CPI excludes fresh food and energy prices which include rice, flour, cereal products, vegetables, fruits, electricity charges, cooking gas, and gasoline.

Other countries with Official Core Measures

Japan

CPI excluding fresh food

Peru

CPI excluding 9 volatile items like food, fruits and vegetables, urban transport, about 21.2 per cent)

United States

CPI excluding food and energy

Chile

CPI excluding 20 per cent with higher (-) variations and 8 per cent with higher (+) variations

Philippines

CPI excluding rice corn fruits vegetables, LPG, Kerosene, Oil, Gasoline, Diesel

Korea

CPI excluding non-grain agricultural products and petroleum products

Columbia

CPI excluding agricultural food, public services and transport

Spain

CPI excluding energy and unprocessed food

Netherlands

CPI (ULI) excluding fruits, vegetables and energy

Portugal

CPI (ULI) excluding energy and unprocessed food

III. Review of Literature

The properties of core inflation measures have been studied extensively, both internationally and domestically. At the global level, research has been conducted for core inflation based on both exclusion-based criteria and statistical measures for a number of countries, by OECD (2005), Rich and Steindel (2005), Blinder and Reis (2005), Marques, et al (2003) and Bryan and Cecchetti (1994). While Bryan and Cecchetti (1993) popularised the trimmed mean or weighted median techniques for computing core inflation, estimating core inflation through a structural vector auto regression (SVAR) method was suggested by Quah and Vahey (1995). In the official literature, however, a number of central banks report core inflation by excluding most volatile sub-components from CPI. For countries where core inflation is a target variable such as Canada, periodic research is conducted to evaluate the performance of existing exclusion-based core inflation measures as well as to suggest new exclusion-based measures that perform well in terms of various criteria (Armour, 2006; Lafleche et al, 2006). The components which are generally considered for exclusion are fresh food items, energy prices and mortgage interest payments. Recognising that all of these measures have pros and cons, Wynne (1999) observed that howsoever, core inflation is measured, for them to be useful for monetary policy formulation, it is crucial that they should be computable in real time and have some predictive power for future inflation.

In the Indian context also, the properties of core inflation measures have been tested by some researchers (Samanta (1999), Mohanty, et al (2000), Kar (2009), Das, et al (2009)). Samanta (1999) computed core inflation following the exclusion principle and concluded that the measure of core inflation, excluding primary food and non-food articles and administered items (combined weight of 46 per cent in WPI), was less volatile during the period 1993-94 to 1998-99. Mohanty, et al (2000) made a comprehensive attempt to examine core inflation measures: trimmed mean, weighted median and exclusion-based (by excluding energy, a number of fresh food items and a few manufactured items related to primary articles). They observed that 20 per cent trimmed mean WPI was an appropriate core inflation indicator for India. Looking at some recent studies on the subject, while Kar (2009) focussed only on statistical measures of core inflation, Das, et al (2009) analysed both exclusion-based measures, excluding food articles, and fuel group, individually and together and some statistical measures such as mean-SD, trimmed mean, median, reweighting, HP filter, Wavelet filter and structural VAR measures. Based on his findings, Kar (2009) showed that geometric exponential smoothing and weighted percentile were most suitable tools for core inflation. Das, et al (2009), on the other hand, found that there was no individual measure of core inflation that could be considered superior to other measures. Both these studies were based on the old WPI series with 1993-94 as base.

IV. Core Inflation - Empirical Evidence in India

Most countries use CPI as a measure of headline inflation. Therefore, core inflation measure in most countries are based on CPI.

However, in India, there is no single measure of inflation which captures economy-wide inflationary pressures in the economy. It is the year on year percentage change in wholesale price index (WPI), which is used as an indicator of headline inflation. Although there are four consumer price indices (CPIs), they are targeted at different population groups and none of them captures economy-wide inflationary pressures. CPI (Rural), CPI (Urban) and CPI (All India) have been launched recently, yet time series data in respect of new CPI series are not available. In view of these constraints, therefore, the Reserve Bank monitors an array of measures of inflation, both overall and disaggregated components, in the context of the evolving macroeconomic situation to assess the underlying inflationary pressures (RBI, 2010).

As indicated earlier, there is no unique way of measuring core inflation. Nevertheless, for policy purposes, most central banks use exclusion-based measures of core inflation as they are easy to explain. Recognising the simplicity and the practical significance of exclusion-based measures for monetary policy purpose, this study tests a variety of exclusion-based measures for core inflation based on the WPI and CPI-IW indices. For WPI, the monthly data for the period April 2004 to July 2011 (2004-05=100) were used. In the case of CPI-IW, the monthly series with 2001 as base year was used.

A good measure of core inflation should have three properties. First, core measure of inflation should be more stable or less volatile than headline inflation. Second, over a long period of time, average rate of core inflation should match the average rate of headline inflation and there should be no systematic divergence between the two. Also, core inflation should be able to track the trend rate of inflation. Third, if core inflation represents the underlying trend of inflation, it should then be better able to predict total or headline inflation. It is expected that core measures contain more information about the future trend of inflation than the headline inflation.

An attempt, therefore, was made to test all these properties of various exclusion-based measures.

Volatility Groups/Sub-Groups in WPI

To help us test the exclusion based measure of core inflation, volatility of various groups/sub-groups in WPI and CPI was first tested using (i) standard deviation that represents the dispersion around the mean and is the most commonly used measure; and (ii) mean of the absolute monthly change in year-over-year inflation. This volatility measure scores over the others in terms of its less direct dependence on the persistence of inflation (Lafleche, 2006, Khettry, 2006). At a group level, ‘fuel and power group turned out to be the most volatile group as expected, followed by ‘primary articles’ group. Within ‘primary articles’, non-food articles and minerals were highly volatile. Interestingly, food sub-group within primary articles was least volatile. Within fuel group, coal and mineral oils were very volatile. The ‘manufactured products’ group was least volatile. However, within manufactured groups, basic metals alloys and metal products exhibited high volatility (Table 3).

Table 3: Volatility in WPI Inflation: April 2005 to July 2011

Commodity Group/Sub-Group

Weight

Mean

SD

Mean Abs. Chg

ALL COMMODITIES

100.0

6.4

3.0

0.7

I PRIMARY ARTICLES

20.1

10.8

5.2

1.6

(A) FOOD ARTICLES

14.3

10.3

5.1

1.8

(B) NON-FOOD ARTICLES

4.3

9.8

9.5

2.3

(C) MINERALS

1.5

18.1

13.8

7.6

II FUEL & POWER

14.9

7.4

7.9

1.8

(A) COAL

2.1

9.5

10.6

1.6

(B) MINERAL OILS

9.4

8.8

10.7

2.5

(C) ELECTRICITY

3.5

2.0

2.4

0.6

III. MANUFACTURED PRODUCTS

65.0

4.6

2.2

0.5

(A) FOOD PRODUCTS

10.0

6.1

4.5

1.2

(B) BEVERAGES, TOBACCO & TOB. PRODUCTS

1.8

6.8

2.3

0.7

(C ) TEXTILES

7.3

3.7

5.2

0.6

(D) WOOD & WOOD PRODUCTS

0.6

6.9

2.8

1.3

(E) PAPER & PAPER PRODUCTS

2.0

4.0

1.5

0.5

(F) LEATHER & LEATHER PRODUCTS

0.8

3.9

3.1

0.9

(G) RUBBER & PLASTIC PRODUCTS

3.0

4.2

2.8

0.7

(H) CHEMICALS & CHEMICAL PRODUCTS

12.0

3.9

2.5

0.6

(I ) NON-METALLIC MINERAL PRODUCTS

2.6

6.3

4.3

0.9

(J) BASIC METALS, ALLOYS & METAL PRODUCT

10.7

6.3

7.8

1.7

(K) MACHINERY & MACHINE TOOLS

8.9

3.3

1.8

0.3

(L) TRANSPORT, EQUIPMENT & PARTS

5.2

3.1

1.3

0.5

The volatility in domestic metal prices increased in the 2000s vis-a-vis the 1990s (Table 4), reflecting the sharp co-movements with global metal prices that have remained high and volatile during the 2000s due to significant demand supply imbalances. The volatility in domestic metal prices such as iron and steel and aluminum was particularly more pronounced during last 3-4 years. Non-food primary articles such as cotton, rubber, oilseeds showed large volatility in the 2000s in line with global prices as trade in such commodities increased. Domestic prices of certain minerals such as copper, zinc, iron ore and crude petroleum also remained volatile due to both domestic demand-supply imbalances and international price shocks.

Table 4: Metal Group Inflation in India

 

Mean
(per cent)

Standard Deviation

1994-2000

4.5

3.6

2001-2011

7.6

8.6

2008-2011

5.8

9.6

Exclusion Based Core Inflation Measures

Based on the observed volatility of sub-groups as presented in Table 3, six measures of core inflation were derived:

  1. WPI excluding Food: Ex-food

  2. WPI excluding fuel group: Ex-fuel

  3. WPI excluding food and fuel: Ex-food fuel

  4. Non food manufacturing inflation: Non-food mfg

  5. WPI excluding fuel group and basic metals and metal products group: Ex-fuel metal

  6. WPI excluding fuel group, metal group and non food primary articles: Ex-fuel metal non-foodprim

While the first three represent the simplest and usually applied core inflation measures in most empirical studies, the fourth one, i.e., non-food manufacturing has been recently used by the Reserve Bank as an indicator of core. The fifth and sixth measures of core inflation were constructed by excluding the highly volatile sub-groups. Apart from ‘fuel group’ that is the most volatile, two other sub-groups, viz., metals sub-group (under manufactured products group) and non-food primary articles sub-group (under primary articles) were found to be highly volatile. Recognising this, two additional exclusion based measures – one excluding fuel group and metals sub-group and another excluding fuel group, metals sub-group and non-food primary articles sub group – were also considered in the study.

Volatility of Some Exclusion-based Measures

Four different techniques were used to assess volatility of exclusion based measures of core inflation: (i) standard deviation, (ii) coefficient of variation, (iii) mean absolute change and (iv) volatility around trend (Table 5). Standard deviation and mean absolute change have been explained earlier. Coefficient of variation is essentially the standard deviation normalised by mean. The last criterion following Clark (2001) was calculated by taking the volatility around the trend inflation in any given month that is simply estimated by a three-year average of overall inflation, with the average centered on the given month. It may be noted that apart from being a measure of volatility, this measure is also a guide as to how best core inflation tracks the trend rate of inflation. When trend inflation rises, for example, core inflation should increase com­mensurately. The accuracy with which core inflation tracks trend inflation is measured, on a monthly basis, as the standard deviation of the difference between core and trend infla­tion. For a core indicator that moves closely with trend, differences tend to be small, so the standard deviation is low.

Table 5: Summary Statistics of Exclusion based Core Inflation WPI Measures
(April 2005 to July 2011)

 

Weight

Mean

Standard Deviation

Coefficient of Variation

Mean Absolute Change

Volatility around Trend

Headline WPI

100.0

6.4

3.0

0.5

0.7

3.3

WPI Excluding Food

75.7

5.6

3.7

0.7

0.8

4.2

WPI excluding food and fuel group

60.8

5.2

3.2

0.6

0.7

3.3

WPI excluding Fuel Group

85.1

6.2

2.6

0.4

0.6

2.5

Non-food Manufacturing

55.0

4.4

2.6

0.6

0.5

3.0

WPI excluding fuel group and metal products

74.3

6.3

2.4

0.4

0.6

2.0

WPI excluding fuel group, Metal Products and non-food prim. articles

68.6

5.7

1.9

0.3

0.5

1.8

Normally, for unbiasedness, the core mean should not be very different from the mean of headline inflation. The standard test of equality of means was conducted and it was accepted for 4 out of 6 measures of core (Table 6). With regard to two remaining measures, excluding food and fuel and non-food manufacturing, it was observed that they are biased with core mean being significantly different from that of headline. Notwithstanding the bias that was observed, the difference between headline and core inflation rates over a long period (during the 2000s) was tested to be stationary (Table 7) indicating that core in respect of these also could be considered as a useful predictor of headline inflation (BIS, 2008). Thus, all six measures could be considered to have broadly satisfied the unbiasedness criterion.

Table 6: Test of Equality of Means

Core Inflation Measures

T statistics@

P-value

WPI Excluding Food

1.43

0.16

WPI excluding Fuel Group

0.28

0.78

WPI Excluding Food and Fuel group

2.40

0.01

Non-food Manufacturing

4.33

0.00

WPI excluding Fuel group and Basic Metal and Metal Products group

-0.23

0.82

WPI excluding Fuel group, Basic Metal and Metal Products group and non-food primary articles

1.46

0.15

@: Test of equality of means based on H0 : μ1=μ2 where μ1 and μ2 are means of headline and core inflation measure respectively.


Table 7: Unit Root Test for the Headline Minus Core (HMC) Variable

Core Inflation Measure

ADF Test statistic

P value

WPI excluding Food and fuel group

-3.1

0.03

Non-food Manufacturing

-3.2

0.02

Although most evaluations of core inflation measures do not include breadth as a criterion, some observers believe that, in general, monetary policy should focus on a core measure that even after exclusion is broad enough to capture at least more than 50 per cent of the commodity base. Going by that logic, all the six exclusion-based measures that were considered in this paper, including non-food manufacturing are broad enough to be considered as candidates for core inflation measures.

Of the six measures of core inflation that were tested, two measures, viz., WPI ex-food, and ex-food and fuel were more volatile than headline inflation by all criteria. Volatility around trend of these two measures was also higher than the headline WPI. However, of the other four, three core measures viz., ex-fuel, ex-fuel metal, and ex-fuel metal non-foodprim were less volatile than headline inflation and they also performed well in tracking the trend behaviour. In the case of non-food manufacturing, while volatility was lower in terms of three measures – standard deviation, mean absolute change and volatility around trend – it was higher than headline in terms of coefficient of variation (CV). Notwithstanding the importance of CV, it may be noted that most studies on core inflation emphasise the other three measures of volatility. CV is generally considered better than standard deviation when the units are different so that the standard deviations are normalised by their means. However, given the same unit in the case of core measures, standard deviation that essentially represents the volatility around mean is considered more direct and appropriate for comparison purposes and further normalisation of it by mean is not felt necessary. Overall, considering that non-food manufacturing has performed better than headline in three out of four criteria of volatility, it could be considered as a candidate for core measure along with the other three as long as it satisfies the other core properties well as tested in the following sections.

Table 8 sets out the graphical representation of the 6 exclusion-based core inflation measures.

It is interesting to note that even though food as a group (including both food articles and food products) is more volatile than headline WPI, the core inflation measure arrived at by excluding food from headline is more volatile. This is essentially because of the potentially offsetting co-movements among the other components and the volatility of the food component is not high enough to compensate for that. This has also been observed in case of other economies (Khettry, 2006). Clark (2001) also argues that it could be reasonable to include food prices in a measure of core inflation given that it is less volatile. If the food and energy price shocks are transitory, with upward spikes that are quickly reversed leaving the medium-term aggregate price path unchanged, policymakers would probably like to ignore them. However the latest behaviour shows that food prices in India are developing a structural component because of rise in aggregate demand for certain protein items such as pulses, egg, meat, fish and milk with higher incomes (Gokarn, 2010). Also, because of better distribution systems, the volatility that was observed in food prices, say, in the 1990s was not observed in the 2000s. Thus, the reduction in volatility in food prices is due to two reasons: reduction in volatility in the transitory component and increasing importance of the structural component determined by increasing income levels. It is the latter that has led food inflation to stabilise at a higher mean level providing the justification for their inclusion in core measures. A study by BIS (2008) also has shown that given the persistent nature of food price inflation in many EMEs, food price inflation’s predictability of headline inflation is increasing, which is not the same as that of energy.

As CPI is the headline inflation in many countries and a core CPI based on exclusion-based criteria is usually used by policymakers, an attempt was also made to examine the volatility of CPI-IW (industrial workers) (base 2001) vis-a-vis core CPI-IW measure by excluding food, fuel and both (Table 9). Exclusion-based measures of CPI core in India were found to be more volatile than overall CPI inflation in terms of all three criteria, thus suggesting that they cannot be used as core inflation measures.

Table 9: Headline and Exclusion based Core Inflation CPI Measures
(April 2005 to May 2011)

 Measure

Weight

Mean

Standard Deviation

Mean Absolute Change

Volatility around Trend

Headline CPI- IW

100.0

9.40

2.85

0.72

1.88

CPI Excluding Food

51.5

7.42

3.37

1.05

1.66

CPI excluding Fuel Group

93.6

9.65

3.01

0.79

2.04

CPI excluding food and fuel group

45.1

7.67

3.71

1.20

2.06

Tests of Predictability

An important feature of a good core inflation measure is that it should help predict future headline inflation. In the literature, the most common approach to judge predictability is based on the idea that if current headline inflation differs from the core, overtime headline inflation should move toward core. Divergence between headline and core should be temporary. For example, when current headline inflation is below core, in future, headline inflation should rise and if current headline inflation is above core, it should decline in future.

The predictive content in alternative measures of core inflation can be gauged from regressions of the change in headline inflation from today to some point in the future on the current gap between core and headline inflation. This standard regression model has been used extensively in many other studies on core inflation (Clark, 2001, Cogley 2002, Lafleche, et al, 2006). The advantage of this gap approach model over other studies which simply regress headline inflation on lags of headline and core is that the difficulties posed by persistence in inflation are taken care of by taking the gap approach instead of taking lagged inflation as independent variables (Clark, 2001). The model used is specified below:

where x denotes certain activity variable.

Regression (1) was tested with 3 months, 6 months, 9 months, 12 months and 24 months line horizons. However, the best results were obtained with a forecasting horizon of 9 months (Table 10). The t-statistic for regression coefficients indicate that, among the exclusion based measures that were considered, co-efficient β was positive and statistically significant for WPI Ex-fuel, non-food mfg, WPI Ex-fuel metal and WPI Ex-fuel metal non-foodprim. Equations with last two variables had relatively high predictive power. Non-food manufacturing inflation also has some predictive power, but lower than the others. Results with 12 month horizon are set out in Annexure I.

Table 10: Testing for Headline Reverting to Core

Core Inflation Measures (h=9)

Coefficient β
(t-statistic)

Regression R2

WPI Excluding Food

-0.11 (0.18)

0.20

WPI excluding Fuel Group

2.53 (5.6)***

0.33

WPI Excluding Food and Fuel group

0.35 (0.82)

0.02

Non-food Manufacturing

0.72 (2.1)**

0.12

WPI excluding Fuel group and Basic Metal and Metal Products group

2.03 (8.7)***

0.55

WPI excluding Fuel group, Basic Metal and Metal Products group and non-food primary articles

1.77 (9.7)***

0.60

Figures in parenthesis indicate the t-statistic.
*, ** and *** indicates significance level at 10, 5 and 1 per cent level.

While β>0 is generally tested to examine whether headline reverts to core inflation (Clark, 2001, Lafleche et al, 2006), some studies also specifically test for the condition that α=0 and β=1, i.e., the hypothesis that headline inflation fully reverts to core within the desired time horizon. If α is not equal to 0, it allows for core inflation to be a biased predictor of headline inflation over a given sample period, for example, in cases where commodity price shocks are predominantly on the upside or downside. In the current study as well, we reject at the 5 per cent significance level the joint hypothesis that α equals zero and β equals 1, i.e., the hypothesis that headline inflation fully reverts to core within the time horizon of 9 months. This result is in line with the BIS (2008) study that has shown that for many EMEs, sharp movements in commodity prices have kept the headline higher than core, with α being significantly positive.

Attempt was made to check the predictive power of model (2) by incorporating change in lagged IIP growth as one of the activity variables along with 9-month forecasting horizon (results not reported here). With the inclusion of lagged IIP growth as an activity variable, predictive power increased in all cases, which was on expected lines.

We further investigated whether core inflation measures show any tendency of reverting to headline by considering the reverse regression (as tested in Lafleche et al, 2006 and BIS, 2008) i.e., whether current gap between core and headline impacts the gap between future and current core measure of inflation over the same 9 month horizon. Generally, if headline is higher than core measure due to some transient shock, it should not affect the core inflation in future, implying that β should be statistically insignificant. As can be seen from Table 11, all core inflation measures, except non-food manufacturing, show that β coefficient is statistically significantly albeit with lower explanatory power than previous regression. This shows that headline inflation is able to explain three measures of core, viz., Ex-fuel, Ex-fuel metal, and Ex-fuel metal non-foodprim. Thus, these three measures revert to headline inflation as well, thereby suggesting that after some time the supply shocks spill over to these measures of core. In view of this, they cannot serve as good measures of core inflation, despite their high predictive power. However, in the case of non-food manufacturing, it is not possible to reject the hypothesis that β = 0 suggesting that headline inflation does not influence the non-food manufacturing measure of core.

Table 11: Testing for Core Reverting to Headline

Core Inflation Measures (h=9)

Coefficient β
(t-statistic)

Regression R2

WPI excluding Fuel Group

1.39 (3.6)***

0.17

Non-food Manufacturing

-0.09 (-0.35)

0.02

WPI excluding Fuel group and Basic Metal and Metal Products group

0.81 (4.9)***

0.20

WPI excluding Fuel group, Basic Metal and Metal Products group and non-food primary articles

0.43 (0.16)***

0.16

Figures in parentheses indicate the t-statistic.
*, ** and *** indicates significance level at 10, 5 and 1 per cent level.

Causality

To further confirm the previous results, an attempt was also made to test whether there was any causal relationship between the headline and the core measures of inflation. This is referred to in the literature as the attraction conditions, i.e., core inflation should cause headline but not vice versa (Rodrigues, et al, 2002). In other words, headline inflation tomorrow will fall towards core inflation today. However, if inflation expectations are not stable, core inflation tomorrow will converge to headline today. If the inflation series are stationary, these attraction conditions can also be checked. Given that the headline and core inflation measures are stationary, Granger causality tests were performed to assess the direction of causality in the long run. Core inflation measure should generally be able to Granger cause headline and not vice versa. Of the six measures tested, only non-food manufacturing WPI satisfies the condition at 5 per cent level of significance. While some other measures such as Ex-food and Ex-food fuel do not Granger cause headline inflation, some other measures such as Ex-fuel metal and Ex-fuel metal non-foodprim exhibited a two-way causality, suggesting that shock to headline inflation get translated into expectation of higher inflation down the road leading to generalised inflation (Table 12). Thus, Granger causality suggest that only non-food manufacturing satisfies the attraction condition and corroborate the earlier findings.

Table 12: Granger Causality Test Results

Core Inflation Measures

Headline does not Granger cause Core

Core does not Granger cause Headline

F statistic

P value

F statistic

P value

WPI Excluding Food

0.94 (5)

0.46

1.65 (5)

0.16

WPI excluding Fuel Group

1.84 (4)

0.13

2.10 (4)

0.09*

WPI Excluding Food and Fuel group

1.86 (3)

0.14

1.87 (3)

0.14

Non-food Manufacturing

1.86 (3)

0.14

2.85 (3)

0.05**

WPI Excluding Fuel group and Basic metals and metal products

5.06 (3)

0.00***

4.06 (3)

0.02**

WPI Excluding Fuel group, Basic metals and metal products and non food primary articles

3.58 (3)

0.02**

2.10 (3)

0.09*

Lag length in the model chosen on the basis of SIC criteria in given in the brackets.
*, ** and *** indicates significance level at 10, 5 and 1 per cent level.

A inflation persistence test was also performed for the core inflation measures in respect of which headline inflation tended to converge to core inflation (Annex II). It was observed that while all the four measures of core inflation tested were persistent, non-food manufacturing inflation turned out to be relatively more persistent than others.

Thus, non-food manufacturing inflation that represents 55 per cent of the weight in headline WPI satisfies all the laid down criteria of a core measure. This also implies that 45 per cent of inflation component consists of non-core component which represents price movements caused by temporary shocks. In most countries, the share of core component is much larger, generally about 80 per cent (Table 13). This suggests that inflation management for monetary policy is much more challenging in India than in other countries.

Table 13: Weight of the Core Inflation measure: Cross Country

Country

Core Inflation Measure

Weight

Canada

CPIX that excludes 8 most volatile components like fruits, vegetables, gasoline, natural gas, fuel oil, mortgage interest costs, intercity transportation and tobacco products

82.2

Thailand

Core CPI excludes fresh food and energy prices which include rice, flour, cereal products, vegetables, fruits, electricity charges, cooking gas, and gasoline.

75.95

Philippines

CPI excluding rice corn fruits vegetables, LPG, Kerosene, Oil, Gasoline, Diesel

81.6

Peru

CPI excluding 9 volatile items like food, fruits and vegetables, urban transport

78.8

United States

CPI excluding food and energy

77.7

Chile

CPI excluding 20 per cent with higher (-) variations and 8 per cent with higher (+) variations

72

V. Summing Up and Some Final Reflections

The headline inflation rate is not wholly under the control of the central bank in the short run. Various economic developments beyond the control of the central bank may generate short run or transitory changes in the inflation rate. Hence, policymakers in many countries focus on the more persistent movements in inflation, called core inflation. A good measure of core inflation helps separate the noise from signals about current and future trends in inflation. If price fluctuations from non-monetary sources can be excluded, the resulting core inflation could be regarded as a measure of the inflation that is the outcome of policy. Therefore, measure of core inflation could be considered more controllable by the monetary authority than published inflation rates.

The Indian economy has been subject to repeated and significant supply side shocks from time to time. Examples of such shocks in the recent period were drought of 2009 and oil price rise in 2010 that kept headline inflation persistently high. In this context, the use of core inflation for monetary policy purposes assumes importance, particularly in terms of communicating to public what the Reserve Bank’s actions are trying to achieve. The Reserve Bank uses non-food manufacturing as an indicator of core. However, several other countries use headline minus food and energy as measures of core inflation. In this backdrop, an attempt was made to assess as to which measure of core inflation is more relevant in the Indian context.

For a core inflation measure to serve its purpose, it must satisfy three criteria. One, core measure of inflation should be more stable or less volatile than headline inflation. Two, over a long period of time, average rate of core inflation should match the average rate of headline inflation and there should be no system divergence below the two. Also, core inflation should be able to track with the trend rate of inflation. Three, if core inflation represents the underlying trend of inflation, it should then be better able to predict total or headline inflation.

‘Fuel and power’ group of WPI was found to be the most volatile in India, followed by primary articles and manufactured products. Within fuel group, mineral oils and coal sub-groups were more volatile. Within primary articles, minerals and non-food articles sub-groups were more volatile. Significantly, food articles sub-group was about half volatile as the non-food articles group and about 1/3rd less volatile than the minerals sub-group. Interestingly, studies in respect of many other EMEs also find that food inflation is less volatile than fuel. This perhaps is due to the reason that the rise in food prices is becoming persistent.

Considering the volatility of various groups/sub-groups of WPI, the study tested six measures of core inflation, viz., WPI excluding food; WPI excluding fuel; WPI excluding food and fuel; non-food manufacturing; WPI excluding fuel and basic metals and metal products; and WPI excluding fuel, metal group and non-food primary articles. These measures were tested for three properties, viz., (i) volatility; (ii) unbiasedness and capability to track trend; and (iii) predictive power. A measure of core inflation, if it is to serve its intended purpose, should not revert to headline inflation. This condition, therefore, was also tested.

Of the six measures tested, two measures, viz., WPI excluding food; and WPI excluding food and fuel did not perform well in terms of volatility, tracking the trend behaviour and predicting the future inflation. However, the remaining four measures, viz., WPI excluding fuel; non-food manufacturing inflation; WPI excluding fuel and basic metals and metal products; and WPI excluding fuel, metal and non-food primary articles broadly satisfied the conditions relating to volatility, unbiasedness and tracking the trend, and predictability.

To test the predictive power of measures of core inflation, change in headline inflation from today to some points in the future (6 months, 9 months, 12 months and 24 months) was regressed on the current gap between core and headline inflation. This is the standard regression model used in many other studies. Although equations were tested with different time horizons results, however, were better with a time horizon of 9 months. That is, headline inflation converges to four core measures of inflation in nine months. However, headline does not covert to core measure of inflation fully. It appears that sharp movements in commodity and food prices have kept the headline higher than the core. Explanatory power of regression with only core as the independent variable is generally not very high. The explanatory power of the equation improves with the inclusion of activity variables.

Even though the four measures satisfied the three criteria, it needs to be ensured that these measures of core inflation also satisfy the condition they do not revert to headline inflation. It will suggest that supply side shocks spill over to these core measures of inflation as a result of which inflation becomes generalised. Such measures, as a result, could not serve as good measures of core inflation even though they may have high predictive power. Of the four measures, it was found that all measures, except non-food manufacturing, revert to headline inflation. Non-food manufacturing measure does not reverse to headline. It, thus, performed well in all the criteria.

Granger causality tests were also carried out to confirm that headline inflation reverts to core and not vice versa. Of the six measures tested, only non-food manufacturing WPI satisfies the causality condition. While two measures, viz., WPI ex-food and WPI ex-food and fuel do not Granger cause headline inflation, two other measures such as WPI ex-fuel metal; and WPI ex-fuel metal non-food primary articles, exhibited a two-way causality, suggesting that shock to headline inflation get translated into expectation of higher core inflation (WPI ex-fuel metal and WPI ex-fuel metal non-food primary articles) down the road leading to generalised inflation. Inflation persistence tests showed that while all the four core inflation measures considered were all persistent, non-food manufacturing inflation turned out to be relatively more persistent than others.

The Study also tested core measure of inflation based on CPI-IW by excluding food and fuel separately and also together. However, measures of inflation based on CPI-IW were more volatile than the CPI inflation.

Thus, non-food manufacturing, which the Reserve Bank uses as a measure of demand side pressures, is the only measure which satisfies all the properties of a core measure. One weakness of exclusion-based measure is that the appropriate components to exclude may change overtime. Central banks, therefore, have tried to address this concern by two ways: (i) periodically re-evaluating the behaviour of prices and (ii) tracking a range of core inflation measures instead of one. It, therefore, is felt that all the six measures of core and some others as are considered necessary need to be reviewed periodically.

Non-food manufacturing component represents 55 per cent of the weight in WPI. It, therefore, suggests that 45 per cent of inflation component consists of non-core component which represents price movements caused by temporary shocks. In most countries, the share of core component is much larger, generally about 80 per cent. This suggests that inflation management by monetary policy is much more challenging in India than in other countries.

A core measure of inflation is not an end in itself, but rather a means to achieve low and stable inflation. Given the loss of information content in the construction of core inflation and the relatively greater public acceptability of the headline inflation, the core measures are useful only as indicators of the underlying inflationary process rather than as policy targets. The measure of headline inflation is based on the solid theory of the cost of living. Thus, containing headline inflation, and not core inflation, should be the focus of monetary policy. That is, core inflation measures cannot substitute for measures of headline inflation. However, they could be used to serve as a short-term operational guide for monetary policy.


References:

Armour, J. 2006 “An Evaluation of Core Inflation Measures.” Bank of Canada Working Paper No. 2006–10.

Bryan, M.F., and Cecchetti S.G. 'Measuring Core Inflation', NBER Working Paper No.4303, March 1993.

Cecchetti, S and R Moessner (2008), “Commodity prices and inflation dynamics”, BIS Quaterly Review, December

Cogley, T. 2002. “A Simple Adaptive Inflation.” Journal of Money, 43 (1): 94–113. Hogan, S., M. Johnson, and T.

Clark, T (2001): “Comparing measures of core inflation”, Federal Reserve Bank of Kansas City Economic Review, vol 86, no 2 (second quarter), pp 5–31.

Das Abhiman, J Joice, S Singh (2009), “Measuring Core Inflation in India”, Indian Economic Review, Vol 44, No 2, July- December

Eckstein, Otto. 1981. Core Inflation. Englewood Cliffs, N.J.: Prentice-Hall.

Gokarn Subir (2010), “The Price of Protein”, Inaugural Address by Deputy Governor, Reserve Bank of India at Special Conference in honour of Dr. Kirit Parikh at IGIDR, Mumbai on October 26.

Kar Sujata (2009), “Statistical tools as Measures of Core Inflation for India”, Indian Economic Review, Vol 44, No 2, July- December

Laflèche, T and J Armour (2006): “Evaluating measures of core inflation”, Bank of Canada Review, Summer, pp 19–29.

Marques, C, P Duarte Neves and L Sarmento (2003): “Evaluating core inflation indicators”, Economic Modelling, vol 20, pp 765–75.

Mohanty, Deepak, D. P. Rath and M. Ramaiah. 'Measures of Core Inflation for India'. Economic and Political Weekly, February 2000.

Mohanty, Deepak, “Changing Inflation Dynamics in India”, Speech delivered at the Motilal Nehru National Institute of Technology, Allahabad on 13th August.

Quah D., and Vahey, S.P. 'Measuring Core Inflation', The Economic Journal, 105, September 1995.

Reserve Bank of India (2005), Report of Currency and Finance, 2003-04.

...........(2006) - Annual Policy Statement, 2006-07.

..........(2010) – Annual Policy Statement, 2010-11.

Roger, Scot. 'A Robust Measure of Core Inflation in New Zealand', Reserve Bank of New Zealand Discussion Paper, G97/7,1997.

Samanta, G. P. 'Core Inflation in India: Measurement and Policy Perspectives'. Reserve Bank of India Occasional Papers, Vol. 20, No. 1, Summer, 1999.

Shiratsuka S (1997). 'Inflation Measures for Monetary Policy: Measuring the Underlying Inflation Trend and its Implication for Monetary Policy Implementation', Institute for Monetary and Economic Studies, Bank of Japan, December 1997.

Wynne, Mark A. 1999. “Core Inflation: A Review of Some Conceptual Issues,” Federal Reserve Bank of Dallas, research department working paper no. 99-03, June.

Wynne, M. 'Core Inflation: Some Conceptual Issues', Bank of England Quarterly Bulletin, Vol. 39, No.4, November 1999.


Annex I: Testing for Headline reverting to Core at 12 months horizon

Core Inflation Measures (h=12)

Coefficient β
(t-statistic)

Regression R2

WPI Excluding Food

-0.11 (0.18)

0.20

WPI excluding Fuel Group

2.6 (5.8)***

0.36

WPI Excluding Food and Fuel group

-0.01 (0.10)

0.01

Non-food Manufacturing

0.68 (1.90)*

0.10

WPI excluding Fuel group and Basic Metal and Metal Products group

2.17 (9.64)***

0.61

WPI excluding Fuel group, Basic Metal and Metal Products group and non-food primary articles

1.97 (12.4)***

0.72

Figures in parenthesis indicate the t-statistic.
*, ** and *** indicates significance level at 10, 5 and 1 per cent level.


Annex II: Persistence of the Exclusion-Based Core Measures

Core Inflation Measures

Based on SBC

Based on AIC

Extent of Persistence

Chosen lag length

Extent of Persistence

Chosen lag length

WPI excluding Fuel Group

0.926*

1

-

1

Non-food Manufacturing

0.942*

1

-

1

WPI excluding Fuel group and Basic Metal and Metal Products group

0.929*

1

0.917*

12

WPI excluding Fuel group, Basic Metal and Metal Products group and non-food primary articles

0.922*

1

0.919*

12

* Denotes significance at 1 per cent level.


*Dr. Janak Raj and Sangita Misra are Adviser-in-Charge and Assistant Adviser, respectively, in the Monetary Policy Department. The views expressed in the article are completely of the Authors and do not reflect the RBI’s view. The usual disclaimer applies. The authors can be reached at janakraj@rbi.org.in and sangitamisra@rbi.org.in.

 
 
© Reserve Bank of India. All Rights Reserved.
Best viewed in 1024x768 resolution in IE 5 and above.