MEASURING CORE INFLATION IN ROMANIA

MEASURING CORE INFLATION IN MEASURING CORE INFLATION IN ROMANIAROMANIA

Dissertation Paper

Student: ANGELA-MONICA MĂRGĂRITSupervisor: Professor MOISĂ ALTĂR

July 2003

ACADEMY OF ECONOMIC STUDIES BUCHARESTDOCTORAL SCHOOL OF FINANCE AND BANKING

I. INTRODUCTION II. THEORETICAL BACKGROUND

III. DATA AND ECONOMETRIC ESTIMATION

IV. EVALUATING CORE INFLATION INDICATORS

V. CONCLUDING REMARKS

I. INTRODUCTION

Reasons of using CORE INFLATION indicators : -- inflation targeting strategy -- better controlled by the monetary authority -- good predictor of future inflation

CORE INFLATION= the persistent component; the trend of CPI inflation; the common component of all prices

Different definitions of core inflation different methods of estimation.

GOAL: estimating and choosing the best core inflation measure for Romania, considering the established criteria

1. Central-bank approacha) “Zero-weighting” technique

• often used in practice and easy explainable to the public• excludes volatile items of CPI: administrated prices, seasonal or interest rate sensitive components

• disadvantage: arbitrary basis in removing CPI items

b) Trimmed mean method (Bryan &Cecchetti-1994)

• argument : distribution of individual price change is skewed & leptokurtic

• cuts % from both tails of price change distribution• theoretical model: price setting with costly price adjustment (Ball & Mankiw -1994)

II. THEORETICAL BACKGROUND

Core inflation= persistent component of measured price index, which is tied in some way to money growth (Bryan &Cecchetti - 1994,1997)

core=m*

i firms where ei (shock in production costs) exceeds the

“menu costs”: i=m*+ei

The change of aggregate price level depends on the shape of shocks (supply shocks) distribution:

- symmetricalCPI inflation= c - asymmetricalCPI inflation> or< core

2. Quah & Vahey approach and extensions

Core inflation= the component of measured inflation that has no impact on real output in the medium-long run (Quah & Vahey -1995).

on the basis of vertical long run Phillips Curve

• placing long- run restrictions on a VAR system in: real output and inflation

• Blachard& Quah decomposition for identifying the 2 structural shocks: -- non-core shock

-- core shock

Identification steps:

Step 1: Reduced form VAR in first differences of real output & CPI : Xt =+ B(L)et , var(et) =ee’=

Step 2: Xt = +C(L)t, var(t) = ; Cot = et; CoCo’ = Step 3: Identifying Co:• orthogonality and unit variance of t: n(n+1)/2 restrictions.• n(n-1)/2 long run restrictions C(1) triangular Step 4: Core inflation recovered considering non-core zero recomputed shocks from t = Co-1 et. For 2 variables:

...

..

2221

1211

0 jtcore

jtnoncore

jcjc

jcjc

Pt

Yt

j

jtcorejcccorejcjj

..*2201200

Long run restriction:

Extensions of Quah & Vahey method

• more variables: adding a monetary indicator

• Core shocks: -- monetary shocks -- real demand shocks

• Blix(1995),Fase&Folkertsma (2002)monetary aggregate

• Gartner & Wehinger (1998), Dewachter & Lustig(1997) short term interest rate

jtdemjcjtmonjcccorejcjcjcjj jjj

.*33.*32023,013,01200 000

III. DATA AND ECONOMETRIC ESTIMATIONSAMPLE 1996:01 - 2002:12

Lxy is natural logarithm of xy variable ( LCPI = ln(CPI)); DLxy is the first difference of Lxy ( DLCPI(t) = LCPI(t) – LCPI(t-1) is the monthly inflation rate). Ixy index as against January 1996)

ESTIMATION RESULTS:1. “Zero - weighting” methodCORE0

Excluded items (26.27% of CPI basket): • Administrated prices (18.77%) - electric energy, gas, central heating - water, salubrity - mail & telecommunications - urban & interurban transport • Seasonal prices (7.5%) - fruits & tinned fruits - vegetables & tinned vegetables

0

4

8

12

16

20

24

28

32

1996 1997 1998 1999 2000 2001 2002

CORE0 CPI inflation

%

2. Trimmed mean estimationTRIM

0

5

10

15

20

25

30

0.00 0.05 0.10 0.15 0.20 0.25

Series: DLCPISample 1996:01 2002:12Observations 84

Mean 0.034762Median 0.025760Maximum 0.267542Minimum 0.003860Std. Dev. 0.036256Skewness 4.104993Kurtosis 23.98776

Jarque-Bera 1777.615Probability 0.000000

DLCPI (CPI inflation) series

• highly asymmetric and leptokurtic inflation distribution• Average weighted skewness=1.0439• Average weighted kurtosis = 19.784

• Symmetric trimming: 5%, 10%, 15%, 18%, 30%• Trimming a higher percent more stable indicator of core

inflation

-5

0

5

10

15

20

25

30

1996 1997 1998 1999 2000 2001 2002

CPI inflationTRIM5TRIM10TRIM15TRIM18TRIM30

%2. Trimmed mean estimationTRIM

3. Quah & Vahey approachCORE

a) SVAR 1: DLY_SA, DLCPI and a constantCORE2

SVAR1 tests: stability, lag length & residuals

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Inverse Roots of AR Characteristic Polynomial

-30

-20

-10

0

10

20

30

1996 1998 1999 2000 2001 2002

CUSUM 5% Significance

-30

-20

-10

0

10

20

30

1996 1998 1999 2000 2001 2002

CUSUM 5% Significance

.00

.05

.10

.15

-.04

-.02

.00

.02

.04

1996 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

.00

.05

.10

.15-.08

-.04

.00

.04

.08

1996 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

Parameters stability tests:Eq. DLY_SA Eq. DLCPI

b) SVAR 2: DLY_SA,DLCPI,constant & Dummy March 1997CORE2d

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Inverse Roots of AR Characteristic Polynomial

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1997 1998 1999 2000 2001 2002

CUSUM of Squares 5% Significance

.00

.05

.10

.15-.04

-.02

.00

.02

.04

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1997 1998 1999 2000 2001 2002

CUSUM of Squares 5% Significance

.00

.05

.10

.15-.04

-.02

.00

.02

.04

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

SVAR2 parameters stability:Eq DLY_SA Eq DLCPI

b) SVAR 3: DLY_SA, DLM2_SA, DLCPI, constant CORE3

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Inverse Roots of AR Characteristic Polynomial

Parameters stability tests:Eq DLY_SA Eq DLM2_SA Eq DLCPI

-30

-20

-10

0

10

20

30

1997 1998 1999 2000 2001 2002

CUSUM 5% Significance

-30

-20

-10

0

10

20

30

1997 1998 1999 2000 2001 2002

CUSUM 5% Significance

-30

-20

-10

0

10

20

30

1997 1998 1999 2000 2001 2002

CUSUM 5% Significance

.00

.05

.10

.15-.04

-.02

.00

.02

.04

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

.00

.05

.10

.15

-.08

-.04

.00

.04

.08

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

.00

.05

.10

.15-.08

-.04

.00

.04

.08

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

CHSQ(1) =0.831 [0.361]; CHSQ(1)=1.130 [0.252]; CHSQ(1)=0.104 [0.745] (Ramsey RESET test 1 fitted term)

b) SVAR 4: DLY_SA, DLM2_SA, DLCPI, constant, Dummy March 1997 CORE3d

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5

Inverse Roots of AR Characteristic Polynomial

Parameters stability tests:Eq DLY_SA Eq DLM2_SA Eq DLCPI

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1997 1998 1999 2000 2001 2002

CUSUM of Squares 5% Significance

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1997 1998 1999 2000 2001 2002

CUSUM of Squares 5% Significance

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1997 1998 1999 2000 2001 2002

CUSUM of Squares 5% Significance

.00

.05

.10

.15-.04

-.02

.00

.02

.04

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

.00

.05

.10

.15-.08

-.04

.00

.04

.08

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

.00

.05

.10

.15-.06

-.04

-.02

.00

.02

.04

.06

1997 1998 1999 2000 2001 2002

N-Step Probability Recursive Residuals

CHSQ=1.718 [0.189] CHSQ=2.180 [0.139] CHSQ=0.458 [0.497] (Ramsey RESET test 1 fitted term)

IV. EVALUATING CORE INFLATION INDICATORSA) Quah & Vahey core inflation measures & economic content

SVAR1 CORE2

0

4

8

12

16

20

24

28

32

1996 1997 1998 1999 2000 2001 2002

CPI inflation CORE2 -SSV

VA

AR

R1

1))

%

.00

.01

.02

.03

.04

2 4 6 8 10 12 14 16 18 20

Accumulated Response of DLY_SA to ShockNON-CORE

.00

.01

.02

.03

.04

2 4 6 8 10 12 14 16 18 20

Accumulated Response of DLY_SA to Shock CORE

-.02

-.01

.00

.01

.02

.03

2 4 6 8 10 12 14 16 18 20

Accumulated Response of DLCPI to Shock NON-CORE

-.02

-.01

.00

.01

.02

.03

2 4 6 8 10 12 14 16 18 20

Accumulated Response of DLCPI to Shock CORE

Accumulated Response to Structural One S.D. Innovations

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10

Shock NON-CORE Shock CORE

Variance Decomposition of DLY_SA

0

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10

Shock NON-CORE Shock CORE

Variance Decomposition of DLCPI

Non-core shocks supply shocks;Core shocks demand shocks

96%

88%

SVAR2 CORE2d

0

4

8

12

16

20

24

28

32

1996 1997 1998 1999 2000 2001 2002

CPI inflation CORE2d (SVAR2)

%

• strong inertial character of

inflation • administrated & seasonal prices or supply shocks are not determinant inflationary sources

SVAR3 CORE3

-5

0

5

10

15

20

25

30

35

1996 1997 1998 1999 2000 2001 2002

CPI inflation CORE3 (SVAR 3)

%

60

80

100

120

140

160

1996 1997 1998 1999 2000 2001 2002

Y_SA WL_SA EMPL

%

LNONCORE3= DLCPI - LCORE3-3

-2

-1

0

1

2

3

4

1996 1997 1998 1999 2000 2001 2002

NONCORE3

%

Test statistics: 1. Serial correlation LM: F-statistic 0.593 [0.837]; Obs*R-squared 7.229 [0.842] 2. White heteroskedasticity: F-statistic 0.595 [0.857]; Obs*R-squared 9.168 [0.820][ [ ] P-VALUE 3. Ramsey’s test (2 fitted): F-statistic 0.042 [0.958]; Loglikelihood ratio 0.098 [0.951] 4. Normality: Jarque-Bera 0.777168 [0.678016]

SVAR4 CORE3d

-4

0

4

8

12

16

20

24

28

32

1996 1997 1998 1999 2000 2001 2002

CPI inflation CORE3d (SVAR4)

%

B) Choosing the best core inflation indicator

CRITERIA: Bryan & Cecchetti (1994), Roger(1997), Marques (2000),

Valkovszky & Vincze(2000), H. Mio (2001)

1. Core & CPI inflation correlation2. Cointegration condition3. Moving average methods & efficient core indicators4. Core measures & the correlation with money growth

1. Core & CPI inflation correlation

• Correlation coefficients: higher for TRIM• Granger causality tests DLCPI - CORE indicators

2. Cointegration condition

LICPI96=0.884023*LICORE3+0.257784 Speed of adjustment (-0.114694, –0.099032)

Long run relation (4 lags in differences):

LICPI96 & LICORE3 (log of index base Jan. ‘96)

3. Moving average methods & efficient core indicators

n

itt MACORE

NRMSE

1

2)(1

n

itt MACORE

nMAD

1

1

TRIM18 - The best core indicatorCORE3 - the best among Quah & Vahey core indicators

4. Core measures & the correlation with money growth

• Granger causality tests CORE measures - DLM2_SA - Core should be Granger caused by money growth & not reverse

TRIM18 performs better in the long run

• Inflation indicators variability

V. CONCLUDING REMARKS

• Core inflation indicators closely follow the CPI inflation

• Decreasing variability of TRIM & Exclusion methods;

• TRIM18 would be recommended as the optimal core indicator

• Quah & Vahey indicators perform less successful, but are signaling links in economic variables