Inflation Persistence

21
Grup Riset Ekonomi Departemen Kebijakan Ekonomi dan Moneter Desember 2016 MHA Ridhwan Angsoka Yorintha Paundralingga Melva Viva Grace Juan Samuel CR/2/GRE-DKEM/2016 CATATAN RISET Regional Inflation Persistence Dynamics: A Case of Indonesia Alt: A Regional Analysis of Inflation Persistence Dynamics in Indonesia

Transcript of Inflation Persistence

Page 1: Inflation Persistence

Grup Riset Ekonomi

Departemen Kebijakan Ekonomi dan Moneter

Desember 2016

MHA Ridhwan

Angsoka Yorintha Paundralingga

Melva Viva Grace

Juan Samuel

CR/2/GRE-DKEM/2016

CATATAN RISET

Regional Inflation Persistence

Dynamics: A Case of Indonesia Alt: A Regional Analysis of Inflation Persistence Dynamics in Indonesia

Page 2: Inflation Persistence

1-1

1. Introduction

As an archipelagic economy, Indonesia’s economy has historically experienced

relatively high and persistent inflation rate (Ridhwan, 2014). Nonetheless, in early

2000s, a number of Asian central banks, including Indonesia, have adopted explicit

inflation targeting approach as their monetary policy frameworks (Filardo & Genberg,

2010a). This new anchor for monetary policy following the abandonment of a fixed

exchange rate regime was needed to tame this high inflation.

Second properties of Indonesia’s inflation, high persistence, is also a central

issue to understand the inflation process. Undeniably, quantifying inflation

persistence is essential for central bank to achieve its goal. Meller and Nautz (2012)

survey monetary policies taken by central banks where inflation persistence had been

increasingly used as an indicator of policy effectiveness.

Benati (2008) evaluates estimated persistence across different monetary

regimes from several developed countries over long samples. His finding is that regimes

that clearly anchor inflation generate less persistence inflation. The clear anchor

inflation could be in the form of gold standard or inflation targeting framework. His

research is in-line with Levin, Natalucci & Piger (2004) who found that inflation has

become less persistent in countries that implement inflation targeting framework.

Therefore, after the implementation of inflation targeting framework, the measure of

inflation persistence serves other purposes, namely the ability of the central bank to

control inflation. In this regard, Phiri (2016) argues that high inflation persistence

indicates the inability of central banks to control inflation such that any deviations of

inflation from its steady-state will ensure that inflation does not easily adjust back [to]

its long-run equilibrium. In other words, high inflation persistence has

profound consequences, since it can weaken the central bank’s credibility and public

trust in the monetary policymakers. Jain (2013) explains that if the public did not believe

that the central bank’s policy is credible, inflation that is driven by simple shock and

Page 3: Inflation Persistence

1-2

supposed to only rise temporarily, then fall back, would instead rise for a long time,

despite the central bank’s attempt to control it.

Roache (2014) compares inflation targeting countries to find relationship

between persistence and ability of monetary policy to accommodate temporary price

shocks, which later be defined as policy space. He found that the higher the persistence,

the smaller the policy space.

Several researchers, such as Fuhrer (1994), Mishkin (2007), Ascari and Ropele

(2012) introduced the term “sacrifice ratio” which measures output growth lost

required to reduce inflation by a percentage point. In short, this ratio represents lower

costs of disinflation. They argue that higher level of inflation persistence is related to

higher sacrifice ratio. Furthermore, this ratio has implications for the output cost of

monetary policy, in which higher persistence implies much costly disinflation monetary

policy.

Cecchetti and Debelle (2006) argue that if inflation persistence has changed to

a lower level, monetary policy’s attempt to reduce inflation will become easier, since

people’s tendency to anchor inflationary expectations to the past will be lessened. In

other words, lower persistent inflation will sustain low inflation rate given that a shock

boosting inflation now would have a less extended impact on the rate of later inflation.

Thus, lower persistent inflation will reduce the cost of disinflation. Fundamentally, the

nature of inflation persistence dynamics indicates important implications for monetary

policy.

It is also important for the central bank to have a good understanding of the

underlying patterns and determinants of regional inflation persistence. A good

understanding of the underlying patterns and determinants which led to the

heterogeneity across region would enable the monetary policymaker to more

accurately forecast regional inflation and construct appropriate monetary policy. It

could also help the regional government in constructing complementary structural

policies in regions whose inflationary processes are less responsive to a common

monetary policy. Inflation persistence varies in its level, magnitude, and variance

Page 4: Inflation Persistence

2-3

among regions. Then explaining this phenomenon remains a challenge, particularly for

a country that is as large and spatially diverse as Indonesia. This heterogeneity reflects

different regional structural rigidities which could reduce each region’s capacity to

adjust to common and national shocks and the corresponding policy responses. Several

literatures suggest that inefficiency in a regional labor market may explain different

responsiveness of inflation to a common shock.

In this context, the objective of this paper is to estimate the degree of

persistence for Indonesia’s inflation. The degree of persistence in inflation in every

region provides an important test concerning the effectiveness of monetary policy in

controlling inflation. Unfortunately, only few papers investigate the heterogeneity in

the inflation process at a regional level. Several papers, such as Cecchetti et al. (2002),

study the regional dynamics of relative prices across the US. Beck and Weber (2005a,b)

analyze inflation rate dispersion across regions in the US, Japan, Canada and EMU

regions. Studying Italy, Vaona (2009) was the first researcher to study regional inflation

persistence. With its huge disparity between northern and southern region, Italy is an

ideal laboratory to study how inflation persistence may interact with regional

disparities. Benigno (2004) argues that central banks should overweight regions, within

their target index, with stickier price developments and underweight more flexible

regions in order to avoid the former ones that bear a disproportionate part of the

adjustment process following a monetary shock.

The rest of the paper is organized as follows. The next section explains our

methodology used to examine the regional inflation persistence. Section 3 presents the

empirical results. Section 4 concludes.

2. Estimating Persistence at the Regional Level

Fuhrer (2009) explains the persistence using physics theory of inertia, which is

defined as the tendency of a body with some mass to resist acceleration and to remain

at rest, unless acted upon by a force. The greater the mass, the greater the inertia, i.e.

the greater the force required to accelerate the body. Similarly, inflation shows a

Page 5: Inflation Persistence

2-4

tendency to stay near its equilibrium value. To move inflation from its current level, it

requires economic “force”. Thus the greater the economic force required to move it

from its recent level, the more persistent it is. Willis (2003) defines persistence as the

speed of adjustment of inflation returns to its equilibrium after a shock or after a

deviation. Technically, European Central Bank establishes Inflation Persistence Network

(IPN) research project and define inflation persistence as “the tendency of inflation to

converge slowly (or sluggishly) towards its long-run value following a shock which has

led inflation away from its long-run value”. Vladova and Pachedjiev (2008) argue that

this definition implies to two important subject, i.e., the speed of adjustment and the

equilibrium level of inflation.

The measurement of inflation persistence has two types of analysis. First,

structural persistence represents inflation persistence that comes from known

economic disturbances. Second, reduced-form persistence refers to the empirical and

econometric property of inflation without any economic interpretation. In this paper,

we study reduced-form persistence which is represented using a univariate

autoregressive (AR) function.1 This approach lies on the assumption of stationary

inflationary process. Technically, a non-stationary variable will vary a lot and cross the

estimated mean infrequently. Halunga, Osborn and Senseir, (2009) argue that inflation

targeting is meaningless when inflation is more likely to contain a unit root. If inflation

follows a random walk pattern, then the best forecast of tomorrow’s inflation is today’s

inflation as the most recent observed inflation. Nevertheless, the average value could

not be the predictor of inflation level. Hence, it has been the standard practice in

the empirical works to firstly diagnose the integration properties of a univariate

autoregressive series of inflation. Moreover, several measures of persistence will be

derived from its autocorrelation function for inflation. Examples of such measures

include the integer order of integration, half-life of responses to shocks, the largest

autoregressive root, and the sum of the autoregressive coefficients. Pivetta and Reis

1 While the reduced-form analysis is related with the univariate model, the structural analysis is associated with the multivariate models, such as vector autoregressive models. To obtain measures of persistence in multivariate models, other variables are also used to make inference about the persistence of inflation.

Page 6: Inflation Persistence

2-5

(2001) define half-life as the number of periods in which inflation is at least half as large

as the initial shock. Other researchers, such as Cogley and Sargent (2001), use the

largest autoregressive root as persistent indicator. Andrews and Chen (1994) suggest

that the sum of autoregressive coefficients (SARC) is still the best scalar measure of

persistence. The inflation series is said to contain a unit root if SARC is greater than or

equal to unity. In other words, shocks to inflation has a permanent effect on the price

level. In this case, the inflation is persistent and will never returns to its mean value or

its equilibrium level. Therefore, following Altissimo, Ehrmann and Smets (2006), we

defined inflation persistence as a tendency for the inflation to converge into the long-

run equilibrium slowly after a shock that brings inflation afar from its long-run

equilibrium.

Inflation Persistence

Inflation persistence that exists in every province in Indonesia is measured

assuming that inflation follows a stationary autoregressive process of order or AR(m) as

follow:

𝜋𝑖,𝑡 = 𝛼𝑖 + ∑ 𝛽𝑖,𝑗𝜋𝑖,𝑡−𝑗 + 휀𝑖,𝑡

𝑚

𝑗=1

(1)

In which:

𝜋𝑖,𝑡 = Inflation rate in a province 𝑖 at time 𝑡 (% yoy)

𝑚 = Selected optimal lag length by means of a standard information criterion (SIC)

휀𝑡 = Random disturbance error term

Afterwards, to look at the degree of inflation persistence in the context of AR coefficient

summation, equation (1) can be written as follow:

𝜋𝑖,𝑡 = 𝛼𝑖 + 𝜌𝑖𝜋𝑖,𝑡−1 + ∑ 𝛿𝑖,𝑗∆𝜋𝑖,𝑡−𝑗 + 휀𝑖,𝑡

𝑚−1

𝑗=1

(2)

In which:

𝜌𝑖 = ∑ 𝛽𝑖,𝑘

𝑚

𝑘=1

= Sum of AR coefficient (SARC), inflation persistence parameter that is achieved from the result of AR coefficient calculation using time series model

Page 7: Inflation Persistence

2-6

𝛿𝑖,𝑗 = − ∑ 𝛽𝑖,𝑘

𝑚

𝑘=1+𝑗

= Simple form of 𝛽

Parameter 𝜌 can be interpreted as an inflation velocity to return to the equilibrium after

a shock. Equation (2) can be reparameterized as follow:

∆𝜋𝑖,𝑡 = (𝜌𝑖 − 1) [𝜋𝑖,𝑡−1 − 𝜏𝑖] + ∑ 𝛿𝑖,𝑗∆𝜋𝑡−𝑗 + 휀𝑡

𝑚−1

𝑗=1

(3)

where 𝜏𝑖 =𝛼𝑖

1−𝜌𝑖. Equation (3) has a basic specification from the Augmented Dickey

Fuller (ADF) to assess the unit root existence. Zero hypothesis to test ADF is a process

to get unit root (𝜌𝑖 = 1), when 𝜌𝑖 = 1 has a series that are random walk, non-

stationary, and has high persistency because there is an autocorrelation with that lag

series. A shock that occurs will achieve a long-run effect and need time to converge into

its long-run equilibrium or be divergent from its long-run path permanently. Dorsche

(2005) said if 𝜌𝑖 value < 0.5, then inflation data can be assumed to have low persistency.

To test the robustness’ result for the test that was done, Levin and Piger (2004)

consider a change in the average inflation rate after calculating the structural changes

that happen all the time. To explain the structural changes, we can add time- varying

mean as follow:

∆𝑍𝑖,𝑡 = (𝜌𝑖 − 1) 𝑍𝑖,𝑡−1 + ∑ 𝛿𝑖,𝑗∆𝑍𝑖,𝑡−𝑗 + 휀𝑖,𝑡

𝑚−1

𝑗=1

(4)

where 𝑍𝑖,𝑡 = 𝜋𝑖,𝑡 − 𝜇𝑖,𝑡 , a difference in inflation using time-varying mean. To get time-

varying mean, we use Hodrick- Prescott (HP) filter for inflation data, get subtracted by

inflation data, and then the filtered data will be used as an input data. Afterwards, half-

life from a shock to inflation will be calculated using this formula.

−ln (2)

ln (𝜌𝑖) (5)

To evaluate the persistence of inflation, there are two problems that need to be

considered: the first is the number of lags and the second is time variation. Roache

Page 8: Inflation Persistence

2-7

(2014) demonstrates that the persistence change does not happen suddenly, but by

degrees as credibility in the new policy regime is accumulated over time. To capture the

time variation in persistence, we use regression estimated over 36 overlapping rolling

period samples of monthly data. This approach is better known as “rolling regression”.

The window size was chosen to balance two objectives, i.e., adequately assessing time-

variation and sufficiently estimating large sample periods. In particular, the sample of

the first series is January 2003 through December 2005. The second series then starts

and finishes one month later, from February 2003 through January 2005 and so on.

Regional Inflation Persistence Determinant

To identify inflation persistence determinant, we use panel data model to

explain determinant variation of inflation persistence among provinces for a research

time period as follow:

𝜌𝑖,𝑡 = 𝛼 + 𝛽′𝑋𝑖𝑡 + 𝑢𝑖𝑡 (6)

Where 𝜌𝑖,𝑡

is the inflation persistence level in province i for t time period, 𝑋𝑖𝑡 is an

observation vector from the explanation variable. The advantages of this data panel,

according to Baltagi (2004), are the ability to control unobserved heterogeneities among

individual units, can avoid omitted variables problems that are not changing over time,

and the low possibility to encounter autocorrelation and multicollinearity problems, like

in time-series data.

There are two types of the basic data panel, fixed effect model and random

effect model. For fixed effect model, we assume that individual heterogeneity is caught

by intercept. Each individual has an intercept, but the slope from each individual is

considered constant. For the fixed effect model, 𝜇𝑖 can be assumed by fixed Model fixed

effect to be written like:

𝑌𝑖,𝑡 = 𝛼 + 𝛽′𝑋𝑖,𝑡 + 𝑢𝑖,𝑡 (6)

𝑢𝑖,𝑡 = 𝜇𝑖 + 𝑣𝑖,𝑡

Page 9: Inflation Persistence

2-8

Then, fixed effect model can be written using dummy variable to sign an

individual as follow:

𝜌𝑖,𝑡 = ∑ 𝛼𝑗

𝑁

𝑗=1

𝑑𝑖,𝑡𝑗

+ 𝛽′𝑋𝑖,𝑡 + 𝑢𝑖,𝑡

where 𝑑𝑖𝑡𝑗

is a dummy, 𝑑𝑖,𝑡𝑗

= 1 if i=j, else is 0. The Parameter can be estimated using OLS.

Estimator of 𝛽′is called Least Square dummy variable (LSDV) estimator. LSDV estimator

includes a bunch of dummy variable for N-1 to sign each individual. Explanation variable

that has not changed over time will be dropped because of perfect collinear and

individual dummy variable.

Data

Data that is used for this research includes time period from 2002 until 2015

with monthly, quarterly, and yearly frequencies for 33 provinces in Indonesia as it is

written on the table below.

Table 1. Variable and Data

Number Variable Data Highlights Source of Data

1 Inflation

Monthly inflation level (year-on-year) that is measured by Consumer Price Index for 33 provinces in Indonesia. Beside the general inflation level, we use raw food category, processed food inflation category, and transportation inflation category.

BPS

2 Shock

Shock is measured using the growth rate of GRDP for 33 provinces that are estimated using AR, then there will be calculation of the standard deviation from the residual.

BPS

3 Competition Competition is measured by the price dispersion. The calculation using standard deviation from the price index of each

BPS

Page 10: Inflation Persistence

3-9

price category that is normalized into first period.

4 Rainfall Rainfall in 33 provinces BMKG

5 Infrastructure % of village with paved road BPS

6 Institution Institution is estimated by the child mortality level

BPS

7 Output Gap Output Gap is gotten from HP Filter approach to get actual Output from the farming GRDP

BPS

3. Regional Inflation Persistence Analysis

To analyze the degree of persistency in every region, we use rolling regression

technique in order to discover whether the inflation persistence is changing over time.

The variation degree of inflation persistence that exists among regions can give

important input for constructing a structural policy in every region. From the estimation

result, we get the persistency parameter for a full sample from 2005 until 2015 and a

sub sample that is divided into three periods in every province: 2003-2006, 2007-2010,

and 2011-2015.

Based on the result of a full sample period, the general inflation persistence

parameter is around 0.7 – 0.9 with half-life value around 3-8 months. Most of the

provinces have high inflation persistence that is above 0.8. The highest inflation

persistence is located at West Java province, following DIY and DKI Jakarta. The different

persistence level among regions shows different structural rigidity level that exists in

every province and reflects the respond of any change or shock that occurs.

The estimation result in a sub sample period shows the tendency of decreasing

level of inflation persistence in most provinces. Even though there is an increase level

of inflation persistent in most provinces (except Aceh) in 2007-2010 sub sample, there

were certain number of shocks that happened in all provinces, such as the rapid

increase in domestic oil price and international commodities. In the sub-sample 2011-

Page 11: Inflation Persistence

3-10

2015 period, many provinces experienced simultaneous decrease in the degree of

persistency. Some provinces, arguably, were no longer persistent, such as Gorontalo

and South Kalimantan whose degree of persistence between 0.4-0.5. The decline of

persistence level is also reflected by the decline of half life that captures time period

that is needed for a province to get its inflation back to the long-run equilibrium.

Table 2. General Inflation Persistence

Province Full Sample 2003-2006 2007-2010 2011-2015

Half Life

Half Life

Half Life

Half Life

Aceh 0.857 4.5 0.967 20.7 0.862 4.7 0.862 4.7

Babel 0.827 3.6 0.850 4.3 0.946 12.6 0.633 1.5

Bali 0.857 4.5 0.843 4.1 0.901 6.7 0.825 3.6

Banten 0.901 6.6 0.935 10.3 0.967 20.9 0.900 6.6

Bengkulu 0.874 5.1 0.944 12.0 0.943 11.9 0.875 5.2

Jateng 0.891 6.0 0.952 14.1 0.962 17.8 0.865 4.8

Kalteng 0.846 4.2 0.900 6.6 0.957 15.6 0.636 1.5

Sulteng 0.859 4.6 0.963 18.4 0.921 8.5 0.763 2.6

Jatim 0.891 6.0 0.944 12.0 0.964 18.7 0.841 4.0

Kaltim 0.901 6.6 0.954 14.7 0.950 13.4 0.863 4.7

NTT 0.856 4.5 0.900 6.6 0.898 6.4 0.767 2.6

Gorontalo 0.752 2.4 0.956 15.4 0.928 9.2 0.471 0.9

Jakarta 0.904 6.9 0.954 14.7 0.958 16.1 0.908 7.1

Jambi 0.829 3.7 0.870 5.0 0.940 11.1 0.787 2.9

Lampung 0.899 6.5 0.948 13.0 0.953 14.4 0.845 4.1

Maluku 0.820 3.5 0.919 8.2 0.878 5.3 0.717 2.1

Malut 0.827 3.6 0.904 6.9 0.931 9.7 0.813 3.3

Sulut 0.855 4.4 0.961 17.4 0.940 11.2 0.837 3.9

Sumut 0.880 5.4 0.935 10.3 0.954 14.7 0.888 5.8

Papua 0.791 3.0 0.823 3.6 0.908 7.2 0.820 3.5

Riau 0.873 5.1 0.894 6.2 0.963 18.3 0.880 5.4

Kepri 0.901 6.7 0.934 10.2 0.947 12.7 0.939 11.0

Kalsel 0.795 3.0 0.898 6.4 0.914 7.7 0.555 1.2

Sulsel 0.882 5.5 0.950 13.5 0.927 9.2 0.869 4.9

Sumsel 0.876 5.3 0.905 6.9 0.980 35.0 0.843 4.1

Sultra 0.818 3.5 0.921 8.4 0.934 10.2 0.756 2.5

Jabar 0.925 8.9 0.973 25.3 0.973 25.0 0.905 6.9

Kalbar 0.847 4.2 0.945 12.3 0.958 16.0 0.783 2.8

Page 12: Inflation Persistence

3-11

Province Full Sample 2003-2006 2007-2010 2011-2015

Half Life

Half Life

Half Life

Half Life

NTB 0.860 4.6 0.932 9.8 0.941 11.5 0.787 2.9

Sumbar 0.837 3.9 0.899 6.5 0.940 11.2 0.835 3.8

Yogyakarta 0.913 7.7 0.973 25.3 0.973 24.9 0.880 5.4

Besides measuring general inflation persistence level, we also measure CPI

inflation persistence level based on certain inflation categories: Raw Food, Processed

Food, and Transportation. These three groups are chosen because of the high inflation

level in the inflation basket.

Based on the estimation result from the full sample period, the inflation

persistence degree of raw food category in some provinces, particularly in Java and

Sumatera, have high level of inflation, which is above 0.8. On the other hand, in KTI

sector, the persistence level is considered low. This situation is also reflected by the half

life level, which means that the more persistent level of inflation in a province, the half-

life is going to be high.

For the sub sample period, the estimation result shows a similar pattern like the

general inflation, which has a decreasing pattern in the degree of inflation persistence.

There were a tendency of high level of persistence in 2003-2006 period and in 2007-

2010 period, and then reached low level of persistence in 2011-2015 period. Provinces

in Java island generally have higher degree of persistence compare to other regions.

Table 3. Inflation Persistence in the Raw Food Category

Province Full Sample 2003-2006 2007-2010 2011-2015

Half Life

Half Life

Half Life

Half Life

Aceh 0.806 3.2 0.962 17.7 0.742 2.3 0.802 3.1

Babel 0.762 2.6 0.634 1.5 0.930 9.5 0.623 1.5

Bali 0.819 3.5 0.775 2.7 0.827 3.7 0.834 3.8

Banten 0.870 5.0 0.882 5.5 0.956 15.4 0.800 3.1

Bengkulu 0.831 3.7 0.881 5.5 0.881 5.5 0.810 3.3

Jateng 0.860 4.6 0.929 9.4 0.953 14.5 0.820 3.5

Kalteng 0.849 4.2 0.913 7.6 0.914 7.7 0.655 1.6

Sulteng 0.756 2.5 0.792 3.0 0.866 4.8 0.578 1.3

Jatim 0.886 5.7 0.896 6.3 0.964 19.1 0.877 5.3

Page 13: Inflation Persistence

3-12

Province Full Sample 2003-2006 2007-2010 2011-2015

Half Life

Half Life

Half Life

Half Life

Kaltim 0.853 4.4 0.931 9.8 0.934 10.2 0.731 2.2

NTT 0.832 3.8 0.924 8.8 0.849 4.2 0.683 1.8

Gorontalo 0.721 2.1 0.894 6.2 0.818 3.5 0.542 1.1

Jakarta 0.877 5.3 0.948 13.1 0.952 14.1 0.863 4.7

Jambi 0.818 3.5 0.760 2.5 0.938 10.9 0.755 2.5

Lampung 0.869 4.9 0.823 3.6 0.941 11.4 0.847 4.2

Maluku 0.724 2.1 0.815 3.4 0.785 2.9 0.674 1.8

Malut 0.689 1.9 0.668 1.7 0.872 5.1 0.423 0.8

Sulut 0.793 3.0 0.876 5.2 0.881 5.5 0.716 2.1

Sumut 0.827 3.6 0.879 5.4 0.927 9.2 0.827 3.7

Papua 0.637 1.5 0.899 6.5 0.918 8.2 0.319 0.6

Riau 0.824 3.6 0.882 5.5 0.937 10.7 0.807 3.2

Kepri 0.799 3.1 0.888 5.9 0.908 7.2 0.855 4.4

Kalsel 0.816 3.4 0.897 6.4 0.851 4.3 0.618 1.4

Sulsel 0.837 3.9 0.901 6.6 0.880 5.4 0.859 4.6

Sumsel 0.851 4.3 0.857 4.5 0.962 17.8 0.768 2.6

Sultra 0.752 2.4 0.823 3.6 0.927 9.2 0.721 2.1

Jabar 0.882 5.5 0.947 12.7 0.947 12.8 0.871 5.0

Kalbar 0.793 3.0 0.813 3.4 0.917 8.0 0.604 1.4

NTB 0.848 4.2 0.900 6.6 0.914 7.7 0.819 3.5

Sumbar 0.809 3.3 0.840 4.0 0.898 6.4 0.769 2.6

Yogyakarta 0.877 5.3 0.896 6.3 0.966 20.0 0.911 7.4

For the processed food category, the estimated persistence in a full sample is varied

between 0.7 and 0.9. Similar to raw food category, provinces in Java Island are more

persistent than other regions. Banten and Kepri, which are known as industrial cities,

have high level of persistence. On the other hand, Bali and Gorontalo have low level of

persistence. In sub-sample period, the estimation result shows that in 2003-2006

period, almost all provinces tend to increase in the level of persistence, decrease in

2007-2010 period, and bounce back in 2011-2015 period. The persistence level in Bali

and Kepri for those three sub-sample periods continue to increase.

Tabel 4. Inflation Persistence in the Processed Food category

Page 14: Inflation Persistence

3-13

Province Full Sample 2003-2006 2007-2010 2011-2015

Half Life

Half Life

Half Life

Half Life

Aceh 0.841 4.0 0.898 6.5 0.944 12.1 0.715 2.1

Babel 0.835 3.8 0.894 6.2 0.932 9.8 0.717 2.1

Bali 0.791 3.0 0.748 2.4 0.842 4.0 0.778 2.8

Banten 0.931 9.7 0.860 4.6 0.928 9.3 0.960 16.8

Bengkulu 0.805 3.2 0.972 24.7 0.946 12.6 0.449 0.9

Jateng 0.912 7.5 0.981 35.6 0.892 6.1 0.929 9.4

Kalteng 0.837 3.9 0.829 3.7 0.896 6.3 0.846 4.1

Sulteng 0.886 5.7 0.866 4.8 0.934 10.1 0.894 6.2

Jatim 0.902 6.7 0.981 36.8 0.936 10.4 0.949 13.2

Kaltim 0.902 6.7 0.971 23.7 0.865 4.8 0.853 4.3

NTT 0.860 4.6 0.858 4.5 0.703 2.0 0.866 4.8

Gorontalo 0.730 2.2 0.751 2.4 0.897 6.4 0.518 1.1

Jakarta 0.916 7.9 0.975 27.5 0.833 3.8 0.962 17.7

Jambi 0.897 6.3 0.919 8.2 0.867 4.9 0.908 7.2

Lampung 0.833 3.8 0.944 12.1 0.785 2.9 0.856 4.4

Maluku 0.849 4.2 0.928 9.3 0.877 5.3 0.893 6.2

Malut 0.871 5.0 0.930 9.5 0.877 5.3 0.897 6.4

Sulut 0.887 5.8 0.910 7.3 0.888 5.8 0.918 8.1

Sumut 0.857 4.5 0.934 10.1 0.882 5.5 0.771 2.7

Papua 0.882 5.5 0.730 2.2 0.889 5.9 0.960 16.9

Riau 0.845 4.1 0.907 7.1 0.954 14.7 0.883 5.6

Kepri 0.919 8.2 0.850 4.3 0.950 13.4 0.972 24.6

Kalsel 0.833 3.8 0.958 16.1 0.775 2.7 0.861 4.6

Sulsel 0.869 4.9 0.954 14.6 0.927 9.2 0.848 4.2

Sumsel 0.891 6.0 0.941 11.4 0.960 17.0 0.955 15.2

Sultra 0.887 5.8 0.984 43.8 0.946 12.6 0.909 7.2

Jabar 0.889 5.9 0.870 5.0 0.923 8.7 0.900 6.6

Kalbar 0.879 5.4 0.915 7.8 0.886 5.7 0.942 11.7

NTB 0.899 6.5 0.940 11.1 0.892 6.1 0.931 9.7

Sumbar 0.842 4.0 0.967 20.9 0.898 6.4 0.895 6.2

Yogyakarta 0.910 7.3 0.993 96.6 0.931 9.7 0.933 10.0

The estimated result of the inflation persistence in transportation category in

full sample period shows that Java and other region in Sumatera are more persistent

than other region. But for the sub sample period, all provinces show high level of

Page 15: Inflation Persistence

3-14

persistence in 2003-2006 period, which was above 0.9. This level of persistence, in

general, tends to decrease in 2007-2010 sub sample period and 2011-2015.

Tabel 5. Inflation Persistence in Transportation Category

Province Full Sample 2003-2006 2007-2010 2011-2015

Half Life

Half Life

Half Life

Half Life

Aceh 0.840 4.0 0.916 7.9 0.750 2.4 0.874 5.1

Babel 0.817 3.4 0.924 8.8 0.840 4.0 0.729 2.2

Bali 0.871 5.0 0.917 8.0 0.883 5.6 0.898 6.4

Banten 0.875 5.2 0.916 7.9 0.898 6.4 0.909 7.2

Bengkulu 0.870 5.0 0.928 9.3 0.859 4.6 0.907 7.1

Jateng 0.863 4.7 0.926 9.0 0.867 4.9 0.881 5.5

Kalteng 0.873 5.1 0.915 7.8 0.902 6.8 0.877 5.3

Sulteng 0.837 3.9 0.940 11.2 0.880 5.4 0.892 6.1

Jatim 0.868 4.9 0.926 9.0 0.877 5.3 0.899 6.5

Kaltim 0.880 5.4 0.933 10.1 0.864 4.7 0.903 6.8

NTT 0.846 4.1 0.925 8.9 0.932 9.9 0.783 2.8

Gorontalo 0.863 4.7 0.930 9.6 0.873 5.1 0.848 4.2

Jakarta 0.876 5.2 0.933 10.0 0.861 4.6 0.900 6.6

Jambi 0.889 5.9 0.940 11.2 0.882 5.5 0.911 7.5

Lampung 0.882 5.5 0.931 9.7 0.889 5.9 0.899 6.5

Maluku 0.874 5.2 0.918 8.1 0.892 6.1 0.878 5.3

Malut 0.861 4.6 0.930 9.6 0.835 3.8 0.847 4.2

Sulut 0.861 4.6 0.928 9.3 0.834 3.8 0.915 7.8

Sumut 0.879 5.4 0.930 9.5 0.913 7.7 0.876 5.3

Papua 0.849 4.2 0.907 7.1 0.818 3.5 0.892 6.1

Riau 0.872 5.1 0.934 10.1 0.861 4.6 0.899 6.5

Kepri 0.866 4.8 0.922 8.5 0.868 4.9 0.894 6.2

Kalsel 0.856 4.4 0.921 8.4 0.879 5.4 0.858 4.5

Sulsel 0.865 4.8 0.925 9.0 0.853 4.4 0.890 5.9

Sumsel 0.867 4.9 0.918 8.1 0.824 3.6 0.907 7.1

Sultra 0.813 3.4 0.936 10.4 0.796 3.0 0.892 6.1

Jabar 0.892 6.1 0.911 7.5 0.939 11.1 0.911 7.4

Kalbar 0.775 2.7 0.925 9.0 0.855 4.4 0.635 1.5

NTB 0.890 5.9 0.926 9.0 0.903 6.8 0.911 7.4

Sumbar 0.849 4.2 0.931 9.7 0.845 4.1 0.889 5.9

Yogyakarta 0.882 5.5 0.962 17.9 0.841 4.0 0.868 4.9

Page 16: Inflation Persistence

3-15

Determinant Analysis of Regional Inflation

To explain the persistence variation that exists among region, we do the

estimation using panel data method in three time periods, which are 2003-2006, 2007-

2010, and 2015-2022. This estimation is done in 33 provinces using seconder data with

the model as follow:

𝝆𝒊𝒕 = 𝜶 + 𝜷𝟏𝑺𝒉𝒐𝒄𝒌 + 𝜷𝟐𝑷𝒓𝒊𝒄𝒆𝑫𝒊𝒔 + 𝜷𝟑𝑰𝒏𝒔 + 𝜷𝟒𝑮𝒂𝒑 + 𝜷𝟓𝑹𝒂𝒊𝒏 + 𝜷𝟔𝑻𝒓𝒂𝒏𝒔

+ 𝜷𝟕𝑫𝒖𝒎𝒎𝒚 + 𝜺

Where, 𝝆𝒊𝒕 is the persistence level of raw food, 𝑺𝒉𝒐𝒄𝒌 reflects the shock that exists in

each province’s economy, 𝑷𝒓𝒊𝒄𝒆𝑫𝒊𝒔 is the price dispersion that consists of competition

proxy, 𝑰𝒏𝒔 is an institution variable, 𝑮𝒂𝒑 is an output gap from farming sector, 𝑹𝒂𝒊𝒏 is

the level of rainfall, 𝑻𝒓𝒂𝒏𝒔 is a transportation variable that uses a data of village with

paved roads. This estimation uses OLS Panel, LSDV, and Panel Fixed Effect approaches.

Table 6. OLS Panel Estimation Result

Variable Model 1 Model 2

Shock 0.002

0.00007

(0.002) (0.004)

Disperse -0.759 *** -0.633 **

(0.237) (0.276)

Curjan -0.002

0.016

Transport -0.00001

(-0.0005)

Inst -0.002 **

(0.001)

Agri_gap -0.002

(0.002)

Konstanta 1.078 *** 0.935 ***

(0.073) (0.045)

Number of obs 97 97

R-squared 0.1961 0.0768

Root MSE 0.11035 0.11571

The estimation result with OLS Panel approach shows that variable shock is positively

correlated with inflation persistence level of a province. A shock that occurs in the

economy of a region will impact the inflation, and the average inflation velocity will go

Page 17: Inflation Persistence

3-16

back to its long-run equilibrium. The type and size of the expected shock will influence

inflation persistence level, such as, shock of the rise in domestic oil price in 2005

increases the level of inflation persistence in that period, including general inflation and

group inflation. Besides, supply shocks oftentimes increase persistence level in several

regions that experience deficit budget.

A price dispersion variable also proves to significantly influence the inflation

persistence level. Price dispersion is a proxy from the competitive market that has been

indicated to impact positively on the inflation persistence. But the estimation result

shows that a price dispersion variable is negatively correlated with the level of

persistence. Similar result is also found by Babetski, Coricelli, dan Horvath (2007)1 in

their research about inflation persistence determinant using price dispersion.

Transportation variable that uses a percentage sum of villages with paved road

compare to the total villages is found to negatively correlated with inflation persistence

level. This is in accordance to the first hypothesis, which believes that the good quality

of existing transportation will expedite good flow of goods and services. The

smoothness of goods distribution will stabilize price level and reduce supply shock that,

at the end, creates low inflation rate and impersistent result.

Another approaches that are used to estimate inflation persistent determinant

are LSDV and Panel Fixed Effect. These approaches are useful to discover another

variable that affects inflation rate by adding island dummy that represents idiosyncratic

factor from each region.

Table 7. LSDV Panel and Fixed Effect Estimation Result

Variable LSDV Fixed Effect Panel

shock 0.003 0.006

0.003 0.004

disperse -0.587 ** -0.822 ***

0.246 0.304

curjan 0.003

0.017

transport -0.001 -0.005 **

0.001 0.002

d_sumatra -0.045

0.028

Page 18: Inflation Persistence

3-17

Variable LSDV Fixed Effect Panel

d_kalimantan -0.094 *

0.052

d_sulawesi -0.114 ***

0.039

d_kti -0.158 ***

0.046

Konstanta 1.032 *** 1.232 ***

0.076 0.125

Number of obs 97 97

R-squared 0.2307 0.25

Root MSE 0.10917

The estimation result using LSDV and Panel Fixed Effect approaches confirms the

estimation result using OLS Panel. Shock variable, price dispersion, and transportation

influence the inflation persistence level. Rainfall variable is positively correlated with

inflation persistence among regions. High rainfall level will impact planting season and

harvest season of various farming commodities in a region. Also, the high rainfall level

could influence production result and commodity availabilities that are important to

fulfill society daily needs. The common level of defective harvest, as a result of high

rainfall, will impact the commodity supply, which also create the high level of

persistence.

Idiosyncratic factor that shows the characteristic of region can influence the

level of persistence in regions that are represented by the island dummy variable. This

shows that the economy characteristic of a region will influence the inflation

persistence variation that occurs among regions. This is confirmed by the calculation of

inflation persistence in the region with high level of processed food category, such as

Banten, a province with industrial region. On the other hand, the persistence level of

inflation in raw food category in Java Island, which is the central production of farming

commodities, is relatively higher than other regions.

Page 19: Inflation Persistence

4-18

4. Conclusion

In this paper, we analyzed Indonesia’s regional inflation persistence using

univariate models of inflation over the period 2003-2016. In this univariate framework,

we examined whether regional inflation goes back to its equilibrium value after any

shock. We also analyzed the determinant of inflation persistence.

The results on regional inflation persistence in Indonesia can be summarized as

follows: (a) First, using a simple univariate approach, this study shows a tendency of

decreasing level of inflation persistence in most provinces, (b) Second, provinces in Java

Island are more persistent than other regions. This is not only true for general inflation,

but also for raw food inflation and processed food inflation, (c) Third, factors which

determined high persistence are shock variable, price dispersion, and transportation

variable. This is not unexpected since prices of Indonesia’s commodity adjust sluggishly

to change in aggregate Indonesia, especially food. This rigidity is an important

characteristics of Indonesia’s price function. High rigidity and persistent are also related

with high prices. In a limited supply situation and a crawling upward of the commodity

price, the sellers who have an acceptable profit margin in mind, will keep the price in

high level in order to save their stocks, since they already bought their commodities

when the prices, at the moment, are in overbought condition. This finding is also in-line

with our recent survey that the wholesaler is the price maker in the food market. Thus,

this study indicates high market power of the middlemen.

Page 20: Inflation Persistence

5-19

5. References

Alogoskoufis, G. S., and Smith, R. P., 1991. “The Phillips curve, the persistence of inflation, and

the Lucas critique: evidence from exchange rate regimes,” American Economic Review,

81(6):1254-1275.

Altissimo, F., Ehrmann, M., and Smets, F., 2006. “Inflation Persistence and Price-setting

Behavior in the Euro Area: A Summary of the IPN Evidence,” ECB Occasional Paper No. 46.

Andrews, D., and Chen W., 1994. “Approximately Median-Unbiased Estimation of

Autoregressive Models,” Journal of Business and Economic Statistics, 12: 187 - 204.

Ascari, G., and Sbordone, A. M., 2014. “The Macroeconomics of Trend Inflation,” Journal of

Economic Literature,” 52(3), 679–739.

Baltagi, Badi H., Bresson, Georges, and Pirotte, A., (2004). “Tobin q: Forecast performance for

hierarchical Bayes, shrinkage, heterogeneous and homogeneous panel data estimators,”

Empirical Economics, Vol. 29, pp. 107-113.

Beck, G. W., and Weber, A. A., 2005a. “Price Stability, Inflation Convergence and Diversity in

EMU: Does One Size Fit All?” CFS Working Paper, No. 2005/30.

Beck, G. W., and Weber, A. A., 2005b. “Inflation Dispersion and Convergence in Monetary and

Economic Unions: Lessons for the ECB,” CFS Working Paper, No. 2005/31, forthcoming in the

International Journal of Central Banking.

Benati, L., 2008. “Investigating Inflation Persistence across Monetary Regimes,” Quarterly

Journal of Economics, 123 (3): 1005-1060.

Benigno, P., 2004. “Optimal Monetary Policy in a Currency Area,” Journal of International

Economics 63, pp. 293-320.

Cecchetti, S. G., Mark, N. C., and Sonora, R., 2002. “Price level convergence among United States

cities: Lessons for the European Central Bank,” International Economic Review, Vol. 43, pp.

1081–1099.

Cecchetti, S., and Debelle, G., 2006. “Has the inflation process changed?” Economic Policy 21,

311-316.

Cogley, T., and Sargent, T. J., 2001. “Evolving post-World War II U.S. inflation dynamics,” NBER

Macroeconomics Annual 16, 331–373.

Dossche, M., and Everaert, G., 2005. “Measuring Inflation Persistence: A Structural Series

Approach,” National Bank of Belgium, July 2005.

Page 21: Inflation Persistence

5-20

Filardo, A and Genberg, H., 2010. "Targeting inflation in Asia and the Pacific: lessons from the

recent past,” in Cobham, D., Eitrheim, Gerlach, S., and Qvigstad, J., F., (eds.), “Inflation targeting

twenty years on,” Cambridge University Presss.

Fuhrer, J., 2009, “Inflation Persistence,” Federal Reserve Bank of Boston Working Paper No. 09-

14.

Fuhrer, J. C., 2009. “Inflation Persistence,” Conference Series; [Proceedings], Federal Reserve

Bank of Boston, 38, 43-84.

Halunga A., Osborn D., and Sensier M., 2009. “Changes in the order of integration of US and UK

inflation,” Economic Letters, 102(1), 30-32.

Jain, M., 2013. “Perceived Inflation Persistence,” Bank of Canada Working Paper No. 43.

Levin, A. T., and Piger, J. M., 2004. “Is inflation persistence intrinsic in industrial economy,”

European Central Bank Working Paper No. 334.

Levin, A. T., Natalucci, F. M., and Piger, J. M., 2004. “The macroeconomic effects of inflation

targeting,” Federal Reserve Bank of St. Louis Review 86, 51–80.

Meller, B., and Nautz, D., 2012. “Inflation persistence in the Euro area before and after the

European Monetary Union,” Economic Modelling 29, 1170-1176.

Mishkin, F.S., 2007. “Inflation Dynamics, Speech, Annual Macro Conference,” Federal Reserve

Bank of San Francisco.

Phiri, A., 2016. “Inflation Persistence in African Countries: Does Inflation Targeting Matter?”

MPRA Paper No. 69155.

Pivetta, F., and Reis R., 2001. “The Persistence of Inflation in the United States,” Manuscript,

Economics Department, Harvard University, Harvard.

Roache, S.K., 2014. “Inflation Persistence in Brazil - A Cross Country Comparison,” IMF Working

Paper WP/14/55.

Vladova, Z., and Pachedjiev, S., 2008. “Empirical Analysis of Inflation Persistence and Price

Dynamics in Bulgaria,” Bulgarian National Bank.

Willis, J. L., 2003. “Implications of Structural Changes in the U.S. Economy for Pricing Behavior

and Inflation Dynamics”, Economic Review, First Quarter 2003, Federal Reserve Bank of Kansas

City.

Vaona, A., and Ascari, G., 2012. “Regional inflation persistence: evidence from Italy.” Reg Stud

46:509–523