Pakistan Denoised Inflation: A New Measure of Core Inflation

Transcription

1. SBP Working Paper Series No . 102 January, 2019 Denoised Inflation: A New Measure of Core Inflation Muhammad Nadim Hanif Javed Iqbal Syed Hamza Ali Muhammad Abdus Salam STATE BANK OF PAKISTAN
2. SBP Working Paper Series Editor : Sajawal Khan The objective of the SBP Working Paper Series is to stimulate and generate discussions on different aspects of macroeconomic issues among the staff members of the State Bank of Pakistan. Papers published in this series are subject to intense internal review process. The views expressed in these papers are those of the author(s) not State Bank of Pakistan. © State Bank of Pakistan. Price per Working Paper (print form) Pakistan: Rs 50 (inclusive of postage) Foreign: US$ 20 (inclusive of postage) Purchase orders, accompanied with cheques/drafts drawn in favor of State Bank of Pakistan, should be sent to: Chief Spokesperson, External Relations Department, State Bank of Pakistan, I.I. Chundrigar Road, P.O. Box No. 4456, Karachi 74000. Pakistan. Soft copy is downloadable for free from SBP website: http://www.sbp.org.pk For all other correspondence: Postal: Editor, SBP Working Paper Series, Research Department, State Bank of Pakistan, I.I. Chundrigar Road, P.O. Box No. 4456, Karachi 74000. Pakistan. Email: wps@sbp.org.pk ISSN 1997-3802 (Print) ISSN 1997-3810 (Online) Published by State Bank of Pakistan, Karachi, Pakistan. Printed at the SBP BSC (Bank) – Printing Press, Karachi, Pakistan
3. Denoised Inflation : A New Measure of Core Inflation Muhammad Nadim Hanif 1, Javed Iqbal 2, Syed Hamza Ali 3 & Muhammad Abdus Salam 4 Abstract Existing measures of core inflation ignore a part of ‘should be’ the core inflation. Exclusion based measures ‘exclude’ a part of persistent inflation inherently existing in the excluded part whereas filter based measures ‘filter-out’ the cyclical part also rather than the irregular component only. This study proposes a new idea to define and measure core inflation – noise free inflation or denoised inflation. As against considering only trend to define core inflation, this study proposes using cyclical component also to be part of core inflation. If core inflation is to be useful, for monetary policy making, as an indicator of underlying inflation, it has to include demand related component of inflation associated with current economic cycle. By using wavelet analysis approach to decompose seasonally adjusted price index into noise, cyclical component and trend, we estimate a denoised inflation series for Pakistan for the period July 1992 to June 2017. Since denoised inflation passes ‘statistical’ as well as ‘theoretical’ tests necessary for a series to be core inflation, we think it can be used as a new core inflation measure for Pakistan. This can also be estimated and tested for any country. JEL Classification: C19, E31, E52 Key Words: Estimation of Cyclical Component, Inflation, Monetary Policy Acknowledgments Authors would like to thank Mahmood ul Hassan Khan, Khurram Shahzad Mughal and an anonymous reviewer for valuable comments provided to improve this paper. Contact for correspondence: Muhammad Nadim Hanif Lead Economist Research Department State Bank of Pakistan I.I. Chundrigar Road Karachi 74000. Email: nadeem.hanif@sbp.org.pk 1 Lead Economist, Research Department, State Bank of Pakistan, Karachi (nadeem.hanif@sbp.org.pk) Joint Director, Research Department, State Bank of Pakistan, Karachi (javed.iqbal6@sbp.org.pk) 3 Economic Analyst, Research Department, State Bank of Pakistan, Karachi (hamza.ali@sbp.org.pk) 4 Joint Director, Statistics & Data Warehouse Department, State Bank of Pakistan, Karachi (muhammad.abdussalam@sbp.org.pk) 2 -1-
4. Non-technical Summary State Bank of Pakistan (SBP) is responsible to achieve given targets of inflation and real GDP growth along with ensuring financial stability in the country. SBP targets 12-month average of year on year (YoY) consumer price inflation in Pakistan. It suggests that SBP cares for a sort of underlying inflation. Central banks use the concept of core inflation to estimate underlying (trend) inflation. Currently, SBP uses non-food non-energy (NFNE) inflation, 20% trimmed inflation, and inflation in relatively stable component (RSC) of CPI as measures of core inflation in Pakistan. Existing literature defines core inflation as that part of inflation which can be associated with the long term result of monetary dynamics. It is separately defined from the demand related component of inflation – the one that is associated with the gap between actual and potential level of economic activity and is main concern of current monetary policy making – and the exogenous supply shock(s) related component of inflation. This core inflation itself is measured as an underlying inflation rate and not an indicator of underlying inflation trend. If it is to be useful for (short term) monetary policy making as an indicator of underlying inflation trend, it has to include the demand related component of inflation that is associated with the current economic heat/slack in the country; and to drop only exogenous supply shocks related part of inflation. Conventionally, core inflation is measured through some exclusion-based method or through some filter-based smoothing. Both the approaches ignore a part of ‘should be’ the core inflation in fact: exclusion based measures ‘exclude’ a part of the persistent inflation inherently existing in the excluded part; and filter based measures ‘filter-out’ the cyclical part also (rather than subtracting the irregular component only). This study proposes a new idea to define and measure core inflation – noise free inflation or denoised inflation. As against considering only trend to define core inflation, this study proposes using cyclical component also to be part of core inflation. By using wavelet analysis approach to decompose seasonally adjusted price index into noise, cyclical component and trend, we estimate a denoised inflation series for Pakistan for the period July 1992 to June 2017. All the aforementioned measures of core inflation currently being used in Pakistan fail, at least partially, and our denoised inflation passes ‘statistical’ as well as ‘theoretical’ tests necessary for a series to be core inflation. We suggest using denoised inflation as a new core inflation measure for Pakistan. -2-
5. 1 . Introduction During the last couple of decades, price stability has become the only objective of many central banks. For the central banks pursuing multiple objectives, it is has become the main objective. State Bank of Pakistan (SBP) is responsible to achieve government’s targets of real GDP growth and inflation along with ensuring financial stability. SBP targets 12-month average of year on year (YoY) consumer price inflation in Pakistan.1 Pakistan Bureau of Statistics (PBS) compiles the data of consumer price index (CPI) on monthly basis by collecting prices of 487 commodities from 76 Markets of 40 cities of Pakistan. CPI, like many other macroeconomic time series, is composed of underlying trend, cyclical component, seasonal fluctuation and irregular movements. SBP’s targeting of 12-month average clearly indicates that SBP cares for a sort of underlying/trend inflation. This indicator is less sensitive to transient changes in prices (compared to headline inflation). Reason is simple that monetary policy is forward looking and that one-off shock(s) in prices may have temporary effect on inflation that need not necessarily require any policy action as emphasized in Bank for International Settlements (1999)2. SBP uses Non-food Nonenergy (NFNE) inflation, 20% trimmed inflation, and inflation in relatively stable component (RSC) of CPI as measures of core inflation in Pakistan. Economics researchers, national statistical agencies and monetary policy makers use the concept of core inflation widely. The Bureau of Labour Statistics (USA) has been publishing CPI ex food and energy since 1978. The reporting of CPI ex-food has even longer history in developed countries. Regarding core inflation in Pakistan, the first published study was Tahir (2003). When it comes to conceptual definition of core inflation it is yet to find a universally agreed upon definition, however. There exists vast literature on how to define and measure inflation that is free of oneoff and/or temporary shocks. According to Eckstein (1981) core inflation …[is]… “the rate that would occur on the economy’s long term growth path, provided the path were free of shocks, and on the state of demand were neutral in the sense that markets were in the long run equilibrium”. In other words, core inflation is steady state concept and deviations from the steady state are either demand fluctuations or exogenous shocks (Parkin, 1984). Thus, Eckstein (1981) defines core inflation as that part of inflation which can be associated with the long term result of monetary dynamics. It is separately defined from the demand related component of inflation – the one that is associated with the gap between actual and potential level of economic activity and is main concern of current monetary policy making – and the exogenous supply shock(s) related component of inflation. Eckstein suggests econometric approach to estimate the core inflation. So, by Eckstein definition, core inflation itself is an underlying steady state inflation rate; and not an indicator of underlying inflation trend as explained by Clark (2001). If a measure of core inflation is to be useful for (short term) monetary policy making as an indicator of underlying inflation, it has to include the demand related component of inflation that is associated with the current economic heat/slack in the country. Short term monetary policy decision(s) will set the path of 1 Like South African Reserve Bank. In South Africa, the inflation target is specified as an average rate of increase in consumer prices. 2 “Central banks should only resist persistent sources of inflationary pressures and not be concerned with short term and reversible movements in prices and the inflation rate.” Foreword in Bank for International Settlements (1999). -3-
6. future steady state inflation rate and thus to have any measure of core inflation , we need to drop only that part of inflation which is the outcome of exogenous supply shocks and retain the demand related component. There is another strand of literature related to the concept of core inflation - through persistent component of inflation. This part of inflation is the one that has no long-term impact on real output (Quah and Vahey, 1995) and can be measured through different ways. Generally ‘transient’ part is removed either through some exclusion (of commodities with generally volatile prices) approach or through smoothing using some statistical filtering. We believe, both the approaches ignore a part of ‘should be’ the core inflation in fact: exclusion based measures ‘exclude’ a part of the persistent inflation inherently existing in the excluded part; and filter based measures ‘filter-out’ the cyclical part also rather than subtracting the irregular component only. This can be checked by simply testing if the usually excluded component food and energy inflation in the country - is persistent, i.e. not that temporary to exclude. In case food & energy inflation is persistent, which in fact is the case of Pakistan; our exclusion-based measure of core inflation will be ‘leaving out’ the persistent part of inflation in an attempt to estimate the core inflation that conceptually should be the ‘overall’ persistent component of inflation in the country. In such attempt, we may exclude an information permanently which contains not only noise but signal as well. Moreover, exogenous supply shock are not always related to any specific part of the CPI basket, like food and energy. Exclusion based measures of core inflation ‘retain’ the outcome of exogenous supply shocks in the retained part of the CPI basket, like in NFNE basket. If underlying trend in overall inflation is being set by food and energy inflation, as had been the case for some countries (Cecchetti and Moessner, 2008), exclusion-based measures of inflation excludes a part of the ‘should be’ core inflation. When food & energy price changes become persistent (rather than temporary) then they are very likely to impact inflation over the horizon relevant to monetary authorities (Cecchetti and Moessner, 2008). Food & energy inflation also affect the inflation expectations, as we have observed in the case of Pakistan3, and thus creates second round effects and becomes important for monetary authorities to pay attention. Thus, we think there is need for clearly defining what is meant by core inflation and how it should be measured. This is even more important when core inflation is measured using some statistical filtering. In statistical filtering, not only the irregular part is removed from a seasonally adjusted prices series but the cyclical component is also dropped and only the trend is considered as a one and used as core inflation. There is no question on seasonal adjustment. However, we think, with reference to use as an estimate of core inflation, only the irregular component merits to be excluded as a temporary part and not the cyclical component – cycles are not that temporary. From the literature on booms and busts in US economy, we know that in 33 business cycles in USA during 1854-2009 expansion and contraction periods averaged 38.7 and 17.5 months respectively4. If it the case for income, similar duration periods should hold for prices and there is some evidence in the literature in this context which has been documented by Cashin et al. (2002) - slumps (in global commodity prices) last longer than booms. Therefore, we define core 3 4 Results are in Table 11 of the Appendix III. http://www.nber.org/cycles/ -4-
7. inflation as a ‘denoised inflation measured from a seasonally adjusted consumer price index’5. In addition to redefining the theoretical concept of what is in the core inflation, we also contribute to the literature by implementing a rarely used approach to measure core inflation – wavelet analysis approach (WAN). But, we use this approach based upon our own idea to estimate the non-temporary part – cycle and trend (CAT) constitute the core inflation – in the case of Pakistan using monthly consumer price index from July 1992 to June 2017. As against other statistical approaches to decompose a time series, WAN does not require strong assumptions pertaining to noise or the trend in a time series and thus has no problem in analyzing a time series with non-stationarity, regime shifts and isolated shocks (Dowd et al. 2011). Moreover, unlike other conventional filters, including Fully Modified HP filter of Iqbal & Hanif (2017), WAN does not contaminate cyclical component with noise as it separately estimates the noise part. There could have been different names from WAN inflation to CAT inflation, but we preferred to call it ‘denoised’ inflation to emphasize the fact that only the ‘noise’ or ‘irregular’ or ‘transient’ or ‘one-off component’ is excluded. We test the estimated denoised inflation measured for Pakistan to usual statistical diagnostic procedure for it to qualify to be core inflation, as described in (Marques et al. 2003), and found it passing all such tests. Three exclusion based core inflation measures which are currently being reported in Pakistan – NFNE, Trimmed, and ‘Relatively Stable Components of CPI’ (RSC) measures – are found unsuccessful in this statistical evaluation process. Moreover, we also investigated if the denoised inflation is related to monetary policy stance in Pakistan6. We found it ‘linked’ compared to absence of any such link of any of the three exclusion based core inflation measures for the case of Pakistan. After the introduction, this study is organised as follows. Section 2 discusses data used in this study. In section 3, we explain the methodology for the estimation of denoised inflation. In section 4, estimated denoised inflation is presented. Section 5 contains the results of evaluation of denoised inflation and existing measures of core inflation by statistical as well as theoretical standards. Section 6 concludes the paper. 2. Data We used the data of CPI for Pakistan from July 1992 to June 2017 that is compiled by PBS to estimate core inflation. For the purpose of comparison of our estimated core inflation with the currently used measures of core inflation, we used NFNE, 20% trimmed mean and RSC core inflation series from SBP 7. The data for last two measures is available only from July 2008 to June 2017. In order to make We think our idea of defining what is ‘core inflation’ may have vast implications for macroeconomics literature. For example, there may be a need to rethink to proxy permanent income in the context of Permanent Income Hypothesis. ‘Cycle is the trend’ concept (Aguiar and Gopinath, 2007) lends significant support to our suggestion to exclude ‘seasonal and irregular’ components only and proxy the ‘permanent’ income by joint evolution of ‘cyclical and trend’ components. To test our argument one may explore the behaviour of consumption during various stages of business cycles and see consumption in not immune from cyclicality in income. 6 In comparison to the statistical evaluation of Marques et al. 2003, we call this exploration of core inflation measure and monetary policy stance nexus as ‘theoretical evaluation’ of core inflation measures. 7 The data for last two measures is available only for Jul 2008 to June 2017. 5 -5-
8. ‘theoretical’ evaluation of the new and existing measures of core inflation, we also used data of call money rate (CMR) and broad money supply (M2), which is collected from Statistical Bulletin of SBP. 3. Methodology While revisiting Eckstein (1981) idea to exclude both the demand as well as supply related fluctuations in price increases to come up with core inflation, we propose to exclude only the exogenous supply shock(s) related component of inflation to estimate core inflation (as an indicator of underlying inflation). This cannot be estimated using an exclusion based measure of inflation as it will either ‘exclude’ some part of ‘persistent inflation or ‘retain’ exogenous supply shock related inflation in the retained basket. To measure core inflation in conformity with our idea, we proposed to estimate ‘denoised’ inflation from a seasonally adjusted CPI. To extract noise we decomposed the seasonally adjusted CPI by the using wavelet analysis. Wavelets are mathematical expansions that transform data from the time domain into different layers of frequency levels. Conventionally, economists consider only two scales in a time series: the short run and the long run. There are actually more time scales in between the short run and the long run horizon of a time series (Dalkir, 2004). Through wavelet decomposition analysis, we can split a time series into different frequency zones, very high frequency (noise part) and very low frequency (smooth part) and moderate frequency zone (cyclical part). [Multiresolution] wavelet analysis is a useful tool to decompose an economic time series into trend, cycle, and noise (Yogo, 2008). A seasonally adjusted time series P