Inflation dynamics in the post-crisis period: Korea s experience

Transcription

1 Inflaion dynamics in he pos-crisis period: Korea s experience Min Chang, Changho Choi and Keunhyeong Park 3 Absrac This noe presens preliminary evidence of recen srucural changes o inflaion dynamics in Korea, and ses ou some possible causes and policy implicaions. Firs, he esimaed Phillips curve changed significanly during 0, which helps o beer explain he prolonged low inflaion in he pos-0 period. Second, various measures of rend inflaion are esimaed o have fallen o abou % in he pos-0 period from abou 3% before 0, which provides furher evidence of changes in he inflaion process during he recen pas. Third, we focus on he role of changes in economic srucures following he global financial crisis in driving changes in he inflaion dynamics in he pos-0 period, and discuss some implicaions for moneary policy. Keywords: Inflaion dynamics, Phillips curve, rend inflaion, changes in economic srucures, moneary policy JEL classificaion: E3, E5 Direcor General, Research Deparmen, he Bank of Korea. Senior Economis, Research Deparmen, he Bank of Korea. 3 Junior Economis, Paymen and Selemen Sysems Deparmen, he Bank of Korea. BIS Papers No 89

2 . Moivaion Inflaion dynamics in Korea appear o have undergone subsanial changes during he recen pas. While inflaion raes flucuaed during and immediaely afer he global financial crisis, as one would expec, heir subsequen pos-crisis evoluion is much less compaible wih sandard models of he inflaion process. In fac, inflaion raes in Korea have weakened more rapidly and persisenly since 0 han can be explained by variaions in economic slack and oher inflaion deerminans. The low inflaion ha has emerged from 0 onward is unprecedened in erms of is duraion and severiy: inflaion has been running below he arge range or he olerance inerval for 37 consecuive monhs from, November 0 o November 05, wih he average deviaion amouning o.4 percenage poin. 4 This marks he longes period over which inflaion has remained below arge since he adopion of inflaion argeing in Indeed, he recen decline in Korea s inflaion is wice he OECD average. While his low inflaion is in large par due o supply shocks, especially from he second half of 04, is iming and severiy raise he quesion of wheher srucural changes may have influenced he inflaion process in Korea in he pos-crisis period. Inflaion raes in Korea Graph Recen changes in inflaion Table B A (A) (B) Korea OECD average s Year-on-year. The shaded blue area represens eiher he arge range or he olerance inerval. Average CPI inflaion raes over he specified period. Our purpose in his noe is o provide a preliminary assessmen of poenial changes in inflaion dynamics during he recen pas in order o beer undersand he prolonged low inflaion since 0, and o presen some possible causes and policy implicaions of hese changes. To his end, Secion ses up he analysis of wheher inflaion dynamics have changed during he recen pas using Phillips curves. Secion 3 presens hree measures of rend inflaion, and ess for srucural breaks in heir level. Secion 4 discusses possible causes of recen changes in inflaion 4 The inflaion arge was he 3% midpoin of he olerance inerval of 4% during 00, while he arge range was.5 3.5% during The second longes period of he deviaion below he arge range or he inerval is 7 monhs, from July 005 o Sepember 007, wih he average deviaion being only 0.5 percenage poin. BIS Papers No 89

3 dynamics. Finally, Secion 5 summarises he main findings, and considers poenial policy challenges.. Analysis of changes in inflaion dynamics using Phillips curves This secion examines poenial changes in he Phillips curve relaionship so as o beer undersand he prolonged low inflaion from 0 onward. The firs subsecion esimaes a sandard Phillips curve o documen how he model fares wih acual inflaion raes. The second subsecion hen esimaes an augmened Phillips curve allowing for parameer change, o analyse he curve s abiliy o improve he empirical fi of he model, paricularly during he recen pas... Sandard Phillips curve Following he specificaions in Friedrich (04), we esimae a sandard Phillips curve for he headline inflaion raes: π : CPI inflaion, e π = α + βπ + γy + δimp + ε, e π : Household inflaion expecaions (one year) y : GDP gap, imp: Impor price inflaion (in KRW) where curren inflaion is explained by inflaion expecaions, he GDP gap, impor price inflaion, and oher residual facors. The use of household inflaion expecaions is in line wih Coibion and Gorodnichenko (05), who find ha household inflaion expecaions help o explain missing disinflaion in he Unied Saes during he Grea Recession. Due o he availabiliy of daa, he sample used for he esimaion is in quarerly frequency, and spans he period from Q 003 o Q4 04. Table shows he esimaed coefficiens of he sandard specificaion, and Graph illusraes he resuling in-sample fi. As shown in Table, he coefficiens of inflaion expecaions, he GDP gap, and impor price inflaion are saisically significan, and heir signs are consisen wih economic heory. However, as depiced in Graph, while he sandard Phillips curve relaionship generally does a good job in predicing inflaion raes in he period before 0, i does no do well in predicing inflaion raes from 0 onward. A closer examinaion reveals ha he in-sample predicion is consisenly higher han acual ouurns in he pos-0 period, suggesing ha some aspecs of inflaion dynamics are likely o have changed around he period 0. BIS Papers No 89 3

4 Esimaed coefficiens using he sandard Phillips curve Table In-sample fi using he sandard Phillips curve Graph Variable Coefficien Consan (α ) 0.39 Inflaion expecaions ( β ) 0.87 *** GDP gap (γ ) 0.3 ** Impor price inflaion (δ ) 0.05 *** *, ** and *** indicae saisical significance a he 0%, 5% and % levels, respecively. Year-on-year... Augmened Phillips curve In order o accoun for poenial changes in inflaion dynamics parsimoniously, we esimae an augmened Phillips curve which adds a pos-0 dummy and ineracs i wih he oher explanaory variables: e e π = α + β π + γ y + δ imp + α D + β π D + γ y D + δ imp D + ε, π : CPI inflaion, e π : Household inflaion expecaions (one year), y : GDP gap, imp: Impor price inflaion (in KRW), D : Pos-0 dummy variable ( from Q 0 onward) where he dummy akes on he value of from Q 0 onward, and 0 oherwise. Ineracing he pos-crisis dummy wih oher explanaory variables in he equaion allows he effecs of he consan erm, inflaion expecaions, he GDP gap and impor price inflaion o differ beween he pre-0 and pos-0 periods. The definiion of he remaining variables and he sample period are he same as hose wih he sandard Phillips curve. Table 3 indicaes ha inflaion is more sensiive o inflaion expecaions bu less sensiive o impor price inflaion during he pos-0 period. Furher, he consan erm in he inflaion process is much smaller, implying ha he Phillips curve may have shifed downward in he pos-0 period. However, unlike he coefficiens of he oher variables, here is no significan change in he sensiiviy o he oupu gap during he pos-0 period. 4 BIS Papers No 89

5 Esimaed coefficiens using he augmened Phillips curve Table 3 In-sample fi using he augmened Phillips curve Graph 3 Variables Coefficien Consan ( α ) 0.9 Inflaion expecaions ( β ) 0.76 *** GDP gap ( γ ) 0.06 Impor price inflaion ( δ ) 0.05 *** ( α ).5 ** Dummy variable ( β ) 0.53 * ( γ ) 0.68 ( δ ) 0. ** *, ** and *** indicae saisical significance a he 0%, 5% and % levels, respecively. Year-on-year. Graph 3 depics he resuling in-sample fi based on he augmened Phillips curve specificaion. I shows ha adding he pos-0 dummy and ineracing i wih sandard deerminans subsanially improves he explanaory power of he model, especially during he recen pas. Viewed as a whole, he esimaion resuls sugges ha inflaion dynamics in Korea are likely o have undergone subsanial changes around he period 0, and accouning for his change helps o beer explain he prolonged low inflaion during he pos-0 period. Furher, he downward shif in he Phillips curve may have played a prominen role in driving changes in he inflaion process. Before scruinizing he lower-frequency inflaion process furher, i is helpful o quanify he conribuion of srucural change facors in explaining he inflaion flucuaions during he pos-0 period. To his end, we conduc a hisorical decomposiion of he deerminans conained in he augmened Phillips curve, and define srucural change facors as he sum of he erms associaed wih he pos- 0 dummy ineraced wih a consan erm and oher explanaory variables. Graph 4 shows ha inflaion expecaions explain a large par of he inflaion process over he enire period, which is similar o Friedrich (04). Meanwhile, he GDP gap and impor price inflaion generally played a role in boosing inflaion raes in he pre-0 period, whereas hey have played a role in reducing inflaion raes since 0, albei by a smaller magniude relaive o inflaion expecaions. Furher, since 0, srucural change facors have played a role in lowering inflaion raes by percenage poin, a magniude ha is greaer han hose of he GDP gap and impor price inflaion calculaed based on he pre-0 coefficiens. BIS Papers No 89 5

6 Conribuions of individual inflaion deerminans Graph 4 s Resuls are based on he augmened Phillips curve. Srucural change facors consis of he pos-0 dummy ineraced wih a consan erm and oher explanaory variables. 3. Analysis of changes in inflaion dynamics using rend inflaion This secion invesigaes poenial changes in he evoluion of rend inflaion in order o furher our undersanding of he proraced low inflaion during he pos-0 period. Trend inflaion measures he underlying rend in inflaion afer removing ransiory price movemens due o shor-run disurbances. Thus, i is reasonable o aribue is persisen shif o srucural changes in he inflaion process. We sar by esimaing rend inflaion using hree economeric mehods: vecor auoregressions by Cogley and Sbordone (008) (henceforh CS), an unobserved componen model by Harvey and Koopman (009) (henceforh UC), and a dynamic facor model by Giannone and Maheson (006) (henceforh DFM). The use of muliple mehods is o check for he robusness of our resuls o alernaive mehods for esimaing rend inflaion. Specifically, he CS model consrucs four-variable vecor auoregressions consising of inflaion, he GDP gap, he shor-erm ineres rae and uni labour coss, and hen defines rend inflaion as he level o which inflaion is expeced o sele afer shor-run flucuaions die ou. Nex, he UC model decomposes inflaion ino rend and cyclical componens using a sae-space model, and hen akes he rend componen as rend inflaion. Finally, he DFM model exracs common facors from various disaggregaed price inflaion raes, and derives rend inflaion by removing any shor-erm flucuaions from common facors. Meanwhile, in line wih he availabiliy of daa, he sample used for he esimaion is a a quarerly frequency from Q 000 o Q 05 for he CS and UC models, and is a a monhly frequency from January 000 o March 05 for he DFM model. Graph 5 shows he evoluion of rend inflaion since 000 esimaed using he CS and UC models, while Graph 6 shows he evoluion of rend inflaion esimaed using he DFM model. In he period before 0, he hree measures of rend inflaion are esimaed o have been hovering around 3%, which is he midpoin of he inflaion 6 BIS Papers No 89

7 arge over his period. However, hey are esimaed o have undergone persisen decline over he years 0, descending o abou he % level on average during he recen pas. This evidence echoes he finding in he Phillips curve analysis ha srucural changes in he inflaion process seem o have occurred around he period 0. Compared across differen esimaion mehods, rend inflaion esimaed using he DFM model shows a relaively smaller decline recenly, while showing more flucuaions in he pre-0 period. Trend inflaion from CS and UC models, Graph 5 Trend inflaion from DFM Model, Graph 6 s CS indicaes a model by Cogley and Sbordone (008), while UC indicaes an unobserved componen model by Harvey and Koopman (009). The CS and UC rend inflaions are esimaed using quarerly daa for he period from Q 000 o Q 05. s DFM indicaes a dynamic facor model by Giannone and Maheson (006). The DFM rend inflaion is esimaed using monhly daa for he period from January 000 o March 05. As he nex sep, we es for srucural breaks in he level of rend inflaion as developed in Bai and Perron (003) o invesigae wheher he rend inflaion process underwen srucural changes since 000. Table 4 shows he resuling iming of significan srucural breaks occurring for each measure of rend inflaion. The esimaion resuls show ha he hree measures of rend inflaion experienced srucural breaks during he period 0. Specifically, he CS and UC rend inflaion show srucural breaks in Q3 0 and Q3 0, respecively, while he DFM rend inflaion shows srucural changes in June 0 as well as in April 005, July 007 and Ocober 009. Tes for srucural breaks in rend inflaion Table 4 Timing of srucural changes CS rend inflaion Q3 0 UC rend inflaion Q3 0 DFM rend inflaion April 005, July 007, Ocober 009, June 0 The iming of srucural breaks is deermined based upon values minimising he Schwarz es saisic modified by Liu, Wu and Zidek (997). BIS Papers No 89 7

8 4. Possible causes of changes in inflaion dynamics In he ligh of our findings ha he inflaion process may have changed during he recen pas, his secion considers possible causes of he decline in rend inflaion, and discusses some policy implicaions. A radiional macroeconomic view saes ha long-run inflaion is fundamenally a moneary phenomenon, and is deermined by moneary policy and is effecs on long-erm inflaion expecaions. 6 However, given ha economic sagnaion has persised despie excepional moneary policy simulus following he financial crisis, an alernaive view has gained racion, namely ha economic srucures oher han moneary facors may influence inflaion over a longer period of ime. For example, BIS (05) decomposes inflaion drivers ino shor-run, cyclical, and secular facors, and shows ha secular drivers, such as globalisaion, wage srucure, and echnology, have significan impacs on he low-frequency componen of inflaion. Faus and Leeper (05) argue ha mos of he unusual inflaion dynamics, known as disparae confounding dynamics, could be explained by aking proper accoun of srucural economic facors, such as deb srucure, demographics, ec. Our main focus in his secion is he role of economic srucures in explaining longrun inflaion in Korea. One fundamenal background fac is ha he Korean economy has been in he course of gradual ransiion from a fas- o a slow-growh model saring since he mid-000s, a ransiion ha acceleraed in he afermah of he global financial crisis. In erms of growh accouning, he increase in he labour force is consrained by rapid populaion ageing, and capial accumulaion has slowed due o weak domesic invesmen opporuniies. A slowdown in produciviy growh furher reinforces he deceleraion in growh poenial. In principle, hese facors could have boh inflaionary and disinflaionary consequences: hey reduce underlying inflaion by weakening domesic demand, bu hey also end o increase inflaion by limiing producion capaciy. However, he overall impac seems o be disinflaionary, especially because he growh deceleraion ineracs wih oher srucural vulnerabiliies, such as he labour marke dualiy and he inadequacy of he social insurance sysem. The oher imporan facor is he rapid increase in economic inegraion following he global financial crisis. Increased economic inegraion could pu downward pressure on aggregae inflaion by inensifying compeiion beween domesic and foreign markes (see Rogoff (003), and Guerrieri e al (00) for more deail). Furher, higher rade openness is likely o magnify and prolong he ransmission of exernal shocks ino domesic inflaion. 7 Thus, small open economies such as Korea s are more likely o experience prolonged low inflaion in imes of lacklusre global growh in he wake of he financial crisis. 6 See Friedman (963) among ohers. 7 See Milani and Park (04) for more deail. 8 BIS Papers No 89

9 Oupu growh raes Graph 7 Trade openness Graph 8 Sources: Bank of Korea, IMF. Trade openness is measured as he raio of he sum of expors and impors over GDP muliplied by 00. Source: IMF. Despie our focus on economic srucures, moneary policy may have played a par in prolonging low inflaion during he recen pas. For example, moneary auhoriies in small open economies may find i desirable no o respond o persisenly negaive inflaion shocks from abroad as aggressively as hey have o oher ypes of shock. 8 Obviously, furher sudy is necessary so as o beer undersand he precise role of moneary policy in explaining low inflaion. 5. Summary and policy challenges This noe makes a preliminary assessmen of recen changes in inflaion dynamics in order o beer undersand he prolonged low inflaion from 0 onwards, and o se ou is possible causes. The analysis yields hree main findings. Firs, he Phillips curve is esimaed o have changed significanly during he period 0, and aking ino accoun his change beer explains he low inflaion in he pos-0 period. Second, various measures of rend inflaion are esimaed o have fallen o abou % in he pos-0 period from abou 3% before 0, which provides furher evidence of recen changes in he inflaion process. Third, we focus on he role of changes in economic srucures in driving changes in he inflaion dynamics in he pos-0 period, highlighing ha boh srucural ransiion and high economic inegraion may ac as persisen disinflaionary forces, paricularly in imes of lacklusre global growh in he afermah of he global financial crisis. Given ha rend inflaion appears persisenly lower, wo furher issues arise wih regard o he conduc of moneary policy. One issue is he need o provide beer measures of poenial oupu and underlying inflaion in real ime. The oher issue is o analyse how he decline in rend inflaion affecs he relaionship beween real aciviy and inflaion, namely he slope of Phillips curve. Clarificaion of hese issues 8 See Kim e al (05) for more deail. BIS Papers No 89 9

10 would seem o be imporan for he beer calibraion of moneary policy in an environmen of changing rend inflaion. References Bai, J and P Perron (003): Compuaion and analysis of muliple srucural change models, Journal of Applied Economerics, vol 8, pp. Bank for Inernaional Selemens (05): 85h Annual Repor, Chaper 4, Anoher year of moneary policy accommodaion, pp 65 8, June. Cogley, T and A Sbordone (008): Trend inflaion, indexaion, and inflaion persisence in he New Keynesian Phillips curve, American Economic Review, vol 98, no 5, pp 0 6. Coibion, O and Y Gorodnichenko (05): Is he Phillips curve alive and well afer all? Inflaion expecaions and he missing disinflaion, American Economic Journal: Macroeconomics, vol 7, no, pp Faus, J and E Leeper (05): The myh of normal: he bumpy sory of inflaion and moneary policy, Prepared for he Federal Reserve Bank of Kansas Ciy conference, Jackson Hole, 7 8 Augus. Friedman, M (963): Inflaion: Causes and Consequences, Asian Publishing House. Friedrich, C. (04): Global inflaion dynamics in he pos-crisis period: wha explains he win puzzle?, Bank of Canada Working Paper, no Giannone, D and T Maheson (006): A new core inflaion indicaor for New Zealand, Reserve Bank of New Zealand, Discussion Papers. Guerrieri, L, C Gus and D Lopez-Salido (00): Inernaional compeiion and inflaion: A New Keynesian perspecive, American Economic Journal: Macroeconomics, vol, no 4, pp Harvey, C and S Koopman (009): Unobserved componens models in economics and finance: he role of he Kalman filer in ime series economerics, IEEE Conrol Sysems Magazine, vol 9, no 6, pp 7 8. Kim, J, H Bahng and K Park (05): Global rend inflaion and moneary policy, Prepared for he Bank of Korea s conference on Macroeconomic Policy and Price Measuremen Issues in a Low Inflaion Environmen, Seoul, Korea, 0 Augus. Liu, J, S Wu and J Zidek (997): On segmened mulivariae regression, Saisica Sinica 7, pp Milani, F and S Park (04): The effecs of globalizaion on macroeconomic dynamics in a rade-dependen economy: he case of Korea, Economic Modelling, no 48(C), pp Rogoff, K (003): Globalizaion and global disinflaion, Prepared for he Federal Reserve Bank of Kansas Ciy conference, Jackson Hole, 8 30 Augus. 30 BIS Papers No 89