Risk Premium and Central Bank Intervention. Pınar Özlü

Transcription

1 Cenral Bank Review ISSN prin / online 006 Cenral Bank of he Republic of Turkey hp:// Risk Premium and Cenral Bank Inervenion Pınar Özlü Research and Moneary Policy Deparmen Cenral Bank of he Republic of Turkey İsiklal Caddesi No: Ankara, Turkey Pinar.Ozbay@cmb.gov.r Phone: Absrac This sudy examines he relaion beween he risk premium and cenral bank inervenion. Forward raes are calculaed for he Turkish Lira-USD exchange marke and hen he effec of cenral bank inervenion on he risk premium is esimaed. Using high qualiy daily inervenion daa from he Cenral Bank of Turkey as well as implied forward raes, an MA (1)-GARCH (1,1) model is esimaed. Boh purchases and sales of US dollars by he Cenral Bank of Turkey appear o have no effec on he size of risk premium for TL/USD for he free floa period. Similar resuls are found for he managed floa period. Empirical suppor was weak for he heoreical model, wih inervenion having a significan effec on he risk premium. JEL classificaion codes: C, E31. Keywords: Cenral Bank Inervenion, Risk Premium. I hank he Cenral Bank of he Republic of Turkey and Isanbul Sock Exchange for providing daa on inervenions, ineres raes and especially Emrah Ekşi and Ayaç Ersan. The views expressed in his paper do no necessarily represen hose of he Cenral Bank of he Republic of Turkey. All remaining errors belong o me.

2 66 1. Inroducion Pınar Özlü / Cenral Bank Review 1 (006) The lieraure concerning he effecs of inervenion on currency markes and he moivaions for such inervenions is enormous. The resuls are mixed and depend on he exchange rae regime, he sample period and also he inervenion sraegy followed. Alhough he risk premium is no necessarily he inended arge of he inervenion, he empirical evidence indicaes ha some ypes of inervenion can affec he risk premium in forward markes. The forward exchange rae is a conracual exchange rae esablished a he ime of a ransacion ha will ake place a he mauriy ime +1 and usually regarded as he unbiased predicor of he fuure spo exchange rae. Conrary o popular heory, empirical evidence shows ha he forward rae is a biased predicor of he fuure spo rae and/or is evidence of a risk premium as indicaed by Hansen and Hodrick (1980), Hakkio (1981), Baillie e al. (1983), and Baillie (1989). One of he mos imporan unresolved paradoxes in inernaional finance is he forward premium anomaly, where he currency of he counry wih he higher rae of ineres is more likely o appreciae han depreciae. Numerous explanaions have been proposed o explain he forward premium anomaly, bu, he empirical evidence given in Evans and Lewis (1995), Kaminski (1993), Lewis (1988), Frankel and Froo (1987), Lewis (1989), Ellio and Io (1995) has no been saisfacory. In a sudy by Baillie and Oserberg (1997), Hodrick s model (Hodrick, 1989) is exended o allow cenral bank inervenion o have a direc effec on he risk premium. Baillie and Oserberg (1997) find ha purchases of US dollars by he Federal Reserve Bank appear o significanly increase he excess dollar denominaed reurns for boh he DM-USD and he Yen-USD markes. Consisen wih his sudy, Baillie and Oserberg (000) found ha he inervenion variables affec he risk premium in an analysis where he relaionship beween daily deviaions from uncovered ineres rae pariy and inervenion are invesigaed by using daily overnigh euro-currency deposi raes. Cenral banks use inervenion as a policy insrumen. Despie is frequen use, inervenion coninues o be debaed as a policy ool due o he conroversy over wheher i can achieve he policy goals of eiher changing he level of nominal exchange raes or reducing is volailiy. The sudies invesigaing he impac of inervenion direcly on he levels of exchange raes generally has found ha inervenion has no saisically significan effec. This paper aims o invesigae he effec of inervenion on risk premium and o assess wheher inervenion helps o

3 Pınar Özlü / Cenral Bank Review 1 (006) explain he forward premium anomaly, as found by Baillie and Oserberg (1997, 000). The analysis is done for Turkish economy, where he economy is small and has high inflaion. Secion describes he deails of he model, Secion 3 gives he daa, Secion 4 presens he esimaion oupu and Secion 5 discusses he resuls.. Deails of he model: Risk Premium and Inervenion The Covered Ineres Rae Pariy Condiion gives he relaionship beween spo raes, forward raes and ineres raes. ( f, l s + l ) = ( i, l i *, l ) (.1) s and f, l corresponds o logarihmic values of spo and forward exchange raes, respecively. Also i, l denoes he domesic currency reurn on an l-period risk free * bond, denominaed in erms of domesic currency where as i, l is he foreign currency reurn on a risk free bond denominaed in erms of he foreign currency. I implies ha he counry wih he higher rae of ineres has experienced an expeced depreciaion of currency. The relaionship beween forward raes and fuure spo raes may be simply expressed in erms of he forward rae as being an unbiased predicor of he fuure spo exchange rae and given by f l = E s (.), + l where s + l is he logarihm of he spo exchange rae and f, l is he logarihm of he forward rae for mauriy in ime +l. This is widely rejeced by he empirical sudies as in Hansen and Hodrick (1980), Hakkio (1981), Baillie e al. (1983), Baillie (1989). This has led o a ype of model f l = Es + l + ρ (.3), Where ρ is a ime dependen risk premium. The dependen variable in his sudy is he forward rae forecas error or in oher words risk premium, defined as (s +k - f ). Noe ha ( s+ k s ) ( i i * ) = s + k s ( f s ) = s + k f (.4) Hence, ( s + k f ) = ρ + u + k (.5) where u +k is he raional expecaions error associaed wih using he forward rae o predic he spo rae k periods and u is serially uncorrelaed for lags greaer han k, so E(u u +h ) = 0 for h > k. This resricion is consisen wih u, following a moving average process of order k-1.

4 68 Pınar Özlü / Cenral Bank Review 1 (006) Baillie and Oserberg (1997) exend Hodrick s 1989 model based on a consumpion based asse pricing model, where risk premium depends on he condiional variance of producion, money growh raes, consumpion s share of producion and inervenion variables. ρ * * * * = α1 σy α σy + α3σ Ω α4 σω + α5σ ς α6 σ α7τ α ς + 8τ (.6) Where σ y and σ Ω are he condiional variances of logarihms of producion and he money growh respecively. The variable σ ς denoes he condiional variance of he share of he currency used for inervenion. The inervenion * variable τ = M / M, is defined as he share of currency held by a foreign governmen for inervenion operaions. Aserisks denoe foreign counry equivalens. The difference beween his and Hodrick s model is he addiion of he condiional means and variances of he wo inervenion variables in he risk premium. The model does no impose any resricions on wheher or no serilizaion occurs. The model is esimaed from daily daa in order o deermine he relaively shor-lived effec of inervenion on risk premium. Hence i is no possible o include he variances of producion, money growh raes, and foreign currency holdings as a proporion of money sock. The spo exchange rae, he forward exchange and he inervenion variables, which are all observed daily, are he variables included in he esimaed model. Hence he risk premium ρ in (.6) is considered o be deermined by b s s f = ε + θ j b b US b US k j 1,1 j ε (.7) + = 0 1 ε ~ N ( 0, σ ) (.8) σ = ω + αε + βσ (.9) 1 1 The firs wo erms on he righ hand side of Eq (.7) corresponds o u and ε is a serially uncorrelaed whie noise processes, and θ j are he moving average parameers. The explanaory variables b s US, US include he inervenion variables. Condiional variance in equaion (.9) is represened by a linear GARCH (1,1) process. Bollerslev (1986) inroduced he GARCH (Generalized Auoregressive Condiional Heeroscedasiciy process, which exends he ARCH model o make σ a funcion of lagged values of σ as well as he lagged values of ε. Bollerslev (1986) required all he coefficiens o be posiive o ensure ha he

5 Pınar Özlü / Cenral Bank Review 1 (006) condiional variance is never negaive and he sum of coefficiens is less han 1 o avoid explosiveness of condiional variance. The quasi-maximum likelihood esimaion is used. 3. Daa In his sudy, he Cenral Bank of Turkey, he Isanbul Sock Exchange and Federal Reserve Bank Board of Governors provide he daa. The November 1993 hrough December 00 daa sample consiss of daily spo offer raes, inerbank overnigh ineres raes, Treasury bill raes, 30-day euro dollar raes and daily inervenion variables. Inervenion values are millions of US dollars. This sudy uses he daily amoun of ne dollar purchases (sales), daily spo offer raes and ineres raes. The analysis separaely covers boh he managed floa and free floa period in erms of exchange rae regime. The developmen of a fuures marke is very new o he Turkish Economy. The forward exchange rae is calculaed. The implied forward rae 1, is * 30 S (1 + i ( )) (3.1) F, 30 = 30 (1 + i ( )) where F, 30 is he daily 30-day forward rae, S is he daily spo rae as TL/USD see Figure 1 and Figure. i * is a proxy he 30-day reasury-bill ineres raes for Turkey. Daily ineres raes for reasury bills raded in he secondary marke are obained from he Isanbul Sock Exchange. The ineres rae of which he Treasury bill has he closes mauriy o 30 days is chosen for each day. i is a proxy for 30- day euro dollar raes day is assumed as he basis for ineres quoaions insead of 365, see Grabbe (1996). The Implied Forward Rae is calculaed as given in Grabbe (1996).

6 70 Pınar Özlü / Cenral Bank Review 1 (006) Nov-0 Figure 1. Forward Exchange Rae TL/USD (*) Nov-93 Mar-94 Jul-94 Nov-94 Mar-95 Jul-95 Nov-95 Mar-96 Jul-96 Nov-96 Mar-97 Jul-97 Nov-97 Mar-98 Jul-98 Nov-98 Mar-99 Jul-99 Nov-99 Mar-00 Jul-00 Nov-00 Mar-01 Jul-01 Nov-01 Mar-0 Jul-0 (*) Daily values for he log of forward rae of TL/USD including he full period of 1 November 1993 hrough 31 December 00.

7 Pınar Özlü / Cenral Bank Review 1 (006) Nov-0 Jul-0 Figure. Spo Exchange Rae TL/USD (*) Nov-93 Mar-94 Jul-94 Nov-94 Mar-95 Jul-95 Nov-95 Mar-96 Jul-96 Nov-96 Mar-97 Jul-97 Nov-97 Mar-98 Jul-98 Nov-98 Mar-99 Jul-99 Nov-99 Mar-00 Jul-00 Nov-00 Mar-01 Jul-01 Nov-01 Mar-0 (*) Daily values for he log of spo rae of TL/USD including he full period of 1 November 1993 hrough 31 December 00.

8 7 Pınar Özlü / Cenral Bank Review 1 (006) In his sudy, he forward rae quoaions are mached wih he fuure spo rae so ha boh represen conracs ha would be delivered on he same day. The deails of selemen procedures in he spo and forward markes are discussed in deail in a sudy by Riehl and Rodriquez (1977). The imporan aspec here is he number of working days in he conrac period varies. One reason is ha delivery delays ofen occur around he firs of he monh. Conracs also are no seled on weekends or on holidays in eiher of he wo counries for a given exchange rae. This exac maching reveals ha for he daa used in his sudy, k, he number of working days from he day of he forward quoe o he ime of selemen in he spo marke varies from 0 o 6. Since he mos common value of k in our sample is, u, he forecas error is esimaed as an MA (1) process. This analysis has been done for wo sub-periods due o he difference in economic policies. The firs sub-period covers beween Augus 1, 1994 and November 30, 1999 and he second one beween February, 001 and December 31, 00. The forecas error for wo subperiods are shown in Figure 3 and in Figure 4.

9 Pınar Özlü / Cenral Bank Review 1 (006) Oc-99 Aug-99 Figure 3. Forward Rae Forecas Error (*) Aug-94 Oc-94 Dec-94 Feb-95 Apr-95 Jun-95 Aug-95 Oc-95 Dec-95 Feb-96 Apr-96 Jun-96 Aug-96 Oc-96 Dec-96 Feb-97 Apr-97 Jun-97 Aug-97 Oc-97 Dec-97 Feb-98 Apr-98 Jun-98 Aug-98 Oc-98 Dec-98 Feb-99 Jun-99 Apr-99 (*) Daily Forward rae forecas error of TL/USD including he full period of 1 Augus, 1994 hrough 30 November, 1999.

10 74 Pınar Özlü / Cenral Bank Review 1 (006) Dec-0 Nov-0 Figure 4. Forward Rae Forecas Error (*) Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01 Aug-01 Sep-01 Oc-01 Nov-01 Dec-01 Jan-0 Feb-0 Mar-0 Apr-0 May-0 Jun-0 Jul-0 Aug-0 Sep-0 Oc-0 (*) Daily Forward rae forecas error of TL/USD including he full period of February 001 hrough 31 December 00.

11 Pınar Özlü / Cenral Bank Review 1 (006) Esimaes The deails of he esimaed model from he daily risk premium and inervenion daa are given in Table 1 and Table. The model possesses esimaed moving average coefficiens ha approximaely decline linearly wih he lag. Diagnosic esing of he model fails o provide evidence for a higher order moving average process. Also a linear GARCH (1,1) process is found o be an adequae represenaion of he condiional second momens for he managed floa period and a linear inegraed GARCH (1,1) is adequae for free floa period. Q 0 and Q 0 are he Q-saisics for he L-Jung-Box es of whie noise for he linear and squared sandardized residuals. The mos ineresing aspecs of he esimaed models in Tables 1 and concerns he coefficiens of he variables associaed wih inervenion. In paricular, unlike Baillie and Oserberg (1997), boh purchases and sales of US dollars by he Cenral Bank of Turkey appear o have no effec on he size of risk premium for TL/USD for he free floa period. Similar resuls are found for he managed floa period bu he buying of US dollars appears o have significan effec a a 0 percen significance level. This finding is expeced o be he resul of high inflaion in Turkish Economy. Effors of disinflaion were no successful hrough 1990s and sabiliy in he foreign exchange marke was uncommon. Under hese circumsances, facors affecing he ineres raes are relaed o sabiliy in boh he domesic marke and governmen deb managemen. The Cenral Bank of Turkey aimed a achieving sabiliy in he markes. Under hese circumsances, no relaion is expeced beween risk premium and inervenion.

12 76 Pınar Özlü / Cenral Bank Review 1 (006) Table 1: Esimaion of Inervenion/Risk Premium M odel: TL/$ (a) Condiional M ean Parameers Coefficien Sandard Error Coefficien Sandard Error b *** *** b 1 (U S b ) b (U S s ) θ *** *** θ 0.73 *** *** 0.06 θ *** *** θ *** *** 0.11 θ *** *** 0.11 θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** 0.04 Condiional Variance Parameers ω * 0.00 α ** 0.13 β 0.60 ** *** 0.10 Skewness Kurosis Q Q T (a) (*) Full period of 1 Augus, 1994 hrough 30 November, Denoes 10% significance level. (**) Denoes 5% significance level. (***) Denoes 1% significance level. b (s +k -f ) = s ε + θ jε j + b0 + b1us + bus j= 1,1 ε ~ N ( 0, σ ) σ = ω + αε 1 + βσ 1

13 Pınar Özlü / Cenral Bank Review 1 (006) (a) Table : Esimaion of Inervenion/Risk Premium M odel: TL/$ (a) Condiional M ean Parameers Coefficien Sandard Error Coefficien Sandard Error b ** ** b 1 (US b ) b (US s ) θ *** *** θ 0.80 *** *** θ *** *** θ *** *** 0.08 θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** 0.09 θ *** *** θ *** *** θ *** *** 0.08 θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** θ *** *** 0.07 Condiional Variance Parameers ω α β 0.13 ** ** Skewness Kurosis Q Q T Full period of February 001 hrough 31 December 00. (*) Denoes 10% significance level. (**) Denoes 5% significance level. (***) Denoes 1% significance level. b (s +k -f ) = s ε + θ jε j + b0 + b1us + bus j= 1,1 ε ~ N ( 0, σ ) σ = ω + αε 1 + βσ 1

14 78 Pınar Özlü / Cenral Bank Review 1 (006) Conclusions This paper is concerned wih he relaion beween he risk premium and cenral bank inervenion. Forward raes are calculaed for he Turkish Lira-USD exchange marke and hen he effec of cenral bank inervenion on he risk premium is presened. Using high qualiy daily inervenion daa from he Cenral Bank of Turkey as well as implied forward raes, an MA (1)-GARCH (1,1) model is esimaed. Boh purchases and sales of US dollars by he Cenral Bank of Turkey appear o have no effec on he size of risk premium for TL/USD for he free floa period. Similar resuls are found for he managed floa period. Empirical suppor was weak for he heoreical model, wih inervenion having a significan effec on he risk premium.

15 Pınar Özlü / Cenral Bank Review 1 (006) References Baillie, R.T., Lippens, R.E. and McMahon, P.C Tesing Raional Expecaions and Efficiency in he Foreign Exchange Marke, Economerica 51, Baillie, R.T Economeric Tess of Raionaliy and Marke Efficiency, Economeric Reviews 8, Baillie, R.T. and Oserberg, W.P Cenral Bank Inervenion and Risk in he Forward Marke, Journal of Inernaional Economics 43, Deviaions from Daily Uncovered Ineres Rae Pariy and he Role of Inervenion, Journal of Inernaional Financial Markes, Insiuions and Money 10, Baillie, R.T., Humpage, O.F. and Oserberg, W.P Inervenion from an Informaion Perspecive, Journal of Inernaional Financial Markes, Insiuions and Money 10, Bollerslev, T Generalized Auoregressive Condiional Heeroscedasiciy, Journal of Economerics, 31, Ellio, G. and Io, T Heerogeneous Expecaions and Tess of Raionaliy in he US Dollar/Yen Forward Foreign Exchange Rae Marke, Working Paper, Deparmen of Economics, UC-San Diego. Evans, M.D. and Lewis, K.K Do Long Term Swings in he Dollar Affec Esimaes of he Risk Premia?, Review of Financial Sudies 8, Frankel, J.A and Froo, K.A Using Survey Daa o Tes Sandard Proposiions Regarding Exchange Rae Expecaions, American Economic Review 77, Hakkio, C.S Expecaions and he Forward Exchange Rae, Inernaional Economic Review, Hansen, L.P. and Hodrick, R.J Forward Exchange Raes as Opimal Predicors of Fuure Spo Raes, Journal of Poliical Economy 88, Hodrick, R.T Risk, Uncerainy and Exchange Raes, Journal of Moneary Economics 3, Kaminski, G Is here a Peso problem? Evidence from he Dollar/Pound Exchange Rae, American Economic Review 83, Lewis, K.K Tesing he Porfolio Balance Model: A Muli-laeral Approach, Journal of Inernaional Economics 4(1-), The Persisence of he Peso Problem when Policy is Noisy, Journal of Inernaional Money and Finance 7, Riehl, H. and Rodriguez, R Foreign Exchange Markes, McGraw Hill: New York.