Contents lists available at SciVerse ScienceDirect
Economic Modelling
j ourna l homepage: www.e lsev ie r .com/ locate /ecmod
Estimating inflation compensation for Turkey using yield curves☆
Murat Duran ⁎, Eda GülşenCentral Bank of Turkey, Research and Monetary Policy Dep., İstiklal Cad. No: 10, 06100, Ulus, Ankara, Turkey
☆ This paper has benefited from Refet Gürkaynak's vations. The authors are also grateful to Soner Başkaya, HPınar Özlü, Deren Ünalmış, Tolga Tiryaki and the paheld at the Central Bank of Turkey. The views expressthe authors and do not necessarily reflect the opinions⁎ Corresponding author. Tel.: +90 312 5075422; fax:
1 We use the words “conventional bond” and “nothroughout the text.
2 Nominal bond markets are more developed than reaknowledge regarding inflation-indexedbonds is limiteddueintroduced recently in Turkey. Detailed information on thecurves can be found in Akıncı et al. (2007). Hence, in thisinflation-indexed bonds and estimation of real yield curves
Keywords:Inflation compensationReal yieldEvent study
Inflation compensation derived from nominal and real bond yields containsmarket based, real time informationregarding the inflation expectations and the pricing of inflation risks. In this study, we calculate inflation com-pensation for Turkey by using nominal and real yield curves. The findings of event study analysis on inflationcompensation indicate that changes in the term structure of inflation compensation contain information regard-ing the credibility of monetary authority. Moreover, we find that, at daily frequency, liquidity conditions haveno significant effect on inflation compensation and hence the effects of events such asmonetary policy decisionsand inflation surprises on inflation compensation can be attributed mainly to changes in inflation expectationsand pricing of inflation uncertainty.
Acquiring accurate information regarding inflation expectations isimportant for economic agents in order to make healthy consumptionand investment decisions. Surveys are the main sources of informationregarding inflation expectations. However, using survey data has somedisadvantages since the number of participants of a survey is limitedand there is no penalty for incorrect responses. On the other hand, de-riving inflation compensation directly from financial market data over-comes such problems and can be considered as an alternative indicatorof inflation expectations because it is calculated by using the yields ofconventional1 and inflation-indexed bonds traded in markets and con-tains market-based information regarding future inflation.
In this study, we aim to derive an indicator of inflation expecta-tions of market participants by using yield curves fitted to conven-tional and inflation-indexed bonds issued by the Turkish Treasuryand traded on the Istanbul Stock Exchange (ISE).2 In this context,
luable comments and sugges-arun Alp, Doruk Küçüksaraç,
rticipants of internal seminared in this paper are those ofof the Central Bank of Turkey.+90 312 5075732.an), [email protected]
minal bond” interchangeably
l bond markets in Turkey andto the fact that these bondswereestimation of the nominal yieldstudy, we emphasize more onthan nominal ones.
rights reserved.
firstly, inflation-indexed bond markets in Turkey and in the worldare introduced and real yield curves are estimated using Turkishdata. Then, the concept of inflation compensation is defined and it iscalculated for different maturities by using yield curves. Next, dataon inflation expectations obtained from the Central Bank of Turkey's(CBT) Survey of Expectations are compared with inflation compen-sation calculated using yield curves. Finally, liquidity effects on infla-tion compensation are examined and event studies to investigatethe effects of inflation surprises and monetary policy on inflationcompensation are carried out.
2. Inflation-indexed bond market in Turkey
The Turkish Treasury has started issuing exchange-traded inflation-indexed bonds in February 2007. Inflation-indexed bonds are issued for5 or 10 year maturities with semiannual coupon payments. Principaland coupon payments of these bonds are protected against inflation.The calculation of inflation premium in principal and coupon pay-ments is based on Consumer Price Index (CPI) data announced by theTurkStat.3
Real yields on nominal (conventional) bonds decline in case ofunanticipated increases in inflation whereas inflation-indexed bondsguarantee the real yields and protect the investors from inflation. Inother words, real yield on an inflation-indexed bond does not declinewith inflation. Therefore, investors prefer these securities since they pro-vide predictable returns which are unresponsive to changes in inflation.
Issuance of inflation-indexed bonds enables the CBT to extract in-formation regarding the term structure of real yields and hence the
3 For detailed information regarding inflation-indexed bonds market in Turkey,please see Treasury of Turkey (2009).
* Bonds with maturity less than 1.5 years are classified as short term and those with maturity 1.5 to 5 years are classified asmedium term. Source: ISE.
0%
2%
4%
6%
8%
10%
12%
14%
16%
18%
0
5
10
15
20
25
30
35
40
45
02-0
705
-07
08-0
711
-07
02-0
805
-08
08-0
811
-08
02-0
905
-09
08-0
911
-09
02-1
005
-10
08-1
011
-10
02-1
105
-11
Monthly Trading Volumes
(Billions of TL)Inflation-indexed
Nominal
Indexed over Nominal (right scale)
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
10-0
9
12-0
9
02-1
0
04-1
0
06-1
0
08-1
0
10-1
0
12-1
0
02-1
1
04-1
1
06-1
1
Monthly Average Turnover Rates
Short term
Medium term
Overall
Chart 1. Liquidity of inflation-indexed bonds (monthly average turnover rate of a bond is calculated as the ratio of total monthly trading volume of the bond to its outstandingamount).
593M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
term structure of inflation expectations. In addition to this, issuanceof inflation-indexed bonds contributes to the diversification of trea-sury debt securities, broadening the investor base and extending thematurity of treasury debt (Treasury of Turkey, 2009). In fact, nominalamount of inflation-indexed bonds issued by the Turkish Treasury hasreached to almost 63 billion TL as of August 2011.
Trading volume of inflation-indexed bonds on the ISEwas limited inthe first three years but started to rise afterwards. Recently it is around5% of the trading volume of the overall bondmarket on the ISE (Chart 1,left panel).
We think that one of the reasons behind the rise in the tradingvolume of inflation-indexed bonds in 2010 may be the increase inthe supply of inflation-indexed bonds in that period. Newly issued 5and 10 year inflation-indexed bonds along with existing ones filledthe term space of inflation-indexed bonds to some extent. Thisresulted in more options on investment horizons and has contributedto the increase in trading volume of inflation-indexed bonds.
The rising trend in the trading volume of inflation-indexed bondsimplies that the liquidity of these bonds is also improving. Monthlytrading volumes of inflation-indexed bonds were less than 0.1% ofthe nominal value of their outstanding amount before 2010. Withthe help of new issuances and longer maturities, monthly turnoverrate of inflation-indexed bonds has increased and reached to 7.5% in2010 (Chart 1, right panel).
Although the trading volume of inflation indexed bonds is increas-ing, their liquidity is still far from being level with that of the nominalbonds. Therefore, there may be an additional liquidity premium in theyields of inflation-indexed bonds.
4 For more detailed information regarding the yield curve fitting methods used byvarious central banks, please see BIS (2005).
5 For more detailed information regarding the data that are suitable for yield curveestimation in Turkey, please see Akıncı et al. (2007).
6 Having four different inflation-indexed bonds does not guarantee the estimation ofreal yield curve. In any day at least four different inflation-indexed bonds must be trad-ed in the bond market so that there exist four different prices for four different matu-rities. Only then the NS method can be used. In some dates after October 2009, therewere adequate numbers of different inflation-indexed bonds outstanding but inade-quate number of prices because some of the inflation-indexed bonds were not tradedin the markets that day. So real yield curves could not be estimated for those dates.
3. Inflation-indexed bond markets in the world
Inflation-indexed bonds are more preferable for emerging countriesthan for advanced economies since their main features (extending thematurity of treasury debt, diversification of debt instruments, eliminatinginflation risk for investors) aremore valuable for treasury departments ofemerging countries. Therefore, these markets are more widespread inemerging countries (Chart 2). In particular, inflation-indexed bondsconstitute very large portions of the overall bond markets in Israel andin some Latin American countries. On the contrary, these instrumentsare less widespread in advanced economies and Eastern Europeancountries. When it comes to Turkey, amount of inflation-indexed bondsis increasing relative to other emerging countries.
4. Inflation-indexed bonds and real yield curve
Issuance of inflation-indexed bonds enables the measurement ofmarket expectations regarding the real interest rates. In this study,we estimate the real yield curve using inflation-indexed bond data.Estimation is based on the prices of inflation-indexed bonds with dif-ferent maturities, that are traded in the secondary market.
The most conventional methods of yield curve fitting suggested inthe related literature are Nelson–Siegel (NS), Svensson (ExtendedNelson–Siegel) and non-parametric spline function methods.4 Splinemethods fit the observations very well but are very vulnerable to out-liers. Parametric methods such as NS and Svensson enable estimationof yield curves which fit a prespecified functional form instead of con-centrating on fitting every single point. This makes the yield curvessmoother and rules out specific factors affecting the individual securi-ties (Gürkaynak et al., 2010b). In this study, we prefer parametricmethods which are more suitable to macroeconomic policy analysisand estimate the real yield curve by the NS method.
Our yield curve estimations are based on minimizing the differencebetween observed prices and yield curve implied prices similar to theapproach adopted by Akıncı et al. (2007) for the nominal yield curve.Estimated real yield curve plots real yields on zero-coupon bonds fordifferent maturities. Prices of inflation-indexed bonds whose valuedates are the sameday are obtained from the ISE daily bondmarket bul-letins. Real yield curve estimations are performed by using these data.5
The NS method involves the estimation of 4 parameters and this re-quires observing the prices of at least 4 different bonds. In Turkey, thefourth inflation-indexed bond was issued in October 2009, thereforereal yield curves can be estimated after that date.6 Time plots of 1, 2, 3and 4-year real yields estimated by the NS method are illustrated inChart 3.
Source: Bank for International Settlements (BIS). * BIS data includes inflation-indexed bonds issued before 2007by the Turkish Treasury that are not market-traded.
0
20
40
60
80
100
Chi
le
Isra
el
Arg
entin
a
Col
ombi
a
Bra
zil
Mex
ico
Tur
key
S. A
fric
a
Per
u
Pol
and
Rus
sia
Indo
nesi
a
Cze
ch R
ep.
Em
ergi
ng C
.
Adv
ance
d C
.
2000 2010
0
5
10
15
20
25
30
0
5
10
15
20
25
30
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Latin AmericaEastern EuropeTurkey*Advanced Countries
Chart 2. Ratio of inflation-indexed bonds to all domestic debt securities.
0
1
2
3
4
5
61 year2 years3 years4 years
594 M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
The short end of the yield curves cannot be estimated accurately be-cause of some technical shortcomings of the NS method and due tonon-existence of data on 1 year and shorter term inflation-indexedbonds. In addition, since the sensitivity of shorter term yields to index-ation lag is higher, time plot of 1-year real yield seems more volatilethan longer term real yields.7 When the number of inflation-indexedbonds supplied will increase and fill the maturity-space, inaccuracy inthe short terms will be fixed partially. But this will not help to repairthe indexation lag problem. Thus, at present, 3 or 4-year real yieldsare more reliable than 1 or 2-year real yields.
-1
10-0
9
12-0
9
02-1
0
04-1
0
06-1
0
08-1
0
10-1
0
12-1
0
02-1
1
04-1
1
06-1
1
Chart 3. Real yields estimated by the NS method.
5. Inflation compensation
Inflation compensation is the inflation rate which, if realized, wouldleave an investor indifferent between holding an inflation-indexedbond and a conventional bond (Gürkaynak et al., 2010b). According tothis definition, the difference between the yield of a conventionalbond and the yield of an inflation-indexed bond which have the samematurity and other features, gives the inflation compensation.
It is possible to decompose the components of the yield on aconventional bond as follows:
rn ¼ r þ πe þ θr þ θπ þ θc þ ‘n þ εn: ð1Þ
Here, rn stands for yield on conventional bond, and the rest are r:expected real interest rate, πe: expected inflation rate, θr: pricing ofreal interest rate variation risk, θπ: pricing of inflation risk, θc: pricingof credit (counterparty) risk, ‘n: liquidity premium of conventionalbond and εn: idiosyncratic factors regarding the conventional bond.
7 Principal and coupon payments of inflation-indexed bonds are calculated based onreference indices. Therefore, the reference index affects the market price of inflation-indexed bonds. For the 1st day of any month, the reference index is the CPI of 3 monthsago. But for the rest of the days in any month, the reference index is calculated bylinear interpolation of CPIs of 3 months ago and 2 months ago (Treasury of Turkey,2009). Hence by the time a new CPI is announced, the reference index excludes theCPI of 3 months ago (actually 4 months minus 1 day), CPI of 2 months ago becomesthe new CPI of 3 months ago and the CPI of 1 month ago is included as the new CPIof 2 months ago. This creates a jump if CPI excluded from the calculation is significant-ly different than the CPI included. Since the average indexation lag is 2.5 months, a1-year real yield is affected by 2.5/48 = 21% of the jump whereas a 4-year real yieldis affected by only 2.5/48 = 5% of the jump.
Similarly the yield on an inflation-indexed bond can be decomposedas follows:
rr ¼ r þ θr þ θc þ l r þ εr: ð2Þ
Here, rr stands for the yield on an inflation-indexed bond, and therest are ‘r: liquidity premium of inflation-indexed bond and εr: idio-syncratic factors regarding the inflation-indexed bond. Since aninflation-indexed bond protects its investor against the actualized in-flation; inflation expectations and inflation risks are not reflected inits price or yield.8
Consequently, inflation compensation as the difference betweennominal and real yields9 can be expressed as follows;
IC ¼ rn−rr ¼ πe þ θπ þ ‘n−‘rð Þ þ εn−εrð Þ: ð3Þ
Assuming the expected value of εn−εrð Þ equals to zero, Eq. (3)shows that inflation compensation is composed of inflation expecta-tions, inflation uncertainty and the difference between the liquiditypremiums of conventional and inflation-indexed bonds. Since
8 In this study, indexation lag which can affect shorter term real yields is ignored.9 Yields used here are continuously compounded and therefore the inflation com-
pensation is expressed also as continuously compounded. According to the FisherEquation, the expected inflation rate is expressed as πe ¼ 1þrn
1þrr−1. The relationship be-
tween a continuously compounded yield and its discretely compounded equivalentcan be expressed as ercontinuous ¼ 1þ rdiscrete . Hence, continuously compounded expectedinflation rate becomes; eπ
e ¼ ernerr ¼ ern−rr→πe ¼ rn−rr. So, expected inflation rate is cal-
culated by the subtraction operation when yields are expressed as continuouslycompounded.
Real yield curve
Nominal yield curve
Inflation compensa
tion
0
1
2
3
4
5
6
7
8
9
10
1 1.5 2 2.5 3 3.5 4
Term (years)
Inflation Compensation Curve
(Feb.25, 2011)
2
3
4
5
6
7
8
9
Inflation Compensation Time Series
1 year 2 years
3 years 4 years
10-0
9
12-0
9
02-1
0
04-1
0
06-1
0
08-1
0
10-1
0
12-1
0
02-1
1
04-1
1
06-1
1
Chart 4. Inflation compensation.
595M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
inflation uncertainty is a positive component in the yield of a conven-tional bond, it pushes inflation compensation upwards, above expectedinflation level. On the other hand, liquidity of inflation-indexed bonds isusually poor compared to the liquidity of conventional bonds indicatingthat the liquidity premium of inflation-indexed bonds is usually higherthan that of the conventional bonds. Therefore, the difference betweenliquidity premiums pulls inflation compensation downwards, belowthe expected inflation level. In the end, these two factors push the infla-tion compensation in opposite directions.
5.1. Calculation of the inflation compensation
In order to calculate the inflation compensation, the yield of aconventional bond and the yield of an inflation-indexed bond withthe same maturity are required. However, calculating the inflationcompensation simply by comparing the yields of one conventionalbond and one inflation-indexed bond has some deficiencies. This isbecause, instead of reflecting the general view and expectations re-garding the bond market, individual bonds may be affected by specif-ic factors and thus traded overpriced or underpriced. This may lead toa miscalculation of inflation compensation. Furthermore, the usage ofone conventional bond and one inflation-indexed bond which havesame maturity but different cash flow structures (different couponrate or payment period) prevents the proper calculation of inflationcompensation. For instance, at day t, the difference between theyields of a zero-coupon conventional bond whose time to maturityis n and an inflation-indexed bond whose time to maturity is also nbut which has some coupon payments until the maturity is notequal to the n-day inflation compensation. This is because the dura-tion of the inflation-indexed bond is shorter than n days since thereare payments before maturity. Another disadvantage of using actualyields is that it is not possible to obtain a constant maturity time se-ries of inflation compensation since the maturity of actual bondsused becomes 1 day shorter every day. Besides, it is not possible to
observe actual yields every day in every different maturity and ob-serving one conventional and one inflation-indexed bond with matu-rities that match each other is completely a matter of coincidence. Inorder to overcome this problem, calculating the inflation compensa-tion by using actual bonds that have maturities close to each otherseems practical but this also leads to a miscalculation and weakensthe analytical validity and usefulness of inflation compensation.10
Yield curves are preferred to actual bond yields in calculation of in-flation compensation, because of these disadvantages of actual yields.Yield curves provide comparable nominal and real yields which arefree of bond-specific factors, for each maturity. This enables propercalculation of inflation compensation every day in every maturity.
In this study, inflation compensation is calculated on a daily basis asthe difference of yields implied by the zero-coupon nominal yieldcurves estimated using the Svensson method by the CBT and yields im-plied by the zero-coupon real yield curves estimatedusing theNSmeth-od. Nominal yield curve, real yield curve and term structure of inflationcompensation for February 25, 2011 are illustrated in Chart 4, left panelas an example. Note that, in this chart, inflation compensation for anymaturity shows the annualized level of expected inflation and inflationuncertainty from present to that maturity.
Time series plots of 1, 2, 3 and 4-year inflation compensationscalculated using yield curves, are plotted in Chart 4, right panel.Longer term inflation compensations seem to be less volatile becauseof the technical reasons stated before for real yield curves and alsobecause shorter term inflation compensation is more responsive toevents such as monetary policy and inflation announcements.
Inflation compensation is also calculated by central banks of somecountries which have active inflation-indexed bond markets. Further-more, central banks of the United States, England and Israel publishtheir inflation compensation calculations periodically on their websites.
In England, the Bank of England reports 5-year forward 5-year in-flation compensation11 as an indicator of inflation expectations in itsinflation report. The n-year forward m-year inflation compensation isa measure that is completely free of the inflation compensation offirst n-year period and hence completely reflects the market's viewsregarding the remaining m-year period. Long range forward inflationcompensations such as 5-year forward 5-year inflation compensationconsists of the inflation expectations and uncertainty of future,assuming the effect of liquidity conditions is negligible. Increases(decreases) in this indicator mean that the inflationary pressure is in-creasing (decreasing) regardless of the source of the change (expecta-tions or uncertainty). Also, this measure is completely attributable tomonetary policy stance since it eliminates the initial (current) condi-tions of the economy and rules out short-runfluctuations and economic
10 It is not always possible to have conventional and inflation-indexed bonds with thesame maturity because of the issue calendar of the Treasury.11 See Appendix B for details and calculation of forward inflation compensation.
Table 1Inflation compensation and market liquidity.
Sample period: 01.01.2010–24.05.2011.Estimated using Newey–West heteroskedasticity and autocorrelation consistent covariance matrix, standard errors are given in parenthesis.(*): significant at 10% level, (**): significant at 5% level, (***): significant at 1% level.
*2-year average inflation expectations are calculated by compounding annual inflation expectations of 12-months future and 24months future. Difference between the 1st and 3rd quartiles of 2-year average inflation expectations of respondents is used as ameasure of inflation uncertainty (Mankiw et al., 2003).
Source: CBT, Authors’ calculations.
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
8
2-Year Inflation Compensation
2-Year Average Inflation Expectation (Survey)
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
8
2-Year Inflation Compensation
Dispersion in 2-Year Inflation Expectations(right axis)
10-0
9
12-0
9
02-1
0
04-1
0
06-1
0
08-1
0
10-1
0
12-1
0
02-1
1
04-1
1
06-1
1
10-0
9
12-0
9
02-1
0
04-1
0
06-1
0
08-1
0
10-1
0
12-1
0
02-1
1
04-1
1
06-1
1
Chart 6. 2-year inflation compensation and expectation survey results (Mankiw et al., 2003)*.
596 M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
problems waiting in the near future. Therefore, long range forwardinflation compensation is treated as ameasure of inflationfighting cred-ibility of the central bank by academics (Gürkaynak et al., 2005b) andmarket analysts (Shen, 2006). However, it should be taken into accountthat the changes in this measuremay bemore informative than its levelbecause of the existence of a liquidity premium.
In Turkey, bondswithmaturities longer than 4 or 5 years are relative-ly illiquid and their information content is relatively poor. Thus, 3-yearforward 1-year inflation compensationmay be a better measure of infla-tion fighting credibility than the 5-year forward inflation compensationeven if a 3-year period may not be enough to rule out the effects of a cy-clical fluctuation.12 In Turkey, this measure oscillated around 5.5% untilthe last quarter of 2010, but it shifted a bit upwards since then, probablybecause the CBT began to explicitly highlight the increasing importanceof financial stability in its monetary policy stance (Chart 5).
6. Comparison of inflation compensation and survey results
Most common sources of the measurement of inflation expectationsare surveys and financial markets.13 However, each source has some
12 In a full-credibility case, after a period long enough to rule out the effects of tem-porary shocks in the short run, expected inflation rate should be equal to the centralbank's inflation target and there should be no doubts regarding the attainability of thistarget. In such a case inflation uncertainty is zero and hence the forward inflation com-pensation is equal to the inflation target. On the other hand, expected inflation exceed-ing the target and/or existence of inflation uncertainty indicate that the inflationfighting credibility of the central bank is decreasing.13 Most widely used market-based inflation expectation measure is inflationcompensation.
advantages anddisadvantages over the other. Survey results reflect infla-tion expectations directly since they do not contain some other factorssuch as liquidity conditions and inflation uncertainty. On the otherhand, there are some shortcomings of survey results. First of all, surveysare often conducted at low frequencies such as quarterly, monthly or bi-weekly. Also, there are some time lags between data collection and theannouncement of survey results. Inflation forecasting ability of respon-dents is assumed as equal and hence they are weighted equally. Besides,respondents may response different than what they think in order tohide private information (Cunningham et al., 2010). In addition tothese, reliability of survey results is limited since the respondents arenot liable with their responses (Deacon and Derry, 1994). On the con-trary, inflation expectations extracted fromnominal and real yield curvesare derived from the bondpriceswhich are formedby the transactions ofmarket participants and hence the market participants are liable withtheir forecasts. They lose money if they are not skillful or if they actagainst their forecasts. Thereforemarket-basedmeasures of inflation ex-pectations are more reliable in this sense. Another advantage ofmarket-based measures is that they can be observed at daily or even atintra-daily frequency (Christensen et al., 2004). However, market-basedmeasures also have some shortcomings such as existence of time-varying liquidity premium, inflation risk premium and indexation lag. Inaddition, flight-to-quality phenomenon affects conventional bondsmore during crisis periods, and hence pushes inflation compensationupwards, away from expected inflation level. Inflation compensationconverges to actual inflation expectations as these effects are removed.
Most widely used method of comparing these two approaches isbased on the comparison of actual inflation forecasting performanceof the measures. However, this method cannot be used in this study
15 For detailed information on the measurement of market liquidity in Turkey, please
5
5.5
6
6.5
7
7.5
1 1.5 2 2.5 3 3.5 4
Term (year)1 1.5 2 2.5 3 3.5 4
Term (year)
Survey: 1.30% Actual: 1.83%
02.11.2010
03.11.20105.5
6
6.5
7
7.5
8 Survey: 0.97% Actual: 0.03%
02.12.2010
03.12.2010
Chart 7. The effects of inflation surprises on inflation compensation.
Table 2Estimation results of regression of inflation compensation on inflation surprises.
Sample period: March 2010–June 2011 (real yield curves for the inflation announcementdates of July 2010 and May 2011 cannot be estimated due to the lack of observations ofinflation-indexed bond prices and hence these two months are excluded from the analy-sis). Standard errors are given in parenthesis.(*): significant at 10% level, (**): significant at 5% level, (***): significant at 1% level.
597M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
due to the absence of enough data tomake a reliable comparison for thetime being.
Time series plots of 2-year inflation compensation and survey expec-tations regarding 2-year inflation indicate that inflation compensation ismore volatile than survey results (Chart 6, left panel). This phenomenonis not specific to Turkish data and is valid for other countries as well. Oneexplanation of this is that the source of excess volatility is not the varia-tion in inflation expectations but the variation in inflation uncertainty orliquidity conditions.14 Moreover, measurement errors in surveys alsocontribute to this difference in volatility, since the expectations obtainedfrom survey results do not necessarily reflect the exact level of expectedinflation which are essentially unobservable. In fact, inter-quartile rangeof the distribution of the CBT survey responses as a measure of inflationuncertainty moves together with inflation compensation most of thetime (Chart 6, right panel). In other words, inflation compensation usu-ally increases when inflation uncertainty increases.
7. Effects of liquidity conditions on inflation compensation
One of the main components of inflation compensation is the li-quidity premium spread as stated in Section 5. If the liquidity effectscan be removed from inflation compensation, only the informationregarding inflation expectations and inflation uncertainty will be
14 As an example, please see D'Amico et al. (2008). The authors provide the samecomparison by using the U.S. data and find similar results.
left. This enables the extraction of information regarding the distribu-tion of inflation expectations. In this context, we define two measuresof liquidity conditions which are the relative trading volume anddaily turnover rates of inflation-indexed bonds.15 Linear OLS regres-sions are estimated in order to analyze the effects of both measureson 2, 3 and 4-year inflation compensation. According to the regres-sion results given in Table 1, at daily frequency, effects of both mea-sures on inflation compensation are found to be statisticallyinsignificant. Moreover, the coefficients of determination of thesemodels indicate that the explanatory power of liquidity conditionson inflation compensation is very low.
In fact, it is not surprising that the daily measures of liquidity condi-tions are unable to explain variations in inflation compensation sincedaily changes in liquidity are not likely to be reflected in the pricing ofliquidity premium immediately. A lower-frequency liquidity measuresuch as the monthly turnover rate may have significant explanatorypower on inflation compensation, but data at such a lower frequencyis not employed in this study due to the absence of enough number ofobservations to perform a reliable regression analysis for the timebeing and it is left to further studies on the subject.
Statistically insignificant coefficient estimates are not completelymeaningless for our purposes in this study since they enable us to at-tribute the daily variations in inflation compensation mainly tochanges in inflation expectations and uncertainty. In other words, in-flation compensation derived from yield curves is suitable for theevent-study methodology.
8. Event studies
An event study is a method to assess the effects of an event on thevalue of some variable of interest by comparing its value just before andimmediately after the event so that the change in the value can be attrib-uted mainly to the event.16 If noises do not exist between the dates (ortimes) compared, the analysis become completely reliable and unbiased.Thus, the methodology is more likely to reflect purely the effects of theevent on the variable of interest when the time span is the lowest possi-ble. In this study, we perform event studies by comparing the inflationcompensation curves before and after the event takes place.
Inflation compensation is mainly determined by inflation expecta-tions, inflation uncertainty and liquidity conditions as previouslyshown in this paper. Among these factors, liquidity conditions do notsignificantly change in one day or a few days (as shown in Section 7),but inflation expectations and inflation uncertainty can be affected byevents such asmonetary policy decisions and inflation surprises. For in-stance, an unanticipated interest rate cut may lead to an increase in the
see Yıldırım (2011).16 For detailed information regarding event study methodology, please seeGürkaynak et al. (2005a).
Chart 8. Relation between inflation surprises and inflation compensation.
598 M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
inflation expectations since an interest rate cut is expected to increasethe aggregate demand. Similarly, an unusual monetary policy actionsuch as increasing the required reserve ratios along with an interestrate cut as implemented by the CBT on 16 December 2010,may confusethe investors and lead to an increase in inflation uncertainty.
Since both inflation expectations and inflation uncertainty areimportant for monetary authorities in decision making, informationextracted from event studies performed using inflation compensa-tion is valuable. However, it should be taken into account that some-times bond-specific or term-specific factors17 may becomenon-negligible or liquidity conditions may change quickly. There-fore, while performing event-studies; one should be careful about li-quidity conditions, potential noises and potential bond-specific orterm-specific factors before coming to a conclusion.18 In this section ofthe study, the effects of some inflation surprises and events related tomonetary policy that took place in 2010 and 2011 on inflation compen-sation are analyzed as examples in order to demonstrate the use ofevent study methodology.
8.1. Inflation surprises
An inflation surprise can be defined as the significant difference be-tween the expected inflation level and the actual inflation.19 October2010, November 2010 and June 2011 inflation announcements inTurkey are discussed as examples of inflation surprises in this sectionand event studies on these dates are illustrated.
Actual monthly inflation in October 2010 was 53 bps higher thanexpected. This shifted the forward 1-year inflation compensation upin short terms but the effects of the surprise disappear after twoyears (Chart 7, left panel).20 Hence, the event study suggests thatthis surprise was perceived by market participants as a temporary
17 Factors that are defined as εn−εrð Þ in Eq. (3).18 No other important event should take place in the analyzed period in order for theevent study to be reliable.19 In this study, we identify an inflation announcement as surprise if there is at least50 basis points difference between actual monthly inflation level and the average infla-tion forecast of the Bloomberg survey which is conducted shortly before the inflationannouncement. According to this criterion there are 2 upward and 3 downward sur-prises between January 2010 and June 2011. Data regarding Bloomberg survey resultsand actual monthly inflation in Turkey are given in Table A.3.20 On the 3rd day of each month at 10:00 a.m., previous month's inflation data is re-leased by TurkStat. Therefore that day's closing prices contain effects of inflationannouncement.
shock. A close look at the price developments in Turkey at that timeconfirms this finding since the unexpected increase in inflation inOctober 2010 was stemmed from the fluctuations in food prices(CBT, 2011a). Moreover, this surprise was later corrected in November2010. This 94-bp downside surprise in monthly inflation shifted for-ward inflation compensation curve downwards in terms up to 2 years(Chart 7, right panel). Similar to the previous month, this surprise wascaused by the correction in the unprocessed food prices (CBT, 2011a).Hence, market participants perceived this as a temporary shock as well.
In addition to the graphical analysis, the effects of inflation surpriseson inflation compensation are also analyzed by regressing the change ininflation compensation on inflation surprise. In this context, a univari-ate linear regression model is employed and estimated using datafrom March 2010 to June 2011. The model is in the following form;
ΔIC ¼ β0 þ β1ISþ ε: ð4Þ
Here, ΔIC stands for the change in inflation compensation betweenthe closest dates before and after inflation announcement, and IS standsfor the difference of actual inflation from survey forecast.
This regression is estimated separately for inflation compensation ofdifferent terms from 1 year to 4 years along with the 3-year forward1-year inflation compensation and the results are given in Table 2.
According to the results, inflation surprises have significant effectson inflation compensation in all terms. All coefficient estimates arepositive as expected. However, the coefficients decline as the termincreases and the explanatory power of inflation surprises on inflationcompensation falls (Chart 8, left panel). For instance, a 100 bps upwardsurprise in inflation causes a 30 bps increase in 2-year inflationcompensation, but the increase is only 11 bps in 4-year inflationcompensation (Chart 8, right panel). In addition, approximately half ofthe variations in 2-year inflation compensation can be explained byinflation surprises where the explained portion is only one third for4-year inflation compensation.21 In fact, effect of inflation surprises on3-year forward 1-year inflation compensation is statistically insignificant.Although the number of observations is very limited, the regression
21 As a robustness check, this analysis is also performed controlling for liquidity in themodel. In order to do so, change in liquidity conditions between the closest dates be-fore and after inflation announcement is added as an additional explanatory variableto the model. So, the model is estimated in the form ΔIC ¼ β0 þ β1ISþ β2ΔL þ ε. Here,ΔL is both measured as the change in turnover rate and the change in relative tradingvolume of inflation-indexed bonds. Estimation results are consistent with the resultsreported in Table 1. These results are reported in Table A.1 and Table A.2
6
6.5
7
7.5
8
8.5
1 1.5 2 2.5 3 3.5 4
Term (year)
1 1.5 2 2.5 3 3.5 4
Term (year)
Deputy Governor's speech on December 11, 2010
10.12.2010
13.12.2010
6
6.5
7
7.5
8
8.5
December 2010 MPC Meeting
16.12.2010
17.12.2010
Chart 9. Effects of monetary policy communication on inflation compensation.
599M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
results are in line with graphical analyses and imply that the effects of in-flation surprises on inflation compensation are not persistent. Therefore itcan be concluded that the inflation expectations are contained.22
7.5
8
8.2. Monetary policy
Central banks influence inflation expectations and inflation uncer-tainty through monetary policy measures as well as their communi-cation tools such as speeches and inflation reports by which centralbanks publicize their inflation forecasts and guidance on future mon-etary policy stance. In this section it is shown by some examples thatinflation compensation is a useful tool to analyze the effects of mon-etary policy announcements on inflation expectations.
In the last quarter of 2010, increased concerns on financial stabil-ity pushed the CBT to adjust its monetary policy orientation towardsfinancial stability without giving up its price stability goal. As part ofthis change, the CBT adopted a new policy mix to attain these twogoals. Deputy Governor Erdem Basci introduced the details of thisnew policy mix in his speech on December 11, 2010 and signaled in-terest rate cuts in the coming period. This led to a slight increase ininflation compensation curve at all terms (Chart 9, left panel). De-creased importance of inflation in the objective function of the CBTprobably caused the investors to think that the CBT would be lesswilling to fight inflation. As a result, both the level and uncertaintyof future inflation expectations shifted up.
In the first MPC meeting after this speech, the CBT cut the policyinterest rate by 50 basis points to distract capital inflows and hikedthe TL required reserve ratios to curb the credit growth rate in thecountry. However, these measures did not have significant effectson inflation compensation (Chart 9, right panel). Most likely, contrac-tionary effect of required reserve increases and expansionary effect ofinterest rate cut balanced each other in terms of inflationary pres-sures as noted by the CBT after the MPC meeting.
In addition to monetary policy decisions and communication, the CBTmay also influence the inflation expectations via its inflation forecasts.The CBT usually shares its forecasts in its quarterly inflation report. Asan example, in the CBT's April 2011 inflation report, due to soaring energyprices and renewed tariffs which are beyond the control of monetary
22 The inflation expectations are said to be anchored if the short-term inflation expec-tations can not affect long-term inflation expectations. On the contrary, if short-termexpectations pass through long-term one-to-one (fully), then the inflation expecta-tions are said to be unmoored. Between these two extreme situations, inflation expec-tations are said to be contained if short-term expectations affect the long term lessthan one-to-one and if the effect fades with the term. These three situations have im-portant implications for the credibility and implementation of monetary policy. For de-tailed information, please see Gürkaynak et al. (2010a), Potter and Rosenberg (2007)and Jochmann et al. (2009).
policy, the year-end-2011 inflation forecast was revised up by 1percentage point (CBT, 2011b). Consequently, while short term inflationcompensation increased slightly, there was no significant change in longterm inflation compensation (Chart 10). Since the increase in theinflation forecast was caused by temporary factors, financial marketparticipants increased their inflation expectations only in shorter terms.
9. Conclusion
Inflation compensation derived from yield curves contains infor-mation on inflation expectations of financial market participantsand acts as market-based alternative to surveys. As shown in previoussections, calculating inflation compensation using yield curves hasseveral advantages over using actual bond returns since the latterprovides limited and inaccurate information. Moreover, the formerenables calculation of future forward inflation compensation whichhas important implications for monetary policy.
Another important advantage of inflation compensation is its fre-quency. Even though inflation compensation is more volatile thansurvey results, it enables the comparison of two dates just beforeand after an event. Since liquidity conditions do not significantly af-fect the pricing of bonds used in the calculation of inflation compen-sation, daily changes in the inflation compensation can be attributedmainly to changes in inflation expectations and inflation uncertainty.In this context, the effects of events such as inflation surprises andmonetary policy announcements on inflation expectations can beexamined through event studies. However, in order to be cautious,while performing event studies of inflation compensation, changesin the liquidity conditions and other factors that may influence thepricing of bonds should be taken into account. In this way, thishigh-frequency and market-based measure will possibly prove veryuseful in economic decision making process of central banks, othergovernment institutions, banks, firms and other investors.
5.5
6
6.5
7
1 1.5 2 2.5 3 3.5 4
Term (year)
27.04.2011
28.04.2011
Chart 10. The effect of inflation forecast updates on inflation compensation.
Table A.1Inflation compensation and inflation surprises (turnover rate is taken as indicator of market liquidity).
Dependent variable Inflation surprise Turnover rate of inflation-indexed bonds
Constant R2 Number ofobservations
1-year inflation compensation 0.682***(0.181)
−51.090(57.476)
−0.121(0.124)
0.571 14
1.5-year inflation compensation 0.457***(0.119)
−25.851(37.799)
−0.082(0.082)
0.585 14
2-year inflation compensation 0.314***(0.088)
−14.488(27.826)
−0.036(0.060)
0.555 14
2.5-year inflation compensation 0.226***(0.072)
−10.765(22.902)
−0.003(0.049)
0.489 14
3-year inflation compensation 0.174**(0.062)
−11.123(19.670)
0.016(0.042)
0.429 14
3.5-year inflation compensation 0.145**(0.054)
−13.448(17.263)
0.027(0.037)
0.397 14
4-year inflation compensation 0.131**(0.049)
−16.478(15.656)
0.031(0.034)
0.392 14
(*): significant at 10% level, (**): significant at 5% level, (***): significant at 1% level.
Table A.2Inflation compensation and inflation surprises (relative trading volume is taken as market liquidity indicator).
(*): significant at 10% level, (**): significant at 5% level, (***): significant at 1% level.
Appendix A
Table A.3Actual monthly inflation and survey results. Inflation announcements recognized as surprises (difference greater than 50 bps) are highlighted in bold.
600 M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
Appendix B. Forward inflation compensation and its derivation
An n-year forward m-year rate is defined as the expected m-yearrate at time n-year from now. Assuming zero-arbitrage condition;an n-year forwardm-year rate, rn,m can be calculated from the follow-ing expression;
A 1þ r0;nþm
� �nþm ¼ A 1þ r0;n� �n
1þ rn;m� �m
:
The left hand side of this equation is the future value of an amountA invested in an n + m year security at time 0 and the right hand sideof this equation is the future value of the same amount first investedin an n-year security and then re-invested in an m-year security attime n. If one side exceeds the other, an arbitrage opportunity arisesbut such disequilibrium is thought to be removed quickly by arbitra-geurs so that the equality holds again.
If yields r0;nþm and r0;n can be obtained from a zero-yield curve,n-year forward m-year yield that is rn;m can easily be computed.
If the yields are continuously compounded, the expression be-comes as follows;
Ae nþmð Þr0;nþm ¼ Aenr0;n emrn;m
⇒e nþmð Þr0;nþm ¼ enr0;nþmrn;m
⇒ nþmð Þr0;nþm ¼ nr0;n þmrn;m
⇒rn;m ¼ nþmð Þr0;nþm−nr0;nm
:
So, for instance, a three-year forward one-year yield can be calcu-lated as;
r3;1 ¼ 4r0;4−3r0;34−3
¼ 4r0;4−3r0;3:
601M. Duran, E. Gülşen / Economic Modelling 32 (2013) 592–601
A nominal yield can be expressed as the sum of the real yield andthe inflation compensation;
r ¼ rrþ IC:
Subtracting an n-year forward m-year real yield from an n-yearforward m-year nominal yield equals n-year forward m-year inflationcompensation;
nþmð Þrr0;nþm−nrr0;n þ nþmð ÞIC0;nþm−nIC0;n
m−
nþmð Þrr0;nþm−nrr0;nm
¼ nþmð ÞIC0;nþm−nIC0;n
m¼ ICn;m:
This measure is completely free of the inflation compensation offirst n years and hence completely reflects the market's views regard-ing the remaining m-year period.
References
Akıncı, Ö., Gürcihan, B., Gürkaynak, R., Özel, Ö., 2007. Devlet İç Borçlanma Senetleri İçinGetiri Eğrisi Tahmini. İktisat İşletme ve Finans 22 (252), 5–25.
Bank for International Settlements, 2005. Zero-coupon Yield Curves: TechnicalDocumentation. BIS Paper, No:25.
Central Bank of Turkey, 2011a. Inflation Report 2011-I.Central Bank of Turkey, 2011b. Inflation Report 2011-II.Christensen, I., Dion, F., Reid, C., 2004. Real return bonds, inflation expectations, and the
break-even inflation rate. Bank of Canada Working Paper, No: 2004-43.Cunningham, R., Desroches, B., Santor, E., 2010. Inflation expectations and the conduct
of monetary policy: a review of recent evidence and experience. Bank of CanadaReview 13–26 (Spring).
D'Amico, S., Kim, D.H., Wei, M., 2008. TIPS from TIPS: the informational content oftreasury inflation-protected security prices. FEDS Working Paper, No: 19.
Deacon, M., Derry, A., 1994. Estimating market interest rate and inflation expectationsfrom the prices of UK government bonds. Bank of England, Quarterly Bulletin 34(3), 232–240 (August).
Gürkaynak, R., Sack, B., Swanson, E., 2005a. The sensitivity of long-term interest rates toeconomic news: evidence and implications for macroeconomic models. AmericanEconomic Review 95, 425–436.
Gürkaynak, R., Sack, B., Swanson, E., 2005b. Do actions speak louder than words?International Journal of Central Banking 1 (1), 55–93.
Gürkaynak, R., Levin, A., Swanson, E., 2010a. Does Inflation targeting anchor long-runinflation expectations? Evidence from long-term bond yields in the U.S., U.K., andSweden. Journal of the European Economic Association 8 (6), 1208–1242.
Gürkaynak, R., Sack, B., Wright, J.H., 2010b. The TIPS yield curve and inflation compensation.American Economic Journal: Macroeconomics 2 (1), 70–92 (January).
Jochmann, M., Koop, G., Potter, S.M., 2009. Modeling the dynamics of inflation compensa-tion. Journal of Empirical Finance 17, 157–167.
Mankiw, N.G., Reis, R., Wolfers, J., 2003. Disagreement about inflation expectations.National Bureau of Economic Research (NBER) Working Papers, No: 9796.
Potter, S.M., Rosenberg, J., 2007. Are US inflation expectations anchored, contained orunmoored? Federal Reserve Bank of New York, unpublished working paper.
Shen, P., 2006. Liquidity risk premia and breakeven inflation rates. Federal ReserveBank of Kansas City, Economic Review 91 (2), 29–54 (Second Quarter).
Treasury of Turkey, 2009. Consumer Price Index (CPI) indexed government bonds.Investors' Guide.
Yıldırım, B.D., 2011. Türkiye'nin Finansal Piyasa Likiditesi, Ölçümü ve Analizi. CentralBank Review 11 (1), 11–28.
Murat DuranMurat Duran was born in Düzce, Turkey in 1984 andearned his bachelor's degrees in Business in 2006 andEconometrics (double major) in 2007 from Istanbul Uni-versity. After graduation, he began working for Finansbankas an auditor. Since 2009, he has worked for the CentralBank of Turkey (CBT) as a researcher and he is currentlyseeking an MSc degree in Economics from Middle EastTechnical University. He has papers that are published inEconomics Letters, Central Bank Review and CBT WorkingPaper Series. His main research fields of interest includeMonetary Economics, Financial Economics, Econometrics,
Operations Research and Simulation.
Publications
• Measuring the Impact of Monetary Policy on Asset Prices in Turkey (Murat Duran,Gülserim Özcan, Pınar Özlü, Deren Ünalmış)—Economics Letters—2011
• Türkiye İçin Getiri Eğrileri Kullanılarak Enflasyon Telafisi Tahmin Edilmesi (MuratDuran, Eda Gülşen, Refet Gürkaynak)—Central Bank of Turkey, Working Papers, No.11/22—2011 (In Turkish)
• Türkiye'de Endeksli Bonolar Kullanılarak Enflasyon Telafisi Ölçülmesi (Murat Duran,Eda Gülşen, Refet Gürkaynak)—Central Bank of Turkey, Economic Notes, No. 11/15—2011 (In Turkish)
• TCMB Faiz Kararlarının Hisse Senedi Piyasaları Üzerine Etkisi (Murat Duran, PınarÖzlü, Deren Ünalmış)—Central Bank Review—2010 (In Turkish)
• Measuring the Impact of Monetary Policy on Asset Prices in Turkey (Murat Duran,Gülserim Özcan, Pınar Özlü, Deren Ünalmış)—Central Bank of Turkey, WorkingPapers, No. 10/17—2010
• TCMB Faiz Kararlarının Piyasa Faizleri Ve Hisse Senedi Piyasaları Üzerine Etkisi(Murat Duran, Refet Gürkaynak, Pınar Özlü, Deren Ünalmış)—Central Bank of Turkey,Economic Notes, No. 10/08—2010 (In Turkish)
Eda GülşenBorn in 1983, Eda Gülşen graduated from the Departmentof Economics at Middle East Technical University (METU)in 2006. She received MS degree in Economics fromMETU where she worked as a research and teaching assis-tant between 2006 and 2009. She is currently working as aresearcher in Research and Monetary Policy Department ofCentral Bank of Turkey. Her main research fields of interestinclude Monetary Economics, International Macroeco-nomics and Macro econometrics.
