Download Identifying and Analyzing the TIPS Liquidity Premium: A State-Space Model Approach and more Study notes Banking and Finance in PDF only on Docsity!
TIPS Liquidity Premium and Quantitative Easing
Laura Coroneo∗
23rd April 2018
Abstract We assess the effect of the QE2 program on the TIPS liquidity premium using a latent factor approach and a counterfactual exercise. In the context of a state-space model for nominal and TIPS yields, we identify the TIPS liquidity premium as the common component in TIPS yields that is unspanned by nominal yields. We then construct a counterfactual TIPS liquidity premium that exploits suitable conditioning information. Results indicate that the QE2 program had only limited effect on the TIPS liquidity premium.
JEL classification codes: C33, C53, E43, G12.
Keywords: TIPS; Liquidity Premium; Factor models; Quantitative Easing
∗University of York, Department of Economics and Related Studies, Laura.Coroneo@york.ac.uk. The author thanks Adam Golinski, Paulo Santos Monteiro, Peter Spencer and seminar participants at the Bank of England, the University of Nottingham, the University of York, the Financial Econometrics and Empirical Asset Pricing Conference (University of Lancaster), the 2016 EEA-ESEM Congress, 2016 RCEA Macro-Money-Finance Workshop, the 10th International Conference on Computational and Financial Econometric, the 2015 Money, Macro and Finance Conference and the 11th BMRC/DEMS Conference for useful comments. The support of the ESRC grant ES/K001345/1 is gratefully acknowledged.
1 Introduction
Following the 2008 financial crisis, the Federal Reserve conducted large-scale asset purchases, known as quantitative easing (QE), in order to lower long-term interest rates and spur economic growth. A large literature focusses on the impact of these large-scale asset purchases on nominal Treasuries, and there is a consensus that QE successfully reduced yields on long-maturity nominal Treasuries, see among others Krishnamurthy and Vissing-Jorgensen (2011), Gagnon, Raskin, Remache, Sack et al. (2011) and Joyce, Lasaosa, Stevens and Tong (2011). However, little attention has been paid to the effect of the QE program on the Treasury Inflation-Protected Securities (TIPS) liquidity premium. TIPS are fixed-income securities with coupons and principal payments indexed to the non- seasonally-adjusted CPI for all urban consumers. They were introduced in 1997 and now consti- tute 10% of the outstanding U.S. Treasury debt. Given that TIPS have a smaller and less liquid market than US nominal Treasury bonds, they pay a liquidity premium, representing the compen- sation required by investors to hold a security that is less liquid than its nominal counterpart, see G¨urkaynak, Sack and Wright (2010), D’Amico, Kim and Wei (2018) and Campbell, Shiller and Viceira (2009), among others. The QE program included purchases of large amounts of TIPS, and assessing the effects of TIPS purchases on the TIPS liquidity premium may allow to further understand the benefits and risks of QE programs. Large-scale TIPS purchases may have affected the TIPS liquidity premium through two opposite channels: the liquidity channel and the scarcity channel. The liquidity channel implies that large- scale asset purchases may have reduced the liquidity premium required by investors to buy TIPS, as the Federal Reserve’s purchases may have made it less costly for investors to sell TIPS. However, this effect of the QE program on the TIPS liquidity premium depends on the Federal Reserve’s flow of purchases and, therefore, it should have been limited to the duration of the QE program. On the other hand, the scarcity channel implies that the Federal Reserve’s TIPS purchases may have reduced the already scarce stock of TIPS available to investors and this instead may have increased
estimates of the TIPS liquidity premium factor by Quasi-Maximum Likelihood using the Kalman filter and the EM algorithm following the approach proposed by Coroneo, Giannone and Modugno (2016). We then analyze the effect on the TIPS liquidity premium of the QE2 program announced on November, 3 2010 and implemented from November, 23 2010 to June, 17 2011. This program involved $600 billion purchases of Treasury securities, of which $26 billions were TIPS purchases, implying that the Federal Reserve’s purchases of Treasuries within the QE2 program have been larger and concentrated in a shorter time spam with respect to the other large-scale asset purchases. We assess the effect of the QE2 programme on the TIPS liquidity premium by performing a counter- factual analysis. This is easily implementable in our framework regardless of the dimensionality of the conditioning variables, thanks to the use of the state-space representation and Kalman filtering, see Ba´nbura, Giannone and Lenza (2015). Our objective is to construct a counterfactual path for the TIPS liquidity premium factor that does not incorporate the QE2 program, but that exploits suitable conditional information. In our framework, a conditioning variable is suitable for the construction of a counterfactual TIPS liquidity premium if it has predictive ability for the TIPS liquidity premium, but it is not directly affected by the QE nor by the TIPS liquidity premium itself. We find that measures of financial stress, such as the market-wide illiquidity measure of Hu, Pan and Wang (2013) and the corporate spreads, satisfy these conditions. The resulting counterfactual TIPS liquidity factor is on average higher than the realized TIPS liquidity factor, however the difference is only marginally significant when taking into account the accuracy of the counterfactual. We therefore conclude that the QE program had only a marginal effect on the TIPS liquidity premium, and that the liquidity channel was only marginally stronger than scarcity channel. This paper is related to Christensen and Gillan (2013) and Abrahams, Adrian, Crump, Moench and Yu (2016), that also analyze the impact on the QE2 on the TIPS liquidity premium. Christensen and Gillan (2013) find that the QE2 reduced the TIPS liquidity premium, implying that the liquidity channel has been stronger than the scarcity channel, while Abrahams et al. (2016) do not find any
effect of the QE2 program on the TIPS liquidity premium. The main difference with our work is how the TIPS liquidity premium is estimated. Abrahams et al. (2016) use an observable liquidity proxy, while Christensen and Gillan (2013) use replicating portfolios. Our counterfactual results reconcile the results in Abrahams et al. (2016) with the ones is Christensen and Gillan (2013). In particular, as Abrahams et al. (2016) we find that overall the QE2 program did not affect the TIPS liquidity premium, however as Christensen and Gillan (2013) we also document that the effect of the QE2 program on the TIPS liquidity premium peaked during the middle of the program in April 2011, when the realized TIPS liquidity premium was significantly smaller than the counterfactuals. In in this short time period, the liquidity channel was significant, and stronger than the scarcity channel. The paper is organized as follows. In Section 2, we define and identify the liquidity premium in the TIPS market. Section 3 introduces the data set and presents some preliminary evidence. Section 4 outlines the estimation procedure. In Section 5 we report estimation results. Section 6 describes the Quantitative Easing programme. Section 7 shows the counterfactual results and Section 8 contains robustness results using an alternative data set and model specification. Finally, Section 9 concludes.
2 Identifying the TIPS liquidity premium
2.1 Decomposing nominal and TIPS yields
The yield of a nominal zero-coupon Treasury bond of any maturity can be decomposed into the underlying real yield, the inflation expectation over the remaining life of the bond and the inflation risk premium. Denoting by yt,τN the nominal yield with maturity τ , we can write
yt,τN = yRt,τ + πet,t+τ + IPt,t+τ (1)
where aLτ is the maturity-specific intercept, BLτ contains the factor loadings of the liquidity premium of the TIPS with maturity τ on the liquidity premium factor and εLt,τ represents the maturity specific component of the liquidity premium of the TIPS with maturity τ. In order to extract the TIPS liquidity premium factor Lt in Equation (3), we model nominal and TIPS yields using a dynamic factor model. We assume that the yield curve of nominal Treasuries is described by a few latent yield curve factors, while the TIPS yield curve is driven by both the yield curve factors and the TIPS liquidity factor. In practice, following Equations (1)–(2), we identify the liquidity premium in the TIPS market Lt as the common component in TIPS yields that is unspanned by nominal yields. Formally, we assume that nominal yields with different time to maturities are driven by a vector of rX × 1 of common latent factors Xt as follows
yt,τN = aNτ + BτN Xt + εNt,τ (4)
where aNτ is the maturity-specific intercept, BNτ contains the factor loadings of the nominal yield with maturity τ on the latent factors (common across maturities) and εNt,τ represent the maturity- specific component of the nominal yield with maturity τ. In the same way, for real yields we have yRt,τ = aRτ + BτR Xt + εRt,τ (5)
where aRτ is the maturity-specific intercept, BRτ contains the factor loadings and εRt,τ is the maturity- specific component of the real yield with maturity τ. Let the vector ytN = (yNt,τ 1 N ,... , yNt,τ (^) nN )′^ collect the nominal yields with maturities (τ 1 N , ,... , τ (^) nN ) and the vector yTt = (yt,τT 1 T ,... , yt,τT (^) mT )′^ collect the TIPS yields with maturities (τ 1 T ,... , τ (^) mT). Follow- ing equations (2)–(5), the joint model for nominal and TIPS yields can be written as yNt ytT
aN aT
BN^0
BR^ BL
Xt Lt
εNt εTt
where aT^ = aR^ + aL^ and εTt = εRt + εLt. Equation (6) identifies two sets of latent factors: the yield curve factors Xt and the TIPS liquidity factor Lt. The yield curve factors are in line with the literature that exploits the high level of comovement of yields with different maturities to provide a parsimonious representation of the yield curve. This literature has proven that yield curve factor models are very successful in fitting the yield curve of nominal interest rates, see Litterman and Scheinkman (1991), Duffee (2002) and Coroneo, Nyholm and Vidova-Koleva (2011). As for the liquidity factor, by noticing that TIPS are less liquid than nominal Treasuries, see G¨urkaynak and Wright (2012) and Fleckenstein, Longstaff and Lustig (2014), we are able to identify the TIPS liquidity premium factor Lt as the driver of the wedge between real and TIPS yields. We implement this identification condition through the zero factor loading restrictions in Equation (6), which imply that the liquidity factor does not affect the current nominal yield curve, as shown in Equations (1)–(2). This approach is in line with D’Amico et al. (2018), that jointly model nominal and TIPS yields in the framework of an affine term structure model, and allow TIPS yields to be driven by a factor that is unspanned by nominal Treasuries. We allow the (n + m) idiosyncratic components collected in εt = [(εNt )′, (εTt )′]′^ to follow inde- pendent univariate AR(1) processes
εt = Aεt− 1 + vt, vt ∼ N (0, R) (7)
where A and R are diagonal matrices, implying that the common factors fully account for the joint correlation of the observations. The (rX + 1) × 1 vector of zero mean latent factors follow a VAR(1) Xt Lt
ΦXX^ ΦXL
ΦLX ΦLL
Xt−^1 Lt− 1
uXt uLt
where the innovations ut = [(uXt )′, (uLt )′]′^ are normally distributed with zero mean and variance
In Table 1 we report the percentage of variance of nominal yields and of jointly nominal and TIPS yields explained by the first five principal components extracted, respectively, from nominal yields and jointly from nominal and TIPS yields. Results in Table 1 show that two factors fully explain the cross-section of nominal yields, but when considering nominal and TIPS yields jointly an additional factor should be included in the analysis. This indicates that TIPS yields are driven by a factor that is unspanned by nominal yields which accounts for the liquidity premium in the TIPS market. Accordingly, in our analysis we use two factors to explain the cross-section of nominal yields, i.e. rX = 2, and one factor to explain the liquidity premium in the TIPS market. This choice is due to the peculiarity of our data set, namely daily yields from mid to very long maturities. In Section 8, we report results on weekly data with short to long maturities for which we estimate a model with three yield curve factors, in line with Litterman and Scheinkman (1991). To compare our estimates of the TIPS liquidity premium, we construct a liquidity proxy from inflation swap rates. We use mid-quotes of inflation swap rates with maturity 3, 5, 8, 10, 15 and 20 years from Datastream converted to continuously compounded basis. Following Haubrich, Pennacchi and Ritchken (2012), we compute real rates as the difference between equivalent maturity nominal Treasury yields and inflation swap rates. We then construct a liquidity proxy as the average, across maturities, of the difference between equivalent maturity TIPS yields and real rates constructed using inflation swaps. As possible conditioning variables for the counterfactual analysis, we consider the TED spread (defined as the spread between the three month LIBOR and the three-month Tbill rates), the Chicago Board Options Exchange Volatility Index (VIX), the Cleveland Financial Stress Index, the corporate spread (defined as the spread between the Baa corporate and the ten-year Treasury rates), the bid-ask spread on the three-month Tbill and the illiquidity measure of Hu et al. (2013) which is a market-wide measure of illiquidity. Data for all variables is obtained from the FRED database, except for the illiquidity measure of Hu et al. (2013), available from the authors.^1 (^1) The illiquidity measure of Hu et al. (2013) is available at http://www.mit.edu/ junpan/
4 Estimation
The joint model for nominal and TIPS yields in Equations (6)–(8) is a restricted dynamic factor model with autocorrelated idiosyncratic components. In order to cast the model in a state-space form, we augment the vector of state variables with the vector of idiosyncratic components εt and an additional state variable ct restricted to one at every period (by fixing its initial value to one and the variance of its innovations to zero). We then rewrite the measurement equation as
ytN ytT
BN^0 aN^ In^0 BR^ BL^ aT^0 Im
Xt Lt ct εNt εTt
vNt vTt
where Xt = (X 1 ,t, X 2 ,t)′, ((vtN )′, (vtT )′)′^ ∼ N (0, In+m) and is a coefficient that we fix to 1−^12. In the same way, we write the state equation as
Xt Lt ct εNt εTt
ΦXX ΦXL 0 0 0
ΦLX ΦLL 0 0 0
0 0 0 AN^0
0 0 0 0 AT
Xt− 1 Lt− 1 ct− 1 εNt− 1 εTt− 1
uXt uLt νt vNt vTt
with ((uXt )′, (uLt )′, νt, (vtN )′, (vtT )′)′^ ∼ N (0, blkdiag(Q, , R)) and A = diag(AN^ , AT^ ). The model in (9)–(10) is a restricted state-space model for which maximum likelihood estimators of the parameters are not available in closed form. Conditionally on the factors, the model reduces to a set of linear regressions. As consequence, we compute Maximum Likelihood estimates using the Expectation Maximization (EM) algorithm introduced by Shumway and Stoffer (1982) and Watson and Engle (1983). This estimator is feasible when the number of variables is large, and robust to
In practice, we are interested in extracting the TIPS liquidity factor for t ≥ t 0 from a data set of TIPS, nominal yields and conditioning variables, where TIPS and nominal yields are unobserved from t 0. Both the counterfactual TIPS liquidity factor in (11) and its accuracy in (12) can be easily computed in our framework regardless of the dimensionality of the conditioning variables, thanks to the use of the state-space representation and Kalman filtering, see Ba´nbura et al. (2015). Following Durbin and Koopman (2012), we use a modified state-space model where the dimensionality of the vector of observable variables varies over time. In practice, after t 0 only the rows that refer to the conditioning variables will enter into the measurement equation. We construct the (1 − α) confidence interval for the counterfactual TIPS liquidity premium factor as
CI(L∗ t|T ) 1 −α =
L∗ t|T − Φ−^1 (1 − α/2)
V (^) t∗|T , L∗ t|T + Φ−^1 (1 − α/2)
V (^) t∗|T
, t ≥ t 0 (13)
where Φ−^1 (1 − α/2) denotes the (1 − α/2) quantile of the standard normal distribution.
5 Estimated TIPS liquidity factor
In this section, we report estimation results for the joint model for nominal and TIPS yields in (9)- (10) using the full sample of data, i.e. from January, 2 2005 to December, 31 2014, with a particular focus on the estimated TIPS liquidity factor. Table 2 reports the percentage of variance of nominal and TIPS yields explained by the estimated latent factors. The table shows that the model has a good fit for both nominal and TIPS yields, with at least 96% of the variance of yields explained by the latent factors. The first latent factor explains the bulk of the variation in both nominal and TIPS yields, while the second yield curve factor explains up to 14% of the variance of nominal and TIPS yields. The TIPS liquidity factor by construction does not affect nominal yields but it has substantial explanatory power for TIPS yields. It explains up to 22.9% of the variance of TIPS yields and its explanatory power is higher for shorter maturities, implying that investors require higher compen-
Table 2: Variance explained by the latent factors Nominal Yields TIPS Yields Mat X 1 X 2 Total X 1 X 2 L Total 3 0.850 0.146 0.996 0.722 0.005 0.229 0. 5 0.947 0.053 1.000 0.810 0.001 0.179 0. 8 0.999 0.000 0.999 0.841 0.017 0.141 1. 10 0.991 0.009 1.000 0.840 0.032 0.125 1. 15 0.945 0.048 0.993 0.809 0.078 0.098 0. 20 0.901 0.058 0.960 0.744 0.135 0.088 0. Note: this table reports the percentage of variance of nominal yields (left panel) and of TIPS yields (right panel) explained by the esti- mated latent factors of the model in (9)-(10) for each observed matu- rity.
sation for holding shorter-term TIPS rather than longer-term ones. This might be due to the fact that growth of the TIPS markets has not occurred uniformly, see Shen (2006). For example, the Treasury has issued 10-year TIPS every year since the TIPS program began in 1997. On the con- trary, the 5-year TIPS, were issued in 1997 and 1998, but then not again until 2005. An additional explanation could be due to the presence of different types of investors in different segments of the TIPS yield curve, i.e. preferred-habitat investors as in Vayanos and Vila (2009). In particular, investors in the long end of the TIPS yield curve, e.g. pension funds, are more likely to buy and hold and, thus, do not require a liquidity premium in order to invest in TIPS since they will rarely need to turn over their positions. Figure 1 reports the estimated yield curve factors (top panel) and the TIPS liquidity premium factor (bottom panel). The first yield curve factor has a decreasing pattern in our sample due to a general decline in interest rates in this period. The second yield curve factor is more volatile than the first and is higher in the middle of the sample. The bottom plot of Figure 1 reports the TIPS liquidity factor and a proxy for the average liquidity in the TIPS market, constructed as the standardized average, across maturities, of the difference between TIPS yields and real rates computed using inflation swaps and nominal yields of the same maturity, as described in Section 3. The figure shows that the TIPS liquidity factor is highly correlated with this empirical
proxy. The pairwise correlation coefficient is 75%. We can also notice that the proxy for the average liquidity in the TIPS market constructed using inflation swaps is more volatile than the estimated TIPS liquidity premium factor. This may be due to the fact that inflation swaps are also subject to liquidity frictions, see Fleming and Sporn (2013), and therefore the liquidity proxy constructed using inflation swaps measures the liquidity premium in both TIPS and inflation swaps, see Christensen and Gillan (2012). Figure 1 also shows that during the subprime crisis, the TIPS liquidity premium became more volatile and, in September 2008, following the Lehman Brothers collapse, the liquidity premium in the TIPS market increased substantially. This indicates that in this period investors required a higher compensation in order to invest in an instrument that is less liquid than its nominal counterpart. The TIPS liquidity premium remained at this higher level until mid-2009, when it returned to pre-2008 levels.
6 Quantitative Easing
Following the 2008 financial crisis, the Federal Reserve conducted massive asset purchases know as quantitative easing (QE) to lower long-term interest rates and spur economic growth. On November 25, 2009 the Federal Open Market Committee (FOMC) announced the first quantitative easing program (QE1) which would involve purchases in government-sponsored en- terprises (GSEs) debt and in mortgage-backed securities (MBS). On March 18, 2009 the FOMC announced that the QE1 program would involve additional purchases in GSEs and MBS. It was also announced that the program would involve purchases of $300 billion in long-term Treasury securities. The Treasury purchases ended on October 29, 2009 and involved $6.1 billion in TIPS purchases. The second quantitative easing program (QE2) was announced on November 3, 2010 with the target of expanding the Federal Reserve’s balance sheet by $600 billions through Treasury security purchases over an eight-month period. The gross purchases of Treasury securities from November 3, 2010 until June 30, 2011 amounted to nearly $750 billion, of which about $26 billion were TIPS
purchases. On September 21, 2011 the FOMC announced the third QE program, know as maturity exten- sion program (MEP), which would involve Fed purchases of $400 billion in long-term Treasuries and equivalent sales in short-term Treasuries. On June 20, 2012 the FOMC announced that purchases of long-term bonds and the sales of short-term bonds would continue through 2012 and would involve a total of $600 billion in purchases and sales of securities. On December 12, 2012 the Fed announced that it would continue to purchase $45 billion in long-term Treasuries per month but without the sale of short-term Treasuries to sterilize purchases. The MEP involved TIPS purchases for a total of $27.1 billion, all in TIPS with more than 6 years to maturity. The TIPS sales within the MEP totaled $13.4 billion and only included TIPS with less than 3.25 years to maturity. In Figure 2 we report the Federal Reserve’s outright holdings of Treasury securities and TIPS. The figure shows that the Federal Reserve’s outright holdings of both nominal Treasuries and TIPS sharply increased during the QE2 program. This because the Fed purchases of Treasuries within the QE2 program have been larger and concentrated in a shorter time spam with respect to the QE1 and the MEP programs. Such massive purchases of TIPS may have affected the TIPS liquidity premium through two opposite channels. First, the liquidity channel implies that the QE2 programme may have reduced the liquidity premia required by investors to buy TIPS, as the Federal Reserve’s purchases may have made it less costly for investors to sell TIPS. However, this effect of the QE2 programme on the TIPS liquidity premium depends on the Federal Reserve’s flow of purchases and, therefore, it should have been limited to the duration of the large-scale asset purchases programme. Second, the scarcity channel implies that the Federal Reserve’s TIPS purchases may have reduced the already scarce stock of TIPS available to investors, and this instead may have increased the liquidity premium in the TIPS market. The liquidity channel is closely related to the market functioning channel described by Gagnon et al. (2011) and it implies that QE can improve market functioning. The scarcity channel instead is related to the local supply channel of Vayanos and Vila (2009) and D’Amico et al. (2012), and it
Table 3: QE2 TIPS purchase Dates Amount Average (Mill.) Maturity 0 03-Nov- 1 23-Nov-10 $1,821 9. 2 08-Dec-10 $1,778 8. 3 21-Dec-10 $1,725 16. 4 04-Jan-11 $1,729 16. 5 18-Jan-11 $1,812 14. 6 01-Feb-11 $1,831 13. 7 14-Feb-11 $1,589 14. 8 04-Mar-11 $1,589 11. 9 18-Mar-11 $1,653 17. 10 29-Mar-11 $1,640 18. 11 20-Apr-11 $1,729 23. 12 04-May-11 $1,679 13. 13 16-May-11 $1,660 20. 14 07-Jun-11 $1,589 14. 15 17-Jun-11 $2,129 5. Average $1,730 14. Note: this table reports the QE2 TIPS pur- chase operations dates along with the amount (in millions) and the (weighted) average ma- turity.
implies that the purchase by the Federal Reserve of assets with a specific maturity leads to lower yields of assets with similar maturities. However, in the presence of a liquidity premium component, this mechanism implies a higher liquidity premium for the purchased assets. This trade-off with liquidity is not considered in the local supply channel, as noted by Ferdinandusse et al. (2017), and it could represent an important risk of asset purchase policies. Table 3 contains the exact dates of the TIPS purchases along with the amount and the average maturity. As a preliminary assessment of the effect of the QE2 program on the liquidity in the TIPS market, we report in Table 4 the cumulative responses of the estimated liquidity premium in the TIPS market around the days of the QE2 announcement (Nov. 3, 2010) and operations (from Nov. 23, 2010 to Jun. 17, 2011) over different time windows (from one day to five days change). The table shows that on the day of the QE2 program announcement the estimated TIPS
Table 4: Cumulative responses of the estimated TIPS liquidity premium around QE2 events 1 2 3 4 5 0 03-Nov-10 -0.262 0.082 0.172 0.108 -0. 1 23-Nov-10 0.191 -0.306 0.163 0.146 0. 2 08-Dec-10 -0.272 -0.006 -0.420 -0.287 -0. 3 21-Dec-10 -0.081 -0.193 -0.297 -0.092 -0. 4 04-Jan-11 0.168 -0.199 -0.234 0.057 0. 5 18-Jan-11 0.069 0.298 0.083 0.317 0. 6 01-Feb-11 -0.235 -0.457 -0.856 -1.141 -1. 7 14-Feb-11 0.223 0.087 0.037 0.119 0. 8 04-Mar-11 0.284 0.310 0.199 0.505 0. 9 18-Mar-11 -0.042 0.021 -0.020 -0.126 -0. 10 29-Mar-11 -0.053 -0.101 -0.174 -0.227 -0. 11 20-Apr-11 -0.137 -0.144 0.180 0.145 0. 12 04-May-11 0.146 0.337 0.294 0.241 0. 13 16-May-11 -0.079 0.044 -0.137 -0.121 -0. 14 07-Jun-11 0.043 -0.039 -0.305 -0.163 -0. 15 17-Jun-11 -0.096 -0.214 -0.264 -0.475 -0. Mean -0.008 -0.030 -0.099 -0.062 -0. Median -0.047 -0.022 -0.078 -0.017 -0. Note: this table reports the cumulative changes in the estimated TIPS liquidity premium factor around the days of the QE2 an- nouncement (Nov. 3, 2010) and operations (from Nov. 23, 2010 to Jun. 17, 2011) over different window sizes (from one day change to five days change). The changes in the estimated TIPS liquidity factor are computed from the day before the event, i.e. the two day change for the QE2 announcement is the difference in the es- timated TIPS liquidity premium factor between November 4, 2010 and November 2, 2010.
liquidity premium declined. In addition, on eleven out of fifteen TIPS operation dates of the QE program, the estimated liquidity premium in the TIPS market declined either the same day or the following day. The average and median of the changes of the liquidity factor on the QE2 dates are negative, indicating that the program may have lowered the liquidity premium required by market participants in order to invest in TIPS. To formally assess the impact of the QE2 program on the liquidity premium in the TIPS market, in the next section we perform a counterfactual analysis.
