The Response of Asset Prices to Unconventional Monetary Policy

Transcription

1 The Response of Asset Prices to Unconventional Monetary Policy Alexander Kurov and Raluca Stan * Abstract This paper investigates the impact of US unconventional monetary policy on asset prices at the zero lower bound (ZLB), and examines the number of factors that would be necessary to adequately characterize such impacts. The results indicate that two factors, identified as the future path of the monetary policy factor and the quantitative easing (QE) factor would capture these effects properly. After the ZLB was reached in December 2008, the traditional target factor identified by Gu rkaynak, Sack and Swanson (2005) has dissipated, while the path factor of future monetary policy has becomes the main dimension of the unconventional monetary policy. In addition to the path factor, the QE factor seems to amplify the reaction of medium and long-term yields to unconventional monetary policy. We investigate the effects of unconventional monetary policy on bond yields and stock returns using intraday futures data from December 16, 2008 until September 17, The results show that the path factor itself explains between 82% and 97% of the response of short-term and medium-term interest rates to monetary policy, but the magnitude of this impact declines with longer maturity. On the other hand, the QE factor can explain up to 34% of the response of medium and long-term interest rates to monetary policy, in addition to the variation explained by the path factor. The QE factor does not affect short-term interest rates (such as the Eurodollar futures rates with 4-5, 8-9 quarters to expiration), but instead amplifies the impact of forward guidance on medium and long-term interest rates (5-year, 10-year and 30-year Treasury yields). The results also indicate that the stock market responds differently to the path and the QE factors. More precisely, the path factor has a significant negative impact on the E-mini S&P 500 futures returns, while the QE factor impact on the same variable is insignificant. * Department of Finance, College of Business and Economics, West Virginia University, P.O. Box 6025, Morgantown, WV 26506, Tel: , Raluca.Stan@mail.wvu.edu. Any errors or omissions are our responsibility. Alexander Kurov is an associate professor of finance in the Department of Finance, West Virginia University, Morgantown, West Virginia Raluca Stan is a doctoral student in finance in the Department of Finance, West Virginia University, Morgantown, West Virginia.

2 1. Introduction Under conventional monetary policy, the central bank can stimulate the economy by buying short-term government bonds and thus decreasing the short-term interest rates. Previous research shows that under conventional monetary policy, the reaction of asset prices to monetary policy is well captured by two factors: the current federal funds rate target and the future path of policy factor (Gu rkaynak, Sack and Swanson, 2005). More precisely, the first factor can be regarded as a measure of federal funds target surprises, while the second factor captures the aspects of the FOMC announcements that move futures rates for the upcoming year without changing the current federal funds rate. However, when the short-term interest rates reach the zero lower bound (ZLB), as happened in December 2008 in the US, the conventional monetary policy becomes ineffective, and the current federal funds rate target factor previously mentioned is equal to zero. At the ZLB, forward guidance is still implemented, and in fact becomes the leading factor capturing the impact of monetary policy on asset prices. If the ZLB is not binding simultaneously on intermediate-yields, the Fed can still influence people s expectations about future short-term rates through guidance about future monetary policy. At the ZLB, the central bank can also purchase bonds with longer-maturity in order to reduce long-term interest rates and stimulate the economy. This program is known as quantitative easing (QE). The aim of this paper is to investigate how many factors would be necessary to adequately describe the impact of unconventional monetary policy on asset prices at the ZLB. The analysis considers the unconventional monetary policy implemented by the Federal Reserve in the US after the ZLB was reached in December Such study would help us understand whether the Fed officials effort to provide information about the future path of policy through

3 forward guidance is efficient or not. That is, does the path factor have a significant impact on financial markets under unconventional monetary policy, and is it the main policy tool of the Federal Reserve at the ZLB? The results in this paper indicate that the path factor is indeed the main policy tool through which the unconventional monetary policy affects the markets. More precisely, the path factor itself explains between 82% and 97% of the response of short-term and medium-term interest rates to monetary policy. However, the magnitude of this impact declines for longer-term yields. On the other hand, the QE factor can explain up to 34% of the response of medium and long-term interest rates to monetary policy, in addition to the variation explained by the path factor. The QE factor does not affect shorter-term interest rates, such as the Eurodollar futures rates with 4-5, 8-9 quarters to expiration, but instead it amplifies the impact of forward guidance on the 5-year, 10-year and 30-year Treasury yields. The results also indicate that the two dimensions of the unconventional monetary policy have different impacts on the stock market: the path factor has a significant negative impact on the E-mini S&P futures return, while the QE factor has an insignificant impact on the same variable. Several papers mention that QE could be regarded as a signal confirming the FOMC s commitment to its interest rate guidance (Woodford (2012), Smith and Becker (2015)). In other words, QE can be thought as an additional dimension to forward guidance, through which unconventional monetary policy impacts the markets. This paper aims to contribute to the literature by providing evidence that the effect of unconventional monetary policy on the term structure of interest rates can adequately be described by two factors: the path of the future monetary policy and the quantitative easing.

4 2. Related Literature and the Proposed Research Question 2.1.Market Reaction to Quantitative Easing Quantitative easing is used by central banks when their policy rates approach the zero lower bound, and is implemented in order to stimulate the economy and ease financial conditions, by putting downward pressure on the longer-term yields. Several papers investigate the impact of QE on asset prices, and find that QE leads to lower long-term yields. For example, Christensen and Rudebush (2012) find that the QE announcements of the Federal Reserve and the Bank of England lead to a decline in the government bond yields. Bernanke, Reinhart and Sack (2004) investigate the QE in Japan and show that the Japanese yields were around 50 basis points lower than expected during QE. Wright (2012) finds that unconventional monetary policy has a significant impact on the stock, bond and foreign exchange markets. Krishnamurthy and Vissing-Jorgensen (2011) study the effect of Fed s purchase of long-term Treasuries and other long-term bonds on interest rates. They find that focusing on Treasury rates as a policy target might be inappropriate since QE works through several channels, and affects particular assets differently. They find evidence for a signaling channel, a unique demand for long-term safe assets, and an inflation channel for both QE1 and QE2. Some papers see the QE as a signal confirming the FOMC s commitment to its interest rate guidance (Woodford (2012), Smith and Becker (2015)). Therefore, the QE can be thought as an additional dimension to forward guidance, through which unconventional monetary policy impacts the markets. This seems realistic because many forward guidance statements are also accompanied by changes in the Federal Reserve s balance sheet.

5 The reaction of Treasury yields to forward guidance can be amplified by QE through two channels: the signaling theory channel and the portfolio balance channel. Following Bauer and Rudebush (2013), Treasury yields can be decomposed into a term premium 1 and a riskneutral rate 2. Therefore, the impact of QE on Treasury yields can happen through its impact on both these components. That is, QE can decrease long-term Treasury yields, by decreasing the term premium by reducing the amount of longer term bonds in the private-sector portfolios (the portfolio balance channel). Also, QE can reduce yields by affecting the risk-neutral component: by signaling its commitment to maintain exceptionally low levels of the target federal funds rate in the future (the signaling channel). This paper aims to contribute to the literature by providing evidence that the QE and the path factors are two main dimensions through which the unconventional monetary policy affects the term structure of interest rates. Our results indicate that the QE factor does not affect short-term interest rates, but has a positive significant impact on longer-term yields (5, 10 and 30-year Treasury yields). 2.2.The Effects of FOMC Forward Guidance under Unconventional Monetary Policy Expectations about future monetary policy influence the way markets react to the actual implementation of monetary policy, conventional or unconventional. To that extent, Fed officials have tried to provide information about the future path of monetary policy both before and after December Although forward guidance was implemented under conventional monetary policy as well, the guidance policy implemented at the ZLB (i.e. after December 2008) contains a few main changes: it specifies a length of time for which the target federal funds rate is 1 The term premium contains a maturity specific term-premium (reflecting the pricing of interest risk), and an idiosyncratic instrument-specific term premium (reflecting the demand and supply imbalances of the security). 2 The average level of short-term interest rates over the maturity of the bond.

6 expected to remain exceptionally low, it can include guidance based on future economic conditions (by actually providing a target upper limit for the unemployment rate- from December 2012 until March 2014) to name just a few. On January 25 th, 2012, the FOMC started publishing its participants projections of the target federal funds rate for up to three years ahead. However, given all the effort to provide information about the future path of policy, at the end, is this forward guidance efficient? That is, does it have a significant impact on financial markets, and is it the main policy tool of the Fed under unconventional policy? Smith and Becker (2015) show that forward guidance implemented by the Fed under unconventional monetary policy (since 2008) has had similar effects on the economy as changing the target federal funds rate under conventional monetary policy. More precisely, they find that signaling that the federal funds rate would remain lower in the future than previously expected leads to an increase in employment and prices. Thus, forward guidance is thought to function through an interest rate channel, without requiring an actual change in the current target federal funds rate. When the FOMC statements indicate that the policy rate is likely to remain low for an extended period of time, that will decrease both components of long-term rates: the term premium and the expected path of future interest rates. However, the impact of the forward guidance on the economy depends on how the public interprets this forward guidance. This paper aims to contribute to this literature by providing evidence that the path factor is the main policy tool through which unconventional monetary policy can affect asset prices (together with the QE factor). Our results indicate that the path factor has a significant positive impact on the term structure of interest rates, but the magnitude of this impact declines for longer maturities.

7 2.3. The Dimensions of Conventional vs. Unconventional Monetary Policy Gu rkaynak, Sack and Swanson (2005) (GSS from now on) study the effects of US conventional monetary policy on asset prices from 1990 until 2004, and conclude that these effects are well captured by two factors: a current federal funds rate target factor and a future path of policy factor. Their results suggest that both factors have important and different effects on asset prices, with the future policy path factor having a bigger impact on the longer-term Treasury yields. In this paper we investigate a similar question to the one proposed by GSS (2005), but in presence of unconventional monetary policy. More precisely, we investigate the factors necessary to capture the monetary policy implemented by the Federal Reserve in the US, when the US economy is at the zero lower bound. The focus on the post- December 2008 period is helpful for understanding the difference in the way monetary policy is implemented, and through which factors monetary policy affects the markets the most, under the conventional vs. the unconventional scenario. The post-2008 period is unique for such analysis, since it is a period when the target federal funds rate is at the zero lower bound, and the central bank can no longer stimulate the economy through the GSS target factor. Then, is the GSS path factor the only dimension characterizing the US monetary policy after December 2008, or is there an additional dimension through which the Fed can stimulate the markets? One potential additional dimension could be a quantitative easing (QE) factor that would capture the impact of monetary policy on longer-term interest rates. The analysis in this paper tries to answer the following question: Research Question: How many factors would be necessary to adequately characterize the unconventional monetary policy implemented by the Federal Reserve in the United States since December 2008?

8 3. Data 3.1.Monetary Policy Announcements For studying the research question above, the analysis is performed over a sample period characterized by unconventional monetary policy in the US. More precisely, we put together a list of dates and times of monetary policy announcements from December 16, 2008 until September 17, The dates and time of the announcements are particularly those of the FOMC press releases after the FOMC meetings and every change in the policy. We use five-minute futures data for the E-mini S&P 500, Eurodollar futures with 4-5, 8-9 and quarters to expiration, and 5-year, 10-year and 30-year Treasury futures. The analysis consists of an event-study methodology, where we use the continuously compounded return of the E-mini S&P 500 futures, the change in the rate for the Eurodollar futures and the change in the Treasury yields (of different maturities), from 10 minutes before each FOMC announcement to 1 hour and 50 minutes afterwards Methodology and Empirical Results How many Factors adequately describe the Unconventional US Monetary Policy? The question that we try to answer is how many underlying latent factors are sufficient to capture the response of asset prices to monetary policy at the zero lower bound. To address this, we perform a similar variable reduction technique as in Gu rkaynak, Sack, and Swanson (2005). More precisely, we denote by X the T*n matrix, whose rows correspond to the monetary policy announcements, and the columns correspond to different asset prices. The elements of X denote 3 The event study methodology, and therefore the use of a window around each FOMC release is helpful for studying how asset prices respond to monetary policy. It reduces the possibility that the movements in asset prices are driven by other factors unrelated to monetary policy (no other macroeconomic announcement is released during the selected window).

9 a change in the corresponding asset price in a two hours window around the corresponding FOMC announcement (from 10 minutes before, to 1 hour and 50 minutes after the announcement). X can be written as: X = FΛ + η (*) Where F is the T*k matrix of latent factors (k<n), while Λ is the k*n matrix of factor loadings, and η is the T*n matrix of white noise disturbances. We want to identify the number of the dimensions that would be necessary to adequately describe X. To address this, we follow GSS (2005) and analyze the structure of the data in X, by using the matrix rank test of Cragg and Donald (1997) 4. We test the null hypothesis that X is described by k 0 common factors, against the alternative hypothesis that X is described by k > k 0 factors. The rank test measures the minimum distance between Cov(X) and the covariance matrices of all possible factor models in (*) with k 0 factors. After an appropriate normalization, this distance has a limiting χ 2 distribution with (n k 0 )(n k o + 1)/2 n degrees of freedom. The results from the Cragg and Donald (1997) test are summarized in Panels A and B of Table 1. Panel A contains the results when applying the test to Eurodollar futures rates with 4-5, 8-9, Quarters to expiration, Treasury yields with maturity 5-year, 10-year, and 30-year, and stock return (the returns of the E-mini S&P 500 futures). The results indicate that two factors are needed to characterize the response of asset prices to unconventional monetary policy. Similar results are observed in Panel B, where we apply the test to the same variables, excepting the stock returns. In fact, when looking at the response of the term structure of interest rates itself 4 We thank Eric Swanson for providing us the code for the Cragg and Donald (1997) test, code that he used in the paper Do Actions Speak Louder than Words? The Response of Asset Prices to Monetary Policy Actions and Statements, co-authored with Refet Gu rkaynak and Brian Sack.

10 to unconventional monetary policy (Panel B), it can be noticed that the hypotheses of 0 or 1 factor being sufficient to adequately explain the response of asset prices to monetary policy are rejected at the 10% significance level. In addition, the null hypothesis of 2 factors describing the asset prices response to policy announcements cannot be rejected at 10% significance level. [Insert Table 1 about here] An important consequence of this finding is that under unconventional monetary policy, the path factor itself is not sufficient to explain the response of asset prices to monetary policy announcements. In the next section, we try to estimate, understand, interpret this additional dimension of US monetary policy, and estimate its impact on stock returns and Treasury yields Estimation of the Two Factors For estimating the two factors, we perform a factor analysis to matrix X, and decompose the matrix X into a set of orthogonal factors F 1 and F 2. The first factor F 1 is a vector of length T with maximum explanatory power for X, while F 2 is another vector that has maximum explanatory power for the residuals of X, after projecting each column of F 1. Considering the results from the factor rank test, we focus on the first two estimated factors. We estimate F 1 and F 2 by using the Eurodollar futures rates with 4-5, 8-9 quarters to expiration, and the Treasury yields with maturity 5, 10 and 30 years. Focusing on the interest rates for estimating the factors makes the structural interpretation of the factors, and the interpretation of their effects on yields and stock returns more clear (similar to GSS (2005)). The estimates of the factors from using the stock returns and the interest rates are similar.

11 [Inset Table 2 about here] Although the two factors explain a maximal fraction of the variance of X, they do not have a structural interpretation. For facilitating the interpretation, one could perform oblique or orthogonal rotations in the factors. We start with performing a promax 5 oblique rotation of the two factors first, to gain simplicity in the interpretation of the factors (by relaxing the orthogonality constraint). The results reported in Table 2 indicate the individual and the nonredundant contribution of each factor in explaining each considered variable (Eurodollar futures rates with 4-5, 8-9 quarters to expiration (denoted as ed4 and ed8)), Treasury yields with maturity 5-year, 10-year, and 30-year (denoted as T5Y, T10Y, T30Y)). As noticed, the variables significantly loaded on the first factor are the short and medium-term interest rates (4-5 Quarters Eurodollar Futures rate, 8-9 Quarters Eurodollar Futures rate, 5-Year Treasury yield, and to some extent the 10-Year Treasury yield as well), while the variables significantly loaded on the second factor are the longer-term interest rates (30-Year Treasury yield, 10-Year Treasury yield, and to some extent 5-Year Treasury yield as well). The graphical representation of the two rotated factors provided in Figure 1 emphasizes better these results. That is, the 4-5 Quarters Eurodollar Futures rate and the 8-9 Quarters Eurodollar Futures rate are significantly loaded on the first factor. Then, yields with longer time to maturity (5, 10, or 30-Year yields) have a smaller loading on the first factor, and a greater loading on the second factor. [Insert Figure 1 about here] However, although this oblique rotation is helpful to visualize what variables the factors tend to explain the best, the analysis in the paper is based on an orthogonal rotation. Such rotation allows 5 The name of such rotation technique comes from the procrustean rotation, where two shapes are rotated around the origin until the sum of the squared distances among them is minimized.

12 for a more structural interpretation of the factors and allows the construction of convenient restrictions that would be helpful for interpretation. We perform a rotation of the factors F 1 and F 2 that yields new factors, which we call Z 1 and Z 2. We start by normalizing F 1 and F 2 to have zero mean and unit variance, and then we construct Z 1 and Z 2 to be orthogonal and explain the matrix X in the same way as F 1 and F 2. More precisely, we define a 55*2 6 matrix Z as: Z = [F 1 F 2 ] U, where U = [ α 1 β 1 α 2 β 2 ] is a 2*2 matrix identified by four restrictions that we impose. The rotation method is similar to GSS (2005), but the restrictions imposed to the matrix U are adjusted to this analysis. That is, first, the columns of matrix U are normalized to have unit length (so that Z 1 and Z 2 have unit variance 7 ). Then, the orthogonality condition of Z 1 and Z 2 requires that E(Z 1 Z 2 ) = α 1 β 1 + α 2 β 2 = 0. We also impose that Z 2 does not impact the Eurodollar futures rate with 8-9 quarters to expiration. If λ 1 and λ 2 denote the factor loadings of ed8 on F 1 and F 2, and because F 1 = α 2 Z 2 ] and F 2 = 1 (α 1 β 2 α 2 β 1 ) [β 2Z 1 1 (α 1 β 2 α 2 β 1 ) [ β 1Z 1 + α 1 Z 2 ], the last restriction is equivalent to ( α 2 λ 1 + α 1 λ 2 ) = 0. Due to this transformation, the change in the 2-year ahead Eurodollar futures rate can be seen as driven only by Z 1, while Z 2 captures all characteristics of the FOMC announcements that move longer term interest rates, without affecting the 2-year ahead rates. Z 1 is more related to the future path of the monetary policy, while Z 2 captures the effects on the longer-term yields. For this reason we will refer to Z 1 and Z 2 as the path and the QE factors, respectively. 6 Z is a 55 *2 matrix because there are 55 FOMC announcements in the sample period. The two columns of the matrix correspond to the two different factors. 7 These two restrictions are: α α 2 2 = 1 and β β 2 2 = 1.

13 In fact, when running simple OLS regressions of ed4 and ed8 on Z 1, the estimated slope coefficients are close to unity. To facilitate the interpretation of Z 1, we scale it so that a change of 0.01 in Z 1 corresponds to exactly 1 basis point change in the 8-9 quarters ahead Eurodollar futures rate. We follow the procedure in GSS (2005), and rescale Z 2 so that it has the same magnitude effect on the 10-Year treasury yield as Z 1 (namely 0.535, as reported later in Table 3) How do Asset Prices Respond to the Path and the QE Factors? Understanding the impact of these two dimensions of monetary policy on asset prices is important for central banks implementing unconventional monetary policy, because it can provide some insights about what policy channel can be the most effective in affecting asset prices. Also, understanding such impact can be useful for investors, who can adjust their investments appropriately, by correctly anticipating the impact of policy actions on asset prices. In this section we estimate the impact of the path and QE factors on asset prices. To proceed so, we have an estimate of the path factor (Z 1t ) and of the QE factor (Z 2t ) for each monetary policy announcement from December 16, 2008 until September 17, For each FOMC announcement, we compute Δy t as either the change in the Eurodollar futures rates with 4-5 or 8-9 quarters to expirations, the change in the Treasury yields with 5, 10 and 30 years to expiration, or the return of the E-mini S&P500 futures for a 2 hours window around each FOMC announcement (10 minutes before up to 1 hour and 50 minutes after each announcement). For estimating the impact of the path and the QE factors on asset prices, we run the regression in (2) and report the estimated results in Table 3: Δy t = α + β Z 1,t + γ Z 2,t + ε t, (2)

14 As expected from the close relationship between the path factor and the 8-9 quarters ahead Eurodollar futures rate, and also from the orthogonality of Z 1 with Z 2, the estimated coefficient for the path factor is one in the 8-9 quarters Eurodollar futures rate regressions with a single or with both identified factors. Panel A of Table 3 reports the estimated coefficients when regressing Δy t on the path factor only. The results indicate that the path factor has a significant and positive impact on the term structure of interest rates, but the impact declines in magnitude for longer term maturities. In addition, the path factor has a significant and negative impact on stock returns under unconventional monetary policy. Panel B of Table 3 takes into consideration the QE factor as well, and reports the estimated results from regressing Δy t on both the path and the QE factors. As noticed, the path factor still has a significant positive impact on the term structure of interest rates, and the magnitude of its impact declines for longer term maturities. In fact, the path factor does not have a significant effect on the longer-term yields, such as the 30-Year Treasury yield. On the other hand, monetary policy seems to be able to affect longer-term yields through the QE factor. The QE factor does not significantly affect short-term interest rates, but it does have a significant and positive impact on longer-term yields. Moreover, the magnitude of the QE impact on interest rates increases with maturity. Most of the variation in the Eurodollar rates, Treasury yields and the stock returns driven by monetary policy is explained by the path factor, as suggested by the reported R 2 in Table 3. The QE factor does not affect the short-term interest rates (by definition), but adds significant explanatory power for the changes in Treasury yields with 5, 10, and 30 years to maturity.

15 Regarding the stock market reaction to unconventional monetary policy, it seems that the stock market responds differently to the path and the QE factors. More precisely, the path factor has a significant negative impact on the E-mini S&P futures return, the negative impact being consistent with the findings of GSS (2005), who studied such relationship under conventional monetary policy. However, GSS find that the impact is insignificant under conventional policy, while this analysis indicates the impact becomes stronger under unconventional policy. The results indicate that the QE factor does not significantly affect the stock returns. This finding is a little surprising, given the positive impact of the QE factor on interest rates (medium and long-term). However, a potential explanation for this could be that a large positive (negative) realization of the QE factor could signal that the FOMC sees greater (lower) output, employment, inflation in the future than the market expects. Investors could then revise their expectations of future earnings, dividends. This would potentially mitigate/counteract the decline (rise) in stock prices associated with greater (lower) interest rates. 5. Conclusion When the short-term interest rates reach the zero lower bound (ZLB), and the economy can no longer be stimulated through conventional monetary policy, the central bank uses a combination of unconventional strategies to influence people s expectations about future short-term interest rates. This is still possible, if the ZLB is not binding simultaneously on the intermediate-yields. This paper investigates the factors that would be necessary to adequately capture the reaction of asset prices to unconventional monetary policy at the ZLB. The results indicate that at the ZLB the impact of monetary policy on asset prices is mainly captured by two factors: the

16 path factor of future policy and the quantitative easing factor. The path factor itself explains between 82% and 97% of the response of short-term and medium-term interest rates to monetary policy. However, the magnitude of this impact on yields declines with longer maturity. On the other hand, the quantitative easing factor can explain up to 34% of the response of medium and long-term interest rates to monetary policy, in addition to the variation explained by the path factor.

17 References Bauer, M., Rudebush, G., 2013, The signaling channel for Federal Reserve Bond Purchases. Federal Reserve Bank of San Francisco, Working paper. Bernanke, B., Reinhart, V., Sack, B., Monetary policy alternatives at the zero bound: An empirical assessment. Brookings Papers on Economic Activity, 35, Christensen, J., Rudebusch, G., The response of interest rates to U.S. and U.K. quantitative easing. Economic Journal, 122, Cragg, J., Donald, S., 1977, Inferring the rank of a matrix, Journal of Econometrics,76, Gürkaynak, R. S., Sack, B., Swanson, E., 2005, Do Actions Speak Louder Than words? The Response of Asset Prices to Monetary Policy Actions and Statements. International Journal of Central Banking, 1, Gürkaynak, R. S., Sack, B., Swanson, E., Market-based measures of monetary policy expectations. Journal of Business and Economic Statistics, 25, Krishnamurthy A., Vissing-Jorgensen, A., 2011, The effects of quantitative easing on interest rates: channels and implications for policy, NBER Working Paper. Smith, A., L., Becker, T., 2015, Has Forward Guidance Been Effective?, The Macro Bulletin. Woodford, M., 2012, Methods of Policy Accommodation at the Interest-Rate Lower Bound, Federal Reserve Bank of Kansas City, the Changing policy landscape: 2012 Jackson Hole Symposium Wright, J. H., What does monetary policy do to long-term interest rates at the zero lower bound? Economic Journal, 122,

18 Table 1. Tests of the Number of Factors characterizing the US Unconventional Monetary Policy H0: Number of Factors Equals: Panel A 4-5Q, 8-9Q Eurodollar rates, 5Y, 10Y, 30Y Treasury Yields and Stock Prices χ2 Degrees of p-value Freedom Wald Statistics Number of Obs. Panel B 4-5Q, 8-9Q Eurodollar rates, 5Y, 10Y, 30Y Treasury Yields χ2 Degrees of p-value Freedom Wald Statistics Number of Obs. This table reports the results from the Cragg and Donald test (1997), when testing the null hypothesis of N H0 factors against the alternative of N > N H0 factors. The sample period covered is from 12/16/2008 until 09/17/2015. Panel A contains results for the 4-5 Quarters, 8-9 Quarters Eurodollar rates, 5-year, 10-year, and 30-year Treasury yields, and stock prices. Panel B contains the results for the 4-5 Quarters, 8-9 Quarters Eurodollar rates, 5-year, 10-year, and 30-year Treasury yields.

19 Table 2. (Oblique) Rotated Factor Pattern (Standardized Regression Coefficients) Factor1 Factor2 4-5 Quarters Eurodollar FUT Rate Quarters Eurodollar FUT Rate Year Treasury Yield Year Treasury Yield Year Treasury Yield This table reports the factor pattern matrix that can be used to interpret the meaning of the factors. The reported numbers represent the individual and nonredundant contribution that each factor is making to predict each subset. The subsets significantly loaded on the first factor are the short and medium-term interest rates: 4-5 Quarters Eurodollar Futures rate, 8-9 Quarters Eurodollar Futures rate, 5-Year Treasury yield, and to some extent the 10-Year Treasury yield as well. The subsets significantly loaded on the second factor are longer-term interest rates: 30-Year Treasury yield, 10-Year Treasury yield, and to some extent 5-Year Treasury yield as well.

20 Table 3. Response of Asset Prices to Path and QE Factors (orthogonal factors) Panel A) One Factor Panel B) Two factors Intercept Path Factor R (t-stat) (t-stat) Intercept Path Factor QE Factor (t-stat) (t-stat) (t-stat) R Quarters Eurodollar FUT Rate *** 82.59% *** % (0.00) (7.96) (0.00) (7.94) (0.30) 8-9 Quarters Eurodollar FUT Rate *** 97.10% *** % (0.00) (23.99) (0.00) (18.38) (-0.04) 5-Year Treasury Yield *** 96.69% *** 0.253*** 99.78% (0.00) (18.01) (0.00) (53.87) (14.16) 10-Year Treasury Yield *** 85.97% *** 0.535*** 99.80% (0.00) (8.69) (0.00) (37.63) (31.52) 30-Year Treasury Yield *** 54.97% *** 88.80% (0.00) (5.38) (0.00) (1.27) (9.52) E-mini S&P FUT Return *** 36.19% *** % (0.00) (-4.81) (0.00) (-3.76) (1.49) This table summarizes the estimates for the following models: Δy t = α + β Z 1,t + ε t in Panel A, and Δy t = α + β Z 1,t + γ Z 2,t + ε t in Panel B. Z 1,t and Z 2,t represent the Path and the QE factors, while Δy t can be either the Eurodollar futures rate with 4-5 quarters or with 8-9 quarters to expiration, the Treasury yields with 5 years, 10 years or 30 years maturity, or the E-mini S&P500 futures return. The sample period covered is from 12/16/2008 until 09/17/2015. The regression is estimated using OLS with White (1980) heteroskedasticity consistent covariance matrix. *, **, *** denote significance at 10 percent, 5 percent, and 1 percent, respectively.

21 Figure 1. The Oblique Procrustes Rotation of the Factors This figure presents a graphical representation of an oblique promax rotation of the two factors. Factor 1 explains more of the changes in the 8-9 and 4-5 quarters-ahead Eurodollar futures rates (symbols ed8 and ed4 respectively) and somewhat of the changes in the 5-year Treasury yield (symbol T5Y) in response to monetary policy announcements. The changes in longer term interest rates, such as the 10-year and 30- year Treasury yields (symbols T10Y, T30Y) are associated with Factor 2.