Un intento fallido: La propuesta de Gustave Moynier

In order to understand the manner in which our monetary policy shocks affect global flows and valuations, we first need to better understand the nature of the revisions in expectations housed

11_{Although a number of contracts satisfy the identifying assumptions needed to extract a surprise, we converge on}

five-year bond futures for a number of reasons, the first being precedence in the literature (see Rogers, Scotti and Wright 2014). Furthermore, using the five-year bond futures allows for us to extract a monetary policy shock that exhibits variation in both the conventional and unconventional monetary policy periods, where the latter period it accounts for the explicit targeting of long-term interest rates by the Federal Reserve.

12_{For robustness, we examine a variety of monetary policy shocks extracted from different futures instruments such}

in our variable. As mentioned, monetary policy potentially influences both the expected path of short-term interest rates and the term premium. However, from mid-2008 until as recently as mid-2015, the Fed was not expected to deviate from zero short-term interest rates; it is not unreasonable, therefore, to suspect that monetary policy is qualitatively different in the periods of QE and of LSAP tapering in the sense that the relationship between monetary policy and the term structure of interest rates is altered. In this section of the paper, we explore the relationship between our monetary policy surprise measure and the decomposition of the yield curve into a component associated with the expected path of the short interest rate and that associated with the term premium. This disaggregation permits an evaluation of the role for monetary policy surprises across the conventional and unconventional periods.

We appeal to a well-established affine term structure methodology from Kim and Wright (2005) that permits the decomposition of various government bond yields into information about future short rates and term premia. Kim and Wright estimate a standard latent three-factor Gaussian term structure model using zero-coupon Treasury yields from the Grkaynak, Sack, and Wright (GSW, 2007) database. To facilitate empirical implementation, forecast data on the three-month T-bill yield from Blue Chip Financial Forecasts are incorporated into the model estimation. Their model yields a point-in-time daily estimate of the expected short rate over the life of any longer dated bond as well as the risk compensation market participants require for holding that bond.13

To use these components, we separately regress changes (1) in bond yields, (2) in the expected path of the short rate, and (3) in the term premium onto our monetary policy shock to assess its importance on each. We conduct these separate regressions for one, five and ten-year maturity U.S. Treasury bonds, separating the MP shock effects of interest (those arising on relevant FOMC or policy announcement days) across three periods (pre-crisis, March 1994 August 2008, QE, December 2008 April 2013, and tapering, May 2013 June 2016):14

13_{The Kim and Wright yield curve decomposition data are made available at}

http://www.federalreserve.gov/pubs/feds/2005/200533/200533abs.html

14_{The pre-crisis dummy is equal to one in the period March 1994 to July 2008 and zero otherwise. The QE dummy}

∆Y(n),t =α+β1dummypreM Pt+β2dummyqeM Pt+β3dummytaperM Pt+t (2.2)

Where the left-hand side variable Y(n),t is the zero coupon bond yield on an n-year bond, the expectations hypothesis-implied average short rate component of an n-year bond, or the term pre- mium on an n-year bond. We consider the daily change in the dependent variables for event days perfectly coinciding with the day of the MP shock.15 We provide White standard errors to correct for heteroskedasticity (in parentheses). Finally, we allow the monetary policy shock coefficients to vary across the three periods.

Table B.3a (Panel A) shows the daily regressions for the overall yield change, the change in the path of the expected short rate, and the change in the term premium for the Kim and Wright model (2005), and Panel B shows the same daily yield change regressions for the same event dates using the Adrian, Crump and Moench (2013) model.16 First, focusing on the coefficients associated with the conventional, pre-crisis period, we find that a positive FF5 shock is significantly associated with bond yield changes across maturities both for the one-day event (Panel A) though this effect appears to diminish with the 10-year bond across both event windows. To provide a sense of the economic magnitude, a one-standard deviation FF5 shock would be associated, on average, with a 4.75 basis point increase in the ten-year bond yield across the one-day event window over the

May 2013 when Ben Bernanke first mentioned the possibility of tapering LSAP purchases. The period between the collapse of Lehman Brothers and the beginning of QE was marked by global “flight to safety“ and its inclusion in either neighboring sub-period contaminates the analysis in that it is a period of extraordinary uncertainty and cannot clearly be classified as the QE or the conventional periods. To ensure that we isolate the “flight to safety“ flows, we include in our robustness checks and specification wherein the QE dummy begins on March 2009, as March 2009 marks the month during which equity markets in the United States began their recovery. While the quantitative effects for the emerging equity markets and exchange rates are smaller, we continue to observe statistical and economic significance. Results are available upon request.

15_{We also considered regressions based on a three-day event window. While there still appears to be an important}

role for shocks housed in the Treasury futures contracts, the effects on bond yield and their yield components do start to diminish by day three.

16_{To capture any relevant slow-moving market effects, we also estimated both models for two-day windows finding}

conventional policy period. For comparison, a one-standard deviation daily ten-year bond yield change is 5.76 basis points over this period.

Next, we decompose the overall yield changes into changes in the expected path of the short rate and in the relevant term premia. The regression results suggest that over the conventional mon- etary policy period, FF5 shocks play a role in altering the expected path of the short rate. As one might anticipate, this effect diminishes sharply with maturity across both measurement windows. We also uncover an important role for FF5 shocks in altering term premia, where the risk com- pensation effects are relatively stable across different maturities following a hump-shaped pattern that peaks at the five-year horizon declining thereafter. In sum, during the period of conventional monetary policy, our measured FF5 shock has implications for both revisions in expectations about the path of future short rates as well as risk premia.

We now turn to the coefficients associated with FF5 shocks during the period of unconventional monetary policy (both QE and eventual policy tapering).17 _{First, overall yield changes appear to}

be significantly affected by FF5 shocks across both the QE and tapering periods for both the one and two-day event windows. As an example, a one- standard deviation FF5 shock during the QE period would be associated, on average, with a 11.7 basis point increase in the ten-year bond yield across the one-day event window. For comparison, a one-standard deviation daily ten-year bond yield change is 7.10 basis points over the QE period. Similarly, sizeable effects are present during the tapering period.

One interesting point to note is the fact that, unlike the conventional period, the FF5 shock effects on bond yields during the unconventional periods monotonically increase over time, i.e., the hump-shaped pattern documented for the pre-crisis period is no longer in evidence. Since the effect of a shock on the expected path of future short rates is likely to berelativelyshort-lived over the life of a long-term maturity bond, we can speculate on the manner in which the FF5 shocks

17_{Although Kim and Wrights model does not account for the binding quality of the Zero Lower Bound (ZLB), a}

dynamic term structure model with the ZLB included will likely put more weight on the term premium for matching observed yields. Thus, results in the unconventional monetary policy period may very well understate the effect of shocks on the term premium.

map into revisions in the compensation for interest rate risk. Indeed, despite a role for the FF5 shocks altering the expected path of future short rates during the unconventional QE and tapering periods, the largest effects are associated with sizeable and statistically significant revisions in term premia. Further, these risk premia effects monotonically increase with maturity.

A potential concern with using term structure models during the unconventional monetary policy period is the fact that the decompositions lower effect on expected short rates during the zero lower bound period may simply be mechanical and result from the fact that nominal short rates cannot fall any further. To verify that our results are not being driven by our choice of term structure model, Panel B of Table B.3a repeats the exercise using an alternative, but also widely used term structure model from Adrian, Crump and Moench (2013). It is reassuring that the use of an alternative term structure model to decompose the impact of the monetary policy shocks on yields into the expected short rate and the term premia yield very similar patterns. The results are also similar both in magnitudes and across sub-periods. The impact of our monetary policy shock on the expected path of the short rates follows a hump-shaped pattern across maturities, rising from the one-year to the five-year maturities and then declines from the five-year to ten- year maturity. However, we see the magnitudes of the coefficients on the term premia increasing across maturity during the quantitative easing period. Given that this sub-period was characterized by falling yields, the coefficient estimates suggest that the fall in the term premia was greater for longer duration bonds consistent with falling duration risk.

Nakamura and Steisson (2018) suggest that movements in risk premia at the time of FOMC announcements do not play an important role in explaining longer-term real interest rates and that the expectations hypothesis of the term structure is a good approximation in response to monetary shocks extracted using actual yield changes. In contrast, we find that adjustments to the risk premia are an important source of movements in longer-term yields using our measure of monetary policy shocks extracted from futures prices. We conjecture that separately estimating the impact of monetary policy shocks on the two yield components across the pre-crisis conventional, and post-crisis unconventional periods appears to be an important distinction that allows us to discern

the impact of these shocks on risk-premia during the unconventional period. Estimating the model for the full sample may obscure the term-premia effects by averaging them over time. In other words, it is important to consider the differences across monetary policy regimes to observe the time varying effects on the term premia.

Taken together, during the period of unconventional monetary policy, our measured FF5 shock appears to have a greater influence on the variation in the required risk compensation relative to variation in the expected path of the short rate. This result holds despite the fact that high frequency variation in futures contracts are employed in the construction of the FF5 shocks in the first place. The important role for FF5 shocks in describing variation in risk compensation during the period of post-crisis unconventional policy may help us better interpret the manner in which our measured monetary policy shocks affect global flows and valuations in the sections to follow.