«Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Aﬀairs Federal Reserve Board, Washington, D.C. The ...»
Finance and Economics Discussion Series
Divisions of Research & Statistics and Monetary Aﬀairs
Federal Reserve Board, Washington, D.C.
The High-Frequency Impact of News on Long-Term Yields and
Forward Rates: Is It Real?
Meredith J. Beechey and Jonathan H. Wright
NOTE: Staﬀ working papers in the Finance and Economics Discussion Series (FEDS) are preliminary
materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth
are those of the authors and do not indicate concurrence by other members of the research staﬀ or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.
The High-Frequency Impact of News on Long-Term Yields and Forward Rates:
Is it Real?
Meredith J. Beechey* and Jonathan H. Wright* Abstract This paper uses high-frequency intradaily data to estimate the effects of macroeconomic news announcements on yields and forward rates on nominal and index-linked bonds, and on inflation compensation. To our knowledge, it is the first study in the macro announcements literature to use intradaily real yield data, which allow us to parse the effects of news announcements on real rates and inflation compensation far more precisely than we can using daily data. Long-term nominal yields and forward rates are very sensitive to macroeconomic news announcements. We find that inflation compensation is sensitive to announcements about price indices and monetary policy. However, for news announcements about real economic activity, such as nonfarm payrolls, the vast majority of the sensitivity is concentrated in real rates. Accordingly, we conclude that most of the sizeable impact of news about real economic activity on the nominal term structure of interest rates represents changes in expected future real short-term interest rates and/or real risk premia rather than changes in expected future inflation and/or inflation risk premia. This suggests that explanations for the puzzling sensitivity of long-term nominal rates need to look beyond just inflation expectations and toward models that encompass uncertainty about the long-run real rate of interest.
Keywords: intradaily data, news announcements, inflat
* Federal Reserve Board, Washington DC 20551; firstname.lastname@example.org, email@example.com and firstname.lastname@example.org. We are grateful to Joseph Gagnon, Andy Levin and Clara Vega for helpful comments on an earlier draft. The views expressed in this paper are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve System or of any other employee of the Federal Reserve System.
1. Introduction It has long been recognized that the high-frequency reaction of asset prices to macroeconomic news announcements represents a rich source of information about the effects of economic fundamentals on financial markets. Economic theory―and much empirical evidence―indicates that macroeconomic news announcements lead to sudden jumps in the conditional means of asset prices. Studying the relationship between these jumps and the surprise components of the announcements may be the closest thing that we get in macroeconomics to a natural experiment.
In this work, high-frequency data is especially valuable. In a short enough window around a news announcement, the data release should be the only information hitting the market. But macroeconomic news announcements explain only a modest share of the total variation in most asset prices, and so in longer windows the relationship between news announcements and asset prices will be clouded by the impact of other public or private information. For example, exchange rates are notoriously hard to relate to any macroeconomic fundamentals. Nonetheless, Andersen, Bollerslev, Diebold and Vega (2003) find that in short intradaily windows around macroeconomic news announcements, there is a systematic relationship between macroeconomic surprises and exchange rate changes. This relationship cannot be detected at lower frequencies, because macroeconomic announcements explain only a small share of the total variability of exchange rates.
The literature on the reaction of financial markets to macroeconomic news is vast, and we make no attempt to survey it. Many authors have studied the effects of news announcements on the term structure of interest rates, including Fleming and Remolona (1999), Gürkaynak, Sack and Swanson (2005a), and Goldberg and Klein (2005). A striking finding of these papers is that long-term nominal yields and forward rates are remarkably sensitive to incoming data. The release of an employment report can sometimes cause ten-year yields and forward rates to jump over 10 basis points. Nominal yields and forward rates embody inflation expectations, expected future real interest rates, and term premia. Gürkaynak, Sack and Swanson (2005a) assume that news announcements do not affect long-term expectations of future real short-term interest rates or distant-horizon term premia. Hence, they interpret the sensitivity of far-horizon nominal forward rates to news announcements as investors revising their long-run inflation expectations in light of today’s data, indicating that inflation expectations are very poorly anchored.
However, the sensitivity of nominal yields and forward rates to news surprises may owe to more than just inflation expectations: revisions to expected future real short-term interest rates, or to nominal term premia (in turn the sum of inflation risk premia and real term premia) are also candidates, and there are not strong reasons to rule them out a priori. Beechey (2007) used a three-factor affine term structure model to decompose the variation in nominal forward rates in the wake of news announcements into nominal term premia and expectations of future nominal short-term interest rates, and argued that most of the reaction at long horizons was in nominal term premia.
This paper employs a decomposition of the reaction of nominal yields to news into real and inflation components. This decomposition is implemented using a new dataset of intradaily quotes on Treasury Inflation Protected Securities (TIPS)―debt securities for which the coupon and principal payments are indexed to the Consumer Price Index (CPI)—in combination with intradaily data on nominal yields. We regress intradaily changes in nominal yields, real yields, and the spread between these two―inflation compensation―on the surprise component of macroeconomic news announcements. Several recent papers, among them Gürkaynak, Levin and Swanson (2006) and Beechey, Johannsen and Levin (2008), have employed daily-frequency inflation compensation to estimate news announcement regressions for the United States and other countries, highlighting the greater sensitivity of U.S. forward rates of inflation compensation to various types of macroeconomic news announcements.
We group macroeconomic announcements into three broad types: news about prices, news about the real side of the economy (indicators of the labor market, spending, production and sentiment) and news about monetary policy. Armed with intradaily data, we find that different types of news have quite different effects on real rates and rates of inflation compensation. To summarize, price data mainly affect inflation compensation, real-side data have a large impact on real yields and forward rates―but no measurable effect on forward rates of inflation compensation—and surprise tightenings of monetary policy push real rates up but inflation compensation down. We demonstrate the efficiency gains from running these regressions with intradaily data, rather than data at the daily frequency. That news buffets forward rates of inflation compensation is consistent with the conclusions of others that either long-run inflation expectations or inflation risk premia, or both, are sensitive to news—but the sensitivity appears to be confined to news about prices and monetary policy. Instead, our findings point to a great deal of sensitivity of the real component of long-run interest rates, a possibility that has largely been overlooked in the recent news-announcement literature. This in turn suggests that the puzzling sensitivity of nominal forward rates lurks not just in the inflation component but, importantly, in uncertainty about the long-run real rate of interest.
The plan for the remainder of this paper is as follows. In Section 2, we introduce the data. Section 3 contains the main results, and Section 4 provides robustness checks and comparison with results in the existing literature. Section 5 gives discussion of the economic interpretation of our results, arguing that the degree of sensitivity of long-term real rates to macroeconomic news seems puzzling and hard to rationalize within existing macroeconomic models. Section 6 offers some concluding thoughts.
2. The Data This section briefly describes data sources and construction. We consider fourteen macroeconomic announcements, including the surprise element of FOMC decisions about the target Federal Funds rate. Table 1 lists the announcements, their frequency and release times.
For all announcements other than the target funds rate, the surprise component of each announcement is measured as the real-time actual value less the median expectation from the survey conducted by Money Market Services (MMS) on the previous Friday. For the target Fed Funds rate, we measure the surprise component of the Fed’s decision using intradaily changes in Federal Funds futures prices, following the methodology of Kuttner (2001). We sign the surprises such that positive surprises represent stronger-than-expected growth or higher-thanexpected inflation: for the unemployment rate and initial jobless claims, which are both countercyclical indicators, we flip the sign of the surprise so that positive surprises reflect stronger-than-expected growth for these as well. The surprises are then standardized by their respective standard deviations. Letting Aj,t denote the released value of an announcement of type j at time t, and E j,t denote the ex-ante expectation of this release, the surprise is defined as
where σ (.) denotes the standard deviation. This transformation has the effect of making the units comparable across different types of announcements. Table 1 also lists the standard deviation of the surprises for each announcement type.
We construct a high-frequency dataset consisting of five-minute quotes of yields-tomaturity on nominal Treasury securities and TIPS. We choose securities with around five and ten years remaining time to maturity. The specific bonds used are listed in Table 2, and are generally the most recently issued “on-the-run” five- and ten-year nominal Treasury notes and TIPS.1 The data are available from February 17, 2004 to June 13, 2008. From these, we can use the approximation of Shiller, Campbell and Schoenholtz (1983) to construct five-to-ten year forward nominal and real interest rates.2 And, we can construct measures of five-year, ten-year, and five-to-ten-year forward inflation compensation, defined as the spreads between nominal and comparable-maturity real yields.3 On-the-run nominal securities command a liquidity premium, and so trade at lower yields than their off-the-run counterparts. However, in the narrow windows around news announcements, the relative liquidity of on- and off-the-run bonds is unlikely to change. Likewise, a liquidity premium tends to push up TIPS yields relative to their nominal counterparts; but again, this liquidity premium should difference out under the plausible assumption that macroeconomic news does not affect the relative liquidity of index-linked and nominal securities.
An on-the-run security is the most recently issued bond of a given maturity.
The approximation is that if yn denotes the yield on a bond with a maturity of n years and duration Dn, then the five-to-ten-year forward rate is approximately ( D10 y10 − D5 y5 ) /( D10 − D5 ).
All yields are expressed with semiannual compounding, and the calculation of the spread takes appropriate account of this. That is, if yn and yn denote the nominal and real yield in percentage points, inflation compensation is nom real
To compare our results with those from daily data, we consider two alternative measures of nominal and real yields. The first consists of quotes on the same specific five- and ten-year nominal and TIPS securities recorded at 4pm each day from our intradaily dataset. The second consists of the five- and ten-year par nominal and TIPS yields from the smoothed fitted yield curves estimated late each afternoon by Federal Reserve staff using only off-the-run securities, as described by Gürkaynak, Sack and Wright (2007, 2008).
3. Regressions and Results As a preliminary illustration of how macroeconomic news affects nominal and index-linked bond yields, Figure 1 plots the ten-year nominal and TIPS yields from 8:15am to 9:30am on the days of the two largest employment report surprises in our sample: April 2, 2004, when total payroll growth was 208,000 above survey expectations, and August 6, 2004, when payrolls came in 193,000 below expectations.
As expected in an efficient market, yields jumped up on the stronger-than-expected data and down on the weaker-than-expected data, with the reactions complete within ten minutes.
Such rapid reaction of nominal interest rates is well documented in the news-announcement literature. But what has not been documented before is that TIPS yields move just as rapidly.
The magnitude of the jump in TIPS yields on surprise days was similar to―but a bit smaller than―that in nominal yields. Having illustrated the point with these two large data surprises, we now turn to a more systematic analysis of the relationship between surprises and high-frequency yield changes.
3.1 Regressions of Changes in Yields and Forward Rates on Surprises
The regressions that we consider are of the form: