Macroeconomic Announcements and the Distribution of Price‐Endings in the U.S. Treasury Market

• Date: 01 February 2017
• DOI: http://doi.org/10.1111/fire.12118
• Published date: 01 February 2017

The Financial Review 52 (2017) 69–100

Macroeconomic Announcements and the

Distribution of Price-Endings in the U.S.

Treasury Market

Andrei Nikiforov∗and Eugene Pilotte

Rutgers University

Abstract

We investigate how new information impacts quote clustering in the bond market. We

find that clustering, along with quote activity, price volatility and bid-ask spreads, increases

sharply in the minutes following releases of macroeconomic news. Each returns to near-

normal levels within the hour. Effects are strongest for more liquid on-the-run notes and for

the announcements typically associated with substantial information flow. The strong positive

comovement of clustering, quote activity, price volatility, and bid-ask spreads supports the

conclusion that innovations of these variables are endogenous to the arrivaland incorporation

of information into prices.

Keywords: macroeconomic announcements, price resolution, U.S. Treasury securities, clus-

tering

JEL Classifications: G10, G14, G23, G24

∗Correspondingauthor: Rutgers School of Business, Rutgers University, Camden, NJ 08102; Phone: (856)

225-6594; Fax: (856) 225-6231; E-mail: andnikif@camden.rutgers.edu.

The authors acknowledge partial financial support from the David Whitcomb Center for Research in Fi-

nancial Services of Rutgers University.We thank the Editor, SrinivasanKrishnamurthy, and an anonymous

referee for helpful suggestions. We also thank seminar participants at Annual Meetings of the Eastern

Finance Association (2011) and Financial Management Association (2011) for useful comments and

suggestions on early drafts, especially our discussants Pankaj Jain and Dieter Hess.

C2017 The Eastern Finance Association 69

70 A. Nikiforov and E. Pilotte/The Financial Review 52 (2017) 69–100

1. Introduction

The endings of financial asset prices (the price with the whole number dropped)

are not uniformly distributed across available price-ticks. Similar to the tendency of

retail price-endings to cluster on 99, financial asset price-endings tend to cluster dis-

proportionately on round fractions.1For example, when the minimum price increment

in U.S. equity markets was one-eighth, even multiples of an eighth occurred more

often than odd multiples. Since decimalization, multiples of five and 10 cents are

disproportionately used relative to other multiples of a penny. The fractions that are

typically used are said to define the degree of price resolution (or price discreteness)

for an asset.

Ball, Torusand Tschoegl (1985) argue that the degree of price resolution depends

on the extent to which information can be incorporated to precisely estimate price,

so the use of round quotes will vary with the amount of information, the flow of

information, and the cost of introducing information in a market. They hypothesize

that the process by which a market interprets news and incorporates it into prices

is characterized by heightened market activity, increased price volatility, and greater

than normal clustering on round price-endings. Being a seminal study, Ball, Torus

and Tschoegl (1985) examine the empirical relations of proxy measures for the

amount, flow, and cost of introducing information to clustering in the gold market.

Surprisingly, three decades later, no study directly links clustering to information

arrival.

Our major contribution is empirical results that are the first, to our knowledge, to

establish a direct link between information arrival and the clustering of financial asset

prices. We do so using data from the U.S. Treasurymarket, where Green (2004) finds

that the release of macroeconomic news produces temporary increases in the level of

information asymmetry and the informational role of trading, due to differences in

the abilities of market participants to interpret the information. As predicted by Ball,

Torus and Tschoegl (1985), we find that clustering, along with quote activity, price

volatility, and bid-ask spreads, increases sharply in the minutes following the release

of macroeconomic news. Each variable returns to normal, or near-normal, levels

within the hour.The effect of information arrivalis most pronounced in the on-the-run

segment of the market, where very activetrading and low bid-ask spreads facilitate the

incorporation of information into prices, and for the high-impact announcements for

1Clustering in the U.S. stock markets is documented by Osborne (1962), Niederhoffer(1965, 1966), Harris

(1991), Christie and Schultz (1994), Bessembinder (1997), and Ikenberry and Weston (2007). Grossman,

Miller, Cone, Fischel and Ross (1997) document clustering on the London Stock Exchange, London gold

market, and international foreign exchange market. Huang and Stoll (2001) and As¸c¸ıo˘

glu, Comerton-

Forde and McInish (2007) find clustering on the London and Tokyostock exchanges, respectively. Price

clustering is documented for foreign bond futures contracts by ap Gwilym, Clare and Thomas (1998),

initial public offerings by Kandel, Sarig and Wohl (2001) and ap Gwilym and Verousis (2010), foreign

exchange spot markets by Sopranzetti and Datar (2002), S&P 500 index futures by Schwartz, VanNess

and Van Ness (2004), and real estate quotes and transaction prices by Palmon, Smith and Sopranzetti

(2004).

A. Nikiforov and E. Pilotte/The Financial Review 52 (2017) 69–100 71

which Green (2004) finds there is substantial information flow.2The strong positive

comovement of clustering, quote activity, price volatility and bid-ask spreads on the

release of macroeconomic news supports the conclusion that innovations of these

variables are endogenous to the arrival and incorporation of information into prices.

Our evidence linking bond price clustering to information arrival contrasts with

Ikenberry and Weston’s (2007) statement that stock price clustering is unaffected by

earnings announcements.3

The U.S. Treasury market is an ideal setting for tests of the impact of information

flow on price clustering. Price uncertainty is limited to variation in the risk-free rates

used to value certain cash flows. And information arrivalrelevant to the determination

of those rates is concentrated on days of macroeconomic announcements whose value

relevance and nearly immediate impact are well documented (e.g., Fleming and

Remolona, 1997; Balduzzi, Elton and Green, 2001; Green, 2004). Because macro

news is released at pre-announced times, with low likelihood of advancedleakage, we

are confident that we observe the impact of the process by which news made available

simultaneously to all is incorporated in price.4And aspects of the Treasury market

facilitate controlled tests of the impact of liquidity, through comparisons of highly

liquid on-the-run notes to illiquid off-the-run notes of similar price risk (duration).

Moreover,we are able to control for market design by choosing a period during which

both on-the-run and off-the-run securities traded predominantly in the same market

setting, voice-brokered trading facilitated by interdealer brokers (IDBs). And we are

able to confirm our results with a recent sample from an alternative data source and

markedly different bond market environment.

Treasury bonds are not examined in the prior clustering literature, so before

examining the impact of information arrival, we document the existence of clus-

tering in our sample of voice-brokered data from GovPX. For the full sample and

for subsamples based on seasonedness (on-the-run vs off-the-run status), we find

statistically significant price clustering consistent with a general preference for avoid-

ing the use of small odd fractions in favor of larger even fractions. We find that the

tendency of quotes to cluster on round price-endings decreases when notes transi-

tion to off-the-run. The higher clustering for on-the-run securities is contrary to the

2Green (2004) finds that the order flow in the Treasury market following high-impact announcements is

more informative than order flow that follows low-impactannouncements.

3The published version of their manuscript does not contain details regarding the empirical results on

which Ikenberry and Weston (2007) base this statement, so we have no basis for proposing a cause for

the difference in conclusions. Potential causes include differences in asset characteristics, methodologies,

market structures, and/or the information environments.

4As a caveat to this statement, we note that Bernile, Hu and Tang (2016) find evidence of informed

trading, as measured by order imbalances in some markets during the lockup periods ahead of Federal

Open Market Committee announcements, and Kurov, Sancetta, Strasser and Wolfe (2016) find evidence

that price moves partially anticipate the surprise component of announcements in the minutes prior to some

types of macroeconomic announcements. The later pre-announcement effect is weak prior to January 2008;

strong after.