Challenging the spanning hypothesis at short horizons: Evidence from Norway

• Author: Siri Valseth
• Date: 01 December 2019
• Published date: 01 December 2019
• DOI: http://doi.org/10.1002/for.2594

Received: 29 September 2018 Revised: 3 March 2019 Accepted: 8 March 2019

DOI: 10.1002/for.2594

RESEARCH ARTICLE

Challenging the spanning hypothesis at short horizons:

Evidence from Norway

Siri Valseth

University of Stavanger Business School,

Stavanger,Norway

Correspondence

Siri Valseth, University of Stavanger

Business School, 4036 Stavanger,Norway.

Email: siri.valseth@uis.no

Abstract

The spanning hypothesis states that the yield curve reflects all availableinforma-

tion about future yields. This paper challenges the hypothesis by investigating

the predictive power of bond market order flow while controlling for principal

components extracted from the current yield curve. The results show that order

flows aggregated from interdealer trades in Norwegian government bonds can

predict yield changes and excess returns both in-sample and out-of-sample at

the daily and weekly horizons. This suggests that bond dealers possess informa-

tion about future bond yields that is not yet incorporated into the yield curve.

Forecasts based on microstructure data can thus be valuable for short term

investors.

KEYWORDS

bond risk premia, bond yields, order flows, predictability, spanning hypothesis

1INTRODUCTION

The future level and shape of the yield curve is of great

interest to investors, intermediaries, and monetary policy

makers alike. Fama and Bliss (1987) and Campbell and

Shiller (1991) were the first to establish that future bond

yields can be predicted and that the predictable component

is a time-varying risk premium. Their choice of predictors,

forward rates and yield spreads, rests on the assumption

that in efficient markets the current yield curve fully

reflects or “spans” all information relevant for forecasting

future yields and excess returns. This spanning hypothesis

has been challenged by a number of studies in recent years.

Ludvigson and Ng (2009), Greenwood and Vayanos (2014),

Joslin, Priebsch, and Singleton (2014), and Cieslak and

Povala (2015) find that macroeconomic factors, Treasury

bond supply,economic growth, and inflation expectations,

respectively, have additional power to predict bond yields

beyond that of forward rates and yield spreads. This paper

extends this literature by studying bond market order flow

as a predictor.

By challenging the spanning hypothesis at short hori-

zons using microstructure data this paper makes two

contributions. First, it investigates whether bond trades

contain information about the future level and shape of

the yield curve that is different from the information in

forward rates and yield spreads. Second, unlike previous

papers, which investigate predictability at monthly,yearly,

or longer horizons, this paper focuses on daily and weekly

forecasts. This is important for short-term investors and

bond market intermediaries, as a continuous flow of news

is disseminated to market participants via a number of

news services, and private information in bond markets

tends to be short lived.

The paper investigates the power of order flows to pre-

dict yield changes and bond excess returns both in-sample

and out-of-sample. Out-of-sample predictability is impor-

tant for investors and intermediaries to benefit from

the information in order flows. For lack of sufficient

microstructure data from major bond markets, I employ

data from the Norwegian government bond market for the

period 1999–2015. I aggregate trades into order flows by

Journal of Forecasting. 2019;38:820–832.© 2019 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/for

820