Empirical research on the benefits of investing in inflation-linked bonds usually relies on a limited number of observations due to the relatively recent introduction of these assets. We estimate models for the break-even inflation rate and use these to create hypothetical inflation-linked bond returns. We compare these with the return on actual inflation-linked bond returns on a recent sample and find that surveys of professional forecasters and moving average models perform best. We confirm these findings for a sample of 19 international inflation-linked bond markets. Using surveys of professional forecasters, we create hypothetical inflation-linked bond return series for 41 countries starting in 1987 or later depending on the availability of nominal bond markets. These simulated series can be used by asset allocation researchers, but an average correlation of 0.7 means that the simulated series are at best reasonable proxies for real data on inflation-linked bond returns. This cautionary note is also relevant to appreciate existing research using simulated inflation-linked bond returns.

Introduction

Asset allocation studies typically use historical data on asset classes as inputs for expected returns, standard deviations, or correlations. In addition, models for asset liability management make direct or indirect use of real interest rates or inflation-linked bond returns when pension liabilities contain cost-of-living adjustments. Unfortunately, inflation-linked bonds have only been introduced relatively recently as investment opportunities for the public, which limits the available historical return series that can be used for empirical analyses.1 The longest inflation-linked bond return data series produced by Bloomberg Barclays are available for the United Kingdom (May 1981), Australia (January 1997), Canada (January 1997) the United States (February 1997), and France (September 1998). Since then, many other governments, both from developed and emerging markets, have started issuing inflation-linked bonds; see Swinkels (2012) and King and Low (2014). The existence of inflation-linked bond markets helps economists determine the real interest rates across countries; see Mishkin (1984) and Barro and Sala-i Martin (1990) for historical estimates of real interest rates for developed countries. The available inflation-linked bond return series of developed markets are not well-suited to determine the potential advantage that inflation-linked bonds have in an asset allocation problem, as inflation has been relatively constant around two percent. Brière and Signori (2009) investigate the benefits of inflation-linked bonds using historical data on the U.S. and France over the period 1997 to 2007. They conclude that inflation-linked bonds did not add much value in an investor’s asset allocation after 2003. Cartea et al. (2012) find that inflation-linked bonds can be attractive to US long-term investors with a real investment objective. In order to determine the asset allocation benefits of inflation-linked bonds in times of inflation, Kothari and Shanken (2004) simulate historical inflation-linked bond returns starting in 1953. It does not become clear from their study what the quality of the simulated return series is relative to actual inflation-linked bond returns. Their sample period ends in December 2000, just after the introduction of inflation-linked bonds in the U.S., so called Treasury Inflation-Protected Securities (TIPS). Perhaps the short sample period is the reason they do not compare their simulated returns with observed returns on U.S. inflation-linked bonds.2 We aim to fill this gap in the literature. The use of extended time-series by historical simulation allows researchers to also include inflationary periods in their analyses. Another possibility is to increase the cross-section with observed inflation-linked bond returns. Swinkels (2012) extends the crosssection of inflation-linked bond returns to 9 emerging and 11 developed markets. Inflation in emerging markets has been markedly higher and more volatile than for developed markets, which is a good environment to investigate the benefits of inflation-linked bonds. Unsurprisingly, for samples with more inflation variability than in Brière and Signori (2009), inflation-linked bonds turn out to be a more attractive asset class for investors, mainly because of increased diversification with nominal bond returns in these environments. Barnes et al. (2010) advocate the use of index-linked bonds as an effective hedge against changes in the real yield or inflation shocks, especially for longer holding periods.