Forecasting European interest rates in times of financial crisis – What insights do we get from international survey forecasts?

https://doi.org/10.1016/j.intfin.2017.01.005Get rights and content

Highlights

  • Interest rate forecasts are important for financial market participants as well as monetary and political decision makers.

  • Survey predictions for interest rates are widely used but their accuracy may not be given as certain.

  • The accuracy of interest rate forecasts both for the United Kingdom and Germany is investigated.

  • High-Low-Spreads and Forecast Errors are used as dispersion measures to examine the uncertainty of professional forecasters.

  • In times of financial crisis financial market experts’ uncertainty can be linked to real economic indicators.

Abstract

Interest rate forecasts are widely used in the international financial services industry. For decades, both practitioners and academic researchers question the quality and usefulness of forecasts. Survey predictions do not only deliver point forecasts but also allow to draw conclusions with regard to the variety of forecasts provided by professional analysts. We evaluate the quality of interest rate forecasts for the three months interbank rate in the UK (LIBOR) and Germany (EURIBOR) as well as the corresponding 10Y government bond yields using the root mean squared error as well as the Theil’s U measure and also apply models of time series analysis (i.e. cointegration and causality analysis). Finally, we check for possible implications from uncertainty measures (i.e. High-Low-Spread of forecasts as well as forecast errors) and structural breaks. We are able to find some links to the real economy. Applying our methodological approach both to the UK and Germany we are able to draw conclusions with regard to the quality of international forecasts in times of uncertainty.

Introduction

Interest rate forecasts are important for financial market participants’ investment decisions. Since not all decision makers are willing or able to perform their own forecasts for the relevant term, forecast horizon or region, surveys collecting interest rate predictions are widely used. In addition to that, there exists empirical evidence for the general usefulness of survey forecasts (for a discussion of survey forecasts in general see for example Ang et al., 2007, Pesaran and Weale, 2006, Schmeling and Schrimpf, 2011 as well as Ince and Molodtsova (2017)). Predictions for interest rates are provided both for the short and the long end of the yield curve and for different countries. There already exists a vast literature dealing with the usefulness of the collected forecasts (i.e. both the individual predictions and the survey mean) for example with regard to rationality (see Friedman, 1980, Simon, 1989, Jongen and Verschoor, 2008 as well as Chortareas et al. (2012) and more recently Miah et al. (2016)), unbiasedness (Hafer and Hein (1989) as well as Mitchell and Pearce, 2007, Miah et al., 2016), efficiency (see for example Hafer and Hein (1989)) and accuracy (see for example Kolb and Stekler (1996) as well as Greer (2003)). Furthermore, summing up the results from the relevant literature would inevitably lead to the disillusioning conclusion that especially in times of financial crisis forecasts from professional analysts seem to perform fairly badly. This puts the usefulness of into question. However, in the context of crisis events dispersion measures have been investigated empirically. Dispersion measures may act as an indicator for uncertainty. Hence, a higher level of uncertainty might serve as a bellwether for dawning crisis events.

The contribution of this paper to the existing literature is threefold. Firstly, a long term assessment of survey forecasts for the three months interbank rate for two major European economies (i.e. the United Kingdom and Germany) and for the 10 year government bond yield will be performed using standard evaluation methods. Secondly, we will use more sophisticated methods of time series analysis to investigate the quality of survey forecasts (i.e. the survey mean) in an international context. Finally, and most important in the context of the global financial crisis, we try to get insights from uncertainty measures (i.e. the dispersion of the collected individual forecasts as well as the forecast errors). We try to find crisis related structural changes in the time series of dispersion measures and forecast errors. Since we understand possible structural breaks to be crisis-related the research question behind this approach points in the direction of early warning indicators or at least some kind of bellwether or discretional warning sign. In order to find evidence we link the uncertainty measures to real economic indicators using Granger causality analysis. By applying this methodological approach both to the UK and to Germany we try find parallels and want to rule out possible country specific implications.

To the best of our knowledge this is the first study to address the relevant question of combining uncertainty of interest rate forecast with real economic data in the context of the global financial crisis focusing on Germany and the United Kingdom. In this sense, we both deliver useful results for the forecast evaluation literature and fill a research gap in the context of possible crisis related relationships between forecaster uncertainty and real economic activity. One major advantage of our approach lies in the combination of easy to interpret forecast accuracy measures with time series analysis for a two country data set. In addition to that, we are able to cross link the financial market variables to the real economy. For political decision makers uncertainty measures can under certain circumstances be used as crisis indicators.

The structure of this paper is as follows. Chapter 2 provides a short literature overview whereas in chapter 3 the relevant data set of forecasts will be introduced. In chapter 4 we will briefly present the applied forecast evaluation methods whereas the empirical results will be discussed in chapter 5. Finally, we will shortly present the idea of uncertainty measures in chapter 6. In the same section we will also present some empirical evidence for crisis related structural breaks for the discussed uncertainty measures and check for implications from the uncertainty measures for real economic activity. In chapter 7 we will discuss possible economic implications. Chapter 8 concludes the paper.

Section snippets

Literature review

Not least due to the relevance of the accuracy of forecasts for decision makers the literature dealing with the evaluation of economic as well as financial market forecasts from professional analysts dates back for several decades. This is also the case for the research dealing with the evaluation of interest rate forecasts (see for example Friedman, 1980, Belongia, 1987, Spiwoks et al., 2008, as well as Hafer and Hein (1989)). One important field of empirical research has been the evaluation

Data

In this paper we investigate interest rate forecasts for two major European economies: The United Kingdom (UK) and Germany (GER). For both countries the three months interbank rate (from now on 3M) as well as the yield of 10 Year Government bonds (from now on 10Y) provided by Consensus Economics Inc. will be examined. The surveys for UK and GER are conducted with two forecast horizons: A three months as well as a twelve months ahead forecast. In fact, the actual forecast horizons are four

Methodologies of forecast evaluation

We start the empirical analysis of the UK respectively GER interest rate forecasts with simple evaluation measures. Although the root mean squared error (RMSE) is a crude method in the context of forecast evaluation it is widely used by practitioners in the financial service industry due to the straightforward interpretation and the relevance in statistical methodologies (see for example Hyndman and Koehler (2006)). Since this measure is widely known and in order to preserve space the formula

Evaluation results

Table 2 shows the Theil’s U results for the survey means for the UK data as well as the GER data.

It seems that the survey means are slightly better compared to the naive forecasts as reported in Table 2. However, this result is only statistically significant for the long run interest rates as reported in Table 3 on a level of 10%. As Table 4 below shows the time series for the actual interest rates as well as the survey means both for the UK data and for the GER data are non-stationary on

Examining possible implications from dispersion measures and forecast errors

The evaluation results presented in Chapter 5 above put the forecasting performance of survey forecasts into question. With regard to the survey mean (both for the case of the UK data and the GER data) we did not find strong evidence for the usefulness of professional forecasts. Anyhow, bearing in mind the exceptional market reactions due to past financial respectively economic crisis we want to check for possible insights from forecast errors and dispersion measures in the context of

Policy implications

Dealing with the causes and consequences of financial crisis events is a demanding task for financial market participants as well as political decision makers. Additionally, the global financial and economic crisis did visualize quite impressively that crisis events and the consequences thereof are not regionally restricted and hence have to be analyzed in an international context. For example, monetary policy decisions of the European Central Bank (ECB) have to be considered by the Bank of

Conclusion

In this paper the forecasts of professional analysts for the three months interbank rate as well as the 10Y government bond yield in the United Kingdom and Germany have been evaluated. Applying basic measures of forecast accuracy it has been shown, that the UK survey forecasts in general do not perform better than a simple naive forecast. These results do fit to the findings of earlier studies dealing with survey forecasts for German interest rates. Additionally, evaluating the mean forecasts

References (51)

  • F. Kunze et al.

    Interest rate forecasts in times of financial crisis: what might be interesting to know?

    Eur. J. Polit. Econ.

    (2014)
  • F. Miah et al.

    Further evidence on the rationality of interest rate expectations: a comprehensive study of developed and emerging economies

    Econ. Modell.

    (2016)
  • K. Mitchell et al.

    Professional forecasts of interest rates and exchange rates: evidence from the wall street journals panel of economists

    J. Macroecon.

    (2007)
  • W. Ploberger et al.

    A new test for structural stability in the linear regression model

    J. Econometr.

    (1989)
  • D. Saar et al.

    Forecasting growth and stock performance using government and corporate yield curves: evidence from the European and Asian markets

    J. Int. Finan. Markets Inst. Money

    (2015)
  • M. Schmeling et al.

    Expected inflation, expected stock returns, and money illusion: what can we learn from survey expectations?

    Eur. Econ. Rev.

    (2011)
  • P. Sibbertsen et al.

    Testing for a break in the persistence in yield spreads of EMU government bonds

    J. Bank. Finan.

    (2014)
  • J.R. Stillwagon

    Testing the expectations hypothesis with survey forecasts: the impacts of consumer sentiment and the zero lower bound in an I(2) CVAR

    J. Int. Finan. Markets Inst. Money

    (2015)
  • J.H. Stock et al.

    Interpreting the evidence on money-income causality

    J. Econometr.

    (1989)
  • A. Zeileis

    Implementing a class of structural change tests: an econometric computing approach

    Comput. Stat. Data Anal.

    (2006)
  • A. Zeileis et al.

    Testing and dating of structural changes in practice

    Comput. Stat. Data Anal.

    (2003)
  • J. Bai et al.

    Computation and analysis of multiple structural change models

    J. Appl. Econometr.

    (2003)
  • T. Basse et al.

    Variable-ordering induced problems of impulse-response analysis and other difficulties: the dividend policy of Austrian firms

    Int. J. Comput. Econ. Econometr.

    (2010)
  • M. Belongia

    Predicting interest rates: a comparison of professional and market-based forecasts

  • B. Bernanke

    On the predictive power of interest rates and interest rate spreads. Tech. rep.

    (1990)
  • Cited by (13)

    • Modelling the dynamics of unconventional monetary policies’ impact on professionals’ forecasts

      2020, Journal of International Financial Markets, Institutions and Money
      Citation Excerpt :

      However, Ince and Molodtsova (2017) provided evidence that the exchange rate survey-based forecasts for developing countries are biased at all forecast horizons, while for developed countries long-term forecasts are more accurate than short-term forecasts. Kunze et al. (2017), by using interest rate forecasts for the United Kingdom and Germany, showed that in times of financial crisis financial market experts’ uncertainty can be linked to real economic indicators. Christensen and Rudebusch (2012), by studying QE policy announcements in the United States during 2008 and the United Kingdom during 2009, supported that the drop in interest rates in the former can be attributed to lower expectations about future interest rates (Blue Chip Financial Forecasts estimated what the future expected value of a 3-month Treasury bill would yield for 6 and 12 months ahead), while the latter reflect lower risk premia.

    • Sovereign bond return prediction with realized higher moments

      2019, Journal of International Financial Markets, Institutions and Money
      Citation Excerpt :

      Finally, Chordia et al. (2015) and Choi and Kim (2018) investigate the importance of equity market anomalies in relation to the cross-section of corporate bond returns. Evidence on sovereign bond markets mainly focuses on the in-sample predictability of short to medium-term bond returns using variables such as forward spreads (Fama and Bliss, 1987; Cochrane and Piazzesi, 2005), yield spreads (Campbell and Shiller, 1991), factors extracted from the cross-section of macroeconomic variables (Ludvigson and Ng, 2009), or simply using basic measures of forecast accuracy (Kunze et al., 2017). Fujiwara et al. (2013) study conditional asymmetries in government bond excess returns for five developed markets - Canada, Germany, Japan, U.K. and the U.S. – and find that positive bond returns are associated with negative skewness and that bond liquidity is able to forecast skewness with a positive sign.

    • Optimization of Financial Market Forecasting Model Based on Machine Learning Algorithm

      2023, Proceedings - 2023 International Conference on Networking, Informatics and Computing, ICNETIC 2023
    • Interest rate forecasts in Latin America

      2022, Journal of Economic Studies
    • Sticky Stock Market Analysts

      2021, Journal of Risk and Financial Management
    View all citing articles on Scopus
    View full text