The accuracy of interest rate forecasts in the Asia-Pacific region: opportunities for portfolio management

ABSTRACT

We analysed interest rate forecasts from Australia, China, Hong Kong, India, Indonesia, Malaysia, New Zealand, Singapore, South Korea, Taiwan and Thailand. We assessed 532 forecast time series with a total of 85,264 individual interest rate forecasts. To do so, we carried out a comparison to naïve forecasts and investigated the forecast time series for topically orientated trend adjustments. In addition, we deployed the sign accuracy test and the unbiasedness test. The results are very sobering in part: 95.9% of all forecast time series are characterized by the phenomenon of topically orientated trend adjustments, and 99.4% of all forecast time series proved to be biased. Only a small proportion of the forecast time series (3.6%) reflected the future interest rate trend significantly more precisely than a naïve forecast. However, at the same time some of the results of the study are surprisingly positive. The sign accuracy test revealed that 48.3% of all forecast time series predict the interest rate trend significantly better than a random walk forecast.

KEYWORDS:

• Interest rate forecasts
• survey forecasts
• forecast accuracy
• portfolio management
• topically orientated trend adjustment behaviour

JEL CLASSIFICATION:

• E44
• E47
• F37
• G11
• G15

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ See, for example Throop (1981), Tabak and Feitosa (2008), Baghestani and Marchon (2012), Knüppel and Schultefrankenfeld (2013) and Pierdzioch (2015).

² See, for example Dua (1988), Zarnowitz and Braun (1992), Cho (1996), Gosnell and Kolb (1997), Greer (2003), Scheier and Spiwoks (2006), Goodhart and Lim (2008), Spiwoks, Bedke, and Hein (2008), Chun (2009), Spiwoks, Bedke, and Hein (2010), Jongen, Verschoor, and Wolff (2011), Kunze, Kramer, and Rudschuck (2014), Kunze and Gruppe (2014), Baghestani and Danila (2014), Beechey and Österholm (2014), Oliver and Pasaogullari (2015), and Miah, Khalifa, and Hammoudeh (2016).

³ See, for example Friedman (1980), Belongia (1987), Simon (1989), Hafer and Hein (1989), Francis (1991), Hafer, Hein, and MacDonald (1992), Domian (1992), Ilmanen (1996), Kolb and Stekler (1996), Baghestani, Jung, and Zuchegno (2000), Albrecht (2000), Spiwoks (2003), Brooks and Gray (2004), Benke (2004), Mose (2005), Baghestani (2005), Benke (2006), Spiwoks and Hein (2007), Mitchell and Pearce (2007), Spiwoks, Bedke, and Hein (2009), Gubaydullina, Hein, and Spiwoks (2011), Schwarzbach et al. (2012), Chortareas, Jitmaneeroj, and Wood (2012), Butter and Jansen (2013), Spiwoks, Gubaydullina, and Hein (2015), and Kunze et al. (2017).

⁴ See, for example Gosnell and Kolb (1997), Spiwoks and Hein (2007), Gubaydullina, Hein, and Spiwoks (2011), Jongen, Verschoor, and Wolff (2011), Butter and Jansen (2013), Spiwoks, Gubaydullina, and Hein (2015), and Baghestani, Arzaghi, and Kaya (2015).

⁵ World Economic and Financial Surveys, Regional Economic Outlook, Asia and Pacific, Oct 2018, International Monetary Fund.

⁶ See, for example, the J. P. Morgan Government Bond Index Global, the FTSE World Government Bond Index.

⁷ The Australian interest rate forecasts start in 1990. The other time series only begin later.

⁸ See Diebold and Mariano (1995), Diebold (2015) and Kilian (2015).

⁹ The test is used by a variety of studies, see, for example Baumeister and Kilian (2015), Baghestani, Arzaghi, and Kaya (2015), Miah, Khalifa, and Hammoudeh (2016).

¹⁰ The TOTA coefficient is calculated by a variety of studies, see, for example Bofinger and Schmidt (2003), Baghestani (2010).

¹¹ In the forecasts of the prime lending rate in Hong Kong, it can be seen that 26 out of 30 forecast time series (86.7%) were superior to the naïve forecast. In the case of the forecasts of the 3-month rate in Hong Kong, 24 out of 38 forecast time series (63.2%) were superior to a naïve forecast, while in the forecasts of the 3-month rate in India, at least the forecasts with a 13-month forecast horizon were highly successful: 9 out of 12 forecast time series (75%) were more precise than the corresponding naïve forecast. Among the forecasts of 10-year interest rates in Indonesia, the forecasts with a horizon of 13 months were once again very successful. About 9 out of 12 forecast time series (75%) are more exact than a naïve forecast. Forecasts of the base lending rate in Malaysia were more successful than a naïve forecast in 18 out of 30 cases (60%), which is also the case for forecasts of three-month interest rates in Malaysia in 19 out of 36 cases (52.8%). The forecasts of the three-month rate in New Zealand were more precise than a naïve forecast in 25 out of 36 cases (69.4%). Among the forecasts of the 3-month rate in Thailand with a forecast horizon of 13 months, 9 out of 13 forecast time series (69.2%) were superior to the naïve forecast.