Interest rate expectations based on Taylor rule versus central bank’s survey: which performs better in a large emerging economy?

ABSTRACT

We analyzed rationality, content, and anchoring of the monetary policy interest rate expectations (for 3, 6, 9, and 12 months ahead), taking into account the Brazilian data from January 2003 to July 2020. We consider expectations based on two perspectives: expectations gathered from a Taylor rule and market expectations obtained from the survey carried out by the central bank. The findings point out that, concerning rationality and anchoring, the interest rate expectations based on a Taylor rule performed better than expectations from the Central Bank of Brazil’s (CBB) survey, even when we consider the Top 5 forecasters. Moreover, our analysis shows that the content of monetary policy interest rate expectations based on a Taylor rule and CBB’s survey (including Top 5 forecasters) is different. Thus, they may be seen as complementary sources of information on the future interest rate. Concerning anchoring of the monetary policy interest rate expectations, the CBB’s survey expectations are not anchored, while the result for the expectations based on a Taylor rule shows the opposite for horizons up to 9 months ahead.

KEYWORDS:

• Interest rate expectations
• Taylor rule
• Central bank’s survey
• Rationality
• Anchoring

JEL CLASSIFICATION:

• E42, E47

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 CBB was awarded in the Central Banking FinTech RegTech Global Awards 2018 for Best Data Management Initiative.

2 The inflation rates and interest rates from the G20 countries at the end of 2021 are available from: https://tradingeconomics.com/indicators.

3 On the methodology regarding non-bias and efficiency tests, see, for example, Mincer and Mincer and Zarnowitz (1969), Holden and Peel (Holden and Peel 1990), Beechey and Österholm (2014a), and Chortareas, Jitmaneeroj, and Wood (2012).

4 On tests concerning the comparison in predictive ability, see Diebold and Mariano (1995) and Harvey, Leybourne, and Newbold (1997).

5 The test we will use as a reference is derived from Fair and Shiller (1989).

6 See, for example, Gürkaynak, Levin, and Swanson (2010).

7 According to CBB: “the inflation expectations for the next 12 months are calculated daily for each of the institutions that have projections for all 12 months of the period considered. The calculation is made by accumulating the next 12 expectations for monthly inflation” (https://www.bcb.gov.br/en/monetarypolicy/marketexpectationsfaq).

8 For example, the weighted inflation target (t +12 - next twelve months) in June 2015 results from 6× inflation target 2015 + 6× inflation target 2016.

9 Table A1 (see Appendix) shows the descriptive statistics and unit root tests for all variables in the regressions. The findings indicate that all series are I(0).

10 As pointed out by the CBB, a good part of the participants updates their expectations weekly (https://www.bcb.gov.br/en/monetarypolicy/marketexpectationsfaq). Moreover, the number of participants can be considered high compared to other central banks’ surveys. For example, the surveys of professional forecasters like from the European Central Bank and Philadelphia Federal Reserve have about ninety and forty respondents, respectively.

11 It is important to note that “the samples of institutions that inform each monthly projection and each annual projection are different. Furthermore, the accumulation of the monthly medians would not necessarily be equal to the respective annual median, even if the samples were the same” (CBB 2016, 10)).

12 For a detailed explanation of how the Top 5 institutions are ranked, see https://www.bcb.gov.br/en/monetarypolicy/marketexpectationsfaq.

13 For an analysis regarding efficiency, see Nordhaus (1987).

14 It is noteworthy that all information used to forecast the monetary policy interest rate to t+n, including the realized interest at t, is available to the three types of forecasts under consideration (Taylor rule-based, professional forecasters, and Top 5). Hence, a priori, there is no advantage of information to a specific type of forecast.

15 The level of economic activity is the Central Bank Economic Activity Index (seasonally adjusted), and the inflation rate is the Extended National Consumer Price Index. Both series are available from the Time Series Management System/CBB.

16 For an analysis regarding anchoring inflation expectations in Brazil, see de Mendonça, Garcia, and Vicente (2021).

17 We checked the results from the total sample using two subsample periods. Period 1 – covers the span period from the beginning of the total sample to the climax of the 2008 global financial crisis (January 2003 – September 2008); and period 2 – from the start of the impeachment proceedings against the President of Brazil Rousseff to the beginning of the COVID-19 pandemic (January 2016 – December 2019). In general, the results agree with those found to the total period (see Table A2 – Appendix).