i 
 

Do Domestic Yield Curves in Emerging Market Economies Prove to be Useful in 

Forecasting Future Economic Growth? 

 

 

by 

 

 

Rushai Gosai 

0510713v. 

 

 

 

A RESEARCH REPORT PRESENTED IN PARTIAL FULFILMENT (50%) OF 

THE REQUIREMENTS FOR THE DEGREE OF MASTER OF COMMERCE IN 

BUSINESS ECONOMICS (FINANCE) 

 

in the 

 

SCHOOL OF ECONOMIC AND BUSINESS SCIENCES 

 

at the 

 

UNIVERSITY OF THE WITWATERSRAND, JOHANNESBURG  

 

 

 

 

 

 

 

 

Supervisor: Mr James Britten  

 

 

 

 

 

 



 

ii 
 

Date of submission: 6th of December 2022 



 

iii 
 

Abstract 

 

Much has been said and researched about the term spreads ability to forecast the path 

of Gross Domestic Product (GDP) in developed economies. The relationship holds that 

should the yield spread turn negative that this indicates that future GDP will retract and 

that a recession is eminent. At the back end of 2019, the subject found prominence 

again as the yield spread measured by the ten year government bond and the three 

month Treasury Bill (Tbill) turned negative. The Federal Reserve Bank of America (The 

Fed) lowered interest rates in the hope that lower borrowing costs would stimulate the 

economy and lead to an increase in aggregate demand. It then follows, could the 

domestic yield curve spread perhaps be suitable in forecasting domestic Emerging 

Market (EM) GDP growth? This research highlights the EM experience whilst still 

testing the ability of the yield curve in the US to predict future economic growth. The 

framework  based on the work of Bosner-Neal and Morley (1997),  found over the  

horizon of 1980 to 2020, for the EM countries of Brazil, Russia, India, China and South 

Africa (BRICS) unsupportive evidence that the domestic yield curve spread is a 

suitable indicator to forecast future GDP growth. 

 

 

  



 

iv 
 

 

 

SCHOOL OF ECONOMIC AND 
BUSINESS SCIENCES 

 

 

Declaration Regarding Plagiarism 

 

  

 

I (full names & surname): Rushai Gosai 

Student number: 0510713v 

 

 

Declare the following: 

1. I understand what plagiarism entails and am aware of the University’s policy in this regard. 

2. I declare that this assignment is my own, original work. Where someone else’s work was 
used (whether from a printed source, the Internet or any other source) due 
acknowledgement was given and reference was made according to departmental 
requirements. 

3. I did not copy and paste any information directly from an electronic source (e.g., a web 
page, electronic journal article or CD ROM) into this document. 

4. I did not make use of another student’s previous work and submitted it as my own. 

5. I did not allow and will not allow anyone to copy my work with the intention of presenting it 
as his/her own work. 

Rushai Gosai       30TH of April 2021 

Signature 

 

 

 Date 

 

 

 

 

 

 

 

  



 

v 
 

Acknowledgements 

I would like to take this opportunity to thank the Markets and Treasury staff at ABSA 

for providing me with the necessary data required for this research report as well as 

for the guidance provided to understand yield curve dynamics. 

  



 

vi 
 

Contents 

Abstract ....................................................................................................... iii 

Declaration Regarding Plagiarism ............................................................. iv 

Acknowledgements ..................................................................................... v 

Abbreviations ............................................................................................... x 

1. Introduction .......................................................................................... 1 

1.1. Background and overview .............................................................. 1 

1.2. Motivation of the study ................................................................... 3 

1.3. Research objective ......................................................................... 4 

1.4. Problem statement ......................................................................... 5 

1.5. Hypothesis ..................................................................................... 6 

1.6. Contribution to the study................................................................. 6 

2. Literature review .................................................................................. 8 

2.1. The yield curve as a variable to forecast economic growth ............. 8 

2.2. Forecasting inflation from the term structure ................................. 22 

3. Description of data and research methodology .............................. 36 

3.1. Theoretical framework .................................................................. 37 

3.2. Data ............................................................................................. 38 

3.3. Variables ...................................................................................... 38 

3.4. Limitations .................................................................................... 39 

3.5. Hypotheses .................................................................................. 41 

4. Results ................................................................................................ 42 

4.1 Regression Results ............................................................................ 43 

4.2 Results ............................................................................................... 56 

4.2.1 Results for the USA......................................................................... 56 

4.2.2 Results for SA ................................................................................. 58 

4.2.3 Results for India .............................................................................. 59 

4.2.4 Results for Russia ........................................................................... 60 

4.2.5 Results for China ............................................................................ 61 

4.2.6 Results for Brazil ............................................................................. 62 

5. Conclusion ......................................................................................... 66 



 

vii 
 

6. Future research .................................................................................. 68 

7. References ......................................................................................... 70 

 

  



 

viii 
 

List of Tables 

Table 1 - Probability of recession four quarters ahead ............................................. 25 

Table 2 - Descriptive statistics for The United States of America ............................. 43 

Table 3 - Descriptive statistics for South Africa ........................................................ 46 

Table 4 - Descriptive statistics for India ................................................................... 48 

Table 5 - Descriptive statistics for Russia ................................................................ 50 

Table 6 - Descriptive statistics for China .................................................................. 52 

Table 7 - Descriptive statistics for Brazil .................................................................. 54 

Table 8 - Yield curve spreads and the forecasted GDP reaction ............................. 73 

Table 9 - Beta Coefficients and forecasted real GDP reaction ................................. 74 

 

  



 

ix 
 

List of figures 

Figure 1 - Term Spread 10 year bond less 3 month Tbill (USA) ................................ 2 

Figure 2 - Yield curve spread tracks future output ................................................... 13 

Figure 3 - Beta coefficients for each country for real GDP using the term spread .... 20 

Figure 4 - Beta results for the USA .......................................................................... 57 

Figure 5 - Beta results for South Africa .................................................................... 58 

Figure 6 - Summary of Beta results for countries examined .................................... 63 

 

  



 

x 
 

Abbreviations 

Gross Domestic Product – GDP 

South African Reserve Bank – SARB 

Gross National Product - GNP 

Emerging Market - EM 

Brazil, Russia, India, China, South Africa - BRICS 

The United States of America – The US 

Treasury Bill - Tbill 

The United Kingdom – The UK



 

1 
 

 

1. Introduction  

1.1. Background and overview 

  

Investigating the yield curve term structure between the three month T bill and the ten 

year government bond has at times shown to be inverted. Inverted yield curves arrive 

when short term debt is deemed riskier than long term debt. Though many investors 

may try to time the market, the consensus represented by an inversion is historically 

correct and foreshadows economic woes to come. Harvey (1988) found that the yield 

spreads between US treasuries forecasted real Gross National Product (GNP) activity 

at least five quarters in advance. Estrella and Hardouvelis (1991), one of the seminal 

references in this topic, show that when yield spreads turn negative, this precedes a 

decline in economic activity. Whilst the three-month, ten-year spread may prove to be 

negative at times in the United States. This research investigates whether using the 

same term points to calculate a term point differential provides insight into future 

economic activity for developing economies. 

 

Figure 1 below shows the illustrative history over a twenty-year period of the spread 

between The United States (The US) ten year government bond yield less The US 

three month Tbill rate. The shaded areas show The United States recessions and it 

can be seen that when the term spread turns negative, it precedes a recession both in 

2001 and 2008. Figure 1 is similar to Figure 2 from Estrella and Hardouvelis (1991) 

which shows the same relationship over a different time period.  

 

Central banks around the industrialised world have historically used the term spread 

as a precursor for a declining economic environment. While forecasting time periods 

vary, models used by the St Louis Federal Reserve forecast economic conditions from 

two to six quarters ahead. This is graphically shown below in figure 1. 

 

 

 

 



 

2 
 

 

Figure 1 - Term Spread 10 year bond less 3 month Tbill (USA) 

Source: Adopted from “South Africa’s Taylor’s Rule calculated by, 

“https://fred.stlouisfed.org/series/T10Y3M, January 28, 2021 

 

Recent economic events, such as the Federal Reserve’s emergency cut in interest 

rates and enormous intervention in the markets which was followed by a total of 300 

basis points cut by the South African Reserve Bank (The SARB), has raised the 

argument again whether inverted yield curves signal a fore coming economic 

recession. With all the monetary policy intervention, it is the aim of this research to 

determine whether the yield curve spread provides valuable and accurate forecasting 

information related to economic growth in EM. Rate cuts lowered short term rates as 

a monetary response to create aggregate demand. This intervention in the US did not 

permeate through the longer tenors of the yield curve as investors are of the belief that 

longer term yields and expectations of the future are more grounded in reality than 

those of the current. Given that we have seen yield curve inversion in 2020, which 

preceded a recession, the argument is whether the SARB and other developing 

economies, who similarly cuts rates as well, show a similar relationship when 

comparing the short term rates to long term rates and whether this will hold 

informational content about the economic prospects. 

 

Research from Fama (1984), Harvey (1988) and Estrella and Hardouvelis (1991) to 

name a few have suggested that inverted yield curves have been shown to forecast 

economic downturns and subsequently economic growth. The key element missing, 

and one that this research addresses, is to study whether the term spread of emerging 

economies has informational content that can be used to forecast the path of future 

economic growth. 



 

3 
 

 

The ten year US bond yield (10 Year Treasury Rate, n.d), which is taken as the risk 

free rate for financial transactions across the globe, has dropped below one percent 

since late March 2020.  

 

The yield spread, which is predominantly taken as the difference between the ten year 

government bond yield and the three month Tbill rate, as highlighted in figure 1, turned 

negative in February 2020 and stayed negative into early March 2020. Typically, bonds 

with longer maturities pay more in periodic coupons -this is common sense. However, 

this is also based on the expectation that that the economy will grow in the future. 

 

To illustrate why the yield curve inverts before recession, let’s consider the example of 

the yield curve representing monetary policy expectations. If market participants 

expect growth to slow down in the near future, they start pricing in higher probabilities 

of future rate cuts to support economic growth. This has resulted in long dated bond 

yields decreasing by a higher margin than short dated bonds and hence, the yield 

curve flattens.  

 

Historically, an inverted yield curve has successfully signaled a recession six to 

eighteen months before it happened (Lewis, 2019), thus justifying recent debate on 

whether a US recession is incoming. Market participants keep a close eye on the ten 

year bond, three month Tbill spread. It is this spread and the relation to economic 

growth which literature shows has a forecasting relationship that key market 

participants have picked up on. 

 

1.2. Motivation of the study 

 

The study is motivated by the review of the literature which shows that the bulk of the 

studies in the field of yield curve analysis for forecasting output or inflation is primarily 

centred on The US and other industrialised or first world countries. There have been 

very few studies, when compared to the vast literature on the topic in developed 

economies, that have considered whether the yield curve can forecast the path of 

future output growth in emerging market economies.  



 

4 
 

 

 

 The aim of this research is to determine if the informational properties highlighted in 

previous research related to output and inflation drawn from the yield curve holds for 

emerging market economies. It is the primary outcome of this research to replicate the 

studies based predominantly on first world economies that have gone before to 

determine if the relationship, as seen in the review of the literature, which implies that 

the yield curve does hold forecasting properties related to the path of future output and 

inflation.  

 

1.3. Research objective   

 

Bernard and Gerlach (1998) build on the work by Estrella and Hardouvelis (1991) who 

showed that the yield curve forecasts recessions from using the term spread by as 

much as two years in advance. The general market accepted methodology is that yield 

curve inversion predicts a future recession four quarters ahead.  

 

There are research studies which show that forecasting real economic activity forms 

the basis of many decisions. This can take the form of businesses relying on such 

forecasts in order to plan their production process. Central bankers consume such 

statistical outputs and reports when making decisions that relates to the trajectory of 

monetary policy going forward into the future or when planning for future budgetary 

processes. Therein lies the true question of the research, the value of the choices 

mentioned above depends on the quality of the forecast. 

 

The objective of this paper is to determine whether the domestic yield curves in 

emerging market economies also hold the same informational content which forecast 

economic output. The objective of this research is to make use of the paper by Bosner-

Neal and Morley (1997) and replicate the methodology presented in the paper with the 

change of using emerging market economy data to determine whether the term spread 

in emerging economies has the forecasting ability as mentioned by Estrella and 

Hardouvelis (1991) and Bosner-Neal and Morley (1997).  



 

5 
 

1.4. Problem statement  

 

Bosner-Neal and Morley (1997) illustrate that the application of the yield curve spread 

as a variable to forecast real economic activity is well established for data based on 

The US. The authors ask the question whether this is the case in other countries. There 

are studies which ask the question outside of The US such as Mehl (2009) as well as 

Plosser and Rouwenhorst (1994), initially these studies were limited but have grown 

in significance and importance, and those that have asked the question have limited 

their studies to other industrialised economies as well as the length of the forecast 

horizon. 

 

It must be noted that the studies mentioned in the research by Evgenidis, Papadamou 

and Siriupoulos (2018) are all based on information that is specific to the US. 

Subsequently there have been a number of studies that have investigated whether the 

yield curves forecasting ability is applicable in other countries. The diffusion of the 

study moved from the US and across to Euopean nations to determine if the yield 

curve relationship to GDP held. Whilst this was a natural evolution, the bulk of the 

European nations studied are all developed economies.  Plosser and Rouwenhosrt 

(1994) were among the first to study the yield curves forecasting ability outside the US. 

The results showed that the yield curve spread proved a suitable forecasting tool for 

information from Canada and Germany however the same cannot be said for the UK 

and France. 

 

In summary, the problem statement is centered around the investigation on emerging 

market economies. The research that has been reviewed is clear in that the research 

has not really been done on emerging market economies. This is the aim of this 

research. 

 

  



 

6 
 

1.5. Hypothesis 

 

Estrella and Mishkin (1991), Bernard and Gerlach (1998), Bosner-Neal and Morley 

(1997), Hamilton and Kin (2000) and Evgenidis and Siriopolous (2016), to name a few 

of the studies reviewed for this research, have shown that the yield curve spread holds 

informational content related to the future path of economic output. It has been 

documented that this relationship holds for the US and many other industrialised 

countries. It is the aim of this research to determine if the relationship that the yield 

curve spread holds information content for economic output growth. If so, how far into 

the future does the yield curve spread forecast the future path for economic output and 

whether this forecast will have a statistically and economic significant impact for the 

EM countries that have been selected.  

  

1.6. Contribution to the study 

 

The yield curve spread as an economic indicator for future economic growth has 

become somewhat of a stylized fact over the last few decades. This relationship that 

the spread can forecast economic growth and, in particular, forecast future downturns 

has been shown to be prevalent in developed economies. This research aims to 

determine if the relationship holds for emerging market economies and to what extent.  

It is well known that the US and other developed economies are the major trading 

partners of the emerging economies studied in this research and as such it would be 

a safe approximation that the domestic yield curves of the developed economies will 

impact that of the developing economies. In this research we ignore the impact of 

developed economies' interrelatedness and focus primarily on the domestic yield curve 

informational content. This research will aim to understand to what extent the domestic 

yield curve has informational content about domestic economic growth. 

Core findings show that in the US, the term point differential between the 3 month Tbill 

and 10 year does have explanatory power for future economic growth one, two and 

three years into the future. The results are mixed at best for the BRICS countries with 

the results leading to less than satisfactory. Only South Africa shows some evidence 

that the yield curve has explanatory power for future economic growth but only at the 

one year ahead horizon. 

 



 

7 
 

The remainder of the research proceeds as follows: Section two will introduce a 

literature review on the topic of yield spread and the relationship with forecasting 

economic growth. Section three introduces the methodology that the research 

employed and the variables used in the regression model. Section four highlights the 

results of EM countries and The US. Section five provides a conclusion of the research. 

Section six provides future research with section seven highlighting the references 

used for this research. 

 

  

 

 

  



 

8 
 

2. Literature review 

2.1. The yield curve as a variable to forecast economic growth   

 

Yield inversion is the term used when long term rates are lower than short term rates. 

This happens when investors are nervous about the future and expect short term rates 

to fall. When so many investors think rates are going to fall, they will crowd into the 

longer-dated bonds to try to lock in the 'high' rate for as long as possible. The traditional 

measure of whether or not the yield curve is said to be normal, flat, or inverted is by 

examining the relationship between the 3-month and 10-year rates. The below chart 

shows the value of the 10-year bond (normally high) minus the value of the 3-month 

bond (normally low). We should expect the result to be a positive number, given a 

normally sloped yield curve. This is observed in the chart below. The majority of the 

data points are above 0%, shaded blue. The few periods where the yield curve was 

inverted are shaded red. The grey vertical lines indicate recessions, and occur after 

each inversion. 

 

Figure 1 – Term spread 10 year bond less 3 month Tbill (Present USA) 

Source: Yield curve indicates stock market is fairly valued by, “https:// 

https://www.currentmarketvaluation.com/models/yield-curve.php 

 

The implication of the forecasting ability of the term spread in forecasting economic 

activity has been accepted for The US in papers by Fama (1984), Harvey (1988), 

Estrella and Hardouvelis (1991), Davis and Henry (1994), Bosner-Neal and Morley 

(1997), Davis and Fagan (1997) and Stock and Watson (2003). 



 

9 
 

 

The maxim that when the yield curve inverts, it signals a recession has been formalized 

by a number of studies namely Laurent (1988, 1989), Harvey (1988), Stock and 

Watson (1989) as well as Estrella and Hardouvelis (1991). The studies mentioned here 

focused primarily on using the term spread to forecast future output growth based on 

data from the US.   

 

Fama (1984) examined the one to six month Tbill rates from 1959 through to 1982. 

Fama (1984) found that forward rates forecasted the correct direction of the changes 

in short term rates. Mankiw and Miron, (1986) in their research evidence showing the 

forecasting ability of short term rates on real economic activity, using the three and six 

month Federal Reserve rates. Mankiw and Miron (1986) looked at data during periods 

monetary regimes.  

 

While this is not the purpose of this research, the study is suitable as it does pertain to 

the forecasting power of the term spread. The results interestingly show that the 

forecast ability of short term rates on real economic activity varies through monetary 

regimes. Over the periods studied 1890-1914, 1915-1933, 1934-1951, 1951-1958 the 

results show that there is zero to little evidence of the short term rates having some 

sort of forecast ability on real economic activity. The relationship then changes 

completely from 1959 to 1979. Whether this is due to changes in macroeconomic 

policy or policy regime or integrated financial markets is unclear.  

 

Harvey (1988) found that the yield spreads between the US treasuries forecasted real 

GNP activity at least five quarters in advance. In the same breath, Estrella and 

Hardouvelis (1991) found that the term spread differential between the ten year 

government bond and the three month US Tbill proved suitable in forecasting output 

growth and recessions in the US. Fama and Bliss (1987) found that longer term forward 

rates have forecast ability power for a horizon at least two to four years ahead. 

 

Hardouvelis (1988) examined the forecasting ability of forward rates spanning through 

multiple monetary policy system changes. The research used the three month Tbill 

rates with maturities from one to twenty six weeks. The research found that there was 

no connection between the ability of the Fed to stick to rate targeting and the forecast 



 

10 
 

ability of interest rates. That said, Hardouvelis (1988) reported that the forecasting 

ability of term spread did increase from 1979. 

Harvey (1988) examined the term structure of real interest rates as a variable to 

forecast future consumption - the evidence supported that the slope of the yield curve 

is a better forecasting variable of real economic growth. Harvey (1988) outlined that a 

common approach for research within the financial field is to use or find variables that 

explain movements of prices. Harvey (1988) differed in that expected returns were first 

estimated and subsequently information about future consumption growth was 

extracted.  

 

Harvey (1988) employed the use of quarterly data using a regression model which 

tests for real consumption growth on the expected real yield spread over the period 

1953 to 1987. Yield spread data is used for three, six and nine month Tbills and one 

year Treasury bonds. Harvey (1988) showed that there is informational content about 

future consumption growth in the real term structure. Harvey (1988) went further and 

showed that for the data set and period used, real interest rate variables proved to be 

better in forecasting consumption growth than lagged variables. This was shown in 

both in sample and out of sample testing.  

 

Mishkin (1988) corroborated the evidence found by Fama (1984) who used multiple 

time series regressions with the purpose of determining the informational content in 

forward interest rates. Fama (1984) found evidence that one month forward rates has 

power to forecast the spot rate one month ahead. Hardouvelis (1988) used weekly Tbill 

data with maturities ranging from one to twenty six weeks in a regression model in 

which Hardouvelis (1988) finds that r-squared values indicated little forecast ability. 

This is also similar to the regression results of Estrella and Hardouvelis (1991) as well 

as Bosner-Neal and Morley (1997).  

 

Hardouvelis (1988) found that when The Fed let interest rates fluctuate between the 

period of 1979 and 1982, forecasting power increased substantially from a one week 

ahead to twenty one weeks ahead.  Mishkin (1990) examined yields with maturities 

stretching from one to twelve months and found that most of the data inherent in 

forward rates is about the expectation of future real rates.  

 



 

11 
 

Mishkin (1990) found in the research that the term spread provided in sample 

forecasting content for real interest rates, particularly at shorter horizons. Interestingly, 

Fama (1990) found that an increased spread differential between the five year bond 

and the one year bond forecasted an uptick in the rate of economic growth rate for the 

prevailing five years and a subsequent decrease in the real interest rate between two 

to three years into the future. Fama (1990) employed the use of a regression model 

using spot rate, inflation rate and real return on the five year yield spread to forecast 

the one year spot rate one year ahead into the future. The results show that the yield 

spread shows no forecasting power over the one year ahead spot rate. The yield 

spread evidence however did show strong evidence in forecasting inflation. Studies 

specific to inflation, make the distinction where the importance lies, “inferring where 

the economy is” 

 

There are studies that investigated the forecasting qualities of the yield curve and 

inflation. The Fisher equation illustrates that nominal rates reflect expectations of both 

future inflation as well as the real rates. Mishkin (1990) finds forecasting information of 

the US yield curve and inflation. Further Mishkin and Jorian (1991) get to a similar 

conclusion over ten developed economies.   

 

Estrella and Hardouvelis (1991) showed in the review of the literature that the data is 

comparable with the idea that the gradient of the yield curve has forecasting ability 

about interest rates in the short and long term. Estrella and Hardouvelis (1991) 

provided the strong evidence using their probit regressions.  

 

Estrella and Hardouvelis (1991) observed quarterly real GNP from the 1955 through 

to 1988. In their simple regression model the dependent variable is the annualised 

cumulative percentage change in the seasonal adjusted real GNP numbers. The 

interest rate set of data used for the study is the ten year government bond and the 

three month Tbill rate, where the ten year rate government bond represents the long 

term rate and the Tbill rate represents the short term rate.  

 

It is the aim of this research to follow a similar approach to Estrella and Hardouvelis 

(1991), Bosner-Neal and Morley (1997) provided a refreshed view of the methodology 

employed by Estrella and Hardouvelis (1991). The results of both support the notion 

that the yield holds forecasting ability as it relates to future GDP growth.  



 

12 
 

 

The regression model from Estrella and Hardouvelis (1991) present their regression 

results which illustrates that the findings are consistent with previous theory and 

research in that a steeper or flatter yield curve slope implies that there will be a faster 

or reduced expectation of future economic growth in real output. The data shows that 

for a one hundred basis point spread between the ten year bond and the three month 

Tbill shows that the cumulative change over the course of a year the real GNP is 

forecasted to grow by three percent. This illustrates that there is a relationship between 

the term spread of short term and long term rates and the impact they have on future 

economic productivity. 

 

Estrella and Hardouvelis (1991) show that for a forecast of negative real rates to occur 

from a current quarter to a future quarter, the slope of the yield curve would need to 

be less than -1.31%. More importantly, the results how that the marginal forecasting 

power for market participants will occur six to seven quarters in advance. Estrella and 

Hardouvelis (1991) showed that the term spread explained close to one third of the 

variation in the future real output. 

 

As a visual representation of the result from the study by Estrella and Hardouvelis 

(1991), Figure 2 highlights that the shaded areas are defined by the National Bureau 

of Economic Research as recessions. The illustration shows that when the slope of 

the yield curve spread tracks the future realization of output quite effectively, this 

coincides with the decline in economic activity. 

 



 

13 
 

 

Figure 3 - Yield curve spread tracks future output 

Source: Adapted from “The term structure as a forecaster of real economic activity” by 

Estrella and Hardouvelis (1991) 

The current growth in real GNP and the slope of the yield curve four quarters earlier. 

To summarize, Estrella and Hardouvelis (1991) presented evidence to support the 

notion that the slope of the yield curve can be used to forecast changes in the real 

economic output at least four years into the future. 

 

Mishkin (1991) linked the information of the term structure to the focus at the time of 

central banks on price stability. In order for the central banks to pursue price stability, 

one of the major markers a central bank would need information on is inflation and one 

of the best sources of explanatory information is to analyse the term structure. One of 

the primary reasons why the information carried in the term structure relating to 

inflation requires careful analysis is due to the central bank using the term structure as 

a guide for monetary policy. 

 

Of further importance is the evidence shown by Mishkin (1991) which highlights that 

the term structure of real interest rates has an important role in understanding of asset 

pricing as well as the business cycle. Mishkin (1991) examined the term structure and 

the relation to inflation in The US and nine other industrialised countries.  



 

14 
 

This research notes that it is noteworthy that the term structure informs about future 

inflation in other countries. This is critical to this research as the investigation attempts 

to determine if the term structure hold future forecasting analysis about economic 

growth within the domestic country. 

 

Mishkin (1991) examined the explanatory power of the term structure on inflation using 

short term maturities of twelve months or less. It is important to note that Mishkin 

(1991) illustrated that forecast surveys do better using longer term maturities. Mishkin 

(1990) found that for maturities that are over a year, the differential in the term structure 

for The US appeared to be significantly smaller.  

 

This methodological change shows that using the same methodology, the coefficients 

using the longer maturities are above zero and are statistically significant. Mishkin 

(1991) empirical evidence revealed that every country which includes The US, 

Canada, Belgium, France, West Germany, Italy, Netherlands, Switzerland, Japan bar 

the United Kingdom (UK), showed that there is a great deal of valuable information in 

the term structure of nominal rates at the short end of the curve. The finding has had 

a profound impact in that it suggested that most countries can observe data on a 

nominal scale which provides information about the behaviour of the real term 

structure.  

 

Estrella and Hardouvelis, (1991) showed that historically the information reflected in 

the slope of the yield curve are independent of monetary policy and therefore the slope 

provides suitable information to market participants as well as central bankers and 

policy makers. 

 

The flattening of the yield curve in 1988 and subsequent inversion in 1989 according 

to Estrella and Hardouvelis, (1991) has been interpreted by both market participants 

and financial commentators as data that a downturn in economic conditions recession 

is forthcoming. The implication herein is that when the yield curve flattens and 

forecasts a future drop in interest rates, this inevitably forecasts a lower level of GNP.  

 

 

 



 

15 
 

Estrella and Hardouvelis (1991) reviewed the research at the time (Harvey (1988), 

Hardouvelis (1988) and Mishkin (1991) and showed that the recent work of the time 

confirmed that fluctuations in the term spread do forecast the correct course of future 

expected spot rates, however, the authors do show that at the time there was sparse 

practical work of the forecasting of the changes in economic activity. 

 

Through the review of the literature, a number of investigations provided information 

that showed that the term spread has forecasting ability of the path of future economic 

growth. Plosser and Rouwenhorst (1994) were among the first to study the yield curves 

forecasting ability outside the US. The results showed that the yield curve spread 

proved a suitable forecasting tool for information from Canada and Germany however 

the same cannot be said for the UK and France.  

 

Plosser and Rouwenhorst (1994) used monthly data for The US, The UK and Germany 

over the period 1973 to 1988 where the yield spread was calculated between the one 

year term point and the three and five year term point. This approach is similar to the 

approach taken by Jorion and Mishkin (1991) in terms of the term points referenced.  

 

Plosser and Rouwenhorst (1994) make use of a regression model with the aim of 

determining if the yield spread forecasts real economic growth, the results in the paper 

are consistent with that of Estrella and Hardouvelis (1991) in that the slope of the yield 

curve forecasts real output growth in The US. The results are similar for the German 

set of data, but for the UK, the results are weaker due to the finding, as cited by Plosser 

and Rouwenhorst (1994), that inflation was higher and more variable in the UK sample 

which obscures informational content of the yield curve. 

 

Estrella and Mishkin (1995) showed that in many countries the yield curve differential 

are suitable for forecasting recession at least six to eight quarters in advance. Estrella 

and Mishkin (1995) examine the relationship between the term spread, future inflation, 

the central bank rate and future real activity for France, Germany, the UK, Italy and 

The US. Estrella and Mishkin (1995) employ, much like the authors above, a 

regression model. The results are positive and significant for France, Germany, the 

UK and The US with forecast for real GDP being significant from four to eight quarters 

ahead.  



 

16 
 

Italy is shown to be the only country where the results from a statistical point of view 

are significantly lower.  Estrella and Mishkin (1995) reason that the term spread is 

suitable for forecasting future economic growth in developed in economies as the 

evidence suggests, especially is the US, it provides ample warning of a coming 

recession as the yield curve inverts. The evidence holds true for the UK, Germany, 

Italy but the results are statistically weak for the France dataset.  

 

Tzavalis and Wickens (1996) agree with the general finding from previous research 

studies using The US as a baseline that the greater the time horizon used the more 

information that term structure possess regarding future inflation. In summation, 

Tzavalis and Wickens (1996) agreed with the general findings of the body of literature 

in that the greater the time horizon, the more information the term structure would be 

able to provide related to future inflation and growth. Tzavalis and Wickens (1996) go 

further by showing that the forecasting of abilities of the term spread is poor but that 

the real interest rate contains far more suitable and forecasting information about 

future inflation and growth than the term spread.  

 

In light of the practicality of the term spread as a forecasts for real economic growth, 

Bosner-Neal and Morley (1997) show that forecasts of economic activity can be 

unreliable. The authors illustrate that forecasts based on macroeconomic models are 

often let down by the lack of accurate and timeous data Added to this is the 

unnecessary over complication of models used to create forecasts. Bosner-Neal and 

Morley (1997) postulate that there has been a steady growth in interest of using 

financial variables in order to create forecasts. The pros of using financial variables 

are that they are readily available and easy to implement in models but probably more 

important is that they are less prone to error in measurement. This finding is brought 

to light in other studies surveyed in this literature review.  

 

A financial variable that been particularly successful in forecasting real economic 

activity is the yield spread or the difference between the long term and short term 

interest rates. In simple and general terms a positive spread, is one in which the long 

term interest rates are higher than the short term rates, and is usually associated with 

economic expansion.  

 



 

17 
 

The reverse of this situation in which shorter term rates are higher than long term rates 

usually implies that there is a chance of economic contraction. Important to note is that 

the nature and size of the change of the term spread indicates the scale at which 

economic growth expands or contracts, the relationship indicates that greater the 

deviation of the larger the spread the greater  the consequent impact on the real 

economy. 

 

Bosner-Neal and Morley (1997) illustrated that the application of the yield curve spread 

as a variable to forecast real economic activity is well established for data based on 

The US. The authors asked the question whether this is the case in other industrialised 

countries.  

  

Bosner-Neal and Morley (1997) examined the link between yield spread and 

forecasting real economic activity over eleven industrialised countries. The authors 

found empirical evidence that specifies that the yield spread is a statistically and 

economically important forecasting variable of economic activity in several countries. 

Further, Bosner-Neal and Morley (1997) find that the yield curve spread forecasting 

model generally outperforms other economic variable forecasting models used to 

forecast future real economic activity. 

  

In order to understand the relationship between the yield curve spread and real 

economic growth, an understanding of the yield curve and the movements in the yield 

curve is important. This paragraphs below discuss the relevance of the yield curve 

spread and the yield curve as well as discusses the argument behind why the yield 

curve could be reliably used as an economic variable to forecast economic activity. 

 

There are two possible reasons for this empirical relationship between the yield spread 

and economic activity. The first being that yield spread may reflect the position of 

central bankers with respect to the stance of monetary policy. When the monetary 

policy stance is tightened then short term interest rates increase. What follows is that 

long term rate increase but not by the same degree, the result of which is that the yield 

curve spread narrows or could perhaps turn negative.  

 

 



 

18 
 

The result of this tightening is to impart a tightening of economic activity by making the 

cost of borrowing more expensive and thus is the monetary policy stance to cooling 

the economy down. The consequence of this is that a yield spread that happens to be 

negative is generally a precondition to mean a slowdown of future projected economic 

growth. 

 

The alternate approach is that the yield curve reflects the market expectation of future 

economic growth. If the expectation that real income is set to rise in the future, the 

expectation is that there will be an increase in profitable investment opportunities. For 

market participants armed with this knowledge to take advantage of this movement, 

the investment houses increase their borrowing and issue more bonds. The bonds 

issued will tend to be longer term to lock in investors as well as secure long term 

financing. The knock on impact would be an uptick in the supply of longer dated debt 

securities which has the impact of reducing the price of the bonds but subsequently 

increasing the yield.  

 

Rates on longer term debt instruments will therefore increase when compared to short 

term debt instruments rates and the yield curve will steepen but positively so. Should 

the expectation that real income realise to even a slight increase in real economic 

growth, then the positive steepening of the yield curve will be associated with the 

increase of real economic activity into the future, this according to the research of 

Harvey (1988). This point is further elaborated on in other research that was surveyed 

in the literature review. 

 

Research suggests that both theories presented in their own right above have value. 

Estrella and Hardouvelis (1991) as well as Estrella and Mishkin (1995) show that the 

inclusion of other variables to forecast real economic activity is suitable but not as 

suitable and significant as the yield curve.  

 

Thus the result suggest that the yield curve spread reflects more than just the current 

monetary policy stance of central bankers. The higher short term rates contracts 

lending activity which then has the desired impact of contributing to slowing the 

economy down as cited by Bernanke and Blinder (1992).  

 



 

19 
 

The results of the research by Bosner-Neal and Morley (1997) show that the yield 

curve spread is suitable in forecasting economic activity based on the regression of 

the term spread calculated by using the ten year government bond and three month 

Tbill and real GDP. Bosner-Neal and Morley (1997) go on to show the same conclusion 

for in and out of sample results as well, the former showing the more promising results. 

 

The results of Bosner-Neal and Morley (1997) are presented here and contrasted in 

the results section of this research as this research employs the regression 

methodology as used by Bosner-Neal and Morley (1997). The results of Bosner-Neal 

and Morley (1997) can be seen by the bar charts shown below in Figure 3 which 

highlights the GDP response to a one percent change in the yield spread. The bar 

chart visually illustrates the beta coefficients for each country over the respective 

forecast horizons calculated by Bosner-Neal and Morley (1997) based on their 

regression equation where future real GDP is the dependent variable and the term 

spread is the independent variable.  The bars highlighted in blue are statistically 

significant. 



 

20 
 

 

Figure 4 - Beta coefficients for each country for real GDP using the term spread 

Source: Adapted from Source: Adapted from “Does the yield spread forecast real 

economic activity” by Bosner-Neal and Morley (1997) 

 

The results also show another important relationship, the forecasting ability and 

statistical significance reduces as the forecast horizon is expanded. A more up to date 

investigation by Kim and Limpaphaymon (1997) and Davis and Fagan (1997) found in 

their research that the term spread is a suitable forecaster of output growth in six out 

of the nine European countries that were studied.  

 



 

21 
 

Davis and Fagan (1997), Smets and Tsatsaronis (1997) and Canova and De Nicolo 

(2000) provided evidence against the plethora in support of the yield curve forecasting 

ability. Davis and Fagan (1997) find subpar and erratic evidence for forecasting 

performance in the nine euro countries covered in their research. Smets and 

Tsatsaronis (1997) investigated the slope of the yield curve in The US and Germany 

as leading indicators of future output growth.  

 

Smets and Tsatsaronis (1997) found that monetary policy stance is at the centre of 

attraction in determining the intensity of the relationship between the term structure 

and output growth. Smets and Tsatsaronis (1997) make an interesting point indicating 

that the forecasting content of the yield curve is not policy independent. This gives rise 

to the problem of endogeneity in that monetary authorities do attempt to respond to 

market events as they have impacts on assets prices which are themselves based on 

the anticipation of the central banks current and future policy actions. Smets and 

Tsatsaronis (1997) note that the endogeneity does not reduce or invalidate the 

information inherent in asset prices but imply that a central bank would need to use 

this information in a complementary fashion to make its own independent valuation. 

Canova and De Nicolo (2000) found only limited evidence for forecasting using the 

term premium in Germany, Japan and the UK.  

 

 

 

  



 

22 
 

2.2. Forecasting inflation from the term structure  

 

Inflation has been seen as a suitable tool used to forecast economic activity. The 

general premise is as aggregate demand increases and borrowing increases to 

support growth, inflation rises. The research reviewed below illustrates the distinction 

between inflation and forecasting growth. The research presented above showed the 

use of the yield curve as a variable to forecast economic growth. Given the relationship 

between interest rates and inflation and subsequently inflation and GDP, the research 

presented below illustrates the literature of inflation used as a variable to forecast GDP.  

 

Estrella and Mishkin (1997) make the case that since the forward looking debt 

instruments are intently intertwined to the core debt securities then the term structure 

of interest rate can be shown to contain valuable information about the future 

expectations of the term structure in relation to real economic activity and inflation. 

 

Davis and Fagan (1997) use quarterly data over the period of 1986 to 1992 employing 

a univariate auto regression model for Belgium, Denmark, Germany, Spain, France, 

Ireland, Italy, Netherlands and the UK. At a ten percent significance level the results 

for Germany, Italy, Belgium and Denmark shows that out of sample forecasting shows 

very little to no forecasting ability for future inflation and economic growth. 

 

For output growth, the performance of the model is somewhat better with the yield 

curve being found to be suitable in forecasting for Germany, France, Belgium, 

Denmark, Netherlands and the UK. That said however, the forecasting ability is not 

found to be suitable when trying to satisfy the condition of stability using out of sample 

data.  

 

Overall the study by Davis and Fagan (1997) tested whether a handful of leading 

indicators have forecasting ability over future inflation and output growth. The leading 

indicators that the authors looked at are the yield curve spread, which is relevant to his 

study in particular, stock prices, credit spreads and foreign bond yields.  

 



 

23 
 

Overall, the authors find that there are instances where the yield curve does provide 

forecasting ability of future inflation and output growth, however overall the authors 

find no significant evidence that yield spreads in particular are suitable leading 

indicators for future inflation and output growth.  

 

Bernard and Gerlach (1998) build on the work by Fama (1984), Mankiw and Miron 

(1986) and Estrella and Hardouvelis (1991) who showed that the yield curve forecasts 

recessions from using the term spread by as much as two years in advance. Bernard 

and Gerlach (1998) show that while the slope of the yield curve slope is suitable in 

forecasting macroeconomic growth there are in fact three considerations that make 

this especially striking for the purpose for forecasting real economic activity. 

 

The first is that data of the term spreads are instantaneously available. The second is 

that yield curve data is not revised which is a problem that GDP data suffers from. The 

third and probably most insightful is that interest rate data is available for long 

maturities which allows forecasts to be computed for much longer time horizons. 

 

Bernard and Gerlach (1998) provided a cross country approach to evidence the 

practicality of the term spread and hence the probability of recession within the 

following eight countries, Canada, France, Belgium, Germany, Netherlands, Japan, 

the UK and The US over the period 1972 to 1993. Bernard and Gerlach (1998) noted 

the evidence from Estrella and Mishkin (1995) who showed that the forecasting ability 

of the term spread is found to be the highest in Germany followed by the US and 

Canada and the lowest in Japan.  

 

Bernard and Gerlach (1998) show that there are no definitive findings that which 

evidence that economic leading indicators contain information for extended forecast 

horizons of economic activity into the future. Whilst Bernard and Gerlach (1998) do not 

go into why there is a relationship the slope of the yield curve and real economic 

activity, they do offer two leading hypotheses. The first holds that the relationship 

arises from the relationship between term structure and the effect from monetary 

policy.  

 

 



 

24 
 

To illustrate this relationship Bernard and Gerlach (1998) assume that the central bank 

increases rates on short term debt securities and since the action by the central bank 

is temporary, market participants will revise their expectations upward to reflect 

changes in future short term rate by less than the change in real short term rates.  

 

The knock on impact of this is that long term interest rates also raise by less than the 

current real short term interest rate. The impact of this is that the yield curve evolves 

into a downward slopping curve. Given all of this, and the fact that the transmission 

mechanism takes between twelve to eighteen months to show the impact in the real 

economy, the argument that tightening of interest rate policy is then linked with a 

contraction in future growth and hence an escalation in the chance of economic 

downturn, Bernard and Gerlach (1998).  

 

The second hypothesis put forward in the research by Bernard and Gerlach (1998) 

relied on the expectation of market participants and the view of future economic 

growth. Bernard and Gerlach (1998) illustrated that if market expectations of a 

recession start to materialize, inflation will reduce as a result of a lower growth 

environment, and as such the outlook is expectations are predictably likely to lead to 

a decrease in longer term yields and an increase in shorter term yields. Bernard and 

Gerlach (1998) go on to note that it is somewhat of an exercise to meaningfully by way 

of econometric analysis differentiate between the two hypotheses. That said, the 

research by Bernard and Gerlach (1998) focus to add to the current evidence of the 

ability of the term structure to forecast real economic growth. 

    

Bernard and Gerlach (1998) make mention of a point that is stressed by Estrella and 

Mishkin (1995) in that one reason that the term spread is as suitable as the evidence 

suggests in developed in economies, especially is the US, is because it provides ample 

warning of a coming recession as the yield curve inverts. Bernard and Gerlach (1998) 

ask the question, does the same relationship hold in other developing countries. Table 

1 below shows the empirical results from Bernard and Gerlach (1998) when the 

authors tested for estimated probabilities across other countries by using the term 

spread in each of the countries mentioned earlier.  



 

25 
 

Table 1 - Probability of recession four quarters ahead 

 

Source: Adapted from “Does the term structure forecast recessions? The international 

evidence” by Bernard and Gerlach (1998) 

 

An interpretation of the results shows that for a period of four quarters into the future, 

if the term spread differential for Germany for instance, is 4%, the probability of 

recession is 0.00%. However, more importantly, the results do show that as the rates 

go negative the probability of recession starts to increase.  

 

Table 1 from Bernard and Gerlach (1998) show that data for Germany seems to show 

the most forecasting power of the term spread for real economic activity in to the future.  

Research studies show that forecasting real economic activity forms the basis of many 

decisions. This can take the form of businesses relying on such forecasts in order to 

plan their production process. Policymakers make use of such research reports when 

deciding on a path of monetary policy going forward or when planning the national 

budgeting process. Therein lies the true question of the research, the value of the 

choices mentioned above depends on the quality of the forecast.  

 

Over the past couple decades, financial markets have seen integration that has been 

unprecedented over the past fifty years. This is particularly important for yield curve 

analysis as this process of integration has a profound impact for transmission shocks 

across financial assets and the real economy. 

 

 

 

Spread (%) Belgium Canada France Germany Japan Netherlands UK US

4.00 0.10 0.02 0.06 0.00 0.19 0.21 0.05 0.00

3.00 0.18 0.04 0.11 0.02 0.20 0.28 0.09 0.02

2.00 0.29 0.09 0.17 0.07 0.22 0.36 0.14 0.08

1.00 0.43 0.16 0.25 0.20 0.24 0.44 0.21 0.21

0.00 0.57 0.25 0.34 0.41 0.25 0.53 0.30 0.41

-1.00 0.71 0.37 0.45 0.66 0.27 0.62 0.40 0.64

-2.00 0.82 0.51 0.56 0.85 0.29 0.70 0.51 0.83

-3.00 0.90 0.64 0.66 0.95 0.31 0.77 0.62 0.94

-4.00 0.95 0.76 0.76 0.99 0.33 0.83 0.72 0.98

Probability of a recession four quarters ahead as a function of the current spread



 

26 
 

Neumeyer and Perri (2001) research was motivated by examining real interest rates 

and business cycles in emerging markets. The emerging market economies looked at 

are Argentina, Brazil, Mexico, Korea and the Philippines. The evidence presented 

shows that in emerging economies that interest rates act counter cyclically in that they 

lead business cycles, the opposite was found of developed economies. The important 

relationship is that the yield curve has forecasting ability for economic growth. 

Neumeyer and Perri (2001) find that interest rate shocks are an especially important 

factor in explaining business cycles in emerging economies. 

 

Stock and Watson (2003) introduce the principle that yields on debt securities as well 

as tradeable assets contain valuable information related to future economic prospects 

is embodied in the fundamental foundation concepts of macroeconomics. This relation 

was postulated by Irving Fisher in that the nominal interest rate is calculated by taking 

the real interest rate plus the expected inflation value. If a central bank tightens 

monetary policy by increasing short term interest rates, an inverted yield curve 

generally leads to future economic slowdown. 

 

Stock and Watson (2003) reviewed fifteen years’ worth of research studies that have 

considered forecasting economic activity using asset prices. The important 

consideration is that Stock and Watson (2003) define asset prices as including interest 

rates or more specifically the differential between the short term and long term interest 

rates. During the 1970s and 1980s there were many studies that employed monetary 

aggregates to forecast inflation and output growth.  

 

Stock and Watson (2003) on close examination show that the GDP data for the USA 

successfully forecasted by the short term rates rather than the term spread itself. Stock 

and Watson (2003) examined various leading indicators and examined the data 

leading up to the US 2001 recession, Stock and Watson (2003) showed that the term 

spread did turn negative in advance of the recession. 

 

The fundamental issue, as highlighted by Stock and Watson (2003), with using 

monetary aggregates is that there are constant revisions to the aggregates. In stark 

contrast, tradeable assets and tradeable debt securities are observed in real time and 

there is hardly a revision of these results which proves exceptionally beneficial to 

econometric studies in that the measurement error is reduced greatly. 



 

27 
 

The analysis by Stock and Watson (2003) employed quarterly data for forty three 

variables from France, the UK, Japan, Italy, the US, Canada and Germany for the 

period 1959 to 1999. The evidence suggests that asset prices prove to be more 

suitable in forecasting future output growth than for inflation. The basis of the review 

of the research by Stock and Watson (2003) is centred on the evidence provided on 

using the term spread and the relationship to output growth. Arnwine (2004) adds to 

this by showing that lower long term rates may reflect lower real yields due the 

expectations of the slower output growth. This then leads to an inversion of the yield 

curve. 

 

With the body of evidence presented above, the bulk of which is on developed 

economies, the evidence shows that interest rate shocks can explain economic 

business cycles in developed economies. The evidence summarized the performance 

of real GDP growth and the four quarter horizon. The results as given evidence before, 

are consistent with previous studies. The forecast based on the term spread are of 

particular importance. The results show that, consistent with the literature, the term 

spread is suitable in its forecasting in the first period for the US and Germany, however 

the forecast improved in the second period in Canada and Japan but not for France, 

Italy the UK and the US. 

 

Stock and Watson (2003) show that there are additional studies which provide a more 

holistic view of whether the yield curve as a forecaster of economic activity holds in 

other countries. Harvey (1990), Hu (1993), Bosner-Neal and Morley (1997) and others 

generally all conclude that the yield curve spread has forecasting content for real 

economic activity in countries other than the US.  

 

Uribe and Yue (2005) show through their research that country spreads and business 

cycles are interrelated in emerging economies. Uribe and Yue (2005) found that 

country spreads do drive business cycles in emerging economies, however the 

evidence of the effects are fairly muted over the period of consideration. Specifically, 

Uribe and Yue (2005) find that if country spreads in emerging market economies were 

independent of The US interest rate then variance of the economic activity explained 

by the domestic yield curve drops by sixty six percent.  

 



 

28 
 

This finding is related to the interconnectedness of global economies but more so to 

the fact that developing nations respond more to state of developed nations as a result 

of the balance of the power over global trade that developed economies have due to 

globalization.    

 

The paper examines the term structure of interest rates, monetary policy and the 

consequent impact on real activity in Europe and The US. The yield curve is thus a 

reliable measure that can be used as one piece of suitable information which, along 

with other leading indicators, can be used to assist in directing the course of central 

banks decision making.  

 

Uribe and Yue (2005) show that the business cycle in emerging market economies are 

correlated with the cost of financing that those countries face in the financial markets. 

This is typically shown through a combination of interest rates and economic 

production. The results showed that over the period of 1994 to 2001 periods of low 

interest are associated with economic growth and period of high interest show 

economic decline. This finding is intuitive as lower cost of borrowing makes it cheaper 

for business to expand or pay down outstanding debt.  

 

It is generally conceded to contain some information that may be of use to both market 

participants and to the monetary authority. This paper builds on the research by Uribe 

and Yue (2005) by examining those associations in a multi-country methodology.  

 

Does a Central Bank wield enough influence to directly impact the yield curve and 

thereby influence the yield curve term spread through intervention via the short term 

instrument? The answer, more often than not for credible and independent Central 

banks is that they can directly impact the short end of the yield curve and that to a 

large and impactful manor. The long end, however, will be determined by many other 

considerations, including long term expectations of inflation and real economic activity.  

 

 

 



 

29 
 

The term structure contains valuable information which relates to the markets 

expectations of where future economic growth will go as well as the projection of 

inflation. There are other building blocks that make up a yield, such as credit risk 

premium and liquidity premiums, however these inputs create more noise and make 

the isolation of the future economic expectations and inflation difficult to zero in on.  

 

The out of sample results of using the term spread to forecast economic growth has 

shown more promising results than the results from research using a host of other 

economic indicators to forecast economic growth.  

 

A worthwhile contraction of monetary policy should have varying impacts on both the 

short end debt securities as well as the longer term debt instruments. The major impact 

of a contractionary monetary policy stance is a reduction in the need for borrowing by 

consumers at the short end of the curve. The long end of the curve is determined more 

by expectations of what future inflation and economic growth will look like. If the 

monetary policy intervention is deemed to be worthy this will have an impact of a 

lowering the expectation of long term inflation.  

 

The combined result is that the change in long term rates tends to change by less than 

the change in the short term rates and subsequent the term spread declines. The case 

study suggests that an increase in the monetary authority base rate is probably going 

to have an impact on the yield curve spread if the intervention by the central bank is 

credible and does not lead to the perceived expectation by market participants and 

consumers of further intervention in the near term.  

 

Furthermore, they show that central bank credibility may be an important factor in 

determining the estimated responses of the spread to a change in the central bank 

rate. More important is that Estrella (2005) showed that the yield curve has forecasted 

almost every post war recession in the US.  

 

Interestingly, Estrella (2005) showed that the extent to which the yield curve is a good 

forecaster of future GDP is dependent on the form of monetary policy that is adopted 

by the in country central bank.  



 

30 
 

While this is not within the scope of this research, the form of monetary regime and the 

impact that has on the yield curve and subsequently GDP growth could be followed up 

for future research in emerging economies. 

 

Literature shows that asset prices, interest rates, dividend yield and exchange rate as 

forecasters of growth and inflation, Burn and Mitchell (1935), Stock and Watson (2003) 

have provided the most suitable for forecasting the slope of the yield curve according 

to Mehl (2006). The forecasting of the yield curve has come into much focus due to 

the inversion of The US yield curve. The important consideration is that there is 

evidence to suggest that the inverted yield curve is a signal of a recession Mishkin 

(1991)  

 

The paper by Mehl (2006) investigates the degree to which the slope of the yield 

forecasts domestic inflation and growth, the distinguishing point of research with this 

paper is that it referred to emerging market economies.  

The bulk of the evidence and research in the field of yield curve slope has been done 

primarily on developed economies.  

 

Mehl (2006) noted at the time of publication, international evidence has remained 

scarce to a limited number of industrialised countries. The evidence for emerging 

economies at the time was nil. This was primarily due to domestic bond markets in 

emerging economies only starting to deepen since the turn of the millennium, Mehl 

and Reynaud (2005). On a high level, the research found evidence to support the 

finding that that domestic yield curves within EM economies has forecasting content 

that can be reliably used in some geographies to forecasting economic growth and 

resultant inflation.  

 

Mehl (2006) sampled fourteen emerging market economies with the purpose of 

investigating the appropriateness of the domestic yield curves in each country to 

forecast growth and inflation over the past ten years. The paper by Mehl (2006) 

contributed to existing literature by investigating the practicality of the slope of the yield 

curve as a forecaster of growth and inflation across fourteen emerging economies - 

this is the main interest of this paper.  

 



 

31 
 

The other areas contributed by the author related to whether the yield curve spread in 

the US or the European Union contribute in some way to forecast growth and inflation 

in the fourteen emerging market economies. Mehl (2006) also tested whether there is 

any information contained in the domestic yield curves that can be explained by The 

US and or Euro area. Lastly, Mehl (2006) investigated whether the movements in the 

domestic yield curves are country specific or whether there are other variables that 

can impact the domestic yield curve movements.    

 

There is research which showed empirical evidence that the inversion of the slope of 

the yield curve signals a coming recession This dates back to research by Mishkin 

(1990) as well as Hardouvelis (1991). Mehl (2006) outlined the economic rationale 

behind this - the standard economics understanding is that the slope of the yield curve 

is directly related to monetary policy. As the central bank tightens interest rates, the 

short term interest rates are high relative to the longer term rates.  

 

Mehl's (2006) evidence found across the fourteen emerging market economies, and 

using the methodology prescribed by Stock and Watson (2003), that on average a one 

hundred basis point steepening created the grounds whereby, inflation and lagged 

growth are expected to accelerate by thirty basis points a year ahead. It is important 

to present this evidence here as this is includes results for developing economies, 

some of which are examined in this research. 

 

The results of Mehl (2006) show that the slope of the yield curve in emerging 

economies is found to have information content for future inflation and output growth. 

The results of the slope of the yield curve, which is used as a forecaster of domestic 

inflation and production growth, are summarized below. 

 

The results in Mehl (2006) are reported for all of the thirteen countries from a six month 

forecast horizon to a two year forecast horizon with the interval between horizons being 

six months. The corresponding betas area also shown for lagged inflation as well as 

production growth for a one hundred basis point steepening of the domestic yield 

curve.  

 



 

32 
 

A one hundred basis points steepening of the US yield curve observed over the two 

year forecast horizon is linked with a projected acceleration in inflation by around forty 

basis points over the next six months for The US. The growth response is fairly robust 

in The US with two hundred and seventeen basis point response in relation to a 

steepening of the yield curve by one hundred basis points.  

 

The results for the countries that are examined for this research show that for a one 

hundred basis points steepening would result in an eighty eight and eighty five basis 

points increase in inflation and production growth respectively. India shows a converse 

result with a one hundred and seventy basis points increase in inflation but a ninety 

seven basis points reduction in production growth. The result set for South Africa 

shows that the inflation response is sixty six basis points increase and a one hundred 

and thirty seven basis points increase in production growth. 

 

Similarly, the study by Evgenidis and Siriopolous (2016) added additional evidence 

supporting the finding by Harvey (1988) and Estrella and Hardouvelis (1991) in that 

the yield curve provided suitable forecasting of output growth and inflation from one 

quarter to up to two years in advance. It must be noted that the studies mentioned 

above are all based on information that is specific to The US. Subsequently, there have 

been a number of studies that has investigated whether the yield curves forecasting 

ability is applicable in other countries.  

 

Evgenidis, Papadamou and Siriupoulos (2018) showed what many other research 

papers before them have shown in that policy makers need to have the ability to 

influence the expectation over short and medium term expectations. However, in order 

to execute on this ability, central bankers require a solid foundation and knowledge of 

where the economic trajectory is actually headed. In layman’s terms, central bankers 

need the ability to use reliable forecasts and one of those reliable forecasting tools that 

is most widely used in gaining a feel for the trajectory of the economy is the 

implementation and subsequent analysis of the difference between the difference on 

long and short term treasury securities.  

 

 



 

33 
 

Evgenidis, Papadamou and Siriupoulos (2018) surmise the essence of using the yield 

spread to forecast the trajectory of the economy to be particularly successful in 

forecasting the course of the economy particularly inflation and output growth. 

Evgenidis, Papadamou and Siriupoulos (2018) note that the ability of the yield curve 

to forecast inflation and output growth using the term spread has become a stylized 

fact and remains today as a powerful tool used by central bankers to forecast the 

course of the economy. 

 

That said, it must be noted that there is literature where Evgenidis, Papadamou and 

Siriupoulos (2018) cite Dotsey (1998), Venetis and Paya (2005) and Hamilton and Kim 

(2002) whose research attempted to determine the effectiveness of forecasting 

inflation and output growth using the yield curve but whose results range from 

statistically positively significant to negative consequences.  

 

The authors show that in some countries the ability of the yield curve to forecast output 

growth is strong whereas is other countries the evidence is muted at best. The work of 

Evgenidis, Papadamou and Siriupoulos (2018) provided a view of the preeminent 

research in this field in order to gain a complete view of the literature available on the 

yield curves ability to forecast economic output and inflation. The sample periods cover 

a long period which includes multiple economic shocks such as the Asian Crisis, the 

introduction of the Euro, the Dot Com bubble and the Global Financial Crisis. 

 

Harvey (1988) and Estrella and Hardouvelis (1991) were amongst the first to study the 

impact of the forecasting ability of the yield spread on forecasting economic output and 

inflation.  

 

Evgenidis, Papadamou and Siriupoulos (2018) go further by reviewing studies within 

the previous ten years which studies the forecasting ability of the yield spread and 

whether it has increased or decreased over time. Dotsey (1998), for example, found 

that the yield spreads forecasting power has reduced since the mid 1980’s. Stock and 

Watson find similar evidence in the US from 1985 onwards. To expand on this article, 

Benati and Goodhart (2008) examined data for the US, UK, Euro area, Australia and 

Canada and find that the yield spreads forecasting power has considerably reduced 

after the 1908’s and up until 2005. 



 

34 
 

Evgenidis, Papadamou and Siriupoulos (2018) focused their analysis on the country 

coefficients to paint a clearer picture of where the yield spread is suitable as an 

economic variable for future output growth. The results are depicted in the table below 

which highlights by the sign and the significance of the coefficients that the ability of 

the yield spread to forecast future economic output is valid in the US, Canada and 

Europe. Japan’s coefficient estimates are insignificant and the coefficient estimates for 

Australia are negative. This indicates that the yield spread is not as suitable as an 

economic variable to be used as a forecaster of future economic growth. 

 

Furthermore, the results in the table below notes the differences in magnitude and the 

significance of the coefficients through the decades. The results suggest that from the 

early 2000s to the 2007 Global Financial Crisis, the coefficients have been significantly 

positive. Going back further, the results show that through the 1970s, 1980s and the 

1990s the coefficients were insignificant. This lead the authors to postulate that the 

yield curves ability to forecast future output growth is time varying in nature. 

 

The paper finds that there are five main points that evolve from the investigation. The 

first being that US interest rate shocks explain about twenty percent of the average 

economic activity in emerging economies, second that in-country spread shocks 

explain about twelve percent of the impact on the business cycle in emerging 

economies, third that once the US raises rates the emerging market economies rates 

first fall followed by a delayed overreaction. Fourth that US interest rate shocks impact 

domestic economic variables mostly through the yield spread and finally, that the 

emerging market feedback on term spreads significant exaggerate business cycle 

fluctuations. 

 

The question can be posed as to why a relationship between the yield curve spread 

and GDP should exist. The first plausible reason is that monetary policy is a key 

determinant that goes into the construction of the term structure, especially the short 

end of the curve, as well as an indicator for future real activity. Contractionary monetary 

policy has the impact of flattening the curve which would then lead to a slowdown of 

borrowing which then has the impact of leading to a slowdown of economic activity.  

 

 



 

35 
 

In several research studies, the yield curve has been shown to be a suitable and 

accurate economic variable to for forecast economic activity and subsequently 

inflation, with forecasting periods ranging from one to two years for economic activity 

and extended forecast horizons for inflation. This supports the evidence found by 

Smets and Tsatsaronis (1997) as their research found that the forecasting content of 

the yield curve is not policy independent. 

 

Although the resulting statistical significance tend to be somewhat consistent across 

the countries, the important consideration is that the economic significance is shown 

to vary considerably across the different nations. This relationship is brought to light in 

that the six quarters ahead results the coefficient varies between 0.35 and 0.62. Thus 

if the yield curve spread experiences a change of one percentage point increase this 

will mean that this is linked to an average annualized six quarter real GDP growth 

between thirty five and sixty two basis points higher.  

 

Overall, through the analysis of the research reviewed for this study, it is clear that the 

yield curve is a suitable leading indicator for output growth, especially in The US as 

early research was focused. The research then opened to other developed countries 

which the research reviewed for this study showed that the yield curve in general is a 

suitable indicator for forecasting future economic growth.  

 

The question then follows, does the same relationship between the yield curve spread 

and economic activity hold for developing economies? Mehl (2006) provides detailed 

information of the yield curve as indicators for lagged inflation and industrial output for 

emerging economies. Bosner-Neal and Morley (1997) provide a suitable, simple and 

effective regression model which is used in this research with the change of using 

emerging market economy data for Brazil, Russia, India, China and South Africa. The 

US is examined as well as a base case for the relationship as shown by Bosner-Neal 

and Morley. 

 

 

 

 

 



 

36 
 

3. Description of data and research methodology 

 

The research will be conducted by making use of secondary, longitudinal GDP, three 

month Tbill and ten year government bond data for the period 1980 quarter two to 2020 

quarter two. The preeminent studies and previous research on the forecasting strength 

of the yield spread in The US as well as developed economies has focused on the 

spread between the three month Tbill and the ten year government bond.  

 

This study will replicate the methodology of Bosner-Neal and Morley (1997) and 

employ a regression model which is detailed in section 3.1. The regression will focus 

on the beta coefficient as this will show the nature and the rate of change which 

illustrates to what degree GDP growth changes for a one percentage point change in 

yield spread. As such, the purpose of this analysis is to determine if the forecasting 

power of the yield curve in emerging economies holds using the same term points to 

calculate a spread. A further reason for this is that the three month Tbill represents the 

borrowing costs in the near team and the ten year in the long term.  

 

Research data will be sourced from online repository data in which the researcher 

makes use of secondary data. Secondary data, which is described by Saunders and 

Lewis (2012), is data that has already been collected from other sources and is readily 

available. Quarterly data sampling is consistent with the approach taken by Bosner-

Neal and Morley (1997) as well as Mehrotra and Sanchez-Fung, (2011).  

 

  



 

37 
 

3.1. Theoretical framework 

 

Estrella and Hardouvelis (1991) showed through their research which employed the 

use of the three month Tbill rate  and the ten year government bond rate  was used to 

calculate the term spread by subtracting the long term rate from the short term rate. 

This research will follow the same approach in calculating the term spreads for the 

following countries, the US, South Africa, Brazil, Russia, China and India. 

 

Mehl (2006) does show that due to globalisation, financial markets have become more 

interconnected but more importantly, that financial markets, and specifically bond 

markets in emerging market economies, have also developed at an astonishing pace. 

It is the aim of this study to employ the methodology developed by Bosner-Neal and 

Morley (1997) and to use the term points of the three month Tbill and ten year 

government bond as described in the seminal work by Estrella and Hardouvelis (1991). 

 

The framework employed in this research will adopt the methodology applied by 

Bosner-Neal and Morley (1997) and is shown below as equation 1 in section 4. where 

the dependent variable is the percentage change in real GDP: 

(Y)i = β0 + β1X1 + error 

Where: 

• Yi is the dependent variable 

• β0 is the sample intercept 

• β1is the slope coefficient and X1 is the independent variable 

• error is the random error term 

 

The hypothesis is centred on the assumption that the yield spread between the three 

month Tbill and the ten year government bond has been shown to have forecasting 

power for future economic growth in developed economies. Therefore this research's 

aim is to determine whether the same informational content holds for emerging market 

economies.  

 



 

38 
 

3.2. Data 

 

The basis of this investigation makes use of the following data: The three month Tbill 

rate which represents the short term rate and the ten year government bond rate which 

represents the long term rate. As defined by many of the research studies that were 

reviewed earlier in the literature review, the term spread is calculated by subtracting 

the long term government bond yield from the three month Tbill rate. The last economic 

indicator that will be consumed in this research study is the real GDP numbers for the 

countries in question.   

  

3.3. Variables 

 

The variables listed below are the key inputs into the regression model from equation 

1 mentioned above. The topic of which points on the yield curve to use varies. 

Examples such as the research by Fama (1984) and Harvey (1988), used different 

points in the yield curve, predominantly shorter term maturities. As the research in the 

field developed, the focal point became the spread differential between the ten year 

government bond and the three month Tbill.  

 

Estrella and Hardouvelis (1991) were amongst the first to employ the ten year 

government bond, three month Tbill spread. The same term point were also referenced 

in Bosner-Neal and Morley and it remains to this day an integral term spread indicator 

that market participants frequently reference. When choosing the yields of debt 

securities, the critical factor comes down to choosing securities which are actively 

traded. This research followed suit of Estrella and Hardouvelis (1991) and Bosner-Neal 

and Morley (1997) in an attempt to balance both comparability across the countries as 

well as the availability of the data. 

 

As mentioned above in the literature review, yield curve data proves to be a much 

more reliable and better economic data point since the yield curve data is 

instantaneously available and the data is not revised. That said, this is based on the 

fact that the bulk of the studies that have been done in the field have been on the US 

and first world countries, all of whom have deep and liquid financial markets which 



 

39 
 

provided previous authors with an accurate and large data set that goes back several 

decades.  

 

As for the measure of economic activity to use, this research used real GDP since this 

is the primary measure by which economic activity is measured and for the countries 

studied, this is the measure that was readily available.  

Three month Tbill rate, ten year government bond rate 

Real GDP growth measured as quarter on quarter growth, yearly numbers  

Term spread (calculated by the difference between the ten year government bond yield 

and the three month Tbill). 

 

3.4. Limitations 

 

Early research into the forecasting ability of the term spread employed the use of short 

term maturities such as used in the work of Fama (1984) and Harvey (1988). As the 

research into the field evolved, other term points on the yield curve were used. The 

basis of using the three month Tbill and ten year government bond rate was primarily 

based on the availability of the data as well as the fact that the term points reflect debt 

instruments that are actively and frequently traded. In countries where debt securities 

are not actively traded in a deep liquid market, a substitute should be found. Similarly, 

in countries where the three month short term debt security bill rate is not available, a 

substitute for the three month rate needs to be found - one such rate could be three 

month forward rate notes or negotiable certificates of deposit with the same maturity.   

 

Furthermore, it must be noted that due to the impact of globalization and the impacts 

developed economies have on developing economies through interconnectedness of 

financial markets, the impact that developed economies yield curve may have on the 

developing markets is ignored entirely. This effect has been studied by Mehl (2006). 

The specific limitations related to this research arise due to countries not having deep, 

liquid and longstanding bond markets for history to develop for the ten year 

government bond and three month Tbill. India, China, Brazil and Russia are the 

countries impacted. In light of this limitation, the research used data as far back as 



 

40 
 

possible so as not to use other proxies to keep to the methodology of Bosner-Neal and 

Morley (1997).  



 

41 
 

3.5. Hypotheses 

 

3.5.1 H0 – The term spread has not been optimal in forecasting future real 

economic growth one year ahead 

 

3.5.2 H1 – The terms spread has been optimal in forecasting future real 

economic growth one year ahead 

 

3.5.3 H0 – The term spread has not been optimal in forecasting future real 

economic growth two years ahead 

 

3.5.4 H1 – The terms spread has been optimal in forecasting future real 

economic growth two years ahead 

 

3.5.5 H0 – The term spread has not been optimal in forecasting future real 

economic growth three years ahead 

 

3.5.6 H1 – The terms spread has been optimal in forecasting future real 

economic growth three years ahead 

  



 

42 
 

4. Results  

 

The section below will outline the regression statistics that were produced using the 

regression model outlined by equation 1. The regression was run on all five BRICS 

countries as well as The US. Following on from the regression statistics, the analysis 

will then turn to a results section. 

 

To recap, the literature surveyed has predominantly examined The US and the impact 

of the forecasting ability of the yield curve on future GDP as well as inflation. Harvey 

(1988), Hardouvelis (1988), Estrella and Hardouvelis (1991), Bosner-Neal and Morley 

(1997) and Mehl (2006), to name a few, found evidence that the yield curve is a 

valuable variable for forecasting future GDP. As the maxim that the yield curve 

provided forecasting information about GDP gained momentum, studies were then 

expanded to other developed economies such as the UK, Germany, France, Italy and 

the Netherlands. The results of the expanded research also confirmed that the yield 

curve proves to be a suitable variable in forecasting the path of GDP and inflation.  

 

The purpose of this research is simply to determine whether the yield curve holds 

informational content which can be used to forecast the path of real GDP. Mehl (2006) 

provided a more up to date and econometrically complex methodology, the research 

focused on the lagged response of inflation and production output. On review of the 

methodology used by Bosner-Neal and Morley (1997), the appeal delivered by the 

simplicity, efficiency and replicability of the regression model was what ultimately made 

the decision to use the model. Moreover, the results presented by Bosner-Neal and 

Morley (1997) show impacts on real GDP, which is another reason for employing the 

regression model. Mishkin (1991) indicates that there is more informational content in 

longer term maturities which further supports the use of the ten year government bond. 

 

At this point it is important to stress that the purpose of this research is to determine 

whether the domestic yield curve in the EM countries chosen holds informational 

content related to forecasting the path of real GDP. That said, the research does 

include the results for The US as a baseline. 



 

43 
 

The framework employed in this research will adopt the methodology applied by 

Bosner-Neal and Morley (1997) and is shown below as equation 1 where the 

dependent variable is the percentage change in real GDP: 

(% change in real GDP)t, t+k = α + β * (spread)t  + error 

Where: 

• Percentage change is real GDP is defined as the year on year change from the 

quarterly real GDP number in the previous year to the current year 

• Spread is the difference between the ten year government bond and the three 

month Tbill 

• α designated Alpha is the intercept value 

• β designated Beta measures the change in real GDP for a given change in the 

yield spread 

 

4.1 Regression Results  

 

This section outlines the statistical results of the regression model outlined in section 

3.1. The results show quite a bit of information returned by the excel regression 

function. The key numbers to pay attention to are the R-square, Beta coefficients and 

p-values of the beta coefficients for the five BRICS countries and The US. 

Regression results for The Unites States of America are outlined below. 

Table 2 - Regressionstatistics for The United States of America 



 

44 
 

 

 

The regression statistics above illustrates the results for The US. For the purpose of 

this research the focus is on the R-square, Beta coefficients and p-values of the beta 

coefficients. The R-square numbers produced by the regression model using 

regression equation 1 consumes the real GDP quarter on quarter annualised growth 

number as well as the yield spread which is the differential between the ten year 

government bond rate and the three month Tbill rate. The GDP variable in this analysis 

is the dependent variable and the term spread is the independent variable. 

 

The R-squared result provides a signal of the informational content  of the term spread 

for real GDP growth. The Beta coefficient from equation 1 estimates the rate at which 

real GDP growth changes following this particular research, a one percentage-point 

change in the yield spread. To interpret the results of the model it is important to reflect 

that a positive value of the Beta coefficient from a statistical perspective would infer a 

positive linear relationship between the term spread and future economic growth.  

 

SUMMARY OUTPUT

Regression Statistics One year ahead Two years ahead Three years ahead

Multiple R 0.167730404 0.154401273 0.160177699

R Square 0.028133488 0.023839753 0.025656895

Adjusted R Square 0.024084045 0.019755401 0.021563017

Standard Error 4.410424485 4.417286905 4.421940624

Observations 242 241 240

ANOVA One year ahead

df SS MS F Significance F

Regression 1 135.1415796 135.1415796 6.947494472 0.008940788

Residual 240 4668.442594 19.45184414

Total 241 4803.584174

ANOVA Two years ahead

df SS MS F Significance F

Regression 1 113.8910906 113.8910906 5.836850046 0.01644471

Residual 239 4663.469241 19.5124236

Total 240 4777.360332

ANOVA Three years ahead

df SS MS F Significance F

Regression 1 122.5448192 122.5448192 6.267136325 0.0129708

Residual 238 4653.747014 19.55355888

Total 239 4776.291833

One year ahead

Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept 2.070949958 0.455405877 4.547481849 8.61742E-06 1.173846999 2.968052918 1.173846999 2.968052918

X Variable 1 0.621494095 0.235788653 2.635810022 0.008940788 0.157014585 1.085973605 0.157014585 1.085973605

Two years ahead

Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept 2.165598911 0.457637362 4.732128737 3.80465E-06 1.264081039 3.067116784 1.264081039 3.067116784

X Variable 1 0.571348252 0.236489376 2.415957377 0.01644471 0.105478511 1.037217992 0.105478511 1.037217992

Three years ahead

Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept 2.132333101 0.460137002 4.63412656 5.90724E-06 1.225871706 3.038794495 1.225871706 3.038794495

X Variable 1 0.594049794 0.237294831 2.503424919 0.0129708 0.126583359 1.061516229 0.126583359 1.061516229



 

45 
 

In essence, this can be distilled into the following relationship, the bigger and more 

positive the spread between the ten year government bond spread and the three 

month Tbill, the stronger the real GDP growth will be into the future. 

 

Estimates of the Beta coefficients from equation 1 provide insight into economic 

significance of the yield curve as being a reliable economic variable to forecast future 

real economic growth. Specifically, what is meant is that the Beta coefficient measures 

the change in real GDP growth for a given, in this case a, one-percentage-point, 

change in the yield spread. The beta coefficients produced for The US for the one, two 

and three year ahead forecasts compare favourably with the results shown by Bosner-

Neal and Morley (1997) in that they are significant and show a steady decline in the 

Beta values as the forecast goes further out into the future. 

 

The next section delves in to the result set of South Africa’s regressiojn statistics from 

the regression model outlined in equation 1. Following this, the rest of the countries 

results are summarized accordingly. 

 

 

 

 

 

 

 

 

  



 

46 
 

Table 3 - Regression statistics for South Africa 

 

 

The Regression results of South Africa are presented above. The R-squared 

coefficients are poor and do not necessarily provide the best reading in terms of the 

forecasting power of the model over the forecasting horizon of the one, two and three 

year ahead values being 3.6%, 1.1% and 0.01% respectively.  

 

The Beta coefficient estimates for South Africa, however, reveal results that are 

inconsistent with that of the literature surveyed as well as that of Bosner-Neal and 

Morley (1997). The Beta coefficient in the one year ahead regression is statistically 

significant with a p-value that is les that 0.05. The p-values for the two and three year 

ahead regressions are both above 0.05.  

 

A one-percentage-point change in the yield spread today is associated with a 0.14 

percentage-point change in real GDP growth one year ahead, an annualized 0.082 

percentage-point change in growth two years ahead, and an annualized 0.065 

percentage-point change in real GDP growth three years ahead.  

 

SUMMARY OUTPUT

Regression Statistics One year ahead Two years ahead Three years ahead

Multiple R 0.19210196 0.108032182 0.085895935

R Square 0.036903163 0.011670952 0.007378112

Adjusted R Square 0.030729465 0.005294636 0.000932515

Standard Error 1.846159265 1.871216441 1.879663831

Observations 158 157 156

ANOVA One year ahead

df SS MS F Significance F

Regression 1 20.37307397 20.37307397 5.977481407 0.0156028

Residual 156 531.6954287 3.40830403

Total 157 552.0685027

ANOVA Two years ahead

df SS MS F Significance F

Regression 1 6.408914633 6.408914633 1.830359668 0.178055173

Residual 155 542.7249001 3.501450969

Total 156 549.1338148

ANOVA Three years ahead

df SS MS F Significance F

Regression 1 4.044291608 4.044291608 1.144674724 0.286340729

Residual 154 544.1029619 3.533136116

Total 155 548.1472535

One year ahead Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept 0.282475313 0.166881513 1.692669894 0.092514502 -0.047163655 0.612114281 -0.047163655 0.612114281

X Variable 1 0.146045278 0.059734936 2.444888833 0.0156028 0.028051605 0.264038951 0.028051605 0.264038951

Two years ahead Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept 0.357856624 0.169147773 2.115644902 0.035973348 0.02372429 0.691988959 0.02372429 0.691988959

X Variable 1 0.082926977 0.061295362 1.352907856 0.178055173 -0.038155093 0.204009047 -0.038155093 0.204009047

Three years ahead Coefficients Standard Error t Stat P-value Lower 95% Upper 95% Lower 95.0% Upper 95.0%

Intercept 0.373935936 0.170190815 2.197156976 0.029503166 0.037726014 0.710145858 0.037726014 0.710145858

X Variable 1 0.065902216 0.061596916 1.069894726 0.286340729 -0.055781757 0.187586189 -0.055781757 0.187586189



 

47 
 

South Africa’s dataset is the only emerging economy country that has history equal to 

the full range of the analysis and the results produced are not conclusive to determine 

that the yield curve is a suitable variable to forecast future economic growth. The South 

Africa results is in contradiction with Tzavalis and Wickens (1996). Tzavalis and 

Wickens (1996) show a general finding from a number of research studies using the 

Unites States as a baseline that the greater the time horizon used the more information 

that term structure possess regarding future GDP and inflation. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  



 

48 
 

Table 4 - Regression statistics for India 

 

 

When comparing the result of India, the R-squared coefficients again d