THE RELATIONSHIP BETWEEN THE US TREASURY YIELD AND

ECONMIC GROWTH

A Project

Presented to the

Faculty of

California State Polytechnic University, Pomona

In Partial Fulfillment

Of the Requirements for the Degree

Master of Science

In

Economics

By

Bofei Chen

2020
 SIGNATURE PAGE

PROJECT: THE RELATION BETWEEN THE U.S. TREASURY

YIELD AND ECONOMIC GROWTH

AUTHOR: Bofei Chen

DATE SUBMITTED: Spring 2020

Department of Economics

Dr. Carsten Lange

Thesis Committee Chair
Economics

Dr. Bruce Brown

Economics

Dr. Craig Kerr

Economics

ii
 ACKNOWLEDGEMENTS

Thanks to my kind and generous professor Dr. Carsten Lange. Without his

abundant knowledge in Economics, this paper is hardly to be done by my own.

iii
 ABSTRACT

The term structure of interest is regarded as an important indicator of financial

activity. The yield curve can be drawn according to the maturity of Treasury bonds.

The different shapes of the yield curve represent different economic implications.

Under normal circumstances, the yield curve is upwards, indicating that yields on

long-term bonds are higher than those on short-term ones. Conversely, if the interest

rate on a short-term bond is higher than that on a long-term bond, it will be an

inverted shape, indicating a recession; a flat yield curve shows that interest rates on

long-term and short-term bonds are closely correlated. By studying the relationship

between term structure of interest rate and economic growth, we can further predict

the ability of yield curve to predict economic activity.

This paper studies the forecasting ability of the U.S. Treasury yield curve to

economic growth by using the linear model and analyzes the factors that may affect

the performance of the yield curve model. Specifically, the research mainly simulates

the yield curve based on the linear method of estimating the yield, and further tests the

prediction ability of the yield curve for economic growth. The study found that the

yield curve has the ability to predict the economic activity of GDP growth, especially

in the range of long-term forecast.

iv
 TABLE OF CONTENTS

SIGNATURE PAGE...................................................................................................... ii

ACKNOWLEDGEMENTS.......................................................................................... iii

ABSTRACT.................................................................................................................. iv

LIST OF TABLE...........................................................................................................vi

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

2 Literature Review........................................................................................................ 3

3 Statistical Model..........................................................................................................8

4 Empirical Results.......................................................................................................11

4.1 Data..................................................................................................................11

4.2 Empirical Results............................................................................................ 12

5 Conclusion.................................................................................................................15

Reference......................................................................................................................17

v
 LIST OF TABLE

Table 1: Descriptive statistics of returns between Q1 1986 and Q4 2017............ 11

Table 2：Descriptive statistics of growth rate of GDP between Q1 1986 and Q4

2017................................................................................................................12

Table 3：The results of OLS regression analysis of real GDP.............................13

Table 4: OLS regression analysis results of real GDP after the increase of

3-month Treasury yield.................................................................................. 15

vi
 1 Introduction

In order to better promote the development of the economy, countries around

the world are attaching more and more importance to the intervention ability of the

economy. The precondition for a country to make economic policies is to accurately

judge the trend of economic changes. Therefore, an effective method to predict

economic development is an important prerequisite for a country's economic

development.

As one of the most important price variables in financial market, interest rate

can be used as the main observation variable for monetary authorities to make and

implement monetary policies. In the economic system, various economic variables

directly or indirectly affect the interest rate, so the formed term structure of interest

rate also contains a variety of information of economic variables. The Treasury yield

curve is the quantitative relationship between the annual yield to maturity of a group

of Treasury bonds and their remaining maturity at any given time. Interest term

structure is considered as an important indicator of financial activity. Money, banking,

and financial markets is a book jointly written by Stephen and Kermit. Chapter 6,

“term structure and risk structure of interest rates”, explains that different shapes of

the yield curve indicate different economic implications. In general, the yield curve is

upward, suggesting that long-term bonds yield more than short-term ones.

Researchers such as Abdymomunov (2013) and Ang(2007) used the term structure of

interest rate as an indicator to predict future economic activities. In developed

1
countries with relatively mature financial systems, bond yields and interest rates are

generally closely correlated. In these countries, the maturity of the Treasury yield

curve has been used to predict the economy and as the main basis for monetary policy.

This paper studies the forecasting ability of Treasury yield curve to economic growth

by using linear model, and analyzes the factors affecting the performance of yield

curve model.

2
 2 Literature Review

For the record, the Treasury yield curve is closely linked to a number of

macroeconomic indicators. Kessel (1965) pointed out in his research that the

economic growth of the United States is related to the Treasury yield curve. In

addition, as early as 1896, academics began to study the expected relationship

between the term structure of interest rates and inflation (Fisher, 1896).

The Treasury yield of different maturity can form a complete Treasury yield

curve. In the existing research, the yield curve of some advanced economies is mostly

related to macro economy. Litterman and Scheinkman(1991) pointed out in their

study that although historical data showed that the yield curve had various shapes,

through factor analysis, most of the information contained in the yield curve was

reflected in two variables, one was the level of short-term interest rate, the other was

the spread between long-term interest rate and short-term interest rate. Stephen and

Kermit (2006) pointed out that if the yield curve is upward, that is, the yield of

long-term bonds is higher than that of short-term bonds, indicating that the economy

is in a normal state. When the opposite happens, the economy shows signs of

recession. This suggests that the Treasury yield curve is forward-looking because it

contains expected information. Estrella and Hardouvelis(1991) used interest rate

differentials to study the predictive power of term structures. Their research suggests

that in America the spread between 10-year and three-month Treasury yields predicts

average quarterly GDP growth over the next 1-4 years. If the difference between

3
10-year and three-month yields is large, it means faster economic growth for some

time to come. If it is smaller, it means growth will be relatively slow for some time to

come. Li, Zhu and Jia(2013) found that the us Treasury yield curve is related to future

economic development in three aspects: one is related to economic policies. The

different forms of Treasury yield curve represent different economic conditions, from

which we can see the operation of the overall economy; second, the Treasury yield

curve also correlates with inflation. The inflation message is also reflected in the

Treasury yield curve. Third, it has to do with fiscal policy. In general, a country's

fiscal policy can influence its macroeconomic growth, which in turn affects future

economic expectations, investment preferences, interest rates, and ultimately the

direction of the Treasury yield curve. The third point shows that there is a link

between the Treasury yield curve and macroeconomic growth.

In empirical research, previous studies have used different methods to analyze

the predictive power of the yield curve for economic expansion and recession. Some

studies emphasize that the predictive power of the yield curve is more reflected in the

economic recession (Azamat, 2013); Some studies pay more attention to the accuracy

of established yield curve, etc. (Andrew, Bekaert & Wei, 2007).Generally, the existing

models to study the term structure of interest rate are divided into static model (Liu,

2006) and dynamic model (Duffee & Stanton, 2012). The most common is the linear

model. The model focuses on whether the yield to maturity can fit the entire yield

curve. Estrella and Mishkin(1996) used OLS regression to measure the

4
macroeconomic predictive power of term structures. The general approach is to make

a regression of future real output growth in the selected period. SPREAD is used as a

variable to predict the future trend of economic development. More specifically, a

linear economic growth model with term spread is constructed. In its model, the

dependent variable is the data that measures the change in economic activity over a

given period, and the independent variable is the selected maturity spread, which is

the difference in interest rates between long-term Treasury bonds and short-term

Treasury bills. In previous studies, many scholars have used linear methods. Most of

them measure changes in economic activity by the growth rate of real GDP. Haubrich

and Dombrosky(1996) applied sample regression. They used the entire yield sample

to predict each data, and their results confirmed the predictive power of the yield

curve. Haubrich and Dombrosky (1996) use simple and efficient methods to select

data. Their method avoids studying the nonlinear specification of the yield curve and

focuses on the difference between short-term and long-term interest rates, thus

transforming the study of the nonlinear yield curve into a simple linear approximation.

They focused on the ten most commonly used interest rates, the federal funds rate, the

three-month, six-month and one-year, two-year, three-year, five-year, seven-year, ten

-year and 30-year rates from the FRED database.

In recent decades, the research on Treasury yield curve and its value are not

limited to academic circles. As the predictive power of the Treasury yield curve has

been widely recognized in developed economies, central banks in some countries

5
have begun to put this research into practice. The Bank of England's inflation report

and the US federal reserve's leading economic sentiment index, for example, both use

the term structure of interest rates to predict inflation or estimate trends in economic

growth. Moreover, the data obtained from the term structure of interest rate is still

heavily weighted. The Treasury yield curve is also a good guide to other parts of the

economy. In the field of public administration, the Treasury yield curve can provide a

reference for the effective management of social economy by the government and

social organizations.

To summarize the existing research, this paper mainly has three theoretical

backgrounds: first, the long-term interest rate of Treasury bonds contains people's

expectations of the economic situation at each time point in the future, which is a

comprehensive reflection of short-term forecast at each time point in the future.

Second, long-term interest rates are generally higher than short-term rates. The

difference in interest rates represents a return on the risk people take in the future.

When the long-term interest rate is lower than the short-term interest rate, it indicates

that the market is not optimistic about the future economic development, and lower

expectations appear. Third, if people predict better economic development in the

future, they are more inclined to buy short-term bonds and sell long-term bonds in

order to maintain liquidity and obtain high returns. On the contrary, if people think the

future economic development situation is not good, in order to ensure a certain level

of yield even in a bad economy, people tend to buy long-term bonds and sell

6
short-term bonds.

7
 3 Statistical Model

This paper first analyzes the yield curve of the US Treasury and then the

relationship between it and GDP. On this basis, the study evaluates the ability of the

Treasury yield curve to predict future economic growth.

A common measure of the predictive power of the term structure of the Treasury

yield curve is the regression of future growth in real output. In this paper, a linear

economic growth model with term spread is constructed and OLS regression is used.

Specifically, in this model, the dependent variable is the data that measures the change

in economic activity over a given period; the independent variable is the selected term

spread. This is the difference in interest rates between Treasury bills and Treasury

bills. Sample regression is used to predict each data using the whole return sample.

Through this result, the relationship between the yield curve and economic growth is

confirmed, so as to judge whether it has predictive power. One criterion for assessing

the predictive power of the yield curve is root mean square error (RMSE), and

comparing RMSE values to consider whether out-of-sample regression performs

better than sample regression.

The first step of the study is to estimate the term spread. This paper uses the

difference between the 10-year Treasury yield and the three-month Treasury yield as

the term spread.

Next, gross domestic product (GDP) is used to estimate economic activity.

According to the research of Abdymomunov (2013) and Ang et al. (2007), the revised

8
annualized real GDP (Real Gross Domestic Product, Quarterly, Seasonally Adjusted

Annual Rate) is a better indicator of actual economic growth than the commonly

reported economic growth data. Quarterly figures for real GDP are calculated as

annualized real GDP growth. Calculated according to the following equation:

Gt,tk  400 / k(lnRGDPt+k  lnRGDPt ) (1)

Where, Gt,tk is the annualized real GDP growth; k is the lag of GDP from time t.

More specifically, k represents the predicted level of information indicating the next k

quarters.

The basic model is to combine the selected period with the real GDP growth rate

and carry out OLS regression. The basic equation of regression is:

yt,tk   0  1SPREADt   t (2)

In this equation, the spread represents the spread for the selected term. yt,tk is

one of the measures for the change in economic activity in the subsequent k period.

When the equation is taken into account for specific term spreads and actual GDP

growth, the formula is as follows:

Gt,t k   0,k  1,k (yt (120)  yt (3))   t  t ~ N (0, 2) (3)

Where yt (120)  yt (3) is the difference in interest rates between the 10-year

Treasury note and the three-month Treasury note; Gt,tk is the annualized real GDP

growth rate.

Further testing is done by adding more variables that enhance the predictive

9
performance. Previous research has shown that short-term interest rates and lagging

GDP can themselves be predictive of real GDP growth. In particular, ANG et al.

(2006) found that the predictive power of short-term interest rates exceeded that of

yields. Haubrich and Dombrosky (1996) found that lagging GDP itself has predictive

power for predicting real GDP growth. Therefore, this study makes further empirical

analysis on the lagging GDP and short-term interest rate. Empirically, the regression

method would be to add two additional real GDP growths. One regression takes the

four-quarter lag in GDP and the ten-year and three-month spread as two explanatory

variables. The other is to take the three-month Treasury bond rate and the 10-year and

three-month spread as two explanatory variables. The equation is as follows:

Gt,t k   0,k  1,k (yt (120)  yt (3))   2,k Gt 4   t  t ~ N (0, 2) (4)

Where, Gt4 is GDP with A lag of four quarters.

Gt,t k   0,k  1,k (yt (120)  yt (3))  3,k yt (3)   t  t ~ N (0, 2) (5)

Where, yt (3) is the three-month Treasury yield. In addition,  2,k and  3,k are

the coefficients of the two added variables. The purpose of this process is to determine

whether the ability of the yield curve to predict the future economy can be enhanced

by adding possible additional other variables.

10
 4 Empirical Results

4.1 Data

The study was based on data from the U.S. market. The study was based on data

from the Federal Reserve Economic Data. (Link: https://fred.stlouisfed.org )

The yield data in the study covered the period 1986(Q1) to 2017(Q4). This long

sampling period allows for reasonable in-sample and out-of-sample studies. Moreover,

Estrella and Hardouvelis(1991) and (Azamat, 2013) have shown that the predictive

power of the yield curve is weaker after 1985 than before. So it makes sense to focus

on the period from 1986 to 2017.In terms of the types of yields selected, based on

previous studies, this paper selects five yields, namely the 3-month, 6-month, 2-year,

5-year and 10-year yields. Daily data on yields are converted into quarterly data,

giving them the same temporal basis as measures of economic activity. In addition,

the quarterly averages will smooth out some of the abnormal yield data that can

circulate every month. Table 1 shows the Treasury yield data of different maturities.
Table 1: Descriptive statistics of returns between Q1 1986 and Q4 2017
Maturity Observations Average Standard Deviation Minimum Maximum
3 months 128 4.48 3.12 0.01 13.95
6 months 128 4.70 3.23 0.04 14.16
2 years 128 5.11 3.31 0.19 14.25
5 years 128 5.81 3.17 0.61 14.69
10 years 128 6.32 2.91 1.56 14.50

Table 1 show that there are 128 observations per term. In general, short-term

interest rates are low. The standard deviation of each yield is about 3, suggesting that

yield volatility may not be affected by different maturities.

11
 In this study, the growth rate of real gross domestic product is used as an

indicator of economic activity, rather than the original GDP data. The calculation

method of real GDP growth rate has been introduced in the above research methods.

As with the yield data, the quarterly figure for real US GDP is from Q1 1986 to Q4

2017. The long period covered several experiences of economic expansion and

recession, different types of monetary policy, inflation or deflation. According to the

existing research, the yield curve is indeed affected by the economic cycle changes. In

empirical analysis, this paper analyzes the real GDP growth rate at the end of the next

quarter, the next four quarters, the next eight quarters, the next twelve quarters and the

next sixteen quarters. Table 2 shows real GDP growth in different forecast ranges.

In the first model, the coefficient for violent crime is positive and insignificant,

which is unreasonable in our model. As I analyzed before, the reason for this is that

our model suffers multicollinearity. The correlation coefficient was ~0.99, which can

be visualized by the following image.

Table 2：Descriptive statistics of growth rate of GDP between Q1 1986 and Q4 2017
k quarters Observations Average Standard Deviation Minimum Maximum
1 128 2.80 2.62 -8.70 9.03
4 125 2.84 1.94 -4.18 8.20
8 121 2.84 1.61 -1.84 6.51
12 117 2.79 1.39 -0.54 5.74
16 113 2.77 1.23 0.09 5.03

4.2 Empirical Results

(1) The predictive power of pure term spreads

12
 First, this paper tests the predictive power of term spreads without adding any

other variables. The yield of 10 years and 3 months was taken as the independent

variable, and the growth rate of real GDP was taken as the dependent variable to carry

out OLS regression. The regression results are shown in table 3.

Table 3：The results of OLS regression analysis of real GDP
k quarters Adjusted R2
 0,k 1,k

1 2.350(0.000) 0.243(0.025) 0.012

4 2.230(0.000) 0.342(0.020) 0.043
8 2.037(0.000) 0.441(0.000) 0.108
12 2.094(0.000) 0.390(0.000) 0.113
16 2.174(0.000) 0.337(0.000) 0.099

Where,  0,k is the constant term in the regression and 1,k is the coefficient of

the independent variable, that is, the 10-year and three-month yields. The parameter k

means that the real GDP growth rate is the k quarter ahead. The adjusted R2 is used to

measure how good the model is. As an indicator to test the statistical results, the

corresponding P value of each coefficient is in parentheses. The results in table 3

show that all the coefficients of each OLS regression are statistically significant. This

suggests that the term spread can be an explanatory variable for real GDP growth and

has predictive power for economic activity. Take the level of GDP growth forecast for

the previous quarter: when yields rose by 1%, real GDP grew by 2.6% (2.35+0.234*1).

However ， the adjusted R2 data were not as good as expected. The highest value

generated in the forecast range over the next 12 quarters is 0.113.The results show

that the model based on rate of return does not fit well. This suggests that pure yield

spreads are weak predictors of economic growth.

13
 (2) Out of sample analysis

In-sample testing measures the predictive power of term spreads over the entire

sampling period. This means that the prediction contains real future information that

should not be known at the time of the test. As a result, the regression results may not

be as accurate. Therefore, on this basis, the samples before the test point are selected

for sample regression test. In this paper, the 10-year and three-month Treasury yields

of the term spread are used to predict real GDP growth. Azamat (2013) showed that

samples from the first 20 years (1986:Q1 to 2005:Q4) were used to predict GDP

growth in the out-of-sample test. The purpose of this step is to understand whether the

predicted value of the out-of-sample regression is consistent with the actual value, and

to compare the performance of the in-sample prediction with the out-of-sample

prediction of GDP growth. According to the test results of Azamat (2013), the mean

variance (RMSE) of the out of sample forecast of GDP growth rate is 2.348 in the 12

quarter forecast period based on the spread of 10 years and 3 months. For the same

prediction range and spread, the mean square error (RMSE) of the in-sample

prediction is 1.317.The results show that RMSE of in-sample regression is lower than

that of out-of-sample regression, so the prediction accuracy of sample regression is

better. This may be due to the fact that this is the real GDP value of the non-sample

regression forecast based on the sample data, which can only be used before the time

of each forecast.

(3)Analysis of Additional Variables

14
 As discussed in the research methods section, Haubrich and Dombrosky (1996)

mentioned in their study that the predictive power of the yield curve can be improved

by adding additional explanatory variables, namely short-term interest rate or lagging

real GDP growth rate. This paper considers the case of adding short-term interest rates

to the model. The Table 4 below shows an increase in the 3-month Treasury yield. K

represents the projected future K quarters of real GDP growth. The corresponding P

value for each coefficient is shown in parentheses.

Table 4: OLS regression analysis results of real GDP after the increase of 3-month Treasury
yield

k quarters 1,k  2,k Adjusted R2

1 0.471(1.108) 0.207(0.008) 0.054
4 0.583(0.000) 0.251(0.000) 0.168
8 0.696(0.000) 0.265(0.000) 0.309
12 0.641(0.007) 0.247(0.000) 0.337
16 0.144(0.000) 0.746(0.000) 0.383

According to table 4, all the coefficients except one are statistically significant.

By comparing the adjusted R2, it can be concluded that the model with increased

short-term yield can fit better than the model based solely on yield. This suggests that

the combination of yield spreads and short-term interest rates can improve the ability

to predict economic growth.

15
 5 Conclusion

With the help of simple linear model method, this paper focuses on the

forecasting ability of yield curve to economic growth and recession. Specifically, the

term spread of Treasury bonds is used as an approximation of the yield curve. OLS

regression model was used to determine the term difference of real GDP growth rate.

Using a simple linear method, we find that the pure term spread can predict the

growth of real GDP. However, the empirical results of this simple model also show

that the forecasting ability of pure term spread is relatively poor. When the additional

variable of short-term interest rate was added to the model, the results showed that the

model with an increase in short-term interest rate had better prediction ability than the

model with a pure term spread.

In general, this paper proves that the Treasury yield curve itself can predict future

economic trends. In a mature market economy, this is a good economic forecasting

tool. Therefore, the country should strengthen the management of the benchmark

interest rate, clarify the importance of the yield curve, and improve the credibility of

the monetary policy system.

16
 Reference

Abdymomunov, Azamat (2013). Predicting output using the entire yield

curve.Journal of Macroeconomics, 37: 333-344.

Ang, Andrew, Geert Bekaert and Min Wei (2007). The term structure of real rates

and expected inflation. National Bureau of Economic Research Working

Paper12930

Cecchetti, S. G., Schoenholtz, K. L., & Fackler, J. (2006). Money, banking, and

financial markets (Vol. 4). McGraw-Hill/Irwin.

Dotsey, M. (1998). The predictive content of the interest rate term spread for future

economic growth. FRB Richmond Economic Quarterly, 84(3), 31-51.

Duffee, G. R., & Stanton, R. H. (2012). Estimation of dynamic term structure

models. The Quarterly Journal of Finance, 2(02), 1250008.

Estrella, Arturo and Frederic S. Mishkin (1996). The predictive Power of the Term

Structure of Interest Rates in Europe and the United States: Implications for the

European Central Bank. European Economic Review 41: 1375-1401

Estrella, A., and Hardouvelis, G.A. (1991). The term structure a predictor of real

economic activity. Journal offinance 46 (2): 555-576.

Fountas, Stilianos and Menelaos Karanasos (2006). The relationship between

economic growth and real uncertainty in the G3. Economic Modelling

23:638-647

Fisher, I. (1896). Appreciation and Interest: A Study of the Influence of Monetary

17
 Appreciation and Depreciation on the Rate of Interest with Applications to the

Bimetallic Controversy and the Theory of Interest (Vol. 11, No. 4). American

economic association.

Hai-dong, L. I. U. (2006). The Static Analysis and Dynamic Variation Features of the

Term Structure in China [J]. Journal of Shanxi Finance and Economics

University, 5.

Hamilton, J. D., and Kim, D. H. (2001). A re-examination of the predictability of

economic activity using the yield spread. Journal of Money, Credit and

Banking34: 340-344.

Haubrich, J. G., and Ann M. Dombrosky (1996). Predicting real growth using the

yield curve. Federal Reserve Bank of Cleveland Economic Review 32(1): 26-34.

Haubrich, J. G., & Dombrosky, A. M. (1996). Predicting real growth using the yield

curve. Economic Review, 32(1), 26-35.

Kessel, R. A. (1965). The cyclical behaviour of the term structure of interest rates,

Occasional Paper No. 91 National Bureau of Economic Research.

Li Ju. (2013). Analysis of Treasury yield curve and macroeconomic relationship.

Bonds. Global perspective. April 2013.

Litterman, R., & Scheinkman, J. (1991). Common factors affecting bond returns.

Journal of fixed income, 1(1), 54-61.

The Fed in the Communique, “Federal Reserve issues FOM C statement Press

Release,September21,2011”,URL:NTTP://www.federalreserve. Gov/ news events /

18
Press / monetary / 20110921 a. HTM

19