Digital Commons @ George Fox University

Student Scholarship - School of Business School of Business

1-1-2018

Effects of Bonds & Fed Funds on the Stock Market

Josh Nelson
jnelson15@georgefox.edu

Mason Koch
mkoch16@georgefox.edu

Recommended Citation
Nelson, Josh and Koch, Mason, "Effects of Bonds & Fed Funds on the Stock Market" (2018). Student Scholarship - School of Business.
13.
https://digitalcommons.georgefox.edu/gfsb_student/13

This Paper is brought to you for free and open access by the School of Business at Digital Commons @ George Fox University. It has been accepted for
inclusion in Student Scholarship - School of Business by an authorized administrator of Digital Commons @ George Fox University. For more
information, please contact arolfe@georgefox.edu.
 Effects of Bonds & Fed Funds on the Stock Market
Josh Nelson & Mason Koch
George Fox University

Abstract: This paper is to investigate the relationship that Treasury bonds and Fed Funds have
with the stock market. This research is valuable because the stock market is a major aspect of
America’s economy, although many consider it unpredictable and difficult to understand. The
goal of our research is to lessen this unpredictability, as well as to assess the inverse
relationship that hypothetically exists between the stock market, bonds and Fed Funds. By
employing an OLS regression model, this study finds the effect of Fed Funds is statistically and
economically significant with regard to the stock market, also having the predicted inverse
relationship. However, results for bonds are inconclusive; they are neither statistically or
economically significant, nor do they consistently have an inverse relationship with stocks.
Nevertheless, this paper can begin to show the hypothetical relationship that exists between stock
and bond markets, and provide a basis for future research.

Keywords: Stocks, Bonds, Fed Funds, Effects, Econometrics

JEL Codes: C1, G0, and G12

 2

1 Introduction: Importance of Investing

Investing has become an increasingly important aspect of our economy in the United States,

especially in recent years, as it has become a necessity for many in order to feel their finances are

protected for their futures. It is also an effective way to earn money without putting one’s

savings account at risk. This desire to protect oneself from risk is a major driver of investing,

whether it be in the stock market or bonds, and this element of risk puts a high value

economically on being able to generate profits on money already earned, instead of letting

inflation erode those earnings while they sit in a savings account.

Being able to ‘outthink the market’ has long been considered as impossible by many

inexperienced investors, as a common belief of the market is that stocks take a ‘random walk’

along some trend that is often unpredictable at best. This is commonly known as the efficient

market hypothesis, the idea that you cannot beat the market because it only reacts to new

information. This study is interested in whether we can minimize this unpredictability in

movements of the stock market, as the more an investor can understand the market, the better

their decisions and judgement will be to invest effectively. On account of these interests, we will

be researching changes in the rates of stocks by way of regression analysis. We will conduct this

research by analyzing the relationship bond rates (specifically treasury bonds) and the Effective

Fed Funds Rate have with movements of the stock market.

Our hypothesis is that stock movements should be inversely related to bond and Fed

funds interest rates. Our theory behind this is the idea that as bond prices decrease, stock prices

tend to rise. This is justified by the fact that bonds are safer investments but generally provide or

generate lower returns, while stocks are much riskier investments but consistently have superior

revenues, and investors have to choose between them. In theory, if we can spot one of the two
 3

types of investments deviating ahead of time, we will be able to accurately predict a change in

the deviation of the other. If our hypothesis is correct, it will begin to affirm the theory of inverse

relationships between bonds and stocks, allowing professionals to add into their trading

strategies, and enable individuals to plan their investments around an overall market outlook.

2 Data Overview and Weaknesses

To conduct this research, we will be using data pulled from the Federal Reserve Bank of St.

Louis. We have identified a ten-year time frame for all of our variables, and have given a

monthly time frame for this data. The time frame for our data begins in December of 2008 (12

months after a significant drop in the stock market) and ends October 2018. In total, we have 120

data points for each of these variables, giving us a combined data entry of 840 different data

points. To keep a consistent basis, and to ensure our data is stationary, we have adapted all of our

data to be represented on a monthly percent change basis.

Our dependent variables will be the S&P 500 (SPX) and Dow Jones Industrial Average

(DJIA). We have chosen these as the two have long been considered to be accurate

representations of the stock market as a whole. We also are considering them to be

interchangeable in our regressions as they are extremely correlated (96.7%).

Our independent variables are the Effective Federal Funds Rate (FF), 3-Month Treasury

Constant Maturity Rate (MO), 1-Year Treasury Constant Maturity Rate (YR), 5-Year Treasury

Constant Maturity Rate (YR1) and 10-Year Treasury Constant Maturity Rate (YR2). A side note

is that in contrast to our dependent variables, our short term bond rates vary significantly from

our long-term rates. Short term rates (3-month and 1-year bonds) only correlate with long-term

rates (5 and 10-year bonds) approximately 30%. Our summary stats for all our variables are

outlined in the table below.

 4

Table 1.1 Summary Statistics

Variables S&P *Dow Effective 3-Month 1-Year 5-Year 10-Year

(Percent Δ 500 Jones Fed Funds Maturity Maturity Maturity Maturity
monthly (SPX) Industrial Rate (FF) Rate Rate Rate Rate
basis) Average (MO) (YR) (YR1) (YR2)
(DJIA)

Median 1.38 1.24 0.70 2.80 1.40 0.40 0.29

Mean 1.02 0.96 2.22 14.52 1.73 0.82 0.17

Minimum -10.56 -9.48 -66.54 -85.33 -52.00 -34.14 -30.61

Maximum 12.02 10.46 94.78 792.86 75.85 43.84 19.92

Standard 3.05 2.99 18.07 88.36 15.98 12.29 8.00

Deviation

Notes* Dow Jones & S&P 500 are 96.7% correlated.

Our dependent variables SPX and DJIA have means of 1.02% and 0.96% respectively,

these represent the average percent change for the stock prices on a monthly basis. The medians

of 1.38% and 1.24% tell us that the majority of our data for stocks is above the mean, suggesting

the general stock market is consistently increasing, and standard deviations of 3.05% and 2.99%

show us we have a fairly clustered sample. Our first independent variable is the Effective Fed

Funds rate, and it is a slightly different variable than the actual Fed Funds Rate. The Effective

Fed Funds rate is a weighted average of the rates that banks charge each other (banks can

negotiate with each other for a higher rate than the current Fed rate). In short, it is a more

accurate representation of how the market is affected by changes in the Fed Funds rate. This is

why we have a maximum value of 94.78% and a minimum value of -66.54%, they may seem

excessively large, but it is a more accurate representation in the context of the stock market. Our

mean of 2.22% is close to what the actual Fed Funds rate is currently, but a median of 0.70%
 5

shows most of our data is below our mean, and a standard deviation of 18.07% indicates our

sample data is particularly spread out.

Our bond data for MO, YR, YR1, and YR2 have means of 14.52%, 1.73%, 0.82%, and

0.17%, respectively. This suggests that as the maturity of bonds becomes longer, the average

yield is decreased. Similarly, their standard deviations, equaling 88.36%, 15.98%, 12.29%, and

8.00%, shows that changes in bond rates become more and more consistent with increasing

lengths of maturity. This is consistent with the intuition that long term bonds are typically safer

but have a lower percent yield. MO has a median of 2.80%, implying there are a number of

outliers pulling the mean down, whereas the medians for YR, YR1, and YR2 have values quite

close to their means, suggesting the data is evenly distributed.

When we look at the data we have aggregated, there is an apparent weakness that

presents itself. Although we have a significant number of data points for our test, we are only

looking at a period that contains one recession. The fact that we have been in the longest bull run

in history is a significant contributor to this. Dow Jones’ stock price has grown at a rate

approximately equal to 11% annually since December of 2008, when the expected growth

annually for the stock market is 7% (The same is also true of the S&P 500--their stock has grown

10% annually since 2008). This means our results in theory may be less accurate if applied in a

period of recession, due to the unusually high growth rate as compared to the average return of

the market. In order to find a more accurate depiction we may need to look at a series with

multiple recessions for our findings to be more applicable. Our research question asks if there is

a causal relationship between the change in bond value and the change in stock price and this is

seen throughout a recession more than it can be seen in a period of vast expansion.
 6

As our study is based on time series data we are obligated to run ADF tests (Augmented

Dickey-Fuller Tests) on each of our variables in order to test if they are non-stationary or

stationary. Due to the fact that we have appended our data to be in percent change, we have

likely already adjusted the structure of our data if it was originally non-stationary. To confirm

this intuition, and so that we can trust our results, we employed the level of variable version of

ADF. We used AIC to determine the proper amount of lags to include. These results show that

all of our variables are stationary because the p-values are smaller than critical values at 1%

confidence levels. Therefore, we are able to reject the null hypothesis that our data is non-

stationary. This verifies that we are able to use OLS regression models in our study. Our results

for ADF tests are summarized in the table below.

Table 1.2 ADF Tests

Variable S&P Dow Effective 3-Month 1-Year 5-Year 10-Year

500 Jones Fed Treasury Treasury Treasury Treasury
(SPX) (DJIA) Funds (MO) (YR) (YR1) (YR2)
(FF)

T-Stat -8.405 -8.465 -8.945 -10.278 -5.753 -9.792 -9.106

P-Value 8.1e-16 5.085e-14 9.71e-16 8.853e-14 4.476e-06 5.758e-13 9.534e-12

Notes* Calculated using constant and trend using AIC to test for lags

3 Research Methodology

Our underlying theoretical relationship is that the movement of stocks should be inversely

related to the movement of bonds and Effective Fed Fund rates. We expect the coefficients on

Effective Fed Funds and all our bond data to be negative, as we are predicting that a negative

change on bond rates should cause our dependent variables SPX and DJIA to increase. This is

justifiable because if bond rates decrease, investors are more likely to buy stocks instead of
 7

bonds, leading to an increase in the price and value of whatever stocks they choose to buy.

Similarly, when the Effective Fed Funds rate increases there is a decrease in the money supply in

the market, and this discourages investments in the stock market.

In order to test the theoretical relationship, we will be running regressions on our data by

using OLS (Ordinary Least Squares) analysis, we are also operating under the assumption that

the relationship will be linear. Time series data is typically difficult to evaluate in the context of

OLS, but as we have set up our data to be in a percent change format and have identified all of

our variables to be stationary (by way of ADF tests on each of our variables), OLS is a valid

method to test our theory.

We have defined our OLS multiple regression model to be: Expected Stock Market

Yields (SPX) or (DJIA) = β + β X (FF) + β X (MO) + β X (YR) + β X (YR1) + β

1 1 1 2 2 3 3 4 4 5

X (YR2)
5

SPX and DJIA are the monthly percent change of their respective stocks, FF is the Effective Fed

Funds Rate monthly percent change, MO is the monthly percent change of 3-month treasury

bond maturity rates, YR is the monthly percent change of 1-year treasury bond maturity rates,

YR1 is the monthly percent change of 5-year treasury bond maturity rates, and YR2 is the

monthly percent change of 10-year treasury bond maturity rates. Alongside this model we also

will be running smaller regressions with fewer variables to test our data more thoroughly (i.e.

only FF and DJIA, only Bonds and SPX, etc.) to establish a more accurate understanding of

whether our findings are valid, and to provide context for readers. Based on these models we

have run OLS regressions as shown below.

 8

Table 1.3 Results

Dependent Variable Columns (1) (2) (3) = SPX: Columns (4) (5) (6) = DJIA

Independent (1) (2) (3) (4) (5) (6)

Variables

Effective Fed −0.0406** −0.0450** −0.0414** −0.0447**

Funds Rate (0.019) (0.019) (0.018) (0.019)
(FF)

3-Month (MO) −0.0004 −0.0021 −0.0012 −0.0029

Treasury (0.003) (0.003) (0.004) (0.003)
Maturity Rate

1-Year (YR) −0.0177 0.0114 −0.0194 0.0096

Treasury (0.030) (0.034) (0.030) (0.033)
Maturity Rate

5-Year (YR1) 0.0374 0.0045 0.0283 −0.0044

Treasury (0.068) (0.068) (0.067) (0.066)
Maturity Rate

10-Year 0.0854 0.1100 0.0895 0.1140

(YR2) ( 0.102) (0.097) (0.098) (0.091)
Treasury
Maturity Rate

Intercept 1.1302*** 1.0011*** 1.1213*** 1.0789*** 0.972*** 1.0921***

(0.270) (0.251) (0.267) (0.259) (0.249)

Adjusted R 2
0.051 0.109 0.159 0.055 0.098 0.149

F-Test 3.05 2.88 3.24 3.14

(P-value) (0.019) (0.017) (0.014) (0.010)

Notes* F-Tests in columns (2) and (5) use F (4, 114): Testing significance of our Bond
variable Columns (3) and (6) use F (5, 113): Testing significance of all variables
 9

4 Discussion of Results

For the discussion below we will be referencing columns (3) and (6). Regressions in columns (1)

(2) (4) and (5) are to see how FF and bonds act in the absence of the others and are included for

context and for F-tests of bond variables.

In our full regression models, columns (3) and (6), there are only two variables that are

statistically significant--the constant and the variable FF. The constant is significant at 1% and

FF at 5%. The constant being significant is not particularly interesting, as it only means it differs

significantly from 0, but as we are regressing on percent change of stock prices, a constant higher

than 0 is expected, as the stock market is consistently on an upward trend. A key aspect of our

results is the coefficient on FF, it is statistically significant, as well as negative. This is exactly

what we predicted in our hypothesis.

However, the results from our bond variables are less conclusive, as some coefficients are

negative (as predicted) but others are positive, and none are statistically significant. In column

(3) the coefficient on MO is negative, although it is very small and somewhat insignificant.

However, coefficients on YR, YR1, and YR2 are positive, opposite to what we predicted.

Column (6) has similar results, except for the addition of variable YR1 becoming negative. As

none of these are statistically significant, we cannot say these results prove anything about bonds

being inversely related to stocks, only that our initial intuition is at least partially accurate.

Economic significance is a similar story with regard to our bond coefficients. In column

(3) a 1% increase in bond rates in any of our coefficients will only cause a decrease of 0.002%

for MO, and increases of 0.01% for YR, 0.004% for YR2, and 0.1% for YR2. For any of these to

actually make a difference in stock prices there would have to be a major decrease or increase in

bond rates--a change of at least a 10% would be required. Referencing table 1.1, the average
 10

change in bond rates typically averages less than 5%, and therefore a change of that magnitude is

unlikely. Results from column (6) regressing on DJIA tell the same story.

In contrast to the bonds, economic significance shows extensively in the Effective Fed

Funds Rate. If there is an increase of 1% in the Effective Fed Funds rate, we can expect stock

prices to decrease by an amount of 0.045%. This may not seem like much unless you look at the

percent changes of the Fed Funds rate. The Fed recently increased the Fed Funds rate by 25 basis

points (from 2% to 2.25%). This is a 12.5% percent change, meaning the expected change in the

stock market when calculated in our model is -0.56%. This is a much larger effect, and as the

Effective Fed Funds rate is consistently increasing or decreasing (depending on whether the Fed

is trying to stimulate or slow down the economy), this could easily make a large impact on the

stock market over a long period of time. It is very interesting however that even with all this

being true, the F-stat on our bond variables is significant at 5%, even though none of the

individual variables are statistically significant on their own. One last interesting note on our

results is that short term rates seem to have the negative effect on the stock market (Fed Funds

rate and 3-month bonds) but when we look at longer maturity rates (1-year, 5-year, 10-year) the

opposite seems to be true. This suggests that the stock market is seemingly impacted more by

‘overnight rates’ as opposed to long-term rates.

However, there are some limitations to these results. The fact that none of the coefficients

on our bond data are statistically significant or particularly economically significant makes it

difficult to draw a definitive conclusion about the relationship that bonds have with the stock

market. Even if you were to disregard the economic and statistical significance, the fact that the

sign of our coefficients are not consistently negative or positive suggests that different maturity

lengths for bonds each have their own specific relationship with the stock market. In hindsight
 11

this is validated by our discussion of short-term rates only being roughly 30% correlated with

long-term rates.

A second limitation of our results stems from our discussion on the weaknesses in our

data. Because we are only analyzing data from the last 10 years, (during the longest bull run in

the history of the stock market and only includes a few data points during the recession of the

2000s) it is difficult to conclude with certainty how applicable our results will be if applied to a

period during a recession versus a stage of expansion. The only way to decide whether or not this

is the case would be to pull data over a longer period of time, so as to include multiple

recessions, and compare the results.

5 Conclusion

This paper looks to improve on investment techniques by analyzing how changes in the Effective

Fed Funds rate and Treasury Bond rates directly influence change in the stock market. Using an

OLS (Ordinary Least Squares) regression model with time series data we looked to measure the

underlying hypothesis that treasury bond rates and Fed Fund rates have an inverse relationship

with the stock market.

Our results showed statistical significance in the Effective Fed Funds rate at 5%, as well

as economic significance. A change from 2% to 2.25% in the Fed Funds rate (a 12.5% increase)

can negatively influence the stock market nearly half a percent overnight. In the real world this

means if the Fed is consistently increasing or decreasing rates over a long period of time, the

stock market as a whole could lose or gain 5% to 10% in value, a significant impact. In contrast,

coefficients on our bonds showed little significance economically or statistically, and

interestingly enough our hypothesis that bonds and Fed Funds have an inverse relationship with

the stock market is only partially correct. Only the 3-month bonds and Fed Funds have a
 12

negative coefficient, suggesting that the stock market is only impacted the way our hypothesis

predicted with regard to short-term rates.

Although this study only finds the Effective Fed Funds rate to be economically and

statistically significant, it still has viable findings that would be very interesting to research in the

future. As we have already discussed, a future study could look at the same variables as our

study, but analyze a different range of data, which covers a longer period of time with more

downturns in the economy, so as to get a more complete picture of the stock market’s

relationship with bonds and Fed Funds. A second interesting issue with our results are the

inconsistent sign of treasury bond rates. It would be fascinating to research why different

maturity lengths seem to react with the stock market so differently. In our study it is very unclear

why they differ so drastically. It is our hope that this study can begin to affirm the inverse

relationship that has only hypothetically existed previously between stock and bond markets, and

that it will aid investors in their investment strategies.

 13

References

“Federal Reserve Economic Data | FRED | St. Louis Fed.” FRED, Federal Reserve Bank of St.

Louis, fred.stlouisfed.org/.