Empirical Research Paper The Ability of Previous Quarterly Earnings, Net Interest Margin, and Average Assets to Predict Future Earnings of Regional Pure Play Banks Ryan Holcomb 1. Introduction The goal of this research paper is to establish a regression model that is capable of forecasting quarterly earnings estimates of comparable regional pure-play banks. The regression model is constructed around several financial and economic theories that help to explain what impacts banks future earnings, including research from several other econometric studies that will be discussed in sections 2 and 3. Specifically, my hypothesis is that a bank’s previous quarter earnings, net interest margin, and average assets are jointly significant in predicting future earnings. Earnings = B1 + B2Earnings T −1 + B3 ln Net Interest Margin + B4 ln Average Assets +Ui H0 : B2= B3= B4 =0 H1 : H0 is Not True
28
Embed
The ability of previous quarterly earnings, net interest margin, and average assets to predict future earnings of regional pure play banks
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Empirical Research Paper
The Ability of Previous Quarterly Earnings, Net Interest Margin, andAverage Assets to Predict Future Earnings of Regional Pure Play Banks
Ryan Holcomb
1. Introduction
The goal of this research paper is to establish a regression model that is capable of
forecasting quarterly earnings estimates of comparable regional pure-play banks. The
regression model is constructed around several financial and economic theories that help to
explain what impacts banks future earnings, including research from several other econometric
studies that will be discussed in sections 2 and 3. Specifically, my hypothesis is that a bank’s
previous quarter earnings, net interest margin, and average assets are jointly significant in
predicting future earnings.
Earnings=B1+B2EarningsT −1+ B3 ln Net Interest Margin+B4 ln Average Assets+Ui
H0 : B2=B3=B4=0
H1 : H0 is Not True
The model is beneficial in that it establishes a connection between the X variables and Y
outcomes, but also allows the researcher to incorporate information given by the individual firm
and the economy through the ability to vary net interest margin and average assets.
The motivation for this study is based on the field of financial analysis, where analysts
are charged with following a group of companies in a specific industry or sector. Through time
and research many analysts develop an intricate knowledge of the industry and understand that
the future success of the firm can be broken down to several key variables. The aforementioned
model is a manifestation of this reality through my own research of the regional banking sector
and represents a test of the ability of regression models to predict cash flows for not only banks
but other industries as well. The interesting part of the research project lies in the ability of
regression analysis to confirm or deny the importance of certain variables on the future
earnings of a company. Many analysts have the benefit of following industries for many years
and learning the most important variables, however, for the average investor this is not the
case. Through regression analysis, less experienced investors can test the importance of
different variables and their impact on future earnings which can aid them in investing
decisions. In my own case, I have recently participated in a valuation of the regional banking
sector in which I did not utilize a regression, but rather a multiples approach, and my interest in
this project is to determine the value of regression analysis by comparing the regression
estimates to the multiples approach estimates.
The economic theory behind the regression is based on the point that profitability for
pure-play banks rests on the net-interest margins of the firm and the growth of average assets;
net interest margin is calculated by subtracting the firm’s net interest income by its net interest
expenses and dividing by its interest-earning assets—or average assets because the banks are
pure-play means that the bank’s assets are traditional loans or other interest earning assets.
(Net Interest Income−Net Interest Expense )
(Interest Earning Assets)
Due to the fact that these banks are pure-play banks, their ability to earn more income hinges
on their ability to grow assets and the net interest margin they can attain; the higher the net
interest margin the greater the profits for the bank—this means that the amount of interest the
bank receives on its loans is growing at a higher rate than what it’s paying on its deposits.
Furthermore, there have been several econometric studies that have shown a relationship
between past earnings and future earnings. Thus, theory seems to support the hypothesis that
previous quarter earnings, net interest margin, and average assets are jointly significant in
predicting future earnings.
2. Literature Review
Catherine A. Finger in her research report, “The Ability of Earnings to Predict Future
Earnings and Cash Flow”, outlines her hypothesis that there is a connection between previous
year’s earnings and previous year’s cash flow on future earnings. She maintains that current
cash flow by itself is a better predictor of earnings for shorter time horizons and that current
cash flow combined with current earnings is a better predictor of longer horizons. Since a pure-
play bank’s operating cash-flow is essentially its earnings, my model incorporates both these
hypotheses and subsequently my model should be able to predict both short and long horizons.
My regression differs from Finger’s in that my model is industry specific and incorporates other
variables besides earnings to help explain future earnings; whereas, her model is purely focused
on cash flow and earnings (Finger 210-223). Finger’s report is in response to several other
research reports including Albrecht, Lookabill, and McKeown’s 1977 report on, “The Time-Series
Properties of Annual Earnings”. In their report the authors argue that future earnings are
uncorrelated with previous year earnings and subsequently, earnings exhibit a random walk
model (Albrecht, Lookabill, and McKeown 226-244). Once again, these articles investigate the
ability of past earnings to predict future earnings which represents a component in my
regression, however my regression is unique in that I examine other variables in my model.
The importance of net interest margin on the profitability of banks is not a new
phenomenon, but rather, something that is well known in the finance discipline. This point is
evidenced by Gerald Hanweck, professor of finance at George Mason University, and Lisa Ryu,
Senior Financial Economist for the FDIC, in their report “The Sensitivity of Bank Net Interest
Margins and Profitability to Credit, Interest-Rate, and Term-Structure Shocks Across Bank
Product Specializations”, where they note:
“Despite the rising importance of fee-based income as a proportion of total
income for many banks, net interest margins (NIM) remain one of the
principal elements of bank net cash flows and after-tax earnings. As shown
in figure 1, except for very large institutions and credit card specialists,
noninterest income s till remains a relatively small and usually more stable
component of bank earnings (Hanweck, and Ryu 3).”
The important distinction between my regression and industry knowledge about key variables is
that I utilize industry knowledge to create a model that takes into account many theories in the
hope of finding a model that is jointly significant in predicting future earnings. There is a wide
variety of analysts who place importance on different variables in determining the value of a
bank, and I am trying to create a regression that finds a good combination of these theories
which will give an accurate picture of future earnings.
3. Econometric Model
As aforementioned, the econometric model of my study is that future quarterly earnings
are a function of last quarter’s earnings, net interest margin, and average assets.
Earnings=B1+B2EarningsT −1+ B3 ln Net Interest Margin+B4 ln Average Assets+Ui
The specific hypothesis that I am testing is that these variables are jointly significant in
predicting future quarterly earnings.
H0 : B2=B3=B4=0
H1 : H0 is Not True
I have included last year’s earnings in the model as a way to account for the condition of the
bank that I am analyzing; specifically, the variable provides a way to establish how well the bank
has performed the last quarter and this provides a baseline to establish the future quarter
earnings. Similarly, I included the lagged earning variable because of the past research that was
aforementioned in section 2; this research points out that past earnings are indeed correlated
with future earnings, and as such provides a variable that is capable of forecasting future
earnings. Thus, the variable was included in the model because it provides a way to establish
the general condition of the bank and to include a variable that is correlated with future
earnings.
The net interest margin variable was included in the model because of its significance in
determining the profitability of pure-play banks. Pure-banks are aptly named because most of
their profits come from traditional banking practices such as taking in deposits and offering
loans; this compared to large national banks that take on many other services such as insurance,
brokering, and mergers and acquisitions. Net interest margin, as aforementioned, is calculated
by taking the difference between net interest income and net interest expenses and dividing it
by average interest-earning assets. The net interest income is money the bank earns on loans it
handles, and the net interest expense is the amount of money the bank pays out on deposits.
The important factor in net interest margin is the fact that there is a time period difference
between the loans they issue and the deposits they carry—loans are typically longer and can
reach 30 years in length, whereas deposits such as CDs are variable in length but usually are
between 6 months to a year—and all the while the interest rates are varying so that deposits
are being updated to new interest rates while the bank’s loans are fixed at a certain interest
rate. Thus, the important variable in determining profitability is measuring the difference
between what the bank is paying out and what it is taking in, and this measure is the net
interest margin.
The asset variable was included in the model because asset growth is usually associated
with increased earnings. To determine the profits of a pure-play bank we take the average
assets and multiply this by the net interest margin; thus, the regression model incorporates
both the net interest margin and the estimated average assets the bank will have the following
quarter. The main assets of a pure-play bank are its loans and thus, as the bank increases its
loans it usually increases its earnings. An important characteristic of this variable in the model
is that it allows the analyst to vary asset growth based on industry trends and management
guidance which corresponds with a certain growth in earnings.
4. Data
The data I gathered for the regression was quarterly data based on the last 11 quarters
dating back to the second quarter of 2009. I wanted to use data sets that reflected the recovery
since the financial crisis began in 2007 and, based on some opinions, ended in the second
quarter of 2009. Based on my previous valuation of the sector, I included in my regression data
nine publicly listed regional banks which include First Midwest Bank (FMBI), Umpqua Bank
(UMPQ), MB Financial Bank (MBFI), The Privatebank and Trust Company (PVTB), National Penn
Bank (NPBC), Citizens Bank (CRBC), Banner Bank (BANR), Columbia State Bank (COLB), Sterling
Savings Bank (STSA). My original valuation was for Sterling Savings Bank which required a list of
comparable companies that were determined based on asset size, portfolio similarity, regional
growth, and capital ratings.
The earnings and asset data are from the SEC website where publicly listed companies
are required by law to submit quarterly and annual financial data (" EDGAR "). The net interest
margin data was found on the FDIC website where banks are required to provide additional
information regarding financial performance in uniform bank performance reports (“UBPR”). In
an effort to find the best regression model to predict future earnings, I also ran test regressions
using data on tier one capital levels which can be found in the uniform bank performance
reports and on GDP quarterly growth rates which can be found on the Bureau of Economic
Analysis website ("Bureau of Economic Analysis").
Summary Statistics, using the observations 1:01 - 9:11Variable Mean Median Minimum MaximumEarnings -14620.6 6391.00 -455174. 35978.0
Mean dependent var -14620.58 S.D. dependent var 68503.10Sum squared resid 3.33e+11 S.E. of regression 58853.42R-squared 0.308759 Adjusted R-squared 0.294358F(3, 96) 14.29355 P-value(F) 8.98e-08Log-likelihood -1226.249 Akaike criterion 2458.499Schwarz criterion 2466.284 Hannan-Quinn 2461.649rho 0.039826 Durbin-Watson 1.863716
6. Conclusion
The goal of this regression analysis was to come with a model that was capable of
providing estimates of future earnings for regional pure-play banks and based on my analysis I
would conclude that I have a capable working model. Based on the findings of the model, on
average holding all else equal earnings should increase by $0.45 for every one dollar increase in
previous quarter earnings. Furthermore, on average holding all else equal, a one percent
increase in net interest margin results in a $624,466 increase in earnings. Finally, on average
holding all else equal, a one percent increase in average assets results in a decrease in earnings
of $-57,237.6.
Now that I have a working model the goal of future research is to continue updating and
testing different variables to create the best fitting regression possible. Similarly, the goal of
future research is to create regressions for different industries and sectors that can aid in
predicting future earnings for a multitude of investment opportunities. The process of building
a regression not only helps in determining future earnings but also allows the researcher to test
different theories on the most important variables. Thus, the greatest benefit of the models is
finding out which variables have the greatest impact on future earnings.
Bibliography
-500000
-400000
-300000
-200000
-100000
0
100000
1 1.2 1.4 1.6 1.8
Earn
ings
l_Net_Interes
Earnings versus l_Net_Interes (with least squares fit)
Y = -1.80e+005 + 1.24e+005X
-500000
-450000
-400000
-350000
-300000
-250000
-200000
-150000
-100000
-50000
0
50000
-400000 -300000 -200000 -100000 0
Earn
ings
Lagged_Earnings
Earnings versus Lagged_Earnings (with least squares fit)
Y = -6.07e+003 + 0.502X
Albrecht, Steve, Larry Lookabill , and James McKeown. "The Time-Series Properties of Annual Earnings." Journal of Accounting Research. 15.2 (1977): 226-244. Print. <http://www.jstor.org/stable/2490350>.
Bureau of Economic Analysis. U.S. Department of Commerce, n.d. Web. 6 Mar 2012. <http://www.bea.gov/national/>.
EDGAR . SEC, n.d. Web. 6 Mar 2012. <http://www.sec.gov/edgar/searchedgar/companysearch.html>.
Finger, Catherine. "The Ability of Earnings to Predict Future Earnings and Cash Flow." Journal of Accounting Research. 32.2 (1994): 210-223. Print. <http://www.jstor.org/stable/2491282>.
Hanweck, Gerald , and Lisa Ryu. U.S. FDIC. Sensitivity of Bank Net Interest Margins and Profitability to Credit, Interest-Rate, and Term-Structure Shocks Across Bank Product Specializations. 2005. Print. <http://www.fdic.gov/bank/analytical/working/wp2005/WP2005_2.pdf>.
"UBPR Reports." . FDIC, n.d. Web. 6 Mar 2012. <https://cdr.ffiec.gov/public/ManageFacsimiles.asp&xgt;.
Appendices
-500000
-450000
-400000
-350000
-300000
-250000
-200000
-150000
-100000
-50000
0
50000
14.8 15 15.2 15.4 15.6 15.8 16 16.2
Earn
ings
l_AVG_Assets
Earnings versus l_AVG_Assets (with least squares fit)